3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 3.10.0-1160.6.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.1.1.el8 (RHEL8.3)
16 4.18.0-305.3.1.el8 (RHEL8.4)
17 4.4.120-92.70 (SLES12 SP2)
18 4.4.180-94.100 (SLES12 SP3)
19 4.4.0-131 (Ubuntu 16.04)
20 4.15.0-32 (Ubuntu 18.04)
21 5.4.0-48 (Ubuntu 20.04)
22 vanilla linux 5.4.0 (ZFS + ldiskfs)
23 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
24 * Client primary kernels built and tested during release cycle:
25 3.10.0-1062.18.1.el7 (RHEL7.7)
26 3.10.0-1127.19.1.el7 (RHEL7.8)
27 3.10.0-1160.6.1.el7 (RHEL7.9)
28 4.12.14-95.48 (SLES12 SP4)
29 4.12.14-122.20 (SLES12 SP5)
30 4.12.14-197.75 (SLES15 SP1)
31 5.3.18-24.24 (SLES15 SP2)
32 4.15.0-48 (Ubuntu 18.04)
33 * Other clients known to build on these kernels at some point (others may also work):
34 3.10.0-862.14.4.el7 (RHEL7.5)
35 3.10.0-957.27.2.el7 (RHEL7.6)
36 4.14.0-49.13.1.el7a (RHEL7.5)
37 4.14.0-115.2.2.el7a (RHEL7.6)
38 4.18.0-80.11.2.el8 (RHEL8.0)
39 4.18.0-147.8.1.el8 (RHEL8.1)
40 4.18.0-193.28.1.el8 (RHEL8.2)
41 4.18.0-240.1.1.el8 (RHEL8.3)
42 4.18.0-305.3.1.el8 (RHEL8.4)
43 4.18.0-348.2.1.el8 (RHEL8.5)
44 4.4.120-92.70 (SLES12 SP2)
45 4.4.180-94.100 (SLES12 SP3)
46 4.4.0-131 (Ubuntu 16.04)
47 5.4.0-37 (Ubuntu 20.04)
48 5.4.0 (vanilla kernel.org)
49 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
50 * Recommended ZFS version: 2.0.6
51 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
52 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
53 stack overflow. For more information, please refer to bugzilla 17630.
54 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
55 the re-exporting nodes to be mounted with "32bitapi" mount option
59 --------------------------------------------------------------------------------
63 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
64 for currently supported client and server kernel versions.
65 * Server primary kernels built and tested during release cycle:
66 3.10.0-1062.18.1.el7 (RHEL7.7)
67 3.10.0-1127.19.1.el7 (RHEL7.8)
68 3.10.0-1160.6.1.el7 (RHEL7.9)
69 * Other server kernels known to build and work at some point (others may also work):
70 3.10.0-862.14.4.el7 (RHEL7.5)
71 3.10.0-957.27.2.el7 (RHEL7.6)
72 4.18.0-80.11.2.el8 (RHEL8.0)
73 4.18.0-147.8.1.el8 (RHEL8.1)
74 4.18.0-193.28.1.el8 (RHEL8.2)
75 4.18.0-240.1.1.el8 (RHEL8.3)
76 4.4.120-92.70 (SLES12 SP2)
77 4.4.180-94.100 (SLES12 SP3)
78 4.4.0-131 (Ubuntu 16.04)
79 4.15.0-32 (Ubuntu 18.04)
80 5.4.0-48 (Ubuntu 20.04)
81 vanilla linux 5.4.0 (ZFS + ldiskfs)
82 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
83 * Client primary kernels built and tested during release cycle:
84 3.10.0-1062.18.1.el7 (RHEL7.7)
85 3.10.0-1127.19.1.el7 (RHEL7.8)
86 3.10.0-1160.6.1.el7 (RHEL7.9)
87 4.12.14-95.48 (SLES12 SP4)
88 4.12.14-122.20 (SLES12 SP5)
89 4.12.14-197.75 (SLES15 SP1)
90 5.3.18-24.24 (SLES15 SP2)
91 4.15.0-48 (Ubuntu 18.04)
92 * Other clients known to build on these kernels at some point (others may also work):
93 3.10.0-862.14.4.el7 (RHEL7.5)
94 3.10.0-957.27.2.el7 (RHEL7.6)
95 4.14.0-49.13.1.el7a (RHEL7.5)
96 4.14.0-115.2.2.el7a (RHEL7.6)
97 4.18.0-80.11.2.el8 (RHEL8.0)
98 4.18.0-147.8.1.el8 (RHEL8.1)
99 4.18.0-193.28.1.el8 (RHEL8.2)
100 4.18.0-240.1.1.el8 (RHEL8.3)
101 4.4.120-92.70 (SLES12 SP2)
102 4.4.180-94.100 (SLES12 SP3)
103 4.4.0-131 (Ubuntu 16.04)
104 5.4.0-37 (Ubuntu 20.04)
105 5.4.0 (vanilla kernel.org)
106 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
107 * Recommended ZFS version: 2.0.0
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
114 --------------------------------------------------------------------------------
118 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
119 for currently supported client and server kernel versions.
120 * Server primary kernels built and tested during release cycle:
121 3.10.0-1062.1.1.el7 (RHEL7.7)
122 * Other server kernels known to build and work at some point (others may also work):
123 3.10.0-862.14.4.el7 (RHEL7.5)
124 3.10.0-957.27.2.el7 (RHEL7.6)
125 4.18.0-80.7.1.el8 (RHEL8.0)
126 4.4.120-92.70 (SLES12 SP2)
127 4.4.180-94.100 (SLES12 SP3)
128 4.4.0-131 (Ubuntu 16.04)
129 4.15.0-32 (Ubuntu 18.04)
130 vanilla linux 4.6.7 (ZFS only)
131 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
132 * Client primary unpatched kernels built and tested during release cycle:
133 3.10.0-1062.1.1.el7 (RHEL7.7)
134 4.12.14-95.29 (SLES12 SP4)
135 4.15.0-48 (Ubuntu 18.04)
136 * Other clients known to build on unpatched kernels at some point (others may also work):
137 3.10.0-862.14.4.el7 (RHEL7.5)
138 3.10.0-957.27.2.el7 (RHEL7.6)
139 4.14.0-49.13.1.el7a (RHEL7.5)
140 4.14.0-115.2.2.el7a (RHEL7.6)
141 4.18.0-80.7.1.el8 (RHEL8.0)
142 4.4.120-92.70 (SLES12 SP2)
143 4.4.180-94.100 (SLES12 SP3)
144 4.4.0-131 (Ubuntu 16.04)
145 vanilla linux 4.15.0 (ZFS only)
146 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
147 * Recommended ZFS / SPL version: 0.7.13
148 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
149 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
150 stack overflow. For more information, please refer to bugzilla 17630.
151 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
152 the re-exporting nodes to be mounted with "32bitapi" mount option
153 * Note for Ubuntu 14.04.5:
154 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
155 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
157 --------------------------------------------------------------------------------
161 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
162 for currently supported client and server kernel versions.
163 * Server known to build on the following kernels (others may also work):
164 2.6.32-696.18.7.el6 (RHEL6.9)
165 3.10.0-862.14.4.el7 (RHEL7.5)
166 3.10.0-957.el7 (RHEL7.6)
167 4.4.120-92.70 (SLES12 SP2)
168 4.4.162-94.69 (SLES12 SP3)
169 4.4.0-131 (Ubuntu 16.04)
170 4.15.0-32 (Ubuntu 18.04)
171 vanilla linux 4.6.7 (ZFS only)
172 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
173 * Client known to build on unpatched kernels (others may also work):
174 2.6.32-696.18.7.el6 (RHEL6.9)
175 2.6.32-754.6.3.el6 (RHEL6.10)
176 3.10.0-862.14.4.el7 (RHEL7.5)
177 3.10.0-957.el7 (RHEL7.6)
178 4.14.0-49.13.1.el7a (RHEL7.5)
179 4.14.0-115.2.2.el7a (RHEL7.6)
180 4.4.120-92.70 (SLES12 SP2)
181 4.4.162-94.69 (SLES12 SP3)
182 4.4.0-131 (Ubuntu 16.04)
183 4.15.0-32 (Ubuntu 18.04)
184 vanilla linux 4.15.0 (ZFS only)
185 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
186 * Recommended ZFS / SPL version: 0.7.9
187 * Tested with ZFS / SPL version: 0.7.9
188 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
189 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
190 stack overflow. For more information, please refer to bugzilla 17630.
191 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
192 the re-exporting nodes to be mounted with "32bitapi" mount option
193 * Note for Ubuntu 14.04.5:
194 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
195 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
197 --------------------------------------------------------------------------------
199 2018-04-03 Intel Corporation
201 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
202 for currently supported client and server kernel versions.
203 * Server known to build on patched kernels:
204 2.6.32-431.29.2.el6 (RHEL6.5)
205 2.6.32-504.30.3.el6 (RHEL6.6)
206 2.6.32-573.26.1.el6 (RHEL6.7)
207 2.6.32-642.15.1.el6 (RHEL6.8)
208 2.6.32-696.18.7.el6 (RHEL6.9)
209 3.10.0-693.21.1.el7 (RHEL7.4)
210 3.0.101-0.47.71 (SLES11 SP3)
211 3.0.101-107 (SLES11 SP4)
212 3.12.74-60.64.40 (SLES12 SP1)
213 4.4.114-92.64 (SLES12 SP2)
214 4.4.114-94.11 (SLES12 SP3)
215 3.13.0-101 (Ubuntu 14.04, ZFS only)
216 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
217 4.4.0-116 (Ubuntu 16.04)
218 vanilla linux 4.6.7 (ZFS only)
219 * Client known to build on unpatched kernels:
220 2.6.32-431.29.2.el6 (RHEL6.5)
221 2.6.32-504.30.3.el6 (RHEL6.6)
222 2.6.32-573.26.1.el6 (RHEL6.7)
223 2.6.32-642.15.1.el6 (RHEL6.8)
224 2.6.32-696.18.7.el6 (RHEL6.9)
225 3.10.0-693.21.1.el7 (RHEL7.4)
226 3.0.101-0.47.71 (SLES11 SP3)
227 3.0.101-107 (SLES11 SP4)
228 3.12.74-60.64.40 (SLES12 SP1)
229 4.4.114-92.64 (SLES12 SP2)
230 4.4.114-94.11 (SLES12 SP3)
231 3.13.0-101 (Ubuntu 14.04)
232 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
233 4.4.0-116 (Ubuntu 16.04)
235 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
236 * Recommended ZFS / SPL version: 0.7.6
237 * Tested with ZFS / SPL version: 0.7.6
238 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
239 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
240 stack overflow. For more information, please refer to bugzilla 17630.
241 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
242 the re-exporting nodes to be mounted with "32bitapi" mount option
243 * Note for Ubuntu 14.04.5:
244 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
245 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
247 --------------------------------------------------------------------------------
249 2017-07-05 Intel Corporation
251 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
252 for currently supported client and server kernel versions.
253 * Server known to build on patched kernels:
254 2.6.32-431.29.2.el6 (RHEL6.5)
255 2.6.32-504.30.3.el6 (RHEL6.6)
256 2.6.32-573.26.1.el6 (RHEL6.7)
257 2.6.32-642.15.1.el6 (RHEL6.8)
258 2.6.32-696.3.1.el6 (RHEL6.9)
259 3.10.0-514.21.1.el7 (RHEL7.3)
260 3.0.101-0.47.71 (SLES11 SP3)
261 3.0.101-100 (SLES11 SP4)
262 3.12.74-60.64.40 (SLES12 SP1)
263 4.4.59-92.17 (SLES12 SP2)
264 vanilla linux 4.6.7 (ZFS only)
265 * Client known to build on unpatched kernels:
266 2.6.32-431.29.2.el6 (RHEL6.5)
267 2.6.32-504.30.3.el6 (RHEL6.6)
268 2.6.32-573.26.1.el6 (RHEL6.7)
269 2.6.32-642.15.1.el6 (RHEL6.8)
270 2.6.32-696.3.1.el6 (RHEL6.9)
271 3.10.0-514.21.1.el7 (RHEL7.3)
272 3.0.101-0.47.71 (SLES11 SP3)
273 3.0.101-100 (SLES11 SP4)
274 3.12.74-60.64.40 (SLES12 SP1)
275 4.4.59-92.17 (SLES12 SP2)
277 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
278 * Recommended ZFS / SPL version: 0.6.5.9
279 * Tested with ZFS / SPL version: 0.6.5.9
280 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
281 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
282 stack overflow. For more information, please refer to bugzilla 17630.
283 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
284 the re-exporting nodes to be mounted with "32bitapi" mount option
286 --------------------------------------------------------------------------------
288 12-07-2016 Intel Corporation
290 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
291 for currently supported client and server kernel versions.
292 * Server known to build on patched kernels:
293 2.6.32-431.29.2.el6 (RHEL6.5)
294 2.6.32-504.30.3.el6 (RHEL6.6)
295 2.6.32-573.26.1.el6 (RHEL6.7)
296 2.6.32-642.11.1.el6 (RHEL6.8)
297 3.10.0-514.el7 (RHEL7.3)
298 3.0.101-0.47.71 (SLES11 SP3)
299 3.0.101-80 (SLES11 SP4)
300 3.12.62-60.64.8 (SLES12 SP1)
301 vanilla linux 4.5.7 (ZFS only)
302 * Client known to build on unpatched kernels:
303 2.6.32-431.29.2.el6 (RHEL6.5)
304 2.6.32-504.30.3.el6 (RHEL6.6)
305 2.6.32-573.26.1.el6 (RHEL6.7)
306 2.6.32-642.11.1.el6 (RHEL6.8)
307 3.10.0-514.el7 (RHEL7.3)
308 3.0.101-0.47.71 (SLES11 SP3)
309 3.0.101-80 (SLES11 SP4)
310 3.12.62-60.64.8 (SLES12 SP1)
312 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
313 * Recommended ZFS / SPL version: 0.6.5.8
314 * Tested with ZFS / SPL version: 0.6.5.8
315 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
316 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
317 stack overflow. For more information, please refer to bugzilla 17630.
318 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
319 the re-exporting nodes to be mounted with "32bitapi" mount option
321 --------------------------------------------------------------------------------
323 02-29-2016 Intel Corporation
325 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
326 for currently supported client and server kernel versions.
327 * Server known to build on patched kernels:
328 2.6.32-431.29.2.el6 (RHEL6.5)
329 2.6.32-504.30.3.el6 (RHEL6.6)
330 2.6.32-573.12.1.el6 (RHEL6.7)
331 3.10.0-327.3.1.el7 (RHEL7.2)
332 3.0.101-0.47.71 (SLES11 SP3)
333 3.0.101-68 (SLES11 SP4)
334 vanilla linux 4.2.1 (ZFS only)
335 * Client known to build on unpatched kernels:
336 2.6.32-431.29.2.el6 (RHEL6.5)
337 2.6.32-504.30.3.el6 (RHEL6.6)
338 2.6.32-573.12.1.el6 (RHEL6.7)
339 3.10.0-327.3.1.el7 (RHEL7.2)
340 3.0.101-0.47.71 (SLES11 SP3)
341 3.0.101-68 (SLES11 SP4)
344 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
345 * Recommended ZFS / SPL version: 0.6.4.2
346 * Tested with ZFS / SPL version: 0.6.4.2
347 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
348 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
349 stack overflow. For more information, please refer to bugzilla 17630.
350 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
351 the re-exporting nodes to be mounted with "32bitapi" mount option
353 --------------------------------------------------------------------------------
355 03-10-2015 Intel Corporation
357 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
358 for currently supported client and server kernel versions.
359 * Server known to build on patched kernels:
360 2.6.32-431.29.2.el6 (RHEL6.5)
361 2.6.32-504.8.1.el6 (RHEL6.6)
362 3.0.101-0.46 (SLES11 SP3)
363 * Client known to build on unpatched kernels:
364 2.6.32-431.29.2.el6 (RHEL6.5)
365 2.6.32-504.8.1.el6 (RHEL6.6)
366 3.10.0-123.20.1.el7 (RHEL7)
367 3.0.101-0.46 (SLES11 SP3)
368 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
369 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
370 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
371 stack overflow. For more information, please refer to bugzilla 17630.
372 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
373 the re-exporting nodes to be mounted with "32bitapi" mount option
375 Severity : enhancement
377 Description: control OST-index in IDIF via ROCOMPAT flag.
378 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
379 last_rcvd file. For new formatted OST device, it will be auto set;
380 for the case of upgrading from old OST device, you can enable it
381 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
382 enabled, for new created OST-object, its IDIF-in-LMA will contain
383 the OST-index; for the existing OST-object, the OSD will convert
384 old format IDIF as new format IDIF with OST-index stored in the
385 LMA EA when accessing such OST-object or via OI scrub. Once such
386 flag is enabled, it cannot be reverted back, so the system cannot
387 be downgraded to the orignal incompatible version.
389 --------------------------------------------------------------------------------
391 07-30-2014 Intel Corporation
393 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
394 for currently supported client and server kernel versions.
395 * Server known to build on patched kernels:
396 2.6.32-431.20.3.el6 (RHEL6.5)
397 3.0.101-0.31 (SLES11 SP3)
398 * Client known to build on unpatched kernels:
399 2.6.32-431.20.3.el6 (RHEL6.5)
400 3.0.101-0.31 (SLES11 SP3)
403 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
404 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
405 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
406 stack overflow. For more information, please refer to bugzilla 17630.
407 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
408 the re-exporting nodes to be mounted with "32bitapi" mount option
410 --------------------------------------------------------------------------------
412 10-31-2013 Intel Corporation
414 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
415 * Server builds for kernels:
416 2.6.32-358.18.1.el6 (RHEL6.4)
417 * Client builds for unpatched kernels:
418 2.6.32-358.18.1.el6 (RHEL6.4)
419 3.0.92-0.8 (SLES11 SP3)
421 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
422 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
423 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
424 stack overflow. For more information, please refer to bugzilla 17630.
425 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
426 the re-exporting nodes to be mounted with "32bitapi" mount option
428 --------------------------------------------------------------------------------
430 05-30-2013 Intel Corporation
432 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
433 * Server builds for kernels:
434 2.6.32-358.6.2.el6 (RHEL6.4)
435 * Client builds for unpatched kernels:
436 2.6.32-358.6.2.el6 (RHEL6.4)
437 3.0.74-0.6.6 (SLES11 SP2)
439 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
440 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
441 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
442 stack overflow. For more information, please refer to bugzilla 17630.
443 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
444 the re-exporting nodes to be mounted with "32bitapi" mount option
446 --------------------------------------------------------------------------------
448 9-30-2012 Whamcloud, Inc.
450 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
451 * Server support for kernels:
452 2.6.32-279.5.1.el6 (RHEL6)
453 * Client support for unpatched kernels:
454 2.6.32-279.5.1.el6 (RHEL6)
456 2.6.18-238.19.1.el5 (RHEL5)
457 2.6.16.60-0.69.1 (SLES10)
458 * Recommended e2fsprogs version: 1.41.90.wc4
459 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
460 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
461 stack overflow. For more information, please refer to bugzilla 17630.
462 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
463 the re-exporting nodes to be mounted with "32bitapi" mount option
465 Severity : enhancement
467 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
469 Severity : enhancement
471 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
473 Severity : enhancement
475 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
477 Severity : enhancement
479 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
481 Severity : enhancement
483 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
485 --------------------------------------------------------------------------------
487 04-30-2012 Whamcloud, Inc.
489 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
490 * Server support for kernels:
491 2.6.32-220.4.2.el6 (RHEL6)
492 * Client support for unpatched kernels:
493 2.6.32-220.4.2.el6 (RHEL6)
495 2.6.18-238.19.1.el5 (RHEL5)
496 2.6.16.60-0.69.1 (SLES10)
497 * Recommended e2fsprogs version: 1.41.90.wc4
498 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
499 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
500 stack overflow. For more information, please refer to bugzilla 17630.
501 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
502 the re-exporting nodes to be mounted with "32bitapi" mount option
504 Severity : enhancement
506 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
508 Severity : enhancement
510 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
512 Severity : enhancement
514 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
516 Severity : enhancement
518 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
520 --------------------------------------------------------------------------------
522 09-30-2011 Whamcloud, Inc.
524 * Server support for kernels:
525 2.6.18-238.19.1.el5 (RHEL5)
526 2.6.32-131.6.1.el6 (RHEL6)
527 * Client support for unpatched kernels:
528 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
529 2.6.18-238.19.1.el5 (RHEL5)
530 2.6.32-131.6.1.el6 (RHEL6)
531 2.6.29.4-167.fc11 (FC11)
532 2.6.16.60-0.69.1 (SLES10)
534 * Recommended e2fsprogs version: 1.41.90.wc3
535 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
536 Lustre filesystem with 4K stack may cause a stack overflow. For more
537 information, please refer to bugzilla 17630.
538 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
539 mounted with "32bitapi" mount option
541 Severity : enhancement
542 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
544 Severity : enhancement
545 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
547 Severity : enhancement
548 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
551 Frequency : only with SLES10
553 Description: Use OFED "KMP" provided by Novell
554 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
555 Lustre is now built against this package. That means you need to
556 install the ofed-kmp package from Novell for the patchless client.
557 Note that the ofed-kmp that Novell ships may not exactly match the
558 kernel version but should still be compatible.
560 Severity : enhancement
562 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
563 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
565 --------------------------------------------------------------------------------
567 2010-07-15 Oracle, Inc.
569 * Server support for kernels:
570 2.6.18-164.11.1.el5 (RHEL 5)
571 2.6.18-164.11.1.0.1.el5 (OEL 5)
572 * Client support for unpatched kernels:
573 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
574 2.6.18-164.11.1.el5 (RHEL 5),
575 2.6.18-164.11.1.0.1.el5 (OEL 5)
576 2.6.16.60-0.42.8 (SLES 10),
578 2.6.29.4-167.fc11 (FC11)
579 * Recommended e2fsprogs version: 1.41.10-sun2
581 Severity : enhancement
583 Description: Distribute lfs_migrate script with lustre
584 Details : The lfs_migrate mentioned in the manual is now included in the
585 lustre tree in lustre/scripts and will distributed in lustre RPMs.
587 Severity : enhancement
589 Description: Failover nid list added to import proc file.
591 Severity : enhancement
593 Description: "som_preview" mount option required to use SOM
594 Details : The clients will need to use som_preview mount option
598 Description: Lustre does not do 1MB IOs to HW RAID
599 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
601 Severity : enhancement
603 Description: lfs quota output is not very convenient for awk/sed-parsing
604 Details : Some positions in lfs quota output table could be empty or
605 non-empty which made it hard to parse it with scripts, now
606 a dash is put instead of space where there is not supposed
611 Description: Update to OFED 1.5.1
612 Details : OFED 1.5.1 will be used in environments that don't use
613 vendor supplied OFED.
617 Description: Remove lnet. prefix from lctl params display
618 Details : Also renamed lnet.memused to lnet_memused to
623 Description: Use conf_param -d to remove permanant settings.
627 Description: Add list_param -R
628 Details : lctl list_param now supports recursive mode.
631 Bugzilla : 21829, 22312
632 Description: Fix llobdstat.
633 Details : 1. Now updating the write_bytes counter
634 2. Port script fixes from 1.8.
638 Description: decrease the usage of memory on clients.
639 Details : 1. On clients, recycle dentries and inodes unused.
640 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
645 Description: llobdstat fix and enhancement
646 Details : add a counter to set a limit to how many samples will be returned
647 fix a wildcard in the path to limit to obdfilter stats only
651 Description: lfs check only allowed for root
652 Details : Code cleanups and added a regresion test of non-root use of
657 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
658 Details : Fix hr->hr_index overload.
662 Description: Fix FIEMAP functionality.
666 Description: FC11 patchless client support added
670 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
671 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
675 Description: New acc-sm script and various test-framework enhancments.
679 Description: Previously landed patch was reverted.
683 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
687 Description: Add ldiskfs maxdirsize mount option.
691 Description: Fix mkfs.lustre support for 16TB LUNs.
695 Description: Print mmp_check_interval
696 Details : print mmp_check_interval and make it possible to abort
697 mount operation in case it takes too long.
701 Description: Add cascading_rw to lustre mpi tests.
705 Description: Store SOM data in EA.
706 Details : Also several SOM related cleanups.
710 Description: Add refcount to md_open_data
714 Description: Use refcount on lov_request_set to prevent memory corruption.
718 Description: Statahead: use parent dir's i_mutex for synchronization.
722 Description: loadgen improvements.
726 Description: Enable peer health detection in ksocklnd and ko2iblnd.
730 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
734 Description: Fix race on obd_devcie:md_stats
738 Description: Fix SLES 11 client builds.
739 Details : Resolved uninitalized pointer gcc warnings.
743 Description: Allow non-root access for "lfs check".
744 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
748 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
750 Severity : enhancement
752 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
753 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
757 Description: Sanity test_27p fails.
758 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
759 Sanity tests 27 were cleaned up and improved.
763 Description: Unify req format on client/servers
764 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
765 buffers and swabbers approach..
770 Description: drop join file stuff.
772 Severity : enhancement
774 Description: Add 2.6.27 and SLES11 (patchless client) support
778 Description: Fix in ptlrpc_expire_one_request() to print the signed time
781 Severity : enhancement
783 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
786 Severity : enhancement
788 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
790 Severity : enhancement
792 Description: correctly handle big reply message.
793 Details : send LNet event if reply is bigger than buffer and adjust this buffer
798 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
800 Severity : enhancement
802 Description: Add support for OEL5.
806 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
808 Severity : enhancement
810 Description: "weak-modules" support
811 Details : Implement "weak-modules" support which enables kernel modules
812 to be used with any kernel that implements the same kABI. In
813 order to achieve this modules are now installed in
814 /lib/modules/$(uname -r)/updates/kernel on all distributions.
818 Description: client could not reconnect to OST because of an active request.
819 Details : The client now retries to reconnect to the same server, if a
820 connect request failed with EBUSY or -EAGAIN. Test case 35b
821 conf-sanity.sh has been created.
825 Description: Don't allow make backward step on assignin osc next id.
826 Details : race between allocation next id and ll_sync thread can be cause
827 of set wrong osc next id and can be kill valid ost objects.
831 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
835 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
839 Description: Error handling in osc_statfs_interpret() has been improved.
840 Details : Check in osc_statfs_interpret() for EBADR.
844 Description: Do not send statfs() requests to OSTs disabled by administrator.
845 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
847 Severity : enhancement
849 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
851 Severity : enhancement
853 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
855 Severity : enhancement
857 Description: Enhancement for lfs(1) command to use numeric uid/gid.
861 Description: Improved cleanup procedure for conf-sanity test_46a.
865 Description: Do not start bulk transfer, if client tried to reconnect.
866 Details : Do not start bulk transfer, if client tried to reconnect,
867 but failed due to active rpcs. Check for exp_abort_active_req
868 has been implemented.
872 Description: Prevent inconsistences between linux and lustre mount structures.
873 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
874 Make the sleep interruptible.
876 Severity : enhancement
878 Description: provide server to client comms path
879 Details : server to client communications path via new LDLM_SET_INFO rpc
883 Description: truncate starts GFP_FS allocation under transaction causing deadlock
884 Details : ldiskfs_truncate calls grab_cache_page which may start page
885 allocation under an open transaction. This may lead to
886 calling prune_icache with consequent lustre reentrance.
890 Description: send correctly lsm on open replay
891 Details : MDS is trust to LSM size on replay open, but client can set wrong size
896 Description: enable client interface failover
897 Details : When a child reconnects from another NID, properly update export
898 nid hash position and ldlm reverse import.
900 Severity : enhancement
902 Description: Communicate OST degraded/readonly state via statfs to MDS
903 Details : Flags in the statfs returned from OSTs indicate whether the
904 OST is in a degraded RAID state, or if the filesystem has
905 turned read-only after a filesystem error is detected.
910 Description: don't panic if EPROTO was hit when reading symlink
911 Details : correctly handling request reference in error cases.
913 Severity : enhancement
915 Description: Add LustreNetLink, a kernel-userspace communcation path.
917 Severity : enhancement
919 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
924 Description: access to llog context before init.
925 Details : move handling CATALOGS file at osc layer and forbid access to llog
929 Frequency : rare, only when client is writing to OST on local node
931 Description: Avoid deadlock for local client writes
932 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
933 memory freeing context. This allows OST threads to set the
934 PF_MEMALLOC flag on task structures in order to allocate memory
935 from reserved pools and complete IO.
936 Use GFP_HIGHUSER for OST allocations for non-local client writes,
937 so that the OST threads generate memory pressure and allow
938 inactive pages to be reclaimed.
940 Severity : enhancement
942 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
947 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
948 Details : move ldlm namespace creation in setup phase to avoid grab
949 _lprocfs_lock with cli_sem held.
953 Description: Temporarily disable grant shrink.
954 Details : Disable the feature for debugging.
957 Frequency : only when formatting test filesystems
959 Description: Unable to run several mkfs.lustre on loop devices at the same time
960 Details : mkfs.lustre returns error 256 on the concurrent loop devices
961 formatting. The solution is to proper handle the error.
963 Severity : enhancement
965 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
967 Severity : enhancement
969 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
974 Description: resolve race between obd_disconnect and class_disconnect_exports
975 Details : if obd_disconnect will be called to already disconnected export he
976 forget release one reference and osc module can't unloaded.
978 Severity : enhancement
980 Description: Allow tuning service thread via /proc
981 Details : For each service a new
982 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
983 created that can be used to set min/max thread counts, and get the
984 current number of running threads.
989 Descriptoin: don't return error if have partially created objects for file.
990 Details : lov_update_create_set uses set->set_success as index for created
991 objects, so if some requests will be failed, they will have hole
992 at end of array and qos_shrink_lsm for allocate correct lsm.
994 Severity : enhancement
996 Description: Update OFED support to 1.4.1
1001 Description: Reduce small size read RPC
1002 Details : Set read-ahead limit for every file and only do read-ahead when
1003 available read-ahead pages are bigger than 1M to avoid small size
1006 Severity : enhancement
1008 Description: move AT tunable parameters for more consistent usage
1009 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1011 Severity : enhancement
1013 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1014 Details : allow skip disconnected ost for send statfs request and hide error
1020 Description: fix racy locking of mballoc block bitmaps causing BUG
1021 Details : The locking of the mballoc buddy bitmap and the in-memory
1022 block bitmap was using two different spin locks in some
1023 cases. This made it possible to incorrectly access the
1024 mballoc bitmap while another process was modifying it,
1025 causing a sanity assertion to fail. While no on-disk corruption
1026 was reported, there was some risk of this happening.
1029 Frequency : rare, on llog test 6
1031 Descriptoin: don't allow connect to already connected import
1032 Details : allowing connect to already connected import is hide connecting problem.
1037 Descriptoin: allow kill process which wait statahead result
1038 Details : for some reasons 'ls' can stick in waiting result from statahead,
1039 in this case need way for kill this process.
1041 Severity : enhancement
1043 Description: Add state history info file, enhance import info file
1044 Details : Track import connection state changes in a new osc/mdc proc file;
1045 add overview-type data to the osc/mdc import proc file.
1047 Severity : enhancement
1049 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1051 Severity : enhancement
1053 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1058 Descriptoin: OST grant too much space to client even there are not enough space.
1059 Details : Client will shrink its grant cache to OST if there are no write
1060 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1061 this grant cache if there are already not enough avaible space
1062 (left_space < total_clients * 32M).
1065 Frequency : start MDS on uncleanly shutdowned MDS device
1067 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1068 Details : stay in waiting mds<>ost recovery finished produce random bugs
1069 due race between two ll_sync thread for one lov target. send
1070 ACTIVATE event only if connect realy finished and import have
1074 Frequency : rare, connect and disconnect target at same time
1076 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1077 Details : don't call obd_disconnect under lov_lock. this long time
1078 operation and can block ptlrpcd which answer to connect request.
1083 Descriptoin: don't lose wakeup for imp_recovery_waitq
1084 Details : recover_import_no_retry or invalidate_import and import_close can
1085 both sleep on imp_recovery_waitq, but we was send only one wakeup
1089 Frequency : always with long access acl
1091 Descriptoin: mds can't pack reply with long acl.
1092 Details : mds don't control size of acl but they limited by reint/getattr
1095 Severity : enhancement
1097 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1099 Severity : enhancement
1101 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1104 Frequency : start MDS on uncleanly shutdowned MDS device
1106 Descriptoin: aborting recovery hang on MDS
1107 Details : don't throttle destroy RPCs for the MDT.
1110 Frequency : on remount
1112 Description: external journal device not working after the remount
1113 Details : clear dev_rdonly flag for external journal devices in
1119 Description: shutdown vs evict race
1120 Details : client_disconnect_export vs connect request race.
1121 if client will evicted at this time - we start invalidate
1122 thread without referece to import and import can be freed
1126 Frequency : rare, need acl's on inode.
1128 Description: client can't handle ost additional correctly
1129 Details : if ost was added after client connected to mds client can have
1130 hit lnet_try_match_md ... to big messages to wide striped files.
1131 in this case need teach client to handle config events about add
1132 lov target and update client max ea size at that event.
1134 Severity : enhancement
1136 Description: Changelogs
1137 Details : Changelogs are a lightweight mechanism to track filesystem
1138 metadata and namespace changes. The changelog is recorded
1139 permanently on the MDTs, and is periodically "consumed" / purged
1140 when records are no longer needed.
1142 Severity : enhancement
1144 Description: compact fld format with extents
1145 Details : Store range of seq rather than every seq in FLD. Seq
1146 controller update FLD rather than clients. In Case of CMD, mdt0
1147 has FLD, all other metadata server act as non persistent proxy
1148 for FLD queries and cache fld entries in fld cache.
1153 Description: don't skip ost target if they assigned to file
1154 Details : Drop slow OSCs if we can, but not for requested start idx.
1155 This means "if OSC is slow and it is not the requested
1156 start OST, then it can be skipped, otherwise skip it only
1157 if it is inactive/recovering/out-of-space.
1161 Description: more cleanup in mds_lov
1162 Details : not send LOV EA under replay, we can't know about they size at this
1163 time. Don't allow client connect to mds before any ost connected,
1164 for avoid problems with LOV EA size and returning EIO to client.
1166 Severity : enhancement
1168 Description: Interoperability at server side (Disk interoperability)
1170 Severity : enhancement
1172 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1174 Severity : enhancement
1176 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1178 Severity : enhancement
1180 Description: New client IO stack (CLIO).
1182 Severity : enhancement
1184 Description: Commit on sharing. Eliminate inter-client dependencies between
1185 uncommitted transactions by doing transaction commits.
1186 Thereby clients may recovery independently.
1189 Frequency : Create a symlink file with a very long name
1191 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1192 Details : If there is no extra space in the request for early cancels,
1193 ldlm_req_handles_avail() returns 0 instead of a negative value.
1195 Severity : enhancement
1197 Description: Add /proc entry for import status
1198 Details : The mdc, osc, and mgc import directories now have
1199 an import directory that contains useful import data for debugging
1200 connection problems.
1202 Severity : enhancement
1204 Description: Re-disable certain /proc logging
1205 Details : Enable and disable client's offset_stats, extents_stats and
1206 extents_stats_per_process stats logging on the fly.
1209 Frequency : Only on FC kernels 2.6.22+
1211 Description: oops in statahead
1212 Details : Do not drop reference count for the dentry from VFS when lookup,
1213 VFS will do that by itself.
1215 Severity : enhancement
1217 Description: Generic /proc file permissions
1218 Details : Set /Proc file permissions in a more generic way to enable non-
1219 root users operate on some /proc files.
1223 Description: Hitting mdc_commit_close() ASSERTION
1224 Details : Properly handle request reference release in
1225 ll_release_openhandle().
1229 Description: quota recovery deadlock during mds failover
1230 Details : This patch includes att18982, att18236, att18237 in bz14840.
1232 1. fix osts hang when mds does failover with quotaon
1233 2. prevent watchdog storm when osts threads wait for the
1238 Frequency : only patchless client
1239 Description: add workaround for race between add/remove dentry from hash
1241 Severity : enhancement
1243 Description: Allow OST glimpses to return PW locks
1247 Description: LBUG when llog conf file is full
1248 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1252 Description: Prevent import from entering FULL state when server in recovery
1256 Description: service mount cannot take device name with ":"
1257 Details : Only when device name contains ":/" will mount treat it as
1263 Description: replace ptlrpcd with the statahead thread to interpret the async
1264 statahead RPC callback
1268 Frequency : on recovery
1269 Description: I/O failures after umount during fail back
1270 Details : if client reconnected to restarted server we need join to recovery
1271 instead of find server handler is changed and process self
1272 eviction with cancel all locks.
1274 Severity : enhancement
1276 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1278 Severity : enhancement
1280 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1282 Severity : enhancement
1284 Description: Upcall on Lustre log has been dumped
1285 Details : Allow for a user mode script to be called once a Lustre log has
1286 been dumped. It passes the filename of the dumped log to the
1287 script, the location of the script can be specified via
1288 /proc/sys/lnet/debug_log_upcall.
1293 Description: avoid idr_remove called for id which is not allocated.
1294 Details : Move assigment s_dev for clustered nfs to end of initialization,
1295 to avoid problem with error handling.
1300 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1301 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1302 into CONN_USED_HASH and this prodice warning when put connection
1303 again in unused hash.
1306 Severity : enhancement
1308 Description: File striping can now be set to use an arbitrary pool of OSTs.
1310 Severity : enhancement
1312 Description: Export bytes_read/bytes_write count on OSC/OST.
1316 Description: Early reply size mismatch, MGC loses connection
1317 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1318 the connect flags are properly negotiated.
1323 Description: quotas are not honored with O_DIRECT
1324 Details : all writes with the flag O_DIRECT will use grants which leads to
1325 this problem. Now using OBD_BRW_SYNC to guard this.
1329 Description: add quota statistics
1330 Details : 1. sort out quota proc entries and proc code.
1331 2. add quota statistics
1333 Severity : enhancement
1335 Description: enable quota support for HEAD.
1339 Description: Properly propagate oinfo flags from lov to osc for statfs
1340 Details : restore missing copy oi_flags to lov requests.
1342 Severity : enhancement
1344 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1345 : l_getgroups(8), lst(8), routerstat(8)
1346 Details : included man pages for llobdstat(8), llstat(8),
1347 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1349 Severity : enhancement
1351 Description: configure's --enable-quota should check the
1352 : kernel .config for CONFIG_QUOTA
1353 Details : configure is terminated if --enable-quota is passed but
1354 : no quota support is in kernel
1358 Description: Remove portals compatibility
1359 Details : Remove portals compatibility, not interoperable with releases
1364 Description: Resolve device initialization race
1365 Details : Prevent proc handler from accessing devices added to the
1366 obd_devs array but yet be intialized.
1368 Severity : enhancement
1370 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1372 Severity : enhancement
1374 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1379 Description: Using wrong pointer in osc_brw_prep_request
1380 Details : Access to array[-1] can produce panic if kernel compiled with
1381 CONFIG_PAGE_ALLOC enabled
1385 Description: Client runs out of low memory
1386 Details : Consider only lowmem when counting initial number of llap pages
1390 Description: *optional* service tags registration
1391 Details : if the "service tags" package is installed on a Lustre node
1392 When the filesystem is mounted, a local-node service tag will
1393 be created. See http://inventory.sun.com/ for more information
1394 about the Service Tags asset management system.
1398 Description: Kernel BUG tries to release flock
1399 Details : Lustre does not destroy flock lock before last reference goes
1400 away. So always drop flock locks when client is evicted and
1401 perform unlock regardless of successfulness of speaking to MDS.
1405 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1409 Description: sanity test 65a fails if stripecount of -1 is set
1410 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1415 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1416 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1419 Severity : enhancement
1421 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1425 Description: do not process already freed flock
1426 Details : flock can possibly be freed by another thread before it reaches
1427 to ldlm_flock_completion_ast.
1431 Description: LBUG during stress test
1432 Details : Need properly lock accesses the flock deadlock detection list.
1436 Description: oops in page fault handler
1437 Details : kernel page fault handler can return two special 'pages' in error
1438 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1442 Description: timeout with invalidate import.
1443 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1444 should be handled by ptlrpcd. This produce long age waiting and
1445 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1447 Severity : enhancement
1449 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1453 Description: enable MGS and MDT services start separately
1454 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1455 a co-located MGS without starting the MGS, which is a complement
1456 to 'nosvc' mount option.
1460 Frequency : after recovery
1461 Description: precreate to many object's after del orphan.
1462 Details : del orphan st in oscc last_id == next_id and this triger growing
1463 count of precreated objects. Set flag LOW to skip increase count
1464 of precreated objects.
1468 Frequency : rare, on clear nid stats
1469 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1470 Details : when clean nid stats sometimes try destroy live entry,
1471 and this produce panic in free.
1475 Description: Stack overflow during MDS log replay
1476 ease stack pressure by using a thread dealing llog_process.
1480 Description: wait until IO finished before start new when do lock cancel.
1481 Details : VM protocol want old IO finished before start new, in this case
1482 need wait until PG_writeback is cleared until check dirty flag and
1483 call writepages in lock cancel callback.
1485 Severity : enhancement
1487 Description: using special macro for print time and cleanup in includes.
1491 Description: mds_mfd_close() ASSERTION(rc == 0)
1492 Details : In mds_mfd_close(), we need protect inode's writecount change
1493 within its orphan write semaphore to prevent possible races.
1497 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1501 Frequency : rare, on shutdown ost
1502 Description: don't hit live lock with umount ost.
1503 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1504 use shrink_dcache_sb instead.
1508 Description: don't panic with use echo client
1509 Details : echo client pass NULL as client nid pointer and this produce null
1510 pointer dereference.
1514 Description: fix build on ppc32
1515 Details : compile code with -m64 flag produce wrong object file for ppc32.
1519 Description: add message levels for liblustreapi
1523 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1524 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1525 will result in an -EAGAIN and in the exit call path we call
1526 alloc_rr() which will with increasing aggressiveness attempt to
1527 aquire precreated objects on the minimum number of required OSCs.
1531 Description: Use old size assignment to avoid deadlock
1532 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1533 the scheduling while holding a spinlock. We do not need locking
1534 for size in ll_update_inode() because size is only updated from
1535 the MDS for directories or files without objects, so there is no
1536 other place to do the update, and concurrent access to such inodes
1537 are protected by the inode lock.
1541 Description: resolve "_IOWR redefined" build error on SLES10
1545 Description: dump the memory debugging after all modules are unloaded to
1546 suppress false negative in conf_sanity test 39
1548 Severity : enhancement
1550 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1555 Description: fix SLES kernel versioning
1556 Details : the kernel version for our SLES 10 kernel did not include a "-"
1557 before the "smp" at the end. while this was not a problem in
1558 general, it did mean that software trying to use the kernel
1559 version to try to detect a vendor specific kernel would fail.
1560 this was most evident by the OFED build scripts.
1564 Description: Don't update lov_desc members until making sure they are valid
1565 Details : When updating lov_desc members via proc fs, need fix their
1566 validities before doing the real update.
1570 Description: don't put request into delay list while invalidate in flight.
1571 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1572 import in flight. this produce timeout for invalidate and sometimes
1573 can cause stale data.
1577 Frequency : on ppc only
1578 Description: not convert ost objects for directory because it's not exist.
1579 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1583 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1585 Severity : enhancement
1587 Description: Update to support 2.6.22.14 vanilla kernel.
1591 Frequency : rare, on recovery
1592 Description: read procfs can produce deadlock in some situation
1593 Details : Holding lprocfs lock which send rpc can produce block for destroy
1594 obd objects and this also block reconnect with -EALREADY. This
1595 isn't fix all lprocfs bugs - but make it rare.
1597 Severity : enhancement
1599 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1602 Frequency : frequent on X2 node
1604 Description: mdc_set_open_replay_data LBUG
1605 Details : Set replay data for requests that are eligible for replay.
1609 Description: lustre_mgs: operation 101 on unconnected MGS
1610 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1611 MGC, and late on MGC cannot successfully connect to MGS and a lot
1612 of the error messages complaining that MGS is not connected.
1616 Frequency : on network error
1617 Description: panic with double free request if network error
1618 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1619 but it's true only for synchronus enqueue, for async enqueue
1620 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1623 Severity : enhancement
1625 Description: client-side metadata stat-ahead during readdir(directory readahead)
1626 Details : perform client-side metadata stat-ahead when the client detects
1627 readdir and sequential stat of dir entries therein
1630 Frequency : on start mds
1632 Description: Implement get_info(last_id) in obdfilter.
1635 Frequency : occasional
1637 Description: Correctly check stale fid, not start epoch if ost not support SOM
1638 Details : open with flag O_CREATE need set old fid in op_fid3 because
1639 op_fid2 was overwritten with new generated fid, but mds can answer
1640 with one of these two fids and both is not stale. Setattr
1641 incorrectly started an epoch and assume will be called
1642 done_writing, but without SOM done_writing ever being called.
1645 Frequency : rare, depends on device drivers and load
1647 Description: MDS or OSS nodes crash due to stack overflow
1648 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1649 In some cases, in conjunction with device drivers that use a lot
1650 of stack the MDS (or possibly OSS) service threads could overflow
1651 the stack. One change which was identified to consume additional
1652 stack has been reworked to avoid the extra stack usage.
1655 Frequency : occasional
1657 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1658 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1659 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1660 oscc_has_objects and see OSC has no objests and call
1661 oscc_internal_create to resend create request.
1663 Severity : enhancement
1665 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1667 Severity : enhancement
1669 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1672 Frequency : very rare
1674 Description: Fix replay if there is an un-replied request and open
1675 Details : In some cases, older replay request will revert the
1676 mcd->mcd_last_xid on MDS which is used to record the client's
1677 latest sent request.
1679 Severity : enhancement
1681 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1683 Severity : enhancement
1685 Description: Add rhel5 support to HEAD.
1687 Serverity : enhancement
1689 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1694 Description: Don't allow skipping OSTs if index has been specified.
1695 Details : Don't allow skipping OSTs if index has been specified, make
1696 locking in internal create lots better.
1700 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1701 Details : access to bitfield in structure is always rounded to long
1702 and this produce problem with not atomic change any bit.
1706 Description: Lustre make rpms failed.
1707 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1708 builds Lustre rpms from tarball.
1711 Frequency : rare on shutdown ost
1713 Description: If llog cancel was not send before clean_exports phase, this can
1714 produce deadlock in llog code.
1715 Details : If llog thread has last reference to obd and call class_import_put
1716 this produce deadlock because llog_cleanup_commit_master wait when
1717 last llog_commit_thread exited, but this never success because was
1718 called from llog_commit_thread.
1722 Description: allow userland application know is lost one of stripes.
1723 Details : fill lvb_blocks with error code on ost and return it to
1724 application if error flag found.
1728 Description: NULL lov_tgts causing MDS oops
1729 Details : more safe checks for NULL lov_tgts for avoid oops.
1731 Severity : enhancement
1733 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1737 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1738 Details : If an OST is down the MDS will hang indefinitely in
1739 obd_statfs_rqset() waiting for the statfs data. While for
1740 MDS QOS usage of statfs, it should not stuck in waiting.
1742 Severity : enhancement
1744 Description: remote_acl support
1745 Details : Support ACL-based permission check for remote user.
1746 Support setfacl/getfacl for remote user with the utils
1747 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1748 the system "{s,g}etfacl" utils.
1750 Severity : enhancement
1752 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1754 Severity : enhancement
1756 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1760 Description: make mgs_setparam() handle fsname containing dash
1761 Details : fsname containing a dash does not work with lctl conf_param
1763 Severity : enhancement
1765 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1767 Severity : enhancement
1769 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1773 Frequency : rare, at shutdown
1774 Description: access already free / zero obd_namespace.
1775 Details : if client_disconnect_export was called without force flag set,
1776 and exist connect request in flight, this can produce access to
1777 NULL pointer (or already free pointer) when connect_interpret
1778 store ocd flags in obd_namespace.
1782 Frequency : only at startup
1783 Description: not alloc memory with spinlock held.
1784 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1785 if any spinlock held.
1787 Severity : enhancement
1789 Description: make lustre randomly fail allocating memory
1790 Details : Make lustre randomly failed allocating memory for testing purpose.
1792 Severity : enhancement
1794 Description: lost problems with lov objid file
1795 Details : Fixes some scability and access to not inited memory problems
1796 in work with lov objdid file.
1801 Description: lfs find does not continue on file error
1802 Details : Continue other files processing when a file/dir is absent.
1806 Description: Inconsistent usage of lustre_pack_reply()
1807 Details : Standardize the usage of lustre_pack_reply() such that it
1808 always generate a CERROR on failure.
1811 Frequency : occasional
1813 Description: MDS hang or stay in waiting lock
1814 Details : If client receive lock with CBPENDING flag ldlm need send lock
1815 cancel as separate rpc, to avoid situation when cancel request
1816 can't processed due all i/o threads stay in wait lock.
1820 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1824 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1828 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1832 Description: Update to RHEL4 latest kernel.
1834 Severity : enhancement
1836 Description: Build SLES10 patchless client fails
1837 Details : The configure was broken by run ./configure with
1838 --with-linux-obj=.... argument for patchless client. When the
1839 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1840 can't search header adequately. Use absolute path such as
1841 -I($LINUX)/include instead.
1845 Description: interrupt oig_wait produce painc on resend.
1846 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1847 private set, and this produce situation when rq_ptlrpcd_data not
1848 copyed to new allocated request and triggered LBUG on assert
1849 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1850 ptlrpcd set if request is changed and can be safety changed to use
1855 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1856 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1857 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1858 these macros and replace them with their difinitions instead.
1862 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1863 Details : Update which_patch & target file for SLES10 latest kernel.
1866 Description: Remove client patches from SLES 10 kernel.
1867 Details : This causes SLES 10 clients to behave as patchless clients
1868 even on a Lustre-patched (server) kernel.
1870 Severity : enhancement
1872 Description: self-adjustable client's lru lists
1873 Details : use adaptive algorithm for managing client cached locks lru
1874 lists according to current server load, other client's work
1875 pattern, memory activities, etc. Both, server and client
1876 side namespaces provide number of proc tunables for controlling
1879 Severity : enhancement
1881 Description: light-weight GSS support
1882 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1883 and reduce performance overhead.
1885 Severity : enhancement
1887 Description: Linux keyring support
1888 Details : Support using service of Linux keyring for Lustre GSS internal
1889 context refresh/cache mechanism.
1893 Description: Fix errors in lfs documentation
1894 Details : Fixes man pages
1898 Description: don't use GFP_* in generic Lustre code.
1899 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1903 Description: obdclass is limited by single OBD_ALLOC(idarray)
1904 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1908 Description: warnings with build patchless client with vanila 2.6.19 and up
1909 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1910 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1914 Frequency : Only for SLES
1915 Description: sanity_quota fail test_1
1916 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1917 file, which makes TSTID[2] inunique.
1921 Frequency : Only for SLES9
1922 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1923 Details : Update target/ChangeLog/which_patch .
1927 Frequency : Only for test_10 in sanity-quota.sh
1928 Description: a bug in quota test code
1929 Details : checking a test flag in a wrong place
1933 Frequency : Only for test_13 in sanity-quota.sh
1934 Description: a bug in quota test script
1935 Details : shouldn't check overfull allocated quota
1939 Description: jbd statistics
1940 Details : Port older jbd statistics patch for sles10
1942 Severity : enhancement
1944 Description: Add printing inode info into message about error in writepage.
1949 Description: reply_lock_interpret crash due to race with it and lock cancel.
1950 Details : Do not replay locks that are being cancelled. Do not reference
1951 locks by their address during replay, just by their handle.
1955 Frequency : When flocks are used.
1956 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1957 Details : Flock locks might destroy just granted lock if it could be merged
1958 with another existing flock, this is done in completion handler,
1959 so teach ldlm_cli_enquque_fini that this is a valid case for
1965 Description: Oops in read and write path when failing to allocate lock.
1966 Details : Check if lock allocation failed and return error back.
1970 Description: lstripe command fails for valid OST index
1971 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1972 instead of lov->desc.ld_active_tgt_count.
1976 Description: sanity.sh failed test 103
1977 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1982 Description: df doesn't work properly if diskfs blocksize != 4K
1983 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1987 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1988 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1990 Severity : enhancement
1992 Description: Adaptive timeouts
1993 Details : RPC timeouts adapt to changing server load and network
1994 conditions to reduce resend attempts and improve recovery time.
1998 Description: llapi_file_create() does not allow some changes
1999 Details : add llapi_file_open() that allows specifying the mode and
2000 open flags, and also returns an open file handle.
2004 Description: merge and cleanup kernel patches.
2005 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2009 Description: Add journal checksum support.(Kernel part)
2010 Details : The journal checksum feature adds two new flags i.e
2011 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2012 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2013 indicates that the commit block contains the checksum for
2014 the blocks described by the descriptor blocks. Now commit
2015 record can be sent to disk without waiting for descriptor
2016 blocks to be written to disk. This behavior is controlled
2017 using JBD2_FEATURE_ASYNC_COMMIT flag.
2021 Description: OSS needs mutliple precreate threads
2022 Details : Add ability to start more than one create thread per OSS.
2026 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2027 Details : execute lfs setstripe on client
2031 Description: mds_obd_create error creating tmp object
2032 Details : When the user sets quota on root, llog will be affected and can't
2033 create files and write files.
2036 Frequency : Always on ia64 patchless client, and possibly others.
2038 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2039 Details : This allows the patchless client to be loaded on architectures
2040 without this export.
2044 Description: RedHat Update kernel for RHEL5
2045 Details : Add the kernel config file for RHEL5.
2049 Description: RedHat Update kernel for RHEL5
2050 Details : Modify the kernel config file more closer RHEL5.
2054 Description: Build failure against Centos5 (RHEL5)
2055 Details : Define PAGE_SIZE when it isn't present.
2059 Description: test_7 on sanity-quota.sh can't be used on separate machines
2060 Details : add facet to handle it
2064 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2065 Details : with NFS, the anon dentry's parent was set to itself in
2066 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2067 corresponding dentry other than use rec->ur_name.
2069 Severity : enhancement
2071 Description: lfs setstripe enhancement
2072 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2076 Description: enable data checksumming by default
2077 Details : enable checksum by default, allow --disable-checksum
2078 configure option and "-o nochecksum" mount option. Checksums
2079 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2083 Description: lustre support for RHEL5
2084 Details : Add support for RHEL5.
2088 Description: Client eviction due to failover config
2089 Details : after a connection loss, the lustre client should attempt to
2090 reconnect to the last active server first before trying the
2091 other potential connections.
2095 Description: when mds and osts use different quota unit(32bit and 64bit),
2096 quota will be released repeatly.
2097 Details : void sending multiple quota reqs to mds, which will keep the
2098 status between the reqs.
2102 Description: osts not allocated evenly to files
2103 Details : change the condition to increase offset_idx
2106 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2108 Description: Data corruption for OSTs that are formatted larger than 2TB
2110 Details : When generating the bio request for lustre file writes the
2111 sector number would overflow a temporary variable before being
2112 used for the IO. The data reads correctly from Lustre (which
2113 will overflow in a similar manner) but other file data or
2114 filesystem metadata may be corrupted in some cases.
2118 Description: Tune the kernel for good SCSI performance.
2119 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2120 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2125 Description: rewrite ext2-derived code in obdclass/uuid.c
2126 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2127 from scratch preserving functionality.
2131 Description: "lfs find -obd UUID" prints directories
2132 Details : "lfs find -obd UUID" will return all directory names instead
2133 of just file names. It is incorrect because the directories
2134 do not reside on the OSTs.
2138 Description: use i_size_read and i_size_write in 2.6 port
2139 Details : replace inode->i_size access with i_size_read/write()
2142 Frequency : when removing large files
2144 Description: scheduling issue during removal of large Lustre files
2145 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2146 It causes scheduling issues when removing large files (17TB in the
2150 Frequency : only with liblustre clients on XT3
2152 Description: evictions taking too long
2153 Details : allow llrd to evict clients directly on OSTs
2156 Frequency : only on ppc
2158 Description: /proc/fs/lustre/devices broken on ppc
2159 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2160 We should make sure the type of 'index' is loff_t and then cast
2161 explicitly as needed below. Do not assign an explicitly cast
2165 Frequency : only for rhel5
2167 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2168 Details : Modify the target file & which_kernel.
2170 Severity : enhancement
2172 Description: omit set fsid for export NFS
2173 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2174 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2178 Frequency : Always, for kernels after 2.6.16
2179 Description: Fix warning idr_remove called for id=.. which is not allocated.
2180 Details : Last kernels save old s_dev before kill super and not allow
2181 to restore from callback - restore it before call kill_anon_super.
2185 Description: buffer overruns could theoretically occur
2186 Details : llapi_semantic_traverse() modifies the "path" argument by
2187 appending values to the end of the origin string, and a
2188 overrun may occur. Adding buffer overrun check in liblustreapi.
2192 Description: Fix error on 'ls .' at the top of the Lustre mount.
2193 Details : Don't revalidate dentry if it is a root dentry.
2197 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2198 Details : Modify vm-tunables-rhel4.patch.
2202 Description: Kernel config for 2.6.18-vanilla.
2203 Details : Modify targets/2.6-vanilla.target.in.
2204 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2205 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2206 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2207 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2211 Description: improve handling recoverable errors
2212 Details : if request processig with error which can be recoverable on server
2213 request should be resend, otherwise page released from cache and
2218 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2219 Details : Modify target file & which_patch.
2220 A flaw was found in the IA32 system call emulation provided
2221 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2222 value could be stored in the %RAX register, which could trigger an
2223 out-of-bounds system call table access. An untrusted local user
2224 could exploit this flaw to run code in the kernel
2225 (ie a root privilege escalation). (CVE-2007-4573).
2229 Description: O_DIRECT bypasses client statistics.
2230 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2231 read_ahead_stats were updated. copied the stats section from
2232 osc_send_oap_rpc() into async_internal().
2236 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2240 Description: change order of libsysio includes
2241 Details : '#include sysio.h' should always come before '#include xtio.h'
2245 Description: handle "serious error: objid * already exists" more gracefully
2246 Details : If LAST_ID value on disk is smaller than the objects existing in
2247 the O/0/d* directories, it indicates disk corruption and causes an
2248 LBUG(). If the object is 0-length, then we should use the existing
2249 object. This will help to avoid a full fsck in most cases.
2251 Severity : enhancement
2253 Description: adapt the lustre_config script to support the upgrade case
2254 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2255 server targets to 1.6 in parallel.
2259 Description: To avoid grant space > avaible space when the disk is almost
2260 full. Without this patch you might see the error "grant XXXX >
2261 available" or some LBUG about grant, when the disk is almost
2263 Details : In filter_check_grant, for non_grant cache write, we should
2264 check the left space by if (*left > ungranted + bytes), instead
2265 of (*left > ungranted), because only we are sure the left space
2266 is enough for another "bytes", then the ungrant space should be
2267 increase. In client, we should update cl_avail_grant only there
2268 is OBD_MD_FLGRANT in the reply.
2272 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2273 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2274 described in CVE-2007-4573. This problem could allow untrusted
2275 local users to gain root access.
2278 Frequency : when using O_DIRECT and quotas
2280 Description: Incorrect file ownership on O_DIRECT output files
2281 Details : block usage reported by 'lfs quota' does not take into account
2282 files that have been written with O_DIRECT.
2287 Description: touch file failed when fs is not full
2288 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2289 otherwise we can get ENOSP while fs is not full.
2293 Description: parallel lock callbacks
2294 Details : Instead of sending blocking and completion callbacks as separated
2295 requests, adding them to a set and sending in parallel.
2298 Frequency : only for Cray XT3
2299 Bugzilla : 12829/13455
2300 Description: Changing primary group doesn't change the group lustre assigns to
2302 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2303 overridden with the primary group provided by the group upcall,
2304 whereas the supplied fsgid can be trusted if it is in the list of
2305 supplementary groups returned by the group upcall.
2307 Severity : enhancement
2309 Description: Allow masking D_WARNING, D_ERROR messages from console
2310 Details : Console messages can now be disabled via lnet.printk.
2314 Description: User code with malformed file open parameter crashes client node
2315 Details : Before packing join_file req, all the related reference should be
2316 checked carefully in case some malformed flags cause fake
2317 join_file req on client.
2321 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2322 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2323 mode is equal to the granted mode. However, it should also check
2324 that LDLM_FL_CP_REQD is not set.
2328 Description: Heavy nfs access might result in deadlocks
2329 Details : After ELC code landed, it is now improper to enqueue any mds
2330 locks under och_sem, because enqueue might want to decide to
2331 cancel open locks for same inode we are holding och_sem for.
2335 Description: Client eviction while running blogbench
2336 Details : A lot of unlink operations with concurrent I/O can lead to a
2337 deadlock causing evictions. To address the problem, the number of
2338 oustanding OST_DESTROY requests is now throttled to
2339 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2340 callbacks are processed in priority.
2344 Description: enable ACLs on MDS by default
2345 Details : ACLs must be enabled on MDS by default.
2348 Frequency : PPC/PPC64 only
2350 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2351 Details : fix duplicated definitions between asm-ppc64/types.h and
2352 lustre_types.h on PPC.
2355 Frequency : PPC/PPC64 only
2357 Description: asm-ppc/segment.h does not exist
2358 Details : fix compile issue on PPC.
2362 Description: better handle error messages in extents code
2365 Frequency : RHEL4 only
2367 Description: mkfs is very slow on IA64/RHEL4
2368 Details : A performance regression has been discovered in the MPT Fusion
2369 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2370 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2371 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2373 Severity : enhancement
2375 Description: SNMP support enhancement
2376 Details : Adding total number of sampled request for an MDS node in snmp
2379 Severity : enhancement
2381 Description: Optimize ldlm waiting list processing for PR extent locks
2382 Details : When processing waiting list for read extent lock and meeting read
2383 lock that is same or wider to it that is not contended, skip
2384 processing rest of the list and immediatelly return current
2385 status of conflictness, since we are guaranteed there are no
2386 conflicting locks in the rest of the list.
2390 Description: Time out and refuse to reconnect
2391 Details : When the failover node is the primary node, it is possible
2392 to have two identical connections in imp_conn_list. We must
2393 compare not conn's pointers but NIDs, otherwise we can defeat
2394 connection throttling.
2398 Description: port llog fixes from b1_6 into HEAD
2399 Details : Port llog reference couting and some llog cleanups from b1_6
2400 (bug 10800) into HEAD, for protect from panic and access to already
2401 free llog structures.
2405 Description: Detect stride IO mode in read-ahead
2406 Details : When a client does stride read, read-ahead should detect that and
2407 read-ahead pages according to the detected stride pattern.
2411 Description: data checksumming impacts single node performance
2412 Details : add support for several checksum algorithm. Currently, only CRC32
2413 and Adler-32 are supported. The checksum type can be changed on
2414 the fly via /proc/fs/lustre/osc/*/checksum_type.
2418 Description: use adler32 for page checksums
2419 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2424 Description: build for x2 fails
2425 Details : fix compile issue on Cray systems.
2429 Description: Properly match for duplicate locks
2430 Details : Due to different lock order from skiplists code, we need to
2431 traverse entire list for now
2434 Frequency : only on PPC/SLES10
2436 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2437 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2438 be defined. Add a hack in mkfs_lustre.c to work around this compile
2443 Description: LASSERT on MDS when client holding flock lock dies
2444 Details : ldlm pool logic depends on number of granted locks equal to
2445 number of released locks which is not true for flock locks, so
2446 just exclude such locks from consideration.
2450 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2451 Details : Use fsfilt_sync() for both the whole filesystem sync and
2452 individual file sync to eliminate dangerous inode locking
2453 with I_LOCK that can lead to a deadlock.
2457 Description: performance in 1.6.3
2458 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2459 sure that 1MB requests are not fragmented by the block layer.
2461 Severity : enhancement
2463 Description: organize the server-side client stats on per-nid basis
2464 Details : Change the structure of stats under obdfilter and mds to
2472 The "uuid"s file would list the uuids of _active_ exports.
2473 And the clear entry is to clear all stats and stale nids.
2475 Severity : enhancement
2477 Description: eliminate client locks in face of contention
2478 Details : file contention detection and lockless i/o implementation
2479 for contended files.
2483 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2486 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2488 Description: MDS may lose file striping (and hence file data) in some cases
2489 Details : If there are additional extended attributes stored on the MDS,
2490 in particular ACLs, SELinux, or user attributes (if user_xattr
2491 is specified for the client mount options) then there is a risk
2492 of attribute loss. Additionally, the Lustre file striping
2493 needs to be larger than default (e.g. striped over all OSTs),
2494 and an additional attribute must be stored initially in the
2495 inode and then increase in size enough to be moved to the
2496 external attribute block (e.g. ACL growing in size) for file
2501 Description: skiplist implementation simplification
2502 Details : skiplists are used to group compatible locks on granted list
2503 that was implemented as tracking first and last lock of each
2504 lock group the patch changes that to using doubly linked lists
2508 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2509 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2510 because LASSERT on both the data supplied by a client, and the
2511 data on disk is dangerous and incorrect.
2513 Severity : enhancement
2515 Description: Slow truncate/writes to huge files at high offsets.
2516 Details : Directly associate cached pages to lock that protect those pages,
2517 this allows us to quickly find what pages to write and remove
2518 once lock callback is received.
2522 Description: more ldlm soft lockups
2523 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2524 starve other threads from the resource lock for a long time in
2525 case of long waiting queue, so change the debug level from
2526 D_OTHER to the less frequently used D_INFO.
2528 Severity : enhancement
2530 Description: add -gid, -group, -uid, -user options to lfs find
2534 Description: Hung threads in invalidate_inode_pages2_range
2535 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2536 once one is completed. As a result, some RPCs are stuck in the
2537 queue and are never sent.
2541 Description: filter threads hungs on waiting journal commit
2542 Details : Cleanup filter group llog code, then only filter group llog will
2543 be only created in the MDS/OST syncing process.
2547 Description: Procfs and llog threads access destoryed import sometimes.
2548 Details : Sync the import destoryed process with procfs and llog threads by
2549 the import refcount and semaphore.
2551 Severity : enhancement
2553 Description: openlock cache of b1_6 port to HEAD
2558 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2559 Details : If insertion of an extent fails, then discard the inode
2560 preallocation and free data blocks else it can lead to duplicate
2565 Description: don't always update ctime in ext3_xattr_set_handle()
2566 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2567 In some cases the ctime should not be updated, for example for
2568 2.0->1.8 compatibility it is necessary to delete an xattr and it
2569 should not update the ctime.
2573 Bugzilla : 15713/16362
2574 Description: Assertion in iopen_connect_dentry in 1.6.3
2575 Details : looking up an inode via iopen with the wrong generation number can
2576 populate the dcache with a disconneced dentry while the inode
2577 number is in the process of being reallocated. This causes an
2578 assertion failure in iopen since the inode's dentry list contains
2579 both a connected and disconnected dentry.
2583 Description: assertion failure in ldlm_handle2lock()
2584 Details : fix a race between class_handle_unhash() and class_handle2object()
2585 introduced in lustre 1.6.5 by bug 13622.
2590 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2591 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2596 Description: oops in sd_iostats_seq_show()
2597 Details : unloading/reloading the scsi low level driver triggers a kernel
2598 bug when trying to access the sd iostat file.
2603 Description: Kernel panics during QLogic driver reload
2604 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2605 patch, causing memory corruption.
2610 Description: journal_dev option does not work in b1_6
2611 Details : pass mount option during pre-mount.
2613 Severity : enhancement
2615 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2616 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2617 extent information of a file. It can be used to map logical blocks
2618 in a file to physical blocks in the block device.
2622 Description: LDLM soft lockups - improvement
2623 Details : It is be possible to send the lock handle along with each read
2624 or write request because the client is already doing a lock match
2625 itself so there isn't any reason the OST should have to re-do that
2629 Frequency : only on Cray X2
2631 Description: X2 build failures
2632 Details : fix build failures on Cray X2.
2636 Description: xid & resent requests
2637 Details : Initialize RPC XID from clock at startup (randomly if clock is
2640 Severity : enhancement
2642 Description: Add lustre_start utility to start or stop multiple Lustre servers
2647 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2648 Details : In case of memory pressure, list_del() can be called twice on
2649 req->rq_history_list, causing a kernel oops.
2653 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2654 Details : fix stack overflow in the distributed lock manager by defering
2655 export eviction after a failed AST to the elt thread instead of
2656 handling it in the dlm interpret routine.
2660 Description: Convert some comments to new format.
2661 Details : Update documenting comments to match doxygen conventions.
2665 Description: Grammar fixes.
2666 Details : A couple of trivial spelling fixes.
2670 Description: OSD_COUNTERS-mandatory
2671 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2672 to simplify the code.
2676 Description: simplify cmm_device freeing logic.
2677 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2681 Description: Add lockdep support to dt_object_operations locking interface.
2682 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2683 parameter indicating lock ordering. Update mdd code to use new
2688 Description: Introduce failloc constants for lockless IO tests.
2689 Details : Add two new failloc constants to test lockless IO. Only one of
2690 them in implemented---another is checked in yet to be landed
2695 Description: Add lockdep support for inode mutex.
2696 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2697 in the situations where Lustre has to lock more than one inode
2702 Description: Add optional invariants checking support.
2703 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2704 configure switch. This macro is to be used for consistency and
2705 sanity checks that are too expensive to be left in `production'
2710 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2711 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2716 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2717 Details : Use spin_lock_nested() in (the only) situation where more than
2718 one ldlm_lock is locked simultaneously. Also, fix possible
2719 dead-lock in ldlm_lock_change_resource() by enforcing particular
2724 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2725 Details : Instead of specifying each ldlm_lock call-back through separate
2726 parameter, wrap them into struct ldlm_callback_suite.
2730 Description: Kill join_lru obd method and its callers.
2731 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2732 in memory---a requirement implemented through ->o_join_lru() obd
2733 method in HEAD. Remove this method and its users.
2737 Description: Add asynchronous ldlm ENQUEUE completion handler.
2738 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2739 case that stock ldlm_completion_ast() cannot handle as it waits
2740 until lock is granted. Introduce new ldlm_completion_ast_async()
2741 for this. Also comment ldlm_completion_ast().
2745 Description: ldlm_error <-> errno conversion.
2746 Details : Add functions to map (rather arbitrary) between LDLM error codes
2747 and standard errno values. CLIO needs this to prevent LDLM specific
2748 constants from escaping ldlm and osc.
2752 Description: Kill unused ldlm_handle2lock_ns() function.
2753 Details : Kill unused ldlm_handle2lock_ns() function.
2757 Description: Add lu_ref support to ldlm_lock
2758 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2759 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2760 and ldlm_resource. LDLM interface has to be changed, because code that
2761 releases a reference on a lock, has to "know" what reference this is.
2762 In the most frequent case
2764 lock = ldlm_handle2lock(handle);
2766 LDLM_LOCK_PUT(lock);
2768 no changes are required. When any other reference (received _not_ from
2769 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2770 instead of LDLM_LOCK_PUT().
2772 Arguably, changes are pervasive, and interface requires some discipline
2773 for proper use. On the other hand, it was very instrumental in finding
2774 a few leaked lock references.
2778 Description: Add ldlm_lock_addref_try().
2779 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2780 attempts to addref a lock that might be being canceled
2785 Description: Add ldlm_weigh_callback().
2786 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2787 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2788 hard-coded `number of pages' logic.
2792 Description: Add lockdep annotations to llog code.
2793 Details : Use appropriately tagged _nested() locking calls in the places
2794 where llog takes more than one ->lgh_lock lock.
2798 Description: Add loi_kms_set().
2799 Details : Wrap kms updates into a helper function.
2803 Description: Constify instances of struct lsm_operations.
2804 Details : Constify instances of struct lsm_operations.
2808 Description: lu_conf support.
2809 Details : On a server, a file system object is uniquely identified
2810 by a fid, which is sufficient to locate and load all object
2811 state (inode). On a client, on the other hand, more data are
2812 necessary instantiate an object. Change lu_object_find() and
2813 friends to take additional `lu_conf' argument describing object.
2814 Typically this includes layout information.
2818 Description: lu_context fixes.
2819 Details : Introduce new lu_context functions that are needed on the client
2820 side, where some system threads (ptlrpcd) are shared by multiple
2821 modules, and so cannot be stopped during module shutdown.
2825 Description: Add start and stop methods to lu_device_type_operations.
2826 Details : Introduce two new methods in lu_device_type_operations, that are
2827 invoked when first instance of a given type is created and last one
2828 is destroyed respectively. This is need by CLIO.
2832 Description: Add lu_ref support to struct lu_device.
2833 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2834 track leaked references.
2838 Description: Introduce lu_kmem_descr.
2839 Details : lu_kmem_descr and its companion interface allow to create
2840 and destroy a number of kmem caches at once.
2844 Description: Fix lu_object finalization race.
2845 Details : Fix a race between lu_object_find() finding an object and its
2846 concurrent finalization. This race is (most likely) not possible
2847 on the server, but might happen on the client.
2851 Description: Introduce lu_ref interface.
2852 Details : lu_ref is a debugging module allowing to track references to
2853 a given object. It is quite cpu expensive, and has to be
2854 explicitly enabled with --enable-lu_ref. See usage description
2859 Description: Factor lu_site procfs stats into a separate function.
2860 Details : Separate lu_site stats printing code into a separate function
2861 that can be reused on a client.
2865 Description: Constify instances of struct {lu,dt,md}_device_operations.
2866 Details : Constify instances of struct {lu,dt,md}_device_operations.
2870 Description: Introduce struct md_site and move meta-data specific parts of
2871 struct lu_site here.
2872 Details : Move md-specific fields out of struct lu_site into special struct
2873 md_site, so that lu_site can be used on a client.
2877 Description: Kill mdd_lov_destroy().
2878 Details : Remove unused mdd code.
2882 Description: Add st_block checking to multistat.c.
2883 Details : Add st_block checking to multistat.c.
2887 Description: Add lu_ref support to struct obd_device.
2888 Details : Add lu_ref tracking to obd_device.
2892 Description: Kill obd_set_fail_loc().
2893 Details : Remove unused code.
2897 Description: Add special type for ptlrpc_request interpret functions.
2898 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2899 there. Actual usage will be in CLIO.
2903 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2904 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2905 doesn't work with lockdep.
2909 Description: Add rwv.c test program.
2910 Details : New testing program exercising readv(2) and writev(2) (Qian).
2914 Description: Add sendfile.c test program.
2915 Details : New testing program exercising sendfile(2) (Jay).
2919 Description: Ratelimit a message that can be very frequent.
2920 Details : Ratelimit a memory allocation failure message that can
2925 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2926 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2930 Description: Liblustre build fixes.
2931 Details : Liblustre build fixes.
2935 Description: libcfs: add cfs_{need,cond}_resched() interface.
2936 Details : libcfs: add cfs_{need,cond}_resched() definition and
2937 implementations for Linux, NT, and liblustre.
2939 Severity : enhancement
2941 Description: More exported tunables for mballoc
2942 Details : Add support for tunable preallocation window and new tunables for
2943 large/small requests
2947 Description: Detect corruption of block bitmap and checking for preallocations
2948 Details : Checks validity of on-disk block bitmap. Also it does better
2949 checking of number of applied preallocations. When corruption is
2950 found, it turns filesystem readonly to prevent further corruptions.
2954 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2955 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2956 window to 0. If it does hit cache miss, and read-pattern is still
2957 stride-io mode, does not reset the stride window, but also does not
2958 increase the stride window length in this case.
2962 Frequency : only for big-endian servers
2963 Description: Check if system is big-endian while mounting fs with extents feature
2964 Details : Mounting a filesystem with extents feature will fail on big-endian
2965 systems since ext3-based ldiskfs is not supported on big-endian
2966 systems. This can be over-riden with "bigendian_extents" mount option.
2968 Severity : enhancement
2970 Description: The root squash functionality
2971 Details : A security feature, which is to prevent users from being able
2972 to mount lustre on their desktop, run as root, and delete
2973 all of the files in the filesystem. The goal is accomplished by
2974 remapping user id (UID) and group id (GID) of the root user to
2975 a UID and GID specified by the system administartor via Lustre
2976 configuration management server (MGS). The functionality also
2977 allows to specify sets of clients for which the remapping does
2982 Description: Excessive recovery window
2983 Details : With AT enabled, the recovery window can be excessively long (6000+
2984 seconds). To address this problem, we no longer use
2985 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2986 timeout no longer depends on the service time, it is set to
2987 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2988 time via pb_service_time.
2992 Description: Watchdog triggered on MDS failover
2993 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2994 the OSTs know that the MDS "UUID" can be reused for the same export
2995 from a different NID, so we do not need to wait for the export to be
2999 Frequency : rare, only if using MMP with Linux RAID
3001 Description: MMP doesn't work with Linux RAID
3002 Details : While using HA for Lustre servers with Linux RAID, it is possible
3003 that MMP will not detect multiple mounts. To make this work we
3004 need to unplug the device queue in RAID when the MMP block is being
3005 written. Also while reading the MMP block, we should read it from
3006 disk and not the cached one.
3008 Severity : enhancement
3010 Description: open file using fid
3011 Details : A file can be opened using just its fid, like
3012 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3015 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3018 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3019 Details : add proper locking for creation of jbd_4k slab cache
3023 Description: MMP check in ext3_remount() fails without displaying any error
3024 Details : When multiple mount protection fails during remount, proper error
3027 Severity : enhancement
3029 Description: Allow stripe size to be up to 4G-64k
3030 Details : Fix math logic to allow large stripe sizes.
3034 Description: add check for >8TB ldiskfs filesystems
3035 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3036 Don't allow >8TB ldiskfs filesystems to be mounted without
3037 force_over_8tb mount option
3042 Descriptoin: OSS complains "too many blocks" and crashes
3043 Details : In filter_commitrw_write():
3044 1. lquota_chkquota();
3045 2. obdo_from_inode();
3046 3. lquota_pending_commit();
3047 "2." has changed uid and gid in this case, then caused
3048 lquota_chkquota() and lquota_pending_commit() to use
3049 different uid and gid. That is the root of the bug.
3053 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3054 Details : Detect and print out the 'stripe count = -1' case properly.
3058 Description: Out or order replies might be lost on replay
3059 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3060 request with tid smaller than one currently being added, add it
3061 to the start, not end of the list.
3063 --------------------------------------------------------------------------------
3065 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3067 * Support for kernels:
3068 2.6.5-7.283 (SLES 9),
3069 2.6.9-55.EL (RHEL 4),
3070 2.6.16.46-0.14 (SLES 10),
3071 2.6.18.8 vanilla (kernel.org)
3072 * Client support for unpatched kernels:
3073 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3074 2.6.16 - 2.6.22 vanilla (kernel.org)
3075 * Due to recently discovered recovery problems, we do not recommend
3076 using patchless RHEL 4 clients with this or any earlier release.
3077 * Recommended e2fsprogs version: 1.39.cfs8
3078 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3079 * Starting with this release, the ldiskfs backing filesystem required
3080 by Lustre is now in its own package, lustre-ldiskfs. This package
3081 should be installed. It is versioned separately from Lustre and
3082 may be released separately in future.
3086 Description: block reactivating mgc import until all deactivates complete
3087 Details : Fix race when failing back MDT/MGS to itself (testing)
3090 Frequency : at statup only
3092 Description: mds_lov_synchronize race leads to various problems
3093 Details : simultaneous MDT->OST connections at startup can cause the
3094 sync to abort, leaving the OSC in a bad state.
3096 Severity : enhancement
3098 Description: add optional extra BUILD_VERSION info
3099 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3100 to override the lustre version.
3103 Frequency : 2.6.18 servers only
3105 Description: ll_kern_mount() doesn't release the module reference
3106 Details : The ldiskfs module reference count never drops down to 0
3107 because ll_kern_mount() doesn't release the module reference.
3112 Description: server LBUG when using old ost_num_threads parameter
3113 Details : Accept the old ost_num_threads parameter but warn that it
3114 is deprecated, and fix an off-by-one error that caused an LBUG.
3119 Description: Transient SCSI error results in persistent IO issue
3120 Details : iobuf->dr_error is not reinitialized to 0 between two
3124 Frequency : sometimes when underlying device returns I/O errors
3126 Description: OSTs not going read-only during write failures
3127 Details : OSTs are not remounted read-only when the journal commit threads
3128 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3129 instead of the ext3 wrappers.
3133 Description: poor connect scaling with increasing client count
3134 Details : Don't run filter_grant_sanity_check for more than 100 exports
3135 to improve scaling for large numbers of clients.
3138 Frequency : SLES10 only
3140 Description: sanity-quota.sh quotacheck failed: rc = -22
3141 Details : Quotas cannot be enabled on SLES10.
3144 Frequency : liblustre clients only
3146 Description: getdirentries does not give error when run on compute nodes
3147 Details : getdirentries does not fail when the size specified as an argument
3148 is too small to contain at least one entry
3150 Severity : enhancement
3152 Description: Add LNET router traceability for debug purposes
3153 Details : If a checksum failure occurs with a router as part of the
3154 IO path, the NID of the last router that forwarded the bulk data
3155 is printed so it can be identified.
3160 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3161 Details : Due to a race condition, liblustre clients were occasionally
3162 evicted incorrectly.
3164 Severity : enhancement
3166 Description: lfs setstripe use optional parameters instead of postional
3169 Severity : enhancement
3171 Description: Nanosecond timestamp support for ldiskfs
3172 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3173 resolution timestamps. There is not yet support for this at
3174 the Lustre filesystem level.
3177 Frequency : during server recovery
3179 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3180 Details : request with rq_no_resend flag not awake l_wait_event if they get
3184 Frequency : nfs export on patchless client
3186 Description: connectathon hang when test nfs export over patchless client
3187 Details : Disconnected dentry cannot be found with lookup, so we do not need
3188 to unhash it or make it invalid
3191 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3192 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3193 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3198 Description: open req refcounting wrong on reconnect
3199 Details : If reconnect happened between getting open reply from server and
3200 call to mdc_set_replay_data in ll_file_open, we will schedule
3201 replay for unreferenced request that we are about to free.
3202 Subsequent close will crash in variety of ways.
3203 Check that request is still eligible for replay in
3204 mdc_set_replay_data().
3209 Description: disable writes to filesystem when reading health_check file
3210 Details : the default for reading the health_check proc file has changed
3211 to NOT do a journal transaction and write to disk, because this
3212 can cause reads of the /proc file to hang and block HA state
3213 checking on a healthy but otherwise heavily loaded system. It
3214 is possible to return to the previous behaviour during configure
3215 with --enable-health-write.
3217 Severity : enhancement
3219 Description: 64-bit inode version
3220 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3221 to the inode. This will be required for version-based recovery.
3226 Description: MDS fails to start if a duplicate client export is detected
3227 Details : in some rare cases it was possible for a client to connect to
3228 an MDS multiple times. Upon recovery the MDS would detect this
3229 and fail during startup. Handle this more gracefully.
3231 Severity : enhancement
3233 Description: Add -o localflock option to simulate old noflock
3235 Details : This will achieve local-only flock/fcntl locks
3241 Description: log_commit_thread vs filter_destroy race leads to crash
3242 Details : Take import reference before releasing llog record semaphore
3247 Description: Wrong request locking in request set processing
3248 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3249 imp_delayed_list, in this place should be used imp_lock.
3252 Frequency : when reconnection
3254 Description: Grant Leak when osc reconnect to OST
3255 Details : When osc reconnect ost, OST(filter) should check whether it
3256 should grant more space to client by comparing fed_grant and
3257 cl_avail_grant, and return the granted space to client instead
3258 of "new granted" space, because client will call osc_init_grant
3259 to update the client grant space info.
3262 Frequency : when client reconnect to OST
3264 Description: Grant Leak when osc do resend and replay bulk write
3265 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3266 bulk write request, because the grant info while be sync between
3267 OSC and OST when reconnect, and we should ignore the grant info
3268 these of resend/replay write req.
3273 Description: Grant space more than available left space sometimes.
3274 Details : When then OST is about to be full, if two bulk writing from
3275 different clients came to OST. Accord the avaliable space of the
3276 OST, the first req should be permitted, and the second one
3277 should be denied by ENOSPC. But if the seconde arrived before
3278 the first one is committed. The OST might wrongly permit second
3279 writing, which will cause grant space > available space.
3282 Frequency : when client is evicted
3284 Description: Grant might be wrongly erased when osc is evicted by OST
3285 Details : when the import is evicted by server, it will fork another
3286 thread ptlrpc_invalidate_import_thread to invalidate the
3287 import, where the grant will be set to 0. While the original
3288 thread will update the grant it got when connecting. So if
3289 the former happened latter, the grant will be wrongly errased
3290 because of this race.
3295 Description: Checking Stale with correct fid
3296 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3297 to check whether it is stale, because sometimes, we want the
3298 enqueue happened anyway, and op_data.fid2 will not be initialized.
3300 Severity : enhancement
3302 Description: update patchless client
3303 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3306 Frequency : only with 2.4 kernel
3308 Description: random memory corruption
3309 Details : size of struct ll_inode_info is to big for union inode.u and this
3310 can be cause of random memory corruption.
3315 Description: Memory leak in recovery
3316 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3317 It should also check obd_fail before fsfilt_start, otherwise if
3318 fsfilt_start return -EROFS,(failover mds during mds recovery).
3319 then the req will return with repmsg->transno = 0 and rc = EROFS.
3320 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3321 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3322 be freed no matter whether fsfilt_commit success or not.
3325 Frequency : only with huge count clients
3327 Description: Prevents from taking the superblock lock in llap_from_page for
3329 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3330 for llap_from_page call in ll_removepage prevents from taking the
3331 superblock lock for a soon died page.
3336 Description: Not check open intent error before release open handle
3337 Details : in some rare cases, the open intent error is not checked before
3338 release open handle, which may cause
3339 ASSERTION(open_req->rq_transno != 0), because it tries to release
3340 the failed open handle.
3345 Description: Set cat log bitmap only after create log success.
3346 Details : in some rare cases, the cat log bitmap is set too early. and it
3347 should be set only after create log success.
3351 Description: Accessing a block bevice can re-enable I/O when Lustre is
3352 tearing down a device.
3353 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3358 Description: service threads may hog cpus when there are a lot of requests
3359 Details : Insert cond_resched to give other threads a chance to use some CPU
3364 Description: the cat log was not initialized in recovery
3365 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3366 unlink log on mds will not be initialized until mds post
3367 recovery. And also in mds post recovery, the unlink log will
3368 initialization will be done asynchronausly, so there will be race
3369 between add unlink log and unlink log initialization.
3373 Description: brw_stats were being printed incorrectly
3374 Details : brw_stats were being printed as log2 but all of them were not
3375 recorded as log2. Also remove some code duplication arising from
3376 filter_tally_{read,write}.
3380 Frequency : rare, only in recovery.
3381 Description: ASSERTION(req->rq_type != LI_POISON) failed
3382 Details : imp_lock should be held while iterating over imp_sending_list for
3383 prevent destroy request after get timeout in ptlrpc_queue_wait.
3387 Description: replay-single.sh test 52 fails
3388 Details : A lock's skiplist need to be cleanup when it being unlinked
3389 from its resource list.
3393 Description: Short directio read returns full requested size rather than
3395 Details : Direct I/O operations should return actual amount of bytes
3396 transferred rather than requested size.
3398 Severity : enhancement
3400 Description: metadata RPC reduction (e.g. for rm performance)
3401 Details : decrease the amount of synchronous RPC between clients and servers
3402 by canceling conflicing lock before the operation on the client
3403 and packing thier handles into the main operation RPC to server.
3405 Severity : enhancement
3407 Description: Async OSC create to avoid the blocking unnecessarily.
3408 Details : If an OST has no remaining object, system will block on the
3409 creation when it needs to create a new object on this OST. Now,
3410 ways use pre-created objects when available, instead of blocking on
3411 an empty osc while others are not empty. If we must block, we
3412 block for the shortest possible period of time.
3416 Description: improve handling recoverable errors
3417 Details : if request processig with error which can be recoverable on server
3418 request should be resend, otherwise page released from cache and
3421 Severity : enhancement
3423 Description: refine locking for avoid write wrong info into lov_objid file
3424 Details : fix possible races with add new target and write/update data in
3427 --------------------------------------------------------------------------------
3429 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3434 Frequency : on some architectures
3436 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3437 Details : Uninitialized flags sometimes cause configuration commands to
3441 Frequency : patchless clients only
3443 Description: missing __iget() symbol export
3444 Details : The __iget() symbol export is missing. To avoid the need for
3445 this on patchless clients the deathrow inode reaper is turned
3446 off, and we depend on the VM to clean up old inodes. This
3447 dependency was during via the fix for bug 12181.
3449 --------------------------------------------------------------------------------
3451 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3453 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3454 INTEROPERATE with older versions automatically. In many cases a
3455 special upgrade step is needed. Please read the
3456 user documentation before upgrading any part of a live system.
3457 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3458 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3459 * WARNING: Lustre configuration and startup changes are required with
3460 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3464 Severity : enhancement
3466 Description: MountConf
3467 Details : Lustre configuration is now managed via mkfs and mount
3468 commands instead of lmc and lconf. New obd types (MGS, MGC)
3469 are added for dynamic configuration management. See
3470 https://mail.clusterfs.com/wikis/lustre/MountConf for
3473 Severity : enhancement
3475 Description: dynamic OST addition
3476 Details : OSTs can now be added to a live filesystem
3478 Severity : enhancement
3480 Description: startup order invariance
3481 Details : MDTs and OSTs can be started in any order. Clients only
3482 require the MDT to complete startup.
3484 Severity : enhancement
3486 Description: parallel, asynchronous orphan cleanup
3487 Details : orphan cleanup is now performed in separate threads for each
3488 OST, allowing parallel non-blocking operation.
3490 Severity : enhancement
3492 Description: optimized stripe assignment
3493 Details : stripe assignments are now made based on ost space available,
3494 ost previous usage, and OSS previous usage, in order to try
3495 to optimize storage space and networking resources.
3497 Severity : enhancement
3499 Description: Permanently set tunables
3500 Details : All writable /proc/fs/lustre tunables can now be permanently
3501 set on a per-server basis, at mkfs time or on a live system.
3503 Severity : enhancement
3505 Description: Lustre message v2
3506 Details : Add lustre message format v2.
3508 Severity : enhancement
3510 Description: client OST exclusion list
3511 Details : Clients can be started with a list of OSTs that should be
3512 declared "inactive" for known non-responsive OSTs.
3516 Description: SPEC SFS validation failure on NFS v2 over lustre.
3517 Details : Changes the blocksize for regular files to be 2x RPC size,
3518 and not depend on stripe size.
3520 Severity : enhancement
3522 Description: Multiple MD RPCs in flight.
3523 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3524 To avoid overly-overloading MDS, introduce a limit on number of
3525 MDS RPCs in flight for a single client and add /proc controls
3526 to adjust this limit.
3528 Severity : enhancement
3530 Description: client read/write statistics
3531 Details : Add client read/write call usage stats for performance
3532 analysis of user processes.
3533 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3534 file access. extents_stats shows chunk size distribution.
3535 extents_stats_per_process show chunk size distribution per
3538 Severity : enhancement
3540 Description: mds statistics
3541 Details : Add detailed mds operations statistics in
3542 /proc/fs/lustre/mds/*/stats.
3546 Description: Failure of copying files with lustre special EAs.
3547 Details : Client side always return success for setxattr call for lustre
3548 special xattr (currently only "trusted.lov").
3553 Description: Refcount LNET uuids
3554 Details : The global LNET uuid list grew linearly with every startup;
3555 refcount repeated list entries instead of always adding to
3558 Severity : enhancement
3560 Description: Dynamic service threads
3561 Details : Within a small range, start extra service threads
3562 automatically when the request queue builds up.
3565 Frequency : mixed-endian client/server environments
3567 Description: mixed-endian crashes
3568 Details : The new msg_v2 system had some failures in mixed-endian
3571 Severity : enhancement
3573 Description: Easy OST removal
3574 Details : OSTs can be permanently deactivated with e.g. 'lctl
3575 conf_param lustre-OST0001.osc.active=0'
3577 Severity : enhancement
3579 Description: MGS proc entries
3580 Details : Added basic proc entries for the MGS showing what filesystems
3583 Severity : enhancement
3585 Description: provide MGS failover
3586 Details : Added config lock reacquisition after MGS server failover.
3588 Severity : enhancement
3590 Description: add Linux 2.4 support
3591 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3595 Description: a large application tries to do I/O to the same resource and dies
3596 in the middle of it.
3597 Details : Check the req->rq_arrival time after the call to
3598 ost_brw_lock_get(), but before we do anything about
3599 processing it & sending the BULK transfer request. This
3600 should help move old stale pending locks off the queue as
3601 quickly as obd_timeout.
3604 Frequency : when an incorrect nid is specified during startup
3606 Description: ptlrpc connect to non-existant node causes kernel crash
3607 Details : LNET can't be re-entered from an event callback, which
3608 happened when we expire a message after the export has been
3609 cleaned up. Instead, hand the zombie cleanup off to another
3612 Severity : enhancement
3614 Description: plain/inodebits lock performance improvement
3615 Details : Grouping plain/inodebits in granted list by their request modes
3616 and bits policy, thus improving the performance of search through
3620 Frequency : only if OST filesystem is corrupted
3622 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3623 Details : for a short time RPCs with bulk IO are in the replay list,
3624 but replay of bulk IOs is unimplemented. If the OST filesystem
3625 is corrupted due to disk cache incoherency and then replay is
3626 started it is possible to trip an assertion. Avoid putting
3627 committed RPCs into the replay list at all to avoid this issue.
3630 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3632 Description: System hang on startup
3633 Details : This bug allowed the liblustre (e.g. catamount) client to
3634 return to the app before handling all startup RPCs. This
3635 could leave the node unresponsive to lustre network traffic
3636 and manifested as a server ptllnd timeout.
3638 Severity : enhancement
3640 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3641 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3642 boolean to control whether to print peer debug info when a
3643 client's RPC times out.
3646 Frequency : only for kernels with patches from Lustre below 1.4.3
3648 Description: Remove old rdonly API
3649 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3652 Frequency : only for devices with external journals
3654 Description: Set external device read-only also
3655 Details : During a commanded failover stop, we set the disk device
3656 read-only while the server shuts down. We now also set any
3657 external journal device read-only at the same time.
3660 Frequency : when upgrading from 1.4 while trying to change parameters
3662 Description: The wrong (new) MDC name was used when setting parameters for
3663 upgraded MDT's. Also allows changing of OSC (and MDC)
3664 parameters if --writeconf is specified at tunefs upgrade time.
3667 Frequency : when setting specific ost indicies
3669 Description: QOS code breaks on skipped indicies
3670 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3671 created with --index need not be sequential.
3673 Severity : enhancement
3675 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3676 Details : The uninit_groups feature works in conjunction with the kernel
3677 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3678 up the pass1 processing of e2fsck. This is a read-only feature
3679 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3680 mount filesystems that have had this feature enabled.
3682 Severity : enhancement
3684 Description: Improve multi-block allocation algorithm to avoid fragmentation
3685 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3686 allocation locality and avoid filesystem fragmentation.
3688 ------------------------------------------------------------------------------
3690 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3692 * Support for kernels:
3693 2.4.21-47.0.1.EL (RHEL 3)
3694 2.6.5-7.283 (SLES 9)
3695 2.6.9-42.0.10.EL (RHEL 4)
3696 2.6.12.6 vanilla (kernel.org)
3697 2.6.16.27-0.9 (SLES 10)
3698 * Recommended e2fsprogs version: 1.39.cfs5
3700 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3704 Frequency : occasional, depends on client load and configuration
3705 Bugzilla : 12181, 12203
3706 Description: data loss for recently-modified files
3708 Details : In some cases it is possible that recently written or created
3709 files may not be written to disk in a timely manner (this should
3710 normally be within 30s unless client IO load is very high).
3711 The problem appears as zero-length files or files that are a
3712 multiple of 1MB in size after a client crash or client eviction
3713 that are missing data at the end of the file.
3715 This problem is more likely to be hit on clients where files are
3716 repeatedly created and unlinked in the same directory, clients
3717 have a large amount of RAM, have many CPUs, the filesystem has
3718 many OSTs, the clients are rebooted frequently, and/or the files
3719 are not accessed by other nodes after being written.
3721 The presence of the problem can be detected by looking at
3722 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3723 smaller than the second (nr_unused) then dirty files will not
3724 be flushed automatically to disk. "sync; sleep 10" should be
3725 run several times on the node before unmounting it to update
3726 Lustre (this is also safe to run on nodes without this problem).
3728 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3729 that can cause this same problem, so customers using that kernel
3730 also need to update the kernel in addition to Lustre. In order
3731 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3733 It is normal that files written just before a client crash (less
3734 than 30s) may not yet have been flushed to disk, even for local
3738 Frequency : frequent on thin XT3 nodes
3740 Description: UUID collision on thin XT3 Linux nodes
3741 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3742 randomly, since we relied on an insufficiently-seeded PRNG.
3747 Description: OSS hangs after "All ost request buffers busy"
3748 Details : A deadlock between quota and journal operations caused OSS
3749 hangs after printing "All ost request buffers busy."
3752 Frequency : always on liblustre builds
3754 Description: Cleanup compiler warnings on liblustre
3757 Frequency : always on liblustre builds on XT3
3759 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3763 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3764 > 32000 subdirectories
3765 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3766 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3767 subdirectory count crosses 32000. This will aid e2fsck to
3768 correctly handle more than 32000 subdirectories.
3771 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3773 Description: System hang on startup
3774 Details : This bug allowed the liblustre (e.g. catamount) client to
3775 return to the app before handling all startup RPCs. This
3776 could leave the node unresponsive to lustre network traffic
3777 and manifested as a server ptllnd timeout.
3779 Severity : enhancement
3781 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3782 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3783 boolean to control whether to print peer debug info when a
3784 client's RPC times out.
3789 Description: make O_SYNC working on 2.6 kernels
3790 Details : 2.6 kernels use different method for mark pages for write,
3791 so need add a code to lustre for O_SYNC work.
3796 Description: Failure to close file and release space on NFS
3797 Details : Put inode details into lock acquired in ll_intent_file_open.
3798 Use mdc_intent_lock in ll_intent_open to properly
3799 detect all kind of errors unhandled by mdc_enqueue.
3804 Description: proc file read during shutdown sometimes raced obd removal,
3806 Details : Add lock to prevent obd access after proc file removal.
3809 Frequency : Only for files larger than 4GB on 32-bit clients.
3811 Description: improperly doing page alignment of locks
3812 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3813 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3818 Description: RPCs being resent when they shouldn't be
3819 Details : Some RPCs that should not be resent are being resent. This
3820 can cause inconsistencies in the RPC state machine. Do not
3821 resend such requests.
3824 Frequency : rare, only with NFS export
3826 Description: Crash on NFS re-export node
3827 Details : under very unusual load conditions an assertion is hit in
3828 ll_intent_file_open()
3831 Frequency : only if OST filesystem is corrupted
3833 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3834 Details : for a short time RPCs with bulk IO are in the replay list,
3835 but replay of bulk IOs is unimplemented. If the OST filesystem
3836 is corrupted due to disk cache incoherency and then replay is
3837 started it is possible to trip an assertion. Avoid putting
3838 committed RPCs into the replay list at all to avoid this issue.
3843 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3844 Details : Large single O_DIRECT read and write calls can fail to allocate
3845 a sufficiently large buffer to process the request. In case of
3846 allocation failure the allocation is retried with a smaller
3847 buffer and broken into smaller requests.
3849 Severity : enhancement
3851 Description: Add -o localflock option to simulate old noflock behaviour.
3852 Details : This will achieve local-only flock/fcntl locks coherentness.
3857 Description: versioning check is incomplete
3858 Details : Checking the version difference of client vs. server, report
3859 error if the gap is too big.
3864 Description: add support PG_writeback bit
3865 Details : add support for PG_writeback bit for Lustre, for more carefull
3866 work with page cache in 2.6 kernel. This also fix some deadlocks
3867 and remove hack for work O_SYNC with 2.6 kernel.
3869 Severity : enhancement
3871 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3872 Details : The uninit_groups feature works in conjunction with the kernel
3873 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3874 up the pass1 processing of e2fsck. This is a read-only feature
3875 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3876 mount filesystems that have had this feature enabled.
3878 Severity : enhancement
3880 Description: Improve multi-block allocation algorithm to avoid fragmentation
3881 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3882 allocation locality and avoid filesystem fragmentation.
3884 ------------------------------------------------------------------------------
3886 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3888 * Support for kernels:
3889 2.6.9-42.0.3.EL (RHEL 4)
3890 2.6.5-7.276 (SLES 9)
3891 2.4.21-47.0.1.EL (RHEL 3)
3892 2.6.12.6 vanilla (kernel.org)
3893 2.6.16.21-0.8 (SLES10)
3894 * Recommended e2fsprogs version: 1.39.cfs2-0
3896 * The backwards-compatible /proc/sys/portals symlink has been removed
3897 in this release. Before upgrading, please ensure that you change
3898 any configuration scripts or /etc/sysctl.conf files that access
3899 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3900 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3901 in advance of the upgrade on any system running Lustre 1.4.6 or
3902 newer, since /proc/sys/lnet was added in that version.
3903 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3907 Frequency : only when quota is used
3909 Description: avoid scanning export list for quota master
3910 Details : Change the algorithms to avoid scanning export list in order
3911 to improve the efficiency.
3914 Frequency : MDS failover only, very rarely
3916 Description: "went back in time" messages on mds failover
3917 Details : The greatest transno may be lost when the current operation
3918 finishes with an error (transno==0) and the client's last_rcvd
3919 record is over-written. Save the greatest transno in the
3920 mds_last_transno for this case.
3923 Frequency : always for specific kernels and striping counts
3925 Description: client may get "Matching packet too big" without ACL support
3926 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3927 when trying to access files in certain configurations. The
3928 clients should in fact be denied when mounting because they do
3929 not understand ACLs.
3932 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3934 Description: many clients connecting with IO in progress causes connect timeouts
3935 Details : Avoid synchronous journal commits to avoid delays caused by many
3936 clients connecting/disconnecting when bulk IO is in progress.
3937 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3938 OST_IO_PORTAL to avoid delays behind potentially many pending
3942 Frequency : occasionally with multiple writers to a single file
3944 Description: shared writes to file may result in wrong size reported by stat()
3945 Details : Allow growing of kms when extent lock is cancelled
3948 Frequency : always with random mmap IO to multi-striped file
3950 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3951 Details : Only if the hole is at the end of OST object so that kms is too
3952 small. Fix is to increase kms accordingly in ll_nopage.
3955 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3957 Description: writes to a missing object would leak memory on the OST
3958 Details : If there is an inconsistency between the MDS and OST filesystems,
3959 such that the MDS references an object that doesn't exist, writes
3960 to that object will leak memory due to incorrect cleanup in the
3961 error handling path, eventually running out of memory on the OST.
3966 Description: Creating too long symlink causes lustre errors
3967 Details : Check symlink and name lengths before sending requests to MDS.
3970 Frequency : only if flock is enabled (not on by default)
3972 Description: posix locks not released on fd closure on 2.6.9+
3973 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3974 kernels, this caused such locks not to be released on fd close and
3975 then assertions on fs unmount about still used locks.
3978 Frequency : MDS failover only, very rarely
3980 Description: clients may get ASSERTION(granted_lock != NULL)
3981 Details : When request was taking a long time, and a client was resending
3982 a getattr by name lock request. The were multiple lock requests
3983 with the same client lock handle and
3984 mds_getattr_name->fixup_handle_for_resent_request found one of the
3985 lock handles but later failed with ASSERTION(granted_lock != NULL).
3990 Description: handle->h_buffer_credits > 0, assertion failure
3991 Details : h_buffer_credits is zero after truncate, causing assertion
3992 failure. This patch extends the transaction or creates a new
3996 Frequency : NFS re-export or patchless client
3997 Bugzilla : 11179, 10796
3998 Description: Crash on NFS re-export node (__d_move)
3999 Details : We do not want to hash the dentry if we don't have a lock.
4000 But if this dentry is later used in d_move, we'd hit uninitialised
4001 list head d_hash, so we just do this to init d_hash field but
4002 leave dentry unhashed.
4005 Frequency : NFS re-export or patchless client
4007 Description: NFS exports has problem with symbolic link
4008 Details : lustre client didn't properly install dentry when re-exported
4009 to NFS or running patchless client.
4012 Frequency : NFS re-export or patchless client
4014 Description: Various nfs/patchless fixes.
4015 Details : fixes reuse disconected alias for lookup process - this fixes
4016 warning "find_exported_dentry: npd != pd",
4017 fix permission error with open files at nfs.
4018 fix apply umask when do revalidate.
4021 Frequency : occasional
4023 Description: Crash on NFS re-export node
4024 Details : calling clear_page() on the wrong pointer triggered oops in
4025 generic_mapping_read().
4028 Frequency : rarely, using O_DIRECT IO
4030 Description: unaligned directio crashes client with LASSERT
4031 Details : check for unaligned buffers before trying any requests.
4034 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4036 Description: stale data returned from RAID cache
4037 Details : If only a small amount of IO is done to the RAID device before
4038 reading it again it is possible to get stale data from the RAID
4039 cache instead of reading it from disk.
4042 Frequency : always for sles10 kernel
4044 Description: sles10 support
4045 Details : ll_follow_link: compile fixes and using of nd_set_link
4046 under newer kernels.
4049 Frequency : depends on arch, kernel and compiler version, always on sles10
4052 Description: recursive or deep enough symlinks cause stack overflow
4053 Details : getting rid of large stack-allocated variable in
4057 Frequency : depends on hardware
4059 Description: lustre write performance loss in the SLES10 kernel
4060 Details : the performance loss is caused by using of write barriers in the
4061 ext3 code. The SLES10 kernel turns barrier support on by
4062 default. The fix is to undo that change for ldiskfs.
4064 ------------------------------------------------------------------------------
4066 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4068 * Support for kernels:
4069 2.6.9-42.0.3EL (RHEL 4)
4070 2.6.5-7.276 (SLES 9)
4071 2.4.21-47.0.1.EL (RHEL 3)
4072 2.6.12.6 vanilla (kernel.org)
4076 Frequency : quota enabled and large files being deleted
4078 Description: releasing more than 4GB of quota at once hangs OST
4079 Details : If a user deletes more than 4GB of files on a single OST it
4080 will cause the OST to spin in an infinite loop. Release
4081 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4086 Description: statfs data retrieved from /proc may be stale or zero
4087 Details : When reading per-device statfs data from /proc, in the
4088 {kbytes,files}_{total,free,avail} files, it may appear
4089 as zero or be out of date.
4092 Frequency : systems with MD RAID1 external journal devices
4094 Description: lconf's call to blkid is confused by RAID1 journal devices
4095 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4096 of returning all block devices that match the journal UUID.
4099 Frequency : always, for aggregate stripe size over 4GB
4101 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4102 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4103 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4106 Frequency : always if "lfs find" used on a local file/directory
4108 Description: "lfs find" segfaults if used on a local file/directory
4109 Details : The case where a directory component was not specified wasn't
4110 handled correctly. Handle this properly.
4113 Frequency : always on ppc64
4115 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4116 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4117 perform endianness conversion.
4120 Frequency : rarely (truncate to non-zero file size after write under load)
4121 Bugzilla : 10730, 10687
4122 Description: Files padded with zeros to next 4K multiple
4123 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4124 it is possible that files that are truncated to a non-zero size
4125 immediately after being written are filled with zero bytes beyond
4126 the truncated size. No file data is lost.
4128 Severity : enhancement
4130 Description: Allow recovery/failover for liblustre clients.
4131 Details : liblustre clients were unaware of failover configurations until
4134 Severity : enhancement
4136 Description: user file locks should fail when not mounting with flock option
4137 Details : Set up an error-returning stub in ll_file_operations.lock field
4138 to prevent incorrect behaviour when client is mounted without
4139 flock option. Also, set up properly f_op->flock field for
4143 Frequency : always on ia64
4145 Description: "lfs df" loops on printing out MDS statfs information
4146 Details : The obd_ioctl_data was not initialized and in some systems
4147 this caused a failure during the ioctl that did not return
4148 an error. Initialize the struct and return an error on failure.
4151 Frequency : SLES 9 only
4153 Description: Error of copying files with lustre special EAs as root
4154 Details : Client side always return success for setxattr call for lustre
4155 special xattr (currently only "trusted.lov").
4158 Frequency : rarely on clusters with both ia64+i386 clients
4160 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4161 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4162 clients are doing concurrent writes to the same file it is
4163 possible that the ia64 clients may LASSERT because the OST
4164 extent locks are not PAGE_SIZE aligned. Ensure that grown
4165 locks are always aligned on the request boundary.
4168 Frequency : specific use, occasional
4170 Description: Overwriting in use executable truncates on-disk binary image
4171 Details : If one node attempts to overwrite an executable in use by
4172 another node, we now correctly return ETXTBSY instead of
4173 truncating the file.
4175 Severity : enhancement
4177 Description: Async OSC create to avoid the blocking unnecessarily.
4178 Details : If an OST has no remaining object, system will block on the creating
4179 when need to create a new object on this OST. Now, ways use
4180 pre-created objects when available, instead of blocking on an
4181 empty osc while others are not empty. If we must block, we block
4182 for the shortest possible period of time.
4187 Description: chmod on Lustre root is propagated to other clients
4188 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4189 invalid by the follow_mount time.
4192 Frequency : liblustre clients only
4194 Description: Race in 'instant cancel' lock handling could lead to such locks
4195 never to be granted in case of SMP MDS
4196 Details : Do not destroy not yet granted but cbpending locks in
4200 Frequency : replay/resend of open
4202 Description: non null lock assetion failure in mds_intent_policy
4203 Details : Trying to replay/resend lockless open requests resulted in
4204 mds_open() returning 0 with no lock. Now it sets a flag if
4205 a lock is going to be returned.
4207 Severity : enhancement
4209 Description: Checksum enhancements
4210 Details : New checksum enhancements allow for resending RPCs that failed
4213 Severity : enhancement
4215 Description: Tunables on number of dirty pages in cacche
4216 Details : Allow to set limit on number of dirty pages cached.
4221 Description: client crash on unmount - lock still has references
4222 Details : In some error handling cases it was possible to leak a lock
4223 reference on a client while accessing a file. This was not
4224 harmful to the client during operation, but would cause the
4225 client to crash when the filesystem is unmounted.
4228 Frequency : specific case, rare
4230 Description: ETXTBSY on mds though file not in use by client
4231 Details : ETXTBSY is no longer incorrectly returned when attempting to
4232 chmod or chown a directory that the user previously tried to
4233 execute or a currently-executing binary.
4236 Frequency : extremely rare except on liblustre-based clients
4238 Description: Lustre space not freed when files are deleted
4239 Details : Clean up open-unlinked files after client eviction. Previously
4240 the unlink was skipped and the files remained as orphans.
4245 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4246 Details : In some cases it was possible to send a blocking callback to
4247 a client doing a glimpse, even though that client didn't get
4248 a lock granted. When the glimpse lock is cancelled on the OST
4249 the freed lock is left on the waiting list and corrupted the list.
4252 Frequency : all core dumps
4254 Description: Broke core dumps to lustre
4255 Details : Negative dentry may be unhashed if parent does not have UPDATE
4256 lock, but some callers, e.g. do_coredump, expect dentry to be
4257 hashed after successful create, hash it in ll_create_it.
4259 ------------------------------------------------------------------------------
4261 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4263 * Support for kernels:
4264 2.6.9-42.0.2.EL (RHEL 4)
4265 2.6.5-7.276 (SLES 9)
4266 2.4.21-40.EL (RHEL 3)
4267 2.6.12.6 vanilla (kernel.org)
4271 Frequency : always on RHEL 3
4273 Description: Number of open files grows over time
4274 Details : The number of open files grows over time, whether or not
4275 Lustre is started. This was due to a filp leak introduced
4276 by one of our kernel patches.
4278 ------------------------------------------------------------------------------
4280 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4282 * Support for kernels:
4283 2.6.9-42.EL (RHEL 4)
4284 2.6.5-7.276 (SLES 9)
4285 2.4.21-40.EL (RHEL 3)
4286 2.6.12.6 vanilla (kernel.org)
4291 Bugzilla : 5719, 9635, 9792, 9684
4292 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4293 Details : If a server is under heavy load and cannot reply to new
4294 connection requests before the client resends the (re)connect,
4295 the connection handling code can behave badly if two service
4296 threads are concurrently handing separate (re)connections from
4297 the same client. Add better locking to the connection handling
4298 code, and ensure that only a single connection will be processed
4299 for a given client UUID, even if the lock is dropped.
4301 Severity : enhancement
4303 Description: add TCP zero-copy support to kernel
4304 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4305 sends if the hardware supports scatter-gather and checksumming.
4306 This allows socklnd to do client-write and server-read more
4307 efficiently and reduce CPU utilization from skbuf copying.
4310 Frequency : only if NFS exporting from client
4312 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4313 Details : If setting attributes on a file created under NFS that had
4314 never been opened it would be possible to oops the client
4315 if the file had no objects.
4318 Frequency : always for liblustre
4320 Description: liblustre client does MDS+OSTs setattr RPC for each write
4321 Details : When doing a write from a liblustre client, the client
4322 incorrectly issued an RPC to the MDS and each OST the file was
4323 striped over in order to update the timestamps. When writing
4324 with small chunks and many clients this could overwhelm the MDS
4325 with RPCs. In all cases it would slow down the write because
4326 these RPCs are unnecessary.
4328 Severity : enhancement
4330 Description: allow number of MDS service threads to be changed at module load
4331 Details : It is now possible to change the number of MDS service threads
4332 running. Adding "options mds mds_num_threads={N}" to the MDS's
4333 /etc/modprobe.conf will set the number of threads for the next
4334 time Lustre is restarted (assuming the "mds" module is also
4335 reloaded at that time). The default number of threads will
4336 stay the same, 32 for most systems.
4341 Description: OST crash if filesystem is unformatted or corrupt
4342 Details : If an OST is started on a device that has never been formatted
4343 or if the filesystem is corrupt and cannot even mount then the
4344 error handling cleanup routines would dereference a NULL pointer.
4349 Description: NULL pointer deref in llap_from_page.
4350 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4351 incorrect) mapping if the page was truncated/reclaimed while it was
4352 searched for. Check for this condition and skip such pages when
4353 doing readahead. Introduce extra check to llap_from_page() to
4354 verify page->mapping->host is non-NULL (so page is not anonymous).
4357 Frequency : Sometimes when using sys_sendfile
4359 Description: "page not covered by a lock" warnings from ll_readpage
4360 Details : sendfile called ll_readpage without right page locks present.
4361 Now we introduced ll_file_sendfile that does necessary locking
4362 around call to generic_file_sendfile() much like we do in
4366 Frequency : with certain MDS communication failures at client mount time
4368 Description: NULL pointer deref after failed client mount
4369 Details : a client connection request may delayed by the network layer
4370 and not be sent until after the PTLRPC layer has timed out the
4371 request. If the client fails the mount immediately it will try
4372 to clean up before the network times out the request. Add a
4373 reference from the request import to the obd device and delay
4374 the cleanup until the network drops the request.
4377 Frequency : occasionally during client (re)connect
4379 Description: assertion failure during client (re)connect
4380 Details : processing a client connection request may be delayed by the
4381 client or server longer than the client connect timeout. This
4382 causes the client to resend the connection request. If the
4383 original connection request is replied in this interval, the
4384 client may trip an assertion failure in ptlrpc_connect_interpret()
4385 which thought it would be the only running connect process.
4388 Frequency : only with obd_echo servers and clients that are rebooted
4390 Description: kernel BUG accessing uninitialized data structure
4391 Details : When running an obd_echo server it did not start the ping_evictor
4392 thread, and when a client was evicted an uninitialized data
4393 structure was accessed. Start the ping_evictor in the RPC
4394 service startup instead of the OBD startup.
4396 Severity : enhancement
4397 Bugzilla : 10193 (patchless)
4398 Description: Remove dependency on various unexported kernel interfaces.
4399 Details : No longer need reparent_to_init, exit_mm, exit_files,
4400 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4403 Frequency : rare (only users of deprecated and unsupported LDAP config)
4405 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4406 Details : LDAP apparently contains 'lustreName' attributes instead of
4407 'name'. A simple remapping of the name is sufficient.
4410 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4412 Description: waiting_locks_callback trips kernel BUG if client is evicted
4413 Details : Running with the dump_on_timeout debug flag turned on makes
4414 it possible that the waiting_locks_callback() can try to dump
4415 the Lustre kernel debug logs from an interrupt handler. Defer
4416 this log dumping to the expired_lock_main() thread.
4418 Severity : enhancement
4420 Description: Support NFS exporting on 2.6 kernels.
4421 Details : Implement non-rawops metadata methods for NFS server to use without
4422 changing NFS server code.
4425 Frequency : very rare (synthetic metadata workload only)
4427 Description: two racing renames might cause an MDS thread to deadlock
4428 Details : Running the "racer" program may cause one MDS thread to rename
4429 a file from being the source of a rename to being the target of
4430 a rename at exactly the same time that another thread is doing
4431 so, and the second thread has already enqueued these locks after
4432 doing a lookup of the target and is trying to relock them in
4433 order. Ensure that we don't try to re-lock the same resource.
4436 Frequency : only very large systems with liblustre clients
4438 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4439 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4440 all OSTs in parallel. This allows the eviction of stale liblustre
4441 clients to proceed much faster than if they were done in series,
4442 and also offers similar improvements for other set_info RPCs.
4447 Description: excessive CPU usage during initial read phase on client
4448 Details : During the initial read phase on a client, it would agressively
4449 retry readahead on the file, consuming too much CPU and impacting
4450 performance (since 1.4.5.8). Improve the readahead algorithm
4451 to avoid this, and also improve some other common cases (read
4452 of small files in particular, where "small" is files smaller than
4453 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4458 Description: MDS crash when receiving packet with unknown intent.
4459 Details : Do not LBUG in unknown intent case, just return -EFAULT
4461 Severity : enhancement
4462 Bugzilla : 9293, 9385
4463 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4464 Details : Do not serialize getattr (non-intent version) and statfs.
4467 Frequency : occasional, when OST network is overloaded/intermittent
4469 Description: client evicted by OST after bulk IO timeout
4470 Details : If a client sends a bulk IO request (read or write) the OST
4471 may evict the client if it is unresposive to its data GET/PUT
4472 request. This is incorrect if the network is overloaded (takes
4473 too long to transfer the RPC data) or dropped the OST GET/PUT
4474 request. There is no need to evict the client at all, since
4475 the pinger and/or lock callbacks will handle this, and the
4476 client can restart the bulk request.
4479 Frequency : Always when mmapping file with no objects
4481 Description: client crashes when mmapping file with no objects
4482 Details : Check that we actually have objects in a file before doing any
4483 operations on objects in ll_vm_open, ll_vm_close and
4489 Description: Request leak when working with deleted CWD
4490 Details : Introduce advanced request refcount tracking for requests
4491 referenced from lustre intent.
4493 Severity : Enhancement
4495 Description: Cache open file handles on client.
4496 Details : MDS now will return special lock along with openhandle, if
4497 requested and client is allowed to hold openhandle, even if unused,
4498 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4499 closing files for every read/write openration.
4501 Severity : Enhancement
4503 Description: Cache open negative dentries on client when possible.
4504 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4505 negative dentries on lock revocation.
4510 Description: Remounting a client read-only wasn't possible with a zconf mount
4511 Details : It wasn't possible to remount a client read-only with llmount.
4513 Severity : enhancement
4514 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4515 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4516 MPICH is included to add Lustre-specific ADIO interfaces.
4517 This is based closely on the UFS ADIO layer and only differs
4518 in file creation, in order to allow the OST striping to be set.
4519 This is user-contributed code and not supported by CFS.
4524 Description: extended inode attributes (immutable, append-only) work improperly
4525 when 2.4 and 2.6 kernels are used on client/server or vice versa
4526 Details : Introduce kernel-independent values for these flags.
4528 Severity : enhancement
4531 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4532 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4533 granularity. Now, Fractional MB granularity is be supported,
4534 this is very useful for low memory system.
4536 Severity : enhancement
4538 Description: Getattr by fid
4539 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4540 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4541 have locks and avoids getattr rpc requests for it on every stat.
4544 Frequency : Always, for filesystems larger than 2TB
4546 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4547 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4548 filesystems larger than 2TB to be mounted with mballoc enabled.
4551 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4553 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4554 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4555 filesystems larger than 2TB is mounted then the kernel will
4556 silently corrupt the start of the filesystem. CONFIG_LBD is
4557 enabled for all CFS-supported kernels, but the possibility of
4558 this happening with a modified kernel config exists.
4560 Severity : enhancement
4562 Description: add client O_DIRECT support for 2.6 kernels
4563 Details : It is now possible to do O_DIRECT reads and writes to files
4564 in the Lustre client mountpoint on 2.6 kernel clients.
4566 Severity : enhancement
4568 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4569 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4570 parallel, not waiting for response from every OST before sending
4571 a rpc to the next OST.
4576 Description: setattr vs write race when updating file timestamps
4577 Details : Client processes that update a file timestamp into the past
4578 right after writing to the file (e.g. tar) it is possible that
4579 the updated file modification time can be reset to the current
4580 time due to a race between processing the setattr and write RPC.
4582 Severity : enhancement
4584 Description: Bring 'lfs find' closer in line with regular Linux find.
4585 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4586 -print0 options and obtains all the needed info through the lustre
4589 Severity : enhancement
4591 Description: support up to 1024 configured devices on one node
4592 Details : change obd_dev array from statically allocated to dynamically
4593 allocated structs as they are first used to reduce memory usage
4598 Description: Flush dirty partially truncated pages during truncate
4599 Details : Immediatelly flush partially truncated pages in filter_setattr,
4600 this way we completely avoid having any pages in page cache on OST
4601 and can retire ugly workarounds during writes to flush such pages.
4606 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4607 Details : protect inode from truncation within vfs_unlink() context
4608 just take a reference before calling vfs_unlink() and release it
4609 when parent's i_sem is free.
4614 Description: last_id value checked outside lock on OST caused LASSERT failure
4615 Details : If there were multiple MDS->OST object precreate requests in
4616 flight, it was possible that the OST's last object id was checked
4617 outside a lock and incorrectly tripped an assertion. Move checks
4618 inside locks, and discard old precreate requests.
4621 Frequency : always, if extents are used on OSTs
4623 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4624 Details : OSTs using the extents format would not zero the high 16 bits of
4625 the index physical block number. This is not a problem for any
4626 OST filesystems smaller than 16TB, and no kernels support ext3
4627 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4628 new/modified files) and can be fixed for existing filesystems
4629 with e2fsprogs-1.39-cfs1.
4634 Description: import connection selection may be incorrect if timer wraps
4635 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4636 import connections to be ignored if the 32-bit jiffies counter
4637 wraps. Use a 64-bit jiffies counter.
4640 Frequency : very large clusters immediately after boot
4642 Description: LNET request buffers exhausted under heavy short-term load
4643 Details : If a large number of client requests are generated on a service
4644 that has previously never seen so many requests it is possible
4645 that the request buffer growth cannot keep up with the spike in
4646 demand. Instead of dropping incoming requests, they are held in
4647 the LND until the RPC service can accept more requests.
4650 Frequency : Sometimes during replay
4652 Description: Assertion failure in ll_local_open after replay.
4653 Details : If replay happened on an open request reply before we were able
4654 to set replay handler, reply will become not swabbed tripping the
4655 assertion in ll_local_open. Now we set the handler right after
4656 recognising of open request
4659 Frequency : very rare
4661 Description: kernel reports "badness in vsnprintf"
4662 Details : Reading from the "recovery_status" /proc file in small chunks
4663 may cause a negative length in lprocfs_obd_rd_recovery_status()
4664 call to vsnprintf() (which is otherwise harmless). Exit early
4665 if there is no more space in the output buffer.
4667 Severity : enhancement
4669 Description: clear OBD RPC statistics by writing to them
4670 Details : It is now possible to clear the OBD RPC statistics by writing
4671 to the "stats" file.
4676 Description: Client mtime is not the same on different clients after utimes
4677 Details : In some cases, the client was using the utimes() syscall on
4678 a file cached on another node. The clients now validate the
4679 ctime from the MDS + OSTs to determine which one is right.
4684 Description: Inability to activate failout mode
4685 Details : lconf script incorrectly assumed that in pythong string's numeric
4686 value is used in comparisons.
4689 Frequency : always with multiple stripes per file
4691 Description: Inefficient object allocation for mutli-stripe files
4692 Details : When selecting which OSTs to stripe files over, for files with
4693 a stripe count that divides evenly into the number of OSTs,
4694 the MDS is always picking the same starting OST for each file.
4695 Return the OST selection heuristic to the original design.
4700 Description: mount failures may take full timeout to return an error
4701 Details : Under some heavy load conditions it is possible that a
4702 failed mount can wait for the full obd_timeout interval,
4703 possibly several minutes, before reporting an error.
4704 Instead return an error as soon as the status is known.
4706 Frequency : quota enabled and large files being deleted
4708 Description: releasing more than 4GB of quota at once hangs OST
4709 Details : If a user deletes more than 4GB of files on a single OST it
4710 will cause the OST to spin in an infinite loop. Release
4711 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4716 Description: statfs data retrieved from /proc may be stale or zero
4717 Details : When reading per-device statfs data from /proc, in the
4718 {kbytes,files}_{total,free,avail} files, it may appear
4719 as zero or be out of date.
4722 Frequency : systems with MD RAID1 external journal devices
4724 Description: lconf's call to blkid is confused by RAID1 journal devices
4725 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4726 of returning all block devices that match the journal UUID.
4729 Frequency : always, for aggregate stripe size over 4GB
4731 Description: assertion fails when trying to use 4GB stripe size
4732 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4733 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4736 Frequency : always on ppc64
4738 Description: the first write on an ext3 filesystem with mballoc got stuck
4739 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4740 perform endianness conversion.
4742 ------------------------------------------------------------------------------
4744 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4746 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4747 INTEROPERATE with older versions automatically. Please read the
4748 user documentation before upgrading any part of a live system.
4749 * WARNING: Lustre networking configuration changes are required with
4750 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4753 * Support for kernels:
4754 2.6.9-22.0.2.EL (RHEL 4)
4755 2.6.5-7.244 (SLES 9)
4756 2.6.12.6 vanilla (kernel.org)
4759 Severity : enhancement
4760 Bugzilla : 7981/8208
4761 Description: Introduced Lustre Networking (LNET)
4762 Details : LNET is new networking infrastructure for Lustre, it includes
4763 a reorganized network configuration mode (see the user
4764 documentation for full details) as well as support for routing
4765 between different network fabrics. Lustre Networking Devices
4766 (LNDs) for the supported network fabrics have also been
4767 created for this new infrastructure.
4769 Severity : enhancement
4770 Description: Introduced Access control lists
4771 Details : clients can set ACLs on files and directories in order to have
4772 more fine-grained permissions than the standard Unix UGO+RWX.
4773 The MDS must be started with the "-o acl" mount option.
4775 Severity : enhancement
4776 Description: Introduced filesystem quotas
4777 Details : Administrators may now establish per-user quotas on the
4780 Severity : enhancement
4782 Description: Configuration change for the XT3
4783 The PTLLND is now used to run Lustre over Portals on the XT3
4784 The configure option(s) --with-cray-portals are no longer used.
4785 Rather --with-portals=<path-to-portals-includes> is used to
4786 enable building on the XT3. In addition to enable XT3 specific
4787 features the option --enable-cray-xt3 must be used.
4792 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4793 Details : A race between allocating a new llcd and re-getting the llcd_lock
4794 allowed another thread to grab newly-allocated llcd.
4796 Severity : enhancement
4798 Description: 2.6 OST async journal commit and locking fix to improve performance
4799 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4800 2.6 kernels are now async as they already were in 2.4 kernels so
4801 that they can commit concurrently with the network bulk transfer.
4802 For block-allocated files the filter allocation semaphore is held
4803 to avoid filesystem fragmentation during allocation. BKL lock
4804 removed for 2.6 xattr operations where it is no longer needed.
4809 Description: lconf incorrectly determined whether two IP networks could talk
4810 Details : In some more complicated routing and multiple-network
4811 configurations, lconf will avoid trying to make a network
4812 connection to a disjoint part of the IP space. It was doing the
4813 math incorrectly for one set of cases.
4818 Description: Fix for potential infinite loop processing records in an llog.
4819 Details : If an llog record is corrupted/zeroed, it is possible to loop
4820 forever in llog_process(). Validate the llog record length
4821 and skip the remainder of the block on error.
4824 Frequency : occasional (liblustre only)
4826 Description: liblustre could not open files whose last component is a symlink
4827 Details : sysio_path_walk() would incorrectly pass the open intent to
4828 intermediate path components.
4831 Frequency : rare (liblustre only with non-standard tuning)
4832 Bugzilla : 7201 (7350)
4833 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4834 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4835 after it has been released, only do this for referenced locks
4837 Severity : enhancement
4839 Description: specifying an (invalid) directory default stripe_size of -1
4840 would reset the directory default striping
4841 Details : stripe_size -1 was used internally to signal directory stripe
4842 removal, now use "all default" to signal dir stripe removal
4843 as a directory striping of "all default" is not useful
4846 Frequency : common for large clusters running liblustre clients
4848 Description: doing an ls when liblustre clients are running is slow
4849 Details : sending a glimpse AST to a liblustre client waits for every AST
4850 to time out, as liblustre clients will not respond. Since they
4851 cannot cache data we refresh the OST lock LVB from disk instead.
4853 Severity : enhancement
4855 Description: doing an ls at the same time as file IO can be slow
4856 Details : enqueue and other "small" requests can be blocked behind many
4857 large IO requests. Create a new OST IO portal for non-IO
4858 requests so they can be processed faster.
4861 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4863 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4864 created file with mode 000
4865 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4866 allow an ftruncate() as a non-root user to a file with mode 000.
4867 Lustre now respects this flag to disable mode checks when
4868 truncating a file owned by the user
4871 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4874 Description: Revoking locks from clients that went dead or catatonic might take
4876 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4877 cancellation of such locks instant on servers without waiting for
4878 any reply from clients. Clients drops these locks when cancel
4879 notification from server is received without replying.
4882 Frequency : liblustre-only, when liblustre client dies or becomes busy
4884 Description: Doing ls on Linux clients can take a long time with active
4886 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4887 granting such locks to it in the first place if possible. Locks
4888 are taken by proxy on the OST during the read or write and
4889 dropped immediately afterward. Add connect flags handling, do
4890 not grant locks to liblustre clients for glimpse ASTs.
4892 Severity : enhancement
4894 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4895 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4896 sequential reads and behaves badly with applications doing random
4897 reads. Improve it by reading ahead at least read region, and
4898 avoiding excessive large RPC for small reads.
4900 Severity : enhancement
4902 Description: Creating more than 1000 files for a single job may cause a load
4903 imbalance on the OSTs if there are also a large number of OSTs.
4904 Details : qos_prep_create() uses an OST index reseed value that is an
4905 even multiple of the number of available OSTs so that if the
4906 reseed happens in the middle of the object allocation it will
4907 still utilize the OSTs as uniformly as possible.
4912 Description: OST or MDS may oops in ping_evictor_main()
4913 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4914 but doesn't restart at beginning of obd_exports_timed list
4917 Severity : enhancement
4919 Description: improve by-nid export eviction on the MDS and OST
4920 Details : allow multiple exports with the same NID to be evicted at one
4921 time without re-searching the exports list.
4924 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4926 Description: MDS may oops in groups_free()
4927 Details : in rare race conditions a newly allocated group_info struct is
4928 freed again, and this can be NULL. The 2.4 compatibility code
4929 for groups_free() checked for a NULL pointer, but 2.6 did not.
4932 Frequency : common for liblustre clients doing little filesystem IO
4933 Bugzilla : 9352, 7313
4934 Description: server may evict liblustre clients accessing contended locks
4935 Details : if a client is granted a lock or receives a completion AST
4936 with a blocking AST already set it would not reply to the AST
4937 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4940 Frequency : lfs setstripe, only systems with more than 160 OSTs
4942 Description: unable to set striping with a starting offset beyond OST 160
4943 Details : llapi_create_file() incorrectly limited the starting stripe
4944 index to the maximum single-file stripe count.
4947 Frequency : LDAP users only
4949 Description: lconf did not handle in-kernel recovery with LDAP properly
4950 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4952 Severity : enhancement
4954 Description: bind OST threads to NUMA nodes to improve performance
4955 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4956 node and do their allocations there to localize memory access
4958 Severity : enhancement
4960 Description: llmount can determine client NID directly from Myrinet (GM)
4961 Details : the client NID code from gmnalnid was moved directly into
4962 llmount, removing the need to use this or specifying the
4963 client NID explicitly when mounting GM clients with zeroconf
4966 Frequency : if client is started with down MDS
4968 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4969 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4970 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4971 ptlrpc_import_delayed_req() because we didn't check if the
4972 request was interrupted, and we also didn't break out of the
4973 event loop if there was no timeout
4978 Description: data loss during non-page-aligned writes to a single file from
4979 both multiple nodes and multiple threads on one node at same time
4980 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4981 inconsistency led to false short-reads, that were cached and later
4982 used by ->prepare_write() to fill in partially written page,
4983 leading to data loss.
4986 Frequency : always, if lconf --abort_recovery used
4988 Description: lconf --abort_recovery fails with 'Operation not supported'
4989 Details : lconf was attempting to abort recovery on the MDT device and not
4992 Severity : enhancement
4994 Description: remove cleanup logs
4995 Details : replace lconf-generated cleanup logs with lustre internal
4996 cleanup routines. Eliminates the need for client-cleanup and
4999 Severity : enhancement
5001 Description: add support for EAs (user and system) on lustre filesystems
5002 Details : it is now possible to store extended attributes in the Lustre
5003 client filesystem, and with the user_xattr mount option it
5004 is possible to allow users to store EAs on their files also
5006 Severity : enhancement
5008 Description: Add possibility (config option) to show minimal available OST free
5010 Details : When compiled with --enable-mindf configure option, statfs(2)
5011 (and so, df) will return least minimal free space available from
5012 all OSTs as amount of free space on FS, instead of summary of
5013 free spaces of all OSTs.
5015 Severity : enhancement
5017 Description: do not expand extent locks acquired on OST-side
5018 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5019 by server: they are not cached anyway.
5022 Frequency : when mmap is used/binaries executed from Lustre
5024 Description: Unmmap pages before throwing them away from read cache.
5025 Details : llap_shrink cache now attempts to unmap pages before discarding
5026 them (if unmapping failed - do not discard). SLES9 kernel has
5027 extra checks that trigger if this unmapping is not done first.
5032 Description: lconf didn't resolve symlinks before checking to see whether a
5033 given mountpoint was already in use
5036 Frequency : when migrating failover services
5037 Bugzilla : 6395, 9514
5038 Description: When migrating a subset of services from a node (e.g. failback
5039 from a failover service node) the remaining services would
5040 time out and evict clients.
5041 Details : lconf --force (implied by --failover) sets the global obd_timeout
5042 to 5 seconds in order to quickly disconnect, but this caused
5043 other RPCs to time out too quickly. Do not change the global
5044 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5046 Severity : enhancement
5047 Frequency : if MDS is started with down OST
5048 Bugzilla : 9439,5706
5049 Description: Allow startup/shutdown of an MDS without depending on the
5050 availability of the OSTs.
5051 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5052 Add appropriate locking and lov-osc refcounts for safe
5053 cleaning. Add osc abort_inflight calls in case the
5054 synchronize never started.
5057 Frequency : occasional (Cray XT3 only)
5059 Description: root not authorized to access files in CRAY_PORTALS environment
5060 Details : The client process capabilities were not honoured on the MDS in
5061 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5062 been accessed by an authorized user then root was able to access
5063 the file on the local client also. The root user capabilities
5064 are now allowed on the MDS, as this environment has secure UID.
5067 Frequency : occasional
5069 Description: ldiskfs "too long searching" message happens too often
5070 Details : A debugging message (otherwise harmless) prints too often on
5071 the OST console. This has been reduced to only happen when
5072 there are fragmentation problems on the filesystem.
5077 Description: Division by zero in statfs when all OSCs are inactive
5078 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5079 lov_statfs divides by value returned by lov_get_stripecnt().
5083 Bugzilla : 9489, 3273
5084 Description: First write from each client to each OST was only 4kB in size,
5085 to initialize client writeback cache, which caused sub-optimal
5086 RPCs and poor layout on disk for the first written file.
5087 Details : Clients now request an initial cache grant at (re)connect time
5088 and so that they can start streaming writes to the cache right
5089 away and always do full-sized RPCs if there is enough data.
5090 If the OST is rebooted the client also re-establishes its grant
5091 so that client cached writes will be honoured under the grant.
5096 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5097 Details : Now I/O RPCs go to different portal number and (presumably) fast
5098 lock requests (and glimses) and other RPCs get their own service
5099 threads pool that should be able to service those RPCs
5102 Severity : enhancement
5104 Description: Ability to exchange lustre version between client and servers and
5105 issue warnings at client side if client is too old. Also for
5106 liblustre clients there is ability to refuse connection of too old
5108 Details : New 'version' field is added to connect data structure that is
5109 filled with version info. That info is later checked by server and
5113 Frequency : rare, liblustre only.
5114 Bugzilla : 9296, 9581
5115 Description: Two simultaneous writes from liblustre at offset within same page
5116 might proceed at the same time overwriting eachother with stale
5118 Details : I/O lock withing llu_file_prwv was released too early, before data
5119 actually was hitting the wire. Extended lock-holding time until
5120 server acknowledges receiving data.
5123 Frequency : extremely rare. Never observed in practice.
5125 Description: avoid generating lustre_handle cookie of 0.
5126 Details : class_handle_hash() generates handle cookies by incrementing
5127 global counter, and can hit 0 occasionaly (this is unlikely, but
5128 not impossible, because initial value of cookie counter is
5129 selected randonly). Value of 0 is used as a sentinel meaning
5130 "unassigned handle" --- avoid it. Also coalesce two critical
5131 sections in this function into one.
5133 Severity : enhancement
5135 Description: allow liblustre clients to delegate truncate locking to OST
5136 Details : To avoid overhead of locking, liblustre client instructs OST to
5137 take extent lock in ost_punch() on client's behalf. New connection
5138 flag is added to handle backward compatibility.
5140 Severity : enhancement
5141 Bugzilla : 4928, 7341, 9758
5142 Description: allow number of OST service threads to be specified
5143 Details : a module parameter allows the number of OST service threads
5144 to be specified via "options ost ost_num_threads={N}" in the
5145 OSS's /etc/modules.conf or /etc/modprobe.conf.
5149 Bugzilla : 6146, 9635, 9895
5150 Description: servers crash with bad pointer in target_handle_connect()
5151 Details : In rare cases when a client is reconnecting it was possible that
5152 the connection request was the last reference for that export.
5153 We would temporarily drop the export reference and get a new
5154 one, but this may have been the last reference and the export
5155 was just destroyed. Get new reference before dropping old one.
5157 Severity : enhancement
5158 Frequency : if client is started with failover MDS
5160 Description: Allow multiple MDS hostnames in the mount command
5161 Details : Try to read the configuration from all specified MDS
5162 hostnames during a client mount in case the "primary"
5165 Severity : enhancement
5167 Description: Stop sending data to evicted clients as soon as possible.
5168 Details : Check if the client we are about to send or are sending data to
5169 was evicted already. (Check is done every second of waiting,
5170 for which l_wait_event interface was extended to allow checking
5171 of exit condition at specified intervals).
5174 Frequency : rare, normally only when NFS exporting is done from client
5176 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5178 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5181 Frequency : always, if srand() is called before liblustre initialization
5183 Description: Liblustre uses system PRNG disturbing its usage by user application
5184 Details : Introduce internal to lustre fast and high-quality PRNG for
5185 lustre usage and make liblustre and some other places in generic
5186 lustre code to use it.
5188 Severity : enhancement
5189 Bugzilla : 9477, 9557, 9870
5190 Description: Verify that the MDS configuration logs are updated when xml is
5191 Details : Check if the .xml configuration logs are newer than the config
5192 logs stored on the MDS and report an error if this is the case.
5193 Request --write-conf, or allow starting with --old_conf.
5195 Severity : enhancement
5197 Description: Handle symlinks in the path when checking if Lustre is mounted.
5198 Details : Resolve intermediate symlinks when checking if a client has
5199 mounted a filesystem to avoid duplicate client mounts.
5204 Description: lconf can hit an error exception but still return success.
5205 Details : The lconf command catches the Command error exception at the top
5206 level script context and will exit with the associated exit
5207 status, but doesn't ensure that this exit status is non-zero.
5212 Description: failure of ptlrpc thread startup can cause oops
5213 Details : Starting a ptlrpc service thread can fail if there are a large
5214 number of threads or the server memory is very fragmented.
5215 Handle this without oopsing.
5218 Frequency : always, only if liblustre and non-default acceptor port was used
5220 Description: liblustre cannot connect to servers with non-default acceptor port
5221 Details : tcpnal_set_default_params() was not called and was therefore
5222 ignoring the environment varaible TCPNAL_PORT, as well as other
5223 TCPNAL_ environment variables
5228 Description: two objects could be created on the same OST for a single file
5229 Details : If an OST is down, in some cases it was possible to create two
5230 objects on a single OST for a single file. No problems other
5231 than potential performance impact and spurious error messages.
5235 Bugzilla : 5681, 9562
5236 Description: Client may oops in ll_unhash_aliases
5237 Details : Client dcache may become inconsistent in race condition.
5238 In some cases "getcwd" can fail if the current directory is
5244 Description: Inode refcounting problems in NFS export code
5245 Details : link_raw functions used to call d_instantiate without obtaining
5246 extra inode reference first.
5250 Bugzilla : 9942, 9903
5251 Description: Referencing freed requests leading to crash, memleaks with NFS.
5252 Details : We used to require that call to ll_revalidate_it was always
5253 followed by ll_lookup_it. Also with revalidate_special() it is
5254 possible to call ll_revalidate_it() twice for the same dentry
5255 even if first occurence returned success. This fix changes semantic
5256 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5257 reference on a request referred from the intent.
5258 ll_intent_release() then releases such a request.
5261 Frequency : rare, normally benchmark loads only
5263 Description: unlinked inodes were kept in memory on the client
5264 Details : If a client is repeatedly creating and unlinking files it
5265 can accumulate a lot of stale inodes in the inode slab cache.
5266 If there is no other client load running this can cause the
5267 client node to run out of memory. Instead flush old inodes
5268 from client cache that have the same inode number as a new inode.
5271 Frequency : SLES9 2.6.5 kernel and long filenames only
5272 Bugzilla : 9969, 10379
5273 Description: utime reports stale NFS file handle
5274 Details : SLES9 uses out-of-dentry names in some cases, which confused
5275 the lustre dentry revalidation. Change it to always use the
5279 Frequency : rare, unless heavy write-truncate concurrency is continuous
5280 Bugzilla : 4180, 6984, 7171, 9963, 9331
5281 Description: OST becomes very slow and/or deadlocked during object unlink
5282 Details : filter_destroy() was holding onto the parent directory lock
5283 while truncating+unlinking objects. For very large objects this
5284 may block other threads for a long time and slow overall OST
5285 responsiveness. It may also be possible to get a lock ordering
5286 deadlock in this case, or run out of journal credits because of
5287 the combined truncate+unlink. Solution is to do object truncate
5288 first in one transaction without parent lock, and then do the
5289 final unlink in a new transaction with the parent lock. This
5290 reduces the lock hold time dramatically.
5293 Frequency : rare, 2.4 kernels only
5295 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5296 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5297 before unmounting their filesystem in order to avoid deadlock.
5298 The kernel_locked() function in 2.4 kernels only checks whether
5299 the kernel lock is held, not whether it is this process that is
5300 holding it as 2.6 kernels do.
5305 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5306 Details : The client ptlrpc code may be trying to reconnect to a down
5307 server before a previous connection attempt has timed out.
5308 Increase the reconnect interval to be longer than the connection
5309 timeout interval to avoid sending duplicate connections to
5313 Frequency : echo_client brw_test command
5315 Description: fix echo_client to work with OST preallocated code
5316 Details : OST preallocation code (5137) didn't take echo_client IO path
5317 into account: echo_client calls filter methods outside of any
5318 OST thread and, hence, there is no per-thread preallocated
5319 pages and buffers to use. Solution: hijack pga pages for IO. As
5320 a byproduct, this avoids unnecessary data copying.
5324 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5325 Description: Client can oops in mdc_commit_close() after open replay
5326 Details : It was possible for the MDS to return an open request with no
5327 transaction number in mds_finish_transno() if the client was
5328 evicted, but without actually returning an error. Clients
5329 would later try to replay that open and may trip an assertion
5330 Simplify the client close codepath, and always return an error
5331 from the MDS in case the open is not successful.
5334 Frequency : rare, 2.6 OSTs only
5336 Description: OST may deadlock under high load on fragmented files
5337 Details : If there was a heavy load and highly-fragmented OST filesystems
5338 it was possible to have all the OST threads deadlock waiting on
5339 allocation of biovecs, because the biovecs were not released
5340 until the entire RPC IO was completed. Instead, release biovecs
5341 as soon as they are complete to ensure forward IO progress.
5343 Severity : enhancement
5345 Description: Support for specifying external journal device at mount
5346 Details : If an OST or MDS device is formatted with an external journal
5347 device, this device major/minor is stored in the ext3 superblock
5348 and may not be valid for failover. Allow detecting and
5349 specifying the external journal at mount time.
5354 Description: Mounting an MDS with pending unlinked files may cause oops
5355 Details : target_finish_recovery() calls mds_postrecov() which returned
5356 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5357 considers this an error and immediately begins cleaning up the
5358 lov, just after starting the mds_lov process
5360 Severity : enhancement
5362 Description: Implement 'lfs df' to report actual free space on per-OST basis
5363 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5364 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5365 usage of objects; '-h' to report in human readable format.
5367 ------------------------------------------------------------------------------
5369 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5376 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5377 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5378 at startup during orphan unlinking. Instead, skip buddy update
5379 before setup, as it will be regenerated after recovery is complete.
5384 Description: If an OST is inactive, its locks might reference stale inodes.
5385 Details : lov_change_cbdata() must iterate over all namespaces, even if
5386 they are inactive to clear inode references from the lock.
5388 Severity : enhancement
5389 Frequency : occasional, if non-standard max_dirty_mb used
5391 Description: Client will block write RPCs if not enough grant
5392 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5393 then the client will block writes while waiting for another RPC
5394 to complete instead of consuming its dirty limit. With change
5395 we get improved performance when max_dirty_mb is small.
5397 Severity : enhancement
5398 Bugzilla : 3389, 6253
5399 Description: Add support for supplementary groups on the MDS.
5400 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5401 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5402 lookups for user supplementary groups into a cache.
5406 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5407 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5408 for liblustre O_EXCL is enforced by MDS.
5412 Description: readdir never returns NULL in liblustre.
5413 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5414 dentry in struct dirent.
5418 Frequency : liblustre only, depends on application IO pattern
5419 Description: liblustre clients evicted if not contacting servers
5420 Details : Don't put liblustre clients into the ping_evictor list, so
5421 they will not be evicted by the pinger ever.
5423 Severity : enhancement
5425 Description: Add ability to evict clients by NID from MDS.
5426 Details : By echoing "nid:$NID" string into
5427 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5428 $NID would be instantly evicted from this MDS and from all active
5429 OSTs connected to it.
5433 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5434 Details : lookup_it_finish() used to query file size from OSTs that was not
5439 Description: service threads change working directory to that of init
5440 Details : Starting lustre service threads may pin the working directory
5441 of the parent thread, making that filesystem busy. Threads
5442 now change to the working directory of init to avoid this.
5446 Frequency : during shutdown only
5447 Description: shutdown with a failed MDS or OST can cause unmount to hang
5448 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5453 Frequency : 2.6 only
5454 Description: chmod/chown may include an extra supplementary group
5455 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5456 supplementary group and if none is specified this is used.
5459 Bugzilla : 5479 (6816)
5461 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5462 Details : If lookup is in progress on a file that is unlinked we might try
5463 to revalidate the inode and fail in revalidate after lookup is
5464 complete and ll_file_open() enqueues the open again but
5465 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5468 Frequency : always, if lconf --abort_recovery used
5470 Description: lconf --abort_recovery fails with 'Operation not supported'
5471 Details : lconf was attempting to abort recovery on the MDT device and not
5474 ------------------------------------------------------------------------------
5476 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5481 Frequency : rare (only unsupported configurations with a node running as an
5483 Bugzilla : 6514, 5137
5484 Description: Mounting a Lustre file system on a node running as an OST could
5486 Details : OSTs now preallocates memory needed to write out data at
5487 startup, instead of when needed, to avoid having to
5488 allocate memory in possibly low memory situations.
5489 Specifically, if the file system is mounted on on OST,
5490 memory pressure could force it to try to write out data,
5491 which it needed to allocate memory to do. Due to the low
5492 memory, it would be unable to do so and the node would
5493 become unresponsive.
5495 Severity : enhancement
5497 Description: Addition of lconf --service command line option
5498 Details : lconf now accepts a '--service <arg>' option, which is
5499 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5501 Severity : enhancement
5503 Description: Failover mode is now the default for OSTs.
5504 Details : By default, OSTs will now run in failover mode. To return to
5505 the old behaviour, add '--failout' to the lmc line for OSTs.
5507 Severity : enhancement
5509 Description: Health checks are now provided for MDS and OSTs
5510 Details : Additional detailed health check information on MSD and OSTs
5511 is now provided through the procfs health_check value.
5514 Frequency : occasional, depends on IO load
5516 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5517 after numerous create/delete cycles
5518 Details : The ext3 inode allocation policy would not allocate new inodes
5519 very well on the OSTs because there are no new directories
5520 being created. Instead we look for groups with free space if
5521 the parent directories are nearly full.
5526 Description: Network or server problems during mount may cause partially
5527 mounted clients instead of returning an error.
5528 Details : The config llog parsing code may overwrite the error return
5529 code during mount error handling, returning success instead
5535 Description: MDS can fail to allocate large reply buffers
5536 Details : After long uptimes the MDS can fail to allocate large reply
5537 buffers (e.g. zconf client mount config records) due to memory
5538 fragmentation or consumption by the buffer cache. Preallocate
5539 some large reply buffers so that these replies can be sent even
5540 under memory pressure.
5544 Frequency : rare (liblustre)
5545 Description: fsx running with liblustre complained that using truncate() to
5546 extend the file doesn't work. This patch corrects that issue.
5547 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5548 ATTR_SIZE and lsm use in llu_setattr_raw.
5552 Frequency : rare, only 2.6 kernels
5553 Description: Unusual file access patterns on the MDS may result in inode
5554 data being lost in very rare circumstances.
5555 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5556 code in the 2.6 kernel caused the inode and/or EA data not to
5557 be read from disk, causing single-file corruption.
5561 Frequency : rare, only 2.6 filesystems using extents
5562 Description: Heavy concurrent write and delete load may cause data corruption.
5563 Details : It was possible under high-load situations to have an extent
5564 metadata block in the block device cache from a just-unlinked
5565 file overwrite a newly-allocated data block. We now unmap any
5566 metadata buffers that alias just-allocated data blocks.
5570 Frequency : filesystems with default stripe_count larger than 77
5571 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5572 Details : lconf specifies an inode size of 4096 bytes when the default
5573 stripe_count is larger than 77. This conflicts with the default
5574 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5575 this case to avoid pinning too much memory for large EAs.
5577 ------------------------------------------------------------------------------
5579 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5584 Frequency : rare (extremely heavy IO load with hundreds of clients)
5586 Description: Client is evicted, gets IO error writing to file
5587 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5588 caused clients to be evicted for AST timeouts. The fixes in
5589 bug 5192 mean we no longer need to have such short AST timeouts
5590 so ldlm_timeout has been increased.
5593 Frequency : occasional during --force or --failover shutdown under load
5594 Bugzilla : 5949, 4834
5595 Description: Server oops/LBUG if stopped with --force or --failover under load
5596 Details : a collection of import/export refcount and cleanup ordering
5597 issues fixed for safer force cleanup
5600 Frequency : only filesystems larger than 120 OSTs
5601 Bugzilla : 5990, 6223
5602 Description: lfs getstripe would oops on a very large filesystem
5603 Details : lov_getconfig used kfree on vmalloc'd memory
5606 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5607 Bugzilla : 6242, 6243
5608 Description: reading from files that had been truncated to a non-zero size
5609 but never opened returned no data
5610 Details : ll_file_read() reads zeros from no-object files to EOF
5615 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5616 Details : There's little harm in aborting MDS/OSS recovery and letting it
5617 try again, so I removed the LASSERT and return an error instead.
5619 Severity : enhancement
5621 Description: New debugging infrastructure for tracking down data corruption
5622 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5623 (b) cover more of the client-side code path, and (c) try to narrow
5624 down where problems occurred
5628 Bugzilla : 3819, 4364, 4397, 6313
5629 Description: Racing close and eviction MDS could cause assertion in mds_close
5630 Details : It was possible to get multiple mfd references during close and
5631 client eviction, leading to one thread referencing a freed mfd.
5633 Severity: : enhancement
5634 Bugzilla : 3262, 6359
5635 Description: Attempts to reconnect to servers are now more aggressive.
5636 Details : This builds on the enhanced upcall-less recovery that was added
5637 in 1.4.2. When trying to reconnect to servers, clients will
5638 now try each server in the failover group every 10 seconds. By
5639 default, clients would previously try one server every 25 seconds.
5644 Description: After recovery, certain operations trigger a failed
5645 assertion on a client.
5646 Details : Failing over an mds, using lconf -d --failover, while a
5647 client was doing a readdir() call would cause the client to
5648 LBUG after recovery completed and the readdir() was resent.
5650 Severity : enhancement
5652 Description: Default groups are now added by lconf
5653 Details : You can now run lconf --group <servicename> without having to
5654 manually add groups with lmc.
5657 Frequency : occasional
5659 Description: Nodes with an elan id of 0 trigger a failed assertion
5662 Frequency : always when accessing e.g. tty/console device nodes
5664 Description: tty and some other devices nodes cannot be used on lustre
5665 Details : file's private_data field is used by device data and lustre
5666 values in there got lost. New field was added to struct file to
5667 store fs-specific private data.
5670 Frequency : when exporting Lustre via NFS
5672 Description: NFSD failed occasionally when looking up a path component
5673 Details : NFSD is looking up ".." which was broken in ext3 directories
5674 that had grown large enough to become hashed.
5677 Frequency : Clusters with multiple interfaces not on the same subnet
5679 Description: Nodes will repeatedly try to reconnect to an interface which it
5680 cannot reach and report an error to the log.
5681 Details : Extra peer list entries will be created by lconf with some peers
5682 unreachable. lconf now validates the peer before adding it.
5685 Frequency : Only if a default stripe is set on the filesystem root.
5687 Description: Setting a default stripe on the filesystem root prevented the
5688 filesystem from being remounted.
5689 Details : The client was sending extra request flags in the root getattr
5690 request and did not allocate a reply buffer for the dir EA.
5693 Frequency : occasional, higher if lots of files are accessed by one client
5694 Bugzilla : 6159, 6097
5695 Description: Client trips assertion regarding lsm mismatch/magic
5696 Details : While revalidating inodes the VFS looks up inodes with ifind()
5697 and in rare cases can find an inode that is being freed.
5698 The ll_test_inode() code will free the lsm during ifind()
5699 when it finds an existing inode and then the VFS later attaches
5700 this free lsm to a new inode.
5704 Bugzilla : 6422, 7030
5705 Description: MDS deadlock between mkdir and client eviction
5706 Details : Creating a new file via mkdir or mknod (starting a transaction
5707 and getting the ns lock) can deadlock with client eviction
5708 (gets ns lock and trying to finish a synchronous transaction).
5711 Frequency : occasional
5712 Description: While starting a server, the fsfilt_ext3 module could not be
5714 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5715 kernels. Previously, lconf would still use the ext3 name
5716 when trying to load modules. Now, it will correctly use
5717 ext3 on 2.4 and ldiskfs on 2.6.
5719 Severity : enhancement
5720 Description: The default stripe count has been changed to 1
5721 Details : The interpretation of the default stripe count (0, to lfs
5722 or lmc) has been changed to mean striping across a single
5723 OST, rather than all available. For general usage we have
5724 found a stripe count of 1 or 2 works best.
5726 Severity : enhancement
5727 Description: Add support for compiling against Cray portals.
5728 Details : Conditional compiling for some areas that are different
5732 Frequency : occasional
5733 Bugzilla : 6409, 6834
5734 Description: Creating files with an explicit stripe count may lead to
5735 a failed assertion on the MDS
5736 Details : If some OSTs are full or unavailable, creating files may
5737 trigger a failed assertion on the MDS. Now, Lustre will
5738 try to use other servers or return an error to the
5742 Frequency : occasional
5744 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5745 caused lock timeout during readahead (since 1.4.2).
5746 Details : Processes doing readahead might match a lock that hasn't been
5747 granted yet if there are overlapping and conflicting lock
5748 requests. The readahead process waits on ungranted lock
5749 (original lock is CBPENDING), while OST waits for that process
5750 to cancel CBPENDING read lock and eventually evicts client.
5752 Severity : enhancement
5754 Description: Initial enabling of flock support for clients
5755 Details : Implements fcntl advisory locking and file status functions.
5756 This feature is provided as an optional mount flag (default
5757 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5758 locking are implemented yet, and those that are are not guaranteed
5759 to be completely correct in production environments.
5760 mount -t lustre -o [flock|noflock] ...
5763 Frequency : occasional
5765 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5766 assertion in the ext3 JBD (journaling) layer.
5767 Details : The b_committed_data struct is protected by the big kernel lock
5768 in 2.4 kernels, serializing journal_commit_transaction() and
5769 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5770 access to this struct. In 2.6 kernels there is finer grained
5771 locking to improve SMP performance of the JBD layer.
5775 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5780 Description: merge and cleanup kernel patches.
5783 -----------------------------------------------------------------------------
5785 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5787 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5788 versions of Lustre. Please update all servers and clients to
5789 version 1.4.2 or later at the same time. You must also run
5790 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5791 to update the configuration logs.
5793 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5794 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5795 - swab llog records of type '0' so we get proper header size/idx (5861)
5796 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5797 - fix rename of one directory over another leaking an inode (5953)
5798 - avoid SetPageDirty on 2.6 (5981)
5799 - don't re-add just-being-destroyed locks to the waiting list (5653)
5800 - when creating new directories, inherit the parent's custom
5801 striping settings if present parent (3048)
5802 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5803 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5804 - don't hold client locks on temporary worklist from l_lru (5666)
5805 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5806 - automatically evict dead clients (5921)
5807 - Update file size properly in create+truncate+fstat case (6196)
5808 - Do not unhash mountpoint dentries, do not allow removal of
5810 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5811 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5812 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5813 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5814 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5815 - Correctly handle reads of files with no objects (6243)
5816 - lctl recover will also mark a device active if deactivate used (5933)
5818 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5819 - allow --write-conf on an MDS with different nettype than client (5619)
5820 - don't write config llogs to MDS for mounts not from that MDS (5617)
5821 - lconf should create multiple TCP connections from a client (5201)
5822 - init scripts are now turned off by default; run chkconfig --on
5823 lustre and chkconfig --on lustrefs to use them
5824 - upcalls are no longer needed for clients to recover to failover
5826 - add --abort-recovery option to lconf to abort recovery on device
5828 - add support for an arbitrary number of OSTs (3026)
5829 - Quota support protocol changes.
5830 - forward compatibility changes to wire structs (6007)
5831 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5832 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5833 - improved "lustre status" script
5834 - initialize blocksize for non-regular files (6062)
5835 - added --disable-server and --disable-client configure options (5782)
5836 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5837 - Vanilla 2.4.29 support
5838 - increase maximum number of obd devices to 520 (6242)
5839 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5840 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5841 SLES 9 2.6.5 kernels
5842 - limit stripes per file to 160 (the maximum EA size) (6093)
5844 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5847 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5848 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5849 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5850 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5851 - small ext3 extents cleanups and fixes (5733)
5852 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5853 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5854 - avoid needless client->OST connect, fix handle mismatch (5317)
5855 - fix DLM error path that led to out-of-sync client, long delays (5779)
5856 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5857 - fix several locking issues related to i_size (5492,5624,5654,5672)
5858 - don't move pending lock onto export if it is already evicted (5683)
5859 - fix kernel oops when creating .foo in unlinked directory (5548)
5860 - fix deadlock in obdfilter statistics vs. object create (5811)
5861 - use time_{before,after} to avoid timer jiffies wrap (5882)
5862 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5863 - Do not leak request if server was not able to process it (5154)
5864 - If mds_open unable to find parent dir, make that negative lookup(5154)
5865 - don't create new directories with extent-mapping (5909, 5936)
5867 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5868 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5869 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5870 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5871 - support for RHEL4 kernel on i686 (5773)
5872 - provide error messages when incompatible logs are encountered (5898)
5874 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5875 * version 1.4.0.10 (1.4.1 release candidate 1)
5877 - don't keep a lock reference when lock is not granted (4238)
5878 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5879 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5880 - drop import inflight refcount on signal_completed_replay error (5255)
5881 - unlock page after async write error during send (3677)
5882 - handle missing objects in filter_preprw_read properly (5265)
5883 - no transno return for symlink open, don't save no-trasno open (3440)
5884 - don't try to complete elan receive that already failed (4012)
5885 - free RPC server reply state on error (5406)
5886 - clean up thread from ptlrpc_start_thread() on error (5160)
5887 - readahead could read extra page into cache that wasn't ejected (5388)
5888 - prevent races in class_attach/setup/cleanup/detach (5260)
5889 - don't dereference de->d_inode after l_dput of de (5458)
5890 - use "int" for stripe value returned from lock_to_stripe (5544)
5891 - mballoc allocation and error-checking fixes in 2.6 (5504)
5892 - block device patches to fix I/O request sizes in 2.6 (5482)
5893 - look up hostnames for IB nals (5602)
5894 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5895 - don't start multiple acceptors for the same port (5277)
5896 - fix incorrect LASSERT in mds_getattr_name (5635)
5897 - export a proc file for general "ping" checking (5628)
5898 - fix "lfs check" to not block when the MDS is down (5628)
5900 - service request history (4965)
5901 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5902 - create an "evict_client" /proc entry on OSTs, like the MDS has
5903 - fix mount usage message, return errors per mount(8) (5168)
5904 - change grep [] to grep "[]" in tests so they work in more UMLs
5905 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5906 - remove ancient LOV_MAGIC_V0 EA support (5047)
5907 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5908 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5909 - allow readahead allocations to fail when low on memory (5383)
5910 - mmap locking landed again, after considerable improvement (2828)
5911 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5913 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5916 - send OST transaction number in read/write reply to free req (4966)
5917 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5918 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5919 - the watchdog thread now runs as interruptible (5246)
5920 - flock/lockf fixes (but it's still disabled, pending 5135)
5921 - don't use EXT3 constants in llite code (5094)
5922 - memory shortage at startup could cause assertion (5176)
5924 - reorganization of lov code
5925 - single portals codebase
5927 - add extents/mballoc support (5025)
5928 - direct I/O reads in the obdfilter (4048)
5929 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5931 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5934 - send OST transaction number in read/write reply to free req (4966)
5935 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5936 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5937 - the watchdog thread now runs as interruptible (5246)
5938 - handle missing objects in filter_preprw_read properly (5265)
5939 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5940 - drop import inflight refcount on signal_completed_replay error (5255)
5941 - unlock page after async write error during send (3677)
5942 - return original error code on reconstructed replies (3761)
5943 - no transno return for symlink open, don't save no-trasno open (3440)
5945 - add pid to ldlm debugging output (4922)
5946 - bump the watchdog timeouts -- we can't handle 30sec yet
5947 - extra debugging for orphan dentry/inode bug (5259)
5949 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5952 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5953 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5954 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5955 - fix NULL dereference in /proc/sys/portals/routes (4827)
5956 - allow failed mdc_close() operations to be interrupted (4561)
5957 - stop precreate on OST before MDS would time out on it (4778)
5958 - don't send partial-page writes before EOF from client (4410)
5959 - discard client grant for sub-page writes on large-page clients (4520)
5960 - don't free dentries not owned by NFS code, check generation (4806)
5961 - fix lsm leak if mds_create_objects() fails (4801)
5962 - limit debug_daemon file size, always print CERROR messages (4789)
5963 - use transno after validating reply (3892)
5964 - process timed out requests if import state changes (3754)
5965 - update mtime on OST during writes, return in glimpse (4829)
5966 - add mkfsoptions to LDAP (4679)
5967 - use ->max_readahead method instead of zapping global ra (5039)
5968 - don't interrupt __l_wait_event() during strace
5970 - add software watchdogs to catch hung threads quickly (4941)
5971 - make lustrefs init script start after nfs is mounted
5972 - fix CWARN/ERROR duplication (4930)
5973 - return async write errors to application if possible (2248)
5974 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5975 - print NAL number in %x format (4645)
5976 - update barely-supported suse-2.4.21-171 series (4842)
5977 - support for sles 9 %post scripts
5978 - support for building 2.6 kernel-source packages
5979 - support for sles km_* packages
5981 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5984 - ignore -ENOENT errors in osc_destroy (3639)
5985 - notify osc create thread that OSC is being cleaned up (4600)
5986 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5987 - reconstruct ost_handle() like mds_handle() (4657)
5988 - create a new thread to do import eviction to avoid deadlock (3969)
5989 - let lconf resolve symlinked-to devices (4629)
5990 - don't unlink "objects" from directory with default EA (4554)
5991 - hold socknal file ref over connect in case target is down (4394)
5992 - allow more than 32000 subdirectories in a single directory (3244)
5993 - fix blocks count for O_DIRECT writes (3751)
5994 - OST returns ENOSPC from object create when no space left (4539)
5995 - don't send truncate RPC if file size isn't changing (4410)
5996 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5997 - bind to privileged port in socknal and tcpnal (3689)
5999 - rate limit CERROR/CWARN console message to avoid overload (4519)
6000 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6001 - basic mmap support (3918)
6002 - kernel patch series update from b1_4 (4711)
6004 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6007 - avoid crash during MDS cleanup with OST shut down (2775)
6008 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6009 - don't use bad inodes on the MDS (3744)
6010 - dynamic object preallocation to improve recovery speed (4236)
6011 - don't hold spinlock over lock dumping or change debug flags (4401)
6012 - don't zero obd_dev when it is force cleaned (3651)
6013 - print grants to console if they go negative (4431)
6014 - "lctl deactivate" will stop automatic recovery attempts (3406)
6015 - look for existing locks in ldlm_handle_enqueue() (3764)
6016 - don't resolve lock handle twice in recovery avoiding race (4401)
6017 - revalidate should check working dir is a directory (4134)
6019 - don't always mark "slow" obdfilter messages as errors (4418)
6021 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6024 - don't close LustreDB during write_conf until it is done (3860)
6025 - fix typo in lconf for_each_profile (3821)
6026 - allow dumping logs from multiple threads at one time (3820)
6027 - don't allow multiple threads in OSC recovery (3812)
6028 - fix debug_size parameters (3864)
6029 - fix mds_postrecov to initialize import for llog ctxt (3121)
6030 - replace config semaphore with spinlock (3306)
6031 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6032 - don't allow enqueue to complete on a destroyed export (3822)
6033 - down write_lock before checking llog header bitmap (3825)
6034 - recover from lock replay timeout (3764)
6035 - up llog sem before sending rpc (3652)
6036 - reduce ns lock hold times when setting kms (3267)
6037 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6038 - fix NULL deref and obd_dev leak on setup error (3312)
6039 - replace some LBUG about llog ops with error handling (3841)
6040 - don't match INVALID dentries from d_lookup and spin (3784)
6041 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6042 - fix invalid assertion in ptlrpc_set_wait (3880)
6044 - add libwrap support for the TCP acceptor (3996)
6045 - add /proc/sys/portals/routes for non-root route listing (3994)
6046 - allow setting MDS UUID in .xml (2580)
6047 - print the stack of a process that LBUGs (4228)
6049 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6052 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6053 - make sure to unset replay flag from failed open requests (3440)
6054 - if default stripe count is 0, use OST count for inode size (3636)
6055 - update parent mtime/ctime on client for create/unlink (2611)
6056 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6057 - free recovery state on server during a forced cleanup (3571)
6058 - unregister_reply for resent reqs (3063)
6059 - loop back devices mounting and status check on 2.6 (3563)
6060 - fix resource-creation race that can provoke i_size == 0 (3513)
6061 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6062 - more debugging for page-accounting assertion (3746)
6063 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6064 - avoid deadlock after precreation failure (3758)
6065 - fix race and lock order deadlock in orphan handling (3450, 3750)
6066 - add validity checks when grabbing inodes from l_ast_data (3599)
6068 - add /proc/.../recovery_status to obdfilter (3428)
6069 - lightweight CDEBUG infrastructure, debug daemon (3668)
6070 - change default OSC RPC parameters to be better on small clusters
6071 - turn off OST read cache for files smaller than 32MB
6072 - install man pages and include them in rpms (3100)
6073 - add new init script for (un)mounting lustre filesystems (2593)
6074 - run chkconfig in %post for init scripts (3701)
6075 - drop scimac NAL (unmaintained)
6077 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6080 - clean kiobufs before and after use (3485)
6081 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6082 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6083 - change init script to fail more clearly if not run as root (1528)
6084 - allow clients to reconnect during replay (1742)
6085 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6086 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6087 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6088 - fix cond_resched (3554)
6089 - only evict unfinished clients after recovery (3515)
6090 - allow bulk resend, prevent data loss (3570)
6091 - dynamic ptlrpc request buffer allocation (2102)
6092 - don't allow unlinking open directory if it isn't empty (2904)
6093 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6094 - remove extraneous obd dereference causing LASSERT failure (3334)
6095 - don't use get_cycles() when creating temp. files on the mds (3156)
6096 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6097 - handle EEXIST for set-stripe, set proper directory name (3336)
6099 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6100 - fix ksocknal_fmb_callback() error messages (2918)
6102 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6105 - don't copy lvb into (possibly NULL) reply on error (2983)
6106 - don't deref dentry after dput, don't free lvb on error (2922)
6107 - use the kms to determine writeback rpc length (2947)
6108 - increment oti_logcookies when osc is inactive (2948)
6109 - update client's i_blocks count via lvb messages (2543)
6110 - handle intent open/close of special files properly (1557)
6111 - mount MDS with errors=remount-ro, like obdfilter (2009)
6112 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6113 - don't use cancelling-locks' kms values (2947)
6114 - use highest lock extent for kms, not last one (2925)
6115 - don't dereference ERR_PTR() dentry in error handling path (3107)
6116 - fix thread race in portals_debug_dumplog() (3122)
6117 - create lprocfs device entries at setup instead of at attach (1519)
6118 - common AST error handler, don't evict client on completion race (3145)
6119 - zero nameidata in detach_mnt in 2.6 (3118)
6120 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6121 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6122 - initialize RPC timeout timer earlier for 2.6 (3219)
6123 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6124 - print nal/nid for unknown nid (3258)
6125 - additional checks for oscc recovery before doing precreate (3284)
6126 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6127 - don't crash in mdc_close for bad permissions on open (3285)
6128 - zero i_rdev for non-device files (3147)
6129 - clear page->private before handing to FS, better assertion (3119)
6130 - tune the read pipeline (3236)
6131 - fix incorrect decref of invalidated dentry (2350)
6132 - provide read-ahead stats and refine rpc in flight stats (3328)
6133 - don't hold journal transaction open across create RPC (3313)
6134 - update atime on MDS at close time (3265)
6135 - close LDAP connection when recovering to avoid server load (3315)
6136 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6137 - don't leak open file on MDS after open resend (3325)
6138 - serialize filter_precreate and filter_destroy_precreated (3329)
6139 - loop device shouldn't call sync_dev() for nul device (3092)
6140 - clear page cache after eviction (2766)
6141 - resynchronize MDS->OST in background (2824)
6142 - refuse to mount the same filesystem twice on same mountpoint (3394)
6143 - allow llmount to create routes for mounting behind routers (3320)
6144 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6145 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6146 - fix rare problem with rename on htree directories (3417)
6148 - allow default OST striping configuration per directory (1414)
6149 - fix compilation for qswnal for 2.6 kernels (3125)
6150 - increase maximum number of MDS request buffers for large systems
6151 - change liblustreapi to be useful for external progs like lfsck (3098)
6152 - increase local configuration timeout for slow disks (3353)
6153 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6155 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6158 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6159 - don't overwrite extent policy data in reply if lock was blocked (2901)
6160 - drop filter export grants atomically with removal from device (2663)
6161 - del obd_self_export from work_list in class_disconnect_exports (2908)
6162 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6163 - swab reply message in mdc_close, other PPC fixes (2464)
6164 - fix destroying of named logs (2325)
6165 - overwrite old logs when running lconf --write_conf (2264)
6166 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6167 - fix race in target_handle_connect (2898)
6168 - mds_reint_create() should take same inode create lock (2926)
6169 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6170 - don't close files for self_export to avoid uninitialized obd (2936)
6171 - allow MDS with the same name as client node (2939)
6172 - hold dentry reference for closed log files for unlink (2325)
6173 - reserve space for all logs during transactions (2059)
6174 - don't evict page beyond end of stripe extent (2925)
6175 - don't oops on a deleted current working directory (2399)
6176 - handle hard links to targets without a parent properly (2517)
6177 - don't dereference NULL lock when racing during eviction (2867)
6178 - don't grow lock extents when lots of conflicting locks (2919)
6180 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6183 - account for cache space usage on clients to avoid data loss (974)
6184 - lfsck support in lustre kernel code (2349)
6185 - reduce journal credits needed for BRW writes (2370)
6186 - orphan handling to avoid losing space on client/server crashes
6187 - ptlrpcd can be blocked, stopping ALL progress (2477)
6188 - use lock value blocks to assist in proper KMS, faster stat (1021)
6189 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6190 - recovery for initial connections (2355)
6191 - fixes for mds_cleanup_orphans (1934)
6192 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6193 - block all file creations until orphan recovery completes (1901)
6194 - client remove rq_connection from request struct (2423)
6195 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6196 - recovery timer race (2670)
6197 - mdc_close recovey bug (2532)
6198 - ptlrpc cleanup bug (2710)
6199 - mds timeout on local locks (2588)
6200 - namespace lock held during RPCs (2431)
6201 - handle interrupted sync write properly (2503)
6202 - don't try to handle a message that hasn't been replied to (2699)
6203 - client assert failure during cleanup after abort recovery (2701)
6204 - leak mdc device after failed mount (2712)
6205 - ptlrpc_check_set allows timedout requests to complete (2714)
6206 - wait for inflight reqs when ptlrpcd finishes (2710)
6207 - make sure unregistered services are removed from the srv_list
6208 - reset bulk XID's when resending them (caught by 1138 test)
6209 - unregister_bulk after timeout
6210 - fix lconf error (2694)
6211 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6212 - readahead locks pages, leaves pending causing memory pressure (2673)
6213 - increase OST request buffers to 4096 on large machines (2729)
6214 - fix up permission of existing directories in simple_mkdir (2661)
6215 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6216 - don't assign transno to errored transactions (2742)
6217 - don't delete objects on OST if given a bogus objid from MDS (2751)
6218 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6219 - if rq_no_resend, then timeout request after recovery (2432)
6220 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6221 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6222 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6223 - MDS recovery shouldn't delete objects that it has given out (2730)
6224 - if enqueue arrives after completion, don't clobber LVB (2819)
6225 - don't unlock pages twice when trigger_group_io returns error (2814)
6226 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6227 - don't write pages to disk if there was an error (1450)
6228 - don't ping imports that have recovery disabled (2676)
6229 - take buffered bytes into account when balancing socknal conn (2817)
6230 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6231 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6232 - remove little-endian swabbing of llog records (1987)
6233 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6234 - retry reposting request buffers if they fail (1191)
6235 - grow extent at grant time to avoid granting a revoked lock (2809)
6236 - lock revoke doesn't evict page if covered by a second lock (2765)
6237 - disable VM readahead to avoid reading outside lock extents (2805)
6239 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6240 - updated kernel patches for hp-2.4.20 kernel (2681)
6242 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6245 - fix truncated write corruption (2366)
6246 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6248 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6249 - ensure len(uuid) < 37 in lmc (1171)
6250 - fix ia64 OOPS in llog_test (2255)
6251 - zero end of page at obdfilter for partial page writes (2648)
6252 - don't leave stale dentries around after renames (bug 2428)
6253 - fix timeouts when evicting a client with a single lock held (2642)
6254 - set deadline for the initial HELLO message to drain (2634)
6255 - print out dotted-quad IP addresses in the socknal (2302)
6257 - additional debugging for MDS client eviction problem (2443)
6258 - fix mkfsoptions support for osts (2603, 2604)
6260 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6263 - add series for the vanilla 2.6.0 kernel
6264 - add series for the vanilla 2.4.24 kernel
6265 - add series for a cray x86/64 UL kernel drop
6266 - fix xattr patches for the vanilla 2.4.19 series
6268 - generate true UUIDs in lmc (1171)
6269 - have portals stack dumping break in UML (2466)
6270 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6271 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6272 - wake callback waiting threads on client eviction (2460)
6273 - Add --ptldebug and --subsystem to lmc (1719)
6274 - update assertion to allow safe interrupt allocation
6275 - set rq_no_resend for cancel requests (2432)
6276 - recalculate ptlrpcd timeout after resend (2494)
6277 - call vfs_rmdir when removing pending directories (2368)
6278 - fix renaming a file to itself (2429)
6279 - lmc creates a default one-stripe lov (2454)
6280 - expand procfs space to handle large clusters (2326)
6281 - increase UML stack to avoid overflow
6282 - update lconf's list of debug and subsystem masks
6283 - fix lfs find --obd (2510)
6284 - /proc tunable for disabling filter read caching (2591)
6285 - stop rpm packages from altering slapd.conf (2301)
6286 - disable nagle in the socknal under 0conf (2578)
6287 - choose mds inode size based on stripe count (2572)
6288 - fix kernel-source rpm problems (2516)
6290 - add --disable-doc to avoid pdf generation (2421)
6291 - update documentation, tests, type-os, comments
6292 - avoid format warnings on ia64
6293 - remove the TOE NAL
6294 - tiny code cleanups by removing unused fields
6296 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6299 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6300 - remove the most blatant lies from BUILDING file (2371)
6301 - change default debug level to reasonable production setting
6302 - reduce client side cache size to reduce cache flush time
6303 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6304 - make TCP zerocopy and pinger support enabled by default (2476)
6305 - sync writes completed after process exits caused crashes (2319)
6306 - maintain correct mount count on the MDS (2356)
6307 - backout 1557, because 2316 wasn't really fixed
6308 - better file I/O statistics gathering in /proc
6309 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6310 - another kernel patch to fix zero-copy TCP function export
6311 - don't take duplicate lock when processing re-sent getattr (2420)
6312 - lctl uses obd_self_export instead of creating new conn (2353)
6313 - MDS/OST recovery case which requires object creation asserted (2425)
6314 - move lfs from /usr/sbin to /usr/bin in packages
6315 - fix race between mds_client_add and mds_client_free (2417)
6316 - use kmalloc instead of slabs in portals (2430)
6317 - don't create duplicate records when a failover MDS is present (2442)
6318 - remove unnecessary mount age check (2332)
6319 - don't remove directory inodes from locks prematurely (2451)
6320 - don't break if MDS service name is the same as hostname (2103)
6321 - fix races in client write RPC generation when cache full (2482)
6323 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6326 - remove now-unused request->rq_obd (278)
6327 - if an allocation fails, print out how much memory we've used (1933)
6328 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6329 - add the "configurable stack size" patch to most series files (1256)
6330 - ability to write large log records, for 100+ OST configs (2306)
6331 - fix NULL deref when filter_prep fails (2314)
6332 - fix operator precedence error in filter_sync
6333 - dynamic allocation of socknal TX descriptors (2315)
6334 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6335 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6336 - ensure that CWARN messages in Portals always get to the syslog
6337 - __init/__exit are not for prototype decls (ldlm_init/exit)
6338 - x86-64 compile warning fixes
6339 - fix gateway LMC keyword conflict (2318)
6340 - fix MDS lock inversions in getattr/reint paths (1844)
6341 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6342 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6343 - DLM race condition prevented some lock evictions (2328)
6344 - ENOMEM detection and retry on socknal sends (2230)
6345 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6346 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6347 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6348 - better error messages when a client is rejected during recovery (1505)
6349 - avoid cancelling locks which were never granted, after failure (2330)
6350 - fix i_sem/journal inversion in mds_client_add (2333)
6351 - fix truncate/getattr lock cycle deadlock (2334)
6352 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6353 - fix two rare exit paths which could leak an l_lock() ref (2321)
6354 - fencepost error in MDS/OST orphan recovery (2226)
6355 - make log record alignment 8 bytes (1988)
6356 - lstripe now fails when requested offset > ost_count (2237)
6357 - ensure that all kernel series have a complete list.h (1607)
6358 - fix crashes in special-file operations (2316)
6359 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6361 - allow configurable automake binary, for testing new versions
6362 - small update to the lfs documentation
6364 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6366 * fix negative export reference count in fsfilt_sync (2312)
6368 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6369 * release candidate 0.9.1
6371 - orphans are moved into the PENDING directory for possible recovery
6372 - replayed opens now open by fid for orphan/rename safety (1042)
6373 - last close of an orphan inode generates a transno (683)
6374 - chdir() and mount() now pin the directory entry (1020)
6375 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6376 - discard very old requests without processing them (1502)
6377 - remove some common, well-understood CERRORs (1505)
6378 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6379 - clear "grant" flags in OST replies until OST grant code lands (1644)
6380 - fix read performance by not clobbering i_blksize on client (1598)
6381 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6382 - make LRU size a /proc tunable, clears locks when reduced (707)
6383 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6384 - server locks take references on exports now (1558)
6385 - build fixes for 2.4.20-rh trees (1663)
6386 - return an error from lov_create if all OSCs are inactive (1751)
6387 - fix import levels when a reconnect happens without a timeout (1597)
6388 - exit early from mds_open if we get a lookup error (1749)
6389 - partial page read at EOF wouldn't wait for disk before sending (1642)
6390 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6391 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6392 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6393 - write/truncate lock inversion (1639)
6394 - Don't auto-load obdclass, portals modules during cleanup (1495)
6395 - fix timestamps from jumping to "now" (1763)
6396 - extra journal assertions (1648)
6397 - add an extra multiunlink test (1771)
6398 - fix read_record/write_record API (1776)
6399 - fix leak of offset_extent, possible incorrect i_size later (1772)
6400 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6401 - Debugging for the kqswnal_get_idle_tx problems (1820)
6402 - Allow recovery to be attempted multiple times (1536)
6403 - Write out MDS last_rcvd file after it is first created (1600)
6404 - Fix tx_descriptor leak in failed transmit situations (1827)
6405 - ext3 journaling fixes for assertion failure after IO error (1871)
6406 - class_export_put() on freed export after completion AST error (1896)
6407 - Fix revalidate looping in VFS (1322)
6408 - Don't access a freed export during MDS_REINT timeout (1521)
6409 - Add open-unlink recovery support on the MDS (1673,1764)
6410 - Return an error if no MDS data was read from last_rcvd (1946)
6411 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6412 - Don't setup a disk device that doesn't match exported UUID (317)
6413 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6414 - avoid committing NULL handle in force close
6415 - local.sh is now a one-stripe LOV configuration
6416 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6417 - fix ext3 htree duplicate directory entry corruption (1516)
6418 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6419 - update to new LOV EA format (2097)
6420 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6422 2003-06-15 Phil Schwan <phil@clusterfs.com>
6425 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6426 - per-import/export recovery handling (958, 931, 959)
6427 - multiple last-rcvd slots, for serving multiple FSes (949)
6428 - connections are again shared between multiple imp/exports (963, 964)
6429 - "umount -f" would hang if any requests needed to be sent (393, 978)
6430 - avoid pinning large req buffer by copying for queued messages (989)
6431 - add "uuid" to "lctl device" command to help upcalls (991)
6432 - "open" RPCs with transnos would confuse recovery counters (1037)
6433 - do proper endian conversion of all wire messages (288, 340, 891)
6434 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6435 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6436 - fix ptlrpc_connection leak in target_handle_connect (1174)
6437 - fix import refcounting bug in OST and MDS cleanup (1134)
6438 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6439 - fix very unlikely obd_types race condition (501)
6440 - remove osc_open hack for echo_client (1187)
6441 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6442 - a failure in read_inode2 leads to deadlock (1139)
6443 - cancel ack-locks as soon as transaction is committed (1072)
6444 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6445 - make sure to commitrw after any preprw to avoid deadlock (1162)
6446 - failing to execute a file in a lustre FS would lock inode (1203)
6447 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6448 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6449 - obd_uuid2tgt would walk off the end of the list (1255)
6450 - on IA64 the osc would give portals incorrect bulk size (1258)
6451 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6452 - fix lock inversion caused by new llite matching code (1282)
6453 - limit the number of dirty pages on a client to 10MB (1286)
6454 - timed out locks were not being corrected cancelled (1289)
6455 - fix O_DIRECT above 4GB on IA-32 (1292)
6456 * major user-visible changes
6457 - fail out/fail over policy now controlled by the upcall (993)
6459 - add OBD_PING to check server availability and failure (954)
6460 - lustre messages are now sent in sending host order (288, 340, 891)
6461 - add eadatalen to MDS getattr reply (340)
6462 - OST read replies may contain second buffer, with per-page status (593)
6464 2003-03-11 Phil Schwan <phil@clusterfs.com>
6467 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6468 - failed open()s could cause deadlock; fixed (867, 869)
6469 - stop cancelling OST locks when files are closed (481)
6470 - overlapping XID spaces caused network corruption (851, 853)
6471 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6472 - setattr_raw added, to do single-RPC, server-side setattrs
6473 - lmc/lconf syntax change for OST UUIDs
6474 - fix crashy race condition between ptlrpc_free_req and osc_close
6475 - don't use request in mdc_enqueue if we hit a timeout (889)
6476 - don't set the inode i_size for regular files from the MDS (896)
6477 - handle out of order completion AST (842)
6478 - don't LBUG if a lock request times out after receiving AST (913)
6479 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6480 - if a bad lock AST arrives, send an error instead of dropping entirely
6481 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6482 - fix leak in bulk IO when only partially completed (899, 900, 926)
6483 - fix O_DIRECT for ia64 (55)
6484 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6485 - C-z after timeout could hang a process forever; fixed (977)
6487 - client-side I/O cache (678, 924, 929, 941, 970)
6489 - READPAGE and SETATTRs which don't take server-side locks get
6492 2003-02-11 Phil Schwan <phil@clusterfs.com>
6495 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6496 - Fix fsfilt_extN_readpage() to read a full page of directory
6497 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6498 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6499 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6500 - Most intent operations were removed, in favour of a new RPC mode
6501 that does a single RPC to the server and bypasses most of the VFS
6502 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6503 - Aggressively cancel local locks on DLM servers
6504 - mds_reint_unlink sends EA to the client if it's the last nlink.
6505 client uses that EA to unlink OST objects.
6506 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6507 - recursive symlinks were fixed (439)
6508 - fixed NULL deref in DEBUG_REQ
6509 - filter_update_lastobjid no longer calls sync, which annoyed extN
6510 - fixed multi-client small-writes to a single file problem (445)
6511 - fixed mtime updates during file writes (607)
6512 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6513 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6514 - fixed Linux OST size reporting problem (444, 656)
6515 - OST now updates object mtime with writes or setattr (607, 619)
6516 - client verifies file size before zeroing page past EOF (445)
6517 - OST now writes last allocated objid to disk with allocation (108)
6518 - LOV on echo now works (409)
6520 - mds_reint_unlink sends a new buffer, with the EA included. this
6521 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6522 set if a regular file was being unlinked, and it was the last link
6523 - use PtlGet from the target for bulk writes (315)
6524 - OST now updates object mtime with writes or setattr (607, 619)
6525 - LDLM now has a grant-time callback to revalidate locked items, if
6527 - Many MDS operations were reorganized to combat race conditions
6529 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6530 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6532 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6535 - Fully reactivate OST imports after reconnection (512, others)
6536 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6537 - More graceful error handling for truncating on dead OST (515)
6538 - Don't error out unless we're actually accessing dead stripes (474)
6539 - Fix garbage sizes when stripes are missing (410)
6540 - LRU counters were broken, causing constant lock purge (433, 432)
6541 - garbage on read from stripes with failed OSTs (441)
6542 - mark OSCs as active before reconnecting during recovery (438)
6543 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6544 - lfind did not preserve OST order in output (443)
6545 - symlinks cause hung clients, incorrect data (439)
6546 - stop dereferencing request after dropping refcount (457)
6547 - don't LASSERT(spin_is_locked) on non-SMP (455)
6548 - fixes for many rename() bugs
6549 - fstat didn't correctly synchronize attributes (399)
6550 - server must handle lock cancellation during blocking AST prep (487)
6551 - bulk descriptors were free()d too soon (511)
6552 - fix paths in lconf, which would load incorrect modules (451, 507)
6553 - fix confusing lconf 'host not found' error message (386)
6554 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6555 - fix race condition in mdc_blocking_ast() for inode access (526)
6556 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6557 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6558 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6559 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6560 - fix lstripe in several ways (kernel side) (527)
6561 - fix request leak in ldlm_cli_enqueue (262)
6562 - incorrect OSC was marked inactive after OST failure
6563 - call mds_fs_cleanup before unmounting filesystem (524)
6564 - fix races between taking ns_lock and ldlm_lock_change_resource
6565 - fix races updating LOV export open file list
6566 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6567 - fix recovery NULL deref in ldlm_cli_cancel_unused
6568 - fix some DLM races by using new hash table for lock handles (419)
6569 - permit the client to specify desired inodes, at replay
6570 - duplicate requests when we queue them for replay reintegration
6571 - fix last_rcvd offset calculation
6572 - sync after each recovered transaction, so we always make progress
6573 - never, not always, ERESTART requests without transnos
6574 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6576 - skip replay if the MDS says that the client is already connected
6577 - don't check for a recovery-enabled export to match lctl's UUID
6578 - don't INC_USE_COUNT for phantom exports
6579 - don't crash when cleaning up phantom exports (567)
6580 - don't double-finish or set replay data for errored mdc_open requests
6581 - abort requests when they time out, so we don't get old replies
6582 - send/receive replies for AST messages again
6583 - if the client says that it doesn't have the lock, cancel it on the
6585 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6586 mark it as destroyed, it will all work out when decref is called
6587 - fix module use counts (22, 581)
6589 - ASTs now expect a reply (server cancels lock on error reply)
6591 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6594 - fix many simultaneous client startup (392)
6595 - fix dentry->d_it clobbering
6596 - credentials weren't being shipped for readdir/getattr operations
6597 - remove invalid assertions triggered during some concurrent MD
6599 - proper Lustre versions added (336, 389)
6600 - fix memory leak for create error case (398)
6601 - fix LOV locking bug that would get cli/srv out of sync
6602 - fix echo client over LOV (409)
6603 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6604 - fix double-O_EXCL intent crash (424)
6605 - avoid sending multiple lock CANCELs (352)
6607 - MDS can do multi-client recovery (modulo bugs in new code)
6609 - many updates, edits, cleanups
6611 2002-11-18 Phil Schwan <phil@clusterfs.com>
6614 - fix null d_it dereference (346)
6615 - fix full OST/dbench hang (333)
6616 - fix permission problem with file removal (286)
6617 - fix removal of OSCs from LOV when they fail
6618 - fix NULL deref during bulk timeout (214)
6619 - fix problems related to multiple filesystems on one MDS (241)
6620 - fixed serious subtle metadata locking bugs
6621 - free locks on clients when inodes are removed due to memory
6623 - fix inode pointer in lock data (285)
6624 - partial support for multiple MDS on a single host (241)
6625 - data locks weren't cancelled at clear_inode time (290, 311)
6626 - intent locks could lead to unbounded lock growth (205)
6627 - added a maximum lock count, an LRU list, and a flusher
6628 - fix multiple rename (365)
6629 - properly abstracted the echo client
6630 - OSC locked 1 byte too many; fixed
6631 - rewrote brw callback code:
6632 - fixed recovery bugs related to LOVs (306)
6633 - fixed too-many-pages-in-one-write crash (191)
6634 - fixed (again) crash in sync_io_timeout (214)
6635 - probably fixed callback-related race (385)
6637 - Add capability to MDS protocol
6638 - LDLM cancellations and callbacks on different portals
6640 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6643 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6644 - defer open object destruction to close time (601981, 138)
6645 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6646 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6647 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6648 - don't call obd_setattr in ll_file_release for destroyed objects
6649 * protocol change to lustre_msg: move |version| and add |flags|
6650 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6651 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6652 * added replay of create, unlink, link and rename operations during
6653 MDS failover; recovery should be much more robust now
6654 * remove failed OSCs from LOVs (only lov_create uses this so far)
6655 * the lustre-HOWTO was brought (more) up to date (582544)
6657 2002-10-23 Phil Schwan <phil@clusterfs.com>
6660 - in-use dentries weren't being reused properly (617851)
6661 - prevent multiple LDLM setup (599178)
6662 - fix LOV size calculations for truncate (617853)
6663 - fix client handling of MDS intent errors (POSIX)
6664 - fix permission bug in lovstripe.c test (624321)
6665 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6666 - truncate past end of file could corrupt data
6667 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6668 - a race in recovery could return ETIMEDOUT to apps (623947)
6669 - building outside the source directory was fixed
6670 * the lustre-HOWTO was brought (more) up to date (582544)
6671 * major progress was made on recovery functionality
6673 2002-10-10 Phil Schwan <phil@clusterfs.com>
6676 - recovery deadlock fix
6677 - rm -rf causes LBUG fix (617817)
6678 - file open by multiple tasks fix (618962)
6679 - directory permissions bugs (602707 and 620007)
6680 - journal_stop fixed with locking (611313)
6681 - O_APPEND failures resolved (618273, perhaps 614459)
6682 - lconf PATH fix (619770)
6683 - IA64 build fix (621450)
6684 - RPC buffer sizes scale with amount of memory
6686 2002-10-01 Phil Schwan <phil@clusterfs.com>
6689 - locks would be cancelled without throwing away data pages,
6690 resulting in inconsistent data (605627)
6691 - inode attributes were not always being refreshed (605627, 612449)
6692 - lconf now continues to cleanup after lctl reports an error
6693 - MDS now enforces user permissions (602707)
6694 - lprocfs cleanup fixed, but not yet enabled (614157)
6695 - fixed infinite server hang, should a client not respond to an AST
6696 - avoid going into recovery if user calls readlink() with a buffer
6697 that's too small (613941)
6698 - AST RPCs no longer require replies (614867) -- this may be changed
6699 - don't crash server if client sends an IOV that's too big (611336)
6700 - fixed lock conversion deadlock (611892)
6701 - fixed the following of symlinks (614622)
6702 * recovery: the server can remove locks from a client that dies, other
6703 clients can make progress
6704 * more extN patch fixes
6705 * compile-time configurable ptlrpc buffer allocations
6707 - collaborative read cache document
6708 - Lustre Lite Performance CDR document-in-progress
6710 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6713 - fix typo in patch-2.4.18
6715 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6718 - clear ptlrpc request each time in handle_incoming_request()
6719 - unlink of files now destroys the object on the OST
6721 2002-09-19 Peter Braam <braam@clusterfs.com>
6723 * add hard link support
6724 * change obdfile creation method
6725 * kernel patch changed
6727 2002-09-19 Peter Braam <braam@clusterfs.com>
6730 - stack overflow bug in extN fixed
6732 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6734 * documentation updates
6735 - add man pages for config tools
6736 - update tests/README to describe testing with new config tools
6737 - finish metadata API descriptions
6738 * bug fixes and cleanups
6739 - statfs workaround for 16TB limit
6740 - LOV stripe allocation improved, can stripe on subset of OSTs
6741 - LOV file size/IO offset was wrong for files > 4GB in size
6742 - object EA data was being dropped, caused files to be unreadable
6743 - memory overflow with non-LOV OST caused memory corruption
6744 - fixed regression tests to work with new config tools, obdfilter
6745 - fixed bug when directory size became larger than 1 block
6746 - fixed bug (for single client case) when PWD was deleted
6747 - invalidate local directory pages when doing intent-based ops
6748 - avoid LDLM oops when lock callback contained bad data
6750 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6752 * documentation updates
6753 * bug fixes and cleanups
6754 - configuration tools
6757 - 64-bit compile warnings
6758 - 64-bit internal statfs data
6760 * test_brw on persistent OST devices
6762 * lprocfs (disabled)
6764 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6766 * documentation updates
6767 * bug fixes and cleanups
6768 * configuration tools
6770 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6773 * many small fixes to 0_5_4
6774 * io/network handling
6775 * thinkos in MDS operations
6777 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6780 * crucial basic fixes to 0.5.3
6781 * IOR, Iozone work over Elan
6784 2002-08-07 Phil Schwan <phil@clusterfs.com>
6785 * version 0_5_3, our first alpha
6786 * we use the new Portals iovs
6787 * documentation updates
6788 * bug fixes and cleanups
6789 * small changes in the DLM wire protocol
6791 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6792 * version 0_5_1 with some initial stability,
6793 * locking on MD and file I/O.
6794 * documentation updates
6795 * several bug fixes since 0.5.0
6796 * small changes in wire protocol
6798 2002-07-18 Phil Schwan <phil@clusterfs.com>
6800 * delivered as Lustre Light Alpha
6801 * fixed a crash after handling invalid MDS requests
6802 * fixed directory pages for architectures with non-4k pages sizes
6804 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6805 * release version v0_4_4
6806 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6808 * Fixes a number of interruption problems with OST operations.
6809 * Update documentation for portals header changes
6810 * Move all wire protocol structs/defines to lustre_idl.h
6811 * Fixes symlink length bug.
6812 * Add tcpdump to repository.
6814 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6815 * release version v0_4_3
6816 * Fixes statfs for inodes on extN.
6817 * Fixes bug in runtests which would delete /etc/hosts.
6818 * Use 64-bit object IDs wherever possible (not into VFS though)
6819 Remove ost_get_info, which is unused by lustre, and out of date.
6821 2002-07-03 Peter Braam <braam@clusterfs.com>
6822 * release version v0_4_2 Fixes a lookup error (type not passed)
6823 * move forward to head of Portals
6824 * move forward to latest Lustre kernel
6826 2002-06-25 Peter Braam <braam@clusterfs.com>
6827 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob