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 3.10.0-1160.36.2.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.22.1.el8 (RHEL8.3)
16 4.18.0-305.3.1.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 vanilla linux 5.4.21 (ZFS + ldiskfs)
24 vanilla linux 5.4.136 (ZFS + ldiskfs)
25 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
26 * Client primary kernels built and tested during release cycle:
27 3.10.0-1062.18.1.el7 (RHEL7.7)
28 3.10.0-1127.19.1.el7 (RHEL7.8)
29 3.10.0-1160.36.2.el7 (RHEL7.9)
30 4.12.14-95.48 (SLES12 SP4)
31 4.12.14-122.74 (SLES12 SP5)
32 4.12.14-197.75 (SLES15 SP1)
33 5.3.18-24.61 (SLES15 SP2)
34 4.15.0-48 (Ubuntu 18.04)
35 * Other clients known to build on these kernels at some point (others may also work):
36 3.10.0-862.14.4.el7 (RHEL7.5)
37 3.10.0-957.27.2.el7 (RHEL7.6)
38 4.14.0-49.13.1.el7a (RHEL7.5)
39 4.14.0-115.2.2.el7a (RHEL7.6)
40 4.18.0-80.11.2.el8 (RHEL8.0)
41 4.18.0-147.8.1.el8 (RHEL8.1)
42 4.18.0-193.28.1.el8 (RHEL8.2)
43 4.18.0-240.22.1.el8 (RHEL8.3)
44 4.18.0-305.3.1.el8 (RHEL8.4)
45 4.4.120-92.70 (SLES12 SP2)
46 4.4.180-94.100 (SLES12 SP3)
47 4.4.0-131 (Ubuntu 16.04)
48 5.4.0-37 (Ubuntu 20.04)
49 5.4.0 (vanilla kernel.org)
50 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
51 * Recommended ZFS version: 2.0.0
52 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
53 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
54 stack overflow. For more information, please refer to bugzilla 17630.
55 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
56 the re-exporting nodes to be mounted with "32bitapi" mount option
58 --------------------------------------------------------------------------------
62 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
63 for currently supported client and server kernel versions.
64 * Server primary kernels built and tested during release cycle:
65 3.10.0-1062.18.1.el7 (RHEL7.7)
66 3.10.0-1127.19.1.el7 (RHEL7.8)
67 3.10.0-1160.6.1.el7 (RHEL7.9)
68 * Other server kernels known to build and work at some point (others may also work):
69 3.10.0-862.14.4.el7 (RHEL7.5)
70 3.10.0-957.27.2.el7 (RHEL7.6)
71 4.18.0-80.11.2.el8 (RHEL8.0)
72 4.18.0-147.8.1.el8 (RHEL8.1)
73 4.18.0-193.28.1.el8 (RHEL8.2)
74 4.18.0-240.1.1.el8 (RHEL8.3)
75 4.4.120-92.70 (SLES12 SP2)
76 4.4.180-94.100 (SLES12 SP3)
77 4.4.0-131 (Ubuntu 16.04)
78 4.15.0-32 (Ubuntu 18.04)
79 5.4.0-48 (Ubuntu 20.04)
80 vanilla linux 5.4.0 (ZFS + ldiskfs)
81 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
82 * Client primary kernels built and tested during release cycle:
83 3.10.0-1062.18.1.el7 (RHEL7.7)
84 3.10.0-1127.19.1.el7 (RHEL7.8)
85 3.10.0-1160.6.1.el7 (RHEL7.9)
86 4.12.14-95.48 (SLES12 SP4)
87 4.12.14-122.20 (SLES12 SP5)
88 4.12.14-197.75 (SLES15 SP1)
89 5.3.18-24.24 (SLES15 SP2)
90 4.15.0-48 (Ubuntu 18.04)
91 * Other clients known to build on these kernels at some point (others may also work):
92 3.10.0-862.14.4.el7 (RHEL7.5)
93 3.10.0-957.27.2.el7 (RHEL7.6)
94 4.14.0-49.13.1.el7a (RHEL7.5)
95 4.14.0-115.2.2.el7a (RHEL7.6)
96 4.18.0-80.11.2.el8 (RHEL8.0)
97 4.18.0-147.8.1.el8 (RHEL8.1)
98 4.18.0-193.28.1.el8 (RHEL8.2)
99 4.18.0-240.1.1.el8 (RHEL8.3)
100 4.4.120-92.70 (SLES12 SP2)
101 4.4.180-94.100 (SLES12 SP3)
102 4.4.0-131 (Ubuntu 16.04)
103 5.4.0-37 (Ubuntu 20.04)
104 5.4.0 (vanilla kernel.org)
105 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
106 * Recommended ZFS version: 2.0.0
107 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
108 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
109 stack overflow. For more information, please refer to bugzilla 17630.
110 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
111 the re-exporting nodes to be mounted with "32bitapi" mount option
113 --------------------------------------------------------------------------------
117 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
118 for currently supported client and server kernel versions.
119 * Server primary kernels built and tested during release cycle:
120 3.10.0-1062.1.1.el7 (RHEL7.7)
121 * Other server kernels known to build and work at some point (others may also work):
122 3.10.0-862.14.4.el7 (RHEL7.5)
123 3.10.0-957.27.2.el7 (RHEL7.6)
124 4.18.0-80.7.1.el8 (RHEL8.0)
125 4.4.120-92.70 (SLES12 SP2)
126 4.4.180-94.100 (SLES12 SP3)
127 4.4.0-131 (Ubuntu 16.04)
128 4.15.0-32 (Ubuntu 18.04)
129 vanilla linux 4.6.7 (ZFS only)
130 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
131 * Client primary unpatched kernels built and tested during release cycle:
132 3.10.0-1062.1.1.el7 (RHEL7.7)
133 4.12.14-95.29 (SLES12 SP4)
134 4.15.0-48 (Ubuntu 18.04)
135 * Other clients known to build on unpatched kernels at some point (others may also work):
136 3.10.0-862.14.4.el7 (RHEL7.5)
137 3.10.0-957.27.2.el7 (RHEL7.6)
138 4.14.0-49.13.1.el7a (RHEL7.5)
139 4.14.0-115.2.2.el7a (RHEL7.6)
140 4.18.0-80.7.1.el8 (RHEL8.0)
141 4.4.120-92.70 (SLES12 SP2)
142 4.4.180-94.100 (SLES12 SP3)
143 4.4.0-131 (Ubuntu 16.04)
144 vanilla linux 4.15.0 (ZFS only)
145 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
146 * Recommended ZFS / SPL version: 0.7.13
147 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
148 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
149 stack overflow. For more information, please refer to bugzilla 17630.
150 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
151 the re-exporting nodes to be mounted with "32bitapi" mount option
152 * Note for Ubuntu 14.04.5:
153 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
154 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
156 --------------------------------------------------------------------------------
160 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
161 for currently supported client and server kernel versions.
162 * Server known to build on the following kernels (others may also work):
163 2.6.32-696.18.7.el6 (RHEL6.9)
164 3.10.0-862.14.4.el7 (RHEL7.5)
165 3.10.0-957.el7 (RHEL7.6)
166 4.4.120-92.70 (SLES12 SP2)
167 4.4.162-94.69 (SLES12 SP3)
168 4.4.0-131 (Ubuntu 16.04)
169 4.15.0-32 (Ubuntu 18.04)
170 vanilla linux 4.6.7 (ZFS only)
171 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
172 * Client known to build on unpatched kernels (others may also work):
173 2.6.32-696.18.7.el6 (RHEL6.9)
174 2.6.32-754.6.3.el6 (RHEL6.10)
175 3.10.0-862.14.4.el7 (RHEL7.5)
176 3.10.0-957.el7 (RHEL7.6)
177 4.14.0-49.13.1.el7a (RHEL7.5)
178 4.14.0-115.2.2.el7a (RHEL7.6)
179 4.4.120-92.70 (SLES12 SP2)
180 4.4.162-94.69 (SLES12 SP3)
181 4.4.0-131 (Ubuntu 16.04)
182 4.15.0-32 (Ubuntu 18.04)
183 vanilla linux 4.15.0 (ZFS only)
184 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
185 * Recommended ZFS / SPL version: 0.7.9
186 * Tested with ZFS / SPL version: 0.7.9
187 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
188 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
189 stack overflow. For more information, please refer to bugzilla 17630.
190 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
191 the re-exporting nodes to be mounted with "32bitapi" mount option
192 * Note for Ubuntu 14.04.5:
193 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
194 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
196 --------------------------------------------------------------------------------
198 2018-04-03 Intel Corporation
200 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
201 for currently supported client and server kernel versions.
202 * Server known to build on patched kernels:
203 2.6.32-431.29.2.el6 (RHEL6.5)
204 2.6.32-504.30.3.el6 (RHEL6.6)
205 2.6.32-573.26.1.el6 (RHEL6.7)
206 2.6.32-642.15.1.el6 (RHEL6.8)
207 2.6.32-696.18.7.el6 (RHEL6.9)
208 3.10.0-693.21.1.el7 (RHEL7.4)
209 3.0.101-0.47.71 (SLES11 SP3)
210 3.0.101-107 (SLES11 SP4)
211 3.12.74-60.64.40 (SLES12 SP1)
212 4.4.114-92.64 (SLES12 SP2)
213 4.4.114-94.11 (SLES12 SP3)
214 3.13.0-101 (Ubuntu 14.04, ZFS only)
215 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
216 4.4.0-116 (Ubuntu 16.04)
217 vanilla linux 4.6.7 (ZFS only)
218 * Client known to build on unpatched kernels:
219 2.6.32-431.29.2.el6 (RHEL6.5)
220 2.6.32-504.30.3.el6 (RHEL6.6)
221 2.6.32-573.26.1.el6 (RHEL6.7)
222 2.6.32-642.15.1.el6 (RHEL6.8)
223 2.6.32-696.18.7.el6 (RHEL6.9)
224 3.10.0-693.21.1.el7 (RHEL7.4)
225 3.0.101-0.47.71 (SLES11 SP3)
226 3.0.101-107 (SLES11 SP4)
227 3.12.74-60.64.40 (SLES12 SP1)
228 4.4.114-92.64 (SLES12 SP2)
229 4.4.114-94.11 (SLES12 SP3)
230 3.13.0-101 (Ubuntu 14.04)
231 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
232 4.4.0-116 (Ubuntu 16.04)
234 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
235 * Recommended ZFS / SPL version: 0.7.6
236 * Tested with ZFS / SPL version: 0.7.6
237 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
238 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
239 stack overflow. For more information, please refer to bugzilla 17630.
240 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
241 the re-exporting nodes to be mounted with "32bitapi" mount option
242 * Note for Ubuntu 14.04.5:
243 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
244 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
246 --------------------------------------------------------------------------------
248 2017-07-05 Intel Corporation
250 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
251 for currently supported client and server kernel versions.
252 * Server known to build on patched kernels:
253 2.6.32-431.29.2.el6 (RHEL6.5)
254 2.6.32-504.30.3.el6 (RHEL6.6)
255 2.6.32-573.26.1.el6 (RHEL6.7)
256 2.6.32-642.15.1.el6 (RHEL6.8)
257 2.6.32-696.3.1.el6 (RHEL6.9)
258 3.10.0-514.21.1.el7 (RHEL7.3)
259 3.0.101-0.47.71 (SLES11 SP3)
260 3.0.101-100 (SLES11 SP4)
261 3.12.74-60.64.40 (SLES12 SP1)
262 4.4.59-92.17 (SLES12 SP2)
263 vanilla linux 4.6.7 (ZFS only)
264 * Client known to build on unpatched kernels:
265 2.6.32-431.29.2.el6 (RHEL6.5)
266 2.6.32-504.30.3.el6 (RHEL6.6)
267 2.6.32-573.26.1.el6 (RHEL6.7)
268 2.6.32-642.15.1.el6 (RHEL6.8)
269 2.6.32-696.3.1.el6 (RHEL6.9)
270 3.10.0-514.21.1.el7 (RHEL7.3)
271 3.0.101-0.47.71 (SLES11 SP3)
272 3.0.101-100 (SLES11 SP4)
273 3.12.74-60.64.40 (SLES12 SP1)
274 4.4.59-92.17 (SLES12 SP2)
276 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
277 * Recommended ZFS / SPL version: 0.6.5.9
278 * Tested with ZFS / SPL version: 0.6.5.9
279 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
280 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
281 stack overflow. For more information, please refer to bugzilla 17630.
282 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
283 the re-exporting nodes to be mounted with "32bitapi" mount option
285 --------------------------------------------------------------------------------
287 12-07-2016 Intel Corporation
289 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
290 for currently supported client and server kernel versions.
291 * Server known to build on patched kernels:
292 2.6.32-431.29.2.el6 (RHEL6.5)
293 2.6.32-504.30.3.el6 (RHEL6.6)
294 2.6.32-573.26.1.el6 (RHEL6.7)
295 2.6.32-642.11.1.el6 (RHEL6.8)
296 3.10.0-514.el7 (RHEL7.3)
297 3.0.101-0.47.71 (SLES11 SP3)
298 3.0.101-80 (SLES11 SP4)
299 3.12.62-60.64.8 (SLES12 SP1)
300 vanilla linux 4.5.7 (ZFS only)
301 * Client known to build on unpatched kernels:
302 2.6.32-431.29.2.el6 (RHEL6.5)
303 2.6.32-504.30.3.el6 (RHEL6.6)
304 2.6.32-573.26.1.el6 (RHEL6.7)
305 2.6.32-642.11.1.el6 (RHEL6.8)
306 3.10.0-514.el7 (RHEL7.3)
307 3.0.101-0.47.71 (SLES11 SP3)
308 3.0.101-80 (SLES11 SP4)
309 3.12.62-60.64.8 (SLES12 SP1)
311 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
312 * Recommended ZFS / SPL version: 0.6.5.8
313 * Tested with ZFS / SPL version: 0.6.5.8
314 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
315 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
316 stack overflow. For more information, please refer to bugzilla 17630.
317 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
318 the re-exporting nodes to be mounted with "32bitapi" mount option
320 --------------------------------------------------------------------------------
322 02-29-2016 Intel Corporation
324 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
325 for currently supported client and server kernel versions.
326 * Server known to build on patched kernels:
327 2.6.32-431.29.2.el6 (RHEL6.5)
328 2.6.32-504.30.3.el6 (RHEL6.6)
329 2.6.32-573.12.1.el6 (RHEL6.7)
330 3.10.0-327.3.1.el7 (RHEL7.2)
331 3.0.101-0.47.71 (SLES11 SP3)
332 3.0.101-68 (SLES11 SP4)
333 vanilla linux 4.2.1 (ZFS only)
334 * Client known to build on unpatched kernels:
335 2.6.32-431.29.2.el6 (RHEL6.5)
336 2.6.32-504.30.3.el6 (RHEL6.6)
337 2.6.32-573.12.1.el6 (RHEL6.7)
338 3.10.0-327.3.1.el7 (RHEL7.2)
339 3.0.101-0.47.71 (SLES11 SP3)
340 3.0.101-68 (SLES11 SP4)
343 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
344 * Recommended ZFS / SPL version: 0.6.4.2
345 * Tested with ZFS / SPL version: 0.6.4.2
346 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
347 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
348 stack overflow. For more information, please refer to bugzilla 17630.
349 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
350 the re-exporting nodes to be mounted with "32bitapi" mount option
352 --------------------------------------------------------------------------------
354 03-10-2015 Intel Corporation
356 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
357 for currently supported client and server kernel versions.
358 * Server known to build on patched kernels:
359 2.6.32-431.29.2.el6 (RHEL6.5)
360 2.6.32-504.8.1.el6 (RHEL6.6)
361 3.0.101-0.46 (SLES11 SP3)
362 * Client known to build on unpatched kernels:
363 2.6.32-431.29.2.el6 (RHEL6.5)
364 2.6.32-504.8.1.el6 (RHEL6.6)
365 3.10.0-123.20.1.el7 (RHEL7)
366 3.0.101-0.46 (SLES11 SP3)
367 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
368 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
369 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
370 stack overflow. For more information, please refer to bugzilla 17630.
371 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
372 the re-exporting nodes to be mounted with "32bitapi" mount option
374 Severity : enhancement
376 Description: control OST-index in IDIF via ROCOMPAT flag.
377 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
378 last_rcvd file. For new formatted OST device, it will be auto set;
379 for the case of upgrading from old OST device, you can enable it
380 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
381 enabled, for new created OST-object, its IDIF-in-LMA will contain
382 the OST-index; for the existing OST-object, the OSD will convert
383 old format IDIF as new format IDIF with OST-index stored in the
384 LMA EA when accessing such OST-object or via OI scrub. Once such
385 flag is enabled, it cannot be reverted back, so the system cannot
386 be downgraded to the orignal incompatible version.
388 --------------------------------------------------------------------------------
390 07-30-2014 Intel Corporation
392 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
393 for currently supported client and server kernel versions.
394 * Server known to build on patched kernels:
395 2.6.32-431.20.3.el6 (RHEL6.5)
396 3.0.101-0.31 (SLES11 SP3)
397 * Client known to build on unpatched kernels:
398 2.6.32-431.20.3.el6 (RHEL6.5)
399 3.0.101-0.31 (SLES11 SP3)
402 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
403 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
404 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
405 stack overflow. For more information, please refer to bugzilla 17630.
406 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
407 the re-exporting nodes to be mounted with "32bitapi" mount option
409 --------------------------------------------------------------------------------
411 10-31-2013 Intel Corporation
413 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
414 * Server builds for kernels:
415 2.6.32-358.18.1.el6 (RHEL6.4)
416 * Client builds for unpatched kernels:
417 2.6.32-358.18.1.el6 (RHEL6.4)
418 3.0.92-0.8 (SLES11 SP3)
420 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
421 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
422 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
423 stack overflow. For more information, please refer to bugzilla 17630.
424 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
425 the re-exporting nodes to be mounted with "32bitapi" mount option
427 --------------------------------------------------------------------------------
429 05-30-2013 Intel Corporation
431 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
432 * Server builds for kernels:
433 2.6.32-358.6.2.el6 (RHEL6.4)
434 * Client builds for unpatched kernels:
435 2.6.32-358.6.2.el6 (RHEL6.4)
436 3.0.74-0.6.6 (SLES11 SP2)
438 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
439 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
440 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
441 stack overflow. For more information, please refer to bugzilla 17630.
442 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
443 the re-exporting nodes to be mounted with "32bitapi" mount option
445 --------------------------------------------------------------------------------
447 9-30-2012 Whamcloud, Inc.
449 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
450 * Server support for kernels:
451 2.6.32-279.5.1.el6 (RHEL6)
452 * Client support for unpatched kernels:
453 2.6.32-279.5.1.el6 (RHEL6)
455 2.6.18-238.19.1.el5 (RHEL5)
456 2.6.16.60-0.69.1 (SLES10)
457 * Recommended e2fsprogs version: 1.41.90.wc4
458 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
459 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
460 stack overflow. For more information, please refer to bugzilla 17630.
461 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
462 the re-exporting nodes to be mounted with "32bitapi" mount option
464 Severity : enhancement
466 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
468 Severity : enhancement
470 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
472 Severity : enhancement
474 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
476 Severity : enhancement
478 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
480 Severity : enhancement
482 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
484 --------------------------------------------------------------------------------
486 04-30-2012 Whamcloud, Inc.
488 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
489 * Server support for kernels:
490 2.6.32-220.4.2.el6 (RHEL6)
491 * Client support for unpatched kernels:
492 2.6.32-220.4.2.el6 (RHEL6)
494 2.6.18-238.19.1.el5 (RHEL5)
495 2.6.16.60-0.69.1 (SLES10)
496 * Recommended e2fsprogs version: 1.41.90.wc4
497 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
498 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
499 stack overflow. For more information, please refer to bugzilla 17630.
500 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
501 the re-exporting nodes to be mounted with "32bitapi" mount option
503 Severity : enhancement
505 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
507 Severity : enhancement
509 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
511 Severity : enhancement
513 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
515 Severity : enhancement
517 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
519 --------------------------------------------------------------------------------
521 09-30-2011 Whamcloud, Inc.
523 * Server support for kernels:
524 2.6.18-238.19.1.el5 (RHEL5)
525 2.6.32-131.6.1.el6 (RHEL6)
526 * Client support for unpatched kernels:
527 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
528 2.6.18-238.19.1.el5 (RHEL5)
529 2.6.32-131.6.1.el6 (RHEL6)
530 2.6.29.4-167.fc11 (FC11)
531 2.6.16.60-0.69.1 (SLES10)
533 * Recommended e2fsprogs version: 1.41.90.wc3
534 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
535 Lustre filesystem with 4K stack may cause a stack overflow. For more
536 information, please refer to bugzilla 17630.
537 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
538 mounted with "32bitapi" mount option
540 Severity : enhancement
541 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
543 Severity : enhancement
544 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
546 Severity : enhancement
547 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
550 Frequency : only with SLES10
552 Description: Use OFED "KMP" provided by Novell
553 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
554 Lustre is now built against this package. That means you need to
555 install the ofed-kmp package from Novell for the patchless client.
556 Note that the ofed-kmp that Novell ships may not exactly match the
557 kernel version but should still be compatible.
559 Severity : enhancement
561 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
562 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
564 --------------------------------------------------------------------------------
566 2010-07-15 Oracle, Inc.
568 * Server support for kernels:
569 2.6.18-164.11.1.el5 (RHEL 5)
570 2.6.18-164.11.1.0.1.el5 (OEL 5)
571 * Client support for unpatched kernels:
572 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
573 2.6.18-164.11.1.el5 (RHEL 5),
574 2.6.18-164.11.1.0.1.el5 (OEL 5)
575 2.6.16.60-0.42.8 (SLES 10),
577 2.6.29.4-167.fc11 (FC11)
578 * Recommended e2fsprogs version: 1.41.10-sun2
580 Severity : enhancement
582 Description: Distribute lfs_migrate script with lustre
583 Details : The lfs_migrate mentioned in the manual is now included in the
584 lustre tree in lustre/scripts and will distributed in lustre RPMs.
586 Severity : enhancement
588 Description: Failover nid list added to import proc file.
590 Severity : enhancement
592 Description: "som_preview" mount option required to use SOM
593 Details : The clients will need to use som_preview mount option
597 Description: Lustre does not do 1MB IOs to HW RAID
598 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
600 Severity : enhancement
602 Description: lfs quota output is not very convenient for awk/sed-parsing
603 Details : Some positions in lfs quota output table could be empty or
604 non-empty which made it hard to parse it with scripts, now
605 a dash is put instead of space where there is not supposed
610 Description: Update to OFED 1.5.1
611 Details : OFED 1.5.1 will be used in environments that don't use
612 vendor supplied OFED.
616 Description: Remove lnet. prefix from lctl params display
617 Details : Also renamed lnet.memused to lnet_memused to
622 Description: Use conf_param -d to remove permanant settings.
626 Description: Add list_param -R
627 Details : lctl list_param now supports recursive mode.
630 Bugzilla : 21829, 22312
631 Description: Fix llobdstat.
632 Details : 1. Now updating the write_bytes counter
633 2. Port script fixes from 1.8.
637 Description: decrease the usage of memory on clients.
638 Details : 1. On clients, recycle dentries and inodes unused.
639 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
644 Description: llobdstat fix and enhancement
645 Details : add a counter to set a limit to how many samples will be returned
646 fix a wildcard in the path to limit to obdfilter stats only
650 Description: lfs check only allowed for root
651 Details : Code cleanups and added a regresion test of non-root use of
656 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
657 Details : Fix hr->hr_index overload.
661 Description: Fix FIEMAP functionality.
665 Description: FC11 patchless client support added
669 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
670 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
674 Description: New acc-sm script and various test-framework enhancments.
678 Description: Previously landed patch was reverted.
682 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
686 Description: Add ldiskfs maxdirsize mount option.
690 Description: Fix mkfs.lustre support for 16TB LUNs.
694 Description: Print mmp_check_interval
695 Details : print mmp_check_interval and make it possible to abort
696 mount operation in case it takes too long.
700 Description: Add cascading_rw to lustre mpi tests.
704 Description: Store SOM data in EA.
705 Details : Also several SOM related cleanups.
709 Description: Add refcount to md_open_data
713 Description: Use refcount on lov_request_set to prevent memory corruption.
717 Description: Statahead: use parent dir's i_mutex for synchronization.
721 Description: loadgen improvements.
725 Description: Enable peer health detection in ksocklnd and ko2iblnd.
729 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
733 Description: Fix race on obd_devcie:md_stats
737 Description: Fix SLES 11 client builds.
738 Details : Resolved uninitalized pointer gcc warnings.
742 Description: Allow non-root access for "lfs check".
743 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
747 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
749 Severity : enhancement
751 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
752 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
756 Description: Sanity test_27p fails.
757 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
758 Sanity tests 27 were cleaned up and improved.
762 Description: Unify req format on client/servers
763 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
764 buffers and swabbers approach..
769 Description: drop join file stuff.
771 Severity : enhancement
773 Description: Add 2.6.27 and SLES11 (patchless client) support
777 Description: Fix in ptlrpc_expire_one_request() to print the signed time
780 Severity : enhancement
782 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
785 Severity : enhancement
787 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
789 Severity : enhancement
791 Description: correctly handle big reply message.
792 Details : send LNet event if reply is bigger than buffer and adjust this buffer
797 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
799 Severity : enhancement
801 Description: Add support for OEL5.
805 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
807 Severity : enhancement
809 Description: "weak-modules" support
810 Details : Implement "weak-modules" support which enables kernel modules
811 to be used with any kernel that implements the same kABI. In
812 order to achieve this modules are now installed in
813 /lib/modules/$(uname -r)/updates/kernel on all distributions.
817 Description: client could not reconnect to OST because of an active request.
818 Details : The client now retries to reconnect to the same server, if a
819 connect request failed with EBUSY or -EAGAIN. Test case 35b
820 conf-sanity.sh has been created.
824 Description: Don't allow make backward step on assignin osc next id.
825 Details : race between allocation next id and ll_sync thread can be cause
826 of set wrong osc next id and can be kill valid ost objects.
830 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
834 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
838 Description: Error handling in osc_statfs_interpret() has been improved.
839 Details : Check in osc_statfs_interpret() for EBADR.
843 Description: Do not send statfs() requests to OSTs disabled by administrator.
844 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
846 Severity : enhancement
848 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
850 Severity : enhancement
852 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
854 Severity : enhancement
856 Description: Enhancement for lfs(1) command to use numeric uid/gid.
860 Description: Improved cleanup procedure for conf-sanity test_46a.
864 Description: Do not start bulk transfer, if client tried to reconnect.
865 Details : Do not start bulk transfer, if client tried to reconnect,
866 but failed due to active rpcs. Check for exp_abort_active_req
867 has been implemented.
871 Description: Prevent inconsistences between linux and lustre mount structures.
872 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
873 Make the sleep interruptible.
875 Severity : enhancement
877 Description: provide server to client comms path
878 Details : server to client communications path via new LDLM_SET_INFO rpc
882 Description: truncate starts GFP_FS allocation under transaction causing deadlock
883 Details : ldiskfs_truncate calls grab_cache_page which may start page
884 allocation under an open transaction. This may lead to
885 calling prune_icache with consequent lustre reentrance.
889 Description: send correctly lsm on open replay
890 Details : MDS is trust to LSM size on replay open, but client can set wrong size
895 Description: enable client interface failover
896 Details : When a child reconnects from another NID, properly update export
897 nid hash position and ldlm reverse import.
899 Severity : enhancement
901 Description: Communicate OST degraded/readonly state via statfs to MDS
902 Details : Flags in the statfs returned from OSTs indicate whether the
903 OST is in a degraded RAID state, or if the filesystem has
904 turned read-only after a filesystem error is detected.
909 Description: don't panic if EPROTO was hit when reading symlink
910 Details : correctly handling request reference in error cases.
912 Severity : enhancement
914 Description: Add LustreNetLink, a kernel-userspace communcation path.
916 Severity : enhancement
918 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
923 Description: access to llog context before init.
924 Details : move handling CATALOGS file at osc layer and forbid access to llog
928 Frequency : rare, only when client is writing to OST on local node
930 Description: Avoid deadlock for local client writes
931 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
932 memory freeing context. This allows OST threads to set the
933 PF_MEMALLOC flag on task structures in order to allocate memory
934 from reserved pools and complete IO.
935 Use GFP_HIGHUSER for OST allocations for non-local client writes,
936 so that the OST threads generate memory pressure and allow
937 inactive pages to be reclaimed.
939 Severity : enhancement
941 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
946 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
947 Details : move ldlm namespace creation in setup phase to avoid grab
948 _lprocfs_lock with cli_sem held.
952 Description: Temporarily disable grant shrink.
953 Details : Disable the feature for debugging.
956 Frequency : only when formatting test filesystems
958 Description: Unable to run several mkfs.lustre on loop devices at the same time
959 Details : mkfs.lustre returns error 256 on the concurrent loop devices
960 formatting. The solution is to proper handle the error.
962 Severity : enhancement
964 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
966 Severity : enhancement
968 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
973 Description: resolve race between obd_disconnect and class_disconnect_exports
974 Details : if obd_disconnect will be called to already disconnected export he
975 forget release one reference and osc module can't unloaded.
977 Severity : enhancement
979 Description: Allow tuning service thread via /proc
980 Details : For each service a new
981 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
982 created that can be used to set min/max thread counts, and get the
983 current number of running threads.
988 Descriptoin: don't return error if have partially created objects for file.
989 Details : lov_update_create_set uses set->set_success as index for created
990 objects, so if some requests will be failed, they will have hole
991 at end of array and qos_shrink_lsm for allocate correct lsm.
993 Severity : enhancement
995 Description: Update OFED support to 1.4.1
1000 Description: Reduce small size read RPC
1001 Details : Set read-ahead limit for every file and only do read-ahead when
1002 available read-ahead pages are bigger than 1M to avoid small size
1005 Severity : enhancement
1007 Description: move AT tunable parameters for more consistent usage
1008 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1010 Severity : enhancement
1012 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1013 Details : allow skip disconnected ost for send statfs request and hide error
1019 Description: fix racy locking of mballoc block bitmaps causing BUG
1020 Details : The locking of the mballoc buddy bitmap and the in-memory
1021 block bitmap was using two different spin locks in some
1022 cases. This made it possible to incorrectly access the
1023 mballoc bitmap while another process was modifying it,
1024 causing a sanity assertion to fail. While no on-disk corruption
1025 was reported, there was some risk of this happening.
1028 Frequency : rare, on llog test 6
1030 Descriptoin: don't allow connect to already connected import
1031 Details : allowing connect to already connected import is hide connecting problem.
1036 Descriptoin: allow kill process which wait statahead result
1037 Details : for some reasons 'ls' can stick in waiting result from statahead,
1038 in this case need way for kill this process.
1040 Severity : enhancement
1042 Description: Add state history info file, enhance import info file
1043 Details : Track import connection state changes in a new osc/mdc proc file;
1044 add overview-type data to the osc/mdc import proc file.
1046 Severity : enhancement
1048 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1050 Severity : enhancement
1052 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1057 Descriptoin: OST grant too much space to client even there are not enough space.
1058 Details : Client will shrink its grant cache to OST if there are no write
1059 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1060 this grant cache if there are already not enough avaible space
1061 (left_space < total_clients * 32M).
1064 Frequency : start MDS on uncleanly shutdowned MDS device
1066 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1067 Details : stay in waiting mds<>ost recovery finished produce random bugs
1068 due race between two ll_sync thread for one lov target. send
1069 ACTIVATE event only if connect realy finished and import have
1073 Frequency : rare, connect and disconnect target at same time
1075 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1076 Details : don't call obd_disconnect under lov_lock. this long time
1077 operation and can block ptlrpcd which answer to connect request.
1082 Descriptoin: don't lose wakeup for imp_recovery_waitq
1083 Details : recover_import_no_retry or invalidate_import and import_close can
1084 both sleep on imp_recovery_waitq, but we was send only one wakeup
1088 Frequency : always with long access acl
1090 Descriptoin: mds can't pack reply with long acl.
1091 Details : mds don't control size of acl but they limited by reint/getattr
1094 Severity : enhancement
1096 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1098 Severity : enhancement
1100 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1103 Frequency : start MDS on uncleanly shutdowned MDS device
1105 Descriptoin: aborting recovery hang on MDS
1106 Details : don't throttle destroy RPCs for the MDT.
1109 Frequency : on remount
1111 Description: external journal device not working after the remount
1112 Details : clear dev_rdonly flag for external journal devices in
1118 Description: shutdown vs evict race
1119 Details : client_disconnect_export vs connect request race.
1120 if client will evicted at this time - we start invalidate
1121 thread without referece to import and import can be freed
1125 Frequency : rare, need acl's on inode.
1127 Description: client can't handle ost additional correctly
1128 Details : if ost was added after client connected to mds client can have
1129 hit lnet_try_match_md ... to big messages to wide striped files.
1130 in this case need teach client to handle config events about add
1131 lov target and update client max ea size at that event.
1133 Severity : enhancement
1135 Description: Changelogs
1136 Details : Changelogs are a lightweight mechanism to track filesystem
1137 metadata and namespace changes. The changelog is recorded
1138 permanently on the MDTs, and is periodically "consumed" / purged
1139 when records are no longer needed.
1141 Severity : enhancement
1143 Description: compact fld format with extents
1144 Details : Store range of seq rather than every seq in FLD. Seq
1145 controller update FLD rather than clients. In Case of CMD, mdt0
1146 has FLD, all other metadata server act as non persistent proxy
1147 for FLD queries and cache fld entries in fld cache.
1152 Description: don't skip ost target if they assigned to file
1153 Details : Drop slow OSCs if we can, but not for requested start idx.
1154 This means "if OSC is slow and it is not the requested
1155 start OST, then it can be skipped, otherwise skip it only
1156 if it is inactive/recovering/out-of-space.
1160 Description: more cleanup in mds_lov
1161 Details : not send LOV EA under replay, we can't know about they size at this
1162 time. Don't allow client connect to mds before any ost connected,
1163 for avoid problems with LOV EA size and returning EIO to client.
1165 Severity : enhancement
1167 Description: Interoperability at server side (Disk interoperability)
1169 Severity : enhancement
1171 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1173 Severity : enhancement
1175 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1177 Severity : enhancement
1179 Description: New client IO stack (CLIO).
1181 Severity : enhancement
1183 Description: Commit on sharing. Eliminate inter-client dependencies between
1184 uncommitted transactions by doing transaction commits.
1185 Thereby clients may recovery independently.
1188 Frequency : Create a symlink file with a very long name
1190 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1191 Details : If there is no extra space in the request for early cancels,
1192 ldlm_req_handles_avail() returns 0 instead of a negative value.
1194 Severity : enhancement
1196 Description: Add /proc entry for import status
1197 Details : The mdc, osc, and mgc import directories now have
1198 an import directory that contains useful import data for debugging
1199 connection problems.
1201 Severity : enhancement
1203 Description: Re-disable certain /proc logging
1204 Details : Enable and disable client's offset_stats, extents_stats and
1205 extents_stats_per_process stats logging on the fly.
1208 Frequency : Only on FC kernels 2.6.22+
1210 Description: oops in statahead
1211 Details : Do not drop reference count for the dentry from VFS when lookup,
1212 VFS will do that by itself.
1214 Severity : enhancement
1216 Description: Generic /proc file permissions
1217 Details : Set /Proc file permissions in a more generic way to enable non-
1218 root users operate on some /proc files.
1222 Description: Hitting mdc_commit_close() ASSERTION
1223 Details : Properly handle request reference release in
1224 ll_release_openhandle().
1228 Description: quota recovery deadlock during mds failover
1229 Details : This patch includes att18982, att18236, att18237 in bz14840.
1231 1. fix osts hang when mds does failover with quotaon
1232 2. prevent watchdog storm when osts threads wait for the
1237 Frequency : only patchless client
1238 Description: add workaround for race between add/remove dentry from hash
1240 Severity : enhancement
1242 Description: Allow OST glimpses to return PW locks
1246 Description: LBUG when llog conf file is full
1247 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1251 Description: Prevent import from entering FULL state when server in recovery
1255 Description: service mount cannot take device name with ":"
1256 Details : Only when device name contains ":/" will mount treat it as
1262 Description: replace ptlrpcd with the statahead thread to interpret the async
1263 statahead RPC callback
1267 Frequency : on recovery
1268 Description: I/O failures after umount during fail back
1269 Details : if client reconnected to restarted server we need join to recovery
1270 instead of find server handler is changed and process self
1271 eviction with cancel all locks.
1273 Severity : enhancement
1275 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1277 Severity : enhancement
1279 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1281 Severity : enhancement
1283 Description: Upcall on Lustre log has been dumped
1284 Details : Allow for a user mode script to be called once a Lustre log has
1285 been dumped. It passes the filename of the dumped log to the
1286 script, the location of the script can be specified via
1287 /proc/sys/lnet/debug_log_upcall.
1292 Description: avoid idr_remove called for id which is not allocated.
1293 Details : Move assigment s_dev for clustered nfs to end of initialization,
1294 to avoid problem with error handling.
1299 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1300 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1301 into CONN_USED_HASH and this prodice warning when put connection
1302 again in unused hash.
1305 Severity : enhancement
1307 Description: File striping can now be set to use an arbitrary pool of OSTs.
1309 Severity : enhancement
1311 Description: Export bytes_read/bytes_write count on OSC/OST.
1315 Description: Early reply size mismatch, MGC loses connection
1316 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1317 the connect flags are properly negotiated.
1322 Description: quotas are not honored with O_DIRECT
1323 Details : all writes with the flag O_DIRECT will use grants which leads to
1324 this problem. Now using OBD_BRW_SYNC to guard this.
1328 Description: add quota statistics
1329 Details : 1. sort out quota proc entries and proc code.
1330 2. add quota statistics
1332 Severity : enhancement
1334 Description: enable quota support for HEAD.
1338 Description: Properly propagate oinfo flags from lov to osc for statfs
1339 Details : restore missing copy oi_flags to lov requests.
1341 Severity : enhancement
1343 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1344 : l_getgroups(8), lst(8), routerstat(8)
1345 Details : included man pages for llobdstat(8), llstat(8),
1346 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1348 Severity : enhancement
1350 Description: configure's --enable-quota should check the
1351 : kernel .config for CONFIG_QUOTA
1352 Details : configure is terminated if --enable-quota is passed but
1353 : no quota support is in kernel
1357 Description: Remove portals compatibility
1358 Details : Remove portals compatibility, not interoperable with releases
1363 Description: Resolve device initialization race
1364 Details : Prevent proc handler from accessing devices added to the
1365 obd_devs array but yet be intialized.
1367 Severity : enhancement
1369 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1371 Severity : enhancement
1373 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1378 Description: Using wrong pointer in osc_brw_prep_request
1379 Details : Access to array[-1] can produce panic if kernel compiled with
1380 CONFIG_PAGE_ALLOC enabled
1384 Description: Client runs out of low memory
1385 Details : Consider only lowmem when counting initial number of llap pages
1389 Description: *optional* service tags registration
1390 Details : if the "service tags" package is installed on a Lustre node
1391 When the filesystem is mounted, a local-node service tag will
1392 be created. See http://inventory.sun.com/ for more information
1393 about the Service Tags asset management system.
1397 Description: Kernel BUG tries to release flock
1398 Details : Lustre does not destroy flock lock before last reference goes
1399 away. So always drop flock locks when client is evicted and
1400 perform unlock regardless of successfulness of speaking to MDS.
1404 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1408 Description: sanity test 65a fails if stripecount of -1 is set
1409 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1414 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1415 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1418 Severity : enhancement
1420 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1424 Description: do not process already freed flock
1425 Details : flock can possibly be freed by another thread before it reaches
1426 to ldlm_flock_completion_ast.
1430 Description: LBUG during stress test
1431 Details : Need properly lock accesses the flock deadlock detection list.
1435 Description: oops in page fault handler
1436 Details : kernel page fault handler can return two special 'pages' in error
1437 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1441 Description: timeout with invalidate import.
1442 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1443 should be handled by ptlrpcd. This produce long age waiting and
1444 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1446 Severity : enhancement
1448 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1452 Description: enable MGS and MDT services start separately
1453 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1454 a co-located MGS without starting the MGS, which is a complement
1455 to 'nosvc' mount option.
1459 Frequency : after recovery
1460 Description: precreate to many object's after del orphan.
1461 Details : del orphan st in oscc last_id == next_id and this triger growing
1462 count of precreated objects. Set flag LOW to skip increase count
1463 of precreated objects.
1467 Frequency : rare, on clear nid stats
1468 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1469 Details : when clean nid stats sometimes try destroy live entry,
1470 and this produce panic in free.
1474 Description: Stack overflow during MDS log replay
1475 ease stack pressure by using a thread dealing llog_process.
1479 Description: wait until IO finished before start new when do lock cancel.
1480 Details : VM protocol want old IO finished before start new, in this case
1481 need wait until PG_writeback is cleared until check dirty flag and
1482 call writepages in lock cancel callback.
1484 Severity : enhancement
1486 Description: using special macro for print time and cleanup in includes.
1490 Description: mds_mfd_close() ASSERTION(rc == 0)
1491 Details : In mds_mfd_close(), we need protect inode's writecount change
1492 within its orphan write semaphore to prevent possible races.
1496 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1500 Frequency : rare, on shutdown ost
1501 Description: don't hit live lock with umount ost.
1502 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1503 use shrink_dcache_sb instead.
1507 Description: don't panic with use echo client
1508 Details : echo client pass NULL as client nid pointer and this produce null
1509 pointer dereference.
1513 Description: fix build on ppc32
1514 Details : compile code with -m64 flag produce wrong object file for ppc32.
1518 Description: add message levels for liblustreapi
1522 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1523 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1524 will result in an -EAGAIN and in the exit call path we call
1525 alloc_rr() which will with increasing aggressiveness attempt to
1526 aquire precreated objects on the minimum number of required OSCs.
1530 Description: Use old size assignment to avoid deadlock
1531 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1532 the scheduling while holding a spinlock. We do not need locking
1533 for size in ll_update_inode() because size is only updated from
1534 the MDS for directories or files without objects, so there is no
1535 other place to do the update, and concurrent access to such inodes
1536 are protected by the inode lock.
1540 Description: resolve "_IOWR redefined" build error on SLES10
1544 Description: dump the memory debugging after all modules are unloaded to
1545 suppress false negative in conf_sanity test 39
1547 Severity : enhancement
1549 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1554 Description: fix SLES kernel versioning
1555 Details : the kernel version for our SLES 10 kernel did not include a "-"
1556 before the "smp" at the end. while this was not a problem in
1557 general, it did mean that software trying to use the kernel
1558 version to try to detect a vendor specific kernel would fail.
1559 this was most evident by the OFED build scripts.
1563 Description: Don't update lov_desc members until making sure they are valid
1564 Details : When updating lov_desc members via proc fs, need fix their
1565 validities before doing the real update.
1569 Description: don't put request into delay list while invalidate in flight.
1570 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1571 import in flight. this produce timeout for invalidate and sometimes
1572 can cause stale data.
1576 Frequency : on ppc only
1577 Description: not convert ost objects for directory because it's not exist.
1578 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1582 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1584 Severity : enhancement
1586 Description: Update to support 2.6.22.14 vanilla kernel.
1590 Frequency : rare, on recovery
1591 Description: read procfs can produce deadlock in some situation
1592 Details : Holding lprocfs lock which send rpc can produce block for destroy
1593 obd objects and this also block reconnect with -EALREADY. This
1594 isn't fix all lprocfs bugs - but make it rare.
1596 Severity : enhancement
1598 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1601 Frequency : frequent on X2 node
1603 Description: mdc_set_open_replay_data LBUG
1604 Details : Set replay data for requests that are eligible for replay.
1608 Description: lustre_mgs: operation 101 on unconnected MGS
1609 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1610 MGC, and late on MGC cannot successfully connect to MGS and a lot
1611 of the error messages complaining that MGS is not connected.
1615 Frequency : on network error
1616 Description: panic with double free request if network error
1617 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1618 but it's true only for synchronus enqueue, for async enqueue
1619 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1622 Severity : enhancement
1624 Description: client-side metadata stat-ahead during readdir(directory readahead)
1625 Details : perform client-side metadata stat-ahead when the client detects
1626 readdir and sequential stat of dir entries therein
1629 Frequency : on start mds
1631 Description: Implement get_info(last_id) in obdfilter.
1634 Frequency : occasional
1636 Description: Correctly check stale fid, not start epoch if ost not support SOM
1637 Details : open with flag O_CREATE need set old fid in op_fid3 because
1638 op_fid2 was overwritten with new generated fid, but mds can answer
1639 with one of these two fids and both is not stale. Setattr
1640 incorrectly started an epoch and assume will be called
1641 done_writing, but without SOM done_writing ever being called.
1644 Frequency : rare, depends on device drivers and load
1646 Description: MDS or OSS nodes crash due to stack overflow
1647 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1648 In some cases, in conjunction with device drivers that use a lot
1649 of stack the MDS (or possibly OSS) service threads could overflow
1650 the stack. One change which was identified to consume additional
1651 stack has been reworked to avoid the extra stack usage.
1654 Frequency : occasional
1656 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1657 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1658 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1659 oscc_has_objects and see OSC has no objests and call
1660 oscc_internal_create to resend create request.
1662 Severity : enhancement
1664 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1666 Severity : enhancement
1668 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1671 Frequency : very rare
1673 Description: Fix replay if there is an un-replied request and open
1674 Details : In some cases, older replay request will revert the
1675 mcd->mcd_last_xid on MDS which is used to record the client's
1676 latest sent request.
1678 Severity : enhancement
1680 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1682 Severity : enhancement
1684 Description: Add rhel5 support to HEAD.
1686 Serverity : enhancement
1688 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1693 Description: Don't allow skipping OSTs if index has been specified.
1694 Details : Don't allow skipping OSTs if index has been specified, make
1695 locking in internal create lots better.
1699 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1700 Details : access to bitfield in structure is always rounded to long
1701 and this produce problem with not atomic change any bit.
1705 Description: Lustre make rpms failed.
1706 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1707 builds Lustre rpms from tarball.
1710 Frequency : rare on shutdown ost
1712 Description: If llog cancel was not send before clean_exports phase, this can
1713 produce deadlock in llog code.
1714 Details : If llog thread has last reference to obd and call class_import_put
1715 this produce deadlock because llog_cleanup_commit_master wait when
1716 last llog_commit_thread exited, but this never success because was
1717 called from llog_commit_thread.
1721 Description: allow userland application know is lost one of stripes.
1722 Details : fill lvb_blocks with error code on ost and return it to
1723 application if error flag found.
1727 Description: NULL lov_tgts causing MDS oops
1728 Details : more safe checks for NULL lov_tgts for avoid oops.
1730 Severity : enhancement
1732 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1736 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1737 Details : If an OST is down the MDS will hang indefinitely in
1738 obd_statfs_rqset() waiting for the statfs data. While for
1739 MDS QOS usage of statfs, it should not stuck in waiting.
1741 Severity : enhancement
1743 Description: remote_acl support
1744 Details : Support ACL-based permission check for remote user.
1745 Support setfacl/getfacl for remote user with the utils
1746 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1747 the system "{s,g}etfacl" utils.
1749 Severity : enhancement
1751 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1753 Severity : enhancement
1755 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1759 Description: make mgs_setparam() handle fsname containing dash
1760 Details : fsname containing a dash does not work with lctl conf_param
1762 Severity : enhancement
1764 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1766 Severity : enhancement
1768 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1772 Frequency : rare, at shutdown
1773 Description: access already free / zero obd_namespace.
1774 Details : if client_disconnect_export was called without force flag set,
1775 and exist connect request in flight, this can produce access to
1776 NULL pointer (or already free pointer) when connect_interpret
1777 store ocd flags in obd_namespace.
1781 Frequency : only at startup
1782 Description: not alloc memory with spinlock held.
1783 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1784 if any spinlock held.
1786 Severity : enhancement
1788 Description: make lustre randomly fail allocating memory
1789 Details : Make lustre randomly failed allocating memory for testing purpose.
1791 Severity : enhancement
1793 Description: lost problems with lov objid file
1794 Details : Fixes some scability and access to not inited memory problems
1795 in work with lov objdid file.
1800 Description: lfs find does not continue on file error
1801 Details : Continue other files processing when a file/dir is absent.
1805 Description: Inconsistent usage of lustre_pack_reply()
1806 Details : Standardize the usage of lustre_pack_reply() such that it
1807 always generate a CERROR on failure.
1810 Frequency : occasional
1812 Description: MDS hang or stay in waiting lock
1813 Details : If client receive lock with CBPENDING flag ldlm need send lock
1814 cancel as separate rpc, to avoid situation when cancel request
1815 can't processed due all i/o threads stay in wait lock.
1819 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1823 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1827 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1831 Description: Update to RHEL4 latest kernel.
1833 Severity : enhancement
1835 Description: Build SLES10 patchless client fails
1836 Details : The configure was broken by run ./configure with
1837 --with-linux-obj=.... argument for patchless client. When the
1838 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1839 can't search header adequately. Use absolute path such as
1840 -I($LINUX)/include instead.
1844 Description: interrupt oig_wait produce painc on resend.
1845 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1846 private set, and this produce situation when rq_ptlrpcd_data not
1847 copyed to new allocated request and triggered LBUG on assert
1848 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1849 ptlrpcd set if request is changed and can be safety changed to use
1854 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1855 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1856 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1857 these macros and replace them with their difinitions instead.
1861 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1862 Details : Update which_patch & target file for SLES10 latest kernel.
1865 Description: Remove client patches from SLES 10 kernel.
1866 Details : This causes SLES 10 clients to behave as patchless clients
1867 even on a Lustre-patched (server) kernel.
1869 Severity : enhancement
1871 Description: self-adjustable client's lru lists
1872 Details : use adaptive algorithm for managing client cached locks lru
1873 lists according to current server load, other client's work
1874 pattern, memory activities, etc. Both, server and client
1875 side namespaces provide number of proc tunables for controlling
1878 Severity : enhancement
1880 Description: light-weight GSS support
1881 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1882 and reduce performance overhead.
1884 Severity : enhancement
1886 Description: Linux keyring support
1887 Details : Support using service of Linux keyring for Lustre GSS internal
1888 context refresh/cache mechanism.
1892 Description: Fix errors in lfs documentation
1893 Details : Fixes man pages
1897 Description: don't use GFP_* in generic Lustre code.
1898 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1902 Description: obdclass is limited by single OBD_ALLOC(idarray)
1903 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1907 Description: warnings with build patchless client with vanila 2.6.19 and up
1908 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1909 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1913 Frequency : Only for SLES
1914 Description: sanity_quota fail test_1
1915 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1916 file, which makes TSTID[2] inunique.
1920 Frequency : Only for SLES9
1921 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1922 Details : Update target/ChangeLog/which_patch .
1926 Frequency : Only for test_10 in sanity-quota.sh
1927 Description: a bug in quota test code
1928 Details : checking a test flag in a wrong place
1932 Frequency : Only for test_13 in sanity-quota.sh
1933 Description: a bug in quota test script
1934 Details : shouldn't check overfull allocated quota
1938 Description: jbd statistics
1939 Details : Port older jbd statistics patch for sles10
1941 Severity : enhancement
1943 Description: Add printing inode info into message about error in writepage.
1948 Description: reply_lock_interpret crash due to race with it and lock cancel.
1949 Details : Do not replay locks that are being cancelled. Do not reference
1950 locks by their address during replay, just by their handle.
1954 Frequency : When flocks are used.
1955 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1956 Details : Flock locks might destroy just granted lock if it could be merged
1957 with another existing flock, this is done in completion handler,
1958 so teach ldlm_cli_enquque_fini that this is a valid case for
1964 Description: Oops in read and write path when failing to allocate lock.
1965 Details : Check if lock allocation failed and return error back.
1969 Description: lstripe command fails for valid OST index
1970 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1971 instead of lov->desc.ld_active_tgt_count.
1975 Description: sanity.sh failed test 103
1976 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1981 Description: df doesn't work properly if diskfs blocksize != 4K
1982 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1986 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1987 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1989 Severity : enhancement
1991 Description: Adaptive timeouts
1992 Details : RPC timeouts adapt to changing server load and network
1993 conditions to reduce resend attempts and improve recovery time.
1997 Description: llapi_file_create() does not allow some changes
1998 Details : add llapi_file_open() that allows specifying the mode and
1999 open flags, and also returns an open file handle.
2003 Description: merge and cleanup kernel patches.
2004 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2008 Description: Add journal checksum support.(Kernel part)
2009 Details : The journal checksum feature adds two new flags i.e
2010 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2011 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2012 indicates that the commit block contains the checksum for
2013 the blocks described by the descriptor blocks. Now commit
2014 record can be sent to disk without waiting for descriptor
2015 blocks to be written to disk. This behavior is controlled
2016 using JBD2_FEATURE_ASYNC_COMMIT flag.
2020 Description: OSS needs mutliple precreate threads
2021 Details : Add ability to start more than one create thread per OSS.
2025 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2026 Details : execute lfs setstripe on client
2030 Description: mds_obd_create error creating tmp object
2031 Details : When the user sets quota on root, llog will be affected and can't
2032 create files and write files.
2035 Frequency : Always on ia64 patchless client, and possibly others.
2037 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2038 Details : This allows the patchless client to be loaded on architectures
2039 without this export.
2043 Description: RedHat Update kernel for RHEL5
2044 Details : Add the kernel config file for RHEL5.
2048 Description: RedHat Update kernel for RHEL5
2049 Details : Modify the kernel config file more closer RHEL5.
2053 Description: Build failure against Centos5 (RHEL5)
2054 Details : Define PAGE_SIZE when it isn't present.
2058 Description: test_7 on sanity-quota.sh can't be used on separate machines
2059 Details : add facet to handle it
2063 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2064 Details : with NFS, the anon dentry's parent was set to itself in
2065 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2066 corresponding dentry other than use rec->ur_name.
2068 Severity : enhancement
2070 Description: lfs setstripe enhancement
2071 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2075 Description: enable data checksumming by default
2076 Details : enable checksum by default, allow --disable-checksum
2077 configure option and "-o nochecksum" mount option. Checksums
2078 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2082 Description: lustre support for RHEL5
2083 Details : Add support for RHEL5.
2087 Description: Client eviction due to failover config
2088 Details : after a connection loss, the lustre client should attempt to
2089 reconnect to the last active server first before trying the
2090 other potential connections.
2094 Description: when mds and osts use different quota unit(32bit and 64bit),
2095 quota will be released repeatly.
2096 Details : void sending multiple quota reqs to mds, which will keep the
2097 status between the reqs.
2101 Description: osts not allocated evenly to files
2102 Details : change the condition to increase offset_idx
2105 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2107 Description: Data corruption for OSTs that are formatted larger than 2TB
2109 Details : When generating the bio request for lustre file writes the
2110 sector number would overflow a temporary variable before being
2111 used for the IO. The data reads correctly from Lustre (which
2112 will overflow in a similar manner) but other file data or
2113 filesystem metadata may be corrupted in some cases.
2117 Description: Tune the kernel for good SCSI performance.
2118 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2119 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2124 Description: rewrite ext2-derived code in obdclass/uuid.c
2125 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2126 from scratch preserving functionality.
2130 Description: "lfs find -obd UUID" prints directories
2131 Details : "lfs find -obd UUID" will return all directory names instead
2132 of just file names. It is incorrect because the directories
2133 do not reside on the OSTs.
2137 Description: use i_size_read and i_size_write in 2.6 port
2138 Details : replace inode->i_size access with i_size_read/write()
2141 Frequency : when removing large files
2143 Description: scheduling issue during removal of large Lustre files
2144 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2145 It causes scheduling issues when removing large files (17TB in the
2149 Frequency : only with liblustre clients on XT3
2151 Description: evictions taking too long
2152 Details : allow llrd to evict clients directly on OSTs
2155 Frequency : only on ppc
2157 Description: /proc/fs/lustre/devices broken on ppc
2158 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2159 We should make sure the type of 'index' is loff_t and then cast
2160 explicitly as needed below. Do not assign an explicitly cast
2164 Frequency : only for rhel5
2166 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2167 Details : Modify the target file & which_kernel.
2169 Severity : enhancement
2171 Description: omit set fsid for export NFS
2172 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2173 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2177 Frequency : Always, for kernels after 2.6.16
2178 Description: Fix warning idr_remove called for id=.. which is not allocated.
2179 Details : Last kernels save old s_dev before kill super and not allow
2180 to restore from callback - restore it before call kill_anon_super.
2184 Description: buffer overruns could theoretically occur
2185 Details : llapi_semantic_traverse() modifies the "path" argument by
2186 appending values to the end of the origin string, and a
2187 overrun may occur. Adding buffer overrun check in liblustreapi.
2191 Description: Fix error on 'ls .' at the top of the Lustre mount.
2192 Details : Don't revalidate dentry if it is a root dentry.
2196 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2197 Details : Modify vm-tunables-rhel4.patch.
2201 Description: Kernel config for 2.6.18-vanilla.
2202 Details : Modify targets/2.6-vanilla.target.in.
2203 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2204 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2205 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2206 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2210 Description: improve handling recoverable errors
2211 Details : if request processig with error which can be recoverable on server
2212 request should be resend, otherwise page released from cache and
2217 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2218 Details : Modify target file & which_patch.
2219 A flaw was found in the IA32 system call emulation provided
2220 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2221 value could be stored in the %RAX register, which could trigger an
2222 out-of-bounds system call table access. An untrusted local user
2223 could exploit this flaw to run code in the kernel
2224 (ie a root privilege escalation). (CVE-2007-4573).
2228 Description: O_DIRECT bypasses client statistics.
2229 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2230 read_ahead_stats were updated. copied the stats section from
2231 osc_send_oap_rpc() into async_internal().
2235 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2239 Description: change order of libsysio includes
2240 Details : '#include sysio.h' should always come before '#include xtio.h'
2244 Description: handle "serious error: objid * already exists" more gracefully
2245 Details : If LAST_ID value on disk is smaller than the objects existing in
2246 the O/0/d* directories, it indicates disk corruption and causes an
2247 LBUG(). If the object is 0-length, then we should use the existing
2248 object. This will help to avoid a full fsck in most cases.
2250 Severity : enhancement
2252 Description: adapt the lustre_config script to support the upgrade case
2253 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2254 server targets to 1.6 in parallel.
2258 Description: To avoid grant space > avaible space when the disk is almost
2259 full. Without this patch you might see the error "grant XXXX >
2260 available" or some LBUG about grant, when the disk is almost
2262 Details : In filter_check_grant, for non_grant cache write, we should
2263 check the left space by if (*left > ungranted + bytes), instead
2264 of (*left > ungranted), because only we are sure the left space
2265 is enough for another "bytes", then the ungrant space should be
2266 increase. In client, we should update cl_avail_grant only there
2267 is OBD_MD_FLGRANT in the reply.
2271 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2272 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2273 described in CVE-2007-4573. This problem could allow untrusted
2274 local users to gain root access.
2277 Frequency : when using O_DIRECT and quotas
2279 Description: Incorrect file ownership on O_DIRECT output files
2280 Details : block usage reported by 'lfs quota' does not take into account
2281 files that have been written with O_DIRECT.
2286 Description: touch file failed when fs is not full
2287 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2288 otherwise we can get ENOSP while fs is not full.
2292 Description: parallel lock callbacks
2293 Details : Instead of sending blocking and completion callbacks as separated
2294 requests, adding them to a set and sending in parallel.
2297 Frequency : only for Cray XT3
2298 Bugzilla : 12829/13455
2299 Description: Changing primary group doesn't change the group lustre assigns to
2301 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2302 overridden with the primary group provided by the group upcall,
2303 whereas the supplied fsgid can be trusted if it is in the list of
2304 supplementary groups returned by the group upcall.
2306 Severity : enhancement
2308 Description: Allow masking D_WARNING, D_ERROR messages from console
2309 Details : Console messages can now be disabled via lnet.printk.
2313 Description: User code with malformed file open parameter crashes client node
2314 Details : Before packing join_file req, all the related reference should be
2315 checked carefully in case some malformed flags cause fake
2316 join_file req on client.
2320 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2321 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2322 mode is equal to the granted mode. However, it should also check
2323 that LDLM_FL_CP_REQD is not set.
2327 Description: Heavy nfs access might result in deadlocks
2328 Details : After ELC code landed, it is now improper to enqueue any mds
2329 locks under och_sem, because enqueue might want to decide to
2330 cancel open locks for same inode we are holding och_sem for.
2334 Description: Client eviction while running blogbench
2335 Details : A lot of unlink operations with concurrent I/O can lead to a
2336 deadlock causing evictions. To address the problem, the number of
2337 oustanding OST_DESTROY requests is now throttled to
2338 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2339 callbacks are processed in priority.
2343 Description: enable ACLs on MDS by default
2344 Details : ACLs must be enabled on MDS by default.
2347 Frequency : PPC/PPC64 only
2349 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2350 Details : fix duplicated definitions between asm-ppc64/types.h and
2351 lustre_types.h on PPC.
2354 Frequency : PPC/PPC64 only
2356 Description: asm-ppc/segment.h does not exist
2357 Details : fix compile issue on PPC.
2361 Description: better handle error messages in extents code
2364 Frequency : RHEL4 only
2366 Description: mkfs is very slow on IA64/RHEL4
2367 Details : A performance regression has been discovered in the MPT Fusion
2368 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2369 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2370 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2372 Severity : enhancement
2374 Description: SNMP support enhancement
2375 Details : Adding total number of sampled request for an MDS node in snmp
2378 Severity : enhancement
2380 Description: Optimize ldlm waiting list processing for PR extent locks
2381 Details : When processing waiting list for read extent lock and meeting read
2382 lock that is same or wider to it that is not contended, skip
2383 processing rest of the list and immediatelly return current
2384 status of conflictness, since we are guaranteed there are no
2385 conflicting locks in the rest of the list.
2389 Description: Time out and refuse to reconnect
2390 Details : When the failover node is the primary node, it is possible
2391 to have two identical connections in imp_conn_list. We must
2392 compare not conn's pointers but NIDs, otherwise we can defeat
2393 connection throttling.
2397 Description: port llog fixes from b1_6 into HEAD
2398 Details : Port llog reference couting and some llog cleanups from b1_6
2399 (bug 10800) into HEAD, for protect from panic and access to already
2400 free llog structures.
2404 Description: Detect stride IO mode in read-ahead
2405 Details : When a client does stride read, read-ahead should detect that and
2406 read-ahead pages according to the detected stride pattern.
2410 Description: data checksumming impacts single node performance
2411 Details : add support for several checksum algorithm. Currently, only CRC32
2412 and Adler-32 are supported. The checksum type can be changed on
2413 the fly via /proc/fs/lustre/osc/*/checksum_type.
2417 Description: use adler32 for page checksums
2418 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2423 Description: build for x2 fails
2424 Details : fix compile issue on Cray systems.
2428 Description: Properly match for duplicate locks
2429 Details : Due to different lock order from skiplists code, we need to
2430 traverse entire list for now
2433 Frequency : only on PPC/SLES10
2435 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2436 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2437 be defined. Add a hack in mkfs_lustre.c to work around this compile
2442 Description: LASSERT on MDS when client holding flock lock dies
2443 Details : ldlm pool logic depends on number of granted locks equal to
2444 number of released locks which is not true for flock locks, so
2445 just exclude such locks from consideration.
2449 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2450 Details : Use fsfilt_sync() for both the whole filesystem sync and
2451 individual file sync to eliminate dangerous inode locking
2452 with I_LOCK that can lead to a deadlock.
2456 Description: performance in 1.6.3
2457 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2458 sure that 1MB requests are not fragmented by the block layer.
2460 Severity : enhancement
2462 Description: organize the server-side client stats on per-nid basis
2463 Details : Change the structure of stats under obdfilter and mds to
2471 The "uuid"s file would list the uuids of _active_ exports.
2472 And the clear entry is to clear all stats and stale nids.
2474 Severity : enhancement
2476 Description: eliminate client locks in face of contention
2477 Details : file contention detection and lockless i/o implementation
2478 for contended files.
2482 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2485 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2487 Description: MDS may lose file striping (and hence file data) in some cases
2488 Details : If there are additional extended attributes stored on the MDS,
2489 in particular ACLs, SELinux, or user attributes (if user_xattr
2490 is specified for the client mount options) then there is a risk
2491 of attribute loss. Additionally, the Lustre file striping
2492 needs to be larger than default (e.g. striped over all OSTs),
2493 and an additional attribute must be stored initially in the
2494 inode and then increase in size enough to be moved to the
2495 external attribute block (e.g. ACL growing in size) for file
2500 Description: skiplist implementation simplification
2501 Details : skiplists are used to group compatible locks on granted list
2502 that was implemented as tracking first and last lock of each
2503 lock group the patch changes that to using doubly linked lists
2507 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2508 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2509 because LASSERT on both the data supplied by a client, and the
2510 data on disk is dangerous and incorrect.
2512 Severity : enhancement
2514 Description: Slow truncate/writes to huge files at high offsets.
2515 Details : Directly associate cached pages to lock that protect those pages,
2516 this allows us to quickly find what pages to write and remove
2517 once lock callback is received.
2521 Description: more ldlm soft lockups
2522 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2523 starve other threads from the resource lock for a long time in
2524 case of long waiting queue, so change the debug level from
2525 D_OTHER to the less frequently used D_INFO.
2527 Severity : enhancement
2529 Description: add -gid, -group, -uid, -user options to lfs find
2533 Description: Hung threads in invalidate_inode_pages2_range
2534 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2535 once one is completed. As a result, some RPCs are stuck in the
2536 queue and are never sent.
2540 Description: filter threads hungs on waiting journal commit
2541 Details : Cleanup filter group llog code, then only filter group llog will
2542 be only created in the MDS/OST syncing process.
2546 Description: Procfs and llog threads access destoryed import sometimes.
2547 Details : Sync the import destoryed process with procfs and llog threads by
2548 the import refcount and semaphore.
2550 Severity : enhancement
2552 Description: openlock cache of b1_6 port to HEAD
2557 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2558 Details : If insertion of an extent fails, then discard the inode
2559 preallocation and free data blocks else it can lead to duplicate
2564 Description: don't always update ctime in ext3_xattr_set_handle()
2565 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2566 In some cases the ctime should not be updated, for example for
2567 2.0->1.8 compatibility it is necessary to delete an xattr and it
2568 should not update the ctime.
2572 Bugzilla : 15713/16362
2573 Description: Assertion in iopen_connect_dentry in 1.6.3
2574 Details : looking up an inode via iopen with the wrong generation number can
2575 populate the dcache with a disconneced dentry while the inode
2576 number is in the process of being reallocated. This causes an
2577 assertion failure in iopen since the inode's dentry list contains
2578 both a connected and disconnected dentry.
2582 Description: assertion failure in ldlm_handle2lock()
2583 Details : fix a race between class_handle_unhash() and class_handle2object()
2584 introduced in lustre 1.6.5 by bug 13622.
2589 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2590 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2595 Description: oops in sd_iostats_seq_show()
2596 Details : unloading/reloading the scsi low level driver triggers a kernel
2597 bug when trying to access the sd iostat file.
2602 Description: Kernel panics during QLogic driver reload
2603 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2604 patch, causing memory corruption.
2609 Description: journal_dev option does not work in b1_6
2610 Details : pass mount option during pre-mount.
2612 Severity : enhancement
2614 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2615 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2616 extent information of a file. It can be used to map logical blocks
2617 in a file to physical blocks in the block device.
2621 Description: LDLM soft lockups - improvement
2622 Details : It is be possible to send the lock handle along with each read
2623 or write request because the client is already doing a lock match
2624 itself so there isn't any reason the OST should have to re-do that
2628 Frequency : only on Cray X2
2630 Description: X2 build failures
2631 Details : fix build failures on Cray X2.
2635 Description: xid & resent requests
2636 Details : Initialize RPC XID from clock at startup (randomly if clock is
2639 Severity : enhancement
2641 Description: Add lustre_start utility to start or stop multiple Lustre servers
2646 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2647 Details : In case of memory pressure, list_del() can be called twice on
2648 req->rq_history_list, causing a kernel oops.
2652 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2653 Details : fix stack overflow in the distributed lock manager by defering
2654 export eviction after a failed AST to the elt thread instead of
2655 handling it in the dlm interpret routine.
2659 Description: Convert some comments to new format.
2660 Details : Update documenting comments to match doxygen conventions.
2664 Description: Grammar fixes.
2665 Details : A couple of trivial spelling fixes.
2669 Description: OSD_COUNTERS-mandatory
2670 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2671 to simplify the code.
2675 Description: simplify cmm_device freeing logic.
2676 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2680 Description: Add lockdep support to dt_object_operations locking interface.
2681 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2682 parameter indicating lock ordering. Update mdd code to use new
2687 Description: Introduce failloc constants for lockless IO tests.
2688 Details : Add two new failloc constants to test lockless IO. Only one of
2689 them in implemented---another is checked in yet to be landed
2694 Description: Add lockdep support for inode mutex.
2695 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2696 in the situations where Lustre has to lock more than one inode
2701 Description: Add optional invariants checking support.
2702 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2703 configure switch. This macro is to be used for consistency and
2704 sanity checks that are too expensive to be left in `production'
2709 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2710 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2715 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2716 Details : Use spin_lock_nested() in (the only) situation where more than
2717 one ldlm_lock is locked simultaneously. Also, fix possible
2718 dead-lock in ldlm_lock_change_resource() by enforcing particular
2723 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2724 Details : Instead of specifying each ldlm_lock call-back through separate
2725 parameter, wrap them into struct ldlm_callback_suite.
2729 Description: Kill join_lru obd method and its callers.
2730 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2731 in memory---a requirement implemented through ->o_join_lru() obd
2732 method in HEAD. Remove this method and its users.
2736 Description: Add asynchronous ldlm ENQUEUE completion handler.
2737 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2738 case that stock ldlm_completion_ast() cannot handle as it waits
2739 until lock is granted. Introduce new ldlm_completion_ast_async()
2740 for this. Also comment ldlm_completion_ast().
2744 Description: ldlm_error <-> errno conversion.
2745 Details : Add functions to map (rather arbitrary) between LDLM error codes
2746 and standard errno values. CLIO needs this to prevent LDLM specific
2747 constants from escaping ldlm and osc.
2751 Description: Kill unused ldlm_handle2lock_ns() function.
2752 Details : Kill unused ldlm_handle2lock_ns() function.
2756 Description: Add lu_ref support to ldlm_lock
2757 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2758 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2759 and ldlm_resource. LDLM interface has to be changed, because code that
2760 releases a reference on a lock, has to "know" what reference this is.
2761 In the most frequent case
2763 lock = ldlm_handle2lock(handle);
2765 LDLM_LOCK_PUT(lock);
2767 no changes are required. When any other reference (received _not_ from
2768 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2769 instead of LDLM_LOCK_PUT().
2771 Arguably, changes are pervasive, and interface requires some discipline
2772 for proper use. On the other hand, it was very instrumental in finding
2773 a few leaked lock references.
2777 Description: Add ldlm_lock_addref_try().
2778 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2779 attempts to addref a lock that might be being canceled
2784 Description: Add ldlm_weigh_callback().
2785 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2786 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2787 hard-coded `number of pages' logic.
2791 Description: Add lockdep annotations to llog code.
2792 Details : Use appropriately tagged _nested() locking calls in the places
2793 where llog takes more than one ->lgh_lock lock.
2797 Description: Add loi_kms_set().
2798 Details : Wrap kms updates into a helper function.
2802 Description: Constify instances of struct lsm_operations.
2803 Details : Constify instances of struct lsm_operations.
2807 Description: lu_conf support.
2808 Details : On a server, a file system object is uniquely identified
2809 by a fid, which is sufficient to locate and load all object
2810 state (inode). On a client, on the other hand, more data are
2811 necessary instantiate an object. Change lu_object_find() and
2812 friends to take additional `lu_conf' argument describing object.
2813 Typically this includes layout information.
2817 Description: lu_context fixes.
2818 Details : Introduce new lu_context functions that are needed on the client
2819 side, where some system threads (ptlrpcd) are shared by multiple
2820 modules, and so cannot be stopped during module shutdown.
2824 Description: Add start and stop methods to lu_device_type_operations.
2825 Details : Introduce two new methods in lu_device_type_operations, that are
2826 invoked when first instance of a given type is created and last one
2827 is destroyed respectively. This is need by CLIO.
2831 Description: Add lu_ref support to struct lu_device.
2832 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2833 track leaked references.
2837 Description: Introduce lu_kmem_descr.
2838 Details : lu_kmem_descr and its companion interface allow to create
2839 and destroy a number of kmem caches at once.
2843 Description: Fix lu_object finalization race.
2844 Details : Fix a race between lu_object_find() finding an object and its
2845 concurrent finalization. This race is (most likely) not possible
2846 on the server, but might happen on the client.
2850 Description: Introduce lu_ref interface.
2851 Details : lu_ref is a debugging module allowing to track references to
2852 a given object. It is quite cpu expensive, and has to be
2853 explicitly enabled with --enable-lu_ref. See usage description
2858 Description: Factor lu_site procfs stats into a separate function.
2859 Details : Separate lu_site stats printing code into a separate function
2860 that can be reused on a client.
2864 Description: Constify instances of struct {lu,dt,md}_device_operations.
2865 Details : Constify instances of struct {lu,dt,md}_device_operations.
2869 Description: Introduce struct md_site and move meta-data specific parts of
2870 struct lu_site here.
2871 Details : Move md-specific fields out of struct lu_site into special struct
2872 md_site, so that lu_site can be used on a client.
2876 Description: Kill mdd_lov_destroy().
2877 Details : Remove unused mdd code.
2881 Description: Add st_block checking to multistat.c.
2882 Details : Add st_block checking to multistat.c.
2886 Description: Add lu_ref support to struct obd_device.
2887 Details : Add lu_ref tracking to obd_device.
2891 Description: Kill obd_set_fail_loc().
2892 Details : Remove unused code.
2896 Description: Add special type for ptlrpc_request interpret functions.
2897 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2898 there. Actual usage will be in CLIO.
2902 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2903 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2904 doesn't work with lockdep.
2908 Description: Add rwv.c test program.
2909 Details : New testing program exercising readv(2) and writev(2) (Qian).
2913 Description: Add sendfile.c test program.
2914 Details : New testing program exercising sendfile(2) (Jay).
2918 Description: Ratelimit a message that can be very frequent.
2919 Details : Ratelimit a memory allocation failure message that can
2924 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2925 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2929 Description: Liblustre build fixes.
2930 Details : Liblustre build fixes.
2934 Description: libcfs: add cfs_{need,cond}_resched() interface.
2935 Details : libcfs: add cfs_{need,cond}_resched() definition and
2936 implementations for Linux, NT, and liblustre.
2938 Severity : enhancement
2940 Description: More exported tunables for mballoc
2941 Details : Add support for tunable preallocation window and new tunables for
2942 large/small requests
2946 Description: Detect corruption of block bitmap and checking for preallocations
2947 Details : Checks validity of on-disk block bitmap. Also it does better
2948 checking of number of applied preallocations. When corruption is
2949 found, it turns filesystem readonly to prevent further corruptions.
2953 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2954 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2955 window to 0. If it does hit cache miss, and read-pattern is still
2956 stride-io mode, does not reset the stride window, but also does not
2957 increase the stride window length in this case.
2961 Frequency : only for big-endian servers
2962 Description: Check if system is big-endian while mounting fs with extents feature
2963 Details : Mounting a filesystem with extents feature will fail on big-endian
2964 systems since ext3-based ldiskfs is not supported on big-endian
2965 systems. This can be over-riden with "bigendian_extents" mount option.
2967 Severity : enhancement
2969 Description: The root squash functionality
2970 Details : A security feature, which is to prevent users from being able
2971 to mount lustre on their desktop, run as root, and delete
2972 all of the files in the filesystem. The goal is accomplished by
2973 remapping user id (UID) and group id (GID) of the root user to
2974 a UID and GID specified by the system administartor via Lustre
2975 configuration management server (MGS). The functionality also
2976 allows to specify sets of clients for which the remapping does
2981 Description: Excessive recovery window
2982 Details : With AT enabled, the recovery window can be excessively long (6000+
2983 seconds). To address this problem, we no longer use
2984 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2985 timeout no longer depends on the service time, it is set to
2986 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2987 time via pb_service_time.
2991 Description: Watchdog triggered on MDS failover
2992 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2993 the OSTs know that the MDS "UUID" can be reused for the same export
2994 from a different NID, so we do not need to wait for the export to be
2998 Frequency : rare, only if using MMP with Linux RAID
3000 Description: MMP doesn't work with Linux RAID
3001 Details : While using HA for Lustre servers with Linux RAID, it is possible
3002 that MMP will not detect multiple mounts. To make this work we
3003 need to unplug the device queue in RAID when the MMP block is being
3004 written. Also while reading the MMP block, we should read it from
3005 disk and not the cached one.
3007 Severity : enhancement
3009 Description: open file using fid
3010 Details : A file can be opened using just its fid, like
3011 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3014 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3017 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3018 Details : add proper locking for creation of jbd_4k slab cache
3022 Description: MMP check in ext3_remount() fails without displaying any error
3023 Details : When multiple mount protection fails during remount, proper error
3026 Severity : enhancement
3028 Description: Allow stripe size to be up to 4G-64k
3029 Details : Fix math logic to allow large stripe sizes.
3033 Description: add check for >8TB ldiskfs filesystems
3034 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3035 Don't allow >8TB ldiskfs filesystems to be mounted without
3036 force_over_8tb mount option
3041 Descriptoin: OSS complains "too many blocks" and crashes
3042 Details : In filter_commitrw_write():
3043 1. lquota_chkquota();
3044 2. obdo_from_inode();
3045 3. lquota_pending_commit();
3046 "2." has changed uid and gid in this case, then caused
3047 lquota_chkquota() and lquota_pending_commit() to use
3048 different uid and gid. That is the root of the bug.
3052 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3053 Details : Detect and print out the 'stripe count = -1' case properly.
3057 Description: Out or order replies might be lost on replay
3058 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3059 request with tid smaller than one currently being added, add it
3060 to the start, not end of the list.
3062 --------------------------------------------------------------------------------
3064 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3066 * Support for kernels:
3067 2.6.5-7.283 (SLES 9),
3068 2.6.9-55.EL (RHEL 4),
3069 2.6.16.46-0.14 (SLES 10),
3070 2.6.18.8 vanilla (kernel.org)
3071 * Client support for unpatched kernels:
3072 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3073 2.6.16 - 2.6.22 vanilla (kernel.org)
3074 * Due to recently discovered recovery problems, we do not recommend
3075 using patchless RHEL 4 clients with this or any earlier release.
3076 * Recommended e2fsprogs version: 1.39.cfs8
3077 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3078 * Starting with this release, the ldiskfs backing filesystem required
3079 by Lustre is now in its own package, lustre-ldiskfs. This package
3080 should be installed. It is versioned separately from Lustre and
3081 may be released separately in future.
3085 Description: block reactivating mgc import until all deactivates complete
3086 Details : Fix race when failing back MDT/MGS to itself (testing)
3089 Frequency : at statup only
3091 Description: mds_lov_synchronize race leads to various problems
3092 Details : simultaneous MDT->OST connections at startup can cause the
3093 sync to abort, leaving the OSC in a bad state.
3095 Severity : enhancement
3097 Description: add optional extra BUILD_VERSION info
3098 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3099 to override the lustre version.
3102 Frequency : 2.6.18 servers only
3104 Description: ll_kern_mount() doesn't release the module reference
3105 Details : The ldiskfs module reference count never drops down to 0
3106 because ll_kern_mount() doesn't release the module reference.
3111 Description: server LBUG when using old ost_num_threads parameter
3112 Details : Accept the old ost_num_threads parameter but warn that it
3113 is deprecated, and fix an off-by-one error that caused an LBUG.
3118 Description: Transient SCSI error results in persistent IO issue
3119 Details : iobuf->dr_error is not reinitialized to 0 between two
3123 Frequency : sometimes when underlying device returns I/O errors
3125 Description: OSTs not going read-only during write failures
3126 Details : OSTs are not remounted read-only when the journal commit threads
3127 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3128 instead of the ext3 wrappers.
3132 Description: poor connect scaling with increasing client count
3133 Details : Don't run filter_grant_sanity_check for more than 100 exports
3134 to improve scaling for large numbers of clients.
3137 Frequency : SLES10 only
3139 Description: sanity-quota.sh quotacheck failed: rc = -22
3140 Details : Quotas cannot be enabled on SLES10.
3143 Frequency : liblustre clients only
3145 Description: getdirentries does not give error when run on compute nodes
3146 Details : getdirentries does not fail when the size specified as an argument
3147 is too small to contain at least one entry
3149 Severity : enhancement
3151 Description: Add LNET router traceability for debug purposes
3152 Details : If a checksum failure occurs with a router as part of the
3153 IO path, the NID of the last router that forwarded the bulk data
3154 is printed so it can be identified.
3159 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3160 Details : Due to a race condition, liblustre clients were occasionally
3161 evicted incorrectly.
3163 Severity : enhancement
3165 Description: lfs setstripe use optional parameters instead of postional
3168 Severity : enhancement
3170 Description: Nanosecond timestamp support for ldiskfs
3171 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3172 resolution timestamps. There is not yet support for this at
3173 the Lustre filesystem level.
3176 Frequency : during server recovery
3178 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3179 Details : request with rq_no_resend flag not awake l_wait_event if they get
3183 Frequency : nfs export on patchless client
3185 Description: connectathon hang when test nfs export over patchless client
3186 Details : Disconnected dentry cannot be found with lookup, so we do not need
3187 to unhash it or make it invalid
3190 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3191 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3192 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3197 Description: open req refcounting wrong on reconnect
3198 Details : If reconnect happened between getting open reply from server and
3199 call to mdc_set_replay_data in ll_file_open, we will schedule
3200 replay for unreferenced request that we are about to free.
3201 Subsequent close will crash in variety of ways.
3202 Check that request is still eligible for replay in
3203 mdc_set_replay_data().
3208 Description: disable writes to filesystem when reading health_check file
3209 Details : the default for reading the health_check proc file has changed
3210 to NOT do a journal transaction and write to disk, because this
3211 can cause reads of the /proc file to hang and block HA state
3212 checking on a healthy but otherwise heavily loaded system. It
3213 is possible to return to the previous behaviour during configure
3214 with --enable-health-write.
3216 Severity : enhancement
3218 Description: 64-bit inode version
3219 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3220 to the inode. This will be required for version-based recovery.
3225 Description: MDS fails to start if a duplicate client export is detected
3226 Details : in some rare cases it was possible for a client to connect to
3227 an MDS multiple times. Upon recovery the MDS would detect this
3228 and fail during startup. Handle this more gracefully.
3230 Severity : enhancement
3232 Description: Add -o localflock option to simulate old noflock
3234 Details : This will achieve local-only flock/fcntl locks
3240 Description: log_commit_thread vs filter_destroy race leads to crash
3241 Details : Take import reference before releasing llog record semaphore
3246 Description: Wrong request locking in request set processing
3247 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3248 imp_delayed_list, in this place should be used imp_lock.
3251 Frequency : when reconnection
3253 Description: Grant Leak when osc reconnect to OST
3254 Details : When osc reconnect ost, OST(filter) should check whether it
3255 should grant more space to client by comparing fed_grant and
3256 cl_avail_grant, and return the granted space to client instead
3257 of "new granted" space, because client will call osc_init_grant
3258 to update the client grant space info.
3261 Frequency : when client reconnect to OST
3263 Description: Grant Leak when osc do resend and replay bulk write
3264 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3265 bulk write request, because the grant info while be sync between
3266 OSC and OST when reconnect, and we should ignore the grant info
3267 these of resend/replay write req.
3272 Description: Grant space more than available left space sometimes.
3273 Details : When then OST is about to be full, if two bulk writing from
3274 different clients came to OST. Accord the avaliable space of the
3275 OST, the first req should be permitted, and the second one
3276 should be denied by ENOSPC. But if the seconde arrived before
3277 the first one is committed. The OST might wrongly permit second
3278 writing, which will cause grant space > available space.
3281 Frequency : when client is evicted
3283 Description: Grant might be wrongly erased when osc is evicted by OST
3284 Details : when the import is evicted by server, it will fork another
3285 thread ptlrpc_invalidate_import_thread to invalidate the
3286 import, where the grant will be set to 0. While the original
3287 thread will update the grant it got when connecting. So if
3288 the former happened latter, the grant will be wrongly errased
3289 because of this race.
3294 Description: Checking Stale with correct fid
3295 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3296 to check whether it is stale, because sometimes, we want the
3297 enqueue happened anyway, and op_data.fid2 will not be initialized.
3299 Severity : enhancement
3301 Description: update patchless client
3302 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3305 Frequency : only with 2.4 kernel
3307 Description: random memory corruption
3308 Details : size of struct ll_inode_info is to big for union inode.u and this
3309 can be cause of random memory corruption.
3314 Description: Memory leak in recovery
3315 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3316 It should also check obd_fail before fsfilt_start, otherwise if
3317 fsfilt_start return -EROFS,(failover mds during mds recovery).
3318 then the req will return with repmsg->transno = 0 and rc = EROFS.
3319 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3320 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3321 be freed no matter whether fsfilt_commit success or not.
3324 Frequency : only with huge count clients
3326 Description: Prevents from taking the superblock lock in llap_from_page for
3328 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3329 for llap_from_page call in ll_removepage prevents from taking the
3330 superblock lock for a soon died page.
3335 Description: Not check open intent error before release open handle
3336 Details : in some rare cases, the open intent error is not checked before
3337 release open handle, which may cause
3338 ASSERTION(open_req->rq_transno != 0), because it tries to release
3339 the failed open handle.
3344 Description: Set cat log bitmap only after create log success.
3345 Details : in some rare cases, the cat log bitmap is set too early. and it
3346 should be set only after create log success.
3350 Description: Accessing a block bevice can re-enable I/O when Lustre is
3351 tearing down a device.
3352 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3357 Description: service threads may hog cpus when there are a lot of requests
3358 Details : Insert cond_resched to give other threads a chance to use some CPU
3363 Description: the cat log was not initialized in recovery
3364 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3365 unlink log on mds will not be initialized until mds post
3366 recovery. And also in mds post recovery, the unlink log will
3367 initialization will be done asynchronausly, so there will be race
3368 between add unlink log and unlink log initialization.
3372 Description: brw_stats were being printed incorrectly
3373 Details : brw_stats were being printed as log2 but all of them were not
3374 recorded as log2. Also remove some code duplication arising from
3375 filter_tally_{read,write}.
3379 Frequency : rare, only in recovery.
3380 Description: ASSERTION(req->rq_type != LI_POISON) failed
3381 Details : imp_lock should be held while iterating over imp_sending_list for
3382 prevent destroy request after get timeout in ptlrpc_queue_wait.
3386 Description: replay-single.sh test 52 fails
3387 Details : A lock's skiplist need to be cleanup when it being unlinked
3388 from its resource list.
3392 Description: Short directio read returns full requested size rather than
3394 Details : Direct I/O operations should return actual amount of bytes
3395 transferred rather than requested size.
3397 Severity : enhancement
3399 Description: metadata RPC reduction (e.g. for rm performance)
3400 Details : decrease the amount of synchronous RPC between clients and servers
3401 by canceling conflicing lock before the operation on the client
3402 and packing thier handles into the main operation RPC to server.
3404 Severity : enhancement
3406 Description: Async OSC create to avoid the blocking unnecessarily.
3407 Details : If an OST has no remaining object, system will block on the
3408 creation when it needs to create a new object on this OST. Now,
3409 ways use pre-created objects when available, instead of blocking on
3410 an empty osc while others are not empty. If we must block, we
3411 block for the shortest possible period of time.
3415 Description: improve handling recoverable errors
3416 Details : if request processig with error which can be recoverable on server
3417 request should be resend, otherwise page released from cache and
3420 Severity : enhancement
3422 Description: refine locking for avoid write wrong info into lov_objid file
3423 Details : fix possible races with add new target and write/update data in
3426 --------------------------------------------------------------------------------
3428 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3433 Frequency : on some architectures
3435 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3436 Details : Uninitialized flags sometimes cause configuration commands to
3440 Frequency : patchless clients only
3442 Description: missing __iget() symbol export
3443 Details : The __iget() symbol export is missing. To avoid the need for
3444 this on patchless clients the deathrow inode reaper is turned
3445 off, and we depend on the VM to clean up old inodes. This
3446 dependency was during via the fix for bug 12181.
3448 --------------------------------------------------------------------------------
3450 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3452 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3453 INTEROPERATE with older versions automatically. In many cases a
3454 special upgrade step is needed. Please read the
3455 user documentation before upgrading any part of a live system.
3456 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3457 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3458 * WARNING: Lustre configuration and startup changes are required with
3459 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3463 Severity : enhancement
3465 Description: MountConf
3466 Details : Lustre configuration is now managed via mkfs and mount
3467 commands instead of lmc and lconf. New obd types (MGS, MGC)
3468 are added for dynamic configuration management. See
3469 https://mail.clusterfs.com/wikis/lustre/MountConf for
3472 Severity : enhancement
3474 Description: dynamic OST addition
3475 Details : OSTs can now be added to a live filesystem
3477 Severity : enhancement
3479 Description: startup order invariance
3480 Details : MDTs and OSTs can be started in any order. Clients only
3481 require the MDT to complete startup.
3483 Severity : enhancement
3485 Description: parallel, asynchronous orphan cleanup
3486 Details : orphan cleanup is now performed in separate threads for each
3487 OST, allowing parallel non-blocking operation.
3489 Severity : enhancement
3491 Description: optimized stripe assignment
3492 Details : stripe assignments are now made based on ost space available,
3493 ost previous usage, and OSS previous usage, in order to try
3494 to optimize storage space and networking resources.
3496 Severity : enhancement
3498 Description: Permanently set tunables
3499 Details : All writable /proc/fs/lustre tunables can now be permanently
3500 set on a per-server basis, at mkfs time or on a live system.
3502 Severity : enhancement
3504 Description: Lustre message v2
3505 Details : Add lustre message format v2.
3507 Severity : enhancement
3509 Description: client OST exclusion list
3510 Details : Clients can be started with a list of OSTs that should be
3511 declared "inactive" for known non-responsive OSTs.
3515 Description: SPEC SFS validation failure on NFS v2 over lustre.
3516 Details : Changes the blocksize for regular files to be 2x RPC size,
3517 and not depend on stripe size.
3519 Severity : enhancement
3521 Description: Multiple MD RPCs in flight.
3522 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3523 To avoid overly-overloading MDS, introduce a limit on number of
3524 MDS RPCs in flight for a single client and add /proc controls
3525 to adjust this limit.
3527 Severity : enhancement
3529 Description: client read/write statistics
3530 Details : Add client read/write call usage stats for performance
3531 analysis of user processes.
3532 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3533 file access. extents_stats shows chunk size distribution.
3534 extents_stats_per_process show chunk size distribution per
3537 Severity : enhancement
3539 Description: mds statistics
3540 Details : Add detailed mds operations statistics in
3541 /proc/fs/lustre/mds/*/stats.
3545 Description: Failure of copying files with lustre special EAs.
3546 Details : Client side always return success for setxattr call for lustre
3547 special xattr (currently only "trusted.lov").
3552 Description: Refcount LNET uuids
3553 Details : The global LNET uuid list grew linearly with every startup;
3554 refcount repeated list entries instead of always adding to
3557 Severity : enhancement
3559 Description: Dynamic service threads
3560 Details : Within a small range, start extra service threads
3561 automatically when the request queue builds up.
3564 Frequency : mixed-endian client/server environments
3566 Description: mixed-endian crashes
3567 Details : The new msg_v2 system had some failures in mixed-endian
3570 Severity : enhancement
3572 Description: Easy OST removal
3573 Details : OSTs can be permanently deactivated with e.g. 'lctl
3574 conf_param lustre-OST0001.osc.active=0'
3576 Severity : enhancement
3578 Description: MGS proc entries
3579 Details : Added basic proc entries for the MGS showing what filesystems
3582 Severity : enhancement
3584 Description: provide MGS failover
3585 Details : Added config lock reacquisition after MGS server failover.
3587 Severity : enhancement
3589 Description: add Linux 2.4 support
3590 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3594 Description: a large application tries to do I/O to the same resource and dies
3595 in the middle of it.
3596 Details : Check the req->rq_arrival time after the call to
3597 ost_brw_lock_get(), but before we do anything about
3598 processing it & sending the BULK transfer request. This
3599 should help move old stale pending locks off the queue as
3600 quickly as obd_timeout.
3603 Frequency : when an incorrect nid is specified during startup
3605 Description: ptlrpc connect to non-existant node causes kernel crash
3606 Details : LNET can't be re-entered from an event callback, which
3607 happened when we expire a message after the export has been
3608 cleaned up. Instead, hand the zombie cleanup off to another
3611 Severity : enhancement
3613 Description: plain/inodebits lock performance improvement
3614 Details : Grouping plain/inodebits in granted list by their request modes
3615 and bits policy, thus improving the performance of search through
3619 Frequency : only if OST filesystem is corrupted
3621 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3622 Details : for a short time RPCs with bulk IO are in the replay list,
3623 but replay of bulk IOs is unimplemented. If the OST filesystem
3624 is corrupted due to disk cache incoherency and then replay is
3625 started it is possible to trip an assertion. Avoid putting
3626 committed RPCs into the replay list at all to avoid this issue.
3629 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3631 Description: System hang on startup
3632 Details : This bug allowed the liblustre (e.g. catamount) client to
3633 return to the app before handling all startup RPCs. This
3634 could leave the node unresponsive to lustre network traffic
3635 and manifested as a server ptllnd timeout.
3637 Severity : enhancement
3639 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3640 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3641 boolean to control whether to print peer debug info when a
3642 client's RPC times out.
3645 Frequency : only for kernels with patches from Lustre below 1.4.3
3647 Description: Remove old rdonly API
3648 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3651 Frequency : only for devices with external journals
3653 Description: Set external device read-only also
3654 Details : During a commanded failover stop, we set the disk device
3655 read-only while the server shuts down. We now also set any
3656 external journal device read-only at the same time.
3659 Frequency : when upgrading from 1.4 while trying to change parameters
3661 Description: The wrong (new) MDC name was used when setting parameters for
3662 upgraded MDT's. Also allows changing of OSC (and MDC)
3663 parameters if --writeconf is specified at tunefs upgrade time.
3666 Frequency : when setting specific ost indicies
3668 Description: QOS code breaks on skipped indicies
3669 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3670 created with --index need not be sequential.
3672 Severity : enhancement
3674 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3675 Details : The uninit_groups feature works in conjunction with the kernel
3676 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3677 up the pass1 processing of e2fsck. This is a read-only feature
3678 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3679 mount filesystems that have had this feature enabled.
3681 Severity : enhancement
3683 Description: Improve multi-block allocation algorithm to avoid fragmentation
3684 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3685 allocation locality and avoid filesystem fragmentation.
3687 ------------------------------------------------------------------------------
3689 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3691 * Support for kernels:
3692 2.4.21-47.0.1.EL (RHEL 3)
3693 2.6.5-7.283 (SLES 9)
3694 2.6.9-42.0.10.EL (RHEL 4)
3695 2.6.12.6 vanilla (kernel.org)
3696 2.6.16.27-0.9 (SLES 10)
3697 * Recommended e2fsprogs version: 1.39.cfs5
3699 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3703 Frequency : occasional, depends on client load and configuration
3704 Bugzilla : 12181, 12203
3705 Description: data loss for recently-modified files
3707 Details : In some cases it is possible that recently written or created
3708 files may not be written to disk in a timely manner (this should
3709 normally be within 30s unless client IO load is very high).
3710 The problem appears as zero-length files or files that are a
3711 multiple of 1MB in size after a client crash or client eviction
3712 that are missing data at the end of the file.
3714 This problem is more likely to be hit on clients where files are
3715 repeatedly created and unlinked in the same directory, clients
3716 have a large amount of RAM, have many CPUs, the filesystem has
3717 many OSTs, the clients are rebooted frequently, and/or the files
3718 are not accessed by other nodes after being written.
3720 The presence of the problem can be detected by looking at
3721 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3722 smaller than the second (nr_unused) then dirty files will not
3723 be flushed automatically to disk. "sync; sleep 10" should be
3724 run several times on the node before unmounting it to update
3725 Lustre (this is also safe to run on nodes without this problem).
3727 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3728 that can cause this same problem, so customers using that kernel
3729 also need to update the kernel in addition to Lustre. In order
3730 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3732 It is normal that files written just before a client crash (less
3733 than 30s) may not yet have been flushed to disk, even for local
3737 Frequency : frequent on thin XT3 nodes
3739 Description: UUID collision on thin XT3 Linux nodes
3740 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3741 randomly, since we relied on an insufficiently-seeded PRNG.
3746 Description: OSS hangs after "All ost request buffers busy"
3747 Details : A deadlock between quota and journal operations caused OSS
3748 hangs after printing "All ost request buffers busy."
3751 Frequency : always on liblustre builds
3753 Description: Cleanup compiler warnings on liblustre
3756 Frequency : always on liblustre builds on XT3
3758 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3762 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3763 > 32000 subdirectories
3764 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3765 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3766 subdirectory count crosses 32000. This will aid e2fsck to
3767 correctly handle more than 32000 subdirectories.
3770 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3772 Description: System hang on startup
3773 Details : This bug allowed the liblustre (e.g. catamount) client to
3774 return to the app before handling all startup RPCs. This
3775 could leave the node unresponsive to lustre network traffic
3776 and manifested as a server ptllnd timeout.
3778 Severity : enhancement
3780 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3781 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3782 boolean to control whether to print peer debug info when a
3783 client's RPC times out.
3788 Description: make O_SYNC working on 2.6 kernels
3789 Details : 2.6 kernels use different method for mark pages for write,
3790 so need add a code to lustre for O_SYNC work.
3795 Description: Failure to close file and release space on NFS
3796 Details : Put inode details into lock acquired in ll_intent_file_open.
3797 Use mdc_intent_lock in ll_intent_open to properly
3798 detect all kind of errors unhandled by mdc_enqueue.
3803 Description: proc file read during shutdown sometimes raced obd removal,
3805 Details : Add lock to prevent obd access after proc file removal.
3808 Frequency : Only for files larger than 4GB on 32-bit clients.
3810 Description: improperly doing page alignment of locks
3811 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3812 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3817 Description: RPCs being resent when they shouldn't be
3818 Details : Some RPCs that should not be resent are being resent. This
3819 can cause inconsistencies in the RPC state machine. Do not
3820 resend such requests.
3823 Frequency : rare, only with NFS export
3825 Description: Crash on NFS re-export node
3826 Details : under very unusual load conditions an assertion is hit in
3827 ll_intent_file_open()
3830 Frequency : only if OST filesystem is corrupted
3832 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3833 Details : for a short time RPCs with bulk IO are in the replay list,
3834 but replay of bulk IOs is unimplemented. If the OST filesystem
3835 is corrupted due to disk cache incoherency and then replay is
3836 started it is possible to trip an assertion. Avoid putting
3837 committed RPCs into the replay list at all to avoid this issue.
3842 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3843 Details : Large single O_DIRECT read and write calls can fail to allocate
3844 a sufficiently large buffer to process the request. In case of
3845 allocation failure the allocation is retried with a smaller
3846 buffer and broken into smaller requests.
3848 Severity : enhancement
3850 Description: Add -o localflock option to simulate old noflock behaviour.
3851 Details : This will achieve local-only flock/fcntl locks coherentness.
3856 Description: versioning check is incomplete
3857 Details : Checking the version difference of client vs. server, report
3858 error if the gap is too big.
3863 Description: add support PG_writeback bit
3864 Details : add support for PG_writeback bit for Lustre, for more carefull
3865 work with page cache in 2.6 kernel. This also fix some deadlocks
3866 and remove hack for work O_SYNC with 2.6 kernel.
3868 Severity : enhancement
3870 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3871 Details : The uninit_groups feature works in conjunction with the kernel
3872 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3873 up the pass1 processing of e2fsck. This is a read-only feature
3874 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3875 mount filesystems that have had this feature enabled.
3877 Severity : enhancement
3879 Description: Improve multi-block allocation algorithm to avoid fragmentation
3880 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3881 allocation locality and avoid filesystem fragmentation.
3883 ------------------------------------------------------------------------------
3885 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3887 * Support for kernels:
3888 2.6.9-42.0.3.EL (RHEL 4)
3889 2.6.5-7.276 (SLES 9)
3890 2.4.21-47.0.1.EL (RHEL 3)
3891 2.6.12.6 vanilla (kernel.org)
3892 2.6.16.21-0.8 (SLES10)
3893 * Recommended e2fsprogs version: 1.39.cfs2-0
3895 * The backwards-compatible /proc/sys/portals symlink has been removed
3896 in this release. Before upgrading, please ensure that you change
3897 any configuration scripts or /etc/sysctl.conf files that access
3898 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3899 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3900 in advance of the upgrade on any system running Lustre 1.4.6 or
3901 newer, since /proc/sys/lnet was added in that version.
3902 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3906 Frequency : only when quota is used
3908 Description: avoid scanning export list for quota master
3909 Details : Change the algorithms to avoid scanning export list in order
3910 to improve the efficiency.
3913 Frequency : MDS failover only, very rarely
3915 Description: "went back in time" messages on mds failover
3916 Details : The greatest transno may be lost when the current operation
3917 finishes with an error (transno==0) and the client's last_rcvd
3918 record is over-written. Save the greatest transno in the
3919 mds_last_transno for this case.
3922 Frequency : always for specific kernels and striping counts
3924 Description: client may get "Matching packet too big" without ACL support
3925 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3926 when trying to access files in certain configurations. The
3927 clients should in fact be denied when mounting because they do
3928 not understand ACLs.
3931 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3933 Description: many clients connecting with IO in progress causes connect timeouts
3934 Details : Avoid synchronous journal commits to avoid delays caused by many
3935 clients connecting/disconnecting when bulk IO is in progress.
3936 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3937 OST_IO_PORTAL to avoid delays behind potentially many pending
3941 Frequency : occasionally with multiple writers to a single file
3943 Description: shared writes to file may result in wrong size reported by stat()
3944 Details : Allow growing of kms when extent lock is cancelled
3947 Frequency : always with random mmap IO to multi-striped file
3949 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3950 Details : Only if the hole is at the end of OST object so that kms is too
3951 small. Fix is to increase kms accordingly in ll_nopage.
3954 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3956 Description: writes to a missing object would leak memory on the OST
3957 Details : If there is an inconsistency between the MDS and OST filesystems,
3958 such that the MDS references an object that doesn't exist, writes
3959 to that object will leak memory due to incorrect cleanup in the
3960 error handling path, eventually running out of memory on the OST.
3965 Description: Creating too long symlink causes lustre errors
3966 Details : Check symlink and name lengths before sending requests to MDS.
3969 Frequency : only if flock is enabled (not on by default)
3971 Description: posix locks not released on fd closure on 2.6.9+
3972 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3973 kernels, this caused such locks not to be released on fd close and
3974 then assertions on fs unmount about still used locks.
3977 Frequency : MDS failover only, very rarely
3979 Description: clients may get ASSERTION(granted_lock != NULL)
3980 Details : When request was taking a long time, and a client was resending
3981 a getattr by name lock request. The were multiple lock requests
3982 with the same client lock handle and
3983 mds_getattr_name->fixup_handle_for_resent_request found one of the
3984 lock handles but later failed with ASSERTION(granted_lock != NULL).
3989 Description: handle->h_buffer_credits > 0, assertion failure
3990 Details : h_buffer_credits is zero after truncate, causing assertion
3991 failure. This patch extends the transaction or creates a new
3995 Frequency : NFS re-export or patchless client
3996 Bugzilla : 11179, 10796
3997 Description: Crash on NFS re-export node (__d_move)
3998 Details : We do not want to hash the dentry if we don't have a lock.
3999 But if this dentry is later used in d_move, we'd hit uninitialised
4000 list head d_hash, so we just do this to init d_hash field but
4001 leave dentry unhashed.
4004 Frequency : NFS re-export or patchless client
4006 Description: NFS exports has problem with symbolic link
4007 Details : lustre client didn't properly install dentry when re-exported
4008 to NFS or running patchless client.
4011 Frequency : NFS re-export or patchless client
4013 Description: Various nfs/patchless fixes.
4014 Details : fixes reuse disconected alias for lookup process - this fixes
4015 warning "find_exported_dentry: npd != pd",
4016 fix permission error with open files at nfs.
4017 fix apply umask when do revalidate.
4020 Frequency : occasional
4022 Description: Crash on NFS re-export node
4023 Details : calling clear_page() on the wrong pointer triggered oops in
4024 generic_mapping_read().
4027 Frequency : rarely, using O_DIRECT IO
4029 Description: unaligned directio crashes client with LASSERT
4030 Details : check for unaligned buffers before trying any requests.
4033 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4035 Description: stale data returned from RAID cache
4036 Details : If only a small amount of IO is done to the RAID device before
4037 reading it again it is possible to get stale data from the RAID
4038 cache instead of reading it from disk.
4041 Frequency : always for sles10 kernel
4043 Description: sles10 support
4044 Details : ll_follow_link: compile fixes and using of nd_set_link
4045 under newer kernels.
4048 Frequency : depends on arch, kernel and compiler version, always on sles10
4051 Description: recursive or deep enough symlinks cause stack overflow
4052 Details : getting rid of large stack-allocated variable in
4056 Frequency : depends on hardware
4058 Description: lustre write performance loss in the SLES10 kernel
4059 Details : the performance loss is caused by using of write barriers in the
4060 ext3 code. The SLES10 kernel turns barrier support on by
4061 default. The fix is to undo that change for ldiskfs.
4063 ------------------------------------------------------------------------------
4065 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4067 * Support for kernels:
4068 2.6.9-42.0.3EL (RHEL 4)
4069 2.6.5-7.276 (SLES 9)
4070 2.4.21-47.0.1.EL (RHEL 3)
4071 2.6.12.6 vanilla (kernel.org)
4075 Frequency : quota enabled and large files being deleted
4077 Description: releasing more than 4GB of quota at once hangs OST
4078 Details : If a user deletes more than 4GB of files on a single OST it
4079 will cause the OST to spin in an infinite loop. Release
4080 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4085 Description: statfs data retrieved from /proc may be stale or zero
4086 Details : When reading per-device statfs data from /proc, in the
4087 {kbytes,files}_{total,free,avail} files, it may appear
4088 as zero or be out of date.
4091 Frequency : systems with MD RAID1 external journal devices
4093 Description: lconf's call to blkid is confused by RAID1 journal devices
4094 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4095 of returning all block devices that match the journal UUID.
4098 Frequency : always, for aggregate stripe size over 4GB
4100 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4101 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4102 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4105 Frequency : always if "lfs find" used on a local file/directory
4107 Description: "lfs find" segfaults if used on a local file/directory
4108 Details : The case where a directory component was not specified wasn't
4109 handled correctly. Handle this properly.
4112 Frequency : always on ppc64
4114 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4115 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4116 perform endianness conversion.
4119 Frequency : rarely (truncate to non-zero file size after write under load)
4120 Bugzilla : 10730, 10687
4121 Description: Files padded with zeros to next 4K multiple
4122 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4123 it is possible that files that are truncated to a non-zero size
4124 immediately after being written are filled with zero bytes beyond
4125 the truncated size. No file data is lost.
4127 Severity : enhancement
4129 Description: Allow recovery/failover for liblustre clients.
4130 Details : liblustre clients were unaware of failover configurations until
4133 Severity : enhancement
4135 Description: user file locks should fail when not mounting with flock option
4136 Details : Set up an error-returning stub in ll_file_operations.lock field
4137 to prevent incorrect behaviour when client is mounted without
4138 flock option. Also, set up properly f_op->flock field for
4142 Frequency : always on ia64
4144 Description: "lfs df" loops on printing out MDS statfs information
4145 Details : The obd_ioctl_data was not initialized and in some systems
4146 this caused a failure during the ioctl that did not return
4147 an error. Initialize the struct and return an error on failure.
4150 Frequency : SLES 9 only
4152 Description: Error of copying files with lustre special EAs as root
4153 Details : Client side always return success for setxattr call for lustre
4154 special xattr (currently only "trusted.lov").
4157 Frequency : rarely on clusters with both ia64+i386 clients
4159 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4160 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4161 clients are doing concurrent writes to the same file it is
4162 possible that the ia64 clients may LASSERT because the OST
4163 extent locks are not PAGE_SIZE aligned. Ensure that grown
4164 locks are always aligned on the request boundary.
4167 Frequency : specific use, occasional
4169 Description: Overwriting in use executable truncates on-disk binary image
4170 Details : If one node attempts to overwrite an executable in use by
4171 another node, we now correctly return ETXTBSY instead of
4172 truncating the file.
4174 Severity : enhancement
4176 Description: Async OSC create to avoid the blocking unnecessarily.
4177 Details : If an OST has no remaining object, system will block on the creating
4178 when need to create a new object on this OST. Now, ways use
4179 pre-created objects when available, instead of blocking on an
4180 empty osc while others are not empty. If we must block, we block
4181 for the shortest possible period of time.
4186 Description: chmod on Lustre root is propagated to other clients
4187 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4188 invalid by the follow_mount time.
4191 Frequency : liblustre clients only
4193 Description: Race in 'instant cancel' lock handling could lead to such locks
4194 never to be granted in case of SMP MDS
4195 Details : Do not destroy not yet granted but cbpending locks in
4199 Frequency : replay/resend of open
4201 Description: non null lock assetion failure in mds_intent_policy
4202 Details : Trying to replay/resend lockless open requests resulted in
4203 mds_open() returning 0 with no lock. Now it sets a flag if
4204 a lock is going to be returned.
4206 Severity : enhancement
4208 Description: Checksum enhancements
4209 Details : New checksum enhancements allow for resending RPCs that failed
4212 Severity : enhancement
4214 Description: Tunables on number of dirty pages in cacche
4215 Details : Allow to set limit on number of dirty pages cached.
4220 Description: client crash on unmount - lock still has references
4221 Details : In some error handling cases it was possible to leak a lock
4222 reference on a client while accessing a file. This was not
4223 harmful to the client during operation, but would cause the
4224 client to crash when the filesystem is unmounted.
4227 Frequency : specific case, rare
4229 Description: ETXTBSY on mds though file not in use by client
4230 Details : ETXTBSY is no longer incorrectly returned when attempting to
4231 chmod or chown a directory that the user previously tried to
4232 execute or a currently-executing binary.
4235 Frequency : extremely rare except on liblustre-based clients
4237 Description: Lustre space not freed when files are deleted
4238 Details : Clean up open-unlinked files after client eviction. Previously
4239 the unlink was skipped and the files remained as orphans.
4244 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4245 Details : In some cases it was possible to send a blocking callback to
4246 a client doing a glimpse, even though that client didn't get
4247 a lock granted. When the glimpse lock is cancelled on the OST
4248 the freed lock is left on the waiting list and corrupted the list.
4251 Frequency : all core dumps
4253 Description: Broke core dumps to lustre
4254 Details : Negative dentry may be unhashed if parent does not have UPDATE
4255 lock, but some callers, e.g. do_coredump, expect dentry to be
4256 hashed after successful create, hash it in ll_create_it.
4258 ------------------------------------------------------------------------------
4260 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4262 * Support for kernels:
4263 2.6.9-42.0.2.EL (RHEL 4)
4264 2.6.5-7.276 (SLES 9)
4265 2.4.21-40.EL (RHEL 3)
4266 2.6.12.6 vanilla (kernel.org)
4270 Frequency : always on RHEL 3
4272 Description: Number of open files grows over time
4273 Details : The number of open files grows over time, whether or not
4274 Lustre is started. This was due to a filp leak introduced
4275 by one of our kernel patches.
4277 ------------------------------------------------------------------------------
4279 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4281 * Support for kernels:
4282 2.6.9-42.EL (RHEL 4)
4283 2.6.5-7.276 (SLES 9)
4284 2.4.21-40.EL (RHEL 3)
4285 2.6.12.6 vanilla (kernel.org)
4290 Bugzilla : 5719, 9635, 9792, 9684
4291 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4292 Details : If a server is under heavy load and cannot reply to new
4293 connection requests before the client resends the (re)connect,
4294 the connection handling code can behave badly if two service
4295 threads are concurrently handing separate (re)connections from
4296 the same client. Add better locking to the connection handling
4297 code, and ensure that only a single connection will be processed
4298 for a given client UUID, even if the lock is dropped.
4300 Severity : enhancement
4302 Description: add TCP zero-copy support to kernel
4303 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4304 sends if the hardware supports scatter-gather and checksumming.
4305 This allows socklnd to do client-write and server-read more
4306 efficiently and reduce CPU utilization from skbuf copying.
4309 Frequency : only if NFS exporting from client
4311 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4312 Details : If setting attributes on a file created under NFS that had
4313 never been opened it would be possible to oops the client
4314 if the file had no objects.
4317 Frequency : always for liblustre
4319 Description: liblustre client does MDS+OSTs setattr RPC for each write
4320 Details : When doing a write from a liblustre client, the client
4321 incorrectly issued an RPC to the MDS and each OST the file was
4322 striped over in order to update the timestamps. When writing
4323 with small chunks and many clients this could overwhelm the MDS
4324 with RPCs. In all cases it would slow down the write because
4325 these RPCs are unnecessary.
4327 Severity : enhancement
4329 Description: allow number of MDS service threads to be changed at module load
4330 Details : It is now possible to change the number of MDS service threads
4331 running. Adding "options mds mds_num_threads={N}" to the MDS's
4332 /etc/modprobe.conf will set the number of threads for the next
4333 time Lustre is restarted (assuming the "mds" module is also
4334 reloaded at that time). The default number of threads will
4335 stay the same, 32 for most systems.
4340 Description: OST crash if filesystem is unformatted or corrupt
4341 Details : If an OST is started on a device that has never been formatted
4342 or if the filesystem is corrupt and cannot even mount then the
4343 error handling cleanup routines would dereference a NULL pointer.
4348 Description: NULL pointer deref in llap_from_page.
4349 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4350 incorrect) mapping if the page was truncated/reclaimed while it was
4351 searched for. Check for this condition and skip such pages when
4352 doing readahead. Introduce extra check to llap_from_page() to
4353 verify page->mapping->host is non-NULL (so page is not anonymous).
4356 Frequency : Sometimes when using sys_sendfile
4358 Description: "page not covered by a lock" warnings from ll_readpage
4359 Details : sendfile called ll_readpage without right page locks present.
4360 Now we introduced ll_file_sendfile that does necessary locking
4361 around call to generic_file_sendfile() much like we do in
4365 Frequency : with certain MDS communication failures at client mount time
4367 Description: NULL pointer deref after failed client mount
4368 Details : a client connection request may delayed by the network layer
4369 and not be sent until after the PTLRPC layer has timed out the
4370 request. If the client fails the mount immediately it will try
4371 to clean up before the network times out the request. Add a
4372 reference from the request import to the obd device and delay
4373 the cleanup until the network drops the request.
4376 Frequency : occasionally during client (re)connect
4378 Description: assertion failure during client (re)connect
4379 Details : processing a client connection request may be delayed by the
4380 client or server longer than the client connect timeout. This
4381 causes the client to resend the connection request. If the
4382 original connection request is replied in this interval, the
4383 client may trip an assertion failure in ptlrpc_connect_interpret()
4384 which thought it would be the only running connect process.
4387 Frequency : only with obd_echo servers and clients that are rebooted
4389 Description: kernel BUG accessing uninitialized data structure
4390 Details : When running an obd_echo server it did not start the ping_evictor
4391 thread, and when a client was evicted an uninitialized data
4392 structure was accessed. Start the ping_evictor in the RPC
4393 service startup instead of the OBD startup.
4395 Severity : enhancement
4396 Bugzilla : 10193 (patchless)
4397 Description: Remove dependency on various unexported kernel interfaces.
4398 Details : No longer need reparent_to_init, exit_mm, exit_files,
4399 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4402 Frequency : rare (only users of deprecated and unsupported LDAP config)
4404 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4405 Details : LDAP apparently contains 'lustreName' attributes instead of
4406 'name'. A simple remapping of the name is sufficient.
4409 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4411 Description: waiting_locks_callback trips kernel BUG if client is evicted
4412 Details : Running with the dump_on_timeout debug flag turned on makes
4413 it possible that the waiting_locks_callback() can try to dump
4414 the Lustre kernel debug logs from an interrupt handler. Defer
4415 this log dumping to the expired_lock_main() thread.
4417 Severity : enhancement
4419 Description: Support NFS exporting on 2.6 kernels.
4420 Details : Implement non-rawops metadata methods for NFS server to use without
4421 changing NFS server code.
4424 Frequency : very rare (synthetic metadata workload only)
4426 Description: two racing renames might cause an MDS thread to deadlock
4427 Details : Running the "racer" program may cause one MDS thread to rename
4428 a file from being the source of a rename to being the target of
4429 a rename at exactly the same time that another thread is doing
4430 so, and the second thread has already enqueued these locks after
4431 doing a lookup of the target and is trying to relock them in
4432 order. Ensure that we don't try to re-lock the same resource.
4435 Frequency : only very large systems with liblustre clients
4437 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4438 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4439 all OSTs in parallel. This allows the eviction of stale liblustre
4440 clients to proceed much faster than if they were done in series,
4441 and also offers similar improvements for other set_info RPCs.
4446 Description: excessive CPU usage during initial read phase on client
4447 Details : During the initial read phase on a client, it would agressively
4448 retry readahead on the file, consuming too much CPU and impacting
4449 performance (since 1.4.5.8). Improve the readahead algorithm
4450 to avoid this, and also improve some other common cases (read
4451 of small files in particular, where "small" is files smaller than
4452 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4457 Description: MDS crash when receiving packet with unknown intent.
4458 Details : Do not LBUG in unknown intent case, just return -EFAULT
4460 Severity : enhancement
4461 Bugzilla : 9293, 9385
4462 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4463 Details : Do not serialize getattr (non-intent version) and statfs.
4466 Frequency : occasional, when OST network is overloaded/intermittent
4468 Description: client evicted by OST after bulk IO timeout
4469 Details : If a client sends a bulk IO request (read or write) the OST
4470 may evict the client if it is unresposive to its data GET/PUT
4471 request. This is incorrect if the network is overloaded (takes
4472 too long to transfer the RPC data) or dropped the OST GET/PUT
4473 request. There is no need to evict the client at all, since
4474 the pinger and/or lock callbacks will handle this, and the
4475 client can restart the bulk request.
4478 Frequency : Always when mmapping file with no objects
4480 Description: client crashes when mmapping file with no objects
4481 Details : Check that we actually have objects in a file before doing any
4482 operations on objects in ll_vm_open, ll_vm_close and
4488 Description: Request leak when working with deleted CWD
4489 Details : Introduce advanced request refcount tracking for requests
4490 referenced from lustre intent.
4492 Severity : Enhancement
4494 Description: Cache open file handles on client.
4495 Details : MDS now will return special lock along with openhandle, if
4496 requested and client is allowed to hold openhandle, even if unused,
4497 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4498 closing files for every read/write openration.
4500 Severity : Enhancement
4502 Description: Cache open negative dentries on client when possible.
4503 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4504 negative dentries on lock revocation.
4509 Description: Remounting a client read-only wasn't possible with a zconf mount
4510 Details : It wasn't possible to remount a client read-only with llmount.
4512 Severity : enhancement
4513 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4514 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4515 MPICH is included to add Lustre-specific ADIO interfaces.
4516 This is based closely on the UFS ADIO layer and only differs
4517 in file creation, in order to allow the OST striping to be set.
4518 This is user-contributed code and not supported by CFS.
4523 Description: extended inode attributes (immutable, append-only) work improperly
4524 when 2.4 and 2.6 kernels are used on client/server or vice versa
4525 Details : Introduce kernel-independent values for these flags.
4527 Severity : enhancement
4530 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4531 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4532 granularity. Now, Fractional MB granularity is be supported,
4533 this is very useful for low memory system.
4535 Severity : enhancement
4537 Description: Getattr by fid
4538 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4539 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4540 have locks and avoids getattr rpc requests for it on every stat.
4543 Frequency : Always, for filesystems larger than 2TB
4545 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4546 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4547 filesystems larger than 2TB to be mounted with mballoc enabled.
4550 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4552 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4553 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4554 filesystems larger than 2TB is mounted then the kernel will
4555 silently corrupt the start of the filesystem. CONFIG_LBD is
4556 enabled for all CFS-supported kernels, but the possibility of
4557 this happening with a modified kernel config exists.
4559 Severity : enhancement
4561 Description: add client O_DIRECT support for 2.6 kernels
4562 Details : It is now possible to do O_DIRECT reads and writes to files
4563 in the Lustre client mountpoint on 2.6 kernel clients.
4565 Severity : enhancement
4567 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4568 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4569 parallel, not waiting for response from every OST before sending
4570 a rpc to the next OST.
4575 Description: setattr vs write race when updating file timestamps
4576 Details : Client processes that update a file timestamp into the past
4577 right after writing to the file (e.g. tar) it is possible that
4578 the updated file modification time can be reset to the current
4579 time due to a race between processing the setattr and write RPC.
4581 Severity : enhancement
4583 Description: Bring 'lfs find' closer in line with regular Linux find.
4584 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4585 -print0 options and obtains all the needed info through the lustre
4588 Severity : enhancement
4590 Description: support up to 1024 configured devices on one node
4591 Details : change obd_dev array from statically allocated to dynamically
4592 allocated structs as they are first used to reduce memory usage
4597 Description: Flush dirty partially truncated pages during truncate
4598 Details : Immediatelly flush partially truncated pages in filter_setattr,
4599 this way we completely avoid having any pages in page cache on OST
4600 and can retire ugly workarounds during writes to flush such pages.
4605 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4606 Details : protect inode from truncation within vfs_unlink() context
4607 just take a reference before calling vfs_unlink() and release it
4608 when parent's i_sem is free.
4613 Description: last_id value checked outside lock on OST caused LASSERT failure
4614 Details : If there were multiple MDS->OST object precreate requests in
4615 flight, it was possible that the OST's last object id was checked
4616 outside a lock and incorrectly tripped an assertion. Move checks
4617 inside locks, and discard old precreate requests.
4620 Frequency : always, if extents are used on OSTs
4622 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4623 Details : OSTs using the extents format would not zero the high 16 bits of
4624 the index physical block number. This is not a problem for any
4625 OST filesystems smaller than 16TB, and no kernels support ext3
4626 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4627 new/modified files) and can be fixed for existing filesystems
4628 with e2fsprogs-1.39-cfs1.
4633 Description: import connection selection may be incorrect if timer wraps
4634 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4635 import connections to be ignored if the 32-bit jiffies counter
4636 wraps. Use a 64-bit jiffies counter.
4639 Frequency : very large clusters immediately after boot
4641 Description: LNET request buffers exhausted under heavy short-term load
4642 Details : If a large number of client requests are generated on a service
4643 that has previously never seen so many requests it is possible
4644 that the request buffer growth cannot keep up with the spike in
4645 demand. Instead of dropping incoming requests, they are held in
4646 the LND until the RPC service can accept more requests.
4649 Frequency : Sometimes during replay
4651 Description: Assertion failure in ll_local_open after replay.
4652 Details : If replay happened on an open request reply before we were able
4653 to set replay handler, reply will become not swabbed tripping the
4654 assertion in ll_local_open. Now we set the handler right after
4655 recognising of open request
4658 Frequency : very rare
4660 Description: kernel reports "badness in vsnprintf"
4661 Details : Reading from the "recovery_status" /proc file in small chunks
4662 may cause a negative length in lprocfs_obd_rd_recovery_status()
4663 call to vsnprintf() (which is otherwise harmless). Exit early
4664 if there is no more space in the output buffer.
4666 Severity : enhancement
4668 Description: clear OBD RPC statistics by writing to them
4669 Details : It is now possible to clear the OBD RPC statistics by writing
4670 to the "stats" file.
4675 Description: Client mtime is not the same on different clients after utimes
4676 Details : In some cases, the client was using the utimes() syscall on
4677 a file cached on another node. The clients now validate the
4678 ctime from the MDS + OSTs to determine which one is right.
4683 Description: Inability to activate failout mode
4684 Details : lconf script incorrectly assumed that in pythong string's numeric
4685 value is used in comparisons.
4688 Frequency : always with multiple stripes per file
4690 Description: Inefficient object allocation for mutli-stripe files
4691 Details : When selecting which OSTs to stripe files over, for files with
4692 a stripe count that divides evenly into the number of OSTs,
4693 the MDS is always picking the same starting OST for each file.
4694 Return the OST selection heuristic to the original design.
4699 Description: mount failures may take full timeout to return an error
4700 Details : Under some heavy load conditions it is possible that a
4701 failed mount can wait for the full obd_timeout interval,
4702 possibly several minutes, before reporting an error.
4703 Instead return an error as soon as the status is known.
4705 Frequency : quota enabled and large files being deleted
4707 Description: releasing more than 4GB of quota at once hangs OST
4708 Details : If a user deletes more than 4GB of files on a single OST it
4709 will cause the OST to spin in an infinite loop. Release
4710 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4715 Description: statfs data retrieved from /proc may be stale or zero
4716 Details : When reading per-device statfs data from /proc, in the
4717 {kbytes,files}_{total,free,avail} files, it may appear
4718 as zero or be out of date.
4721 Frequency : systems with MD RAID1 external journal devices
4723 Description: lconf's call to blkid is confused by RAID1 journal devices
4724 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4725 of returning all block devices that match the journal UUID.
4728 Frequency : always, for aggregate stripe size over 4GB
4730 Description: assertion fails when trying to use 4GB stripe size
4731 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4732 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4735 Frequency : always on ppc64
4737 Description: the first write on an ext3 filesystem with mballoc got stuck
4738 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4739 perform endianness conversion.
4741 ------------------------------------------------------------------------------
4743 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4745 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4746 INTEROPERATE with older versions automatically. Please read the
4747 user documentation before upgrading any part of a live system.
4748 * WARNING: Lustre networking configuration changes are required with
4749 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4752 * Support for kernels:
4753 2.6.9-22.0.2.EL (RHEL 4)
4754 2.6.5-7.244 (SLES 9)
4755 2.6.12.6 vanilla (kernel.org)
4758 Severity : enhancement
4759 Bugzilla : 7981/8208
4760 Description: Introduced Lustre Networking (LNET)
4761 Details : LNET is new networking infrastructure for Lustre, it includes
4762 a reorganized network configuration mode (see the user
4763 documentation for full details) as well as support for routing
4764 between different network fabrics. Lustre Networking Devices
4765 (LNDs) for the supported network fabrics have also been
4766 created for this new infrastructure.
4768 Severity : enhancement
4769 Description: Introduced Access control lists
4770 Details : clients can set ACLs on files and directories in order to have
4771 more fine-grained permissions than the standard Unix UGO+RWX.
4772 The MDS must be started with the "-o acl" mount option.
4774 Severity : enhancement
4775 Description: Introduced filesystem quotas
4776 Details : Administrators may now establish per-user quotas on the
4779 Severity : enhancement
4781 Description: Configuration change for the XT3
4782 The PTLLND is now used to run Lustre over Portals on the XT3
4783 The configure option(s) --with-cray-portals are no longer used.
4784 Rather --with-portals=<path-to-portals-includes> is used to
4785 enable building on the XT3. In addition to enable XT3 specific
4786 features the option --enable-cray-xt3 must be used.
4791 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4792 Details : A race between allocating a new llcd and re-getting the llcd_lock
4793 allowed another thread to grab newly-allocated llcd.
4795 Severity : enhancement
4797 Description: 2.6 OST async journal commit and locking fix to improve performance
4798 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4799 2.6 kernels are now async as they already were in 2.4 kernels so
4800 that they can commit concurrently with the network bulk transfer.
4801 For block-allocated files the filter allocation semaphore is held
4802 to avoid filesystem fragmentation during allocation. BKL lock
4803 removed for 2.6 xattr operations where it is no longer needed.
4808 Description: lconf incorrectly determined whether two IP networks could talk
4809 Details : In some more complicated routing and multiple-network
4810 configurations, lconf will avoid trying to make a network
4811 connection to a disjoint part of the IP space. It was doing the
4812 math incorrectly for one set of cases.
4817 Description: Fix for potential infinite loop processing records in an llog.
4818 Details : If an llog record is corrupted/zeroed, it is possible to loop
4819 forever in llog_process(). Validate the llog record length
4820 and skip the remainder of the block on error.
4823 Frequency : occasional (liblustre only)
4825 Description: liblustre could not open files whose last component is a symlink
4826 Details : sysio_path_walk() would incorrectly pass the open intent to
4827 intermediate path components.
4830 Frequency : rare (liblustre only with non-standard tuning)
4831 Bugzilla : 7201 (7350)
4832 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4833 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4834 after it has been released, only do this for referenced locks
4836 Severity : enhancement
4838 Description: specifying an (invalid) directory default stripe_size of -1
4839 would reset the directory default striping
4840 Details : stripe_size -1 was used internally to signal directory stripe
4841 removal, now use "all default" to signal dir stripe removal
4842 as a directory striping of "all default" is not useful
4845 Frequency : common for large clusters running liblustre clients
4847 Description: doing an ls when liblustre clients are running is slow
4848 Details : sending a glimpse AST to a liblustre client waits for every AST
4849 to time out, as liblustre clients will not respond. Since they
4850 cannot cache data we refresh the OST lock LVB from disk instead.
4852 Severity : enhancement
4854 Description: doing an ls at the same time as file IO can be slow
4855 Details : enqueue and other "small" requests can be blocked behind many
4856 large IO requests. Create a new OST IO portal for non-IO
4857 requests so they can be processed faster.
4860 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4862 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4863 created file with mode 000
4864 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4865 allow an ftruncate() as a non-root user to a file with mode 000.
4866 Lustre now respects this flag to disable mode checks when
4867 truncating a file owned by the user
4870 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4873 Description: Revoking locks from clients that went dead or catatonic might take
4875 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4876 cancellation of such locks instant on servers without waiting for
4877 any reply from clients. Clients drops these locks when cancel
4878 notification from server is received without replying.
4881 Frequency : liblustre-only, when liblustre client dies or becomes busy
4883 Description: Doing ls on Linux clients can take a long time with active
4885 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4886 granting such locks to it in the first place if possible. Locks
4887 are taken by proxy on the OST during the read or write and
4888 dropped immediately afterward. Add connect flags handling, do
4889 not grant locks to liblustre clients for glimpse ASTs.
4891 Severity : enhancement
4893 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4894 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4895 sequential reads and behaves badly with applications doing random
4896 reads. Improve it by reading ahead at least read region, and
4897 avoiding excessive large RPC for small reads.
4899 Severity : enhancement
4901 Description: Creating more than 1000 files for a single job may cause a load
4902 imbalance on the OSTs if there are also a large number of OSTs.
4903 Details : qos_prep_create() uses an OST index reseed value that is an
4904 even multiple of the number of available OSTs so that if the
4905 reseed happens in the middle of the object allocation it will
4906 still utilize the OSTs as uniformly as possible.
4911 Description: OST or MDS may oops in ping_evictor_main()
4912 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4913 but doesn't restart at beginning of obd_exports_timed list
4916 Severity : enhancement
4918 Description: improve by-nid export eviction on the MDS and OST
4919 Details : allow multiple exports with the same NID to be evicted at one
4920 time without re-searching the exports list.
4923 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4925 Description: MDS may oops in groups_free()
4926 Details : in rare race conditions a newly allocated group_info struct is
4927 freed again, and this can be NULL. The 2.4 compatibility code
4928 for groups_free() checked for a NULL pointer, but 2.6 did not.
4931 Frequency : common for liblustre clients doing little filesystem IO
4932 Bugzilla : 9352, 7313
4933 Description: server may evict liblustre clients accessing contended locks
4934 Details : if a client is granted a lock or receives a completion AST
4935 with a blocking AST already set it would not reply to the AST
4936 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4939 Frequency : lfs setstripe, only systems with more than 160 OSTs
4941 Description: unable to set striping with a starting offset beyond OST 160
4942 Details : llapi_create_file() incorrectly limited the starting stripe
4943 index to the maximum single-file stripe count.
4946 Frequency : LDAP users only
4948 Description: lconf did not handle in-kernel recovery with LDAP properly
4949 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4951 Severity : enhancement
4953 Description: bind OST threads to NUMA nodes to improve performance
4954 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4955 node and do their allocations there to localize memory access
4957 Severity : enhancement
4959 Description: llmount can determine client NID directly from Myrinet (GM)
4960 Details : the client NID code from gmnalnid was moved directly into
4961 llmount, removing the need to use this or specifying the
4962 client NID explicitly when mounting GM clients with zeroconf
4965 Frequency : if client is started with down MDS
4967 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4968 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4969 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4970 ptlrpc_import_delayed_req() because we didn't check if the
4971 request was interrupted, and we also didn't break out of the
4972 event loop if there was no timeout
4977 Description: data loss during non-page-aligned writes to a single file from
4978 both multiple nodes and multiple threads on one node at same time
4979 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4980 inconsistency led to false short-reads, that were cached and later
4981 used by ->prepare_write() to fill in partially written page,
4982 leading to data loss.
4985 Frequency : always, if lconf --abort_recovery used
4987 Description: lconf --abort_recovery fails with 'Operation not supported'
4988 Details : lconf was attempting to abort recovery on the MDT device and not
4991 Severity : enhancement
4993 Description: remove cleanup logs
4994 Details : replace lconf-generated cleanup logs with lustre internal
4995 cleanup routines. Eliminates the need for client-cleanup and
4998 Severity : enhancement
5000 Description: add support for EAs (user and system) on lustre filesystems
5001 Details : it is now possible to store extended attributes in the Lustre
5002 client filesystem, and with the user_xattr mount option it
5003 is possible to allow users to store EAs on their files also
5005 Severity : enhancement
5007 Description: Add possibility (config option) to show minimal available OST free
5009 Details : When compiled with --enable-mindf configure option, statfs(2)
5010 (and so, df) will return least minimal free space available from
5011 all OSTs as amount of free space on FS, instead of summary of
5012 free spaces of all OSTs.
5014 Severity : enhancement
5016 Description: do not expand extent locks acquired on OST-side
5017 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5018 by server: they are not cached anyway.
5021 Frequency : when mmap is used/binaries executed from Lustre
5023 Description: Unmmap pages before throwing them away from read cache.
5024 Details : llap_shrink cache now attempts to unmap pages before discarding
5025 them (if unmapping failed - do not discard). SLES9 kernel has
5026 extra checks that trigger if this unmapping is not done first.
5031 Description: lconf didn't resolve symlinks before checking to see whether a
5032 given mountpoint was already in use
5035 Frequency : when migrating failover services
5036 Bugzilla : 6395, 9514
5037 Description: When migrating a subset of services from a node (e.g. failback
5038 from a failover service node) the remaining services would
5039 time out and evict clients.
5040 Details : lconf --force (implied by --failover) sets the global obd_timeout
5041 to 5 seconds in order to quickly disconnect, but this caused
5042 other RPCs to time out too quickly. Do not change the global
5043 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5045 Severity : enhancement
5046 Frequency : if MDS is started with down OST
5047 Bugzilla : 9439,5706
5048 Description: Allow startup/shutdown of an MDS without depending on the
5049 availability of the OSTs.
5050 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5051 Add appropriate locking and lov-osc refcounts for safe
5052 cleaning. Add osc abort_inflight calls in case the
5053 synchronize never started.
5056 Frequency : occasional (Cray XT3 only)
5058 Description: root not authorized to access files in CRAY_PORTALS environment
5059 Details : The client process capabilities were not honoured on the MDS in
5060 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5061 been accessed by an authorized user then root was able to access
5062 the file on the local client also. The root user capabilities
5063 are now allowed on the MDS, as this environment has secure UID.
5066 Frequency : occasional
5068 Description: ldiskfs "too long searching" message happens too often
5069 Details : A debugging message (otherwise harmless) prints too often on
5070 the OST console. This has been reduced to only happen when
5071 there are fragmentation problems on the filesystem.
5076 Description: Division by zero in statfs when all OSCs are inactive
5077 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5078 lov_statfs divides by value returned by lov_get_stripecnt().
5082 Bugzilla : 9489, 3273
5083 Description: First write from each client to each OST was only 4kB in size,
5084 to initialize client writeback cache, which caused sub-optimal
5085 RPCs and poor layout on disk for the first written file.
5086 Details : Clients now request an initial cache grant at (re)connect time
5087 and so that they can start streaming writes to the cache right
5088 away and always do full-sized RPCs if there is enough data.
5089 If the OST is rebooted the client also re-establishes its grant
5090 so that client cached writes will be honoured under the grant.
5095 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5096 Details : Now I/O RPCs go to different portal number and (presumably) fast
5097 lock requests (and glimses) and other RPCs get their own service
5098 threads pool that should be able to service those RPCs
5101 Severity : enhancement
5103 Description: Ability to exchange lustre version between client and servers and
5104 issue warnings at client side if client is too old. Also for
5105 liblustre clients there is ability to refuse connection of too old
5107 Details : New 'version' field is added to connect data structure that is
5108 filled with version info. That info is later checked by server and
5112 Frequency : rare, liblustre only.
5113 Bugzilla : 9296, 9581
5114 Description: Two simultaneous writes from liblustre at offset within same page
5115 might proceed at the same time overwriting eachother with stale
5117 Details : I/O lock withing llu_file_prwv was released too early, before data
5118 actually was hitting the wire. Extended lock-holding time until
5119 server acknowledges receiving data.
5122 Frequency : extremely rare. Never observed in practice.
5124 Description: avoid generating lustre_handle cookie of 0.
5125 Details : class_handle_hash() generates handle cookies by incrementing
5126 global counter, and can hit 0 occasionaly (this is unlikely, but
5127 not impossible, because initial value of cookie counter is
5128 selected randonly). Value of 0 is used as a sentinel meaning
5129 "unassigned handle" --- avoid it. Also coalesce two critical
5130 sections in this function into one.
5132 Severity : enhancement
5134 Description: allow liblustre clients to delegate truncate locking to OST
5135 Details : To avoid overhead of locking, liblustre client instructs OST to
5136 take extent lock in ost_punch() on client's behalf. New connection
5137 flag is added to handle backward compatibility.
5139 Severity : enhancement
5140 Bugzilla : 4928, 7341, 9758
5141 Description: allow number of OST service threads to be specified
5142 Details : a module parameter allows the number of OST service threads
5143 to be specified via "options ost ost_num_threads={N}" in the
5144 OSS's /etc/modules.conf or /etc/modprobe.conf.
5148 Bugzilla : 6146, 9635, 9895
5149 Description: servers crash with bad pointer in target_handle_connect()
5150 Details : In rare cases when a client is reconnecting it was possible that
5151 the connection request was the last reference for that export.
5152 We would temporarily drop the export reference and get a new
5153 one, but this may have been the last reference and the export
5154 was just destroyed. Get new reference before dropping old one.
5156 Severity : enhancement
5157 Frequency : if client is started with failover MDS
5159 Description: Allow multiple MDS hostnames in the mount command
5160 Details : Try to read the configuration from all specified MDS
5161 hostnames during a client mount in case the "primary"
5164 Severity : enhancement
5166 Description: Stop sending data to evicted clients as soon as possible.
5167 Details : Check if the client we are about to send or are sending data to
5168 was evicted already. (Check is done every second of waiting,
5169 for which l_wait_event interface was extended to allow checking
5170 of exit condition at specified intervals).
5173 Frequency : rare, normally only when NFS exporting is done from client
5175 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5177 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5180 Frequency : always, if srand() is called before liblustre initialization
5182 Description: Liblustre uses system PRNG disturbing its usage by user application
5183 Details : Introduce internal to lustre fast and high-quality PRNG for
5184 lustre usage and make liblustre and some other places in generic
5185 lustre code to use it.
5187 Severity : enhancement
5188 Bugzilla : 9477, 9557, 9870
5189 Description: Verify that the MDS configuration logs are updated when xml is
5190 Details : Check if the .xml configuration logs are newer than the config
5191 logs stored on the MDS and report an error if this is the case.
5192 Request --write-conf, or allow starting with --old_conf.
5194 Severity : enhancement
5196 Description: Handle symlinks in the path when checking if Lustre is mounted.
5197 Details : Resolve intermediate symlinks when checking if a client has
5198 mounted a filesystem to avoid duplicate client mounts.
5203 Description: lconf can hit an error exception but still return success.
5204 Details : The lconf command catches the Command error exception at the top
5205 level script context and will exit with the associated exit
5206 status, but doesn't ensure that this exit status is non-zero.
5211 Description: failure of ptlrpc thread startup can cause oops
5212 Details : Starting a ptlrpc service thread can fail if there are a large
5213 number of threads or the server memory is very fragmented.
5214 Handle this without oopsing.
5217 Frequency : always, only if liblustre and non-default acceptor port was used
5219 Description: liblustre cannot connect to servers with non-default acceptor port
5220 Details : tcpnal_set_default_params() was not called and was therefore
5221 ignoring the environment varaible TCPNAL_PORT, as well as other
5222 TCPNAL_ environment variables
5227 Description: two objects could be created on the same OST for a single file
5228 Details : If an OST is down, in some cases it was possible to create two
5229 objects on a single OST for a single file. No problems other
5230 than potential performance impact and spurious error messages.
5234 Bugzilla : 5681, 9562
5235 Description: Client may oops in ll_unhash_aliases
5236 Details : Client dcache may become inconsistent in race condition.
5237 In some cases "getcwd" can fail if the current directory is
5243 Description: Inode refcounting problems in NFS export code
5244 Details : link_raw functions used to call d_instantiate without obtaining
5245 extra inode reference first.
5249 Bugzilla : 9942, 9903
5250 Description: Referencing freed requests leading to crash, memleaks with NFS.
5251 Details : We used to require that call to ll_revalidate_it was always
5252 followed by ll_lookup_it. Also with revalidate_special() it is
5253 possible to call ll_revalidate_it() twice for the same dentry
5254 even if first occurence returned success. This fix changes semantic
5255 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5256 reference on a request referred from the intent.
5257 ll_intent_release() then releases such a request.
5260 Frequency : rare, normally benchmark loads only
5262 Description: unlinked inodes were kept in memory on the client
5263 Details : If a client is repeatedly creating and unlinking files it
5264 can accumulate a lot of stale inodes in the inode slab cache.
5265 If there is no other client load running this can cause the
5266 client node to run out of memory. Instead flush old inodes
5267 from client cache that have the same inode number as a new inode.
5270 Frequency : SLES9 2.6.5 kernel and long filenames only
5271 Bugzilla : 9969, 10379
5272 Description: utime reports stale NFS file handle
5273 Details : SLES9 uses out-of-dentry names in some cases, which confused
5274 the lustre dentry revalidation. Change it to always use the
5278 Frequency : rare, unless heavy write-truncate concurrency is continuous
5279 Bugzilla : 4180, 6984, 7171, 9963, 9331
5280 Description: OST becomes very slow and/or deadlocked during object unlink
5281 Details : filter_destroy() was holding onto the parent directory lock
5282 while truncating+unlinking objects. For very large objects this
5283 may block other threads for a long time and slow overall OST
5284 responsiveness. It may also be possible to get a lock ordering
5285 deadlock in this case, or run out of journal credits because of
5286 the combined truncate+unlink. Solution is to do object truncate
5287 first in one transaction without parent lock, and then do the
5288 final unlink in a new transaction with the parent lock. This
5289 reduces the lock hold time dramatically.
5292 Frequency : rare, 2.4 kernels only
5294 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5295 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5296 before unmounting their filesystem in order to avoid deadlock.
5297 The kernel_locked() function in 2.4 kernels only checks whether
5298 the kernel lock is held, not whether it is this process that is
5299 holding it as 2.6 kernels do.
5304 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5305 Details : The client ptlrpc code may be trying to reconnect to a down
5306 server before a previous connection attempt has timed out.
5307 Increase the reconnect interval to be longer than the connection
5308 timeout interval to avoid sending duplicate connections to
5312 Frequency : echo_client brw_test command
5314 Description: fix echo_client to work with OST preallocated code
5315 Details : OST preallocation code (5137) didn't take echo_client IO path
5316 into account: echo_client calls filter methods outside of any
5317 OST thread and, hence, there is no per-thread preallocated
5318 pages and buffers to use. Solution: hijack pga pages for IO. As
5319 a byproduct, this avoids unnecessary data copying.
5323 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5324 Description: Client can oops in mdc_commit_close() after open replay
5325 Details : It was possible for the MDS to return an open request with no
5326 transaction number in mds_finish_transno() if the client was
5327 evicted, but without actually returning an error. Clients
5328 would later try to replay that open and may trip an assertion
5329 Simplify the client close codepath, and always return an error
5330 from the MDS in case the open is not successful.
5333 Frequency : rare, 2.6 OSTs only
5335 Description: OST may deadlock under high load on fragmented files
5336 Details : If there was a heavy load and highly-fragmented OST filesystems
5337 it was possible to have all the OST threads deadlock waiting on
5338 allocation of biovecs, because the biovecs were not released
5339 until the entire RPC IO was completed. Instead, release biovecs
5340 as soon as they are complete to ensure forward IO progress.
5342 Severity : enhancement
5344 Description: Support for specifying external journal device at mount
5345 Details : If an OST or MDS device is formatted with an external journal
5346 device, this device major/minor is stored in the ext3 superblock
5347 and may not be valid for failover. Allow detecting and
5348 specifying the external journal at mount time.
5353 Description: Mounting an MDS with pending unlinked files may cause oops
5354 Details : target_finish_recovery() calls mds_postrecov() which returned
5355 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5356 considers this an error and immediately begins cleaning up the
5357 lov, just after starting the mds_lov process
5359 Severity : enhancement
5361 Description: Implement 'lfs df' to report actual free space on per-OST basis
5362 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5363 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5364 usage of objects; '-h' to report in human readable format.
5366 ------------------------------------------------------------------------------
5368 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5375 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5376 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5377 at startup during orphan unlinking. Instead, skip buddy update
5378 before setup, as it will be regenerated after recovery is complete.
5383 Description: If an OST is inactive, its locks might reference stale inodes.
5384 Details : lov_change_cbdata() must iterate over all namespaces, even if
5385 they are inactive to clear inode references from the lock.
5387 Severity : enhancement
5388 Frequency : occasional, if non-standard max_dirty_mb used
5390 Description: Client will block write RPCs if not enough grant
5391 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5392 then the client will block writes while waiting for another RPC
5393 to complete instead of consuming its dirty limit. With change
5394 we get improved performance when max_dirty_mb is small.
5396 Severity : enhancement
5397 Bugzilla : 3389, 6253
5398 Description: Add support for supplementary groups on the MDS.
5399 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5400 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5401 lookups for user supplementary groups into a cache.
5405 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5406 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5407 for liblustre O_EXCL is enforced by MDS.
5411 Description: readdir never returns NULL in liblustre.
5412 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5413 dentry in struct dirent.
5417 Frequency : liblustre only, depends on application IO pattern
5418 Description: liblustre clients evicted if not contacting servers
5419 Details : Don't put liblustre clients into the ping_evictor list, so
5420 they will not be evicted by the pinger ever.
5422 Severity : enhancement
5424 Description: Add ability to evict clients by NID from MDS.
5425 Details : By echoing "nid:$NID" string into
5426 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5427 $NID would be instantly evicted from this MDS and from all active
5428 OSTs connected to it.
5432 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5433 Details : lookup_it_finish() used to query file size from OSTs that was not
5438 Description: service threads change working directory to that of init
5439 Details : Starting lustre service threads may pin the working directory
5440 of the parent thread, making that filesystem busy. Threads
5441 now change to the working directory of init to avoid this.
5445 Frequency : during shutdown only
5446 Description: shutdown with a failed MDS or OST can cause unmount to hang
5447 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5452 Frequency : 2.6 only
5453 Description: chmod/chown may include an extra supplementary group
5454 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5455 supplementary group and if none is specified this is used.
5458 Bugzilla : 5479 (6816)
5460 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5461 Details : If lookup is in progress on a file that is unlinked we might try
5462 to revalidate the inode and fail in revalidate after lookup is
5463 complete and ll_file_open() enqueues the open again but
5464 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5467 Frequency : always, if lconf --abort_recovery used
5469 Description: lconf --abort_recovery fails with 'Operation not supported'
5470 Details : lconf was attempting to abort recovery on the MDT device and not
5473 ------------------------------------------------------------------------------
5475 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5480 Frequency : rare (only unsupported configurations with a node running as an
5482 Bugzilla : 6514, 5137
5483 Description: Mounting a Lustre file system on a node running as an OST could
5485 Details : OSTs now preallocates memory needed to write out data at
5486 startup, instead of when needed, to avoid having to
5487 allocate memory in possibly low memory situations.
5488 Specifically, if the file system is mounted on on OST,
5489 memory pressure could force it to try to write out data,
5490 which it needed to allocate memory to do. Due to the low
5491 memory, it would be unable to do so and the node would
5492 become unresponsive.
5494 Severity : enhancement
5496 Description: Addition of lconf --service command line option
5497 Details : lconf now accepts a '--service <arg>' option, which is
5498 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5500 Severity : enhancement
5502 Description: Failover mode is now the default for OSTs.
5503 Details : By default, OSTs will now run in failover mode. To return to
5504 the old behaviour, add '--failout' to the lmc line for OSTs.
5506 Severity : enhancement
5508 Description: Health checks are now provided for MDS and OSTs
5509 Details : Additional detailed health check information on MSD and OSTs
5510 is now provided through the procfs health_check value.
5513 Frequency : occasional, depends on IO load
5515 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5516 after numerous create/delete cycles
5517 Details : The ext3 inode allocation policy would not allocate new inodes
5518 very well on the OSTs because there are no new directories
5519 being created. Instead we look for groups with free space if
5520 the parent directories are nearly full.
5525 Description: Network or server problems during mount may cause partially
5526 mounted clients instead of returning an error.
5527 Details : The config llog parsing code may overwrite the error return
5528 code during mount error handling, returning success instead
5534 Description: MDS can fail to allocate large reply buffers
5535 Details : After long uptimes the MDS can fail to allocate large reply
5536 buffers (e.g. zconf client mount config records) due to memory
5537 fragmentation or consumption by the buffer cache. Preallocate
5538 some large reply buffers so that these replies can be sent even
5539 under memory pressure.
5543 Frequency : rare (liblustre)
5544 Description: fsx running with liblustre complained that using truncate() to
5545 extend the file doesn't work. This patch corrects that issue.
5546 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5547 ATTR_SIZE and lsm use in llu_setattr_raw.
5551 Frequency : rare, only 2.6 kernels
5552 Description: Unusual file access patterns on the MDS may result in inode
5553 data being lost in very rare circumstances.
5554 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5555 code in the 2.6 kernel caused the inode and/or EA data not to
5556 be read from disk, causing single-file corruption.
5560 Frequency : rare, only 2.6 filesystems using extents
5561 Description: Heavy concurrent write and delete load may cause data corruption.
5562 Details : It was possible under high-load situations to have an extent
5563 metadata block in the block device cache from a just-unlinked
5564 file overwrite a newly-allocated data block. We now unmap any
5565 metadata buffers that alias just-allocated data blocks.
5569 Frequency : filesystems with default stripe_count larger than 77
5570 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5571 Details : lconf specifies an inode size of 4096 bytes when the default
5572 stripe_count is larger than 77. This conflicts with the default
5573 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5574 this case to avoid pinning too much memory for large EAs.
5576 ------------------------------------------------------------------------------
5578 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5583 Frequency : rare (extremely heavy IO load with hundreds of clients)
5585 Description: Client is evicted, gets IO error writing to file
5586 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5587 caused clients to be evicted for AST timeouts. The fixes in
5588 bug 5192 mean we no longer need to have such short AST timeouts
5589 so ldlm_timeout has been increased.
5592 Frequency : occasional during --force or --failover shutdown under load
5593 Bugzilla : 5949, 4834
5594 Description: Server oops/LBUG if stopped with --force or --failover under load
5595 Details : a collection of import/export refcount and cleanup ordering
5596 issues fixed for safer force cleanup
5599 Frequency : only filesystems larger than 120 OSTs
5600 Bugzilla : 5990, 6223
5601 Description: lfs getstripe would oops on a very large filesystem
5602 Details : lov_getconfig used kfree on vmalloc'd memory
5605 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5606 Bugzilla : 6242, 6243
5607 Description: reading from files that had been truncated to a non-zero size
5608 but never opened returned no data
5609 Details : ll_file_read() reads zeros from no-object files to EOF
5614 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5615 Details : There's little harm in aborting MDS/OSS recovery and letting it
5616 try again, so I removed the LASSERT and return an error instead.
5618 Severity : enhancement
5620 Description: New debugging infrastructure for tracking down data corruption
5621 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5622 (b) cover more of the client-side code path, and (c) try to narrow
5623 down where problems occurred
5627 Bugzilla : 3819, 4364, 4397, 6313
5628 Description: Racing close and eviction MDS could cause assertion in mds_close
5629 Details : It was possible to get multiple mfd references during close and
5630 client eviction, leading to one thread referencing a freed mfd.
5632 Severity: : enhancement
5633 Bugzilla : 3262, 6359
5634 Description: Attempts to reconnect to servers are now more aggressive.
5635 Details : This builds on the enhanced upcall-less recovery that was added
5636 in 1.4.2. When trying to reconnect to servers, clients will
5637 now try each server in the failover group every 10 seconds. By
5638 default, clients would previously try one server every 25 seconds.
5643 Description: After recovery, certain operations trigger a failed
5644 assertion on a client.
5645 Details : Failing over an mds, using lconf -d --failover, while a
5646 client was doing a readdir() call would cause the client to
5647 LBUG after recovery completed and the readdir() was resent.
5649 Severity : enhancement
5651 Description: Default groups are now added by lconf
5652 Details : You can now run lconf --group <servicename> without having to
5653 manually add groups with lmc.
5656 Frequency : occasional
5658 Description: Nodes with an elan id of 0 trigger a failed assertion
5661 Frequency : always when accessing e.g. tty/console device nodes
5663 Description: tty and some other devices nodes cannot be used on lustre
5664 Details : file's private_data field is used by device data and lustre
5665 values in there got lost. New field was added to struct file to
5666 store fs-specific private data.
5669 Frequency : when exporting Lustre via NFS
5671 Description: NFSD failed occasionally when looking up a path component
5672 Details : NFSD is looking up ".." which was broken in ext3 directories
5673 that had grown large enough to become hashed.
5676 Frequency : Clusters with multiple interfaces not on the same subnet
5678 Description: Nodes will repeatedly try to reconnect to an interface which it
5679 cannot reach and report an error to the log.
5680 Details : Extra peer list entries will be created by lconf with some peers
5681 unreachable. lconf now validates the peer before adding it.
5684 Frequency : Only if a default stripe is set on the filesystem root.
5686 Description: Setting a default stripe on the filesystem root prevented the
5687 filesystem from being remounted.
5688 Details : The client was sending extra request flags in the root getattr
5689 request and did not allocate a reply buffer for the dir EA.
5692 Frequency : occasional, higher if lots of files are accessed by one client
5693 Bugzilla : 6159, 6097
5694 Description: Client trips assertion regarding lsm mismatch/magic
5695 Details : While revalidating inodes the VFS looks up inodes with ifind()
5696 and in rare cases can find an inode that is being freed.
5697 The ll_test_inode() code will free the lsm during ifind()
5698 when it finds an existing inode and then the VFS later attaches
5699 this free lsm to a new inode.
5703 Bugzilla : 6422, 7030
5704 Description: MDS deadlock between mkdir and client eviction
5705 Details : Creating a new file via mkdir or mknod (starting a transaction
5706 and getting the ns lock) can deadlock with client eviction
5707 (gets ns lock and trying to finish a synchronous transaction).
5710 Frequency : occasional
5711 Description: While starting a server, the fsfilt_ext3 module could not be
5713 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5714 kernels. Previously, lconf would still use the ext3 name
5715 when trying to load modules. Now, it will correctly use
5716 ext3 on 2.4 and ldiskfs on 2.6.
5718 Severity : enhancement
5719 Description: The default stripe count has been changed to 1
5720 Details : The interpretation of the default stripe count (0, to lfs
5721 or lmc) has been changed to mean striping across a single
5722 OST, rather than all available. For general usage we have
5723 found a stripe count of 1 or 2 works best.
5725 Severity : enhancement
5726 Description: Add support for compiling against Cray portals.
5727 Details : Conditional compiling for some areas that are different
5731 Frequency : occasional
5732 Bugzilla : 6409, 6834
5733 Description: Creating files with an explicit stripe count may lead to
5734 a failed assertion on the MDS
5735 Details : If some OSTs are full or unavailable, creating files may
5736 trigger a failed assertion on the MDS. Now, Lustre will
5737 try to use other servers or return an error to the
5741 Frequency : occasional
5743 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5744 caused lock timeout during readahead (since 1.4.2).
5745 Details : Processes doing readahead might match a lock that hasn't been
5746 granted yet if there are overlapping and conflicting lock
5747 requests. The readahead process waits on ungranted lock
5748 (original lock is CBPENDING), while OST waits for that process
5749 to cancel CBPENDING read lock and eventually evicts client.
5751 Severity : enhancement
5753 Description: Initial enabling of flock support for clients
5754 Details : Implements fcntl advisory locking and file status functions.
5755 This feature is provided as an optional mount flag (default
5756 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5757 locking are implemented yet, and those that are are not guaranteed
5758 to be completely correct in production environments.
5759 mount -t lustre -o [flock|noflock] ...
5762 Frequency : occasional
5764 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5765 assertion in the ext3 JBD (journaling) layer.
5766 Details : The b_committed_data struct is protected by the big kernel lock
5767 in 2.4 kernels, serializing journal_commit_transaction() and
5768 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5769 access to this struct. In 2.6 kernels there is finer grained
5770 locking to improve SMP performance of the JBD layer.
5774 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5779 Description: merge and cleanup kernel patches.
5782 -----------------------------------------------------------------------------
5784 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5786 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5787 versions of Lustre. Please update all servers and clients to
5788 version 1.4.2 or later at the same time. You must also run
5789 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5790 to update the configuration logs.
5792 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5793 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5794 - swab llog records of type '0' so we get proper header size/idx (5861)
5795 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5796 - fix rename of one directory over another leaking an inode (5953)
5797 - avoid SetPageDirty on 2.6 (5981)
5798 - don't re-add just-being-destroyed locks to the waiting list (5653)
5799 - when creating new directories, inherit the parent's custom
5800 striping settings if present parent (3048)
5801 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5802 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5803 - don't hold client locks on temporary worklist from l_lru (5666)
5804 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5805 - automatically evict dead clients (5921)
5806 - Update file size properly in create+truncate+fstat case (6196)
5807 - Do not unhash mountpoint dentries, do not allow removal of
5809 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5810 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5811 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5812 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5813 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5814 - Correctly handle reads of files with no objects (6243)
5815 - lctl recover will also mark a device active if deactivate used (5933)
5817 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5818 - allow --write-conf on an MDS with different nettype than client (5619)
5819 - don't write config llogs to MDS for mounts not from that MDS (5617)
5820 - lconf should create multiple TCP connections from a client (5201)
5821 - init scripts are now turned off by default; run chkconfig --on
5822 lustre and chkconfig --on lustrefs to use them
5823 - upcalls are no longer needed for clients to recover to failover
5825 - add --abort-recovery option to lconf to abort recovery on device
5827 - add support for an arbitrary number of OSTs (3026)
5828 - Quota support protocol changes.
5829 - forward compatibility changes to wire structs (6007)
5830 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5831 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5832 - improved "lustre status" script
5833 - initialize blocksize for non-regular files (6062)
5834 - added --disable-server and --disable-client configure options (5782)
5835 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5836 - Vanilla 2.4.29 support
5837 - increase maximum number of obd devices to 520 (6242)
5838 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5839 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5840 SLES 9 2.6.5 kernels
5841 - limit stripes per file to 160 (the maximum EA size) (6093)
5843 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5846 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5847 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5848 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5849 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5850 - small ext3 extents cleanups and fixes (5733)
5851 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5852 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5853 - avoid needless client->OST connect, fix handle mismatch (5317)
5854 - fix DLM error path that led to out-of-sync client, long delays (5779)
5855 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5856 - fix several locking issues related to i_size (5492,5624,5654,5672)
5857 - don't move pending lock onto export if it is already evicted (5683)
5858 - fix kernel oops when creating .foo in unlinked directory (5548)
5859 - fix deadlock in obdfilter statistics vs. object create (5811)
5860 - use time_{before,after} to avoid timer jiffies wrap (5882)
5861 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5862 - Do not leak request if server was not able to process it (5154)
5863 - If mds_open unable to find parent dir, make that negative lookup(5154)
5864 - don't create new directories with extent-mapping (5909, 5936)
5866 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5867 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5868 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5869 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5870 - support for RHEL4 kernel on i686 (5773)
5871 - provide error messages when incompatible logs are encountered (5898)
5873 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5874 * version 1.4.0.10 (1.4.1 release candidate 1)
5876 - don't keep a lock reference when lock is not granted (4238)
5877 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5878 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5879 - drop import inflight refcount on signal_completed_replay error (5255)
5880 - unlock page after async write error during send (3677)
5881 - handle missing objects in filter_preprw_read properly (5265)
5882 - no transno return for symlink open, don't save no-trasno open (3440)
5883 - don't try to complete elan receive that already failed (4012)
5884 - free RPC server reply state on error (5406)
5885 - clean up thread from ptlrpc_start_thread() on error (5160)
5886 - readahead could read extra page into cache that wasn't ejected (5388)
5887 - prevent races in class_attach/setup/cleanup/detach (5260)
5888 - don't dereference de->d_inode after l_dput of de (5458)
5889 - use "int" for stripe value returned from lock_to_stripe (5544)
5890 - mballoc allocation and error-checking fixes in 2.6 (5504)
5891 - block device patches to fix I/O request sizes in 2.6 (5482)
5892 - look up hostnames for IB nals (5602)
5893 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5894 - don't start multiple acceptors for the same port (5277)
5895 - fix incorrect LASSERT in mds_getattr_name (5635)
5896 - export a proc file for general "ping" checking (5628)
5897 - fix "lfs check" to not block when the MDS is down (5628)
5899 - service request history (4965)
5900 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5901 - create an "evict_client" /proc entry on OSTs, like the MDS has
5902 - fix mount usage message, return errors per mount(8) (5168)
5903 - change grep [] to grep "[]" in tests so they work in more UMLs
5904 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5905 - remove ancient LOV_MAGIC_V0 EA support (5047)
5906 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5907 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5908 - allow readahead allocations to fail when low on memory (5383)
5909 - mmap locking landed again, after considerable improvement (2828)
5910 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5912 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5915 - send OST transaction number in read/write reply to free req (4966)
5916 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5917 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5918 - the watchdog thread now runs as interruptible (5246)
5919 - flock/lockf fixes (but it's still disabled, pending 5135)
5920 - don't use EXT3 constants in llite code (5094)
5921 - memory shortage at startup could cause assertion (5176)
5923 - reorganization of lov code
5924 - single portals codebase
5926 - add extents/mballoc support (5025)
5927 - direct I/O reads in the obdfilter (4048)
5928 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5930 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5933 - send OST transaction number in read/write reply to free req (4966)
5934 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5935 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5936 - the watchdog thread now runs as interruptible (5246)
5937 - handle missing objects in filter_preprw_read properly (5265)
5938 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5939 - drop import inflight refcount on signal_completed_replay error (5255)
5940 - unlock page after async write error during send (3677)
5941 - return original error code on reconstructed replies (3761)
5942 - no transno return for symlink open, don't save no-trasno open (3440)
5944 - add pid to ldlm debugging output (4922)
5945 - bump the watchdog timeouts -- we can't handle 30sec yet
5946 - extra debugging for orphan dentry/inode bug (5259)
5948 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5951 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5952 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5953 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5954 - fix NULL dereference in /proc/sys/portals/routes (4827)
5955 - allow failed mdc_close() operations to be interrupted (4561)
5956 - stop precreate on OST before MDS would time out on it (4778)
5957 - don't send partial-page writes before EOF from client (4410)
5958 - discard client grant for sub-page writes on large-page clients (4520)
5959 - don't free dentries not owned by NFS code, check generation (4806)
5960 - fix lsm leak if mds_create_objects() fails (4801)
5961 - limit debug_daemon file size, always print CERROR messages (4789)
5962 - use transno after validating reply (3892)
5963 - process timed out requests if import state changes (3754)
5964 - update mtime on OST during writes, return in glimpse (4829)
5965 - add mkfsoptions to LDAP (4679)
5966 - use ->max_readahead method instead of zapping global ra (5039)
5967 - don't interrupt __l_wait_event() during strace
5969 - add software watchdogs to catch hung threads quickly (4941)
5970 - make lustrefs init script start after nfs is mounted
5971 - fix CWARN/ERROR duplication (4930)
5972 - return async write errors to application if possible (2248)
5973 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5974 - print NAL number in %x format (4645)
5975 - update barely-supported suse-2.4.21-171 series (4842)
5976 - support for sles 9 %post scripts
5977 - support for building 2.6 kernel-source packages
5978 - support for sles km_* packages
5980 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5983 - ignore -ENOENT errors in osc_destroy (3639)
5984 - notify osc create thread that OSC is being cleaned up (4600)
5985 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5986 - reconstruct ost_handle() like mds_handle() (4657)
5987 - create a new thread to do import eviction to avoid deadlock (3969)
5988 - let lconf resolve symlinked-to devices (4629)
5989 - don't unlink "objects" from directory with default EA (4554)
5990 - hold socknal file ref over connect in case target is down (4394)
5991 - allow more than 32000 subdirectories in a single directory (3244)
5992 - fix blocks count for O_DIRECT writes (3751)
5993 - OST returns ENOSPC from object create when no space left (4539)
5994 - don't send truncate RPC if file size isn't changing (4410)
5995 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5996 - bind to privileged port in socknal and tcpnal (3689)
5998 - rate limit CERROR/CWARN console message to avoid overload (4519)
5999 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6000 - basic mmap support (3918)
6001 - kernel patch series update from b1_4 (4711)
6003 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6006 - avoid crash during MDS cleanup with OST shut down (2775)
6007 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6008 - don't use bad inodes on the MDS (3744)
6009 - dynamic object preallocation to improve recovery speed (4236)
6010 - don't hold spinlock over lock dumping or change debug flags (4401)
6011 - don't zero obd_dev when it is force cleaned (3651)
6012 - print grants to console if they go negative (4431)
6013 - "lctl deactivate" will stop automatic recovery attempts (3406)
6014 - look for existing locks in ldlm_handle_enqueue() (3764)
6015 - don't resolve lock handle twice in recovery avoiding race (4401)
6016 - revalidate should check working dir is a directory (4134)
6018 - don't always mark "slow" obdfilter messages as errors (4418)
6020 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6023 - don't close LustreDB during write_conf until it is done (3860)
6024 - fix typo in lconf for_each_profile (3821)
6025 - allow dumping logs from multiple threads at one time (3820)
6026 - don't allow multiple threads in OSC recovery (3812)
6027 - fix debug_size parameters (3864)
6028 - fix mds_postrecov to initialize import for llog ctxt (3121)
6029 - replace config semaphore with spinlock (3306)
6030 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6031 - don't allow enqueue to complete on a destroyed export (3822)
6032 - down write_lock before checking llog header bitmap (3825)
6033 - recover from lock replay timeout (3764)
6034 - up llog sem before sending rpc (3652)
6035 - reduce ns lock hold times when setting kms (3267)
6036 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6037 - fix NULL deref and obd_dev leak on setup error (3312)
6038 - replace some LBUG about llog ops with error handling (3841)
6039 - don't match INVALID dentries from d_lookup and spin (3784)
6040 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6041 - fix invalid assertion in ptlrpc_set_wait (3880)
6043 - add libwrap support for the TCP acceptor (3996)
6044 - add /proc/sys/portals/routes for non-root route listing (3994)
6045 - allow setting MDS UUID in .xml (2580)
6046 - print the stack of a process that LBUGs (4228)
6048 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6051 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6052 - make sure to unset replay flag from failed open requests (3440)
6053 - if default stripe count is 0, use OST count for inode size (3636)
6054 - update parent mtime/ctime on client for create/unlink (2611)
6055 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6056 - free recovery state on server during a forced cleanup (3571)
6057 - unregister_reply for resent reqs (3063)
6058 - loop back devices mounting and status check on 2.6 (3563)
6059 - fix resource-creation race that can provoke i_size == 0 (3513)
6060 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6061 - more debugging for page-accounting assertion (3746)
6062 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6063 - avoid deadlock after precreation failure (3758)
6064 - fix race and lock order deadlock in orphan handling (3450, 3750)
6065 - add validity checks when grabbing inodes from l_ast_data (3599)
6067 - add /proc/.../recovery_status to obdfilter (3428)
6068 - lightweight CDEBUG infrastructure, debug daemon (3668)
6069 - change default OSC RPC parameters to be better on small clusters
6070 - turn off OST read cache for files smaller than 32MB
6071 - install man pages and include them in rpms (3100)
6072 - add new init script for (un)mounting lustre filesystems (2593)
6073 - run chkconfig in %post for init scripts (3701)
6074 - drop scimac NAL (unmaintained)
6076 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6079 - clean kiobufs before and after use (3485)
6080 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6081 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6082 - change init script to fail more clearly if not run as root (1528)
6083 - allow clients to reconnect during replay (1742)
6084 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6085 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6086 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6087 - fix cond_resched (3554)
6088 - only evict unfinished clients after recovery (3515)
6089 - allow bulk resend, prevent data loss (3570)
6090 - dynamic ptlrpc request buffer allocation (2102)
6091 - don't allow unlinking open directory if it isn't empty (2904)
6092 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6093 - remove extraneous obd dereference causing LASSERT failure (3334)
6094 - don't use get_cycles() when creating temp. files on the mds (3156)
6095 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6096 - handle EEXIST for set-stripe, set proper directory name (3336)
6098 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6099 - fix ksocknal_fmb_callback() error messages (2918)
6101 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6104 - don't copy lvb into (possibly NULL) reply on error (2983)
6105 - don't deref dentry after dput, don't free lvb on error (2922)
6106 - use the kms to determine writeback rpc length (2947)
6107 - increment oti_logcookies when osc is inactive (2948)
6108 - update client's i_blocks count via lvb messages (2543)
6109 - handle intent open/close of special files properly (1557)
6110 - mount MDS with errors=remount-ro, like obdfilter (2009)
6111 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6112 - don't use cancelling-locks' kms values (2947)
6113 - use highest lock extent for kms, not last one (2925)
6114 - don't dereference ERR_PTR() dentry in error handling path (3107)
6115 - fix thread race in portals_debug_dumplog() (3122)
6116 - create lprocfs device entries at setup instead of at attach (1519)
6117 - common AST error handler, don't evict client on completion race (3145)
6118 - zero nameidata in detach_mnt in 2.6 (3118)
6119 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6120 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6121 - initialize RPC timeout timer earlier for 2.6 (3219)
6122 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6123 - print nal/nid for unknown nid (3258)
6124 - additional checks for oscc recovery before doing precreate (3284)
6125 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6126 - don't crash in mdc_close for bad permissions on open (3285)
6127 - zero i_rdev for non-device files (3147)
6128 - clear page->private before handing to FS, better assertion (3119)
6129 - tune the read pipeline (3236)
6130 - fix incorrect decref of invalidated dentry (2350)
6131 - provide read-ahead stats and refine rpc in flight stats (3328)
6132 - don't hold journal transaction open across create RPC (3313)
6133 - update atime on MDS at close time (3265)
6134 - close LDAP connection when recovering to avoid server load (3315)
6135 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6136 - don't leak open file on MDS after open resend (3325)
6137 - serialize filter_precreate and filter_destroy_precreated (3329)
6138 - loop device shouldn't call sync_dev() for nul device (3092)
6139 - clear page cache after eviction (2766)
6140 - resynchronize MDS->OST in background (2824)
6141 - refuse to mount the same filesystem twice on same mountpoint (3394)
6142 - allow llmount to create routes for mounting behind routers (3320)
6143 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6144 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6145 - fix rare problem with rename on htree directories (3417)
6147 - allow default OST striping configuration per directory (1414)
6148 - fix compilation for qswnal for 2.6 kernels (3125)
6149 - increase maximum number of MDS request buffers for large systems
6150 - change liblustreapi to be useful for external progs like lfsck (3098)
6151 - increase local configuration timeout for slow disks (3353)
6152 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6154 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6157 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6158 - don't overwrite extent policy data in reply if lock was blocked (2901)
6159 - drop filter export grants atomically with removal from device (2663)
6160 - del obd_self_export from work_list in class_disconnect_exports (2908)
6161 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6162 - swab reply message in mdc_close, other PPC fixes (2464)
6163 - fix destroying of named logs (2325)
6164 - overwrite old logs when running lconf --write_conf (2264)
6165 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6166 - fix race in target_handle_connect (2898)
6167 - mds_reint_create() should take same inode create lock (2926)
6168 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6169 - don't close files for self_export to avoid uninitialized obd (2936)
6170 - allow MDS with the same name as client node (2939)
6171 - hold dentry reference for closed log files for unlink (2325)
6172 - reserve space for all logs during transactions (2059)
6173 - don't evict page beyond end of stripe extent (2925)
6174 - don't oops on a deleted current working directory (2399)
6175 - handle hard links to targets without a parent properly (2517)
6176 - don't dereference NULL lock when racing during eviction (2867)
6177 - don't grow lock extents when lots of conflicting locks (2919)
6179 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6182 - account for cache space usage on clients to avoid data loss (974)
6183 - lfsck support in lustre kernel code (2349)
6184 - reduce journal credits needed for BRW writes (2370)
6185 - orphan handling to avoid losing space on client/server crashes
6186 - ptlrpcd can be blocked, stopping ALL progress (2477)
6187 - use lock value blocks to assist in proper KMS, faster stat (1021)
6188 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6189 - recovery for initial connections (2355)
6190 - fixes for mds_cleanup_orphans (1934)
6191 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6192 - block all file creations until orphan recovery completes (1901)
6193 - client remove rq_connection from request struct (2423)
6194 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6195 - recovery timer race (2670)
6196 - mdc_close recovey bug (2532)
6197 - ptlrpc cleanup bug (2710)
6198 - mds timeout on local locks (2588)
6199 - namespace lock held during RPCs (2431)
6200 - handle interrupted sync write properly (2503)
6201 - don't try to handle a message that hasn't been replied to (2699)
6202 - client assert failure during cleanup after abort recovery (2701)
6203 - leak mdc device after failed mount (2712)
6204 - ptlrpc_check_set allows timedout requests to complete (2714)
6205 - wait for inflight reqs when ptlrpcd finishes (2710)
6206 - make sure unregistered services are removed from the srv_list
6207 - reset bulk XID's when resending them (caught by 1138 test)
6208 - unregister_bulk after timeout
6209 - fix lconf error (2694)
6210 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6211 - readahead locks pages, leaves pending causing memory pressure (2673)
6212 - increase OST request buffers to 4096 on large machines (2729)
6213 - fix up permission of existing directories in simple_mkdir (2661)
6214 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6215 - don't assign transno to errored transactions (2742)
6216 - don't delete objects on OST if given a bogus objid from MDS (2751)
6217 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6218 - if rq_no_resend, then timeout request after recovery (2432)
6219 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6220 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6221 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6222 - MDS recovery shouldn't delete objects that it has given out (2730)
6223 - if enqueue arrives after completion, don't clobber LVB (2819)
6224 - don't unlock pages twice when trigger_group_io returns error (2814)
6225 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6226 - don't write pages to disk if there was an error (1450)
6227 - don't ping imports that have recovery disabled (2676)
6228 - take buffered bytes into account when balancing socknal conn (2817)
6229 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6230 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6231 - remove little-endian swabbing of llog records (1987)
6232 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6233 - retry reposting request buffers if they fail (1191)
6234 - grow extent at grant time to avoid granting a revoked lock (2809)
6235 - lock revoke doesn't evict page if covered by a second lock (2765)
6236 - disable VM readahead to avoid reading outside lock extents (2805)
6238 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6239 - updated kernel patches for hp-2.4.20 kernel (2681)
6241 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6244 - fix truncated write corruption (2366)
6245 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6247 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6248 - ensure len(uuid) < 37 in lmc (1171)
6249 - fix ia64 OOPS in llog_test (2255)
6250 - zero end of page at obdfilter for partial page writes (2648)
6251 - don't leave stale dentries around after renames (bug 2428)
6252 - fix timeouts when evicting a client with a single lock held (2642)
6253 - set deadline for the initial HELLO message to drain (2634)
6254 - print out dotted-quad IP addresses in the socknal (2302)
6256 - additional debugging for MDS client eviction problem (2443)
6257 - fix mkfsoptions support for osts (2603, 2604)
6259 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6262 - add series for the vanilla 2.6.0 kernel
6263 - add series for the vanilla 2.4.24 kernel
6264 - add series for a cray x86/64 UL kernel drop
6265 - fix xattr patches for the vanilla 2.4.19 series
6267 - generate true UUIDs in lmc (1171)
6268 - have portals stack dumping break in UML (2466)
6269 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6270 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6271 - wake callback waiting threads on client eviction (2460)
6272 - Add --ptldebug and --subsystem to lmc (1719)
6273 - update assertion to allow safe interrupt allocation
6274 - set rq_no_resend for cancel requests (2432)
6275 - recalculate ptlrpcd timeout after resend (2494)
6276 - call vfs_rmdir when removing pending directories (2368)
6277 - fix renaming a file to itself (2429)
6278 - lmc creates a default one-stripe lov (2454)
6279 - expand procfs space to handle large clusters (2326)
6280 - increase UML stack to avoid overflow
6281 - update lconf's list of debug and subsystem masks
6282 - fix lfs find --obd (2510)
6283 - /proc tunable for disabling filter read caching (2591)
6284 - stop rpm packages from altering slapd.conf (2301)
6285 - disable nagle in the socknal under 0conf (2578)
6286 - choose mds inode size based on stripe count (2572)
6287 - fix kernel-source rpm problems (2516)
6289 - add --disable-doc to avoid pdf generation (2421)
6290 - update documentation, tests, type-os, comments
6291 - avoid format warnings on ia64
6292 - remove the TOE NAL
6293 - tiny code cleanups by removing unused fields
6295 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6298 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6299 - remove the most blatant lies from BUILDING file (2371)
6300 - change default debug level to reasonable production setting
6301 - reduce client side cache size to reduce cache flush time
6302 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6303 - make TCP zerocopy and pinger support enabled by default (2476)
6304 - sync writes completed after process exits caused crashes (2319)
6305 - maintain correct mount count on the MDS (2356)
6306 - backout 1557, because 2316 wasn't really fixed
6307 - better file I/O statistics gathering in /proc
6308 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6309 - another kernel patch to fix zero-copy TCP function export
6310 - don't take duplicate lock when processing re-sent getattr (2420)
6311 - lctl uses obd_self_export instead of creating new conn (2353)
6312 - MDS/OST recovery case which requires object creation asserted (2425)
6313 - move lfs from /usr/sbin to /usr/bin in packages
6314 - fix race between mds_client_add and mds_client_free (2417)
6315 - use kmalloc instead of slabs in portals (2430)
6316 - don't create duplicate records when a failover MDS is present (2442)
6317 - remove unnecessary mount age check (2332)
6318 - don't remove directory inodes from locks prematurely (2451)
6319 - don't break if MDS service name is the same as hostname (2103)
6320 - fix races in client write RPC generation when cache full (2482)
6322 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6325 - remove now-unused request->rq_obd (278)
6326 - if an allocation fails, print out how much memory we've used (1933)
6327 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6328 - add the "configurable stack size" patch to most series files (1256)
6329 - ability to write large log records, for 100+ OST configs (2306)
6330 - fix NULL deref when filter_prep fails (2314)
6331 - fix operator precedence error in filter_sync
6332 - dynamic allocation of socknal TX descriptors (2315)
6333 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6334 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6335 - ensure that CWARN messages in Portals always get to the syslog
6336 - __init/__exit are not for prototype decls (ldlm_init/exit)
6337 - x86-64 compile warning fixes
6338 - fix gateway LMC keyword conflict (2318)
6339 - fix MDS lock inversions in getattr/reint paths (1844)
6340 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6341 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6342 - DLM race condition prevented some lock evictions (2328)
6343 - ENOMEM detection and retry on socknal sends (2230)
6344 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6345 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6346 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6347 - better error messages when a client is rejected during recovery (1505)
6348 - avoid cancelling locks which were never granted, after failure (2330)
6349 - fix i_sem/journal inversion in mds_client_add (2333)
6350 - fix truncate/getattr lock cycle deadlock (2334)
6351 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6352 - fix two rare exit paths which could leak an l_lock() ref (2321)
6353 - fencepost error in MDS/OST orphan recovery (2226)
6354 - make log record alignment 8 bytes (1988)
6355 - lstripe now fails when requested offset > ost_count (2237)
6356 - ensure that all kernel series have a complete list.h (1607)
6357 - fix crashes in special-file operations (2316)
6358 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6360 - allow configurable automake binary, for testing new versions
6361 - small update to the lfs documentation
6363 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6365 * fix negative export reference count in fsfilt_sync (2312)
6367 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6368 * release candidate 0.9.1
6370 - orphans are moved into the PENDING directory for possible recovery
6371 - replayed opens now open by fid for orphan/rename safety (1042)
6372 - last close of an orphan inode generates a transno (683)
6373 - chdir() and mount() now pin the directory entry (1020)
6374 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6375 - discard very old requests without processing them (1502)
6376 - remove some common, well-understood CERRORs (1505)
6377 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6378 - clear "grant" flags in OST replies until OST grant code lands (1644)
6379 - fix read performance by not clobbering i_blksize on client (1598)
6380 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6381 - make LRU size a /proc tunable, clears locks when reduced (707)
6382 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6383 - server locks take references on exports now (1558)
6384 - build fixes for 2.4.20-rh trees (1663)
6385 - return an error from lov_create if all OSCs are inactive (1751)
6386 - fix import levels when a reconnect happens without a timeout (1597)
6387 - exit early from mds_open if we get a lookup error (1749)
6388 - partial page read at EOF wouldn't wait for disk before sending (1642)
6389 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6390 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6391 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6392 - write/truncate lock inversion (1639)
6393 - Don't auto-load obdclass, portals modules during cleanup (1495)
6394 - fix timestamps from jumping to "now" (1763)
6395 - extra journal assertions (1648)
6396 - add an extra multiunlink test (1771)
6397 - fix read_record/write_record API (1776)
6398 - fix leak of offset_extent, possible incorrect i_size later (1772)
6399 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6400 - Debugging for the kqswnal_get_idle_tx problems (1820)
6401 - Allow recovery to be attempted multiple times (1536)
6402 - Write out MDS last_rcvd file after it is first created (1600)
6403 - Fix tx_descriptor leak in failed transmit situations (1827)
6404 - ext3 journaling fixes for assertion failure after IO error (1871)
6405 - class_export_put() on freed export after completion AST error (1896)
6406 - Fix revalidate looping in VFS (1322)
6407 - Don't access a freed export during MDS_REINT timeout (1521)
6408 - Add open-unlink recovery support on the MDS (1673,1764)
6409 - Return an error if no MDS data was read from last_rcvd (1946)
6410 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6411 - Don't setup a disk device that doesn't match exported UUID (317)
6412 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6413 - avoid committing NULL handle in force close
6414 - local.sh is now a one-stripe LOV configuration
6415 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6416 - fix ext3 htree duplicate directory entry corruption (1516)
6417 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6418 - update to new LOV EA format (2097)
6419 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6421 2003-06-15 Phil Schwan <phil@clusterfs.com>
6424 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6425 - per-import/export recovery handling (958, 931, 959)
6426 - multiple last-rcvd slots, for serving multiple FSes (949)
6427 - connections are again shared between multiple imp/exports (963, 964)
6428 - "umount -f" would hang if any requests needed to be sent (393, 978)
6429 - avoid pinning large req buffer by copying for queued messages (989)
6430 - add "uuid" to "lctl device" command to help upcalls (991)
6431 - "open" RPCs with transnos would confuse recovery counters (1037)
6432 - do proper endian conversion of all wire messages (288, 340, 891)
6433 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6434 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6435 - fix ptlrpc_connection leak in target_handle_connect (1174)
6436 - fix import refcounting bug in OST and MDS cleanup (1134)
6437 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6438 - fix very unlikely obd_types race condition (501)
6439 - remove osc_open hack for echo_client (1187)
6440 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6441 - a failure in read_inode2 leads to deadlock (1139)
6442 - cancel ack-locks as soon as transaction is committed (1072)
6443 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6444 - make sure to commitrw after any preprw to avoid deadlock (1162)
6445 - failing to execute a file in a lustre FS would lock inode (1203)
6446 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6447 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6448 - obd_uuid2tgt would walk off the end of the list (1255)
6449 - on IA64 the osc would give portals incorrect bulk size (1258)
6450 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6451 - fix lock inversion caused by new llite matching code (1282)
6452 - limit the number of dirty pages on a client to 10MB (1286)
6453 - timed out locks were not being corrected cancelled (1289)
6454 - fix O_DIRECT above 4GB on IA-32 (1292)
6455 * major user-visible changes
6456 - fail out/fail over policy now controlled by the upcall (993)
6458 - add OBD_PING to check server availability and failure (954)
6459 - lustre messages are now sent in sending host order (288, 340, 891)
6460 - add eadatalen to MDS getattr reply (340)
6461 - OST read replies may contain second buffer, with per-page status (593)
6463 2003-03-11 Phil Schwan <phil@clusterfs.com>
6466 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6467 - failed open()s could cause deadlock; fixed (867, 869)
6468 - stop cancelling OST locks when files are closed (481)
6469 - overlapping XID spaces caused network corruption (851, 853)
6470 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6471 - setattr_raw added, to do single-RPC, server-side setattrs
6472 - lmc/lconf syntax change for OST UUIDs
6473 - fix crashy race condition between ptlrpc_free_req and osc_close
6474 - don't use request in mdc_enqueue if we hit a timeout (889)
6475 - don't set the inode i_size for regular files from the MDS (896)
6476 - handle out of order completion AST (842)
6477 - don't LBUG if a lock request times out after receiving AST (913)
6478 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6479 - if a bad lock AST arrives, send an error instead of dropping entirely
6480 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6481 - fix leak in bulk IO when only partially completed (899, 900, 926)
6482 - fix O_DIRECT for ia64 (55)
6483 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6484 - C-z after timeout could hang a process forever; fixed (977)
6486 - client-side I/O cache (678, 924, 929, 941, 970)
6488 - READPAGE and SETATTRs which don't take server-side locks get
6491 2003-02-11 Phil Schwan <phil@clusterfs.com>
6494 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6495 - Fix fsfilt_extN_readpage() to read a full page of directory
6496 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6497 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6498 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6499 - Most intent operations were removed, in favour of a new RPC mode
6500 that does a single RPC to the server and bypasses most of the VFS
6501 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6502 - Aggressively cancel local locks on DLM servers
6503 - mds_reint_unlink sends EA to the client if it's the last nlink.
6504 client uses that EA to unlink OST objects.
6505 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6506 - recursive symlinks were fixed (439)
6507 - fixed NULL deref in DEBUG_REQ
6508 - filter_update_lastobjid no longer calls sync, which annoyed extN
6509 - fixed multi-client small-writes to a single file problem (445)
6510 - fixed mtime updates during file writes (607)
6511 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6512 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6513 - fixed Linux OST size reporting problem (444, 656)
6514 - OST now updates object mtime with writes or setattr (607, 619)
6515 - client verifies file size before zeroing page past EOF (445)
6516 - OST now writes last allocated objid to disk with allocation (108)
6517 - LOV on echo now works (409)
6519 - mds_reint_unlink sends a new buffer, with the EA included. this
6520 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6521 set if a regular file was being unlinked, and it was the last link
6522 - use PtlGet from the target for bulk writes (315)
6523 - OST now updates object mtime with writes or setattr (607, 619)
6524 - LDLM now has a grant-time callback to revalidate locked items, if
6526 - Many MDS operations were reorganized to combat race conditions
6528 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6529 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6531 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6534 - Fully reactivate OST imports after reconnection (512, others)
6535 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6536 - More graceful error handling for truncating on dead OST (515)
6537 - Don't error out unless we're actually accessing dead stripes (474)
6538 - Fix garbage sizes when stripes are missing (410)
6539 - LRU counters were broken, causing constant lock purge (433, 432)
6540 - garbage on read from stripes with failed OSTs (441)
6541 - mark OSCs as active before reconnecting during recovery (438)
6542 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6543 - lfind did not preserve OST order in output (443)
6544 - symlinks cause hung clients, incorrect data (439)
6545 - stop dereferencing request after dropping refcount (457)
6546 - don't LASSERT(spin_is_locked) on non-SMP (455)
6547 - fixes for many rename() bugs
6548 - fstat didn't correctly synchronize attributes (399)
6549 - server must handle lock cancellation during blocking AST prep (487)
6550 - bulk descriptors were free()d too soon (511)
6551 - fix paths in lconf, which would load incorrect modules (451, 507)
6552 - fix confusing lconf 'host not found' error message (386)
6553 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6554 - fix race condition in mdc_blocking_ast() for inode access (526)
6555 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6556 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6557 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6558 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6559 - fix lstripe in several ways (kernel side) (527)
6560 - fix request leak in ldlm_cli_enqueue (262)
6561 - incorrect OSC was marked inactive after OST failure
6562 - call mds_fs_cleanup before unmounting filesystem (524)
6563 - fix races between taking ns_lock and ldlm_lock_change_resource
6564 - fix races updating LOV export open file list
6565 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6566 - fix recovery NULL deref in ldlm_cli_cancel_unused
6567 - fix some DLM races by using new hash table for lock handles (419)
6568 - permit the client to specify desired inodes, at replay
6569 - duplicate requests when we queue them for replay reintegration
6570 - fix last_rcvd offset calculation
6571 - sync after each recovered transaction, so we always make progress
6572 - never, not always, ERESTART requests without transnos
6573 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6575 - skip replay if the MDS says that the client is already connected
6576 - don't check for a recovery-enabled export to match lctl's UUID
6577 - don't INC_USE_COUNT for phantom exports
6578 - don't crash when cleaning up phantom exports (567)
6579 - don't double-finish or set replay data for errored mdc_open requests
6580 - abort requests when they time out, so we don't get old replies
6581 - send/receive replies for AST messages again
6582 - if the client says that it doesn't have the lock, cancel it on the
6584 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6585 mark it as destroyed, it will all work out when decref is called
6586 - fix module use counts (22, 581)
6588 - ASTs now expect a reply (server cancels lock on error reply)
6590 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6593 - fix many simultaneous client startup (392)
6594 - fix dentry->d_it clobbering
6595 - credentials weren't being shipped for readdir/getattr operations
6596 - remove invalid assertions triggered during some concurrent MD
6598 - proper Lustre versions added (336, 389)
6599 - fix memory leak for create error case (398)
6600 - fix LOV locking bug that would get cli/srv out of sync
6601 - fix echo client over LOV (409)
6602 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6603 - fix double-O_EXCL intent crash (424)
6604 - avoid sending multiple lock CANCELs (352)
6606 - MDS can do multi-client recovery (modulo bugs in new code)
6608 - many updates, edits, cleanups
6610 2002-11-18 Phil Schwan <phil@clusterfs.com>
6613 - fix null d_it dereference (346)
6614 - fix full OST/dbench hang (333)
6615 - fix permission problem with file removal (286)
6616 - fix removal of OSCs from LOV when they fail
6617 - fix NULL deref during bulk timeout (214)
6618 - fix problems related to multiple filesystems on one MDS (241)
6619 - fixed serious subtle metadata locking bugs
6620 - free locks on clients when inodes are removed due to memory
6622 - fix inode pointer in lock data (285)
6623 - partial support for multiple MDS on a single host (241)
6624 - data locks weren't cancelled at clear_inode time (290, 311)
6625 - intent locks could lead to unbounded lock growth (205)
6626 - added a maximum lock count, an LRU list, and a flusher
6627 - fix multiple rename (365)
6628 - properly abstracted the echo client
6629 - OSC locked 1 byte too many; fixed
6630 - rewrote brw callback code:
6631 - fixed recovery bugs related to LOVs (306)
6632 - fixed too-many-pages-in-one-write crash (191)
6633 - fixed (again) crash in sync_io_timeout (214)
6634 - probably fixed callback-related race (385)
6636 - Add capability to MDS protocol
6637 - LDLM cancellations and callbacks on different portals
6639 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6642 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6643 - defer open object destruction to close time (601981, 138)
6644 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6645 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6646 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6647 - don't call obd_setattr in ll_file_release for destroyed objects
6648 * protocol change to lustre_msg: move |version| and add |flags|
6649 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6650 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6651 * added replay of create, unlink, link and rename operations during
6652 MDS failover; recovery should be much more robust now
6653 * remove failed OSCs from LOVs (only lov_create uses this so far)
6654 * the lustre-HOWTO was brought (more) up to date (582544)
6656 2002-10-23 Phil Schwan <phil@clusterfs.com>
6659 - in-use dentries weren't being reused properly (617851)
6660 - prevent multiple LDLM setup (599178)
6661 - fix LOV size calculations for truncate (617853)
6662 - fix client handling of MDS intent errors (POSIX)
6663 - fix permission bug in lovstripe.c test (624321)
6664 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6665 - truncate past end of file could corrupt data
6666 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6667 - a race in recovery could return ETIMEDOUT to apps (623947)
6668 - building outside the source directory was fixed
6669 * the lustre-HOWTO was brought (more) up to date (582544)
6670 * major progress was made on recovery functionality
6672 2002-10-10 Phil Schwan <phil@clusterfs.com>
6675 - recovery deadlock fix
6676 - rm -rf causes LBUG fix (617817)
6677 - file open by multiple tasks fix (618962)
6678 - directory permissions bugs (602707 and 620007)
6679 - journal_stop fixed with locking (611313)
6680 - O_APPEND failures resolved (618273, perhaps 614459)
6681 - lconf PATH fix (619770)
6682 - IA64 build fix (621450)
6683 - RPC buffer sizes scale with amount of memory
6685 2002-10-01 Phil Schwan <phil@clusterfs.com>
6688 - locks would be cancelled without throwing away data pages,
6689 resulting in inconsistent data (605627)
6690 - inode attributes were not always being refreshed (605627, 612449)
6691 - lconf now continues to cleanup after lctl reports an error
6692 - MDS now enforces user permissions (602707)
6693 - lprocfs cleanup fixed, but not yet enabled (614157)
6694 - fixed infinite server hang, should a client not respond to an AST
6695 - avoid going into recovery if user calls readlink() with a buffer
6696 that's too small (613941)
6697 - AST RPCs no longer require replies (614867) -- this may be changed
6698 - don't crash server if client sends an IOV that's too big (611336)
6699 - fixed lock conversion deadlock (611892)
6700 - fixed the following of symlinks (614622)
6701 * recovery: the server can remove locks from a client that dies, other
6702 clients can make progress
6703 * more extN patch fixes
6704 * compile-time configurable ptlrpc buffer allocations
6706 - collaborative read cache document
6707 - Lustre Lite Performance CDR document-in-progress
6709 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6712 - fix typo in patch-2.4.18
6714 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6717 - clear ptlrpc request each time in handle_incoming_request()
6718 - unlink of files now destroys the object on the OST
6720 2002-09-19 Peter Braam <braam@clusterfs.com>
6722 * add hard link support
6723 * change obdfile creation method
6724 * kernel patch changed
6726 2002-09-19 Peter Braam <braam@clusterfs.com>
6729 - stack overflow bug in extN fixed
6731 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6733 * documentation updates
6734 - add man pages for config tools
6735 - update tests/README to describe testing with new config tools
6736 - finish metadata API descriptions
6737 * bug fixes and cleanups
6738 - statfs workaround for 16TB limit
6739 - LOV stripe allocation improved, can stripe on subset of OSTs
6740 - LOV file size/IO offset was wrong for files > 4GB in size
6741 - object EA data was being dropped, caused files to be unreadable
6742 - memory overflow with non-LOV OST caused memory corruption
6743 - fixed regression tests to work with new config tools, obdfilter
6744 - fixed bug when directory size became larger than 1 block
6745 - fixed bug (for single client case) when PWD was deleted
6746 - invalidate local directory pages when doing intent-based ops
6747 - avoid LDLM oops when lock callback contained bad data
6749 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6751 * documentation updates
6752 * bug fixes and cleanups
6753 - configuration tools
6756 - 64-bit compile warnings
6757 - 64-bit internal statfs data
6759 * test_brw on persistent OST devices
6761 * lprocfs (disabled)
6763 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6765 * documentation updates
6766 * bug fixes and cleanups
6767 * configuration tools
6769 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6772 * many small fixes to 0_5_4
6773 * io/network handling
6774 * thinkos in MDS operations
6776 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6779 * crucial basic fixes to 0.5.3
6780 * IOR, Iozone work over Elan
6783 2002-08-07 Phil Schwan <phil@clusterfs.com>
6784 * version 0_5_3, our first alpha
6785 * we use the new Portals iovs
6786 * documentation updates
6787 * bug fixes and cleanups
6788 * small changes in the DLM wire protocol
6790 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6791 * version 0_5_1 with some initial stability,
6792 * locking on MD and file I/O.
6793 * documentation updates
6794 * several bug fixes since 0.5.0
6795 * small changes in wire protocol
6797 2002-07-18 Phil Schwan <phil@clusterfs.com>
6799 * delivered as Lustre Light Alpha
6800 * fixed a crash after handling invalid MDS requests
6801 * fixed directory pages for architectures with non-4k pages sizes
6803 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6804 * release version v0_4_4
6805 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6807 * Fixes a number of interruption problems with OST operations.
6808 * Update documentation for portals header changes
6809 * Move all wire protocol structs/defines to lustre_idl.h
6810 * Fixes symlink length bug.
6811 * Add tcpdump to repository.
6813 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6814 * release version v0_4_3
6815 * Fixes statfs for inodes on extN.
6816 * Fixes bug in runtests which would delete /etc/hosts.
6817 * Use 64-bit object IDs wherever possible (not into VFS though)
6818 Remove ost_get_info, which is unused by lustre, and out of date.
6820 2002-07-03 Peter Braam <braam@clusterfs.com>
6821 * release version v0_4_2 Fixes a lookup error (type not passed)
6822 * move forward to head of Portals
6823 * move forward to latest Lustre kernel
6825 2002-06-25 Peter Braam <braam@clusterfs.com>
6826 * release version v0_4_1. Hopefully stable on single node use.