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 3.10.0-1160.45.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.22.1.el8 (RHEL8.3)
16 4.18.0-305.19.1.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 3.10.0-1160.45.1.el7 (RHEL7.9)
30 4.12.14-95.48 (SLES12 SP4)
31 4.12.14-122.91 (SLES12 SP5)
32 4.12.14-197.75 (SLES15 SP1)
33 5.3.18-24.61 (SLES15 SP2)
34 5.3.18-59.27 (SLES15 SP3)
35 4.15.0-48 (Ubuntu 18.04)
36 * Other clients known to build on these kernels at some point (others may also work):
37 3.10.0-862.14.4.el7 (RHEL7.5)
38 3.10.0-957.27.2.el7 (RHEL7.6)
39 4.14.0-49.13.1.el7a (RHEL7.5)
40 4.14.0-115.2.2.el7a (RHEL7.6)
41 4.18.0-80.11.2.el8 (RHEL8.0)
42 4.18.0-147.8.1.el8 (RHEL8.1)
43 4.18.0-193.28.1.el8 (RHEL8.2)
44 4.18.0-240.22.1.el8 (RHEL8.3)
45 4.18.0-305.19.1.el8 (RHEL8.4)
46 4.18.0-348.2.1.el8 (RHEL8.5)
47 4.4.120-92.70 (SLES12 SP2)
48 4.4.180-94.100 (SLES12 SP3)
49 4.4.0-131 (Ubuntu 16.04)
50 5.4.0-37 (Ubuntu 20.04)
51 5.8.0-53 (Ubuntu 20.04.2 HWE)
52 5.11.0-31 (Ubuntu 20.04.3 HWE)
53 5.11.0 (vanilla kernel.org)
54 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
55 * Recommended ZFS version: 2.0.0
56 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
57 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
58 stack overflow. For more information, please refer to bugzilla 17630.
59 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
60 the re-exporting nodes to be mounted with "32bitapi" mount option
62 --------------------------------------------------------------------------------
66 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
67 for currently supported client and server kernel versions.
68 * Server primary kernels built and tested during release cycle:
69 3.10.0-1062.18.1.el7 (RHEL7.7)
70 3.10.0-1127.19.1.el7 (RHEL7.8)
71 3.10.0-1160.6.1.el7 (RHEL7.9)
72 * Other server kernels known to build and work at some point (others may also work):
73 3.10.0-862.14.4.el7 (RHEL7.5)
74 3.10.0-957.27.2.el7 (RHEL7.6)
75 4.18.0-80.11.2.el8 (RHEL8.0)
76 4.18.0-147.8.1.el8 (RHEL8.1)
77 4.18.0-193.28.1.el8 (RHEL8.2)
78 4.18.0-240.1.1.el8 (RHEL8.3)
79 4.4.120-92.70 (SLES12 SP2)
80 4.4.180-94.100 (SLES12 SP3)
81 4.4.0-131 (Ubuntu 16.04)
82 4.15.0-32 (Ubuntu 18.04)
83 5.4.0-48 (Ubuntu 20.04)
84 vanilla linux 5.4.0 (ZFS + ldiskfs)
85 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
86 * Client primary kernels built and tested during release cycle:
87 3.10.0-1062.18.1.el7 (RHEL7.7)
88 3.10.0-1127.19.1.el7 (RHEL7.8)
89 3.10.0-1160.6.1.el7 (RHEL7.9)
90 4.12.14-95.48 (SLES12 SP4)
91 4.12.14-122.20 (SLES12 SP5)
92 4.12.14-197.75 (SLES15 SP1)
93 5.3.18-24.24 (SLES15 SP2)
94 4.15.0-48 (Ubuntu 18.04)
95 * Other clients known to build on these kernels at some point (others may also work):
96 3.10.0-862.14.4.el7 (RHEL7.5)
97 3.10.0-957.27.2.el7 (RHEL7.6)
98 4.14.0-49.13.1.el7a (RHEL7.5)
99 4.14.0-115.2.2.el7a (RHEL7.6)
100 4.18.0-80.11.2.el8 (RHEL8.0)
101 4.18.0-147.8.1.el8 (RHEL8.1)
102 4.18.0-193.28.1.el8 (RHEL8.2)
103 4.18.0-240.1.1.el8 (RHEL8.3)
104 4.4.120-92.70 (SLES12 SP2)
105 4.4.180-94.100 (SLES12 SP3)
106 4.4.0-131 (Ubuntu 16.04)
107 5.4.0-37 (Ubuntu 20.04)
108 5.4.0 (vanilla kernel.org)
109 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
110 * Recommended ZFS version: 2.0.0
111 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
112 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
113 stack overflow. For more information, please refer to bugzilla 17630.
114 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
115 the re-exporting nodes to be mounted with "32bitapi" mount option
117 --------------------------------------------------------------------------------
121 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
122 for currently supported client and server kernel versions.
123 * Server primary kernels built and tested during release cycle:
124 3.10.0-1062.1.1.el7 (RHEL7.7)
125 * Other server kernels known to build and work at some point (others may also work):
126 3.10.0-862.14.4.el7 (RHEL7.5)
127 3.10.0-957.27.2.el7 (RHEL7.6)
128 4.18.0-80.7.1.el8 (RHEL8.0)
129 4.4.120-92.70 (SLES12 SP2)
130 4.4.180-94.100 (SLES12 SP3)
131 4.4.0-131 (Ubuntu 16.04)
132 4.15.0-32 (Ubuntu 18.04)
133 vanilla linux 4.6.7 (ZFS only)
134 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
135 * Client primary unpatched kernels built and tested during release cycle:
136 3.10.0-1062.1.1.el7 (RHEL7.7)
137 4.12.14-95.29 (SLES12 SP4)
138 4.15.0-48 (Ubuntu 18.04)
139 * Other clients known to build on unpatched kernels at some point (others may also work):
140 3.10.0-862.14.4.el7 (RHEL7.5)
141 3.10.0-957.27.2.el7 (RHEL7.6)
142 4.14.0-49.13.1.el7a (RHEL7.5)
143 4.14.0-115.2.2.el7a (RHEL7.6)
144 4.18.0-80.7.1.el8 (RHEL8.0)
145 4.4.120-92.70 (SLES12 SP2)
146 4.4.180-94.100 (SLES12 SP3)
147 4.4.0-131 (Ubuntu 16.04)
148 vanilla linux 4.15.0 (ZFS only)
149 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
150 * Recommended ZFS / SPL version: 0.7.13
151 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
152 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
153 stack overflow. For more information, please refer to bugzilla 17630.
154 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
155 the re-exporting nodes to be mounted with "32bitapi" mount option
156 * Note for Ubuntu 14.04.5:
157 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
158 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
160 --------------------------------------------------------------------------------
164 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
165 for currently supported client and server kernel versions.
166 * Server known to build on the following kernels (others may also work):
167 2.6.32-696.18.7.el6 (RHEL6.9)
168 3.10.0-862.14.4.el7 (RHEL7.5)
169 3.10.0-957.el7 (RHEL7.6)
170 4.4.120-92.70 (SLES12 SP2)
171 4.4.162-94.69 (SLES12 SP3)
172 4.4.0-131 (Ubuntu 16.04)
173 4.15.0-32 (Ubuntu 18.04)
174 vanilla linux 4.6.7 (ZFS only)
175 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
176 * Client known to build on unpatched kernels (others may also work):
177 2.6.32-696.18.7.el6 (RHEL6.9)
178 2.6.32-754.6.3.el6 (RHEL6.10)
179 3.10.0-862.14.4.el7 (RHEL7.5)
180 3.10.0-957.el7 (RHEL7.6)
181 4.14.0-49.13.1.el7a (RHEL7.5)
182 4.14.0-115.2.2.el7a (RHEL7.6)
183 4.4.120-92.70 (SLES12 SP2)
184 4.4.162-94.69 (SLES12 SP3)
185 4.4.0-131 (Ubuntu 16.04)
186 4.15.0-32 (Ubuntu 18.04)
187 vanilla linux 4.15.0 (ZFS only)
188 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
189 * Recommended ZFS / SPL version: 0.7.9
190 * Tested with ZFS / SPL version: 0.7.9
191 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
192 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
193 stack overflow. For more information, please refer to bugzilla 17630.
194 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
195 the re-exporting nodes to be mounted with "32bitapi" mount option
196 * Note for Ubuntu 14.04.5:
197 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
198 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
200 --------------------------------------------------------------------------------
202 2018-04-03 Intel Corporation
204 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
205 for currently supported client and server kernel versions.
206 * Server known to build on patched kernels:
207 2.6.32-431.29.2.el6 (RHEL6.5)
208 2.6.32-504.30.3.el6 (RHEL6.6)
209 2.6.32-573.26.1.el6 (RHEL6.7)
210 2.6.32-642.15.1.el6 (RHEL6.8)
211 2.6.32-696.18.7.el6 (RHEL6.9)
212 3.10.0-693.21.1.el7 (RHEL7.4)
213 3.0.101-0.47.71 (SLES11 SP3)
214 3.0.101-107 (SLES11 SP4)
215 3.12.74-60.64.40 (SLES12 SP1)
216 4.4.114-92.64 (SLES12 SP2)
217 4.4.114-94.11 (SLES12 SP3)
218 3.13.0-101 (Ubuntu 14.04, ZFS only)
219 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
220 4.4.0-116 (Ubuntu 16.04)
221 vanilla linux 4.6.7 (ZFS only)
222 * Client known to build on unpatched kernels:
223 2.6.32-431.29.2.el6 (RHEL6.5)
224 2.6.32-504.30.3.el6 (RHEL6.6)
225 2.6.32-573.26.1.el6 (RHEL6.7)
226 2.6.32-642.15.1.el6 (RHEL6.8)
227 2.6.32-696.18.7.el6 (RHEL6.9)
228 3.10.0-693.21.1.el7 (RHEL7.4)
229 3.0.101-0.47.71 (SLES11 SP3)
230 3.0.101-107 (SLES11 SP4)
231 3.12.74-60.64.40 (SLES12 SP1)
232 4.4.114-92.64 (SLES12 SP2)
233 4.4.114-94.11 (SLES12 SP3)
234 3.13.0-101 (Ubuntu 14.04)
235 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
236 4.4.0-116 (Ubuntu 16.04)
238 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
239 * Recommended ZFS / SPL version: 0.7.6
240 * Tested with ZFS / SPL version: 0.7.6
241 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
242 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
243 stack overflow. For more information, please refer to bugzilla 17630.
244 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
245 the re-exporting nodes to be mounted with "32bitapi" mount option
246 * Note for Ubuntu 14.04.5:
247 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
248 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
250 --------------------------------------------------------------------------------
252 2017-07-05 Intel Corporation
254 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
255 for currently supported client and server kernel versions.
256 * Server known to build on patched kernels:
257 2.6.32-431.29.2.el6 (RHEL6.5)
258 2.6.32-504.30.3.el6 (RHEL6.6)
259 2.6.32-573.26.1.el6 (RHEL6.7)
260 2.6.32-642.15.1.el6 (RHEL6.8)
261 2.6.32-696.3.1.el6 (RHEL6.9)
262 3.10.0-514.21.1.el7 (RHEL7.3)
263 3.0.101-0.47.71 (SLES11 SP3)
264 3.0.101-100 (SLES11 SP4)
265 3.12.74-60.64.40 (SLES12 SP1)
266 4.4.59-92.17 (SLES12 SP2)
267 vanilla linux 4.6.7 (ZFS only)
268 * Client known to build on unpatched kernels:
269 2.6.32-431.29.2.el6 (RHEL6.5)
270 2.6.32-504.30.3.el6 (RHEL6.6)
271 2.6.32-573.26.1.el6 (RHEL6.7)
272 2.6.32-642.15.1.el6 (RHEL6.8)
273 2.6.32-696.3.1.el6 (RHEL6.9)
274 3.10.0-514.21.1.el7 (RHEL7.3)
275 3.0.101-0.47.71 (SLES11 SP3)
276 3.0.101-100 (SLES11 SP4)
277 3.12.74-60.64.40 (SLES12 SP1)
278 4.4.59-92.17 (SLES12 SP2)
280 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
281 * Recommended ZFS / SPL version: 0.6.5.9
282 * Tested with ZFS / SPL version: 0.6.5.9
283 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
284 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
285 stack overflow. For more information, please refer to bugzilla 17630.
286 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
287 the re-exporting nodes to be mounted with "32bitapi" mount option
289 --------------------------------------------------------------------------------
291 12-07-2016 Intel Corporation
293 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
294 for currently supported client and server kernel versions.
295 * Server known to build on patched kernels:
296 2.6.32-431.29.2.el6 (RHEL6.5)
297 2.6.32-504.30.3.el6 (RHEL6.6)
298 2.6.32-573.26.1.el6 (RHEL6.7)
299 2.6.32-642.11.1.el6 (RHEL6.8)
300 3.10.0-514.el7 (RHEL7.3)
301 3.0.101-0.47.71 (SLES11 SP3)
302 3.0.101-80 (SLES11 SP4)
303 3.12.62-60.64.8 (SLES12 SP1)
304 vanilla linux 4.5.7 (ZFS only)
305 * Client known to build on unpatched kernels:
306 2.6.32-431.29.2.el6 (RHEL6.5)
307 2.6.32-504.30.3.el6 (RHEL6.6)
308 2.6.32-573.26.1.el6 (RHEL6.7)
309 2.6.32-642.11.1.el6 (RHEL6.8)
310 3.10.0-514.el7 (RHEL7.3)
311 3.0.101-0.47.71 (SLES11 SP3)
312 3.0.101-80 (SLES11 SP4)
313 3.12.62-60.64.8 (SLES12 SP1)
315 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
316 * Recommended ZFS / SPL version: 0.6.5.8
317 * Tested with ZFS / SPL version: 0.6.5.8
318 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
319 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
320 stack overflow. For more information, please refer to bugzilla 17630.
321 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
322 the re-exporting nodes to be mounted with "32bitapi" mount option
324 --------------------------------------------------------------------------------
326 02-29-2016 Intel Corporation
328 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
329 for currently supported client and server kernel versions.
330 * Server known to build on patched kernels:
331 2.6.32-431.29.2.el6 (RHEL6.5)
332 2.6.32-504.30.3.el6 (RHEL6.6)
333 2.6.32-573.12.1.el6 (RHEL6.7)
334 3.10.0-327.3.1.el7 (RHEL7.2)
335 3.0.101-0.47.71 (SLES11 SP3)
336 3.0.101-68 (SLES11 SP4)
337 vanilla linux 4.2.1 (ZFS only)
338 * Client known to build on unpatched kernels:
339 2.6.32-431.29.2.el6 (RHEL6.5)
340 2.6.32-504.30.3.el6 (RHEL6.6)
341 2.6.32-573.12.1.el6 (RHEL6.7)
342 3.10.0-327.3.1.el7 (RHEL7.2)
343 3.0.101-0.47.71 (SLES11 SP3)
344 3.0.101-68 (SLES11 SP4)
347 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
348 * Recommended ZFS / SPL version: 0.6.4.2
349 * Tested with ZFS / SPL version: 0.6.4.2
350 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
351 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
352 stack overflow. For more information, please refer to bugzilla 17630.
353 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
354 the re-exporting nodes to be mounted with "32bitapi" mount option
356 --------------------------------------------------------------------------------
358 03-10-2015 Intel Corporation
360 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
361 for currently supported client and server kernel versions.
362 * Server known to build on patched kernels:
363 2.6.32-431.29.2.el6 (RHEL6.5)
364 2.6.32-504.8.1.el6 (RHEL6.6)
365 3.0.101-0.46 (SLES11 SP3)
366 * Client known to build on unpatched kernels:
367 2.6.32-431.29.2.el6 (RHEL6.5)
368 2.6.32-504.8.1.el6 (RHEL6.6)
369 3.10.0-123.20.1.el7 (RHEL7)
370 3.0.101-0.46 (SLES11 SP3)
371 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
372 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
373 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
374 stack overflow. For more information, please refer to bugzilla 17630.
375 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
376 the re-exporting nodes to be mounted with "32bitapi" mount option
378 Severity : enhancement
380 Description: control OST-index in IDIF via ROCOMPAT flag.
381 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
382 last_rcvd file. For new formatted OST device, it will be auto set;
383 for the case of upgrading from old OST device, you can enable it
384 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
385 enabled, for new created OST-object, its IDIF-in-LMA will contain
386 the OST-index; for the existing OST-object, the OSD will convert
387 old format IDIF as new format IDIF with OST-index stored in the
388 LMA EA when accessing such OST-object or via OI scrub. Once such
389 flag is enabled, it cannot be reverted back, so the system cannot
390 be downgraded to the orignal incompatible version.
392 --------------------------------------------------------------------------------
394 07-30-2014 Intel Corporation
396 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
397 for currently supported client and server kernel versions.
398 * Server known to build on patched kernels:
399 2.6.32-431.20.3.el6 (RHEL6.5)
400 3.0.101-0.31 (SLES11 SP3)
401 * Client known to build on unpatched kernels:
402 2.6.32-431.20.3.el6 (RHEL6.5)
403 3.0.101-0.31 (SLES11 SP3)
406 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
407 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
408 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
409 stack overflow. For more information, please refer to bugzilla 17630.
410 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
411 the re-exporting nodes to be mounted with "32bitapi" mount option
413 --------------------------------------------------------------------------------
415 10-31-2013 Intel Corporation
417 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
418 * Server builds for kernels:
419 2.6.32-358.18.1.el6 (RHEL6.4)
420 * Client builds for unpatched kernels:
421 2.6.32-358.18.1.el6 (RHEL6.4)
422 3.0.92-0.8 (SLES11 SP3)
424 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
425 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
426 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
427 stack overflow. For more information, please refer to bugzilla 17630.
428 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
429 the re-exporting nodes to be mounted with "32bitapi" mount option
431 --------------------------------------------------------------------------------
433 05-30-2013 Intel Corporation
435 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
436 * Server builds for kernels:
437 2.6.32-358.6.2.el6 (RHEL6.4)
438 * Client builds for unpatched kernels:
439 2.6.32-358.6.2.el6 (RHEL6.4)
440 3.0.74-0.6.6 (SLES11 SP2)
442 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
443 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
444 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
445 stack overflow. For more information, please refer to bugzilla 17630.
446 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
447 the re-exporting nodes to be mounted with "32bitapi" mount option
449 --------------------------------------------------------------------------------
451 9-30-2012 Whamcloud, Inc.
453 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
454 * Server support for kernels:
455 2.6.32-279.5.1.el6 (RHEL6)
456 * Client support for unpatched kernels:
457 2.6.32-279.5.1.el6 (RHEL6)
459 2.6.18-238.19.1.el5 (RHEL5)
460 2.6.16.60-0.69.1 (SLES10)
461 * Recommended e2fsprogs version: 1.41.90.wc4
462 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
463 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
464 stack overflow. For more information, please refer to bugzilla 17630.
465 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
466 the re-exporting nodes to be mounted with "32bitapi" mount option
468 Severity : enhancement
470 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
472 Severity : enhancement
474 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
476 Severity : enhancement
478 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
480 Severity : enhancement
482 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
484 Severity : enhancement
486 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
488 --------------------------------------------------------------------------------
490 04-30-2012 Whamcloud, Inc.
492 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
493 * Server support for kernels:
494 2.6.32-220.4.2.el6 (RHEL6)
495 * Client support for unpatched kernels:
496 2.6.32-220.4.2.el6 (RHEL6)
498 2.6.18-238.19.1.el5 (RHEL5)
499 2.6.16.60-0.69.1 (SLES10)
500 * Recommended e2fsprogs version: 1.41.90.wc4
501 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
502 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
503 stack overflow. For more information, please refer to bugzilla 17630.
504 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
505 the re-exporting nodes to be mounted with "32bitapi" mount option
507 Severity : enhancement
509 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
511 Severity : enhancement
513 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
515 Severity : enhancement
517 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
519 Severity : enhancement
521 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
523 --------------------------------------------------------------------------------
525 09-30-2011 Whamcloud, Inc.
527 * Server support for kernels:
528 2.6.18-238.19.1.el5 (RHEL5)
529 2.6.32-131.6.1.el6 (RHEL6)
530 * Client support for unpatched kernels:
531 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
532 2.6.18-238.19.1.el5 (RHEL5)
533 2.6.32-131.6.1.el6 (RHEL6)
534 2.6.29.4-167.fc11 (FC11)
535 2.6.16.60-0.69.1 (SLES10)
537 * Recommended e2fsprogs version: 1.41.90.wc3
538 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
539 Lustre filesystem with 4K stack may cause a stack overflow. For more
540 information, please refer to bugzilla 17630.
541 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
542 mounted with "32bitapi" mount option
544 Severity : enhancement
545 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
547 Severity : enhancement
548 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
550 Severity : enhancement
551 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
554 Frequency : only with SLES10
556 Description: Use OFED "KMP" provided by Novell
557 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
558 Lustre is now built against this package. That means you need to
559 install the ofed-kmp package from Novell for the patchless client.
560 Note that the ofed-kmp that Novell ships may not exactly match the
561 kernel version but should still be compatible.
563 Severity : enhancement
565 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
566 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
568 --------------------------------------------------------------------------------
570 2010-07-15 Oracle, Inc.
572 * Server support for kernels:
573 2.6.18-164.11.1.el5 (RHEL 5)
574 2.6.18-164.11.1.0.1.el5 (OEL 5)
575 * Client support for unpatched kernels:
576 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
577 2.6.18-164.11.1.el5 (RHEL 5),
578 2.6.18-164.11.1.0.1.el5 (OEL 5)
579 2.6.16.60-0.42.8 (SLES 10),
581 2.6.29.4-167.fc11 (FC11)
582 * Recommended e2fsprogs version: 1.41.10-sun2
584 Severity : enhancement
586 Description: Distribute lfs_migrate script with lustre
587 Details : The lfs_migrate mentioned in the manual is now included in the
588 lustre tree in lustre/scripts and will distributed in lustre RPMs.
590 Severity : enhancement
592 Description: Failover nid list added to import proc file.
594 Severity : enhancement
596 Description: "som_preview" mount option required to use SOM
597 Details : The clients will need to use som_preview mount option
601 Description: Lustre does not do 1MB IOs to HW RAID
602 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
604 Severity : enhancement
606 Description: lfs quota output is not very convenient for awk/sed-parsing
607 Details : Some positions in lfs quota output table could be empty or
608 non-empty which made it hard to parse it with scripts, now
609 a dash is put instead of space where there is not supposed
614 Description: Update to OFED 1.5.1
615 Details : OFED 1.5.1 will be used in environments that don't use
616 vendor supplied OFED.
620 Description: Remove lnet. prefix from lctl params display
621 Details : Also renamed lnet.memused to lnet_memused to
626 Description: Use conf_param -d to remove permanant settings.
630 Description: Add list_param -R
631 Details : lctl list_param now supports recursive mode.
634 Bugzilla : 21829, 22312
635 Description: Fix llobdstat.
636 Details : 1. Now updating the write_bytes counter
637 2. Port script fixes from 1.8.
641 Description: decrease the usage of memory on clients.
642 Details : 1. On clients, recycle dentries and inodes unused.
643 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
648 Description: llobdstat fix and enhancement
649 Details : add a counter to set a limit to how many samples will be returned
650 fix a wildcard in the path to limit to obdfilter stats only
654 Description: lfs check only allowed for root
655 Details : Code cleanups and added a regresion test of non-root use of
660 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
661 Details : Fix hr->hr_index overload.
665 Description: Fix FIEMAP functionality.
669 Description: FC11 patchless client support added
673 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
674 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
678 Description: New acc-sm script and various test-framework enhancments.
682 Description: Previously landed patch was reverted.
686 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
690 Description: Add ldiskfs maxdirsize mount option.
694 Description: Fix mkfs.lustre support for 16TB LUNs.
698 Description: Print mmp_check_interval
699 Details : print mmp_check_interval and make it possible to abort
700 mount operation in case it takes too long.
704 Description: Add cascading_rw to lustre mpi tests.
708 Description: Store SOM data in EA.
709 Details : Also several SOM related cleanups.
713 Description: Add refcount to md_open_data
717 Description: Use refcount on lov_request_set to prevent memory corruption.
721 Description: Statahead: use parent dir's i_mutex for synchronization.
725 Description: loadgen improvements.
729 Description: Enable peer health detection in ksocklnd and ko2iblnd.
733 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
737 Description: Fix race on obd_devcie:md_stats
741 Description: Fix SLES 11 client builds.
742 Details : Resolved uninitalized pointer gcc warnings.
746 Description: Allow non-root access for "lfs check".
747 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
751 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
753 Severity : enhancement
755 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
756 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
760 Description: Sanity test_27p fails.
761 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
762 Sanity tests 27 were cleaned up and improved.
766 Description: Unify req format on client/servers
767 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
768 buffers and swabbers approach..
773 Description: drop join file stuff.
775 Severity : enhancement
777 Description: Add 2.6.27 and SLES11 (patchless client) support
781 Description: Fix in ptlrpc_expire_one_request() to print the signed time
784 Severity : enhancement
786 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
789 Severity : enhancement
791 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
793 Severity : enhancement
795 Description: correctly handle big reply message.
796 Details : send LNet event if reply is bigger than buffer and adjust this buffer
801 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
803 Severity : enhancement
805 Description: Add support for OEL5.
809 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
811 Severity : enhancement
813 Description: "weak-modules" support
814 Details : Implement "weak-modules" support which enables kernel modules
815 to be used with any kernel that implements the same kABI. In
816 order to achieve this modules are now installed in
817 /lib/modules/$(uname -r)/updates/kernel on all distributions.
821 Description: client could not reconnect to OST because of an active request.
822 Details : The client now retries to reconnect to the same server, if a
823 connect request failed with EBUSY or -EAGAIN. Test case 35b
824 conf-sanity.sh has been created.
828 Description: Don't allow make backward step on assignin osc next id.
829 Details : race between allocation next id and ll_sync thread can be cause
830 of set wrong osc next id and can be kill valid ost objects.
834 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
838 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
842 Description: Error handling in osc_statfs_interpret() has been improved.
843 Details : Check in osc_statfs_interpret() for EBADR.
847 Description: Do not send statfs() requests to OSTs disabled by administrator.
848 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
850 Severity : enhancement
852 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
854 Severity : enhancement
856 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
858 Severity : enhancement
860 Description: Enhancement for lfs(1) command to use numeric uid/gid.
864 Description: Improved cleanup procedure for conf-sanity test_46a.
868 Description: Do not start bulk transfer, if client tried to reconnect.
869 Details : Do not start bulk transfer, if client tried to reconnect,
870 but failed due to active rpcs. Check for exp_abort_active_req
871 has been implemented.
875 Description: Prevent inconsistences between linux and lustre mount structures.
876 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
877 Make the sleep interruptible.
879 Severity : enhancement
881 Description: provide server to client comms path
882 Details : server to client communications path via new LDLM_SET_INFO rpc
886 Description: truncate starts GFP_FS allocation under transaction causing deadlock
887 Details : ldiskfs_truncate calls grab_cache_page which may start page
888 allocation under an open transaction. This may lead to
889 calling prune_icache with consequent lustre reentrance.
893 Description: send correctly lsm on open replay
894 Details : MDS is trust to LSM size on replay open, but client can set wrong size
899 Description: enable client interface failover
900 Details : When a child reconnects from another NID, properly update export
901 nid hash position and ldlm reverse import.
903 Severity : enhancement
905 Description: Communicate OST degraded/readonly state via statfs to MDS
906 Details : Flags in the statfs returned from OSTs indicate whether the
907 OST is in a degraded RAID state, or if the filesystem has
908 turned read-only after a filesystem error is detected.
913 Description: don't panic if EPROTO was hit when reading symlink
914 Details : correctly handling request reference in error cases.
916 Severity : enhancement
918 Description: Add LustreNetLink, a kernel-userspace communcation path.
920 Severity : enhancement
922 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
927 Description: access to llog context before init.
928 Details : move handling CATALOGS file at osc layer and forbid access to llog
932 Frequency : rare, only when client is writing to OST on local node
934 Description: Avoid deadlock for local client writes
935 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
936 memory freeing context. This allows OST threads to set the
937 PF_MEMALLOC flag on task structures in order to allocate memory
938 from reserved pools and complete IO.
939 Use GFP_HIGHUSER for OST allocations for non-local client writes,
940 so that the OST threads generate memory pressure and allow
941 inactive pages to be reclaimed.
943 Severity : enhancement
945 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
950 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
951 Details : move ldlm namespace creation in setup phase to avoid grab
952 _lprocfs_lock with cli_sem held.
956 Description: Temporarily disable grant shrink.
957 Details : Disable the feature for debugging.
960 Frequency : only when formatting test filesystems
962 Description: Unable to run several mkfs.lustre on loop devices at the same time
963 Details : mkfs.lustre returns error 256 on the concurrent loop devices
964 formatting. The solution is to proper handle the error.
966 Severity : enhancement
968 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
970 Severity : enhancement
972 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
977 Description: resolve race between obd_disconnect and class_disconnect_exports
978 Details : if obd_disconnect will be called to already disconnected export he
979 forget release one reference and osc module can't unloaded.
981 Severity : enhancement
983 Description: Allow tuning service thread via /proc
984 Details : For each service a new
985 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
986 created that can be used to set min/max thread counts, and get the
987 current number of running threads.
992 Descriptoin: don't return error if have partially created objects for file.
993 Details : lov_update_create_set uses set->set_success as index for created
994 objects, so if some requests will be failed, they will have hole
995 at end of array and qos_shrink_lsm for allocate correct lsm.
997 Severity : enhancement
999 Description: Update OFED support to 1.4.1
1004 Description: Reduce small size read RPC
1005 Details : Set read-ahead limit for every file and only do read-ahead when
1006 available read-ahead pages are bigger than 1M to avoid small size
1009 Severity : enhancement
1011 Description: move AT tunable parameters for more consistent usage
1012 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1014 Severity : enhancement
1016 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1017 Details : allow skip disconnected ost for send statfs request and hide error
1023 Description: fix racy locking of mballoc block bitmaps causing BUG
1024 Details : The locking of the mballoc buddy bitmap and the in-memory
1025 block bitmap was using two different spin locks in some
1026 cases. This made it possible to incorrectly access the
1027 mballoc bitmap while another process was modifying it,
1028 causing a sanity assertion to fail. While no on-disk corruption
1029 was reported, there was some risk of this happening.
1032 Frequency : rare, on llog test 6
1034 Descriptoin: don't allow connect to already connected import
1035 Details : allowing connect to already connected import is hide connecting problem.
1040 Descriptoin: allow kill process which wait statahead result
1041 Details : for some reasons 'ls' can stick in waiting result from statahead,
1042 in this case need way for kill this process.
1044 Severity : enhancement
1046 Description: Add state history info file, enhance import info file
1047 Details : Track import connection state changes in a new osc/mdc proc file;
1048 add overview-type data to the osc/mdc import proc file.
1050 Severity : enhancement
1052 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1054 Severity : enhancement
1056 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1061 Descriptoin: OST grant too much space to client even there are not enough space.
1062 Details : Client will shrink its grant cache to OST if there are no write
1063 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1064 this grant cache if there are already not enough avaible space
1065 (left_space < total_clients * 32M).
1068 Frequency : start MDS on uncleanly shutdowned MDS device
1070 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1071 Details : stay in waiting mds<>ost recovery finished produce random bugs
1072 due race between two ll_sync thread for one lov target. send
1073 ACTIVATE event only if connect realy finished and import have
1077 Frequency : rare, connect and disconnect target at same time
1079 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1080 Details : don't call obd_disconnect under lov_lock. this long time
1081 operation and can block ptlrpcd which answer to connect request.
1086 Descriptoin: don't lose wakeup for imp_recovery_waitq
1087 Details : recover_import_no_retry or invalidate_import and import_close can
1088 both sleep on imp_recovery_waitq, but we was send only one wakeup
1092 Frequency : always with long access acl
1094 Descriptoin: mds can't pack reply with long acl.
1095 Details : mds don't control size of acl but they limited by reint/getattr
1098 Severity : enhancement
1100 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1102 Severity : enhancement
1104 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1107 Frequency : start MDS on uncleanly shutdowned MDS device
1109 Descriptoin: aborting recovery hang on MDS
1110 Details : don't throttle destroy RPCs for the MDT.
1113 Frequency : on remount
1115 Description: external journal device not working after the remount
1116 Details : clear dev_rdonly flag for external journal devices in
1122 Description: shutdown vs evict race
1123 Details : client_disconnect_export vs connect request race.
1124 if client will evicted at this time - we start invalidate
1125 thread without referece to import and import can be freed
1129 Frequency : rare, need acl's on inode.
1131 Description: client can't handle ost additional correctly
1132 Details : if ost was added after client connected to mds client can have
1133 hit lnet_try_match_md ... to big messages to wide striped files.
1134 in this case need teach client to handle config events about add
1135 lov target and update client max ea size at that event.
1137 Severity : enhancement
1139 Description: Changelogs
1140 Details : Changelogs are a lightweight mechanism to track filesystem
1141 metadata and namespace changes. The changelog is recorded
1142 permanently on the MDTs, and is periodically "consumed" / purged
1143 when records are no longer needed.
1145 Severity : enhancement
1147 Description: compact fld format with extents
1148 Details : Store range of seq rather than every seq in FLD. Seq
1149 controller update FLD rather than clients. In Case of CMD, mdt0
1150 has FLD, all other metadata server act as non persistent proxy
1151 for FLD queries and cache fld entries in fld cache.
1156 Description: don't skip ost target if they assigned to file
1157 Details : Drop slow OSCs if we can, but not for requested start idx.
1158 This means "if OSC is slow and it is not the requested
1159 start OST, then it can be skipped, otherwise skip it only
1160 if it is inactive/recovering/out-of-space.
1164 Description: more cleanup in mds_lov
1165 Details : not send LOV EA under replay, we can't know about they size at this
1166 time. Don't allow client connect to mds before any ost connected,
1167 for avoid problems with LOV EA size and returning EIO to client.
1169 Severity : enhancement
1171 Description: Interoperability at server side (Disk interoperability)
1173 Severity : enhancement
1175 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1177 Severity : enhancement
1179 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1181 Severity : enhancement
1183 Description: New client IO stack (CLIO).
1185 Severity : enhancement
1187 Description: Commit on sharing. Eliminate inter-client dependencies between
1188 uncommitted transactions by doing transaction commits.
1189 Thereby clients may recovery independently.
1192 Frequency : Create a symlink file with a very long name
1194 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1195 Details : If there is no extra space in the request for early cancels,
1196 ldlm_req_handles_avail() returns 0 instead of a negative value.
1198 Severity : enhancement
1200 Description: Add /proc entry for import status
1201 Details : The mdc, osc, and mgc import directories now have
1202 an import directory that contains useful import data for debugging
1203 connection problems.
1205 Severity : enhancement
1207 Description: Re-disable certain /proc logging
1208 Details : Enable and disable client's offset_stats, extents_stats and
1209 extents_stats_per_process stats logging on the fly.
1212 Frequency : Only on FC kernels 2.6.22+
1214 Description: oops in statahead
1215 Details : Do not drop reference count for the dentry from VFS when lookup,
1216 VFS will do that by itself.
1218 Severity : enhancement
1220 Description: Generic /proc file permissions
1221 Details : Set /Proc file permissions in a more generic way to enable non-
1222 root users operate on some /proc files.
1226 Description: Hitting mdc_commit_close() ASSERTION
1227 Details : Properly handle request reference release in
1228 ll_release_openhandle().
1232 Description: quota recovery deadlock during mds failover
1233 Details : This patch includes att18982, att18236, att18237 in bz14840.
1235 1. fix osts hang when mds does failover with quotaon
1236 2. prevent watchdog storm when osts threads wait for the
1241 Frequency : only patchless client
1242 Description: add workaround for race between add/remove dentry from hash
1244 Severity : enhancement
1246 Description: Allow OST glimpses to return PW locks
1250 Description: LBUG when llog conf file is full
1251 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1255 Description: Prevent import from entering FULL state when server in recovery
1259 Description: service mount cannot take device name with ":"
1260 Details : Only when device name contains ":/" will mount treat it as
1266 Description: replace ptlrpcd with the statahead thread to interpret the async
1267 statahead RPC callback
1271 Frequency : on recovery
1272 Description: I/O failures after umount during fail back
1273 Details : if client reconnected to restarted server we need join to recovery
1274 instead of find server handler is changed and process self
1275 eviction with cancel all locks.
1277 Severity : enhancement
1279 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1281 Severity : enhancement
1283 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1285 Severity : enhancement
1287 Description: Upcall on Lustre log has been dumped
1288 Details : Allow for a user mode script to be called once a Lustre log has
1289 been dumped. It passes the filename of the dumped log to the
1290 script, the location of the script can be specified via
1291 /proc/sys/lnet/debug_log_upcall.
1296 Description: avoid idr_remove called for id which is not allocated.
1297 Details : Move assigment s_dev for clustered nfs to end of initialization,
1298 to avoid problem with error handling.
1303 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1304 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1305 into CONN_USED_HASH and this prodice warning when put connection
1306 again in unused hash.
1309 Severity : enhancement
1311 Description: File striping can now be set to use an arbitrary pool of OSTs.
1313 Severity : enhancement
1315 Description: Export bytes_read/bytes_write count on OSC/OST.
1319 Description: Early reply size mismatch, MGC loses connection
1320 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1321 the connect flags are properly negotiated.
1326 Description: quotas are not honored with O_DIRECT
1327 Details : all writes with the flag O_DIRECT will use grants which leads to
1328 this problem. Now using OBD_BRW_SYNC to guard this.
1332 Description: add quota statistics
1333 Details : 1. sort out quota proc entries and proc code.
1334 2. add quota statistics
1336 Severity : enhancement
1338 Description: enable quota support for HEAD.
1342 Description: Properly propagate oinfo flags from lov to osc for statfs
1343 Details : restore missing copy oi_flags to lov requests.
1345 Severity : enhancement
1347 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1348 : l_getgroups(8), lst(8), routerstat(8)
1349 Details : included man pages for llobdstat(8), llstat(8),
1350 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1352 Severity : enhancement
1354 Description: configure's --enable-quota should check the
1355 : kernel .config for CONFIG_QUOTA
1356 Details : configure is terminated if --enable-quota is passed but
1357 : no quota support is in kernel
1361 Description: Remove portals compatibility
1362 Details : Remove portals compatibility, not interoperable with releases
1367 Description: Resolve device initialization race
1368 Details : Prevent proc handler from accessing devices added to the
1369 obd_devs array but yet be intialized.
1371 Severity : enhancement
1373 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1375 Severity : enhancement
1377 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1382 Description: Using wrong pointer in osc_brw_prep_request
1383 Details : Access to array[-1] can produce panic if kernel compiled with
1384 CONFIG_PAGE_ALLOC enabled
1388 Description: Client runs out of low memory
1389 Details : Consider only lowmem when counting initial number of llap pages
1393 Description: *optional* service tags registration
1394 Details : if the "service tags" package is installed on a Lustre node
1395 When the filesystem is mounted, a local-node service tag will
1396 be created. See http://inventory.sun.com/ for more information
1397 about the Service Tags asset management system.
1401 Description: Kernel BUG tries to release flock
1402 Details : Lustre does not destroy flock lock before last reference goes
1403 away. So always drop flock locks when client is evicted and
1404 perform unlock regardless of successfulness of speaking to MDS.
1408 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1412 Description: sanity test 65a fails if stripecount of -1 is set
1413 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1418 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1419 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1422 Severity : enhancement
1424 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1428 Description: do not process already freed flock
1429 Details : flock can possibly be freed by another thread before it reaches
1430 to ldlm_flock_completion_ast.
1434 Description: LBUG during stress test
1435 Details : Need properly lock accesses the flock deadlock detection list.
1439 Description: oops in page fault handler
1440 Details : kernel page fault handler can return two special 'pages' in error
1441 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1445 Description: timeout with invalidate import.
1446 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1447 should be handled by ptlrpcd. This produce long age waiting and
1448 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1450 Severity : enhancement
1452 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1456 Description: enable MGS and MDT services start separately
1457 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1458 a co-located MGS without starting the MGS, which is a complement
1459 to 'nosvc' mount option.
1463 Frequency : after recovery
1464 Description: precreate to many object's after del orphan.
1465 Details : del orphan st in oscc last_id == next_id and this triger growing
1466 count of precreated objects. Set flag LOW to skip increase count
1467 of precreated objects.
1471 Frequency : rare, on clear nid stats
1472 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1473 Details : when clean nid stats sometimes try destroy live entry,
1474 and this produce panic in free.
1478 Description: Stack overflow during MDS log replay
1479 ease stack pressure by using a thread dealing llog_process.
1483 Description: wait until IO finished before start new when do lock cancel.
1484 Details : VM protocol want old IO finished before start new, in this case
1485 need wait until PG_writeback is cleared until check dirty flag and
1486 call writepages in lock cancel callback.
1488 Severity : enhancement
1490 Description: using special macro for print time and cleanup in includes.
1494 Description: mds_mfd_close() ASSERTION(rc == 0)
1495 Details : In mds_mfd_close(), we need protect inode's writecount change
1496 within its orphan write semaphore to prevent possible races.
1500 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1504 Frequency : rare, on shutdown ost
1505 Description: don't hit live lock with umount ost.
1506 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1507 use shrink_dcache_sb instead.
1511 Description: don't panic with use echo client
1512 Details : echo client pass NULL as client nid pointer and this produce null
1513 pointer dereference.
1517 Description: fix build on ppc32
1518 Details : compile code with -m64 flag produce wrong object file for ppc32.
1522 Description: add message levels for liblustreapi
1526 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1527 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1528 will result in an -EAGAIN and in the exit call path we call
1529 alloc_rr() which will with increasing aggressiveness attempt to
1530 aquire precreated objects on the minimum number of required OSCs.
1534 Description: Use old size assignment to avoid deadlock
1535 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1536 the scheduling while holding a spinlock. We do not need locking
1537 for size in ll_update_inode() because size is only updated from
1538 the MDS for directories or files without objects, so there is no
1539 other place to do the update, and concurrent access to such inodes
1540 are protected by the inode lock.
1544 Description: resolve "_IOWR redefined" build error on SLES10
1548 Description: dump the memory debugging after all modules are unloaded to
1549 suppress false negative in conf_sanity test 39
1551 Severity : enhancement
1553 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1558 Description: fix SLES kernel versioning
1559 Details : the kernel version for our SLES 10 kernel did not include a "-"
1560 before the "smp" at the end. while this was not a problem in
1561 general, it did mean that software trying to use the kernel
1562 version to try to detect a vendor specific kernel would fail.
1563 this was most evident by the OFED build scripts.
1567 Description: Don't update lov_desc members until making sure they are valid
1568 Details : When updating lov_desc members via proc fs, need fix their
1569 validities before doing the real update.
1573 Description: don't put request into delay list while invalidate in flight.
1574 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1575 import in flight. this produce timeout for invalidate and sometimes
1576 can cause stale data.
1580 Frequency : on ppc only
1581 Description: not convert ost objects for directory because it's not exist.
1582 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1586 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1588 Severity : enhancement
1590 Description: Update to support 2.6.22.14 vanilla kernel.
1594 Frequency : rare, on recovery
1595 Description: read procfs can produce deadlock in some situation
1596 Details : Holding lprocfs lock which send rpc can produce block for destroy
1597 obd objects and this also block reconnect with -EALREADY. This
1598 isn't fix all lprocfs bugs - but make it rare.
1600 Severity : enhancement
1602 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1605 Frequency : frequent on X2 node
1607 Description: mdc_set_open_replay_data LBUG
1608 Details : Set replay data for requests that are eligible for replay.
1612 Description: lustre_mgs: operation 101 on unconnected MGS
1613 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1614 MGC, and late on MGC cannot successfully connect to MGS and a lot
1615 of the error messages complaining that MGS is not connected.
1619 Frequency : on network error
1620 Description: panic with double free request if network error
1621 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1622 but it's true only for synchronus enqueue, for async enqueue
1623 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1626 Severity : enhancement
1628 Description: client-side metadata stat-ahead during readdir(directory readahead)
1629 Details : perform client-side metadata stat-ahead when the client detects
1630 readdir and sequential stat of dir entries therein
1633 Frequency : on start mds
1635 Description: Implement get_info(last_id) in obdfilter.
1638 Frequency : occasional
1640 Description: Correctly check stale fid, not start epoch if ost not support SOM
1641 Details : open with flag O_CREATE need set old fid in op_fid3 because
1642 op_fid2 was overwritten with new generated fid, but mds can answer
1643 with one of these two fids and both is not stale. Setattr
1644 incorrectly started an epoch and assume will be called
1645 done_writing, but without SOM done_writing ever being called.
1648 Frequency : rare, depends on device drivers and load
1650 Description: MDS or OSS nodes crash due to stack overflow
1651 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1652 In some cases, in conjunction with device drivers that use a lot
1653 of stack the MDS (or possibly OSS) service threads could overflow
1654 the stack. One change which was identified to consume additional
1655 stack has been reworked to avoid the extra stack usage.
1658 Frequency : occasional
1660 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1661 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1662 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1663 oscc_has_objects and see OSC has no objests and call
1664 oscc_internal_create to resend create request.
1666 Severity : enhancement
1668 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1670 Severity : enhancement
1672 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1675 Frequency : very rare
1677 Description: Fix replay if there is an un-replied request and open
1678 Details : In some cases, older replay request will revert the
1679 mcd->mcd_last_xid on MDS which is used to record the client's
1680 latest sent request.
1682 Severity : enhancement
1684 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1686 Severity : enhancement
1688 Description: Add rhel5 support to HEAD.
1690 Serverity : enhancement
1692 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1697 Description: Don't allow skipping OSTs if index has been specified.
1698 Details : Don't allow skipping OSTs if index has been specified, make
1699 locking in internal create lots better.
1703 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1704 Details : access to bitfield in structure is always rounded to long
1705 and this produce problem with not atomic change any bit.
1709 Description: Lustre make rpms failed.
1710 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1711 builds Lustre rpms from tarball.
1714 Frequency : rare on shutdown ost
1716 Description: If llog cancel was not send before clean_exports phase, this can
1717 produce deadlock in llog code.
1718 Details : If llog thread has last reference to obd and call class_import_put
1719 this produce deadlock because llog_cleanup_commit_master wait when
1720 last llog_commit_thread exited, but this never success because was
1721 called from llog_commit_thread.
1725 Description: allow userland application know is lost one of stripes.
1726 Details : fill lvb_blocks with error code on ost and return it to
1727 application if error flag found.
1731 Description: NULL lov_tgts causing MDS oops
1732 Details : more safe checks for NULL lov_tgts for avoid oops.
1734 Severity : enhancement
1736 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1740 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1741 Details : If an OST is down the MDS will hang indefinitely in
1742 obd_statfs_rqset() waiting for the statfs data. While for
1743 MDS QOS usage of statfs, it should not stuck in waiting.
1745 Severity : enhancement
1747 Description: remote_acl support
1748 Details : Support ACL-based permission check for remote user.
1749 Support setfacl/getfacl for remote user with the utils
1750 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1751 the system "{s,g}etfacl" utils.
1753 Severity : enhancement
1755 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1757 Severity : enhancement
1759 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1763 Description: make mgs_setparam() handle fsname containing dash
1764 Details : fsname containing a dash does not work with lctl conf_param
1766 Severity : enhancement
1768 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1770 Severity : enhancement
1772 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1776 Frequency : rare, at shutdown
1777 Description: access already free / zero obd_namespace.
1778 Details : if client_disconnect_export was called without force flag set,
1779 and exist connect request in flight, this can produce access to
1780 NULL pointer (or already free pointer) when connect_interpret
1781 store ocd flags in obd_namespace.
1785 Frequency : only at startup
1786 Description: not alloc memory with spinlock held.
1787 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1788 if any spinlock held.
1790 Severity : enhancement
1792 Description: make lustre randomly fail allocating memory
1793 Details : Make lustre randomly failed allocating memory for testing purpose.
1795 Severity : enhancement
1797 Description: lost problems with lov objid file
1798 Details : Fixes some scability and access to not inited memory problems
1799 in work with lov objdid file.
1804 Description: lfs find does not continue on file error
1805 Details : Continue other files processing when a file/dir is absent.
1809 Description: Inconsistent usage of lustre_pack_reply()
1810 Details : Standardize the usage of lustre_pack_reply() such that it
1811 always generate a CERROR on failure.
1814 Frequency : occasional
1816 Description: MDS hang or stay in waiting lock
1817 Details : If client receive lock with CBPENDING flag ldlm need send lock
1818 cancel as separate rpc, to avoid situation when cancel request
1819 can't processed due all i/o threads stay in wait lock.
1823 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1827 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1831 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1835 Description: Update to RHEL4 latest kernel.
1837 Severity : enhancement
1839 Description: Build SLES10 patchless client fails
1840 Details : The configure was broken by run ./configure with
1841 --with-linux-obj=.... argument for patchless client. When the
1842 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1843 can't search header adequately. Use absolute path such as
1844 -I($LINUX)/include instead.
1848 Description: interrupt oig_wait produce painc on resend.
1849 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1850 private set, and this produce situation when rq_ptlrpcd_data not
1851 copyed to new allocated request and triggered LBUG on assert
1852 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1853 ptlrpcd set if request is changed and can be safety changed to use
1858 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1859 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1860 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1861 these macros and replace them with their difinitions instead.
1865 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1866 Details : Update which_patch & target file for SLES10 latest kernel.
1869 Description: Remove client patches from SLES 10 kernel.
1870 Details : This causes SLES 10 clients to behave as patchless clients
1871 even on a Lustre-patched (server) kernel.
1873 Severity : enhancement
1875 Description: self-adjustable client's lru lists
1876 Details : use adaptive algorithm for managing client cached locks lru
1877 lists according to current server load, other client's work
1878 pattern, memory activities, etc. Both, server and client
1879 side namespaces provide number of proc tunables for controlling
1882 Severity : enhancement
1884 Description: light-weight GSS support
1885 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1886 and reduce performance overhead.
1888 Severity : enhancement
1890 Description: Linux keyring support
1891 Details : Support using service of Linux keyring for Lustre GSS internal
1892 context refresh/cache mechanism.
1896 Description: Fix errors in lfs documentation
1897 Details : Fixes man pages
1901 Description: don't use GFP_* in generic Lustre code.
1902 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1906 Description: obdclass is limited by single OBD_ALLOC(idarray)
1907 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1911 Description: warnings with build patchless client with vanila 2.6.19 and up
1912 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1913 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1917 Frequency : Only for SLES
1918 Description: sanity_quota fail test_1
1919 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1920 file, which makes TSTID[2] inunique.
1924 Frequency : Only for SLES9
1925 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1926 Details : Update target/ChangeLog/which_patch .
1930 Frequency : Only for test_10 in sanity-quota.sh
1931 Description: a bug in quota test code
1932 Details : checking a test flag in a wrong place
1936 Frequency : Only for test_13 in sanity-quota.sh
1937 Description: a bug in quota test script
1938 Details : shouldn't check overfull allocated quota
1942 Description: jbd statistics
1943 Details : Port older jbd statistics patch for sles10
1945 Severity : enhancement
1947 Description: Add printing inode info into message about error in writepage.
1952 Description: reply_lock_interpret crash due to race with it and lock cancel.
1953 Details : Do not replay locks that are being cancelled. Do not reference
1954 locks by their address during replay, just by their handle.
1958 Frequency : When flocks are used.
1959 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1960 Details : Flock locks might destroy just granted lock if it could be merged
1961 with another existing flock, this is done in completion handler,
1962 so teach ldlm_cli_enquque_fini that this is a valid case for
1968 Description: Oops in read and write path when failing to allocate lock.
1969 Details : Check if lock allocation failed and return error back.
1973 Description: lstripe command fails for valid OST index
1974 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1975 instead of lov->desc.ld_active_tgt_count.
1979 Description: sanity.sh failed test 103
1980 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1985 Description: df doesn't work properly if diskfs blocksize != 4K
1986 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1990 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1991 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1993 Severity : enhancement
1995 Description: Adaptive timeouts
1996 Details : RPC timeouts adapt to changing server load and network
1997 conditions to reduce resend attempts and improve recovery time.
2001 Description: llapi_file_create() does not allow some changes
2002 Details : add llapi_file_open() that allows specifying the mode and
2003 open flags, and also returns an open file handle.
2007 Description: merge and cleanup kernel patches.
2008 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2012 Description: Add journal checksum support.(Kernel part)
2013 Details : The journal checksum feature adds two new flags i.e
2014 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2015 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2016 indicates that the commit block contains the checksum for
2017 the blocks described by the descriptor blocks. Now commit
2018 record can be sent to disk without waiting for descriptor
2019 blocks to be written to disk. This behavior is controlled
2020 using JBD2_FEATURE_ASYNC_COMMIT flag.
2024 Description: OSS needs mutliple precreate threads
2025 Details : Add ability to start more than one create thread per OSS.
2029 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2030 Details : execute lfs setstripe on client
2034 Description: mds_obd_create error creating tmp object
2035 Details : When the user sets quota on root, llog will be affected and can't
2036 create files and write files.
2039 Frequency : Always on ia64 patchless client, and possibly others.
2041 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2042 Details : This allows the patchless client to be loaded on architectures
2043 without this export.
2047 Description: RedHat Update kernel for RHEL5
2048 Details : Add the kernel config file for RHEL5.
2052 Description: RedHat Update kernel for RHEL5
2053 Details : Modify the kernel config file more closer RHEL5.
2057 Description: Build failure against Centos5 (RHEL5)
2058 Details : Define PAGE_SIZE when it isn't present.
2062 Description: test_7 on sanity-quota.sh can't be used on separate machines
2063 Details : add facet to handle it
2067 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2068 Details : with NFS, the anon dentry's parent was set to itself in
2069 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2070 corresponding dentry other than use rec->ur_name.
2072 Severity : enhancement
2074 Description: lfs setstripe enhancement
2075 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2079 Description: enable data checksumming by default
2080 Details : enable checksum by default, allow --disable-checksum
2081 configure option and "-o nochecksum" mount option. Checksums
2082 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2086 Description: lustre support for RHEL5
2087 Details : Add support for RHEL5.
2091 Description: Client eviction due to failover config
2092 Details : after a connection loss, the lustre client should attempt to
2093 reconnect to the last active server first before trying the
2094 other potential connections.
2098 Description: when mds and osts use different quota unit(32bit and 64bit),
2099 quota will be released repeatly.
2100 Details : void sending multiple quota reqs to mds, which will keep the
2101 status between the reqs.
2105 Description: osts not allocated evenly to files
2106 Details : change the condition to increase offset_idx
2109 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2111 Description: Data corruption for OSTs that are formatted larger than 2TB
2113 Details : When generating the bio request for lustre file writes the
2114 sector number would overflow a temporary variable before being
2115 used for the IO. The data reads correctly from Lustre (which
2116 will overflow in a similar manner) but other file data or
2117 filesystem metadata may be corrupted in some cases.
2121 Description: Tune the kernel for good SCSI performance.
2122 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2123 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2128 Description: rewrite ext2-derived code in obdclass/uuid.c
2129 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2130 from scratch preserving functionality.
2134 Description: "lfs find -obd UUID" prints directories
2135 Details : "lfs find -obd UUID" will return all directory names instead
2136 of just file names. It is incorrect because the directories
2137 do not reside on the OSTs.
2141 Description: use i_size_read and i_size_write in 2.6 port
2142 Details : replace inode->i_size access with i_size_read/write()
2145 Frequency : when removing large files
2147 Description: scheduling issue during removal of large Lustre files
2148 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2149 It causes scheduling issues when removing large files (17TB in the
2153 Frequency : only with liblustre clients on XT3
2155 Description: evictions taking too long
2156 Details : allow llrd to evict clients directly on OSTs
2159 Frequency : only on ppc
2161 Description: /proc/fs/lustre/devices broken on ppc
2162 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2163 We should make sure the type of 'index' is loff_t and then cast
2164 explicitly as needed below. Do not assign an explicitly cast
2168 Frequency : only for rhel5
2170 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2171 Details : Modify the target file & which_kernel.
2173 Severity : enhancement
2175 Description: omit set fsid for export NFS
2176 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2177 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2181 Frequency : Always, for kernels after 2.6.16
2182 Description: Fix warning idr_remove called for id=.. which is not allocated.
2183 Details : Last kernels save old s_dev before kill super and not allow
2184 to restore from callback - restore it before call kill_anon_super.
2188 Description: buffer overruns could theoretically occur
2189 Details : llapi_semantic_traverse() modifies the "path" argument by
2190 appending values to the end of the origin string, and a
2191 overrun may occur. Adding buffer overrun check in liblustreapi.
2195 Description: Fix error on 'ls .' at the top of the Lustre mount.
2196 Details : Don't revalidate dentry if it is a root dentry.
2200 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2201 Details : Modify vm-tunables-rhel4.patch.
2205 Description: Kernel config for 2.6.18-vanilla.
2206 Details : Modify targets/2.6-vanilla.target.in.
2207 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2208 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2209 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2210 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2214 Description: improve handling recoverable errors
2215 Details : if request processig with error which can be recoverable on server
2216 request should be resend, otherwise page released from cache and
2221 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2222 Details : Modify target file & which_patch.
2223 A flaw was found in the IA32 system call emulation provided
2224 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2225 value could be stored in the %RAX register, which could trigger an
2226 out-of-bounds system call table access. An untrusted local user
2227 could exploit this flaw to run code in the kernel
2228 (ie a root privilege escalation). (CVE-2007-4573).
2232 Description: O_DIRECT bypasses client statistics.
2233 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2234 read_ahead_stats were updated. copied the stats section from
2235 osc_send_oap_rpc() into async_internal().
2239 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2243 Description: change order of libsysio includes
2244 Details : '#include sysio.h' should always come before '#include xtio.h'
2248 Description: handle "serious error: objid * already exists" more gracefully
2249 Details : If LAST_ID value on disk is smaller than the objects existing in
2250 the O/0/d* directories, it indicates disk corruption and causes an
2251 LBUG(). If the object is 0-length, then we should use the existing
2252 object. This will help to avoid a full fsck in most cases.
2254 Severity : enhancement
2256 Description: adapt the lustre_config script to support the upgrade case
2257 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2258 server targets to 1.6 in parallel.
2262 Description: To avoid grant space > avaible space when the disk is almost
2263 full. Without this patch you might see the error "grant XXXX >
2264 available" or some LBUG about grant, when the disk is almost
2266 Details : In filter_check_grant, for non_grant cache write, we should
2267 check the left space by if (*left > ungranted + bytes), instead
2268 of (*left > ungranted), because only we are sure the left space
2269 is enough for another "bytes", then the ungrant space should be
2270 increase. In client, we should update cl_avail_grant only there
2271 is OBD_MD_FLGRANT in the reply.
2275 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2276 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2277 described in CVE-2007-4573. This problem could allow untrusted
2278 local users to gain root access.
2281 Frequency : when using O_DIRECT and quotas
2283 Description: Incorrect file ownership on O_DIRECT output files
2284 Details : block usage reported by 'lfs quota' does not take into account
2285 files that have been written with O_DIRECT.
2290 Description: touch file failed when fs is not full
2291 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2292 otherwise we can get ENOSP while fs is not full.
2296 Description: parallel lock callbacks
2297 Details : Instead of sending blocking and completion callbacks as separated
2298 requests, adding them to a set and sending in parallel.
2301 Frequency : only for Cray XT3
2302 Bugzilla : 12829/13455
2303 Description: Changing primary group doesn't change the group lustre assigns to
2305 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2306 overridden with the primary group provided by the group upcall,
2307 whereas the supplied fsgid can be trusted if it is in the list of
2308 supplementary groups returned by the group upcall.
2310 Severity : enhancement
2312 Description: Allow masking D_WARNING, D_ERROR messages from console
2313 Details : Console messages can now be disabled via lnet.printk.
2317 Description: User code with malformed file open parameter crashes client node
2318 Details : Before packing join_file req, all the related reference should be
2319 checked carefully in case some malformed flags cause fake
2320 join_file req on client.
2324 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2325 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2326 mode is equal to the granted mode. However, it should also check
2327 that LDLM_FL_CP_REQD is not set.
2331 Description: Heavy nfs access might result in deadlocks
2332 Details : After ELC code landed, it is now improper to enqueue any mds
2333 locks under och_sem, because enqueue might want to decide to
2334 cancel open locks for same inode we are holding och_sem for.
2338 Description: Client eviction while running blogbench
2339 Details : A lot of unlink operations with concurrent I/O can lead to a
2340 deadlock causing evictions. To address the problem, the number of
2341 oustanding OST_DESTROY requests is now throttled to
2342 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2343 callbacks are processed in priority.
2347 Description: enable ACLs on MDS by default
2348 Details : ACLs must be enabled on MDS by default.
2351 Frequency : PPC/PPC64 only
2353 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2354 Details : fix duplicated definitions between asm-ppc64/types.h and
2355 lustre_types.h on PPC.
2358 Frequency : PPC/PPC64 only
2360 Description: asm-ppc/segment.h does not exist
2361 Details : fix compile issue on PPC.
2365 Description: better handle error messages in extents code
2368 Frequency : RHEL4 only
2370 Description: mkfs is very slow on IA64/RHEL4
2371 Details : A performance regression has been discovered in the MPT Fusion
2372 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2373 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2374 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2376 Severity : enhancement
2378 Description: SNMP support enhancement
2379 Details : Adding total number of sampled request for an MDS node in snmp
2382 Severity : enhancement
2384 Description: Optimize ldlm waiting list processing for PR extent locks
2385 Details : When processing waiting list for read extent lock and meeting read
2386 lock that is same or wider to it that is not contended, skip
2387 processing rest of the list and immediatelly return current
2388 status of conflictness, since we are guaranteed there are no
2389 conflicting locks in the rest of the list.
2393 Description: Time out and refuse to reconnect
2394 Details : When the failover node is the primary node, it is possible
2395 to have two identical connections in imp_conn_list. We must
2396 compare not conn's pointers but NIDs, otherwise we can defeat
2397 connection throttling.
2401 Description: port llog fixes from b1_6 into HEAD
2402 Details : Port llog reference couting and some llog cleanups from b1_6
2403 (bug 10800) into HEAD, for protect from panic and access to already
2404 free llog structures.
2408 Description: Detect stride IO mode in read-ahead
2409 Details : When a client does stride read, read-ahead should detect that and
2410 read-ahead pages according to the detected stride pattern.
2414 Description: data checksumming impacts single node performance
2415 Details : add support for several checksum algorithm. Currently, only CRC32
2416 and Adler-32 are supported. The checksum type can be changed on
2417 the fly via /proc/fs/lustre/osc/*/checksum_type.
2421 Description: use adler32 for page checksums
2422 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2427 Description: build for x2 fails
2428 Details : fix compile issue on Cray systems.
2432 Description: Properly match for duplicate locks
2433 Details : Due to different lock order from skiplists code, we need to
2434 traverse entire list for now
2437 Frequency : only on PPC/SLES10
2439 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2440 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2441 be defined. Add a hack in mkfs_lustre.c to work around this compile
2446 Description: LASSERT on MDS when client holding flock lock dies
2447 Details : ldlm pool logic depends on number of granted locks equal to
2448 number of released locks which is not true for flock locks, so
2449 just exclude such locks from consideration.
2453 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2454 Details : Use fsfilt_sync() for both the whole filesystem sync and
2455 individual file sync to eliminate dangerous inode locking
2456 with I_LOCK that can lead to a deadlock.
2460 Description: performance in 1.6.3
2461 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2462 sure that 1MB requests are not fragmented by the block layer.
2464 Severity : enhancement
2466 Description: organize the server-side client stats on per-nid basis
2467 Details : Change the structure of stats under obdfilter and mds to
2475 The "uuid"s file would list the uuids of _active_ exports.
2476 And the clear entry is to clear all stats and stale nids.
2478 Severity : enhancement
2480 Description: eliminate client locks in face of contention
2481 Details : file contention detection and lockless i/o implementation
2482 for contended files.
2486 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2489 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2491 Description: MDS may lose file striping (and hence file data) in some cases
2492 Details : If there are additional extended attributes stored on the MDS,
2493 in particular ACLs, SELinux, or user attributes (if user_xattr
2494 is specified for the client mount options) then there is a risk
2495 of attribute loss. Additionally, the Lustre file striping
2496 needs to be larger than default (e.g. striped over all OSTs),
2497 and an additional attribute must be stored initially in the
2498 inode and then increase in size enough to be moved to the
2499 external attribute block (e.g. ACL growing in size) for file
2504 Description: skiplist implementation simplification
2505 Details : skiplists are used to group compatible locks on granted list
2506 that was implemented as tracking first and last lock of each
2507 lock group the patch changes that to using doubly linked lists
2511 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2512 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2513 because LASSERT on both the data supplied by a client, and the
2514 data on disk is dangerous and incorrect.
2516 Severity : enhancement
2518 Description: Slow truncate/writes to huge files at high offsets.
2519 Details : Directly associate cached pages to lock that protect those pages,
2520 this allows us to quickly find what pages to write and remove
2521 once lock callback is received.
2525 Description: more ldlm soft lockups
2526 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2527 starve other threads from the resource lock for a long time in
2528 case of long waiting queue, so change the debug level from
2529 D_OTHER to the less frequently used D_INFO.
2531 Severity : enhancement
2533 Description: add -gid, -group, -uid, -user options to lfs find
2537 Description: Hung threads in invalidate_inode_pages2_range
2538 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2539 once one is completed. As a result, some RPCs are stuck in the
2540 queue and are never sent.
2544 Description: filter threads hungs on waiting journal commit
2545 Details : Cleanup filter group llog code, then only filter group llog will
2546 be only created in the MDS/OST syncing process.
2550 Description: Procfs and llog threads access destoryed import sometimes.
2551 Details : Sync the import destoryed process with procfs and llog threads by
2552 the import refcount and semaphore.
2554 Severity : enhancement
2556 Description: openlock cache of b1_6 port to HEAD
2561 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2562 Details : If insertion of an extent fails, then discard the inode
2563 preallocation and free data blocks else it can lead to duplicate
2568 Description: don't always update ctime in ext3_xattr_set_handle()
2569 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2570 In some cases the ctime should not be updated, for example for
2571 2.0->1.8 compatibility it is necessary to delete an xattr and it
2572 should not update the ctime.
2576 Bugzilla : 15713/16362
2577 Description: Assertion in iopen_connect_dentry in 1.6.3
2578 Details : looking up an inode via iopen with the wrong generation number can
2579 populate the dcache with a disconneced dentry while the inode
2580 number is in the process of being reallocated. This causes an
2581 assertion failure in iopen since the inode's dentry list contains
2582 both a connected and disconnected dentry.
2586 Description: assertion failure in ldlm_handle2lock()
2587 Details : fix a race between class_handle_unhash() and class_handle2object()
2588 introduced in lustre 1.6.5 by bug 13622.
2593 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2594 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2599 Description: oops in sd_iostats_seq_show()
2600 Details : unloading/reloading the scsi low level driver triggers a kernel
2601 bug when trying to access the sd iostat file.
2606 Description: Kernel panics during QLogic driver reload
2607 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2608 patch, causing memory corruption.
2613 Description: journal_dev option does not work in b1_6
2614 Details : pass mount option during pre-mount.
2616 Severity : enhancement
2618 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2619 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2620 extent information of a file. It can be used to map logical blocks
2621 in a file to physical blocks in the block device.
2625 Description: LDLM soft lockups - improvement
2626 Details : It is be possible to send the lock handle along with each read
2627 or write request because the client is already doing a lock match
2628 itself so there isn't any reason the OST should have to re-do that
2632 Frequency : only on Cray X2
2634 Description: X2 build failures
2635 Details : fix build failures on Cray X2.
2639 Description: xid & resent requests
2640 Details : Initialize RPC XID from clock at startup (randomly if clock is
2643 Severity : enhancement
2645 Description: Add lustre_start utility to start or stop multiple Lustre servers
2650 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2651 Details : In case of memory pressure, list_del() can be called twice on
2652 req->rq_history_list, causing a kernel oops.
2656 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2657 Details : fix stack overflow in the distributed lock manager by defering
2658 export eviction after a failed AST to the elt thread instead of
2659 handling it in the dlm interpret routine.
2663 Description: Convert some comments to new format.
2664 Details : Update documenting comments to match doxygen conventions.
2668 Description: Grammar fixes.
2669 Details : A couple of trivial spelling fixes.
2673 Description: OSD_COUNTERS-mandatory
2674 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2675 to simplify the code.
2679 Description: simplify cmm_device freeing logic.
2680 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2684 Description: Add lockdep support to dt_object_operations locking interface.
2685 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2686 parameter indicating lock ordering. Update mdd code to use new
2691 Description: Introduce failloc constants for lockless IO tests.
2692 Details : Add two new failloc constants to test lockless IO. Only one of
2693 them in implemented---another is checked in yet to be landed
2698 Description: Add lockdep support for inode mutex.
2699 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2700 in the situations where Lustre has to lock more than one inode
2705 Description: Add optional invariants checking support.
2706 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2707 configure switch. This macro is to be used for consistency and
2708 sanity checks that are too expensive to be left in `production'
2713 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2714 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2719 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2720 Details : Use spin_lock_nested() in (the only) situation where more than
2721 one ldlm_lock is locked simultaneously. Also, fix possible
2722 dead-lock in ldlm_lock_change_resource() by enforcing particular
2727 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2728 Details : Instead of specifying each ldlm_lock call-back through separate
2729 parameter, wrap them into struct ldlm_callback_suite.
2733 Description: Kill join_lru obd method and its callers.
2734 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2735 in memory---a requirement implemented through ->o_join_lru() obd
2736 method in HEAD. Remove this method and its users.
2740 Description: Add asynchronous ldlm ENQUEUE completion handler.
2741 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2742 case that stock ldlm_completion_ast() cannot handle as it waits
2743 until lock is granted. Introduce new ldlm_completion_ast_async()
2744 for this. Also comment ldlm_completion_ast().
2748 Description: ldlm_error <-> errno conversion.
2749 Details : Add functions to map (rather arbitrary) between LDLM error codes
2750 and standard errno values. CLIO needs this to prevent LDLM specific
2751 constants from escaping ldlm and osc.
2755 Description: Kill unused ldlm_handle2lock_ns() function.
2756 Details : Kill unused ldlm_handle2lock_ns() function.
2760 Description: Add lu_ref support to ldlm_lock
2761 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2762 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2763 and ldlm_resource. LDLM interface has to be changed, because code that
2764 releases a reference on a lock, has to "know" what reference this is.
2765 In the most frequent case
2767 lock = ldlm_handle2lock(handle);
2769 LDLM_LOCK_PUT(lock);
2771 no changes are required. When any other reference (received _not_ from
2772 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2773 instead of LDLM_LOCK_PUT().
2775 Arguably, changes are pervasive, and interface requires some discipline
2776 for proper use. On the other hand, it was very instrumental in finding
2777 a few leaked lock references.
2781 Description: Add ldlm_lock_addref_try().
2782 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2783 attempts to addref a lock that might be being canceled
2788 Description: Add ldlm_weigh_callback().
2789 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2790 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2791 hard-coded `number of pages' logic.
2795 Description: Add lockdep annotations to llog code.
2796 Details : Use appropriately tagged _nested() locking calls in the places
2797 where llog takes more than one ->lgh_lock lock.
2801 Description: Add loi_kms_set().
2802 Details : Wrap kms updates into a helper function.
2806 Description: Constify instances of struct lsm_operations.
2807 Details : Constify instances of struct lsm_operations.
2811 Description: lu_conf support.
2812 Details : On a server, a file system object is uniquely identified
2813 by a fid, which is sufficient to locate and load all object
2814 state (inode). On a client, on the other hand, more data are
2815 necessary instantiate an object. Change lu_object_find() and
2816 friends to take additional `lu_conf' argument describing object.
2817 Typically this includes layout information.
2821 Description: lu_context fixes.
2822 Details : Introduce new lu_context functions that are needed on the client
2823 side, where some system threads (ptlrpcd) are shared by multiple
2824 modules, and so cannot be stopped during module shutdown.
2828 Description: Add start and stop methods to lu_device_type_operations.
2829 Details : Introduce two new methods in lu_device_type_operations, that are
2830 invoked when first instance of a given type is created and last one
2831 is destroyed respectively. This is need by CLIO.
2835 Description: Add lu_ref support to struct lu_device.
2836 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2837 track leaked references.
2841 Description: Introduce lu_kmem_descr.
2842 Details : lu_kmem_descr and its companion interface allow to create
2843 and destroy a number of kmem caches at once.
2847 Description: Fix lu_object finalization race.
2848 Details : Fix a race between lu_object_find() finding an object and its
2849 concurrent finalization. This race is (most likely) not possible
2850 on the server, but might happen on the client.
2854 Description: Introduce lu_ref interface.
2855 Details : lu_ref is a debugging module allowing to track references to
2856 a given object. It is quite cpu expensive, and has to be
2857 explicitly enabled with --enable-lu_ref. See usage description
2862 Description: Factor lu_site procfs stats into a separate function.
2863 Details : Separate lu_site stats printing code into a separate function
2864 that can be reused on a client.
2868 Description: Constify instances of struct {lu,dt,md}_device_operations.
2869 Details : Constify instances of struct {lu,dt,md}_device_operations.
2873 Description: Introduce struct md_site and move meta-data specific parts of
2874 struct lu_site here.
2875 Details : Move md-specific fields out of struct lu_site into special struct
2876 md_site, so that lu_site can be used on a client.
2880 Description: Kill mdd_lov_destroy().
2881 Details : Remove unused mdd code.
2885 Description: Add st_block checking to multistat.c.
2886 Details : Add st_block checking to multistat.c.
2890 Description: Add lu_ref support to struct obd_device.
2891 Details : Add lu_ref tracking to obd_device.
2895 Description: Kill obd_set_fail_loc().
2896 Details : Remove unused code.
2900 Description: Add special type for ptlrpc_request interpret functions.
2901 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2902 there. Actual usage will be in CLIO.
2906 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2907 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2908 doesn't work with lockdep.
2912 Description: Add rwv.c test program.
2913 Details : New testing program exercising readv(2) and writev(2) (Qian).
2917 Description: Add sendfile.c test program.
2918 Details : New testing program exercising sendfile(2) (Jay).
2922 Description: Ratelimit a message that can be very frequent.
2923 Details : Ratelimit a memory allocation failure message that can
2928 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2929 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2933 Description: Liblustre build fixes.
2934 Details : Liblustre build fixes.
2938 Description: libcfs: add cfs_{need,cond}_resched() interface.
2939 Details : libcfs: add cfs_{need,cond}_resched() definition and
2940 implementations for Linux, NT, and liblustre.
2942 Severity : enhancement
2944 Description: More exported tunables for mballoc
2945 Details : Add support for tunable preallocation window and new tunables for
2946 large/small requests
2950 Description: Detect corruption of block bitmap and checking for preallocations
2951 Details : Checks validity of on-disk block bitmap. Also it does better
2952 checking of number of applied preallocations. When corruption is
2953 found, it turns filesystem readonly to prevent further corruptions.
2957 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2958 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2959 window to 0. If it does hit cache miss, and read-pattern is still
2960 stride-io mode, does not reset the stride window, but also does not
2961 increase the stride window length in this case.
2965 Frequency : only for big-endian servers
2966 Description: Check if system is big-endian while mounting fs with extents feature
2967 Details : Mounting a filesystem with extents feature will fail on big-endian
2968 systems since ext3-based ldiskfs is not supported on big-endian
2969 systems. This can be over-riden with "bigendian_extents" mount option.
2971 Severity : enhancement
2973 Description: The root squash functionality
2974 Details : A security feature, which is to prevent users from being able
2975 to mount lustre on their desktop, run as root, and delete
2976 all of the files in the filesystem. The goal is accomplished by
2977 remapping user id (UID) and group id (GID) of the root user to
2978 a UID and GID specified by the system administartor via Lustre
2979 configuration management server (MGS). The functionality also
2980 allows to specify sets of clients for which the remapping does
2985 Description: Excessive recovery window
2986 Details : With AT enabled, the recovery window can be excessively long (6000+
2987 seconds). To address this problem, we no longer use
2988 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2989 timeout no longer depends on the service time, it is set to
2990 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2991 time via pb_service_time.
2995 Description: Watchdog triggered on MDS failover
2996 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2997 the OSTs know that the MDS "UUID" can be reused for the same export
2998 from a different NID, so we do not need to wait for the export to be
3002 Frequency : rare, only if using MMP with Linux RAID
3004 Description: MMP doesn't work with Linux RAID
3005 Details : While using HA for Lustre servers with Linux RAID, it is possible
3006 that MMP will not detect multiple mounts. To make this work we
3007 need to unplug the device queue in RAID when the MMP block is being
3008 written. Also while reading the MMP block, we should read it from
3009 disk and not the cached one.
3011 Severity : enhancement
3013 Description: open file using fid
3014 Details : A file can be opened using just its fid, like
3015 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3018 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3021 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3022 Details : add proper locking for creation of jbd_4k slab cache
3026 Description: MMP check in ext3_remount() fails without displaying any error
3027 Details : When multiple mount protection fails during remount, proper error
3030 Severity : enhancement
3032 Description: Allow stripe size to be up to 4G-64k
3033 Details : Fix math logic to allow large stripe sizes.
3037 Description: add check for >8TB ldiskfs filesystems
3038 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3039 Don't allow >8TB ldiskfs filesystems to be mounted without
3040 force_over_8tb mount option
3045 Descriptoin: OSS complains "too many blocks" and crashes
3046 Details : In filter_commitrw_write():
3047 1. lquota_chkquota();
3048 2. obdo_from_inode();
3049 3. lquota_pending_commit();
3050 "2." has changed uid and gid in this case, then caused
3051 lquota_chkquota() and lquota_pending_commit() to use
3052 different uid and gid. That is the root of the bug.
3056 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3057 Details : Detect and print out the 'stripe count = -1' case properly.
3061 Description: Out or order replies might be lost on replay
3062 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3063 request with tid smaller than one currently being added, add it
3064 to the start, not end of the list.
3066 --------------------------------------------------------------------------------
3068 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3070 * Support for kernels:
3071 2.6.5-7.283 (SLES 9),
3072 2.6.9-55.EL (RHEL 4),
3073 2.6.16.46-0.14 (SLES 10),
3074 2.6.18.8 vanilla (kernel.org)
3075 * Client support for unpatched kernels:
3076 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3077 2.6.16 - 2.6.22 vanilla (kernel.org)
3078 * Due to recently discovered recovery problems, we do not recommend
3079 using patchless RHEL 4 clients with this or any earlier release.
3080 * Recommended e2fsprogs version: 1.39.cfs8
3081 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3082 * Starting with this release, the ldiskfs backing filesystem required
3083 by Lustre is now in its own package, lustre-ldiskfs. This package
3084 should be installed. It is versioned separately from Lustre and
3085 may be released separately in future.
3089 Description: block reactivating mgc import until all deactivates complete
3090 Details : Fix race when failing back MDT/MGS to itself (testing)
3093 Frequency : at statup only
3095 Description: mds_lov_synchronize race leads to various problems
3096 Details : simultaneous MDT->OST connections at startup can cause the
3097 sync to abort, leaving the OSC in a bad state.
3099 Severity : enhancement
3101 Description: add optional extra BUILD_VERSION info
3102 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3103 to override the lustre version.
3106 Frequency : 2.6.18 servers only
3108 Description: ll_kern_mount() doesn't release the module reference
3109 Details : The ldiskfs module reference count never drops down to 0
3110 because ll_kern_mount() doesn't release the module reference.
3115 Description: server LBUG when using old ost_num_threads parameter
3116 Details : Accept the old ost_num_threads parameter but warn that it
3117 is deprecated, and fix an off-by-one error that caused an LBUG.
3122 Description: Transient SCSI error results in persistent IO issue
3123 Details : iobuf->dr_error is not reinitialized to 0 between two
3127 Frequency : sometimes when underlying device returns I/O errors
3129 Description: OSTs not going read-only during write failures
3130 Details : OSTs are not remounted read-only when the journal commit threads
3131 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3132 instead of the ext3 wrappers.
3136 Description: poor connect scaling with increasing client count
3137 Details : Don't run filter_grant_sanity_check for more than 100 exports
3138 to improve scaling for large numbers of clients.
3141 Frequency : SLES10 only
3143 Description: sanity-quota.sh quotacheck failed: rc = -22
3144 Details : Quotas cannot be enabled on SLES10.
3147 Frequency : liblustre clients only
3149 Description: getdirentries does not give error when run on compute nodes
3150 Details : getdirentries does not fail when the size specified as an argument
3151 is too small to contain at least one entry
3153 Severity : enhancement
3155 Description: Add LNET router traceability for debug purposes
3156 Details : If a checksum failure occurs with a router as part of the
3157 IO path, the NID of the last router that forwarded the bulk data
3158 is printed so it can be identified.
3163 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3164 Details : Due to a race condition, liblustre clients were occasionally
3165 evicted incorrectly.
3167 Severity : enhancement
3169 Description: lfs setstripe use optional parameters instead of postional
3172 Severity : enhancement
3174 Description: Nanosecond timestamp support for ldiskfs
3175 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3176 resolution timestamps. There is not yet support for this at
3177 the Lustre filesystem level.
3180 Frequency : during server recovery
3182 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3183 Details : request with rq_no_resend flag not awake l_wait_event if they get
3187 Frequency : nfs export on patchless client
3189 Description: connectathon hang when test nfs export over patchless client
3190 Details : Disconnected dentry cannot be found with lookup, so we do not need
3191 to unhash it or make it invalid
3194 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3195 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3196 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3201 Description: open req refcounting wrong on reconnect
3202 Details : If reconnect happened between getting open reply from server and
3203 call to mdc_set_replay_data in ll_file_open, we will schedule
3204 replay for unreferenced request that we are about to free.
3205 Subsequent close will crash in variety of ways.
3206 Check that request is still eligible for replay in
3207 mdc_set_replay_data().
3212 Description: disable writes to filesystem when reading health_check file
3213 Details : the default for reading the health_check proc file has changed
3214 to NOT do a journal transaction and write to disk, because this
3215 can cause reads of the /proc file to hang and block HA state
3216 checking on a healthy but otherwise heavily loaded system. It
3217 is possible to return to the previous behaviour during configure
3218 with --enable-health-write.
3220 Severity : enhancement
3222 Description: 64-bit inode version
3223 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3224 to the inode. This will be required for version-based recovery.
3229 Description: MDS fails to start if a duplicate client export is detected
3230 Details : in some rare cases it was possible for a client to connect to
3231 an MDS multiple times. Upon recovery the MDS would detect this
3232 and fail during startup. Handle this more gracefully.
3234 Severity : enhancement
3236 Description: Add -o localflock option to simulate old noflock
3238 Details : This will achieve local-only flock/fcntl locks
3244 Description: log_commit_thread vs filter_destroy race leads to crash
3245 Details : Take import reference before releasing llog record semaphore
3250 Description: Wrong request locking in request set processing
3251 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3252 imp_delayed_list, in this place should be used imp_lock.
3255 Frequency : when reconnection
3257 Description: Grant Leak when osc reconnect to OST
3258 Details : When osc reconnect ost, OST(filter) should check whether it
3259 should grant more space to client by comparing fed_grant and
3260 cl_avail_grant, and return the granted space to client instead
3261 of "new granted" space, because client will call osc_init_grant
3262 to update the client grant space info.
3265 Frequency : when client reconnect to OST
3267 Description: Grant Leak when osc do resend and replay bulk write
3268 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3269 bulk write request, because the grant info while be sync between
3270 OSC and OST when reconnect, and we should ignore the grant info
3271 these of resend/replay write req.
3276 Description: Grant space more than available left space sometimes.
3277 Details : When then OST is about to be full, if two bulk writing from
3278 different clients came to OST. Accord the avaliable space of the
3279 OST, the first req should be permitted, and the second one
3280 should be denied by ENOSPC. But if the seconde arrived before
3281 the first one is committed. The OST might wrongly permit second
3282 writing, which will cause grant space > available space.
3285 Frequency : when client is evicted
3287 Description: Grant might be wrongly erased when osc is evicted by OST
3288 Details : when the import is evicted by server, it will fork another
3289 thread ptlrpc_invalidate_import_thread to invalidate the
3290 import, where the grant will be set to 0. While the original
3291 thread will update the grant it got when connecting. So if
3292 the former happened latter, the grant will be wrongly errased
3293 because of this race.
3298 Description: Checking Stale with correct fid
3299 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3300 to check whether it is stale, because sometimes, we want the
3301 enqueue happened anyway, and op_data.fid2 will not be initialized.
3303 Severity : enhancement
3305 Description: update patchless client
3306 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3309 Frequency : only with 2.4 kernel
3311 Description: random memory corruption
3312 Details : size of struct ll_inode_info is to big for union inode.u and this
3313 can be cause of random memory corruption.
3318 Description: Memory leak in recovery
3319 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3320 It should also check obd_fail before fsfilt_start, otherwise if
3321 fsfilt_start return -EROFS,(failover mds during mds recovery).
3322 then the req will return with repmsg->transno = 0 and rc = EROFS.
3323 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3324 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3325 be freed no matter whether fsfilt_commit success or not.
3328 Frequency : only with huge count clients
3330 Description: Prevents from taking the superblock lock in llap_from_page for
3332 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3333 for llap_from_page call in ll_removepage prevents from taking the
3334 superblock lock for a soon died page.
3339 Description: Not check open intent error before release open handle
3340 Details : in some rare cases, the open intent error is not checked before
3341 release open handle, which may cause
3342 ASSERTION(open_req->rq_transno != 0), because it tries to release
3343 the failed open handle.
3348 Description: Set cat log bitmap only after create log success.
3349 Details : in some rare cases, the cat log bitmap is set too early. and it
3350 should be set only after create log success.
3354 Description: Accessing a block bevice can re-enable I/O when Lustre is
3355 tearing down a device.
3356 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3361 Description: service threads may hog cpus when there are a lot of requests
3362 Details : Insert cond_resched to give other threads a chance to use some CPU
3367 Description: the cat log was not initialized in recovery
3368 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3369 unlink log on mds will not be initialized until mds post
3370 recovery. And also in mds post recovery, the unlink log will
3371 initialization will be done asynchronausly, so there will be race
3372 between add unlink log and unlink log initialization.
3376 Description: brw_stats were being printed incorrectly
3377 Details : brw_stats were being printed as log2 but all of them were not
3378 recorded as log2. Also remove some code duplication arising from
3379 filter_tally_{read,write}.
3383 Frequency : rare, only in recovery.
3384 Description: ASSERTION(req->rq_type != LI_POISON) failed
3385 Details : imp_lock should be held while iterating over imp_sending_list for
3386 prevent destroy request after get timeout in ptlrpc_queue_wait.
3390 Description: replay-single.sh test 52 fails
3391 Details : A lock's skiplist need to be cleanup when it being unlinked
3392 from its resource list.
3396 Description: Short directio read returns full requested size rather than
3398 Details : Direct I/O operations should return actual amount of bytes
3399 transferred rather than requested size.
3401 Severity : enhancement
3403 Description: metadata RPC reduction (e.g. for rm performance)
3404 Details : decrease the amount of synchronous RPC between clients and servers
3405 by canceling conflicing lock before the operation on the client
3406 and packing thier handles into the main operation RPC to server.
3408 Severity : enhancement
3410 Description: Async OSC create to avoid the blocking unnecessarily.
3411 Details : If an OST has no remaining object, system will block on the
3412 creation when it needs to create a new object on this OST. Now,
3413 ways use pre-created objects when available, instead of blocking on
3414 an empty osc while others are not empty. If we must block, we
3415 block for the shortest possible period of time.
3419 Description: improve handling recoverable errors
3420 Details : if request processig with error which can be recoverable on server
3421 request should be resend, otherwise page released from cache and
3424 Severity : enhancement
3426 Description: refine locking for avoid write wrong info into lov_objid file
3427 Details : fix possible races with add new target and write/update data in
3430 --------------------------------------------------------------------------------
3432 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3437 Frequency : on some architectures
3439 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3440 Details : Uninitialized flags sometimes cause configuration commands to
3444 Frequency : patchless clients only
3446 Description: missing __iget() symbol export
3447 Details : The __iget() symbol export is missing. To avoid the need for
3448 this on patchless clients the deathrow inode reaper is turned
3449 off, and we depend on the VM to clean up old inodes. This
3450 dependency was during via the fix for bug 12181.
3452 --------------------------------------------------------------------------------
3454 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3456 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3457 INTEROPERATE with older versions automatically. In many cases a
3458 special upgrade step is needed. Please read the
3459 user documentation before upgrading any part of a live system.
3460 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3461 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3462 * WARNING: Lustre configuration and startup changes are required with
3463 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3467 Severity : enhancement
3469 Description: MountConf
3470 Details : Lustre configuration is now managed via mkfs and mount
3471 commands instead of lmc and lconf. New obd types (MGS, MGC)
3472 are added for dynamic configuration management. See
3473 https://mail.clusterfs.com/wikis/lustre/MountConf for
3476 Severity : enhancement
3478 Description: dynamic OST addition
3479 Details : OSTs can now be added to a live filesystem
3481 Severity : enhancement
3483 Description: startup order invariance
3484 Details : MDTs and OSTs can be started in any order. Clients only
3485 require the MDT to complete startup.
3487 Severity : enhancement
3489 Description: parallel, asynchronous orphan cleanup
3490 Details : orphan cleanup is now performed in separate threads for each
3491 OST, allowing parallel non-blocking operation.
3493 Severity : enhancement
3495 Description: optimized stripe assignment
3496 Details : stripe assignments are now made based on ost space available,
3497 ost previous usage, and OSS previous usage, in order to try
3498 to optimize storage space and networking resources.
3500 Severity : enhancement
3502 Description: Permanently set tunables
3503 Details : All writable /proc/fs/lustre tunables can now be permanently
3504 set on a per-server basis, at mkfs time or on a live system.
3506 Severity : enhancement
3508 Description: Lustre message v2
3509 Details : Add lustre message format v2.
3511 Severity : enhancement
3513 Description: client OST exclusion list
3514 Details : Clients can be started with a list of OSTs that should be
3515 declared "inactive" for known non-responsive OSTs.
3519 Description: SPEC SFS validation failure on NFS v2 over lustre.
3520 Details : Changes the blocksize for regular files to be 2x RPC size,
3521 and not depend on stripe size.
3523 Severity : enhancement
3525 Description: Multiple MD RPCs in flight.
3526 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3527 To avoid overly-overloading MDS, introduce a limit on number of
3528 MDS RPCs in flight for a single client and add /proc controls
3529 to adjust this limit.
3531 Severity : enhancement
3533 Description: client read/write statistics
3534 Details : Add client read/write call usage stats for performance
3535 analysis of user processes.
3536 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3537 file access. extents_stats shows chunk size distribution.
3538 extents_stats_per_process show chunk size distribution per
3541 Severity : enhancement
3543 Description: mds statistics
3544 Details : Add detailed mds operations statistics in
3545 /proc/fs/lustre/mds/*/stats.
3549 Description: Failure of copying files with lustre special EAs.
3550 Details : Client side always return success for setxattr call for lustre
3551 special xattr (currently only "trusted.lov").
3556 Description: Refcount LNET uuids
3557 Details : The global LNET uuid list grew linearly with every startup;
3558 refcount repeated list entries instead of always adding to
3561 Severity : enhancement
3563 Description: Dynamic service threads
3564 Details : Within a small range, start extra service threads
3565 automatically when the request queue builds up.
3568 Frequency : mixed-endian client/server environments
3570 Description: mixed-endian crashes
3571 Details : The new msg_v2 system had some failures in mixed-endian
3574 Severity : enhancement
3576 Description: Easy OST removal
3577 Details : OSTs can be permanently deactivated with e.g. 'lctl
3578 conf_param lustre-OST0001.osc.active=0'
3580 Severity : enhancement
3582 Description: MGS proc entries
3583 Details : Added basic proc entries for the MGS showing what filesystems
3586 Severity : enhancement
3588 Description: provide MGS failover
3589 Details : Added config lock reacquisition after MGS server failover.
3591 Severity : enhancement
3593 Description: add Linux 2.4 support
3594 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3598 Description: a large application tries to do I/O to the same resource and dies
3599 in the middle of it.
3600 Details : Check the req->rq_arrival time after the call to
3601 ost_brw_lock_get(), but before we do anything about
3602 processing it & sending the BULK transfer request. This
3603 should help move old stale pending locks off the queue as
3604 quickly as obd_timeout.
3607 Frequency : when an incorrect nid is specified during startup
3609 Description: ptlrpc connect to non-existant node causes kernel crash
3610 Details : LNET can't be re-entered from an event callback, which
3611 happened when we expire a message after the export has been
3612 cleaned up. Instead, hand the zombie cleanup off to another
3615 Severity : enhancement
3617 Description: plain/inodebits lock performance improvement
3618 Details : Grouping plain/inodebits in granted list by their request modes
3619 and bits policy, thus improving the performance of search through
3623 Frequency : only if OST filesystem is corrupted
3625 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3626 Details : for a short time RPCs with bulk IO are in the replay list,
3627 but replay of bulk IOs is unimplemented. If the OST filesystem
3628 is corrupted due to disk cache incoherency and then replay is
3629 started it is possible to trip an assertion. Avoid putting
3630 committed RPCs into the replay list at all to avoid this issue.
3633 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3635 Description: System hang on startup
3636 Details : This bug allowed the liblustre (e.g. catamount) client to
3637 return to the app before handling all startup RPCs. This
3638 could leave the node unresponsive to lustre network traffic
3639 and manifested as a server ptllnd timeout.
3641 Severity : enhancement
3643 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3644 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3645 boolean to control whether to print peer debug info when a
3646 client's RPC times out.
3649 Frequency : only for kernels with patches from Lustre below 1.4.3
3651 Description: Remove old rdonly API
3652 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3655 Frequency : only for devices with external journals
3657 Description: Set external device read-only also
3658 Details : During a commanded failover stop, we set the disk device
3659 read-only while the server shuts down. We now also set any
3660 external journal device read-only at the same time.
3663 Frequency : when upgrading from 1.4 while trying to change parameters
3665 Description: The wrong (new) MDC name was used when setting parameters for
3666 upgraded MDT's. Also allows changing of OSC (and MDC)
3667 parameters if --writeconf is specified at tunefs upgrade time.
3670 Frequency : when setting specific ost indicies
3672 Description: QOS code breaks on skipped indicies
3673 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3674 created with --index need not be sequential.
3676 Severity : enhancement
3678 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3679 Details : The uninit_groups feature works in conjunction with the kernel
3680 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3681 up the pass1 processing of e2fsck. This is a read-only feature
3682 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3683 mount filesystems that have had this feature enabled.
3685 Severity : enhancement
3687 Description: Improve multi-block allocation algorithm to avoid fragmentation
3688 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3689 allocation locality and avoid filesystem fragmentation.
3691 ------------------------------------------------------------------------------
3693 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3695 * Support for kernels:
3696 2.4.21-47.0.1.EL (RHEL 3)
3697 2.6.5-7.283 (SLES 9)
3698 2.6.9-42.0.10.EL (RHEL 4)
3699 2.6.12.6 vanilla (kernel.org)
3700 2.6.16.27-0.9 (SLES 10)
3701 * Recommended e2fsprogs version: 1.39.cfs5
3703 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3707 Frequency : occasional, depends on client load and configuration
3708 Bugzilla : 12181, 12203
3709 Description: data loss for recently-modified files
3711 Details : In some cases it is possible that recently written or created
3712 files may not be written to disk in a timely manner (this should
3713 normally be within 30s unless client IO load is very high).
3714 The problem appears as zero-length files or files that are a
3715 multiple of 1MB in size after a client crash or client eviction
3716 that are missing data at the end of the file.
3718 This problem is more likely to be hit on clients where files are
3719 repeatedly created and unlinked in the same directory, clients
3720 have a large amount of RAM, have many CPUs, the filesystem has
3721 many OSTs, the clients are rebooted frequently, and/or the files
3722 are not accessed by other nodes after being written.
3724 The presence of the problem can be detected by looking at
3725 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3726 smaller than the second (nr_unused) then dirty files will not
3727 be flushed automatically to disk. "sync; sleep 10" should be
3728 run several times on the node before unmounting it to update
3729 Lustre (this is also safe to run on nodes without this problem).
3731 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3732 that can cause this same problem, so customers using that kernel
3733 also need to update the kernel in addition to Lustre. In order
3734 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3736 It is normal that files written just before a client crash (less
3737 than 30s) may not yet have been flushed to disk, even for local
3741 Frequency : frequent on thin XT3 nodes
3743 Description: UUID collision on thin XT3 Linux nodes
3744 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3745 randomly, since we relied on an insufficiently-seeded PRNG.
3750 Description: OSS hangs after "All ost request buffers busy"
3751 Details : A deadlock between quota and journal operations caused OSS
3752 hangs after printing "All ost request buffers busy."
3755 Frequency : always on liblustre builds
3757 Description: Cleanup compiler warnings on liblustre
3760 Frequency : always on liblustre builds on XT3
3762 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3766 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3767 > 32000 subdirectories
3768 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3769 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3770 subdirectory count crosses 32000. This will aid e2fsck to
3771 correctly handle more than 32000 subdirectories.
3774 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3776 Description: System hang on startup
3777 Details : This bug allowed the liblustre (e.g. catamount) client to
3778 return to the app before handling all startup RPCs. This
3779 could leave the node unresponsive to lustre network traffic
3780 and manifested as a server ptllnd timeout.
3782 Severity : enhancement
3784 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3785 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3786 boolean to control whether to print peer debug info when a
3787 client's RPC times out.
3792 Description: make O_SYNC working on 2.6 kernels
3793 Details : 2.6 kernels use different method for mark pages for write,
3794 so need add a code to lustre for O_SYNC work.
3799 Description: Failure to close file and release space on NFS
3800 Details : Put inode details into lock acquired in ll_intent_file_open.
3801 Use mdc_intent_lock in ll_intent_open to properly
3802 detect all kind of errors unhandled by mdc_enqueue.
3807 Description: proc file read during shutdown sometimes raced obd removal,
3809 Details : Add lock to prevent obd access after proc file removal.
3812 Frequency : Only for files larger than 4GB on 32-bit clients.
3814 Description: improperly doing page alignment of locks
3815 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3816 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3821 Description: RPCs being resent when they shouldn't be
3822 Details : Some RPCs that should not be resent are being resent. This
3823 can cause inconsistencies in the RPC state machine. Do not
3824 resend such requests.
3827 Frequency : rare, only with NFS export
3829 Description: Crash on NFS re-export node
3830 Details : under very unusual load conditions an assertion is hit in
3831 ll_intent_file_open()
3834 Frequency : only if OST filesystem is corrupted
3836 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3837 Details : for a short time RPCs with bulk IO are in the replay list,
3838 but replay of bulk IOs is unimplemented. If the OST filesystem
3839 is corrupted due to disk cache incoherency and then replay is
3840 started it is possible to trip an assertion. Avoid putting
3841 committed RPCs into the replay list at all to avoid this issue.
3846 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3847 Details : Large single O_DIRECT read and write calls can fail to allocate
3848 a sufficiently large buffer to process the request. In case of
3849 allocation failure the allocation is retried with a smaller
3850 buffer and broken into smaller requests.
3852 Severity : enhancement
3854 Description: Add -o localflock option to simulate old noflock behaviour.
3855 Details : This will achieve local-only flock/fcntl locks coherentness.
3860 Description: versioning check is incomplete
3861 Details : Checking the version difference of client vs. server, report
3862 error if the gap is too big.
3867 Description: add support PG_writeback bit
3868 Details : add support for PG_writeback bit for Lustre, for more carefull
3869 work with page cache in 2.6 kernel. This also fix some deadlocks
3870 and remove hack for work O_SYNC with 2.6 kernel.
3872 Severity : enhancement
3874 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3875 Details : The uninit_groups feature works in conjunction with the kernel
3876 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3877 up the pass1 processing of e2fsck. This is a read-only feature
3878 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3879 mount filesystems that have had this feature enabled.
3881 Severity : enhancement
3883 Description: Improve multi-block allocation algorithm to avoid fragmentation
3884 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3885 allocation locality and avoid filesystem fragmentation.
3887 ------------------------------------------------------------------------------
3889 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3891 * Support for kernels:
3892 2.6.9-42.0.3.EL (RHEL 4)
3893 2.6.5-7.276 (SLES 9)
3894 2.4.21-47.0.1.EL (RHEL 3)
3895 2.6.12.6 vanilla (kernel.org)
3896 2.6.16.21-0.8 (SLES10)
3897 * Recommended e2fsprogs version: 1.39.cfs2-0
3899 * The backwards-compatible /proc/sys/portals symlink has been removed
3900 in this release. Before upgrading, please ensure that you change
3901 any configuration scripts or /etc/sysctl.conf files that access
3902 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3903 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3904 in advance of the upgrade on any system running Lustre 1.4.6 or
3905 newer, since /proc/sys/lnet was added in that version.
3906 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3910 Frequency : only when quota is used
3912 Description: avoid scanning export list for quota master
3913 Details : Change the algorithms to avoid scanning export list in order
3914 to improve the efficiency.
3917 Frequency : MDS failover only, very rarely
3919 Description: "went back in time" messages on mds failover
3920 Details : The greatest transno may be lost when the current operation
3921 finishes with an error (transno==0) and the client's last_rcvd
3922 record is over-written. Save the greatest transno in the
3923 mds_last_transno for this case.
3926 Frequency : always for specific kernels and striping counts
3928 Description: client may get "Matching packet too big" without ACL support
3929 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3930 when trying to access files in certain configurations. The
3931 clients should in fact be denied when mounting because they do
3932 not understand ACLs.
3935 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3937 Description: many clients connecting with IO in progress causes connect timeouts
3938 Details : Avoid synchronous journal commits to avoid delays caused by many
3939 clients connecting/disconnecting when bulk IO is in progress.
3940 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3941 OST_IO_PORTAL to avoid delays behind potentially many pending
3945 Frequency : occasionally with multiple writers to a single file
3947 Description: shared writes to file may result in wrong size reported by stat()
3948 Details : Allow growing of kms when extent lock is cancelled
3951 Frequency : always with random mmap IO to multi-striped file
3953 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3954 Details : Only if the hole is at the end of OST object so that kms is too
3955 small. Fix is to increase kms accordingly in ll_nopage.
3958 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3960 Description: writes to a missing object would leak memory on the OST
3961 Details : If there is an inconsistency between the MDS and OST filesystems,
3962 such that the MDS references an object that doesn't exist, writes
3963 to that object will leak memory due to incorrect cleanup in the
3964 error handling path, eventually running out of memory on the OST.
3969 Description: Creating too long symlink causes lustre errors
3970 Details : Check symlink and name lengths before sending requests to MDS.
3973 Frequency : only if flock is enabled (not on by default)
3975 Description: posix locks not released on fd closure on 2.6.9+
3976 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3977 kernels, this caused such locks not to be released on fd close and
3978 then assertions on fs unmount about still used locks.
3981 Frequency : MDS failover only, very rarely
3983 Description: clients may get ASSERTION(granted_lock != NULL)
3984 Details : When request was taking a long time, and a client was resending
3985 a getattr by name lock request. The were multiple lock requests
3986 with the same client lock handle and
3987 mds_getattr_name->fixup_handle_for_resent_request found one of the
3988 lock handles but later failed with ASSERTION(granted_lock != NULL).
3993 Description: handle->h_buffer_credits > 0, assertion failure
3994 Details : h_buffer_credits is zero after truncate, causing assertion
3995 failure. This patch extends the transaction or creates a new
3999 Frequency : NFS re-export or patchless client
4000 Bugzilla : 11179, 10796
4001 Description: Crash on NFS re-export node (__d_move)
4002 Details : We do not want to hash the dentry if we don't have a lock.
4003 But if this dentry is later used in d_move, we'd hit uninitialised
4004 list head d_hash, so we just do this to init d_hash field but
4005 leave dentry unhashed.
4008 Frequency : NFS re-export or patchless client
4010 Description: NFS exports has problem with symbolic link
4011 Details : lustre client didn't properly install dentry when re-exported
4012 to NFS or running patchless client.
4015 Frequency : NFS re-export or patchless client
4017 Description: Various nfs/patchless fixes.
4018 Details : fixes reuse disconected alias for lookup process - this fixes
4019 warning "find_exported_dentry: npd != pd",
4020 fix permission error with open files at nfs.
4021 fix apply umask when do revalidate.
4024 Frequency : occasional
4026 Description: Crash on NFS re-export node
4027 Details : calling clear_page() on the wrong pointer triggered oops in
4028 generic_mapping_read().
4031 Frequency : rarely, using O_DIRECT IO
4033 Description: unaligned directio crashes client with LASSERT
4034 Details : check for unaligned buffers before trying any requests.
4037 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4039 Description: stale data returned from RAID cache
4040 Details : If only a small amount of IO is done to the RAID device before
4041 reading it again it is possible to get stale data from the RAID
4042 cache instead of reading it from disk.
4045 Frequency : always for sles10 kernel
4047 Description: sles10 support
4048 Details : ll_follow_link: compile fixes and using of nd_set_link
4049 under newer kernels.
4052 Frequency : depends on arch, kernel and compiler version, always on sles10
4055 Description: recursive or deep enough symlinks cause stack overflow
4056 Details : getting rid of large stack-allocated variable in
4060 Frequency : depends on hardware
4062 Description: lustre write performance loss in the SLES10 kernel
4063 Details : the performance loss is caused by using of write barriers in the
4064 ext3 code. The SLES10 kernel turns barrier support on by
4065 default. The fix is to undo that change for ldiskfs.
4067 ------------------------------------------------------------------------------
4069 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4071 * Support for kernels:
4072 2.6.9-42.0.3EL (RHEL 4)
4073 2.6.5-7.276 (SLES 9)
4074 2.4.21-47.0.1.EL (RHEL 3)
4075 2.6.12.6 vanilla (kernel.org)
4079 Frequency : quota enabled and large files being deleted
4081 Description: releasing more than 4GB of quota at once hangs OST
4082 Details : If a user deletes more than 4GB of files on a single OST it
4083 will cause the OST to spin in an infinite loop. Release
4084 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4089 Description: statfs data retrieved from /proc may be stale or zero
4090 Details : When reading per-device statfs data from /proc, in the
4091 {kbytes,files}_{total,free,avail} files, it may appear
4092 as zero or be out of date.
4095 Frequency : systems with MD RAID1 external journal devices
4097 Description: lconf's call to blkid is confused by RAID1 journal devices
4098 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4099 of returning all block devices that match the journal UUID.
4102 Frequency : always, for aggregate stripe size over 4GB
4104 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4105 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4106 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4109 Frequency : always if "lfs find" used on a local file/directory
4111 Description: "lfs find" segfaults if used on a local file/directory
4112 Details : The case where a directory component was not specified wasn't
4113 handled correctly. Handle this properly.
4116 Frequency : always on ppc64
4118 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4119 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4120 perform endianness conversion.
4123 Frequency : rarely (truncate to non-zero file size after write under load)
4124 Bugzilla : 10730, 10687
4125 Description: Files padded with zeros to next 4K multiple
4126 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4127 it is possible that files that are truncated to a non-zero size
4128 immediately after being written are filled with zero bytes beyond
4129 the truncated size. No file data is lost.
4131 Severity : enhancement
4133 Description: Allow recovery/failover for liblustre clients.
4134 Details : liblustre clients were unaware of failover configurations until
4137 Severity : enhancement
4139 Description: user file locks should fail when not mounting with flock option
4140 Details : Set up an error-returning stub in ll_file_operations.lock field
4141 to prevent incorrect behaviour when client is mounted without
4142 flock option. Also, set up properly f_op->flock field for
4146 Frequency : always on ia64
4148 Description: "lfs df" loops on printing out MDS statfs information
4149 Details : The obd_ioctl_data was not initialized and in some systems
4150 this caused a failure during the ioctl that did not return
4151 an error. Initialize the struct and return an error on failure.
4154 Frequency : SLES 9 only
4156 Description: Error of copying files with lustre special EAs as root
4157 Details : Client side always return success for setxattr call for lustre
4158 special xattr (currently only "trusted.lov").
4161 Frequency : rarely on clusters with both ia64+i386 clients
4163 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4164 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4165 clients are doing concurrent writes to the same file it is
4166 possible that the ia64 clients may LASSERT because the OST
4167 extent locks are not PAGE_SIZE aligned. Ensure that grown
4168 locks are always aligned on the request boundary.
4171 Frequency : specific use, occasional
4173 Description: Overwriting in use executable truncates on-disk binary image
4174 Details : If one node attempts to overwrite an executable in use by
4175 another node, we now correctly return ETXTBSY instead of
4176 truncating the file.
4178 Severity : enhancement
4180 Description: Async OSC create to avoid the blocking unnecessarily.
4181 Details : If an OST has no remaining object, system will block on the creating
4182 when need to create a new object on this OST. Now, ways use
4183 pre-created objects when available, instead of blocking on an
4184 empty osc while others are not empty. If we must block, we block
4185 for the shortest possible period of time.
4190 Description: chmod on Lustre root is propagated to other clients
4191 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4192 invalid by the follow_mount time.
4195 Frequency : liblustre clients only
4197 Description: Race in 'instant cancel' lock handling could lead to such locks
4198 never to be granted in case of SMP MDS
4199 Details : Do not destroy not yet granted but cbpending locks in
4203 Frequency : replay/resend of open
4205 Description: non null lock assetion failure in mds_intent_policy
4206 Details : Trying to replay/resend lockless open requests resulted in
4207 mds_open() returning 0 with no lock. Now it sets a flag if
4208 a lock is going to be returned.
4210 Severity : enhancement
4212 Description: Checksum enhancements
4213 Details : New checksum enhancements allow for resending RPCs that failed
4216 Severity : enhancement
4218 Description: Tunables on number of dirty pages in cacche
4219 Details : Allow to set limit on number of dirty pages cached.
4224 Description: client crash on unmount - lock still has references
4225 Details : In some error handling cases it was possible to leak a lock
4226 reference on a client while accessing a file. This was not
4227 harmful to the client during operation, but would cause the
4228 client to crash when the filesystem is unmounted.
4231 Frequency : specific case, rare
4233 Description: ETXTBSY on mds though file not in use by client
4234 Details : ETXTBSY is no longer incorrectly returned when attempting to
4235 chmod or chown a directory that the user previously tried to
4236 execute or a currently-executing binary.
4239 Frequency : extremely rare except on liblustre-based clients
4241 Description: Lustre space not freed when files are deleted
4242 Details : Clean up open-unlinked files after client eviction. Previously
4243 the unlink was skipped and the files remained as orphans.
4248 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4249 Details : In some cases it was possible to send a blocking callback to
4250 a client doing a glimpse, even though that client didn't get
4251 a lock granted. When the glimpse lock is cancelled on the OST
4252 the freed lock is left on the waiting list and corrupted the list.
4255 Frequency : all core dumps
4257 Description: Broke core dumps to lustre
4258 Details : Negative dentry may be unhashed if parent does not have UPDATE
4259 lock, but some callers, e.g. do_coredump, expect dentry to be
4260 hashed after successful create, hash it in ll_create_it.
4262 ------------------------------------------------------------------------------
4264 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4266 * Support for kernels:
4267 2.6.9-42.0.2.EL (RHEL 4)
4268 2.6.5-7.276 (SLES 9)
4269 2.4.21-40.EL (RHEL 3)
4270 2.6.12.6 vanilla (kernel.org)
4274 Frequency : always on RHEL 3
4276 Description: Number of open files grows over time
4277 Details : The number of open files grows over time, whether or not
4278 Lustre is started. This was due to a filp leak introduced
4279 by one of our kernel patches.
4281 ------------------------------------------------------------------------------
4283 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4285 * Support for kernels:
4286 2.6.9-42.EL (RHEL 4)
4287 2.6.5-7.276 (SLES 9)
4288 2.4.21-40.EL (RHEL 3)
4289 2.6.12.6 vanilla (kernel.org)
4294 Bugzilla : 5719, 9635, 9792, 9684
4295 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4296 Details : If a server is under heavy load and cannot reply to new
4297 connection requests before the client resends the (re)connect,
4298 the connection handling code can behave badly if two service
4299 threads are concurrently handing separate (re)connections from
4300 the same client. Add better locking to the connection handling
4301 code, and ensure that only a single connection will be processed
4302 for a given client UUID, even if the lock is dropped.
4304 Severity : enhancement
4306 Description: add TCP zero-copy support to kernel
4307 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4308 sends if the hardware supports scatter-gather and checksumming.
4309 This allows socklnd to do client-write and server-read more
4310 efficiently and reduce CPU utilization from skbuf copying.
4313 Frequency : only if NFS exporting from client
4315 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4316 Details : If setting attributes on a file created under NFS that had
4317 never been opened it would be possible to oops the client
4318 if the file had no objects.
4321 Frequency : always for liblustre
4323 Description: liblustre client does MDS+OSTs setattr RPC for each write
4324 Details : When doing a write from a liblustre client, the client
4325 incorrectly issued an RPC to the MDS and each OST the file was
4326 striped over in order to update the timestamps. When writing
4327 with small chunks and many clients this could overwhelm the MDS
4328 with RPCs. In all cases it would slow down the write because
4329 these RPCs are unnecessary.
4331 Severity : enhancement
4333 Description: allow number of MDS service threads to be changed at module load
4334 Details : It is now possible to change the number of MDS service threads
4335 running. Adding "options mds mds_num_threads={N}" to the MDS's
4336 /etc/modprobe.conf will set the number of threads for the next
4337 time Lustre is restarted (assuming the "mds" module is also
4338 reloaded at that time). The default number of threads will
4339 stay the same, 32 for most systems.
4344 Description: OST crash if filesystem is unformatted or corrupt
4345 Details : If an OST is started on a device that has never been formatted
4346 or if the filesystem is corrupt and cannot even mount then the
4347 error handling cleanup routines would dereference a NULL pointer.
4352 Description: NULL pointer deref in llap_from_page.
4353 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4354 incorrect) mapping if the page was truncated/reclaimed while it was
4355 searched for. Check for this condition and skip such pages when
4356 doing readahead. Introduce extra check to llap_from_page() to
4357 verify page->mapping->host is non-NULL (so page is not anonymous).
4360 Frequency : Sometimes when using sys_sendfile
4362 Description: "page not covered by a lock" warnings from ll_readpage
4363 Details : sendfile called ll_readpage without right page locks present.
4364 Now we introduced ll_file_sendfile that does necessary locking
4365 around call to generic_file_sendfile() much like we do in
4369 Frequency : with certain MDS communication failures at client mount time
4371 Description: NULL pointer deref after failed client mount
4372 Details : a client connection request may delayed by the network layer
4373 and not be sent until after the PTLRPC layer has timed out the
4374 request. If the client fails the mount immediately it will try
4375 to clean up before the network times out the request. Add a
4376 reference from the request import to the obd device and delay
4377 the cleanup until the network drops the request.
4380 Frequency : occasionally during client (re)connect
4382 Description: assertion failure during client (re)connect
4383 Details : processing a client connection request may be delayed by the
4384 client or server longer than the client connect timeout. This
4385 causes the client to resend the connection request. If the
4386 original connection request is replied in this interval, the
4387 client may trip an assertion failure in ptlrpc_connect_interpret()
4388 which thought it would be the only running connect process.
4391 Frequency : only with obd_echo servers and clients that are rebooted
4393 Description: kernel BUG accessing uninitialized data structure
4394 Details : When running an obd_echo server it did not start the ping_evictor
4395 thread, and when a client was evicted an uninitialized data
4396 structure was accessed. Start the ping_evictor in the RPC
4397 service startup instead of the OBD startup.
4399 Severity : enhancement
4400 Bugzilla : 10193 (patchless)
4401 Description: Remove dependency on various unexported kernel interfaces.
4402 Details : No longer need reparent_to_init, exit_mm, exit_files,
4403 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4406 Frequency : rare (only users of deprecated and unsupported LDAP config)
4408 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4409 Details : LDAP apparently contains 'lustreName' attributes instead of
4410 'name'. A simple remapping of the name is sufficient.
4413 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4415 Description: waiting_locks_callback trips kernel BUG if client is evicted
4416 Details : Running with the dump_on_timeout debug flag turned on makes
4417 it possible that the waiting_locks_callback() can try to dump
4418 the Lustre kernel debug logs from an interrupt handler. Defer
4419 this log dumping to the expired_lock_main() thread.
4421 Severity : enhancement
4423 Description: Support NFS exporting on 2.6 kernels.
4424 Details : Implement non-rawops metadata methods for NFS server to use without
4425 changing NFS server code.
4428 Frequency : very rare (synthetic metadata workload only)
4430 Description: two racing renames might cause an MDS thread to deadlock
4431 Details : Running the "racer" program may cause one MDS thread to rename
4432 a file from being the source of a rename to being the target of
4433 a rename at exactly the same time that another thread is doing
4434 so, and the second thread has already enqueued these locks after
4435 doing a lookup of the target and is trying to relock them in
4436 order. Ensure that we don't try to re-lock the same resource.
4439 Frequency : only very large systems with liblustre clients
4441 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4442 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4443 all OSTs in parallel. This allows the eviction of stale liblustre
4444 clients to proceed much faster than if they were done in series,
4445 and also offers similar improvements for other set_info RPCs.
4450 Description: excessive CPU usage during initial read phase on client
4451 Details : During the initial read phase on a client, it would agressively
4452 retry readahead on the file, consuming too much CPU and impacting
4453 performance (since 1.4.5.8). Improve the readahead algorithm
4454 to avoid this, and also improve some other common cases (read
4455 of small files in particular, where "small" is files smaller than
4456 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4461 Description: MDS crash when receiving packet with unknown intent.
4462 Details : Do not LBUG in unknown intent case, just return -EFAULT
4464 Severity : enhancement
4465 Bugzilla : 9293, 9385
4466 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4467 Details : Do not serialize getattr (non-intent version) and statfs.
4470 Frequency : occasional, when OST network is overloaded/intermittent
4472 Description: client evicted by OST after bulk IO timeout
4473 Details : If a client sends a bulk IO request (read or write) the OST
4474 may evict the client if it is unresposive to its data GET/PUT
4475 request. This is incorrect if the network is overloaded (takes
4476 too long to transfer the RPC data) or dropped the OST GET/PUT
4477 request. There is no need to evict the client at all, since
4478 the pinger and/or lock callbacks will handle this, and the
4479 client can restart the bulk request.
4482 Frequency : Always when mmapping file with no objects
4484 Description: client crashes when mmapping file with no objects
4485 Details : Check that we actually have objects in a file before doing any
4486 operations on objects in ll_vm_open, ll_vm_close and
4492 Description: Request leak when working with deleted CWD
4493 Details : Introduce advanced request refcount tracking for requests
4494 referenced from lustre intent.
4496 Severity : Enhancement
4498 Description: Cache open file handles on client.
4499 Details : MDS now will return special lock along with openhandle, if
4500 requested and client is allowed to hold openhandle, even if unused,
4501 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4502 closing files for every read/write openration.
4504 Severity : Enhancement
4506 Description: Cache open negative dentries on client when possible.
4507 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4508 negative dentries on lock revocation.
4513 Description: Remounting a client read-only wasn't possible with a zconf mount
4514 Details : It wasn't possible to remount a client read-only with llmount.
4516 Severity : enhancement
4517 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4518 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4519 MPICH is included to add Lustre-specific ADIO interfaces.
4520 This is based closely on the UFS ADIO layer and only differs
4521 in file creation, in order to allow the OST striping to be set.
4522 This is user-contributed code and not supported by CFS.
4527 Description: extended inode attributes (immutable, append-only) work improperly
4528 when 2.4 and 2.6 kernels are used on client/server or vice versa
4529 Details : Introduce kernel-independent values for these flags.
4531 Severity : enhancement
4534 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4535 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4536 granularity. Now, Fractional MB granularity is be supported,
4537 this is very useful for low memory system.
4539 Severity : enhancement
4541 Description: Getattr by fid
4542 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4543 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4544 have locks and avoids getattr rpc requests for it on every stat.
4547 Frequency : Always, for filesystems larger than 2TB
4549 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4550 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4551 filesystems larger than 2TB to be mounted with mballoc enabled.
4554 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4556 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4557 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4558 filesystems larger than 2TB is mounted then the kernel will
4559 silently corrupt the start of the filesystem. CONFIG_LBD is
4560 enabled for all CFS-supported kernels, but the possibility of
4561 this happening with a modified kernel config exists.
4563 Severity : enhancement
4565 Description: add client O_DIRECT support for 2.6 kernels
4566 Details : It is now possible to do O_DIRECT reads and writes to files
4567 in the Lustre client mountpoint on 2.6 kernel clients.
4569 Severity : enhancement
4571 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4572 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4573 parallel, not waiting for response from every OST before sending
4574 a rpc to the next OST.
4579 Description: setattr vs write race when updating file timestamps
4580 Details : Client processes that update a file timestamp into the past
4581 right after writing to the file (e.g. tar) it is possible that
4582 the updated file modification time can be reset to the current
4583 time due to a race between processing the setattr and write RPC.
4585 Severity : enhancement
4587 Description: Bring 'lfs find' closer in line with regular Linux find.
4588 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4589 -print0 options and obtains all the needed info through the lustre
4592 Severity : enhancement
4594 Description: support up to 1024 configured devices on one node
4595 Details : change obd_dev array from statically allocated to dynamically
4596 allocated structs as they are first used to reduce memory usage
4601 Description: Flush dirty partially truncated pages during truncate
4602 Details : Immediatelly flush partially truncated pages in filter_setattr,
4603 this way we completely avoid having any pages in page cache on OST
4604 and can retire ugly workarounds during writes to flush such pages.
4609 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4610 Details : protect inode from truncation within vfs_unlink() context
4611 just take a reference before calling vfs_unlink() and release it
4612 when parent's i_sem is free.
4617 Description: last_id value checked outside lock on OST caused LASSERT failure
4618 Details : If there were multiple MDS->OST object precreate requests in
4619 flight, it was possible that the OST's last object id was checked
4620 outside a lock and incorrectly tripped an assertion. Move checks
4621 inside locks, and discard old precreate requests.
4624 Frequency : always, if extents are used on OSTs
4626 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4627 Details : OSTs using the extents format would not zero the high 16 bits of
4628 the index physical block number. This is not a problem for any
4629 OST filesystems smaller than 16TB, and no kernels support ext3
4630 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4631 new/modified files) and can be fixed for existing filesystems
4632 with e2fsprogs-1.39-cfs1.
4637 Description: import connection selection may be incorrect if timer wraps
4638 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4639 import connections to be ignored if the 32-bit jiffies counter
4640 wraps. Use a 64-bit jiffies counter.
4643 Frequency : very large clusters immediately after boot
4645 Description: LNET request buffers exhausted under heavy short-term load
4646 Details : If a large number of client requests are generated on a service
4647 that has previously never seen so many requests it is possible
4648 that the request buffer growth cannot keep up with the spike in
4649 demand. Instead of dropping incoming requests, they are held in
4650 the LND until the RPC service can accept more requests.
4653 Frequency : Sometimes during replay
4655 Description: Assertion failure in ll_local_open after replay.
4656 Details : If replay happened on an open request reply before we were able
4657 to set replay handler, reply will become not swabbed tripping the
4658 assertion in ll_local_open. Now we set the handler right after
4659 recognising of open request
4662 Frequency : very rare
4664 Description: kernel reports "badness in vsnprintf"
4665 Details : Reading from the "recovery_status" /proc file in small chunks
4666 may cause a negative length in lprocfs_obd_rd_recovery_status()
4667 call to vsnprintf() (which is otherwise harmless). Exit early
4668 if there is no more space in the output buffer.
4670 Severity : enhancement
4672 Description: clear OBD RPC statistics by writing to them
4673 Details : It is now possible to clear the OBD RPC statistics by writing
4674 to the "stats" file.
4679 Description: Client mtime is not the same on different clients after utimes
4680 Details : In some cases, the client was using the utimes() syscall on
4681 a file cached on another node. The clients now validate the
4682 ctime from the MDS + OSTs to determine which one is right.
4687 Description: Inability to activate failout mode
4688 Details : lconf script incorrectly assumed that in pythong string's numeric
4689 value is used in comparisons.
4692 Frequency : always with multiple stripes per file
4694 Description: Inefficient object allocation for mutli-stripe files
4695 Details : When selecting which OSTs to stripe files over, for files with
4696 a stripe count that divides evenly into the number of OSTs,
4697 the MDS is always picking the same starting OST for each file.
4698 Return the OST selection heuristic to the original design.
4703 Description: mount failures may take full timeout to return an error
4704 Details : Under some heavy load conditions it is possible that a
4705 failed mount can wait for the full obd_timeout interval,
4706 possibly several minutes, before reporting an error.
4707 Instead return an error as soon as the status is known.
4709 Frequency : quota enabled and large files being deleted
4711 Description: releasing more than 4GB of quota at once hangs OST
4712 Details : If a user deletes more than 4GB of files on a single OST it
4713 will cause the OST to spin in an infinite loop. Release
4714 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4719 Description: statfs data retrieved from /proc may be stale or zero
4720 Details : When reading per-device statfs data from /proc, in the
4721 {kbytes,files}_{total,free,avail} files, it may appear
4722 as zero or be out of date.
4725 Frequency : systems with MD RAID1 external journal devices
4727 Description: lconf's call to blkid is confused by RAID1 journal devices
4728 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4729 of returning all block devices that match the journal UUID.
4732 Frequency : always, for aggregate stripe size over 4GB
4734 Description: assertion fails when trying to use 4GB stripe size
4735 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4736 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4739 Frequency : always on ppc64
4741 Description: the first write on an ext3 filesystem with mballoc got stuck
4742 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4743 perform endianness conversion.
4745 ------------------------------------------------------------------------------
4747 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4749 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4750 INTEROPERATE with older versions automatically. Please read the
4751 user documentation before upgrading any part of a live system.
4752 * WARNING: Lustre networking configuration changes are required with
4753 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4756 * Support for kernels:
4757 2.6.9-22.0.2.EL (RHEL 4)
4758 2.6.5-7.244 (SLES 9)
4759 2.6.12.6 vanilla (kernel.org)
4762 Severity : enhancement
4763 Bugzilla : 7981/8208
4764 Description: Introduced Lustre Networking (LNET)
4765 Details : LNET is new networking infrastructure for Lustre, it includes
4766 a reorganized network configuration mode (see the user
4767 documentation for full details) as well as support for routing
4768 between different network fabrics. Lustre Networking Devices
4769 (LNDs) for the supported network fabrics have also been
4770 created for this new infrastructure.
4772 Severity : enhancement
4773 Description: Introduced Access control lists
4774 Details : clients can set ACLs on files and directories in order to have
4775 more fine-grained permissions than the standard Unix UGO+RWX.
4776 The MDS must be started with the "-o acl" mount option.
4778 Severity : enhancement
4779 Description: Introduced filesystem quotas
4780 Details : Administrators may now establish per-user quotas on the
4783 Severity : enhancement
4785 Description: Configuration change for the XT3
4786 The PTLLND is now used to run Lustre over Portals on the XT3
4787 The configure option(s) --with-cray-portals are no longer used.
4788 Rather --with-portals=<path-to-portals-includes> is used to
4789 enable building on the XT3. In addition to enable XT3 specific
4790 features the option --enable-cray-xt3 must be used.
4795 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4796 Details : A race between allocating a new llcd and re-getting the llcd_lock
4797 allowed another thread to grab newly-allocated llcd.
4799 Severity : enhancement
4801 Description: 2.6 OST async journal commit and locking fix to improve performance
4802 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4803 2.6 kernels are now async as they already were in 2.4 kernels so
4804 that they can commit concurrently with the network bulk transfer.
4805 For block-allocated files the filter allocation semaphore is held
4806 to avoid filesystem fragmentation during allocation. BKL lock
4807 removed for 2.6 xattr operations where it is no longer needed.
4812 Description: lconf incorrectly determined whether two IP networks could talk
4813 Details : In some more complicated routing and multiple-network
4814 configurations, lconf will avoid trying to make a network
4815 connection to a disjoint part of the IP space. It was doing the
4816 math incorrectly for one set of cases.
4821 Description: Fix for potential infinite loop processing records in an llog.
4822 Details : If an llog record is corrupted/zeroed, it is possible to loop
4823 forever in llog_process(). Validate the llog record length
4824 and skip the remainder of the block on error.
4827 Frequency : occasional (liblustre only)
4829 Description: liblustre could not open files whose last component is a symlink
4830 Details : sysio_path_walk() would incorrectly pass the open intent to
4831 intermediate path components.
4834 Frequency : rare (liblustre only with non-standard tuning)
4835 Bugzilla : 7201 (7350)
4836 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4837 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4838 after it has been released, only do this for referenced locks
4840 Severity : enhancement
4842 Description: specifying an (invalid) directory default stripe_size of -1
4843 would reset the directory default striping
4844 Details : stripe_size -1 was used internally to signal directory stripe
4845 removal, now use "all default" to signal dir stripe removal
4846 as a directory striping of "all default" is not useful
4849 Frequency : common for large clusters running liblustre clients
4851 Description: doing an ls when liblustre clients are running is slow
4852 Details : sending a glimpse AST to a liblustre client waits for every AST
4853 to time out, as liblustre clients will not respond. Since they
4854 cannot cache data we refresh the OST lock LVB from disk instead.
4856 Severity : enhancement
4858 Description: doing an ls at the same time as file IO can be slow
4859 Details : enqueue and other "small" requests can be blocked behind many
4860 large IO requests. Create a new OST IO portal for non-IO
4861 requests so they can be processed faster.
4864 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4866 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4867 created file with mode 000
4868 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4869 allow an ftruncate() as a non-root user to a file with mode 000.
4870 Lustre now respects this flag to disable mode checks when
4871 truncating a file owned by the user
4874 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4877 Description: Revoking locks from clients that went dead or catatonic might take
4879 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4880 cancellation of such locks instant on servers without waiting for
4881 any reply from clients. Clients drops these locks when cancel
4882 notification from server is received without replying.
4885 Frequency : liblustre-only, when liblustre client dies or becomes busy
4887 Description: Doing ls on Linux clients can take a long time with active
4889 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4890 granting such locks to it in the first place if possible. Locks
4891 are taken by proxy on the OST during the read or write and
4892 dropped immediately afterward. Add connect flags handling, do
4893 not grant locks to liblustre clients for glimpse ASTs.
4895 Severity : enhancement
4897 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4898 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4899 sequential reads and behaves badly with applications doing random
4900 reads. Improve it by reading ahead at least read region, and
4901 avoiding excessive large RPC for small reads.
4903 Severity : enhancement
4905 Description: Creating more than 1000 files for a single job may cause a load
4906 imbalance on the OSTs if there are also a large number of OSTs.
4907 Details : qos_prep_create() uses an OST index reseed value that is an
4908 even multiple of the number of available OSTs so that if the
4909 reseed happens in the middle of the object allocation it will
4910 still utilize the OSTs as uniformly as possible.
4915 Description: OST or MDS may oops in ping_evictor_main()
4916 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4917 but doesn't restart at beginning of obd_exports_timed list
4920 Severity : enhancement
4922 Description: improve by-nid export eviction on the MDS and OST
4923 Details : allow multiple exports with the same NID to be evicted at one
4924 time without re-searching the exports list.
4927 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4929 Description: MDS may oops in groups_free()
4930 Details : in rare race conditions a newly allocated group_info struct is
4931 freed again, and this can be NULL. The 2.4 compatibility code
4932 for groups_free() checked for a NULL pointer, but 2.6 did not.
4935 Frequency : common for liblustre clients doing little filesystem IO
4936 Bugzilla : 9352, 7313
4937 Description: server may evict liblustre clients accessing contended locks
4938 Details : if a client is granted a lock or receives a completion AST
4939 with a blocking AST already set it would not reply to the AST
4940 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4943 Frequency : lfs setstripe, only systems with more than 160 OSTs
4945 Description: unable to set striping with a starting offset beyond OST 160
4946 Details : llapi_create_file() incorrectly limited the starting stripe
4947 index to the maximum single-file stripe count.
4950 Frequency : LDAP users only
4952 Description: lconf did not handle in-kernel recovery with LDAP properly
4953 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4955 Severity : enhancement
4957 Description: bind OST threads to NUMA nodes to improve performance
4958 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4959 node and do their allocations there to localize memory access
4961 Severity : enhancement
4963 Description: llmount can determine client NID directly from Myrinet (GM)
4964 Details : the client NID code from gmnalnid was moved directly into
4965 llmount, removing the need to use this or specifying the
4966 client NID explicitly when mounting GM clients with zeroconf
4969 Frequency : if client is started with down MDS
4971 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4972 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4973 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4974 ptlrpc_import_delayed_req() because we didn't check if the
4975 request was interrupted, and we also didn't break out of the
4976 event loop if there was no timeout
4981 Description: data loss during non-page-aligned writes to a single file from
4982 both multiple nodes and multiple threads on one node at same time
4983 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4984 inconsistency led to false short-reads, that were cached and later
4985 used by ->prepare_write() to fill in partially written page,
4986 leading to data loss.
4989 Frequency : always, if lconf --abort_recovery used
4991 Description: lconf --abort_recovery fails with 'Operation not supported'
4992 Details : lconf was attempting to abort recovery on the MDT device and not
4995 Severity : enhancement
4997 Description: remove cleanup logs
4998 Details : replace lconf-generated cleanup logs with lustre internal
4999 cleanup routines. Eliminates the need for client-cleanup and
5002 Severity : enhancement
5004 Description: add support for EAs (user and system) on lustre filesystems
5005 Details : it is now possible to store extended attributes in the Lustre
5006 client filesystem, and with the user_xattr mount option it
5007 is possible to allow users to store EAs on their files also
5009 Severity : enhancement
5011 Description: Add possibility (config option) to show minimal available OST free
5013 Details : When compiled with --enable-mindf configure option, statfs(2)
5014 (and so, df) will return least minimal free space available from
5015 all OSTs as amount of free space on FS, instead of summary of
5016 free spaces of all OSTs.
5018 Severity : enhancement
5020 Description: do not expand extent locks acquired on OST-side
5021 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5022 by server: they are not cached anyway.
5025 Frequency : when mmap is used/binaries executed from Lustre
5027 Description: Unmmap pages before throwing them away from read cache.
5028 Details : llap_shrink cache now attempts to unmap pages before discarding
5029 them (if unmapping failed - do not discard). SLES9 kernel has
5030 extra checks that trigger if this unmapping is not done first.
5035 Description: lconf didn't resolve symlinks before checking to see whether a
5036 given mountpoint was already in use
5039 Frequency : when migrating failover services
5040 Bugzilla : 6395, 9514
5041 Description: When migrating a subset of services from a node (e.g. failback
5042 from a failover service node) the remaining services would
5043 time out and evict clients.
5044 Details : lconf --force (implied by --failover) sets the global obd_timeout
5045 to 5 seconds in order to quickly disconnect, but this caused
5046 other RPCs to time out too quickly. Do not change the global
5047 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5049 Severity : enhancement
5050 Frequency : if MDS is started with down OST
5051 Bugzilla : 9439,5706
5052 Description: Allow startup/shutdown of an MDS without depending on the
5053 availability of the OSTs.
5054 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5055 Add appropriate locking and lov-osc refcounts for safe
5056 cleaning. Add osc abort_inflight calls in case the
5057 synchronize never started.
5060 Frequency : occasional (Cray XT3 only)
5062 Description: root not authorized to access files in CRAY_PORTALS environment
5063 Details : The client process capabilities were not honoured on the MDS in
5064 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5065 been accessed by an authorized user then root was able to access
5066 the file on the local client also. The root user capabilities
5067 are now allowed on the MDS, as this environment has secure UID.
5070 Frequency : occasional
5072 Description: ldiskfs "too long searching" message happens too often
5073 Details : A debugging message (otherwise harmless) prints too often on
5074 the OST console. This has been reduced to only happen when
5075 there are fragmentation problems on the filesystem.
5080 Description: Division by zero in statfs when all OSCs are inactive
5081 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5082 lov_statfs divides by value returned by lov_get_stripecnt().
5086 Bugzilla : 9489, 3273
5087 Description: First write from each client to each OST was only 4kB in size,
5088 to initialize client writeback cache, which caused sub-optimal
5089 RPCs and poor layout on disk for the first written file.
5090 Details : Clients now request an initial cache grant at (re)connect time
5091 and so that they can start streaming writes to the cache right
5092 away and always do full-sized RPCs if there is enough data.
5093 If the OST is rebooted the client also re-establishes its grant
5094 so that client cached writes will be honoured under the grant.
5099 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5100 Details : Now I/O RPCs go to different portal number and (presumably) fast
5101 lock requests (and glimses) and other RPCs get their own service
5102 threads pool that should be able to service those RPCs
5105 Severity : enhancement
5107 Description: Ability to exchange lustre version between client and servers and
5108 issue warnings at client side if client is too old. Also for
5109 liblustre clients there is ability to refuse connection of too old
5111 Details : New 'version' field is added to connect data structure that is
5112 filled with version info. That info is later checked by server and
5116 Frequency : rare, liblustre only.
5117 Bugzilla : 9296, 9581
5118 Description: Two simultaneous writes from liblustre at offset within same page
5119 might proceed at the same time overwriting eachother with stale
5121 Details : I/O lock withing llu_file_prwv was released too early, before data
5122 actually was hitting the wire. Extended lock-holding time until
5123 server acknowledges receiving data.
5126 Frequency : extremely rare. Never observed in practice.
5128 Description: avoid generating lustre_handle cookie of 0.
5129 Details : class_handle_hash() generates handle cookies by incrementing
5130 global counter, and can hit 0 occasionaly (this is unlikely, but
5131 not impossible, because initial value of cookie counter is
5132 selected randonly). Value of 0 is used as a sentinel meaning
5133 "unassigned handle" --- avoid it. Also coalesce two critical
5134 sections in this function into one.
5136 Severity : enhancement
5138 Description: allow liblustre clients to delegate truncate locking to OST
5139 Details : To avoid overhead of locking, liblustre client instructs OST to
5140 take extent lock in ost_punch() on client's behalf. New connection
5141 flag is added to handle backward compatibility.
5143 Severity : enhancement
5144 Bugzilla : 4928, 7341, 9758
5145 Description: allow number of OST service threads to be specified
5146 Details : a module parameter allows the number of OST service threads
5147 to be specified via "options ost ost_num_threads={N}" in the
5148 OSS's /etc/modules.conf or /etc/modprobe.conf.
5152 Bugzilla : 6146, 9635, 9895
5153 Description: servers crash with bad pointer in target_handle_connect()
5154 Details : In rare cases when a client is reconnecting it was possible that
5155 the connection request was the last reference for that export.
5156 We would temporarily drop the export reference and get a new
5157 one, but this may have been the last reference and the export
5158 was just destroyed. Get new reference before dropping old one.
5160 Severity : enhancement
5161 Frequency : if client is started with failover MDS
5163 Description: Allow multiple MDS hostnames in the mount command
5164 Details : Try to read the configuration from all specified MDS
5165 hostnames during a client mount in case the "primary"
5168 Severity : enhancement
5170 Description: Stop sending data to evicted clients as soon as possible.
5171 Details : Check if the client we are about to send or are sending data to
5172 was evicted already. (Check is done every second of waiting,
5173 for which l_wait_event interface was extended to allow checking
5174 of exit condition at specified intervals).
5177 Frequency : rare, normally only when NFS exporting is done from client
5179 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5181 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5184 Frequency : always, if srand() is called before liblustre initialization
5186 Description: Liblustre uses system PRNG disturbing its usage by user application
5187 Details : Introduce internal to lustre fast and high-quality PRNG for
5188 lustre usage and make liblustre and some other places in generic
5189 lustre code to use it.
5191 Severity : enhancement
5192 Bugzilla : 9477, 9557, 9870
5193 Description: Verify that the MDS configuration logs are updated when xml is
5194 Details : Check if the .xml configuration logs are newer than the config
5195 logs stored on the MDS and report an error if this is the case.
5196 Request --write-conf, or allow starting with --old_conf.
5198 Severity : enhancement
5200 Description: Handle symlinks in the path when checking if Lustre is mounted.
5201 Details : Resolve intermediate symlinks when checking if a client has
5202 mounted a filesystem to avoid duplicate client mounts.
5207 Description: lconf can hit an error exception but still return success.
5208 Details : The lconf command catches the Command error exception at the top
5209 level script context and will exit with the associated exit
5210 status, but doesn't ensure that this exit status is non-zero.
5215 Description: failure of ptlrpc thread startup can cause oops
5216 Details : Starting a ptlrpc service thread can fail if there are a large
5217 number of threads or the server memory is very fragmented.
5218 Handle this without oopsing.
5221 Frequency : always, only if liblustre and non-default acceptor port was used
5223 Description: liblustre cannot connect to servers with non-default acceptor port
5224 Details : tcpnal_set_default_params() was not called and was therefore
5225 ignoring the environment varaible TCPNAL_PORT, as well as other
5226 TCPNAL_ environment variables
5231 Description: two objects could be created on the same OST for a single file
5232 Details : If an OST is down, in some cases it was possible to create two
5233 objects on a single OST for a single file. No problems other
5234 than potential performance impact and spurious error messages.
5238 Bugzilla : 5681, 9562
5239 Description: Client may oops in ll_unhash_aliases
5240 Details : Client dcache may become inconsistent in race condition.
5241 In some cases "getcwd" can fail if the current directory is
5247 Description: Inode refcounting problems in NFS export code
5248 Details : link_raw functions used to call d_instantiate without obtaining
5249 extra inode reference first.
5253 Bugzilla : 9942, 9903
5254 Description: Referencing freed requests leading to crash, memleaks with NFS.
5255 Details : We used to require that call to ll_revalidate_it was always
5256 followed by ll_lookup_it. Also with revalidate_special() it is
5257 possible to call ll_revalidate_it() twice for the same dentry
5258 even if first occurence returned success. This fix changes semantic
5259 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5260 reference on a request referred from the intent.
5261 ll_intent_release() then releases such a request.
5264 Frequency : rare, normally benchmark loads only
5266 Description: unlinked inodes were kept in memory on the client
5267 Details : If a client is repeatedly creating and unlinking files it
5268 can accumulate a lot of stale inodes in the inode slab cache.
5269 If there is no other client load running this can cause the
5270 client node to run out of memory. Instead flush old inodes
5271 from client cache that have the same inode number as a new inode.
5274 Frequency : SLES9 2.6.5 kernel and long filenames only
5275 Bugzilla : 9969, 10379
5276 Description: utime reports stale NFS file handle
5277 Details : SLES9 uses out-of-dentry names in some cases, which confused
5278 the lustre dentry revalidation. Change it to always use the
5282 Frequency : rare, unless heavy write-truncate concurrency is continuous
5283 Bugzilla : 4180, 6984, 7171, 9963, 9331
5284 Description: OST becomes very slow and/or deadlocked during object unlink
5285 Details : filter_destroy() was holding onto the parent directory lock
5286 while truncating+unlinking objects. For very large objects this
5287 may block other threads for a long time and slow overall OST
5288 responsiveness. It may also be possible to get a lock ordering
5289 deadlock in this case, or run out of journal credits because of
5290 the combined truncate+unlink. Solution is to do object truncate
5291 first in one transaction without parent lock, and then do the
5292 final unlink in a new transaction with the parent lock. This
5293 reduces the lock hold time dramatically.
5296 Frequency : rare, 2.4 kernels only
5298 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5299 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5300 before unmounting their filesystem in order to avoid deadlock.
5301 The kernel_locked() function in 2.4 kernels only checks whether
5302 the kernel lock is held, not whether it is this process that is
5303 holding it as 2.6 kernels do.
5308 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5309 Details : The client ptlrpc code may be trying to reconnect to a down
5310 server before a previous connection attempt has timed out.
5311 Increase the reconnect interval to be longer than the connection
5312 timeout interval to avoid sending duplicate connections to
5316 Frequency : echo_client brw_test command
5318 Description: fix echo_client to work with OST preallocated code
5319 Details : OST preallocation code (5137) didn't take echo_client IO path
5320 into account: echo_client calls filter methods outside of any
5321 OST thread and, hence, there is no per-thread preallocated
5322 pages and buffers to use. Solution: hijack pga pages for IO. As
5323 a byproduct, this avoids unnecessary data copying.
5327 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5328 Description: Client can oops in mdc_commit_close() after open replay
5329 Details : It was possible for the MDS to return an open request with no
5330 transaction number in mds_finish_transno() if the client was
5331 evicted, but without actually returning an error. Clients
5332 would later try to replay that open and may trip an assertion
5333 Simplify the client close codepath, and always return an error
5334 from the MDS in case the open is not successful.
5337 Frequency : rare, 2.6 OSTs only
5339 Description: OST may deadlock under high load on fragmented files
5340 Details : If there was a heavy load and highly-fragmented OST filesystems
5341 it was possible to have all the OST threads deadlock waiting on
5342 allocation of biovecs, because the biovecs were not released
5343 until the entire RPC IO was completed. Instead, release biovecs
5344 as soon as they are complete to ensure forward IO progress.
5346 Severity : enhancement
5348 Description: Support for specifying external journal device at mount
5349 Details : If an OST or MDS device is formatted with an external journal
5350 device, this device major/minor is stored in the ext3 superblock
5351 and may not be valid for failover. Allow detecting and
5352 specifying the external journal at mount time.
5357 Description: Mounting an MDS with pending unlinked files may cause oops
5358 Details : target_finish_recovery() calls mds_postrecov() which returned
5359 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5360 considers this an error and immediately begins cleaning up the
5361 lov, just after starting the mds_lov process
5363 Severity : enhancement
5365 Description: Implement 'lfs df' to report actual free space on per-OST basis
5366 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5367 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5368 usage of objects; '-h' to report in human readable format.
5370 ------------------------------------------------------------------------------
5372 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5379 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5380 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5381 at startup during orphan unlinking. Instead, skip buddy update
5382 before setup, as it will be regenerated after recovery is complete.
5387 Description: If an OST is inactive, its locks might reference stale inodes.
5388 Details : lov_change_cbdata() must iterate over all namespaces, even if
5389 they are inactive to clear inode references from the lock.
5391 Severity : enhancement
5392 Frequency : occasional, if non-standard max_dirty_mb used
5394 Description: Client will block write RPCs if not enough grant
5395 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5396 then the client will block writes while waiting for another RPC
5397 to complete instead of consuming its dirty limit. With change
5398 we get improved performance when max_dirty_mb is small.
5400 Severity : enhancement
5401 Bugzilla : 3389, 6253
5402 Description: Add support for supplementary groups on the MDS.
5403 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5404 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5405 lookups for user supplementary groups into a cache.
5409 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5410 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5411 for liblustre O_EXCL is enforced by MDS.
5415 Description: readdir never returns NULL in liblustre.
5416 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5417 dentry in struct dirent.
5421 Frequency : liblustre only, depends on application IO pattern
5422 Description: liblustre clients evicted if not contacting servers
5423 Details : Don't put liblustre clients into the ping_evictor list, so
5424 they will not be evicted by the pinger ever.
5426 Severity : enhancement
5428 Description: Add ability to evict clients by NID from MDS.
5429 Details : By echoing "nid:$NID" string into
5430 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5431 $NID would be instantly evicted from this MDS and from all active
5432 OSTs connected to it.
5436 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5437 Details : lookup_it_finish() used to query file size from OSTs that was not
5442 Description: service threads change working directory to that of init
5443 Details : Starting lustre service threads may pin the working directory
5444 of the parent thread, making that filesystem busy. Threads
5445 now change to the working directory of init to avoid this.
5449 Frequency : during shutdown only
5450 Description: shutdown with a failed MDS or OST can cause unmount to hang
5451 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5456 Frequency : 2.6 only
5457 Description: chmod/chown may include an extra supplementary group
5458 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5459 supplementary group and if none is specified this is used.
5462 Bugzilla : 5479 (6816)
5464 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5465 Details : If lookup is in progress on a file that is unlinked we might try
5466 to revalidate the inode and fail in revalidate after lookup is
5467 complete and ll_file_open() enqueues the open again but
5468 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5471 Frequency : always, if lconf --abort_recovery used
5473 Description: lconf --abort_recovery fails with 'Operation not supported'
5474 Details : lconf was attempting to abort recovery on the MDT device and not
5477 ------------------------------------------------------------------------------
5479 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5484 Frequency : rare (only unsupported configurations with a node running as an
5486 Bugzilla : 6514, 5137
5487 Description: Mounting a Lustre file system on a node running as an OST could
5489 Details : OSTs now preallocates memory needed to write out data at
5490 startup, instead of when needed, to avoid having to
5491 allocate memory in possibly low memory situations.
5492 Specifically, if the file system is mounted on on OST,
5493 memory pressure could force it to try to write out data,
5494 which it needed to allocate memory to do. Due to the low
5495 memory, it would be unable to do so and the node would
5496 become unresponsive.
5498 Severity : enhancement
5500 Description: Addition of lconf --service command line option
5501 Details : lconf now accepts a '--service <arg>' option, which is
5502 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5504 Severity : enhancement
5506 Description: Failover mode is now the default for OSTs.
5507 Details : By default, OSTs will now run in failover mode. To return to
5508 the old behaviour, add '--failout' to the lmc line for OSTs.
5510 Severity : enhancement
5512 Description: Health checks are now provided for MDS and OSTs
5513 Details : Additional detailed health check information on MSD and OSTs
5514 is now provided through the procfs health_check value.
5517 Frequency : occasional, depends on IO load
5519 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5520 after numerous create/delete cycles
5521 Details : The ext3 inode allocation policy would not allocate new inodes
5522 very well on the OSTs because there are no new directories
5523 being created. Instead we look for groups with free space if
5524 the parent directories are nearly full.
5529 Description: Network or server problems during mount may cause partially
5530 mounted clients instead of returning an error.
5531 Details : The config llog parsing code may overwrite the error return
5532 code during mount error handling, returning success instead
5538 Description: MDS can fail to allocate large reply buffers
5539 Details : After long uptimes the MDS can fail to allocate large reply
5540 buffers (e.g. zconf client mount config records) due to memory
5541 fragmentation or consumption by the buffer cache. Preallocate
5542 some large reply buffers so that these replies can be sent even
5543 under memory pressure.
5547 Frequency : rare (liblustre)
5548 Description: fsx running with liblustre complained that using truncate() to
5549 extend the file doesn't work. This patch corrects that issue.
5550 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5551 ATTR_SIZE and lsm use in llu_setattr_raw.
5555 Frequency : rare, only 2.6 kernels
5556 Description: Unusual file access patterns on the MDS may result in inode
5557 data being lost in very rare circumstances.
5558 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5559 code in the 2.6 kernel caused the inode and/or EA data not to
5560 be read from disk, causing single-file corruption.
5564 Frequency : rare, only 2.6 filesystems using extents
5565 Description: Heavy concurrent write and delete load may cause data corruption.
5566 Details : It was possible under high-load situations to have an extent
5567 metadata block in the block device cache from a just-unlinked
5568 file overwrite a newly-allocated data block. We now unmap any
5569 metadata buffers that alias just-allocated data blocks.
5573 Frequency : filesystems with default stripe_count larger than 77
5574 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5575 Details : lconf specifies an inode size of 4096 bytes when the default
5576 stripe_count is larger than 77. This conflicts with the default
5577 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5578 this case to avoid pinning too much memory for large EAs.
5580 ------------------------------------------------------------------------------
5582 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5587 Frequency : rare (extremely heavy IO load with hundreds of clients)
5589 Description: Client is evicted, gets IO error writing to file
5590 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5591 caused clients to be evicted for AST timeouts. The fixes in
5592 bug 5192 mean we no longer need to have such short AST timeouts
5593 so ldlm_timeout has been increased.
5596 Frequency : occasional during --force or --failover shutdown under load
5597 Bugzilla : 5949, 4834
5598 Description: Server oops/LBUG if stopped with --force or --failover under load
5599 Details : a collection of import/export refcount and cleanup ordering
5600 issues fixed for safer force cleanup
5603 Frequency : only filesystems larger than 120 OSTs
5604 Bugzilla : 5990, 6223
5605 Description: lfs getstripe would oops on a very large filesystem
5606 Details : lov_getconfig used kfree on vmalloc'd memory
5609 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5610 Bugzilla : 6242, 6243
5611 Description: reading from files that had been truncated to a non-zero size
5612 but never opened returned no data
5613 Details : ll_file_read() reads zeros from no-object files to EOF
5618 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5619 Details : There's little harm in aborting MDS/OSS recovery and letting it
5620 try again, so I removed the LASSERT and return an error instead.
5622 Severity : enhancement
5624 Description: New debugging infrastructure for tracking down data corruption
5625 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5626 (b) cover more of the client-side code path, and (c) try to narrow
5627 down where problems occurred
5631 Bugzilla : 3819, 4364, 4397, 6313
5632 Description: Racing close and eviction MDS could cause assertion in mds_close
5633 Details : It was possible to get multiple mfd references during close and
5634 client eviction, leading to one thread referencing a freed mfd.
5636 Severity: : enhancement
5637 Bugzilla : 3262, 6359
5638 Description: Attempts to reconnect to servers are now more aggressive.
5639 Details : This builds on the enhanced upcall-less recovery that was added
5640 in 1.4.2. When trying to reconnect to servers, clients will
5641 now try each server in the failover group every 10 seconds. By
5642 default, clients would previously try one server every 25 seconds.
5647 Description: After recovery, certain operations trigger a failed
5648 assertion on a client.
5649 Details : Failing over an mds, using lconf -d --failover, while a
5650 client was doing a readdir() call would cause the client to
5651 LBUG after recovery completed and the readdir() was resent.
5653 Severity : enhancement
5655 Description: Default groups are now added by lconf
5656 Details : You can now run lconf --group <servicename> without having to
5657 manually add groups with lmc.
5660 Frequency : occasional
5662 Description: Nodes with an elan id of 0 trigger a failed assertion
5665 Frequency : always when accessing e.g. tty/console device nodes
5667 Description: tty and some other devices nodes cannot be used on lustre
5668 Details : file's private_data field is used by device data and lustre
5669 values in there got lost. New field was added to struct file to
5670 store fs-specific private data.
5673 Frequency : when exporting Lustre via NFS
5675 Description: NFSD failed occasionally when looking up a path component
5676 Details : NFSD is looking up ".." which was broken in ext3 directories
5677 that had grown large enough to become hashed.
5680 Frequency : Clusters with multiple interfaces not on the same subnet
5682 Description: Nodes will repeatedly try to reconnect to an interface which it
5683 cannot reach and report an error to the log.
5684 Details : Extra peer list entries will be created by lconf with some peers
5685 unreachable. lconf now validates the peer before adding it.
5688 Frequency : Only if a default stripe is set on the filesystem root.
5690 Description: Setting a default stripe on the filesystem root prevented the
5691 filesystem from being remounted.
5692 Details : The client was sending extra request flags in the root getattr
5693 request and did not allocate a reply buffer for the dir EA.
5696 Frequency : occasional, higher if lots of files are accessed by one client
5697 Bugzilla : 6159, 6097
5698 Description: Client trips assertion regarding lsm mismatch/magic
5699 Details : While revalidating inodes the VFS looks up inodes with ifind()
5700 and in rare cases can find an inode that is being freed.
5701 The ll_test_inode() code will free the lsm during ifind()
5702 when it finds an existing inode and then the VFS later attaches
5703 this free lsm to a new inode.
5707 Bugzilla : 6422, 7030
5708 Description: MDS deadlock between mkdir and client eviction
5709 Details : Creating a new file via mkdir or mknod (starting a transaction
5710 and getting the ns lock) can deadlock with client eviction
5711 (gets ns lock and trying to finish a synchronous transaction).
5714 Frequency : occasional
5715 Description: While starting a server, the fsfilt_ext3 module could not be
5717 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5718 kernels. Previously, lconf would still use the ext3 name
5719 when trying to load modules. Now, it will correctly use
5720 ext3 on 2.4 and ldiskfs on 2.6.
5722 Severity : enhancement
5723 Description: The default stripe count has been changed to 1
5724 Details : The interpretation of the default stripe count (0, to lfs
5725 or lmc) has been changed to mean striping across a single
5726 OST, rather than all available. For general usage we have
5727 found a stripe count of 1 or 2 works best.
5729 Severity : enhancement
5730 Description: Add support for compiling against Cray portals.
5731 Details : Conditional compiling for some areas that are different
5735 Frequency : occasional
5736 Bugzilla : 6409, 6834
5737 Description: Creating files with an explicit stripe count may lead to
5738 a failed assertion on the MDS
5739 Details : If some OSTs are full or unavailable, creating files may
5740 trigger a failed assertion on the MDS. Now, Lustre will
5741 try to use other servers or return an error to the
5745 Frequency : occasional
5747 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5748 caused lock timeout during readahead (since 1.4.2).
5749 Details : Processes doing readahead might match a lock that hasn't been
5750 granted yet if there are overlapping and conflicting lock
5751 requests. The readahead process waits on ungranted lock
5752 (original lock is CBPENDING), while OST waits for that process
5753 to cancel CBPENDING read lock and eventually evicts client.
5755 Severity : enhancement
5757 Description: Initial enabling of flock support for clients
5758 Details : Implements fcntl advisory locking and file status functions.
5759 This feature is provided as an optional mount flag (default
5760 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5761 locking are implemented yet, and those that are are not guaranteed
5762 to be completely correct in production environments.
5763 mount -t lustre -o [flock|noflock] ...
5766 Frequency : occasional
5768 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5769 assertion in the ext3 JBD (journaling) layer.
5770 Details : The b_committed_data struct is protected by the big kernel lock
5771 in 2.4 kernels, serializing journal_commit_transaction() and
5772 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5773 access to this struct. In 2.6 kernels there is finer grained
5774 locking to improve SMP performance of the JBD layer.
5778 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5783 Description: merge and cleanup kernel patches.
5786 -----------------------------------------------------------------------------
5788 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5790 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5791 versions of Lustre. Please update all servers and clients to
5792 version 1.4.2 or later at the same time. You must also run
5793 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5794 to update the configuration logs.
5796 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5797 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5798 - swab llog records of type '0' so we get proper header size/idx (5861)
5799 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5800 - fix rename of one directory over another leaking an inode (5953)
5801 - avoid SetPageDirty on 2.6 (5981)
5802 - don't re-add just-being-destroyed locks to the waiting list (5653)
5803 - when creating new directories, inherit the parent's custom
5804 striping settings if present parent (3048)
5805 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5806 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5807 - don't hold client locks on temporary worklist from l_lru (5666)
5808 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5809 - automatically evict dead clients (5921)
5810 - Update file size properly in create+truncate+fstat case (6196)
5811 - Do not unhash mountpoint dentries, do not allow removal of
5813 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5814 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5815 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5816 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5817 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5818 - Correctly handle reads of files with no objects (6243)
5819 - lctl recover will also mark a device active if deactivate used (5933)
5821 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5822 - allow --write-conf on an MDS with different nettype than client (5619)
5823 - don't write config llogs to MDS for mounts not from that MDS (5617)
5824 - lconf should create multiple TCP connections from a client (5201)
5825 - init scripts are now turned off by default; run chkconfig --on
5826 lustre and chkconfig --on lustrefs to use them
5827 - upcalls are no longer needed for clients to recover to failover
5829 - add --abort-recovery option to lconf to abort recovery on device
5831 - add support for an arbitrary number of OSTs (3026)
5832 - Quota support protocol changes.
5833 - forward compatibility changes to wire structs (6007)
5834 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5835 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5836 - improved "lustre status" script
5837 - initialize blocksize for non-regular files (6062)
5838 - added --disable-server and --disable-client configure options (5782)
5839 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5840 - Vanilla 2.4.29 support
5841 - increase maximum number of obd devices to 520 (6242)
5842 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5843 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5844 SLES 9 2.6.5 kernels
5845 - limit stripes per file to 160 (the maximum EA size) (6093)
5847 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5850 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5851 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5852 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5853 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5854 - small ext3 extents cleanups and fixes (5733)
5855 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5856 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5857 - avoid needless client->OST connect, fix handle mismatch (5317)
5858 - fix DLM error path that led to out-of-sync client, long delays (5779)
5859 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5860 - fix several locking issues related to i_size (5492,5624,5654,5672)
5861 - don't move pending lock onto export if it is already evicted (5683)
5862 - fix kernel oops when creating .foo in unlinked directory (5548)
5863 - fix deadlock in obdfilter statistics vs. object create (5811)
5864 - use time_{before,after} to avoid timer jiffies wrap (5882)
5865 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5866 - Do not leak request if server was not able to process it (5154)
5867 - If mds_open unable to find parent dir, make that negative lookup(5154)
5868 - don't create new directories with extent-mapping (5909, 5936)
5870 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5871 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5872 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5873 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5874 - support for RHEL4 kernel on i686 (5773)
5875 - provide error messages when incompatible logs are encountered (5898)
5877 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5878 * version 1.4.0.10 (1.4.1 release candidate 1)
5880 - don't keep a lock reference when lock is not granted (4238)
5881 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5882 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5883 - drop import inflight refcount on signal_completed_replay error (5255)
5884 - unlock page after async write error during send (3677)
5885 - handle missing objects in filter_preprw_read properly (5265)
5886 - no transno return for symlink open, don't save no-trasno open (3440)
5887 - don't try to complete elan receive that already failed (4012)
5888 - free RPC server reply state on error (5406)
5889 - clean up thread from ptlrpc_start_thread() on error (5160)
5890 - readahead could read extra page into cache that wasn't ejected (5388)
5891 - prevent races in class_attach/setup/cleanup/detach (5260)
5892 - don't dereference de->d_inode after l_dput of de (5458)
5893 - use "int" for stripe value returned from lock_to_stripe (5544)
5894 - mballoc allocation and error-checking fixes in 2.6 (5504)
5895 - block device patches to fix I/O request sizes in 2.6 (5482)
5896 - look up hostnames for IB nals (5602)
5897 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5898 - don't start multiple acceptors for the same port (5277)
5899 - fix incorrect LASSERT in mds_getattr_name (5635)
5900 - export a proc file for general "ping" checking (5628)
5901 - fix "lfs check" to not block when the MDS is down (5628)
5903 - service request history (4965)
5904 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5905 - create an "evict_client" /proc entry on OSTs, like the MDS has
5906 - fix mount usage message, return errors per mount(8) (5168)
5907 - change grep [] to grep "[]" in tests so they work in more UMLs
5908 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5909 - remove ancient LOV_MAGIC_V0 EA support (5047)
5910 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5911 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5912 - allow readahead allocations to fail when low on memory (5383)
5913 - mmap locking landed again, after considerable improvement (2828)
5914 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5916 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5919 - send OST transaction number in read/write reply to free req (4966)
5920 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5921 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5922 - the watchdog thread now runs as interruptible (5246)
5923 - flock/lockf fixes (but it's still disabled, pending 5135)
5924 - don't use EXT3 constants in llite code (5094)
5925 - memory shortage at startup could cause assertion (5176)
5927 - reorganization of lov code
5928 - single portals codebase
5930 - add extents/mballoc support (5025)
5931 - direct I/O reads in the obdfilter (4048)
5932 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5934 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5937 - send OST transaction number in read/write reply to free req (4966)
5938 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5939 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5940 - the watchdog thread now runs as interruptible (5246)
5941 - handle missing objects in filter_preprw_read properly (5265)
5942 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5943 - drop import inflight refcount on signal_completed_replay error (5255)
5944 - unlock page after async write error during send (3677)
5945 - return original error code on reconstructed replies (3761)
5946 - no transno return for symlink open, don't save no-trasno open (3440)
5948 - add pid to ldlm debugging output (4922)
5949 - bump the watchdog timeouts -- we can't handle 30sec yet
5950 - extra debugging for orphan dentry/inode bug (5259)
5952 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5955 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5956 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5957 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5958 - fix NULL dereference in /proc/sys/portals/routes (4827)
5959 - allow failed mdc_close() operations to be interrupted (4561)
5960 - stop precreate on OST before MDS would time out on it (4778)
5961 - don't send partial-page writes before EOF from client (4410)
5962 - discard client grant for sub-page writes on large-page clients (4520)
5963 - don't free dentries not owned by NFS code, check generation (4806)
5964 - fix lsm leak if mds_create_objects() fails (4801)
5965 - limit debug_daemon file size, always print CERROR messages (4789)
5966 - use transno after validating reply (3892)
5967 - process timed out requests if import state changes (3754)
5968 - update mtime on OST during writes, return in glimpse (4829)
5969 - add mkfsoptions to LDAP (4679)
5970 - use ->max_readahead method instead of zapping global ra (5039)
5971 - don't interrupt __l_wait_event() during strace
5973 - add software watchdogs to catch hung threads quickly (4941)
5974 - make lustrefs init script start after nfs is mounted
5975 - fix CWARN/ERROR duplication (4930)
5976 - return async write errors to application if possible (2248)
5977 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5978 - print NAL number in %x format (4645)
5979 - update barely-supported suse-2.4.21-171 series (4842)
5980 - support for sles 9 %post scripts
5981 - support for building 2.6 kernel-source packages
5982 - support for sles km_* packages
5984 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5987 - ignore -ENOENT errors in osc_destroy (3639)
5988 - notify osc create thread that OSC is being cleaned up (4600)
5989 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5990 - reconstruct ost_handle() like mds_handle() (4657)
5991 - create a new thread to do import eviction to avoid deadlock (3969)
5992 - let lconf resolve symlinked-to devices (4629)
5993 - don't unlink "objects" from directory with default EA (4554)
5994 - hold socknal file ref over connect in case target is down (4394)
5995 - allow more than 32000 subdirectories in a single directory (3244)
5996 - fix blocks count for O_DIRECT writes (3751)
5997 - OST returns ENOSPC from object create when no space left (4539)
5998 - don't send truncate RPC if file size isn't changing (4410)
5999 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6000 - bind to privileged port in socknal and tcpnal (3689)
6002 - rate limit CERROR/CWARN console message to avoid overload (4519)
6003 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6004 - basic mmap support (3918)
6005 - kernel patch series update from b1_4 (4711)
6007 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6010 - avoid crash during MDS cleanup with OST shut down (2775)
6011 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6012 - don't use bad inodes on the MDS (3744)
6013 - dynamic object preallocation to improve recovery speed (4236)
6014 - don't hold spinlock over lock dumping or change debug flags (4401)
6015 - don't zero obd_dev when it is force cleaned (3651)
6016 - print grants to console if they go negative (4431)
6017 - "lctl deactivate" will stop automatic recovery attempts (3406)
6018 - look for existing locks in ldlm_handle_enqueue() (3764)
6019 - don't resolve lock handle twice in recovery avoiding race (4401)
6020 - revalidate should check working dir is a directory (4134)
6022 - don't always mark "slow" obdfilter messages as errors (4418)
6024 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6027 - don't close LustreDB during write_conf until it is done (3860)
6028 - fix typo in lconf for_each_profile (3821)
6029 - allow dumping logs from multiple threads at one time (3820)
6030 - don't allow multiple threads in OSC recovery (3812)
6031 - fix debug_size parameters (3864)
6032 - fix mds_postrecov to initialize import for llog ctxt (3121)
6033 - replace config semaphore with spinlock (3306)
6034 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6035 - don't allow enqueue to complete on a destroyed export (3822)
6036 - down write_lock before checking llog header bitmap (3825)
6037 - recover from lock replay timeout (3764)
6038 - up llog sem before sending rpc (3652)
6039 - reduce ns lock hold times when setting kms (3267)
6040 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6041 - fix NULL deref and obd_dev leak on setup error (3312)
6042 - replace some LBUG about llog ops with error handling (3841)
6043 - don't match INVALID dentries from d_lookup and spin (3784)
6044 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6045 - fix invalid assertion in ptlrpc_set_wait (3880)
6047 - add libwrap support for the TCP acceptor (3996)
6048 - add /proc/sys/portals/routes for non-root route listing (3994)
6049 - allow setting MDS UUID in .xml (2580)
6050 - print the stack of a process that LBUGs (4228)
6052 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6055 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6056 - make sure to unset replay flag from failed open requests (3440)
6057 - if default stripe count is 0, use OST count for inode size (3636)
6058 - update parent mtime/ctime on client for create/unlink (2611)
6059 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6060 - free recovery state on server during a forced cleanup (3571)
6061 - unregister_reply for resent reqs (3063)
6062 - loop back devices mounting and status check on 2.6 (3563)
6063 - fix resource-creation race that can provoke i_size == 0 (3513)
6064 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6065 - more debugging for page-accounting assertion (3746)
6066 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6067 - avoid deadlock after precreation failure (3758)
6068 - fix race and lock order deadlock in orphan handling (3450, 3750)
6069 - add validity checks when grabbing inodes from l_ast_data (3599)
6071 - add /proc/.../recovery_status to obdfilter (3428)
6072 - lightweight CDEBUG infrastructure, debug daemon (3668)
6073 - change default OSC RPC parameters to be better on small clusters
6074 - turn off OST read cache for files smaller than 32MB
6075 - install man pages and include them in rpms (3100)
6076 - add new init script for (un)mounting lustre filesystems (2593)
6077 - run chkconfig in %post for init scripts (3701)
6078 - drop scimac NAL (unmaintained)
6080 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6083 - clean kiobufs before and after use (3485)
6084 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6085 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6086 - change init script to fail more clearly if not run as root (1528)
6087 - allow clients to reconnect during replay (1742)
6088 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6089 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6090 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6091 - fix cond_resched (3554)
6092 - only evict unfinished clients after recovery (3515)
6093 - allow bulk resend, prevent data loss (3570)
6094 - dynamic ptlrpc request buffer allocation (2102)
6095 - don't allow unlinking open directory if it isn't empty (2904)
6096 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6097 - remove extraneous obd dereference causing LASSERT failure (3334)
6098 - don't use get_cycles() when creating temp. files on the mds (3156)
6099 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6100 - handle EEXIST for set-stripe, set proper directory name (3336)
6102 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6103 - fix ksocknal_fmb_callback() error messages (2918)
6105 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6108 - don't copy lvb into (possibly NULL) reply on error (2983)
6109 - don't deref dentry after dput, don't free lvb on error (2922)
6110 - use the kms to determine writeback rpc length (2947)
6111 - increment oti_logcookies when osc is inactive (2948)
6112 - update client's i_blocks count via lvb messages (2543)
6113 - handle intent open/close of special files properly (1557)
6114 - mount MDS with errors=remount-ro, like obdfilter (2009)
6115 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6116 - don't use cancelling-locks' kms values (2947)
6117 - use highest lock extent for kms, not last one (2925)
6118 - don't dereference ERR_PTR() dentry in error handling path (3107)
6119 - fix thread race in portals_debug_dumplog() (3122)
6120 - create lprocfs device entries at setup instead of at attach (1519)
6121 - common AST error handler, don't evict client on completion race (3145)
6122 - zero nameidata in detach_mnt in 2.6 (3118)
6123 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6124 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6125 - initialize RPC timeout timer earlier for 2.6 (3219)
6126 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6127 - print nal/nid for unknown nid (3258)
6128 - additional checks for oscc recovery before doing precreate (3284)
6129 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6130 - don't crash in mdc_close for bad permissions on open (3285)
6131 - zero i_rdev for non-device files (3147)
6132 - clear page->private before handing to FS, better assertion (3119)
6133 - tune the read pipeline (3236)
6134 - fix incorrect decref of invalidated dentry (2350)
6135 - provide read-ahead stats and refine rpc in flight stats (3328)
6136 - don't hold journal transaction open across create RPC (3313)
6137 - update atime on MDS at close time (3265)
6138 - close LDAP connection when recovering to avoid server load (3315)
6139 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6140 - don't leak open file on MDS after open resend (3325)
6141 - serialize filter_precreate and filter_destroy_precreated (3329)
6142 - loop device shouldn't call sync_dev() for nul device (3092)
6143 - clear page cache after eviction (2766)
6144 - resynchronize MDS->OST in background (2824)
6145 - refuse to mount the same filesystem twice on same mountpoint (3394)
6146 - allow llmount to create routes for mounting behind routers (3320)
6147 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6148 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6149 - fix rare problem with rename on htree directories (3417)
6151 - allow default OST striping configuration per directory (1414)
6152 - fix compilation for qswnal for 2.6 kernels (3125)
6153 - increase maximum number of MDS request buffers for large systems
6154 - change liblustreapi to be useful for external progs like lfsck (3098)
6155 - increase local configuration timeout for slow disks (3353)
6156 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6158 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6161 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6162 - don't overwrite extent policy data in reply if lock was blocked (2901)
6163 - drop filter export grants atomically with removal from device (2663)
6164 - del obd_self_export from work_list in class_disconnect_exports (2908)
6165 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6166 - swab reply message in mdc_close, other PPC fixes (2464)
6167 - fix destroying of named logs (2325)
6168 - overwrite old logs when running lconf --write_conf (2264)
6169 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6170 - fix race in target_handle_connect (2898)
6171 - mds_reint_create() should take same inode create lock (2926)
6172 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6173 - don't close files for self_export to avoid uninitialized obd (2936)
6174 - allow MDS with the same name as client node (2939)
6175 - hold dentry reference for closed log files for unlink (2325)
6176 - reserve space for all logs during transactions (2059)
6177 - don't evict page beyond end of stripe extent (2925)
6178 - don't oops on a deleted current working directory (2399)
6179 - handle hard links to targets without a parent properly (2517)
6180 - don't dereference NULL lock when racing during eviction (2867)
6181 - don't grow lock extents when lots of conflicting locks (2919)
6183 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6186 - account for cache space usage on clients to avoid data loss (974)
6187 - lfsck support in lustre kernel code (2349)
6188 - reduce journal credits needed for BRW writes (2370)
6189 - orphan handling to avoid losing space on client/server crashes
6190 - ptlrpcd can be blocked, stopping ALL progress (2477)
6191 - use lock value blocks to assist in proper KMS, faster stat (1021)
6192 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6193 - recovery for initial connections (2355)
6194 - fixes for mds_cleanup_orphans (1934)
6195 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6196 - block all file creations until orphan recovery completes (1901)
6197 - client remove rq_connection from request struct (2423)
6198 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6199 - recovery timer race (2670)
6200 - mdc_close recovey bug (2532)
6201 - ptlrpc cleanup bug (2710)
6202 - mds timeout on local locks (2588)
6203 - namespace lock held during RPCs (2431)
6204 - handle interrupted sync write properly (2503)
6205 - don't try to handle a message that hasn't been replied to (2699)
6206 - client assert failure during cleanup after abort recovery (2701)
6207 - leak mdc device after failed mount (2712)
6208 - ptlrpc_check_set allows timedout requests to complete (2714)
6209 - wait for inflight reqs when ptlrpcd finishes (2710)
6210 - make sure unregistered services are removed from the srv_list
6211 - reset bulk XID's when resending them (caught by 1138 test)
6212 - unregister_bulk after timeout
6213 - fix lconf error (2694)
6214 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6215 - readahead locks pages, leaves pending causing memory pressure (2673)
6216 - increase OST request buffers to 4096 on large machines (2729)
6217 - fix up permission of existing directories in simple_mkdir (2661)
6218 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6219 - don't assign transno to errored transactions (2742)
6220 - don't delete objects on OST if given a bogus objid from MDS (2751)
6221 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6222 - if rq_no_resend, then timeout request after recovery (2432)
6223 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6224 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6225 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6226 - MDS recovery shouldn't delete objects that it has given out (2730)
6227 - if enqueue arrives after completion, don't clobber LVB (2819)
6228 - don't unlock pages twice when trigger_group_io returns error (2814)
6229 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6230 - don't write pages to disk if there was an error (1450)
6231 - don't ping imports that have recovery disabled (2676)
6232 - take buffered bytes into account when balancing socknal conn (2817)
6233 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6234 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6235 - remove little-endian swabbing of llog records (1987)
6236 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6237 - retry reposting request buffers if they fail (1191)
6238 - grow extent at grant time to avoid granting a revoked lock (2809)
6239 - lock revoke doesn't evict page if covered by a second lock (2765)
6240 - disable VM readahead to avoid reading outside lock extents (2805)
6242 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6243 - updated kernel patches for hp-2.4.20 kernel (2681)
6245 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6248 - fix truncated write corruption (2366)
6249 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6251 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6252 - ensure len(uuid) < 37 in lmc (1171)
6253 - fix ia64 OOPS in llog_test (2255)
6254 - zero end of page at obdfilter for partial page writes (2648)
6255 - don't leave stale dentries around after renames (bug 2428)
6256 - fix timeouts when evicting a client with a single lock held (2642)
6257 - set deadline for the initial HELLO message to drain (2634)
6258 - print out dotted-quad IP addresses in the socknal (2302)
6260 - additional debugging for MDS client eviction problem (2443)
6261 - fix mkfsoptions support for osts (2603, 2604)
6263 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6266 - add series for the vanilla 2.6.0 kernel
6267 - add series for the vanilla 2.4.24 kernel
6268 - add series for a cray x86/64 UL kernel drop
6269 - fix xattr patches for the vanilla 2.4.19 series
6271 - generate true UUIDs in lmc (1171)
6272 - have portals stack dumping break in UML (2466)
6273 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6274 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6275 - wake callback waiting threads on client eviction (2460)
6276 - Add --ptldebug and --subsystem to lmc (1719)
6277 - update assertion to allow safe interrupt allocation
6278 - set rq_no_resend for cancel requests (2432)
6279 - recalculate ptlrpcd timeout after resend (2494)
6280 - call vfs_rmdir when removing pending directories (2368)
6281 - fix renaming a file to itself (2429)
6282 - lmc creates a default one-stripe lov (2454)
6283 - expand procfs space to handle large clusters (2326)
6284 - increase UML stack to avoid overflow
6285 - update lconf's list of debug and subsystem masks
6286 - fix lfs find --obd (2510)
6287 - /proc tunable for disabling filter read caching (2591)
6288 - stop rpm packages from altering slapd.conf (2301)
6289 - disable nagle in the socknal under 0conf (2578)
6290 - choose mds inode size based on stripe count (2572)
6291 - fix kernel-source rpm problems (2516)
6293 - add --disable-doc to avoid pdf generation (2421)
6294 - update documentation, tests, type-os, comments
6295 - avoid format warnings on ia64
6296 - remove the TOE NAL
6297 - tiny code cleanups by removing unused fields
6299 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6302 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6303 - remove the most blatant lies from BUILDING file (2371)
6304 - change default debug level to reasonable production setting
6305 - reduce client side cache size to reduce cache flush time
6306 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6307 - make TCP zerocopy and pinger support enabled by default (2476)
6308 - sync writes completed after process exits caused crashes (2319)
6309 - maintain correct mount count on the MDS (2356)
6310 - backout 1557, because 2316 wasn't really fixed
6311 - better file I/O statistics gathering in /proc
6312 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6313 - another kernel patch to fix zero-copy TCP function export
6314 - don't take duplicate lock when processing re-sent getattr (2420)
6315 - lctl uses obd_self_export instead of creating new conn (2353)
6316 - MDS/OST recovery case which requires object creation asserted (2425)
6317 - move lfs from /usr/sbin to /usr/bin in packages
6318 - fix race between mds_client_add and mds_client_free (2417)
6319 - use kmalloc instead of slabs in portals (2430)
6320 - don't create duplicate records when a failover MDS is present (2442)
6321 - remove unnecessary mount age check (2332)
6322 - don't remove directory inodes from locks prematurely (2451)
6323 - don't break if MDS service name is the same as hostname (2103)
6324 - fix races in client write RPC generation when cache full (2482)
6326 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6329 - remove now-unused request->rq_obd (278)
6330 - if an allocation fails, print out how much memory we've used (1933)
6331 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6332 - add the "configurable stack size" patch to most series files (1256)
6333 - ability to write large log records, for 100+ OST configs (2306)
6334 - fix NULL deref when filter_prep fails (2314)
6335 - fix operator precedence error in filter_sync
6336 - dynamic allocation of socknal TX descriptors (2315)
6337 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6338 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6339 - ensure that CWARN messages in Portals always get to the syslog
6340 - __init/__exit are not for prototype decls (ldlm_init/exit)
6341 - x86-64 compile warning fixes
6342 - fix gateway LMC keyword conflict (2318)
6343 - fix MDS lock inversions in getattr/reint paths (1844)
6344 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6345 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6346 - DLM race condition prevented some lock evictions (2328)
6347 - ENOMEM detection and retry on socknal sends (2230)
6348 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6349 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6350 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6351 - better error messages when a client is rejected during recovery (1505)
6352 - avoid cancelling locks which were never granted, after failure (2330)
6353 - fix i_sem/journal inversion in mds_client_add (2333)
6354 - fix truncate/getattr lock cycle deadlock (2334)
6355 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6356 - fix two rare exit paths which could leak an l_lock() ref (2321)
6357 - fencepost error in MDS/OST orphan recovery (2226)
6358 - make log record alignment 8 bytes (1988)
6359 - lstripe now fails when requested offset > ost_count (2237)
6360 - ensure that all kernel series have a complete list.h (1607)
6361 - fix crashes in special-file operations (2316)
6362 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6364 - allow configurable automake binary, for testing new versions
6365 - small update to the lfs documentation
6367 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6369 * fix negative export reference count in fsfilt_sync (2312)
6371 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6372 * release candidate 0.9.1
6374 - orphans are moved into the PENDING directory for possible recovery
6375 - replayed opens now open by fid for orphan/rename safety (1042)
6376 - last close of an orphan inode generates a transno (683)
6377 - chdir() and mount() now pin the directory entry (1020)
6378 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6379 - discard very old requests without processing them (1502)
6380 - remove some common, well-understood CERRORs (1505)
6381 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6382 - clear "grant" flags in OST replies until OST grant code lands (1644)
6383 - fix read performance by not clobbering i_blksize on client (1598)
6384 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6385 - make LRU size a /proc tunable, clears locks when reduced (707)
6386 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6387 - server locks take references on exports now (1558)
6388 - build fixes for 2.4.20-rh trees (1663)
6389 - return an error from lov_create if all OSCs are inactive (1751)
6390 - fix import levels when a reconnect happens without a timeout (1597)
6391 - exit early from mds_open if we get a lookup error (1749)
6392 - partial page read at EOF wouldn't wait for disk before sending (1642)
6393 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6394 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6395 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6396 - write/truncate lock inversion (1639)
6397 - Don't auto-load obdclass, portals modules during cleanup (1495)
6398 - fix timestamps from jumping to "now" (1763)
6399 - extra journal assertions (1648)
6400 - add an extra multiunlink test (1771)
6401 - fix read_record/write_record API (1776)
6402 - fix leak of offset_extent, possible incorrect i_size later (1772)
6403 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6404 - Debugging for the kqswnal_get_idle_tx problems (1820)
6405 - Allow recovery to be attempted multiple times (1536)
6406 - Write out MDS last_rcvd file after it is first created (1600)
6407 - Fix tx_descriptor leak in failed transmit situations (1827)
6408 - ext3 journaling fixes for assertion failure after IO error (1871)
6409 - class_export_put() on freed export after completion AST error (1896)
6410 - Fix revalidate looping in VFS (1322)
6411 - Don't access a freed export during MDS_REINT timeout (1521)
6412 - Add open-unlink recovery support on the MDS (1673,1764)
6413 - Return an error if no MDS data was read from last_rcvd (1946)
6414 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6415 - Don't setup a disk device that doesn't match exported UUID (317)
6416 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6417 - avoid committing NULL handle in force close
6418 - local.sh is now a one-stripe LOV configuration
6419 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6420 - fix ext3 htree duplicate directory entry corruption (1516)
6421 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6422 - update to new LOV EA format (2097)
6423 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6425 2003-06-15 Phil Schwan <phil@clusterfs.com>
6428 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6429 - per-import/export recovery handling (958, 931, 959)
6430 - multiple last-rcvd slots, for serving multiple FSes (949)
6431 - connections are again shared between multiple imp/exports (963, 964)
6432 - "umount -f" would hang if any requests needed to be sent (393, 978)
6433 - avoid pinning large req buffer by copying for queued messages (989)
6434 - add "uuid" to "lctl device" command to help upcalls (991)
6435 - "open" RPCs with transnos would confuse recovery counters (1037)
6436 - do proper endian conversion of all wire messages (288, 340, 891)
6437 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6438 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6439 - fix ptlrpc_connection leak in target_handle_connect (1174)
6440 - fix import refcounting bug in OST and MDS cleanup (1134)
6441 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6442 - fix very unlikely obd_types race condition (501)
6443 - remove osc_open hack for echo_client (1187)
6444 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6445 - a failure in read_inode2 leads to deadlock (1139)
6446 - cancel ack-locks as soon as transaction is committed (1072)
6447 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6448 - make sure to commitrw after any preprw to avoid deadlock (1162)
6449 - failing to execute a file in a lustre FS would lock inode (1203)
6450 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6451 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6452 - obd_uuid2tgt would walk off the end of the list (1255)
6453 - on IA64 the osc would give portals incorrect bulk size (1258)
6454 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6455 - fix lock inversion caused by new llite matching code (1282)
6456 - limit the number of dirty pages on a client to 10MB (1286)
6457 - timed out locks were not being corrected cancelled (1289)
6458 - fix O_DIRECT above 4GB on IA-32 (1292)
6459 * major user-visible changes
6460 - fail out/fail over policy now controlled by the upcall (993)
6462 - add OBD_PING to check server availability and failure (954)
6463 - lustre messages are now sent in sending host order (288, 340, 891)
6464 - add eadatalen to MDS getattr reply (340)
6465 - OST read replies may contain second buffer, with per-page status (593)
6467 2003-03-11 Phil Schwan <phil@clusterfs.com>
6470 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6471 - failed open()s could cause deadlock; fixed (867, 869)
6472 - stop cancelling OST locks when files are closed (481)
6473 - overlapping XID spaces caused network corruption (851, 853)
6474 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6475 - setattr_raw added, to do single-RPC, server-side setattrs
6476 - lmc/lconf syntax change for OST UUIDs
6477 - fix crashy race condition between ptlrpc_free_req and osc_close
6478 - don't use request in mdc_enqueue if we hit a timeout (889)
6479 - don't set the inode i_size for regular files from the MDS (896)
6480 - handle out of order completion AST (842)
6481 - don't LBUG if a lock request times out after receiving AST (913)
6482 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6483 - if a bad lock AST arrives, send an error instead of dropping entirely
6484 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6485 - fix leak in bulk IO when only partially completed (899, 900, 926)
6486 - fix O_DIRECT for ia64 (55)
6487 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6488 - C-z after timeout could hang a process forever; fixed (977)
6490 - client-side I/O cache (678, 924, 929, 941, 970)
6492 - READPAGE and SETATTRs which don't take server-side locks get
6495 2003-02-11 Phil Schwan <phil@clusterfs.com>
6498 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6499 - Fix fsfilt_extN_readpage() to read a full page of directory
6500 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6501 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6502 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6503 - Most intent operations were removed, in favour of a new RPC mode
6504 that does a single RPC to the server and bypasses most of the VFS
6505 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6506 - Aggressively cancel local locks on DLM servers
6507 - mds_reint_unlink sends EA to the client if it's the last nlink.
6508 client uses that EA to unlink OST objects.
6509 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6510 - recursive symlinks were fixed (439)
6511 - fixed NULL deref in DEBUG_REQ
6512 - filter_update_lastobjid no longer calls sync, which annoyed extN
6513 - fixed multi-client small-writes to a single file problem (445)
6514 - fixed mtime updates during file writes (607)
6515 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6516 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6517 - fixed Linux OST size reporting problem (444, 656)
6518 - OST now updates object mtime with writes or setattr (607, 619)
6519 - client verifies file size before zeroing page past EOF (445)
6520 - OST now writes last allocated objid to disk with allocation (108)
6521 - LOV on echo now works (409)
6523 - mds_reint_unlink sends a new buffer, with the EA included. this
6524 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6525 set if a regular file was being unlinked, and it was the last link
6526 - use PtlGet from the target for bulk writes (315)
6527 - OST now updates object mtime with writes or setattr (607, 619)
6528 - LDLM now has a grant-time callback to revalidate locked items, if
6530 - Many MDS operations were reorganized to combat race conditions
6532 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6533 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6535 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6538 - Fully reactivate OST imports after reconnection (512, others)
6539 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6540 - More graceful error handling for truncating on dead OST (515)
6541 - Don't error out unless we're actually accessing dead stripes (474)
6542 - Fix garbage sizes when stripes are missing (410)
6543 - LRU counters were broken, causing constant lock purge (433, 432)
6544 - garbage on read from stripes with failed OSTs (441)
6545 - mark OSCs as active before reconnecting during recovery (438)
6546 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6547 - lfind did not preserve OST order in output (443)
6548 - symlinks cause hung clients, incorrect data (439)
6549 - stop dereferencing request after dropping refcount (457)
6550 - don't LASSERT(spin_is_locked) on non-SMP (455)
6551 - fixes for many rename() bugs
6552 - fstat didn't correctly synchronize attributes (399)
6553 - server must handle lock cancellation during blocking AST prep (487)
6554 - bulk descriptors were free()d too soon (511)
6555 - fix paths in lconf, which would load incorrect modules (451, 507)
6556 - fix confusing lconf 'host not found' error message (386)
6557 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6558 - fix race condition in mdc_blocking_ast() for inode access (526)
6559 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6560 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6561 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6562 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6563 - fix lstripe in several ways (kernel side) (527)
6564 - fix request leak in ldlm_cli_enqueue (262)
6565 - incorrect OSC was marked inactive after OST failure
6566 - call mds_fs_cleanup before unmounting filesystem (524)
6567 - fix races between taking ns_lock and ldlm_lock_change_resource
6568 - fix races updating LOV export open file list
6569 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6570 - fix recovery NULL deref in ldlm_cli_cancel_unused
6571 - fix some DLM races by using new hash table for lock handles (419)
6572 - permit the client to specify desired inodes, at replay
6573 - duplicate requests when we queue them for replay reintegration
6574 - fix last_rcvd offset calculation
6575 - sync after each recovered transaction, so we always make progress
6576 - never, not always, ERESTART requests without transnos
6577 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6579 - skip replay if the MDS says that the client is already connected
6580 - don't check for a recovery-enabled export to match lctl's UUID
6581 - don't INC_USE_COUNT for phantom exports
6582 - don't crash when cleaning up phantom exports (567)
6583 - don't double-finish or set replay data for errored mdc_open requests
6584 - abort requests when they time out, so we don't get old replies
6585 - send/receive replies for AST messages again
6586 - if the client says that it doesn't have the lock, cancel it on the
6588 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6589 mark it as destroyed, it will all work out when decref is called
6590 - fix module use counts (22, 581)
6592 - ASTs now expect a reply (server cancels lock on error reply)
6594 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6597 - fix many simultaneous client startup (392)
6598 - fix dentry->d_it clobbering
6599 - credentials weren't being shipped for readdir/getattr operations
6600 - remove invalid assertions triggered during some concurrent MD
6602 - proper Lustre versions added (336, 389)
6603 - fix memory leak for create error case (398)
6604 - fix LOV locking bug that would get cli/srv out of sync
6605 - fix echo client over LOV (409)
6606 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6607 - fix double-O_EXCL intent crash (424)
6608 - avoid sending multiple lock CANCELs (352)
6610 - MDS can do multi-client recovery (modulo bugs in new code)
6612 - many updates, edits, cleanups
6614 2002-11-18 Phil Schwan <phil@clusterfs.com>
6617 - fix null d_it dereference (346)
6618 - fix full OST/dbench hang (333)
6619 - fix permission problem with file removal (286)
6620 - fix removal of OSCs from LOV when they fail
6621 - fix NULL deref during bulk timeout (214)
6622 - fix problems related to multiple filesystems on one MDS (241)
6623 - fixed serious subtle metadata locking bugs
6624 - free locks on clients when inodes are removed due to memory
6626 - fix inode pointer in lock data (285)
6627 - partial support for multiple MDS on a single host (241)
6628 - data locks weren't cancelled at clear_inode time (290, 311)
6629 - intent locks could lead to unbounded lock growth (205)
6630 - added a maximum lock count, an LRU list, and a flusher
6631 - fix multiple rename (365)
6632 - properly abstracted the echo client
6633 - OSC locked 1 byte too many; fixed
6634 - rewrote brw callback code:
6635 - fixed recovery bugs related to LOVs (306)
6636 - fixed too-many-pages-in-one-write crash (191)
6637 - fixed (again) crash in sync_io_timeout (214)
6638 - probably fixed callback-related race (385)
6640 - Add capability to MDS protocol
6641 - LDLM cancellations and callbacks on different portals
6643 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6646 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6647 - defer open object destruction to close time (601981, 138)
6648 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6649 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6650 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6651 - don't call obd_setattr in ll_file_release for destroyed objects
6652 * protocol change to lustre_msg: move |version| and add |flags|
6653 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6654 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6655 * added replay of create, unlink, link and rename operations during
6656 MDS failover; recovery should be much more robust now
6657 * remove failed OSCs from LOVs (only lov_create uses this so far)
6658 * the lustre-HOWTO was brought (more) up to date (582544)
6660 2002-10-23 Phil Schwan <phil@clusterfs.com>
6663 - in-use dentries weren't being reused properly (617851)
6664 - prevent multiple LDLM setup (599178)
6665 - fix LOV size calculations for truncate (617853)
6666 - fix client handling of MDS intent errors (POSIX)
6667 - fix permission bug in lovstripe.c test (624321)
6668 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6669 - truncate past end of file could corrupt data
6670 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6671 - a race in recovery could return ETIMEDOUT to apps (623947)
6672 - building outside the source directory was fixed
6673 * the lustre-HOWTO was brought (more) up to date (582544)
6674 * major progress was made on recovery functionality
6676 2002-10-10 Phil Schwan <phil@clusterfs.com>
6679 - recovery deadlock fix
6680 - rm -rf causes LBUG fix (617817)
6681 - file open by multiple tasks fix (618962)
6682 - directory permissions bugs (602707 and 620007)
6683 - journal_stop fixed with locking (611313)
6684 - O_APPEND failures resolved (618273, perhaps 614459)
6685 - lconf PATH fix (619770)
6686 - IA64 build fix (621450)
6687 - RPC buffer sizes scale with amount of memory
6689 2002-10-01 Phil Schwan <phil@clusterfs.com>
6692 - locks would be cancelled without throwing away data pages,
6693 resulting in inconsistent data (605627)
6694 - inode attributes were not always being refreshed (605627, 612449)
6695 - lconf now continues to cleanup after lctl reports an error
6696 - MDS now enforces user permissions (602707)
6697 - lprocfs cleanup fixed, but not yet enabled (614157)
6698 - fixed infinite server hang, should a client not respond to an AST
6699 - avoid going into recovery if user calls readlink() with a buffer
6700 that's too small (613941)
6701 - AST RPCs no longer require replies (614867) -- this may be changed
6702 - don't crash server if client sends an IOV that's too big (611336)
6703 - fixed lock conversion deadlock (611892)
6704 - fixed the following of symlinks (614622)
6705 * recovery: the server can remove locks from a client that dies, other
6706 clients can make progress
6707 * more extN patch fixes
6708 * compile-time configurable ptlrpc buffer allocations
6710 - collaborative read cache document
6711 - Lustre Lite Performance CDR document-in-progress
6713 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6716 - fix typo in patch-2.4.18
6718 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6721 - clear ptlrpc request each time in handle_incoming_request()
6722 - unlink of files now destroys the object on the OST
6724 2002-09-19 Peter Braam <braam@clusterfs.com>
6726 * add hard link support
6727 * change obdfile creation method
6728 * kernel patch changed
6730 2002-09-19 Peter Braam <braam@clusterfs.com>
6733 - stack overflow bug in extN fixed
6735 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6737 * documentation updates
6738 - add man pages for config tools
6739 - update tests/README to describe testing with new config tools
6740 - finish metadata API descriptions
6741 * bug fixes and cleanups
6742 - statfs workaround for 16TB limit
6743 - LOV stripe allocation improved, can stripe on subset of OSTs
6744 - LOV file size/IO offset was wrong for files > 4GB in size
6745 - object EA data was being dropped, caused files to be unreadable
6746 - memory overflow with non-LOV OST caused memory corruption
6747 - fixed regression tests to work with new config tools, obdfilter
6748 - fixed bug when directory size became larger than 1 block
6749 - fixed bug (for single client case) when PWD was deleted
6750 - invalidate local directory pages when doing intent-based ops
6751 - avoid LDLM oops when lock callback contained bad data
6753 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6755 * documentation updates
6756 * bug fixes and cleanups
6757 - configuration tools
6760 - 64-bit compile warnings
6761 - 64-bit internal statfs data
6763 * test_brw on persistent OST devices
6765 * lprocfs (disabled)
6767 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6769 * documentation updates
6770 * bug fixes and cleanups
6771 * configuration tools
6773 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6776 * many small fixes to 0_5_4
6777 * io/network handling
6778 * thinkos in MDS operations
6780 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6783 * crucial basic fixes to 0.5.3
6784 * IOR, Iozone work over Elan
6787 2002-08-07 Phil Schwan <phil@clusterfs.com>
6788 * version 0_5_3, our first alpha
6789 * we use the new Portals iovs
6790 * documentation updates
6791 * bug fixes and cleanups
6792 * small changes in the DLM wire protocol
6794 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6795 * version 0_5_1 with some initial stability,
6796 * locking on MD and file I/O.
6797 * documentation updates
6798 * several bug fixes since 0.5.0
6799 * small changes in wire protocol
6801 2002-07-18 Phil Schwan <phil@clusterfs.com>
6803 * delivered as Lustre Light Alpha
6804 * fixed a crash after handling invalid MDS requests
6805 * fixed directory pages for architectures with non-4k pages sizes
6807 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6808 * release version v0_4_4
6809 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6811 * Fixes a number of interruption problems with OST operations.
6812 * Update documentation for portals header changes
6813 * Move all wire protocol structs/defines to lustre_idl.h
6814 * Fixes symlink length bug.
6815 * Add tcpdump to repository.
6817 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6818 * release version v0_4_3
6819 * Fixes statfs for inodes on extN.
6820 * Fixes bug in runtests which would delete /etc/hosts.
6821 * Use 64-bit object IDs wherever possible (not into VFS though)
6822 Remove ost_get_info, which is unused by lustre, and out of date.
6824 2002-07-03 Peter Braam <braam@clusterfs.com>
6825 * release version v0_4_2 Fixes a lookup error (type not passed)
6826 * move forward to head of Portals
6827 * move forward to latest Lustre kernel
6829 2002-06-25 Peter Braam <braam@clusterfs.com>
6830 * release version v0_4_1. Hopefully stable on single node use.