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 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.9.1.el7 (RHEL7.7)
8 * Server known to build on the following kernels (others may also work):
9 2.6.32-696.18.7.el6 (RHEL6.9)
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.97 (SLES12 SP3)
13 4.4.0-142 (Ubuntu 16.04)
14 4.15.0-45 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-957.27.2.el7 (RHEL7.6)
19 3.10.0-1062.9.1.el7 (RHEL7.7)
20 4.18.0-80.11.2.el8 (RHEL8.0)
21 4.4.180-94.97 (SLES12 SP3)
22 4.12.14-95.29 (SLES12 SP4)
23 4.15.0-45 (Ubuntu 18.04)
24 * Client known to build on unpatched kernels (others may also work):
25 2.6.32-696.18.7.el6 (RHEL6.9)
26 2.6.32-754.6.3.el6 (RHEL6.10)
27 3.10.0-862.14.4.el7 (RHEL7.5)
28 4.14.0-49.13.1.el7a (RHEL7.5)
29 4.14.0-115.2.2.el7a (RHEL7.6)
30 4.18.0-147.3.1.el8 (RHEL8.1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.0-142 (Ubuntu 16.04)
33 vanilla linux 4.15.0 (ZFS only)
34 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
35 * Recommended ZFS / SPL version: 0.7.13
36 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
37 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
38 stack overflow. For more information, please refer to bugzilla 17630.
39 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
40 the re-exporting nodes to be mounted with "32bitapi" mount option
41 * Note for Ubuntu 14.04.5:
42 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
43 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
47 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
48 for currently supported client and server kernel versions.
49 * Server primary kernels built and tested during release cycle:
50 3.10.0-957.27.2.el7 (RHEL7.6)
51 3.10.0-1062.1.1.el7 (RHEL7.7)
52 * Server known to build on the following kernels (others may also work):
53 2.6.32-696.18.7.el6 (RHEL6.9)
54 3.10.0-862.14.4.el7 (RHEL7.5)
55 4.4.120-92.70 (SLES12 SP2)
56 4.4.180-94.97 (SLES12 SP3)
57 4.4.0-142 (Ubuntu 16.04)
58 4.15.0-45 (Ubuntu 18.04)
59 vanilla linux 4.6.7 (ZFS only)
60 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
61 * Client primary unpatched kernels built and tested during release cycle:
62 3.10.0-957.27.2.el7 (RHEL7.6)
63 3.10.0-1062.1.1.el7 (RHEL7.7)
64 4.18.0-80.7.1.el8 (RHEL8.0)
65 4.4.180-94.97 (SLES12 SP3)
66 4.12.14-95.29 (SLES12 SP4)
67 4.15.0-45 (Ubuntu 18.04)
68 * Client known to build on unpatched kernels (others may also work):
69 2.6.32-696.18.7.el6 (RHEL6.9)
70 2.6.32-754.6.3.el6 (RHEL6.10)
71 3.10.0-862.14.4.el7 (RHEL7.5)
72 4.14.0-49.13.1.el7a (RHEL7.5)
73 4.14.0-115.2.2.el7a (RHEL7.6)
74 4.4.120-92.70 (SLES12 SP2)
75 4.4.0-142 (Ubuntu 16.04)
76 vanilla linux 4.15.0 (ZFS only)
77 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
78 * Recommended ZFS / SPL version: 0.7.13
79 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
80 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
81 stack overflow. For more information, please refer to bugzilla 17630.
82 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
83 the re-exporting nodes to be mounted with "32bitapi" mount option
84 * Note for Ubuntu 14.04.5:
85 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
86 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
90 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
91 for currently supported client and server kernel versions.
92 * Server known to build on the following kernels (others may also work):
93 2.6.32-696.18.7.el6 (RHEL6.9)
94 3.10.0-862.14.4.el7 (RHEL7.5)
95 3.10.0-957.10.1.el7 (RHEL7.6)
96 4.4.120-92.70 (SLES12 SP2)
97 4.4.162-94.69 (SLES12 SP3)
98 4.4.0-142 (Ubuntu 16.04)
99 4.15.0-45 (Ubuntu 18.04)
100 vanilla linux 4.6.7 (ZFS only)
101 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
102 * Client known to build on unpatched kernels (others may also work):
103 2.6.32-696.18.7.el6 (RHEL6.9)
104 2.6.32-754.6.3.el6 (RHEL6.10)
105 3.10.0-862.14.4.el7 (RHEL7.5)
106 3.10.0-957.10.1.el7 (RHEL7.6)
107 4.14.0-49.13.1.el7a (RHEL7.5)
108 4.14.0-115.2.2.el7a (RHEL7.6)
109 4.4.120-92.70 (SLES12 SP2)
110 4.4.162-94.69 (SLES12 SP3)
111 4.12.14-95.13 (SLES12 SP4)
112 4.4.0-142 (Ubuntu 16.04)
113 4.15.0-45 (Ubuntu 18.04)
114 vanilla linux 4.15.0 (ZFS only)
115 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
116 * Recommended ZFS / SPL version: 0.7.13
117 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
118 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
119 stack overflow. For more information, please refer to bugzilla 17630.
120 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
121 the re-exporting nodes to be mounted with "32bitapi" mount option
122 * Note for Ubuntu 14.04.5:
123 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
124 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
129 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
130 for currently supported client and server kernel versions.
131 * Server known to build on the following kernels (others may also work):
132 2.6.32-696.18.7.el6 (RHEL6.9)
133 3.10.0-862.14.4.el7 (RHEL7.5)
134 3.10.0-957.10.1.el7 (RHEL7.6)
135 4.4.120-92.70 (SLES12 SP2)
136 4.4.162-94.69 (SLES12 SP3)
137 4.4.0-142 (Ubuntu 16.04)
138 4.15.0-45 (Ubuntu 18.04)
139 vanilla linux 4.6.7 (ZFS only)
140 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
141 * Client known to build on unpatched kernels (others may also work):
142 2.6.32-696.18.7.el6 (RHEL6.9)
143 2.6.32-754.6.3.el6 (RHEL6.10)
144 3.10.0-862.14.4.el7 (RHEL7.5)
145 3.10.0-957.10.1.el7 (RHEL7.6)
146 4.14.0-49.13.1.el7a (RHEL7.5)
147 4.14.0-115.2.2.el7a (RHEL7.6)
148 4.4.120-92.70 (SLES12 SP2)
149 4.4.162-94.69 (SLES12 SP3)
150 4.12.14-95.13 (SLES12 SP4)
151 4.4.0-142 (Ubuntu 16.04)
152 4.15.0-45 (Ubuntu 18.04)
153 vanilla linux 4.15.0 (ZFS only)
154 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
155 * Recommended ZFS / SPL version: 0.7.13
156 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
157 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
158 stack overflow. For more information, please refer to bugzilla 17630.
159 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
160 the re-exporting nodes to be mounted with "32bitapi" mount option
161 * Note for Ubuntu 14.04.5:
162 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
163 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
167 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
168 for currently supported client and server kernel versions.
169 * Server known to build on the following kernels (others may also work):
170 2.6.32-696.18.7.el6 (RHEL6.9)
171 3.10.0-862.14.4.el7 (RHEL7.5)
172 3.10.0-957.el7 (RHEL7.6)
173 4.4.120-92.70 (SLES12 SP2)
174 4.4.162-94.69 (SLES12 SP3)
175 4.4.0-131 (Ubuntu 16.04)
176 4.15.0-32 (Ubuntu 18.04)
177 vanilla linux 4.6.7 (ZFS only)
178 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
179 * Client known to build on unpatched kernels (others may also work):
180 2.6.32-696.18.7.el6 (RHEL6.9)
181 2.6.32-754.6.3.el6 (RHEL6.10)
182 3.10.0-862.14.4.el7 (RHEL7.5)
183 3.10.0-957.el7 (RHEL7.6)
184 4.14.0-49.13.1.el7a (RHEL7.5)
185 4.14.0-115.2.2.el7a (RHEL7.6)
186 4.4.120-92.70 (SLES12 SP2)
187 4.4.162-94.69 (SLES12 SP3)
188 4.4.0-131 (Ubuntu 16.04)
189 4.15.0-32 (Ubuntu 18.04)
190 vanilla linux 4.15.0 (ZFS only)
191 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
192 * Recommended ZFS / SPL version: 0.7.9
193 * Tested with ZFS / SPL version: 0.7.9
194 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
195 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
196 stack overflow. For more information, please refer to bugzilla 17630.
197 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
198 the re-exporting nodes to be mounted with "32bitapi" mount option
199 * Note for Ubuntu 14.04.5:
200 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
201 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
203 --------------------------------------------------------------------------------
206 2018-04-03 Intel Corporation
208 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
209 for currently supported client and server kernel versions.
210 * Server known to build on patched kernels:
211 2.6.32-431.29.2.el6 (RHEL6.5)
212 2.6.32-504.30.3.el6 (RHEL6.6)
213 2.6.32-573.26.1.el6 (RHEL6.7)
214 2.6.32-642.15.1.el6 (RHEL6.8)
215 2.6.32-696.18.7.el6 (RHEL6.9)
216 3.10.0-693.21.1.el7 (RHEL7.4)
217 3.0.101-0.47.71 (SLES11 SP3)
218 3.0.101-107 (SLES11 SP4)
219 3.12.74-60.64.40 (SLES12 SP1)
220 4.4.114-92.64 (SLES12 SP2)
221 4.4.114-94.11 (SLES12 SP3)
222 3.13.0-101 (Ubuntu 14.04, ZFS only)
223 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
224 4.4.0-116 (Ubuntu 16.04)
225 vanilla linux 4.6.7 (ZFS only)
226 * Client known to build on unpatched kernels:
227 2.6.32-431.29.2.el6 (RHEL6.5)
228 2.6.32-504.30.3.el6 (RHEL6.6)
229 2.6.32-573.26.1.el6 (RHEL6.7)
230 2.6.32-642.15.1.el6 (RHEL6.8)
231 2.6.32-696.18.7.el6 (RHEL6.9)
232 3.10.0-693.21.1.el7 (RHEL7.4)
233 3.0.101-0.47.71 (SLES11 SP3)
234 3.0.101-107 (SLES11 SP4)
235 3.12.74-60.64.40 (SLES12 SP1)
236 4.4.114-92.64 (SLES12 SP2)
237 4.4.114-94.11 (SLES12 SP3)
238 3.13.0-101 (Ubuntu 14.04)
239 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
240 4.4.0-116 (Ubuntu 16.04)
242 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
243 * Recommended ZFS / SPL version: 0.7.6
244 * Tested with ZFS / SPL version: 0.7.6
245 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
246 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
247 stack overflow. For more information, please refer to bugzilla 17630.
248 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
249 the re-exporting nodes to be mounted with "32bitapi" mount option
250 * Note for Ubuntu 14.04.5:
251 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
252 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
254 --------------------------------------------------------------------------------
256 2017-07-05 Intel Corporation
258 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
259 for currently supported client and server kernel versions.
260 * Server known to build on patched kernels:
261 2.6.32-431.29.2.el6 (RHEL6.5)
262 2.6.32-504.30.3.el6 (RHEL6.6)
263 2.6.32-573.26.1.el6 (RHEL6.7)
264 2.6.32-642.15.1.el6 (RHEL6.8)
265 2.6.32-696.3.1.el6 (RHEL6.9)
266 3.10.0-514.21.1.el7 (RHEL7.3)
267 3.0.101-0.47.71 (SLES11 SP3)
268 3.0.101-100 (SLES11 SP4)
269 3.12.74-60.64.40 (SLES12 SP1)
270 4.4.59-92.17 (SLES12 SP2)
271 vanilla linux 4.6.7 (ZFS only)
272 * Client known to build on unpatched kernels:
273 2.6.32-431.29.2.el6 (RHEL6.5)
274 2.6.32-504.30.3.el6 (RHEL6.6)
275 2.6.32-573.26.1.el6 (RHEL6.7)
276 2.6.32-642.15.1.el6 (RHEL6.8)
277 2.6.32-696.3.1.el6 (RHEL6.9)
278 3.10.0-514.21.1.el7 (RHEL7.3)
279 3.0.101-0.47.71 (SLES11 SP3)
280 3.0.101-100 (SLES11 SP4)
281 3.12.74-60.64.40 (SLES12 SP1)
282 4.4.59-92.17 (SLES12 SP2)
284 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
285 * Recommended ZFS / SPL version: 0.6.5.9
286 * Tested with ZFS / SPL version: 0.6.5.9
287 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
288 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
289 stack overflow. For more information, please refer to bugzilla 17630.
290 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
291 the re-exporting nodes to be mounted with "32bitapi" mount option
293 --------------------------------------------------------------------------------
295 12-07-2016 Intel Corporation
297 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
298 for currently supported client and server kernel versions.
299 * Server known to build on patched kernels:
300 2.6.32-431.29.2.el6 (RHEL6.5)
301 2.6.32-504.30.3.el6 (RHEL6.6)
302 2.6.32-573.26.1.el6 (RHEL6.7)
303 2.6.32-642.11.1.el6 (RHEL6.8)
304 3.10.0-514.el7 (RHEL7.3)
305 3.0.101-0.47.71 (SLES11 SP3)
306 3.0.101-80 (SLES11 SP4)
307 3.12.62-60.64.8 (SLES12 SP1)
308 vanilla linux 4.5.7 (ZFS only)
309 * Client known to build on unpatched kernels:
310 2.6.32-431.29.2.el6 (RHEL6.5)
311 2.6.32-504.30.3.el6 (RHEL6.6)
312 2.6.32-573.26.1.el6 (RHEL6.7)
313 2.6.32-642.11.1.el6 (RHEL6.8)
314 3.10.0-514.el7 (RHEL7.3)
315 3.0.101-0.47.71 (SLES11 SP3)
316 3.0.101-80 (SLES11 SP4)
317 3.12.62-60.64.8 (SLES12 SP1)
319 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
320 * Recommended ZFS / SPL version: 0.6.5.8
321 * Tested with ZFS / SPL version: 0.6.5.8
322 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
323 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
324 stack overflow. For more information, please refer to bugzilla 17630.
325 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
326 the re-exporting nodes to be mounted with "32bitapi" mount option
328 --------------------------------------------------------------------------------
330 02-29-2016 Intel Corporation
332 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
333 for currently supported client and server kernel versions.
334 * Server known to build on patched kernels:
335 2.6.32-431.29.2.el6 (RHEL6.5)
336 2.6.32-504.30.3.el6 (RHEL6.6)
337 2.6.32-573.12.1.el6 (RHEL6.7)
338 3.10.0-327.3.1.el7 (RHEL7.2)
339 3.0.101-0.47.71 (SLES11 SP3)
340 3.0.101-68 (SLES11 SP4)
341 vanilla linux 4.2.1 (ZFS only)
342 * Client known to build on unpatched kernels:
343 2.6.32-431.29.2.el6 (RHEL6.5)
344 2.6.32-504.30.3.el6 (RHEL6.6)
345 2.6.32-573.12.1.el6 (RHEL6.7)
346 3.10.0-327.3.1.el7 (RHEL7.2)
347 3.0.101-0.47.71 (SLES11 SP3)
348 3.0.101-68 (SLES11 SP4)
351 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
352 * Recommended ZFS / SPL version: 0.6.4.2
353 * Tested with ZFS / SPL version: 0.6.4.2
354 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
355 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
356 stack overflow. For more information, please refer to bugzilla 17630.
357 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
358 the re-exporting nodes to be mounted with "32bitapi" mount option
360 --------------------------------------------------------------------------------
362 03-10-2015 Intel Corporation
364 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
365 for currently supported client and server kernel versions.
366 * Server known to build on patched kernels:
367 2.6.32-431.29.2.el6 (RHEL6.5)
368 2.6.32-504.8.1.el6 (RHEL6.6)
369 3.0.101-0.46 (SLES11 SP3)
370 * Client known to build on unpatched kernels:
371 2.6.32-431.29.2.el6 (RHEL6.5)
372 2.6.32-504.8.1.el6 (RHEL6.6)
373 3.10.0-123.20.1.el7 (RHEL7)
374 3.0.101-0.46 (SLES11 SP3)
375 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
376 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
377 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
378 stack overflow. For more information, please refer to bugzilla 17630.
379 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
380 the re-exporting nodes to be mounted with "32bitapi" mount option
382 Severity : enhancement
384 Description: control OST-index in IDIF via ROCOMPAT flag.
385 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
386 last_rcvd file. For new formatted OST device, it will be auto set;
387 for the case of upgrading from old OST device, you can enable it
388 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
389 enabled, for new created OST-object, its IDIF-in-LMA will contain
390 the OST-index; for the existing OST-object, the OSD will convert
391 old format IDIF as new format IDIF with OST-index stored in the
392 LMA EA when accessing such OST-object or via OI scrub. Once such
393 flag is enabled, it cannot be reverted back, so the system cannot
394 be downgraded to the orignal incompatible version.
396 --------------------------------------------------------------------------------
398 07-30-2014 Intel Corporation
400 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
401 for currently supported client and server kernel versions.
402 * Server known to build on patched kernels:
403 2.6.32-431.20.3.el6 (RHEL6.5)
404 3.0.101-0.31 (SLES11 SP3)
405 * Client known to build on unpatched kernels:
406 2.6.32-431.20.3.el6 (RHEL6.5)
407 3.0.101-0.31 (SLES11 SP3)
410 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
411 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
412 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
413 stack overflow. For more information, please refer to bugzilla 17630.
414 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
415 the re-exporting nodes to be mounted with "32bitapi" mount option
417 --------------------------------------------------------------------------------
419 10-31-2013 Intel Corporation
421 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
422 * Server builds for kernels:
423 2.6.32-358.18.1.el6 (RHEL6.4)
424 * Client builds for unpatched kernels:
425 2.6.32-358.18.1.el6 (RHEL6.4)
426 3.0.92-0.8 (SLES11 SP3)
428 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
429 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
430 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
431 stack overflow. For more information, please refer to bugzilla 17630.
432 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
433 the re-exporting nodes to be mounted with "32bitapi" mount option
435 --------------------------------------------------------------------------------
437 05-30-2013 Intel Corporation
439 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
440 * Server builds for kernels:
441 2.6.32-358.6.2.el6 (RHEL6.4)
442 * Client builds for unpatched kernels:
443 2.6.32-358.6.2.el6 (RHEL6.4)
444 3.0.74-0.6.6 (SLES11 SP2)
446 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
447 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
448 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
449 stack overflow. For more information, please refer to bugzilla 17630.
450 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
451 the re-exporting nodes to be mounted with "32bitapi" mount option
453 --------------------------------------------------------------------------------
455 9-30-2012 Whamcloud, Inc.
457 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
458 * Server support for kernels:
459 2.6.32-279.5.1.el6 (RHEL6)
460 * Client support for unpatched kernels:
461 2.6.32-279.5.1.el6 (RHEL6)
463 2.6.18-238.19.1.el5 (RHEL5)
464 2.6.16.60-0.69.1 (SLES10)
465 * Recommended e2fsprogs version: 1.41.90.wc4
466 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
467 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
468 stack overflow. For more information, please refer to bugzilla 17630.
469 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
470 the re-exporting nodes to be mounted with "32bitapi" mount option
472 Severity : enhancement
474 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
476 Severity : enhancement
478 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
480 Severity : enhancement
482 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
484 Severity : enhancement
486 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
488 Severity : enhancement
490 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
492 --------------------------------------------------------------------------------
494 04-30-2012 Whamcloud, Inc.
496 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
497 * Server support for kernels:
498 2.6.32-220.4.2.el6 (RHEL6)
499 * Client support for unpatched kernels:
500 2.6.32-220.4.2.el6 (RHEL6)
502 2.6.18-238.19.1.el5 (RHEL5)
503 2.6.16.60-0.69.1 (SLES10)
504 * Recommended e2fsprogs version: 1.41.90.wc4
505 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
506 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
507 stack overflow. For more information, please refer to bugzilla 17630.
508 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
509 the re-exporting nodes to be mounted with "32bitapi" mount option
511 Severity : enhancement
513 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
515 Severity : enhancement
517 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
519 Severity : enhancement
521 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
523 Severity : enhancement
525 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
527 --------------------------------------------------------------------------------
529 09-30-2011 Whamcloud, Inc.
531 * Server support for kernels:
532 2.6.18-238.19.1.el5 (RHEL5)
533 2.6.32-131.6.1.el6 (RHEL6)
534 * Client support for unpatched kernels:
535 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
536 2.6.18-238.19.1.el5 (RHEL5)
537 2.6.32-131.6.1.el6 (RHEL6)
538 2.6.29.4-167.fc11 (FC11)
539 2.6.16.60-0.69.1 (SLES10)
541 * Recommended e2fsprogs version: 1.41.90.wc3
542 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
543 Lustre filesystem with 4K stack may cause a stack overflow. For more
544 information, please refer to bugzilla 17630.
545 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
546 mounted with "32bitapi" mount option
548 Severity : enhancement
549 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
551 Severity : enhancement
552 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
554 Severity : enhancement
555 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
558 Frequency : only with SLES10
560 Description: Use OFED "KMP" provided by Novell
561 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
562 Lustre is now built against this package. That means you need to
563 install the ofed-kmp package from Novell for the patchless client.
564 Note that the ofed-kmp that Novell ships may not exactly match the
565 kernel version but should still be compatible.
567 Severity : enhancement
569 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
570 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
572 --------------------------------------------------------------------------------
574 2010-07-15 Oracle, Inc.
576 * Server support for kernels:
577 2.6.18-164.11.1.el5 (RHEL 5)
578 2.6.18-164.11.1.0.1.el5 (OEL 5)
579 * Client support for unpatched kernels:
580 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
581 2.6.18-164.11.1.el5 (RHEL 5),
582 2.6.18-164.11.1.0.1.el5 (OEL 5)
583 2.6.16.60-0.42.8 (SLES 10),
585 2.6.29.4-167.fc11 (FC11)
586 * Recommended e2fsprogs version: 1.41.10-sun2
588 Severity : enhancement
590 Description: Distribute lfs_migrate script with lustre
591 Details : The lfs_migrate mentioned in the manual is now included in the
592 lustre tree in lustre/scripts and will distributed in lustre RPMs.
594 Severity : enhancement
596 Description: Failover nid list added to import proc file.
598 Severity : enhancement
600 Description: "som_preview" mount option required to use SOM
601 Details : The clients will need to use som_preview mount option
605 Description: Lustre does not do 1MB IOs to HW RAID
606 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
608 Severity : enhancement
610 Description: lfs quota output is not very convenient for awk/sed-parsing
611 Details : Some positions in lfs quota output table could be empty or
612 non-empty which made it hard to parse it with scripts, now
613 a dash is put instead of space where there is not supposed
618 Description: Update to OFED 1.5.1
619 Details : OFED 1.5.1 will be used in environments that don't use
620 vendor supplied OFED.
624 Description: Remove lnet. prefix from lctl params display
625 Details : Also renamed lnet.memused to lnet_memused to
630 Description: Use conf_param -d to remove permanant settings.
634 Description: Add list_param -R
635 Details : lctl list_param now supports recursive mode.
638 Bugzilla : 21829, 22312
639 Description: Fix llobdstat.
640 Details : 1. Now updating the write_bytes counter
641 2. Port script fixes from 1.8.
645 Description: decrease the usage of memory on clients.
646 Details : 1. On clients, recycle dentries and inodes unused.
647 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
652 Description: llobdstat fix and enhancement
653 Details : add a counter to set a limit to how many samples will be returned
654 fix a wildcard in the path to limit to obdfilter stats only
658 Description: lfs check only allowed for root
659 Details : Code cleanups and added a regresion test of non-root use of
664 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
665 Details : Fix hr->hr_index overload.
669 Description: Fix FIEMAP functionality.
673 Description: FC11 patchless client support added
677 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
678 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
682 Description: New acc-sm script and various test-framework enhancments.
686 Description: Previously landed patch was reverted.
690 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
694 Description: Add ldiskfs maxdirsize mount option.
698 Description: Fix mkfs.lustre support for 16TB LUNs.
702 Description: Print mmp_check_interval
703 Details : print mmp_check_interval and make it possible to abort
704 mount operation in case it takes too long.
708 Description: Add cascading_rw to lustre mpi tests.
712 Description: Store SOM data in EA.
713 Details : Also several SOM related cleanups.
717 Description: Add refcount to md_open_data
721 Description: Use refcount on lov_request_set to prevent memory corruption.
725 Description: Statahead: use parent dir's i_mutex for synchronization.
729 Description: loadgen improvements.
733 Description: Enable peer health detection in ksocklnd and ko2iblnd.
737 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
741 Description: Fix race on obd_devcie:md_stats
745 Description: Fix SLES 11 client builds.
746 Details : Resolved uninitalized pointer gcc warnings.
750 Description: Allow non-root access for "lfs check".
751 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
755 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
757 Severity : enhancement
759 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
760 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
764 Description: Sanity test_27p fails.
765 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
766 Sanity tests 27 were cleaned up and improved.
770 Description: Unify req format on client/servers
771 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
772 buffers and swabbers approach..
777 Description: drop join file stuff.
779 Severity : enhancement
781 Description: Add 2.6.27 and SLES11 (patchless client) support
785 Description: Fix in ptlrpc_expire_one_request() to print the signed time
788 Severity : enhancement
790 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
793 Severity : enhancement
795 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
797 Severity : enhancement
799 Description: correctly handle big reply message.
800 Details : send LNet event if reply is bigger than buffer and adjust this buffer
805 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
807 Severity : enhancement
809 Description: Add support for OEL5.
813 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
815 Severity : enhancement
817 Description: "weak-modules" support
818 Details : Implement "weak-modules" support which enables kernel modules
819 to be used with any kernel that implements the same kABI. In
820 order to achieve this modules are now installed in
821 /lib/modules/$(uname -r)/updates/kernel on all distributions.
825 Description: client could not reconnect to OST because of an active request.
826 Details : The client now retries to reconnect to the same server, if a
827 connect request failed with EBUSY or -EAGAIN. Test case 35b
828 conf-sanity.sh has been created.
832 Description: Don't allow make backward step on assignin osc next id.
833 Details : race between allocation next id and ll_sync thread can be cause
834 of set wrong osc next id and can be kill valid ost objects.
838 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
842 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
846 Description: Error handling in osc_statfs_interpret() has been improved.
847 Details : Check in osc_statfs_interpret() for EBADR.
851 Description: Do not send statfs() requests to OSTs disabled by administrator.
852 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
854 Severity : enhancement
856 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
858 Severity : enhancement
860 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
862 Severity : enhancement
864 Description: Enhancement for lfs(1) command to use numeric uid/gid.
868 Description: Improved cleanup procedure for conf-sanity test_46a.
872 Description: Do not start bulk transfer, if client tried to reconnect.
873 Details : Do not start bulk transfer, if client tried to reconnect,
874 but failed due to active rpcs. Check for exp_abort_active_req
875 has been implemented.
879 Description: Prevent inconsistences between linux and lustre mount structures.
880 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
881 Make the sleep interruptible.
883 Severity : enhancement
885 Description: provide server to client comms path
886 Details : server to client communications path via new LDLM_SET_INFO rpc
890 Description: truncate starts GFP_FS allocation under transaction causing deadlock
891 Details : ldiskfs_truncate calls grab_cache_page which may start page
892 allocation under an open transaction. This may lead to
893 calling prune_icache with consequent lustre reentrance.
897 Description: send correctly lsm on open replay
898 Details : MDS is trust to LSM size on replay open, but client can set wrong size
903 Description: enable client interface failover
904 Details : When a child reconnects from another NID, properly update export
905 nid hash position and ldlm reverse import.
907 Severity : enhancement
909 Description: Communicate OST degraded/readonly state via statfs to MDS
910 Details : Flags in the statfs returned from OSTs indicate whether the
911 OST is in a degraded RAID state, or if the filesystem has
912 turned read-only after a filesystem error is detected.
917 Description: don't panic if EPROTO was hit when reading symlink
918 Details : correctly handling request reference in error cases.
920 Severity : enhancement
922 Description: Add LustreNetLink, a kernel-userspace communcation path.
924 Severity : enhancement
926 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
931 Description: access to llog context before init.
932 Details : move handling CATALOGS file at osc layer and forbid access to llog
936 Frequency : rare, only when client is writing to OST on local node
938 Description: Avoid deadlock for local client writes
939 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
940 memory freeing context. This allows OST threads to set the
941 PF_MEMALLOC flag on task structures in order to allocate memory
942 from reserved pools and complete IO.
943 Use GFP_HIGHUSER for OST allocations for non-local client writes,
944 so that the OST threads generate memory pressure and allow
945 inactive pages to be reclaimed.
947 Severity : enhancement
949 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
954 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
955 Details : move ldlm namespace creation in setup phase to avoid grab
956 _lprocfs_lock with cli_sem held.
960 Description: Temporarily disable grant shrink.
961 Details : Disable the feature for debugging.
964 Frequency : only when formatting test filesystems
966 Description: Unable to run several mkfs.lustre on loop devices at the same time
967 Details : mkfs.lustre returns error 256 on the concurrent loop devices
968 formatting. The solution is to proper handle the error.
970 Severity : enhancement
972 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
974 Severity : enhancement
976 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
981 Description: resolve race between obd_disconnect and class_disconnect_exports
982 Details : if obd_disconnect will be called to already disconnected export he
983 forget release one reference and osc module can't unloaded.
985 Severity : enhancement
987 Description: Allow tuning service thread via /proc
988 Details : For each service a new
989 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
990 created that can be used to set min/max thread counts, and get the
991 current number of running threads.
996 Descriptoin: don't return error if have partially created objects for file.
997 Details : lov_update_create_set uses set->set_success as index for created
998 objects, so if some requests will be failed, they will have hole
999 at end of array and qos_shrink_lsm for allocate correct lsm.
1001 Severity : enhancement
1003 Description: Update OFED support to 1.4.1
1008 Description: Reduce small size read RPC
1009 Details : Set read-ahead limit for every file and only do read-ahead when
1010 available read-ahead pages are bigger than 1M to avoid small size
1013 Severity : enhancement
1015 Description: move AT tunable parameters for more consistent usage
1016 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1018 Severity : enhancement
1020 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1021 Details : allow skip disconnected ost for send statfs request and hide error
1027 Description: fix racy locking of mballoc block bitmaps causing BUG
1028 Details : The locking of the mballoc buddy bitmap and the in-memory
1029 block bitmap was using two different spin locks in some
1030 cases. This made it possible to incorrectly access the
1031 mballoc bitmap while another process was modifying it,
1032 causing a sanity assertion to fail. While no on-disk corruption
1033 was reported, there was some risk of this happening.
1036 Frequency : rare, on llog test 6
1038 Descriptoin: don't allow connect to already connected import
1039 Details : allowing connect to already connected import is hide connecting problem.
1044 Descriptoin: allow kill process which wait statahead result
1045 Details : for some reasons 'ls' can stick in waiting result from statahead,
1046 in this case need way for kill this process.
1048 Severity : enhancement
1050 Description: Add state history info file, enhance import info file
1051 Details : Track import connection state changes in a new osc/mdc proc file;
1052 add overview-type data to the osc/mdc import proc file.
1054 Severity : enhancement
1056 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1058 Severity : enhancement
1060 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1065 Descriptoin: OST grant too much space to client even there are not enough space.
1066 Details : Client will shrink its grant cache to OST if there are no write
1067 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1068 this grant cache if there are already not enough avaible space
1069 (left_space < total_clients * 32M).
1072 Frequency : start MDS on uncleanly shutdowned MDS device
1074 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1075 Details : stay in waiting mds<>ost recovery finished produce random bugs
1076 due race between two ll_sync thread for one lov target. send
1077 ACTIVATE event only if connect realy finished and import have
1081 Frequency : rare, connect and disconnect target at same time
1083 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1084 Details : don't call obd_disconnect under lov_lock. this long time
1085 operation and can block ptlrpcd which answer to connect request.
1090 Descriptoin: don't lose wakeup for imp_recovery_waitq
1091 Details : recover_import_no_retry or invalidate_import and import_close can
1092 both sleep on imp_recovery_waitq, but we was send only one wakeup
1096 Frequency : always with long access acl
1098 Descriptoin: mds can't pack reply with long acl.
1099 Details : mds don't control size of acl but they limited by reint/getattr
1102 Severity : enhancement
1104 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1106 Severity : enhancement
1108 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1111 Frequency : start MDS on uncleanly shutdowned MDS device
1113 Descriptoin: aborting recovery hang on MDS
1114 Details : don't throttle destroy RPCs for the MDT.
1117 Frequency : on remount
1119 Description: external journal device not working after the remount
1120 Details : clear dev_rdonly flag for external journal devices in
1126 Description: shutdown vs evict race
1127 Details : client_disconnect_export vs connect request race.
1128 if client will evicted at this time - we start invalidate
1129 thread without referece to import and import can be freed
1133 Frequency : rare, need acl's on inode.
1135 Description: client can't handle ost additional correctly
1136 Details : if ost was added after client connected to mds client can have
1137 hit lnet_try_match_md ... to big messages to wide striped files.
1138 in this case need teach client to handle config events about add
1139 lov target and update client max ea size at that event.
1141 Severity : enhancement
1143 Description: Changelogs
1144 Details : Changelogs are a lightweight mechanism to track filesystem
1145 metadata and namespace changes. The changelog is recorded
1146 permanently on the MDTs, and is periodically "consumed" / purged
1147 when records are no longer needed.
1149 Severity : enhancement
1151 Description: compact fld format with extents
1152 Details : Store range of seq rather than every seq in FLD. Seq
1153 controller update FLD rather than clients. In Case of CMD, mdt0
1154 has FLD, all other metadata server act as non persistent proxy
1155 for FLD queries and cache fld entries in fld cache.
1160 Description: don't skip ost target if they assigned to file
1161 Details : Drop slow OSCs if we can, but not for requested start idx.
1162 This means "if OSC is slow and it is not the requested
1163 start OST, then it can be skipped, otherwise skip it only
1164 if it is inactive/recovering/out-of-space.
1168 Description: more cleanup in mds_lov
1169 Details : not send LOV EA under replay, we can't know about they size at this
1170 time. Don't allow client connect to mds before any ost connected,
1171 for avoid problems with LOV EA size and returning EIO to client.
1173 Severity : enhancement
1175 Description: Interoperability at server side (Disk interoperability)
1177 Severity : enhancement
1179 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1181 Severity : enhancement
1183 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1185 Severity : enhancement
1187 Description: New client IO stack (CLIO).
1189 Severity : enhancement
1191 Description: Commit on sharing. Eliminate inter-client dependencies between
1192 uncommitted transactions by doing transaction commits.
1193 Thereby clients may recovery independently.
1196 Frequency : Create a symlink file with a very long name
1198 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1199 Details : If there is no extra space in the request for early cancels,
1200 ldlm_req_handles_avail() returns 0 instead of a negative value.
1202 Severity : enhancement
1204 Description: Add /proc entry for import status
1205 Details : The mdc, osc, and mgc import directories now have
1206 an import directory that contains useful import data for debugging
1207 connection problems.
1209 Severity : enhancement
1211 Description: Re-disable certain /proc logging
1212 Details : Enable and disable client's offset_stats, extents_stats and
1213 extents_stats_per_process stats logging on the fly.
1216 Frequency : Only on FC kernels 2.6.22+
1218 Description: oops in statahead
1219 Details : Do not drop reference count for the dentry from VFS when lookup,
1220 VFS will do that by itself.
1222 Severity : enhancement
1224 Description: Generic /proc file permissions
1225 Details : Set /Proc file permissions in a more generic way to enable non-
1226 root users operate on some /proc files.
1230 Description: Hitting mdc_commit_close() ASSERTION
1231 Details : Properly handle request reference release in
1232 ll_release_openhandle().
1236 Description: quota recovery deadlock during mds failover
1237 Details : This patch includes att18982, att18236, att18237 in bz14840.
1239 1. fix osts hang when mds does failover with quotaon
1240 2. prevent watchdog storm when osts threads wait for the
1245 Frequency : only patchless client
1246 Description: add workaround for race between add/remove dentry from hash
1248 Severity : enhancement
1250 Description: Allow OST glimpses to return PW locks
1254 Description: LBUG when llog conf file is full
1255 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1259 Description: Prevent import from entering FULL state when server in recovery
1263 Description: service mount cannot take device name with ":"
1264 Details : Only when device name contains ":/" will mount treat it as
1270 Description: replace ptlrpcd with the statahead thread to interpret the async
1271 statahead RPC callback
1275 Frequency : on recovery
1276 Description: I/O failures after umount during fail back
1277 Details : if client reconnected to restarted server we need join to recovery
1278 instead of find server handler is changed and process self
1279 eviction with cancel all locks.
1281 Severity : enhancement
1283 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1285 Severity : enhancement
1287 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1289 Severity : enhancement
1291 Description: Upcall on Lustre log has been dumped
1292 Details : Allow for a user mode script to be called once a Lustre log has
1293 been dumped. It passes the filename of the dumped log to the
1294 script, the location of the script can be specified via
1295 /proc/sys/lnet/debug_log_upcall.
1300 Description: avoid idr_remove called for id which is not allocated.
1301 Details : Move assigment s_dev for clustered nfs to end of initialization,
1302 to avoid problem with error handling.
1307 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1308 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1309 into CONN_USED_HASH and this prodice warning when put connection
1310 again in unused hash.
1313 Severity : enhancement
1315 Description: File striping can now be set to use an arbitrary pool of OSTs.
1317 Severity : enhancement
1319 Description: Export bytes_read/bytes_write count on OSC/OST.
1323 Description: Early reply size mismatch, MGC loses connection
1324 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1325 the connect flags are properly negotiated.
1330 Description: quotas are not honored with O_DIRECT
1331 Details : all writes with the flag O_DIRECT will use grants which leads to
1332 this problem. Now using OBD_BRW_SYNC to guard this.
1336 Description: add quota statistics
1337 Details : 1. sort out quota proc entries and proc code.
1338 2. add quota statistics
1340 Severity : enhancement
1342 Description: enable quota support for HEAD.
1346 Description: Properly propagate oinfo flags from lov to osc for statfs
1347 Details : restore missing copy oi_flags to lov requests.
1349 Severity : enhancement
1351 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1352 : l_getgroups(8), lst(8), routerstat(8)
1353 Details : included man pages for llobdstat(8), llstat(8),
1354 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1356 Severity : enhancement
1358 Description: configure's --enable-quota should check the
1359 : kernel .config for CONFIG_QUOTA
1360 Details : configure is terminated if --enable-quota is passed but
1361 : no quota support is in kernel
1365 Description: Remove portals compatibility
1366 Details : Remove portals compatibility, not interoperable with releases
1371 Description: Resolve device initialization race
1372 Details : Prevent proc handler from accessing devices added to the
1373 obd_devs array but yet be intialized.
1375 Severity : enhancement
1377 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1379 Severity : enhancement
1381 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1386 Description: Using wrong pointer in osc_brw_prep_request
1387 Details : Access to array[-1] can produce panic if kernel compiled with
1388 CONFIG_PAGE_ALLOC enabled
1392 Description: Client runs out of low memory
1393 Details : Consider only lowmem when counting initial number of llap pages
1397 Description: *optional* service tags registration
1398 Details : if the "service tags" package is installed on a Lustre node
1399 When the filesystem is mounted, a local-node service tag will
1400 be created. See http://inventory.sun.com/ for more information
1401 about the Service Tags asset management system.
1405 Description: Kernel BUG tries to release flock
1406 Details : Lustre does not destroy flock lock before last reference goes
1407 away. So always drop flock locks when client is evicted and
1408 perform unlock regardless of successfulness of speaking to MDS.
1412 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1416 Description: sanity test 65a fails if stripecount of -1 is set
1417 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1422 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1423 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1426 Severity : enhancement
1428 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1432 Description: do not process already freed flock
1433 Details : flock can possibly be freed by another thread before it reaches
1434 to ldlm_flock_completion_ast.
1438 Description: LBUG during stress test
1439 Details : Need properly lock accesses the flock deadlock detection list.
1443 Description: oops in page fault handler
1444 Details : kernel page fault handler can return two special 'pages' in error
1445 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1449 Description: timeout with invalidate import.
1450 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1451 should be handled by ptlrpcd. This produce long age waiting and
1452 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1454 Severity : enhancement
1456 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1460 Description: enable MGS and MDT services start separately
1461 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1462 a co-located MGS without starting the MGS, which is a complement
1463 to 'nosvc' mount option.
1467 Frequency : after recovery
1468 Description: precreate to many object's after del orphan.
1469 Details : del orphan st in oscc last_id == next_id and this triger growing
1470 count of precreated objects. Set flag LOW to skip increase count
1471 of precreated objects.
1475 Frequency : rare, on clear nid stats
1476 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1477 Details : when clean nid stats sometimes try destroy live entry,
1478 and this produce panic in free.
1482 Description: Stack overflow during MDS log replay
1483 ease stack pressure by using a thread dealing llog_process.
1487 Description: wait until IO finished before start new when do lock cancel.
1488 Details : VM protocol want old IO finished before start new, in this case
1489 need wait until PG_writeback is cleared until check dirty flag and
1490 call writepages in lock cancel callback.
1492 Severity : enhancement
1494 Description: using special macro for print time and cleanup in includes.
1498 Description: mds_mfd_close() ASSERTION(rc == 0)
1499 Details : In mds_mfd_close(), we need protect inode's writecount change
1500 within its orphan write semaphore to prevent possible races.
1504 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1508 Frequency : rare, on shutdown ost
1509 Description: don't hit live lock with umount ost.
1510 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1511 use shrink_dcache_sb instead.
1515 Description: don't panic with use echo client
1516 Details : echo client pass NULL as client nid pointer and this produce null
1517 pointer dereference.
1521 Description: fix build on ppc32
1522 Details : compile code with -m64 flag produce wrong object file for ppc32.
1526 Description: add message levels for liblustreapi
1530 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1531 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1532 will result in an -EAGAIN and in the exit call path we call
1533 alloc_rr() which will with increasing aggressiveness attempt to
1534 aquire precreated objects on the minimum number of required OSCs.
1538 Description: Use old size assignment to avoid deadlock
1539 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1540 the scheduling while holding a spinlock. We do not need locking
1541 for size in ll_update_inode() because size is only updated from
1542 the MDS for directories or files without objects, so there is no
1543 other place to do the update, and concurrent access to such inodes
1544 are protected by the inode lock.
1548 Description: resolve "_IOWR redefined" build error on SLES10
1552 Description: dump the memory debugging after all modules are unloaded to
1553 suppress false negative in conf_sanity test 39
1555 Severity : enhancement
1557 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1562 Description: fix SLES kernel versioning
1563 Details : the kernel version for our SLES 10 kernel did not include a "-"
1564 before the "smp" at the end. while this was not a problem in
1565 general, it did mean that software trying to use the kernel
1566 version to try to detect a vendor specific kernel would fail.
1567 this was most evident by the OFED build scripts.
1571 Description: Don't update lov_desc members until making sure they are valid
1572 Details : When updating lov_desc members via proc fs, need fix their
1573 validities before doing the real update.
1577 Description: don't put request into delay list while invalidate in flight.
1578 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1579 import in flight. this produce timeout for invalidate and sometimes
1580 can cause stale data.
1584 Frequency : on ppc only
1585 Description: not convert ost objects for directory because it's not exist.
1586 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1590 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1592 Severity : enhancement
1594 Description: Update to support 2.6.22.14 vanilla kernel.
1598 Frequency : rare, on recovery
1599 Description: read procfs can produce deadlock in some situation
1600 Details : Holding lprocfs lock which send rpc can produce block for destroy
1601 obd objects and this also block reconnect with -EALREADY. This
1602 isn't fix all lprocfs bugs - but make it rare.
1604 Severity : enhancement
1606 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1609 Frequency : frequent on X2 node
1611 Description: mdc_set_open_replay_data LBUG
1612 Details : Set replay data for requests that are eligible for replay.
1616 Description: lustre_mgs: operation 101 on unconnected MGS
1617 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1618 MGC, and late on MGC cannot successfully connect to MGS and a lot
1619 of the error messages complaining that MGS is not connected.
1623 Frequency : on network error
1624 Description: panic with double free request if network error
1625 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1626 but it's true only for synchronus enqueue, for async enqueue
1627 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1630 Severity : enhancement
1632 Description: client-side metadata stat-ahead during readdir(directory readahead)
1633 Details : perform client-side metadata stat-ahead when the client detects
1634 readdir and sequential stat of dir entries therein
1637 Frequency : on start mds
1639 Description: Implement get_info(last_id) in obdfilter.
1642 Frequency : occasional
1644 Description: Correctly check stale fid, not start epoch if ost not support SOM
1645 Details : open with flag O_CREATE need set old fid in op_fid3 because
1646 op_fid2 was overwritten with new generated fid, but mds can answer
1647 with one of these two fids and both is not stale. Setattr
1648 incorrectly started an epoch and assume will be called
1649 done_writing, but without SOM done_writing ever being called.
1652 Frequency : rare, depends on device drivers and load
1654 Description: MDS or OSS nodes crash due to stack overflow
1655 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1656 In some cases, in conjunction with device drivers that use a lot
1657 of stack the MDS (or possibly OSS) service threads could overflow
1658 the stack. One change which was identified to consume additional
1659 stack has been reworked to avoid the extra stack usage.
1662 Frequency : occasional
1664 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1665 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1666 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1667 oscc_has_objects and see OSC has no objests and call
1668 oscc_internal_create to resend create request.
1670 Severity : enhancement
1672 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1674 Severity : enhancement
1676 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1679 Frequency : very rare
1681 Description: Fix replay if there is an un-replied request and open
1682 Details : In some cases, older replay request will revert the
1683 mcd->mcd_last_xid on MDS which is used to record the client's
1684 latest sent request.
1686 Severity : enhancement
1688 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1690 Severity : enhancement
1692 Description: Add rhel5 support to HEAD.
1694 Serverity : enhancement
1696 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1701 Description: Don't allow skipping OSTs if index has been specified.
1702 Details : Don't allow skipping OSTs if index has been specified, make
1703 locking in internal create lots better.
1707 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1708 Details : access to bitfield in structure is always rounded to long
1709 and this produce problem with not atomic change any bit.
1713 Description: Lustre make rpms failed.
1714 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1715 builds Lustre rpms from tarball.
1718 Frequency : rare on shutdown ost
1720 Description: If llog cancel was not send before clean_exports phase, this can
1721 produce deadlock in llog code.
1722 Details : If llog thread has last reference to obd and call class_import_put
1723 this produce deadlock because llog_cleanup_commit_master wait when
1724 last llog_commit_thread exited, but this never success because was
1725 called from llog_commit_thread.
1729 Description: allow userland application know is lost one of stripes.
1730 Details : fill lvb_blocks with error code on ost and return it to
1731 application if error flag found.
1735 Description: NULL lov_tgts causing MDS oops
1736 Details : more safe checks for NULL lov_tgts for avoid oops.
1738 Severity : enhancement
1740 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1744 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1745 Details : If an OST is down the MDS will hang indefinitely in
1746 obd_statfs_rqset() waiting for the statfs data. While for
1747 MDS QOS usage of statfs, it should not stuck in waiting.
1749 Severity : enhancement
1751 Description: remote_acl support
1752 Details : Support ACL-based permission check for remote user.
1753 Support setfacl/getfacl for remote user with the utils
1754 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1755 the system "{s,g}etfacl" utils.
1757 Severity : enhancement
1759 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1761 Severity : enhancement
1763 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1767 Description: make mgs_setparam() handle fsname containing dash
1768 Details : fsname containing a dash does not work with lctl conf_param
1770 Severity : enhancement
1772 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1774 Severity : enhancement
1776 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1780 Frequency : rare, at shutdown
1781 Description: access already free / zero obd_namespace.
1782 Details : if client_disconnect_export was called without force flag set,
1783 and exist connect request in flight, this can produce access to
1784 NULL pointer (or already free pointer) when connect_interpret
1785 store ocd flags in obd_namespace.
1789 Frequency : only at startup
1790 Description: not alloc memory with spinlock held.
1791 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1792 if any spinlock held.
1794 Severity : enhancement
1796 Description: make lustre randomly fail allocating memory
1797 Details : Make lustre randomly failed allocating memory for testing purpose.
1799 Severity : enhancement
1801 Description: lost problems with lov objid file
1802 Details : Fixes some scability and access to not inited memory problems
1803 in work with lov objdid file.
1808 Description: lfs find does not continue on file error
1809 Details : Continue other files processing when a file/dir is absent.
1813 Description: Inconsistent usage of lustre_pack_reply()
1814 Details : Standardize the usage of lustre_pack_reply() such that it
1815 always generate a CERROR on failure.
1818 Frequency : occasional
1820 Description: MDS hang or stay in waiting lock
1821 Details : If client receive lock with CBPENDING flag ldlm need send lock
1822 cancel as separate rpc, to avoid situation when cancel request
1823 can't processed due all i/o threads stay in wait lock.
1827 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1831 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1835 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1839 Description: Update to RHEL4 latest kernel.
1841 Severity : enhancement
1843 Description: Build SLES10 patchless client fails
1844 Details : The configure was broken by run ./configure with
1845 --with-linux-obj=.... argument for patchless client. When the
1846 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1847 can't search header adequately. Use absolute path such as
1848 -I($LINUX)/include instead.
1852 Description: interrupt oig_wait produce painc on resend.
1853 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1854 private set, and this produce situation when rq_ptlrpcd_data not
1855 copyed to new allocated request and triggered LBUG on assert
1856 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1857 ptlrpcd set if request is changed and can be safety changed to use
1862 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1863 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1864 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1865 these macros and replace them with their difinitions instead.
1869 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1870 Details : Update which_patch & target file for SLES10 latest kernel.
1873 Description: Remove client patches from SLES 10 kernel.
1874 Details : This causes SLES 10 clients to behave as patchless clients
1875 even on a Lustre-patched (server) kernel.
1877 Severity : enhancement
1879 Description: self-adjustable client's lru lists
1880 Details : use adaptive algorithm for managing client cached locks lru
1881 lists according to current server load, other client's work
1882 pattern, memory activities, etc. Both, server and client
1883 side namespaces provide number of proc tunables for controlling
1886 Severity : enhancement
1888 Description: light-weight GSS support
1889 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1890 and reduce performance overhead.
1892 Severity : enhancement
1894 Description: Linux keyring support
1895 Details : Support using service of Linux keyring for Lustre GSS internal
1896 context refresh/cache mechanism.
1900 Description: Fix errors in lfs documentation
1901 Details : Fixes man pages
1905 Description: don't use GFP_* in generic Lustre code.
1906 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1910 Description: obdclass is limited by single OBD_ALLOC(idarray)
1911 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1915 Description: warnings with build patchless client with vanila 2.6.19 and up
1916 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1917 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1921 Frequency : Only for SLES
1922 Description: sanity_quota fail test_1
1923 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1924 file, which makes TSTID[2] inunique.
1928 Frequency : Only for SLES9
1929 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1930 Details : Update target/ChangeLog/which_patch .
1934 Frequency : Only for test_10 in sanity-quota.sh
1935 Description: a bug in quota test code
1936 Details : checking a test flag in a wrong place
1940 Frequency : Only for test_13 in sanity-quota.sh
1941 Description: a bug in quota test script
1942 Details : shouldn't check overfull allocated quota
1946 Description: jbd statistics
1947 Details : Port older jbd statistics patch for sles10
1949 Severity : enhancement
1951 Description: Add printing inode info into message about error in writepage.
1956 Description: reply_lock_interpret crash due to race with it and lock cancel.
1957 Details : Do not replay locks that are being cancelled. Do not reference
1958 locks by their address during replay, just by their handle.
1962 Frequency : When flocks are used.
1963 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1964 Details : Flock locks might destroy just granted lock if it could be merged
1965 with another existing flock, this is done in completion handler,
1966 so teach ldlm_cli_enquque_fini that this is a valid case for
1972 Description: Oops in read and write path when failing to allocate lock.
1973 Details : Check if lock allocation failed and return error back.
1977 Description: lstripe command fails for valid OST index
1978 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1979 instead of lov->desc.ld_active_tgt_count.
1983 Description: sanity.sh failed test 103
1984 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1989 Description: df doesn't work properly if diskfs blocksize != 4K
1990 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1994 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1995 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1997 Severity : enhancement
1999 Description: Adaptive timeouts
2000 Details : RPC timeouts adapt to changing server load and network
2001 conditions to reduce resend attempts and improve recovery time.
2005 Description: llapi_file_create() does not allow some changes
2006 Details : add llapi_file_open() that allows specifying the mode and
2007 open flags, and also returns an open file handle.
2011 Description: merge and cleanup kernel patches.
2012 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2016 Description: Add journal checksum support.(Kernel part)
2017 Details : The journal checksum feature adds two new flags i.e
2018 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2019 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2020 indicates that the commit block contains the checksum for
2021 the blocks described by the descriptor blocks. Now commit
2022 record can be sent to disk without waiting for descriptor
2023 blocks to be written to disk. This behavior is controlled
2024 using JBD2_FEATURE_ASYNC_COMMIT flag.
2028 Description: OSS needs mutliple precreate threads
2029 Details : Add ability to start more than one create thread per OSS.
2033 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2034 Details : execute lfs setstripe on client
2038 Description: mds_obd_create error creating tmp object
2039 Details : When the user sets quota on root, llog will be affected and can't
2040 create files and write files.
2043 Frequency : Always on ia64 patchless client, and possibly others.
2045 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2046 Details : This allows the patchless client to be loaded on architectures
2047 without this export.
2051 Description: RedHat Update kernel for RHEL5
2052 Details : Add the kernel config file for RHEL5.
2056 Description: RedHat Update kernel for RHEL5
2057 Details : Modify the kernel config file more closer RHEL5.
2061 Description: Build failure against Centos5 (RHEL5)
2062 Details : Define PAGE_SIZE when it isn't present.
2066 Description: test_7 on sanity-quota.sh can't be used on separate machines
2067 Details : add facet to handle it
2071 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2072 Details : with NFS, the anon dentry's parent was set to itself in
2073 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2074 corresponding dentry other than use rec->ur_name.
2076 Severity : enhancement
2078 Description: lfs setstripe enhancement
2079 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2083 Description: enable data checksumming by default
2084 Details : enable checksum by default, allow --disable-checksum
2085 configure option and "-o nochecksum" mount option. Checksums
2086 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2090 Description: lustre support for RHEL5
2091 Details : Add support for RHEL5.
2095 Description: Client eviction due to failover config
2096 Details : after a connection loss, the lustre client should attempt to
2097 reconnect to the last active server first before trying the
2098 other potential connections.
2102 Description: when mds and osts use different quota unit(32bit and 64bit),
2103 quota will be released repeatly.
2104 Details : void sending multiple quota reqs to mds, which will keep the
2105 status between the reqs.
2109 Description: osts not allocated evenly to files
2110 Details : change the condition to increase offset_idx
2113 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2115 Description: Data corruption for OSTs that are formatted larger than 2TB
2117 Details : When generating the bio request for lustre file writes the
2118 sector number would overflow a temporary variable before being
2119 used for the IO. The data reads correctly from Lustre (which
2120 will overflow in a similar manner) but other file data or
2121 filesystem metadata may be corrupted in some cases.
2125 Description: Tune the kernel for good SCSI performance.
2126 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2127 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2132 Description: rewrite ext2-derived code in obdclass/uuid.c
2133 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2134 from scratch preserving functionality.
2138 Description: "lfs find -obd UUID" prints directories
2139 Details : "lfs find -obd UUID" will return all directory names instead
2140 of just file names. It is incorrect because the directories
2141 do not reside on the OSTs.
2145 Description: use i_size_read and i_size_write in 2.6 port
2146 Details : replace inode->i_size access with i_size_read/write()
2149 Frequency : when removing large files
2151 Description: scheduling issue during removal of large Lustre files
2152 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2153 It causes scheduling issues when removing large files (17TB in the
2157 Frequency : only with liblustre clients on XT3
2159 Description: evictions taking too long
2160 Details : allow llrd to evict clients directly on OSTs
2163 Frequency : only on ppc
2165 Description: /proc/fs/lustre/devices broken on ppc
2166 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2167 We should make sure the type of 'index' is loff_t and then cast
2168 explicitly as needed below. Do not assign an explicitly cast
2172 Frequency : only for rhel5
2174 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2175 Details : Modify the target file & which_kernel.
2177 Severity : enhancement
2179 Description: omit set fsid for export NFS
2180 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2181 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2185 Frequency : Always, for kernels after 2.6.16
2186 Description: Fix warning idr_remove called for id=.. which is not allocated.
2187 Details : Last kernels save old s_dev before kill super and not allow
2188 to restore from callback - restore it before call kill_anon_super.
2192 Description: buffer overruns could theoretically occur
2193 Details : llapi_semantic_traverse() modifies the "path" argument by
2194 appending values to the end of the origin string, and a
2195 overrun may occur. Adding buffer overrun check in liblustreapi.
2199 Description: Fix error on 'ls .' at the top of the Lustre mount.
2200 Details : Don't revalidate dentry if it is a root dentry.
2204 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2205 Details : Modify vm-tunables-rhel4.patch.
2209 Description: Kernel config for 2.6.18-vanilla.
2210 Details : Modify targets/2.6-vanilla.target.in.
2211 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2212 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2213 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2214 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2218 Description: improve handling recoverable errors
2219 Details : if request processig with error which can be recoverable on server
2220 request should be resend, otherwise page released from cache and
2225 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2226 Details : Modify target file & which_patch.
2227 A flaw was found in the IA32 system call emulation provided
2228 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2229 value could be stored in the %RAX register, which could trigger an
2230 out-of-bounds system call table access. An untrusted local user
2231 could exploit this flaw to run code in the kernel
2232 (ie a root privilege escalation). (CVE-2007-4573).
2236 Description: O_DIRECT bypasses client statistics.
2237 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2238 read_ahead_stats were updated. copied the stats section from
2239 osc_send_oap_rpc() into async_internal().
2243 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2247 Description: change order of libsysio includes
2248 Details : '#include sysio.h' should always come before '#include xtio.h'
2252 Description: handle "serious error: objid * already exists" more gracefully
2253 Details : If LAST_ID value on disk is smaller than the objects existing in
2254 the O/0/d* directories, it indicates disk corruption and causes an
2255 LBUG(). If the object is 0-length, then we should use the existing
2256 object. This will help to avoid a full fsck in most cases.
2258 Severity : enhancement
2260 Description: adapt the lustre_config script to support the upgrade case
2261 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2262 server targets to 1.6 in parallel.
2266 Description: To avoid grant space > avaible space when the disk is almost
2267 full. Without this patch you might see the error "grant XXXX >
2268 available" or some LBUG about grant, when the disk is almost
2270 Details : In filter_check_grant, for non_grant cache write, we should
2271 check the left space by if (*left > ungranted + bytes), instead
2272 of (*left > ungranted), because only we are sure the left space
2273 is enough for another "bytes", then the ungrant space should be
2274 increase. In client, we should update cl_avail_grant only there
2275 is OBD_MD_FLGRANT in the reply.
2279 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2280 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2281 described in CVE-2007-4573. This problem could allow untrusted
2282 local users to gain root access.
2285 Frequency : when using O_DIRECT and quotas
2287 Description: Incorrect file ownership on O_DIRECT output files
2288 Details : block usage reported by 'lfs quota' does not take into account
2289 files that have been written with O_DIRECT.
2294 Description: touch file failed when fs is not full
2295 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2296 otherwise we can get ENOSP while fs is not full.
2300 Description: parallel lock callbacks
2301 Details : Instead of sending blocking and completion callbacks as separated
2302 requests, adding them to a set and sending in parallel.
2305 Frequency : only for Cray XT3
2306 Bugzilla : 12829/13455
2307 Description: Changing primary group doesn't change the group lustre assigns to
2309 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2310 overridden with the primary group provided by the group upcall,
2311 whereas the supplied fsgid can be trusted if it is in the list of
2312 supplementary groups returned by the group upcall.
2314 Severity : enhancement
2316 Description: Allow masking D_WARNING, D_ERROR messages from console
2317 Details : Console messages can now be disabled via lnet.printk.
2321 Description: User code with malformed file open parameter crashes client node
2322 Details : Before packing join_file req, all the related reference should be
2323 checked carefully in case some malformed flags cause fake
2324 join_file req on client.
2328 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2329 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2330 mode is equal to the granted mode. However, it should also check
2331 that LDLM_FL_CP_REQD is not set.
2335 Description: Heavy nfs access might result in deadlocks
2336 Details : After ELC code landed, it is now improper to enqueue any mds
2337 locks under och_sem, because enqueue might want to decide to
2338 cancel open locks for same inode we are holding och_sem for.
2342 Description: Client eviction while running blogbench
2343 Details : A lot of unlink operations with concurrent I/O can lead to a
2344 deadlock causing evictions. To address the problem, the number of
2345 oustanding OST_DESTROY requests is now throttled to
2346 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2347 callbacks are processed in priority.
2351 Description: enable ACLs on MDS by default
2352 Details : ACLs must be enabled on MDS by default.
2355 Frequency : PPC/PPC64 only
2357 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2358 Details : fix duplicated definitions between asm-ppc64/types.h and
2359 lustre_types.h on PPC.
2362 Frequency : PPC/PPC64 only
2364 Description: asm-ppc/segment.h does not exist
2365 Details : fix compile issue on PPC.
2369 Description: better handle error messages in extents code
2372 Frequency : RHEL4 only
2374 Description: mkfs is very slow on IA64/RHEL4
2375 Details : A performance regression has been discovered in the MPT Fusion
2376 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2377 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2378 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2380 Severity : enhancement
2382 Description: SNMP support enhancement
2383 Details : Adding total number of sampled request for an MDS node in snmp
2386 Severity : enhancement
2388 Description: Optimize ldlm waiting list processing for PR extent locks
2389 Details : When processing waiting list for read extent lock and meeting read
2390 lock that is same or wider to it that is not contended, skip
2391 processing rest of the list and immediatelly return current
2392 status of conflictness, since we are guaranteed there are no
2393 conflicting locks in the rest of the list.
2397 Description: Time out and refuse to reconnect
2398 Details : When the failover node is the primary node, it is possible
2399 to have two identical connections in imp_conn_list. We must
2400 compare not conn's pointers but NIDs, otherwise we can defeat
2401 connection throttling.
2405 Description: port llog fixes from b1_6 into HEAD
2406 Details : Port llog reference couting and some llog cleanups from b1_6
2407 (bug 10800) into HEAD, for protect from panic and access to already
2408 free llog structures.
2412 Description: Detect stride IO mode in read-ahead
2413 Details : When a client does stride read, read-ahead should detect that and
2414 read-ahead pages according to the detected stride pattern.
2418 Description: data checksumming impacts single node performance
2419 Details : add support for several checksum algorithm. Currently, only CRC32
2420 and Adler-32 are supported. The checksum type can be changed on
2421 the fly via /proc/fs/lustre/osc/*/checksum_type.
2425 Description: use adler32 for page checksums
2426 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2431 Description: build for x2 fails
2432 Details : fix compile issue on Cray systems.
2436 Description: Properly match for duplicate locks
2437 Details : Due to different lock order from skiplists code, we need to
2438 traverse entire list for now
2441 Frequency : only on PPC/SLES10
2443 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2444 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2445 be defined. Add a hack in mkfs_lustre.c to work around this compile
2450 Description: LASSERT on MDS when client holding flock lock dies
2451 Details : ldlm pool logic depends on number of granted locks equal to
2452 number of released locks which is not true for flock locks, so
2453 just exclude such locks from consideration.
2457 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2458 Details : Use fsfilt_sync() for both the whole filesystem sync and
2459 individual file sync to eliminate dangerous inode locking
2460 with I_LOCK that can lead to a deadlock.
2464 Description: performance in 1.6.3
2465 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2466 sure that 1MB requests are not fragmented by the block layer.
2468 Severity : enhancement
2470 Description: organize the server-side client stats on per-nid basis
2471 Details : Change the structure of stats under obdfilter and mds to
2479 The "uuid"s file would list the uuids of _active_ exports.
2480 And the clear entry is to clear all stats and stale nids.
2482 Severity : enhancement
2484 Description: eliminate client locks in face of contention
2485 Details : file contention detection and lockless i/o implementation
2486 for contended files.
2490 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2493 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2495 Description: MDS may lose file striping (and hence file data) in some cases
2496 Details : If there are additional extended attributes stored on the MDS,
2497 in particular ACLs, SELinux, or user attributes (if user_xattr
2498 is specified for the client mount options) then there is a risk
2499 of attribute loss. Additionally, the Lustre file striping
2500 needs to be larger than default (e.g. striped over all OSTs),
2501 and an additional attribute must be stored initially in the
2502 inode and then increase in size enough to be moved to the
2503 external attribute block (e.g. ACL growing in size) for file
2508 Description: skiplist implementation simplification
2509 Details : skiplists are used to group compatible locks on granted list
2510 that was implemented as tracking first and last lock of each
2511 lock group the patch changes that to using doubly linked lists
2515 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2516 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2517 because LASSERT on both the data supplied by a client, and the
2518 data on disk is dangerous and incorrect.
2520 Severity : enhancement
2522 Description: Slow truncate/writes to huge files at high offsets.
2523 Details : Directly associate cached pages to lock that protect those pages,
2524 this allows us to quickly find what pages to write and remove
2525 once lock callback is received.
2529 Description: more ldlm soft lockups
2530 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2531 starve other threads from the resource lock for a long time in
2532 case of long waiting queue, so change the debug level from
2533 D_OTHER to the less frequently used D_INFO.
2535 Severity : enhancement
2537 Description: add -gid, -group, -uid, -user options to lfs find
2541 Description: Hung threads in invalidate_inode_pages2_range
2542 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2543 once one is completed. As a result, some RPCs are stuck in the
2544 queue and are never sent.
2548 Description: filter threads hungs on waiting journal commit
2549 Details : Cleanup filter group llog code, then only filter group llog will
2550 be only created in the MDS/OST syncing process.
2554 Description: Procfs and llog threads access destoryed import sometimes.
2555 Details : Sync the import destoryed process with procfs and llog threads by
2556 the import refcount and semaphore.
2558 Severity : enhancement
2560 Description: openlock cache of b1_6 port to HEAD
2565 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2566 Details : If insertion of an extent fails, then discard the inode
2567 preallocation and free data blocks else it can lead to duplicate
2572 Description: don't always update ctime in ext3_xattr_set_handle()
2573 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2574 In some cases the ctime should not be updated, for example for
2575 2.0->1.8 compatibility it is necessary to delete an xattr and it
2576 should not update the ctime.
2580 Bugzilla : 15713/16362
2581 Description: Assertion in iopen_connect_dentry in 1.6.3
2582 Details : looking up an inode via iopen with the wrong generation number can
2583 populate the dcache with a disconneced dentry while the inode
2584 number is in the process of being reallocated. This causes an
2585 assertion failure in iopen since the inode's dentry list contains
2586 both a connected and disconnected dentry.
2590 Description: assertion failure in ldlm_handle2lock()
2591 Details : fix a race between class_handle_unhash() and class_handle2object()
2592 introduced in lustre 1.6.5 by bug 13622.
2597 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2598 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2603 Description: oops in sd_iostats_seq_show()
2604 Details : unloading/reloading the scsi low level driver triggers a kernel
2605 bug when trying to access the sd iostat file.
2610 Description: Kernel panics during QLogic driver reload
2611 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2612 patch, causing memory corruption.
2617 Description: journal_dev option does not work in b1_6
2618 Details : pass mount option during pre-mount.
2620 Severity : enhancement
2622 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2623 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2624 extent information of a file. It can be used to map logical blocks
2625 in a file to physical blocks in the block device.
2629 Description: LDLM soft lockups - improvement
2630 Details : It is be possible to send the lock handle along with each read
2631 or write request because the client is already doing a lock match
2632 itself so there isn't any reason the OST should have to re-do that
2636 Frequency : only on Cray X2
2638 Description: X2 build failures
2639 Details : fix build failures on Cray X2.
2643 Description: xid & resent requests
2644 Details : Initialize RPC XID from clock at startup (randomly if clock is
2647 Severity : enhancement
2649 Description: Add lustre_start utility to start or stop multiple Lustre servers
2654 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2655 Details : In case of memory pressure, list_del() can be called twice on
2656 req->rq_history_list, causing a kernel oops.
2660 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2661 Details : fix stack overflow in the distributed lock manager by defering
2662 export eviction after a failed AST to the elt thread instead of
2663 handling it in the dlm interpret routine.
2667 Description: Convert some comments to new format.
2668 Details : Update documenting comments to match doxygen conventions.
2672 Description: Grammar fixes.
2673 Details : A couple of trivial spelling fixes.
2677 Description: OSD_COUNTERS-mandatory
2678 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2679 to simplify the code.
2683 Description: simplify cmm_device freeing logic.
2684 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2688 Description: Add lockdep support to dt_object_operations locking interface.
2689 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2690 parameter indicating lock ordering. Update mdd code to use new
2695 Description: Introduce failloc constants for lockless IO tests.
2696 Details : Add two new failloc constants to test lockless IO. Only one of
2697 them in implemented---another is checked in yet to be landed
2702 Description: Add lockdep support for inode mutex.
2703 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2704 in the situations where Lustre has to lock more than one inode
2709 Description: Add optional invariants checking support.
2710 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2711 configure switch. This macro is to be used for consistency and
2712 sanity checks that are too expensive to be left in `production'
2717 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2718 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2723 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2724 Details : Use spin_lock_nested() in (the only) situation where more than
2725 one ldlm_lock is locked simultaneously. Also, fix possible
2726 dead-lock in ldlm_lock_change_resource() by enforcing particular
2731 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2732 Details : Instead of specifying each ldlm_lock call-back through separate
2733 parameter, wrap them into struct ldlm_callback_suite.
2737 Description: Kill join_lru obd method and its callers.
2738 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2739 in memory---a requirement implemented through ->o_join_lru() obd
2740 method in HEAD. Remove this method and its users.
2744 Description: Add asynchronous ldlm ENQUEUE completion handler.
2745 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2746 case that stock ldlm_completion_ast() cannot handle as it waits
2747 until lock is granted. Introduce new ldlm_completion_ast_async()
2748 for this. Also comment ldlm_completion_ast().
2752 Description: ldlm_error <-> errno conversion.
2753 Details : Add functions to map (rather arbitrary) between LDLM error codes
2754 and standard errno values. CLIO needs this to prevent LDLM specific
2755 constants from escaping ldlm and osc.
2759 Description: Kill unused ldlm_handle2lock_ns() function.
2760 Details : Kill unused ldlm_handle2lock_ns() function.
2764 Description: Add lu_ref support to ldlm_lock
2765 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2766 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2767 and ldlm_resource. LDLM interface has to be changed, because code that
2768 releases a reference on a lock, has to "know" what reference this is.
2769 In the most frequent case
2771 lock = ldlm_handle2lock(handle);
2773 LDLM_LOCK_PUT(lock);
2775 no changes are required. When any other reference (received _not_ from
2776 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2777 instead of LDLM_LOCK_PUT().
2779 Arguably, changes are pervasive, and interface requires some discipline
2780 for proper use. On the other hand, it was very instrumental in finding
2781 a few leaked lock references.
2785 Description: Add ldlm_lock_addref_try().
2786 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2787 attempts to addref a lock that might be being canceled
2792 Description: Add ldlm_weigh_callback().
2793 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2794 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2795 hard-coded `number of pages' logic.
2799 Description: Add lockdep annotations to llog code.
2800 Details : Use appropriately tagged _nested() locking calls in the places
2801 where llog takes more than one ->lgh_lock lock.
2805 Description: Add loi_kms_set().
2806 Details : Wrap kms updates into a helper function.
2810 Description: Constify instances of struct lsm_operations.
2811 Details : Constify instances of struct lsm_operations.
2815 Description: lu_conf support.
2816 Details : On a server, a file system object is uniquely identified
2817 by a fid, which is sufficient to locate and load all object
2818 state (inode). On a client, on the other hand, more data are
2819 necessary instantiate an object. Change lu_object_find() and
2820 friends to take additional `lu_conf' argument describing object.
2821 Typically this includes layout information.
2825 Description: lu_context fixes.
2826 Details : Introduce new lu_context functions that are needed on the client
2827 side, where some system threads (ptlrpcd) are shared by multiple
2828 modules, and so cannot be stopped during module shutdown.
2832 Description: Add start and stop methods to lu_device_type_operations.
2833 Details : Introduce two new methods in lu_device_type_operations, that are
2834 invoked when first instance of a given type is created and last one
2835 is destroyed respectively. This is need by CLIO.
2839 Description: Add lu_ref support to struct lu_device.
2840 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2841 track leaked references.
2845 Description: Introduce lu_kmem_descr.
2846 Details : lu_kmem_descr and its companion interface allow to create
2847 and destroy a number of kmem caches at once.
2851 Description: Fix lu_object finalization race.
2852 Details : Fix a race between lu_object_find() finding an object and its
2853 concurrent finalization. This race is (most likely) not possible
2854 on the server, but might happen on the client.
2858 Description: Introduce lu_ref interface.
2859 Details : lu_ref is a debugging module allowing to track references to
2860 a given object. It is quite cpu expensive, and has to be
2861 explicitly enabled with --enable-lu_ref. See usage description
2866 Description: Factor lu_site procfs stats into a separate function.
2867 Details : Separate lu_site stats printing code into a separate function
2868 that can be reused on a client.
2872 Description: Constify instances of struct {lu,dt,md}_device_operations.
2873 Details : Constify instances of struct {lu,dt,md}_device_operations.
2877 Description: Introduce struct md_site and move meta-data specific parts of
2878 struct lu_site here.
2879 Details : Move md-specific fields out of struct lu_site into special struct
2880 md_site, so that lu_site can be used on a client.
2884 Description: Kill mdd_lov_destroy().
2885 Details : Remove unused mdd code.
2889 Description: Add st_block checking to multistat.c.
2890 Details : Add st_block checking to multistat.c.
2894 Description: Add lu_ref support to struct obd_device.
2895 Details : Add lu_ref tracking to obd_device.
2899 Description: Kill obd_set_fail_loc().
2900 Details : Remove unused code.
2904 Description: Add special type for ptlrpc_request interpret functions.
2905 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2906 there. Actual usage will be in CLIO.
2910 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2911 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2912 doesn't work with lockdep.
2916 Description: Add rwv.c test program.
2917 Details : New testing program exercising readv(2) and writev(2) (Qian).
2921 Description: Add sendfile.c test program.
2922 Details : New testing program exercising sendfile(2) (Jay).
2926 Description: Ratelimit a message that can be very frequent.
2927 Details : Ratelimit a memory allocation failure message that can
2932 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2933 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2937 Description: Liblustre build fixes.
2938 Details : Liblustre build fixes.
2942 Description: libcfs: add cfs_{need,cond}_resched() interface.
2943 Details : libcfs: add cfs_{need,cond}_resched() definition and
2944 implementations for Linux, NT, and liblustre.
2946 Severity : enhancement
2948 Description: More exported tunables for mballoc
2949 Details : Add support for tunable preallocation window and new tunables for
2950 large/small requests
2954 Description: Detect corruption of block bitmap and checking for preallocations
2955 Details : Checks validity of on-disk block bitmap. Also it does better
2956 checking of number of applied preallocations. When corruption is
2957 found, it turns filesystem readonly to prevent further corruptions.
2961 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2962 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2963 window to 0. If it does hit cache miss, and read-pattern is still
2964 stride-io mode, does not reset the stride window, but also does not
2965 increase the stride window length in this case.
2969 Frequency : only for big-endian servers
2970 Description: Check if system is big-endian while mounting fs with extents feature
2971 Details : Mounting a filesystem with extents feature will fail on big-endian
2972 systems since ext3-based ldiskfs is not supported on big-endian
2973 systems. This can be over-riden with "bigendian_extents" mount option.
2975 Severity : enhancement
2977 Description: The root squash functionality
2978 Details : A security feature, which is to prevent users from being able
2979 to mount lustre on their desktop, run as root, and delete
2980 all of the files in the filesystem. The goal is accomplished by
2981 remapping user id (UID) and group id (GID) of the root user to
2982 a UID and GID specified by the system administartor via Lustre
2983 configuration management server (MGS). The functionality also
2984 allows to specify sets of clients for which the remapping does
2989 Description: Excessive recovery window
2990 Details : With AT enabled, the recovery window can be excessively long (6000+
2991 seconds). To address this problem, we no longer use
2992 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2993 timeout no longer depends on the service time, it is set to
2994 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2995 time via pb_service_time.
2999 Description: Watchdog triggered on MDS failover
3000 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3001 the OSTs know that the MDS "UUID" can be reused for the same export
3002 from a different NID, so we do not need to wait for the export to be
3006 Frequency : rare, only if using MMP with Linux RAID
3008 Description: MMP doesn't work with Linux RAID
3009 Details : While using HA for Lustre servers with Linux RAID, it is possible
3010 that MMP will not detect multiple mounts. To make this work we
3011 need to unplug the device queue in RAID when the MMP block is being
3012 written. Also while reading the MMP block, we should read it from
3013 disk and not the cached one.
3015 Severity : enhancement
3017 Description: open file using fid
3018 Details : A file can be opened using just its fid, like
3019 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3022 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3025 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3026 Details : add proper locking for creation of jbd_4k slab cache
3030 Description: MMP check in ext3_remount() fails without displaying any error
3031 Details : When multiple mount protection fails during remount, proper error
3034 Severity : enhancement
3036 Description: Allow stripe size to be up to 4G-64k
3037 Details : Fix math logic to allow large stripe sizes.
3041 Description: add check for >8TB ldiskfs filesystems
3042 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3043 Don't allow >8TB ldiskfs filesystems to be mounted without
3044 force_over_8tb mount option
3049 Descriptoin: OSS complains "too many blocks" and crashes
3050 Details : In filter_commitrw_write():
3051 1. lquota_chkquota();
3052 2. obdo_from_inode();
3053 3. lquota_pending_commit();
3054 "2." has changed uid and gid in this case, then caused
3055 lquota_chkquota() and lquota_pending_commit() to use
3056 different uid and gid. That is the root of the bug.
3060 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3061 Details : Detect and print out the 'stripe count = -1' case properly.
3065 Description: Out or order replies might be lost on replay
3066 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3067 request with tid smaller than one currently being added, add it
3068 to the start, not end of the list.
3070 --------------------------------------------------------------------------------
3072 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3074 * Support for kernels:
3075 2.6.5-7.283 (SLES 9),
3076 2.6.9-55.EL (RHEL 4),
3077 2.6.16.46-0.14 (SLES 10),
3078 2.6.18.8 vanilla (kernel.org)
3079 * Client support for unpatched kernels:
3080 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3081 2.6.16 - 2.6.22 vanilla (kernel.org)
3082 * Due to recently discovered recovery problems, we do not recommend
3083 using patchless RHEL 4 clients with this or any earlier release.
3084 * Recommended e2fsprogs version: 1.39.cfs8
3085 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3086 * Starting with this release, the ldiskfs backing filesystem required
3087 by Lustre is now in its own package, lustre-ldiskfs. This package
3088 should be installed. It is versioned separately from Lustre and
3089 may be released separately in future.
3093 Description: block reactivating mgc import until all deactivates complete
3094 Details : Fix race when failing back MDT/MGS to itself (testing)
3097 Frequency : at statup only
3099 Description: mds_lov_synchronize race leads to various problems
3100 Details : simultaneous MDT->OST connections at startup can cause the
3101 sync to abort, leaving the OSC in a bad state.
3103 Severity : enhancement
3105 Description: add optional extra BUILD_VERSION info
3106 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3107 to override the lustre version.
3110 Frequency : 2.6.18 servers only
3112 Description: ll_kern_mount() doesn't release the module reference
3113 Details : The ldiskfs module reference count never drops down to 0
3114 because ll_kern_mount() doesn't release the module reference.
3119 Description: server LBUG when using old ost_num_threads parameter
3120 Details : Accept the old ost_num_threads parameter but warn that it
3121 is deprecated, and fix an off-by-one error that caused an LBUG.
3126 Description: Transient SCSI error results in persistent IO issue
3127 Details : iobuf->dr_error is not reinitialized to 0 between two
3131 Frequency : sometimes when underlying device returns I/O errors
3133 Description: OSTs not going read-only during write failures
3134 Details : OSTs are not remounted read-only when the journal commit threads
3135 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3136 instead of the ext3 wrappers.
3140 Description: poor connect scaling with increasing client count
3141 Details : Don't run filter_grant_sanity_check for more than 100 exports
3142 to improve scaling for large numbers of clients.
3145 Frequency : SLES10 only
3147 Description: sanity-quota.sh quotacheck failed: rc = -22
3148 Details : Quotas cannot be enabled on SLES10.
3151 Frequency : liblustre clients only
3153 Description: getdirentries does not give error when run on compute nodes
3154 Details : getdirentries does not fail when the size specified as an argument
3155 is too small to contain at least one entry
3157 Severity : enhancement
3159 Description: Add LNET router traceability for debug purposes
3160 Details : If a checksum failure occurs with a router as part of the
3161 IO path, the NID of the last router that forwarded the bulk data
3162 is printed so it can be identified.
3167 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3168 Details : Due to a race condition, liblustre clients were occasionally
3169 evicted incorrectly.
3171 Severity : enhancement
3173 Description: lfs setstripe use optional parameters instead of postional
3176 Severity : enhancement
3178 Description: Nanosecond timestamp support for ldiskfs
3179 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3180 resolution timestamps. There is not yet support for this at
3181 the Lustre filesystem level.
3184 Frequency : during server recovery
3186 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3187 Details : request with rq_no_resend flag not awake l_wait_event if they get
3191 Frequency : nfs export on patchless client
3193 Description: connectathon hang when test nfs export over patchless client
3194 Details : Disconnected dentry cannot be found with lookup, so we do not need
3195 to unhash it or make it invalid
3198 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3199 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3200 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3205 Description: open req refcounting wrong on reconnect
3206 Details : If reconnect happened between getting open reply from server and
3207 call to mdc_set_replay_data in ll_file_open, we will schedule
3208 replay for unreferenced request that we are about to free.
3209 Subsequent close will crash in variety of ways.
3210 Check that request is still eligible for replay in
3211 mdc_set_replay_data().
3216 Description: disable writes to filesystem when reading health_check file
3217 Details : the default for reading the health_check proc file has changed
3218 to NOT do a journal transaction and write to disk, because this
3219 can cause reads of the /proc file to hang and block HA state
3220 checking on a healthy but otherwise heavily loaded system. It
3221 is possible to return to the previous behaviour during configure
3222 with --enable-health-write.
3224 Severity : enhancement
3226 Description: 64-bit inode version
3227 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3228 to the inode. This will be required for version-based recovery.
3233 Description: MDS fails to start if a duplicate client export is detected
3234 Details : in some rare cases it was possible for a client to connect to
3235 an MDS multiple times. Upon recovery the MDS would detect this
3236 and fail during startup. Handle this more gracefully.
3238 Severity : enhancement
3240 Description: Add -o localflock option to simulate old noflock
3242 Details : This will achieve local-only flock/fcntl locks
3248 Description: log_commit_thread vs filter_destroy race leads to crash
3249 Details : Take import reference before releasing llog record semaphore
3254 Description: Wrong request locking in request set processing
3255 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3256 imp_delayed_list, in this place should be used imp_lock.
3259 Frequency : when reconnection
3261 Description: Grant Leak when osc reconnect to OST
3262 Details : When osc reconnect ost, OST(filter) should check whether it
3263 should grant more space to client by comparing fed_grant and
3264 cl_avail_grant, and return the granted space to client instead
3265 of "new granted" space, because client will call osc_init_grant
3266 to update the client grant space info.
3269 Frequency : when client reconnect to OST
3271 Description: Grant Leak when osc do resend and replay bulk write
3272 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3273 bulk write request, because the grant info while be sync between
3274 OSC and OST when reconnect, and we should ignore the grant info
3275 these of resend/replay write req.
3280 Description: Grant space more than available left space sometimes.
3281 Details : When then OST is about to be full, if two bulk writing from
3282 different clients came to OST. Accord the avaliable space of the
3283 OST, the first req should be permitted, and the second one
3284 should be denied by ENOSPC. But if the seconde arrived before
3285 the first one is committed. The OST might wrongly permit second
3286 writing, which will cause grant space > available space.
3289 Frequency : when client is evicted
3291 Description: Grant might be wrongly erased when osc is evicted by OST
3292 Details : when the import is evicted by server, it will fork another
3293 thread ptlrpc_invalidate_import_thread to invalidate the
3294 import, where the grant will be set to 0. While the original
3295 thread will update the grant it got when connecting. So if
3296 the former happened latter, the grant will be wrongly errased
3297 because of this race.
3302 Description: Checking Stale with correct fid
3303 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3304 to check whether it is stale, because sometimes, we want the
3305 enqueue happened anyway, and op_data.fid2 will not be initialized.
3307 Severity : enhancement
3309 Description: update patchless client
3310 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3313 Frequency : only with 2.4 kernel
3315 Description: random memory corruption
3316 Details : size of struct ll_inode_info is to big for union inode.u and this
3317 can be cause of random memory corruption.
3322 Description: Memory leak in recovery
3323 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3324 It should also check obd_fail before fsfilt_start, otherwise if
3325 fsfilt_start return -EROFS,(failover mds during mds recovery).
3326 then the req will return with repmsg->transno = 0 and rc = EROFS.
3327 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3328 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3329 be freed no matter whether fsfilt_commit success or not.
3332 Frequency : only with huge count clients
3334 Description: Prevents from taking the superblock lock in llap_from_page for
3336 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3337 for llap_from_page call in ll_removepage prevents from taking the
3338 superblock lock for a soon died page.
3343 Description: Not check open intent error before release open handle
3344 Details : in some rare cases, the open intent error is not checked before
3345 release open handle, which may cause
3346 ASSERTION(open_req->rq_transno != 0), because it tries to release
3347 the failed open handle.
3352 Description: Set cat log bitmap only after create log success.
3353 Details : in some rare cases, the cat log bitmap is set too early. and it
3354 should be set only after create log success.
3358 Description: Accessing a block bevice can re-enable I/O when Lustre is
3359 tearing down a device.
3360 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3365 Description: service threads may hog cpus when there are a lot of requests
3366 Details : Insert cond_resched to give other threads a chance to use some CPU
3371 Description: the cat log was not initialized in recovery
3372 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3373 unlink log on mds will not be initialized until mds post
3374 recovery. And also in mds post recovery, the unlink log will
3375 initialization will be done asynchronausly, so there will be race
3376 between add unlink log and unlink log initialization.
3380 Description: brw_stats were being printed incorrectly
3381 Details : brw_stats were being printed as log2 but all of them were not
3382 recorded as log2. Also remove some code duplication arising from
3383 filter_tally_{read,write}.
3387 Frequency : rare, only in recovery.
3388 Description: ASSERTION(req->rq_type != LI_POISON) failed
3389 Details : imp_lock should be held while iterating over imp_sending_list for
3390 prevent destroy request after get timeout in ptlrpc_queue_wait.
3394 Description: replay-single.sh test 52 fails
3395 Details : A lock's skiplist need to be cleanup when it being unlinked
3396 from its resource list.
3400 Description: Short directio read returns full requested size rather than
3402 Details : Direct I/O operations should return actual amount of bytes
3403 transferred rather than requested size.
3405 Severity : enhancement
3407 Description: metadata RPC reduction (e.g. for rm performance)
3408 Details : decrease the amount of synchronous RPC between clients and servers
3409 by canceling conflicing lock before the operation on the client
3410 and packing thier handles into the main operation RPC to server.
3412 Severity : enhancement
3414 Description: Async OSC create to avoid the blocking unnecessarily.
3415 Details : If an OST has no remaining object, system will block on the
3416 creation when it needs to create a new object on this OST. Now,
3417 ways use pre-created objects when available, instead of blocking on
3418 an empty osc while others are not empty. If we must block, we
3419 block for the shortest possible period of time.
3423 Description: improve handling recoverable errors
3424 Details : if request processig with error which can be recoverable on server
3425 request should be resend, otherwise page released from cache and
3428 Severity : enhancement
3430 Description: refine locking for avoid write wrong info into lov_objid file
3431 Details : fix possible races with add new target and write/update data in
3434 --------------------------------------------------------------------------------
3436 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3441 Frequency : on some architectures
3443 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3444 Details : Uninitialized flags sometimes cause configuration commands to
3448 Frequency : patchless clients only
3450 Description: missing __iget() symbol export
3451 Details : The __iget() symbol export is missing. To avoid the need for
3452 this on patchless clients the deathrow inode reaper is turned
3453 off, and we depend on the VM to clean up old inodes. This
3454 dependency was during via the fix for bug 12181.
3456 --------------------------------------------------------------------------------
3458 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3460 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3461 INTEROPERATE with older versions automatically. In many cases a
3462 special upgrade step is needed. Please read the
3463 user documentation before upgrading any part of a live system.
3464 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3465 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3466 * WARNING: Lustre configuration and startup changes are required with
3467 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3471 Severity : enhancement
3473 Description: MountConf
3474 Details : Lustre configuration is now managed via mkfs and mount
3475 commands instead of lmc and lconf. New obd types (MGS, MGC)
3476 are added for dynamic configuration management. See
3477 https://mail.clusterfs.com/wikis/lustre/MountConf for
3480 Severity : enhancement
3482 Description: dynamic OST addition
3483 Details : OSTs can now be added to a live filesystem
3485 Severity : enhancement
3487 Description: startup order invariance
3488 Details : MDTs and OSTs can be started in any order. Clients only
3489 require the MDT to complete startup.
3491 Severity : enhancement
3493 Description: parallel, asynchronous orphan cleanup
3494 Details : orphan cleanup is now performed in separate threads for each
3495 OST, allowing parallel non-blocking operation.
3497 Severity : enhancement
3499 Description: optimized stripe assignment
3500 Details : stripe assignments are now made based on ost space available,
3501 ost previous usage, and OSS previous usage, in order to try
3502 to optimize storage space and networking resources.
3504 Severity : enhancement
3506 Description: Permanently set tunables
3507 Details : All writable /proc/fs/lustre tunables can now be permanently
3508 set on a per-server basis, at mkfs time or on a live system.
3510 Severity : enhancement
3512 Description: Lustre message v2
3513 Details : Add lustre message format v2.
3515 Severity : enhancement
3517 Description: client OST exclusion list
3518 Details : Clients can be started with a list of OSTs that should be
3519 declared "inactive" for known non-responsive OSTs.
3523 Description: SPEC SFS validation failure on NFS v2 over lustre.
3524 Details : Changes the blocksize for regular files to be 2x RPC size,
3525 and not depend on stripe size.
3527 Severity : enhancement
3529 Description: Multiple MD RPCs in flight.
3530 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3531 To avoid overly-overloading MDS, introduce a limit on number of
3532 MDS RPCs in flight for a single client and add /proc controls
3533 to adjust this limit.
3535 Severity : enhancement
3537 Description: client read/write statistics
3538 Details : Add client read/write call usage stats for performance
3539 analysis of user processes.
3540 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3541 file access. extents_stats shows chunk size distribution.
3542 extents_stats_per_process show chunk size distribution per
3545 Severity : enhancement
3547 Description: mds statistics
3548 Details : Add detailed mds operations statistics in
3549 /proc/fs/lustre/mds/*/stats.
3553 Description: Failure of copying files with lustre special EAs.
3554 Details : Client side always return success for setxattr call for lustre
3555 special xattr (currently only "trusted.lov").
3560 Description: Refcount LNET uuids
3561 Details : The global LNET uuid list grew linearly with every startup;
3562 refcount repeated list entries instead of always adding to
3565 Severity : enhancement
3567 Description: Dynamic service threads
3568 Details : Within a small range, start extra service threads
3569 automatically when the request queue builds up.
3572 Frequency : mixed-endian client/server environments
3574 Description: mixed-endian crashes
3575 Details : The new msg_v2 system had some failures in mixed-endian
3578 Severity : enhancement
3580 Description: Easy OST removal
3581 Details : OSTs can be permanently deactivated with e.g. 'lctl
3582 conf_param lustre-OST0001.osc.active=0'
3584 Severity : enhancement
3586 Description: MGS proc entries
3587 Details : Added basic proc entries for the MGS showing what filesystems
3590 Severity : enhancement
3592 Description: provide MGS failover
3593 Details : Added config lock reacquisition after MGS server failover.
3595 Severity : enhancement
3597 Description: add Linux 2.4 support
3598 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3602 Description: a large application tries to do I/O to the same resource and dies
3603 in the middle of it.
3604 Details : Check the req->rq_arrival time after the call to
3605 ost_brw_lock_get(), but before we do anything about
3606 processing it & sending the BULK transfer request. This
3607 should help move old stale pending locks off the queue as
3608 quickly as obd_timeout.
3611 Frequency : when an incorrect nid is specified during startup
3613 Description: ptlrpc connect to non-existant node causes kernel crash
3614 Details : LNET can't be re-entered from an event callback, which
3615 happened when we expire a message after the export has been
3616 cleaned up. Instead, hand the zombie cleanup off to another
3619 Severity : enhancement
3621 Description: plain/inodebits lock performance improvement
3622 Details : Grouping plain/inodebits in granted list by their request modes
3623 and bits policy, thus improving the performance of search through
3627 Frequency : only if OST filesystem is corrupted
3629 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3630 Details : for a short time RPCs with bulk IO are in the replay list,
3631 but replay of bulk IOs is unimplemented. If the OST filesystem
3632 is corrupted due to disk cache incoherency and then replay is
3633 started it is possible to trip an assertion. Avoid putting
3634 committed RPCs into the replay list at all to avoid this issue.
3637 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3639 Description: System hang on startup
3640 Details : This bug allowed the liblustre (e.g. catamount) client to
3641 return to the app before handling all startup RPCs. This
3642 could leave the node unresponsive to lustre network traffic
3643 and manifested as a server ptllnd timeout.
3645 Severity : enhancement
3647 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3648 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3649 boolean to control whether to print peer debug info when a
3650 client's RPC times out.
3653 Frequency : only for kernels with patches from Lustre below 1.4.3
3655 Description: Remove old rdonly API
3656 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3659 Frequency : only for devices with external journals
3661 Description: Set external device read-only also
3662 Details : During a commanded failover stop, we set the disk device
3663 read-only while the server shuts down. We now also set any
3664 external journal device read-only at the same time.
3667 Frequency : when upgrading from 1.4 while trying to change parameters
3669 Description: The wrong (new) MDC name was used when setting parameters for
3670 upgraded MDT's. Also allows changing of OSC (and MDC)
3671 parameters if --writeconf is specified at tunefs upgrade time.
3674 Frequency : when setting specific ost indicies
3676 Description: QOS code breaks on skipped indicies
3677 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3678 created with --index need not be sequential.
3680 Severity : enhancement
3682 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3683 Details : The uninit_groups feature works in conjunction with the kernel
3684 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3685 up the pass1 processing of e2fsck. This is a read-only feature
3686 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3687 mount filesystems that have had this feature enabled.
3689 Severity : enhancement
3691 Description: Improve multi-block allocation algorithm to avoid fragmentation
3692 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3693 allocation locality and avoid filesystem fragmentation.
3695 ------------------------------------------------------------------------------
3697 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3699 * Support for kernels:
3700 2.4.21-47.0.1.EL (RHEL 3)
3701 2.6.5-7.283 (SLES 9)
3702 2.6.9-42.0.10.EL (RHEL 4)
3703 2.6.12.6 vanilla (kernel.org)
3704 2.6.16.27-0.9 (SLES 10)
3705 * Recommended e2fsprogs version: 1.39.cfs5
3707 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3711 Frequency : occasional, depends on client load and configuration
3712 Bugzilla : 12181, 12203
3713 Description: data loss for recently-modified files
3715 Details : In some cases it is possible that recently written or created
3716 files may not be written to disk in a timely manner (this should
3717 normally be within 30s unless client IO load is very high).
3718 The problem appears as zero-length files or files that are a
3719 multiple of 1MB in size after a client crash or client eviction
3720 that are missing data at the end of the file.
3722 This problem is more likely to be hit on clients where files are
3723 repeatedly created and unlinked in the same directory, clients
3724 have a large amount of RAM, have many CPUs, the filesystem has
3725 many OSTs, the clients are rebooted frequently, and/or the files
3726 are not accessed by other nodes after being written.
3728 The presence of the problem can be detected by looking at
3729 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3730 smaller than the second (nr_unused) then dirty files will not
3731 be flushed automatically to disk. "sync; sleep 10" should be
3732 run several times on the node before unmounting it to update
3733 Lustre (this is also safe to run on nodes without this problem).
3735 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3736 that can cause this same problem, so customers using that kernel
3737 also need to update the kernel in addition to Lustre. In order
3738 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3740 It is normal that files written just before a client crash (less
3741 than 30s) may not yet have been flushed to disk, even for local
3745 Frequency : frequent on thin XT3 nodes
3747 Description: UUID collision on thin XT3 Linux nodes
3748 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3749 randomly, since we relied on an insufficiently-seeded PRNG.
3754 Description: OSS hangs after "All ost request buffers busy"
3755 Details : A deadlock between quota and journal operations caused OSS
3756 hangs after printing "All ost request buffers busy."
3759 Frequency : always on liblustre builds
3761 Description: Cleanup compiler warnings on liblustre
3764 Frequency : always on liblustre builds on XT3
3766 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3770 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3771 > 32000 subdirectories
3772 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3773 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3774 subdirectory count crosses 32000. This will aid e2fsck to
3775 correctly handle more than 32000 subdirectories.
3778 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3780 Description: System hang on startup
3781 Details : This bug allowed the liblustre (e.g. catamount) client to
3782 return to the app before handling all startup RPCs. This
3783 could leave the node unresponsive to lustre network traffic
3784 and manifested as a server ptllnd timeout.
3786 Severity : enhancement
3788 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3789 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3790 boolean to control whether to print peer debug info when a
3791 client's RPC times out.
3796 Description: make O_SYNC working on 2.6 kernels
3797 Details : 2.6 kernels use different method for mark pages for write,
3798 so need add a code to lustre for O_SYNC work.
3803 Description: Failure to close file and release space on NFS
3804 Details : Put inode details into lock acquired in ll_intent_file_open.
3805 Use mdc_intent_lock in ll_intent_open to properly
3806 detect all kind of errors unhandled by mdc_enqueue.
3811 Description: proc file read during shutdown sometimes raced obd removal,
3813 Details : Add lock to prevent obd access after proc file removal.
3816 Frequency : Only for files larger than 4GB on 32-bit clients.
3818 Description: improperly doing page alignment of locks
3819 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3820 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3825 Description: RPCs being resent when they shouldn't be
3826 Details : Some RPCs that should not be resent are being resent. This
3827 can cause inconsistencies in the RPC state machine. Do not
3828 resend such requests.
3831 Frequency : rare, only with NFS export
3833 Description: Crash on NFS re-export node
3834 Details : under very unusual load conditions an assertion is hit in
3835 ll_intent_file_open()
3838 Frequency : only if OST filesystem is corrupted
3840 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3841 Details : for a short time RPCs with bulk IO are in the replay list,
3842 but replay of bulk IOs is unimplemented. If the OST filesystem
3843 is corrupted due to disk cache incoherency and then replay is
3844 started it is possible to trip an assertion. Avoid putting
3845 committed RPCs into the replay list at all to avoid this issue.
3850 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3851 Details : Large single O_DIRECT read and write calls can fail to allocate
3852 a sufficiently large buffer to process the request. In case of
3853 allocation failure the allocation is retried with a smaller
3854 buffer and broken into smaller requests.
3856 Severity : enhancement
3858 Description: Add -o localflock option to simulate old noflock behaviour.
3859 Details : This will achieve local-only flock/fcntl locks coherentness.
3864 Description: versioning check is incomplete
3865 Details : Checking the version difference of client vs. server, report
3866 error if the gap is too big.
3871 Description: add support PG_writeback bit
3872 Details : add support for PG_writeback bit for Lustre, for more carefull
3873 work with page cache in 2.6 kernel. This also fix some deadlocks
3874 and remove hack for work O_SYNC with 2.6 kernel.
3876 Severity : enhancement
3878 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3879 Details : The uninit_groups feature works in conjunction with the kernel
3880 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3881 up the pass1 processing of e2fsck. This is a read-only feature
3882 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3883 mount filesystems that have had this feature enabled.
3885 Severity : enhancement
3887 Description: Improve multi-block allocation algorithm to avoid fragmentation
3888 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3889 allocation locality and avoid filesystem fragmentation.
3891 ------------------------------------------------------------------------------
3893 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3895 * Support for kernels:
3896 2.6.9-42.0.3.EL (RHEL 4)
3897 2.6.5-7.276 (SLES 9)
3898 2.4.21-47.0.1.EL (RHEL 3)
3899 2.6.12.6 vanilla (kernel.org)
3900 2.6.16.21-0.8 (SLES10)
3901 * Recommended e2fsprogs version: 1.39.cfs2-0
3903 * The backwards-compatible /proc/sys/portals symlink has been removed
3904 in this release. Before upgrading, please ensure that you change
3905 any configuration scripts or /etc/sysctl.conf files that access
3906 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3907 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3908 in advance of the upgrade on any system running Lustre 1.4.6 or
3909 newer, since /proc/sys/lnet was added in that version.
3910 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3914 Frequency : only when quota is used
3916 Description: avoid scanning export list for quota master
3917 Details : Change the algorithms to avoid scanning export list in order
3918 to improve the efficiency.
3921 Frequency : MDS failover only, very rarely
3923 Description: "went back in time" messages on mds failover
3924 Details : The greatest transno may be lost when the current operation
3925 finishes with an error (transno==0) and the client's last_rcvd
3926 record is over-written. Save the greatest transno in the
3927 mds_last_transno for this case.
3930 Frequency : always for specific kernels and striping counts
3932 Description: client may get "Matching packet too big" without ACL support
3933 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3934 when trying to access files in certain configurations. The
3935 clients should in fact be denied when mounting because they do
3936 not understand ACLs.
3939 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3941 Description: many clients connecting with IO in progress causes connect timeouts
3942 Details : Avoid synchronous journal commits to avoid delays caused by many
3943 clients connecting/disconnecting when bulk IO is in progress.
3944 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3945 OST_IO_PORTAL to avoid delays behind potentially many pending
3949 Frequency : occasionally with multiple writers to a single file
3951 Description: shared writes to file may result in wrong size reported by stat()
3952 Details : Allow growing of kms when extent lock is cancelled
3955 Frequency : always with random mmap IO to multi-striped file
3957 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3958 Details : Only if the hole is at the end of OST object so that kms is too
3959 small. Fix is to increase kms accordingly in ll_nopage.
3962 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3964 Description: writes to a missing object would leak memory on the OST
3965 Details : If there is an inconsistency between the MDS and OST filesystems,
3966 such that the MDS references an object that doesn't exist, writes
3967 to that object will leak memory due to incorrect cleanup in the
3968 error handling path, eventually running out of memory on the OST.
3973 Description: Creating too long symlink causes lustre errors
3974 Details : Check symlink and name lengths before sending requests to MDS.
3977 Frequency : only if flock is enabled (not on by default)
3979 Description: posix locks not released on fd closure on 2.6.9+
3980 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3981 kernels, this caused such locks not to be released on fd close and
3982 then assertions on fs unmount about still used locks.
3985 Frequency : MDS failover only, very rarely
3987 Description: clients may get ASSERTION(granted_lock != NULL)
3988 Details : When request was taking a long time, and a client was resending
3989 a getattr by name lock request. The were multiple lock requests
3990 with the same client lock handle and
3991 mds_getattr_name->fixup_handle_for_resent_request found one of the
3992 lock handles but later failed with ASSERTION(granted_lock != NULL).
3997 Description: handle->h_buffer_credits > 0, assertion failure
3998 Details : h_buffer_credits is zero after truncate, causing assertion
3999 failure. This patch extends the transaction or creates a new
4003 Frequency : NFS re-export or patchless client
4004 Bugzilla : 11179, 10796
4005 Description: Crash on NFS re-export node (__d_move)
4006 Details : We do not want to hash the dentry if we don't have a lock.
4007 But if this dentry is later used in d_move, we'd hit uninitialised
4008 list head d_hash, so we just do this to init d_hash field but
4009 leave dentry unhashed.
4012 Frequency : NFS re-export or patchless client
4014 Description: NFS exports has problem with symbolic link
4015 Details : lustre client didn't properly install dentry when re-exported
4016 to NFS or running patchless client.
4019 Frequency : NFS re-export or patchless client
4021 Description: Various nfs/patchless fixes.
4022 Details : fixes reuse disconected alias for lookup process - this fixes
4023 warning "find_exported_dentry: npd != pd",
4024 fix permission error with open files at nfs.
4025 fix apply umask when do revalidate.
4028 Frequency : occasional
4030 Description: Crash on NFS re-export node
4031 Details : calling clear_page() on the wrong pointer triggered oops in
4032 generic_mapping_read().
4035 Frequency : rarely, using O_DIRECT IO
4037 Description: unaligned directio crashes client with LASSERT
4038 Details : check for unaligned buffers before trying any requests.
4041 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4043 Description: stale data returned from RAID cache
4044 Details : If only a small amount of IO is done to the RAID device before
4045 reading it again it is possible to get stale data from the RAID
4046 cache instead of reading it from disk.
4049 Frequency : always for sles10 kernel
4051 Description: sles10 support
4052 Details : ll_follow_link: compile fixes and using of nd_set_link
4053 under newer kernels.
4056 Frequency : depends on arch, kernel and compiler version, always on sles10
4059 Description: recursive or deep enough symlinks cause stack overflow
4060 Details : getting rid of large stack-allocated variable in
4064 Frequency : depends on hardware
4066 Description: lustre write performance loss in the SLES10 kernel
4067 Details : the performance loss is caused by using of write barriers in the
4068 ext3 code. The SLES10 kernel turns barrier support on by
4069 default. The fix is to undo that change for ldiskfs.
4071 ------------------------------------------------------------------------------
4073 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4075 * Support for kernels:
4076 2.6.9-42.0.3EL (RHEL 4)
4077 2.6.5-7.276 (SLES 9)
4078 2.4.21-47.0.1.EL (RHEL 3)
4079 2.6.12.6 vanilla (kernel.org)
4083 Frequency : quota enabled and large files being deleted
4085 Description: releasing more than 4GB of quota at once hangs OST
4086 Details : If a user deletes more than 4GB of files on a single OST it
4087 will cause the OST to spin in an infinite loop. Release
4088 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4093 Description: statfs data retrieved from /proc may be stale or zero
4094 Details : When reading per-device statfs data from /proc, in the
4095 {kbytes,files}_{total,free,avail} files, it may appear
4096 as zero or be out of date.
4099 Frequency : systems with MD RAID1 external journal devices
4101 Description: lconf's call to blkid is confused by RAID1 journal devices
4102 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4103 of returning all block devices that match the journal UUID.
4106 Frequency : always, for aggregate stripe size over 4GB
4108 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4109 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4110 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4113 Frequency : always if "lfs find" used on a local file/directory
4115 Description: "lfs find" segfaults if used on a local file/directory
4116 Details : The case where a directory component was not specified wasn't
4117 handled correctly. Handle this properly.
4120 Frequency : always on ppc64
4122 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4123 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4124 perform endianness conversion.
4127 Frequency : rarely (truncate to non-zero file size after write under load)
4128 Bugzilla : 10730, 10687
4129 Description: Files padded with zeros to next 4K multiple
4130 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4131 it is possible that files that are truncated to a non-zero size
4132 immediately after being written are filled with zero bytes beyond
4133 the truncated size. No file data is lost.
4135 Severity : enhancement
4137 Description: Allow recovery/failover for liblustre clients.
4138 Details : liblustre clients were unaware of failover configurations until
4141 Severity : enhancement
4143 Description: user file locks should fail when not mounting with flock option
4144 Details : Set up an error-returning stub in ll_file_operations.lock field
4145 to prevent incorrect behaviour when client is mounted without
4146 flock option. Also, set up properly f_op->flock field for
4150 Frequency : always on ia64
4152 Description: "lfs df" loops on printing out MDS statfs information
4153 Details : The obd_ioctl_data was not initialized and in some systems
4154 this caused a failure during the ioctl that did not return
4155 an error. Initialize the struct and return an error on failure.
4158 Frequency : SLES 9 only
4160 Description: Error of copying files with lustre special EAs as root
4161 Details : Client side always return success for setxattr call for lustre
4162 special xattr (currently only "trusted.lov").
4165 Frequency : rarely on clusters with both ia64+i386 clients
4167 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4168 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4169 clients are doing concurrent writes to the same file it is
4170 possible that the ia64 clients may LASSERT because the OST
4171 extent locks are not PAGE_SIZE aligned. Ensure that grown
4172 locks are always aligned on the request boundary.
4175 Frequency : specific use, occasional
4177 Description: Overwriting in use executable truncates on-disk binary image
4178 Details : If one node attempts to overwrite an executable in use by
4179 another node, we now correctly return ETXTBSY instead of
4180 truncating the file.
4182 Severity : enhancement
4184 Description: Async OSC create to avoid the blocking unnecessarily.
4185 Details : If an OST has no remaining object, system will block on the creating
4186 when need to create a new object on this OST. Now, ways use
4187 pre-created objects when available, instead of blocking on an
4188 empty osc while others are not empty. If we must block, we block
4189 for the shortest possible period of time.
4194 Description: chmod on Lustre root is propagated to other clients
4195 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4196 invalid by the follow_mount time.
4199 Frequency : liblustre clients only
4201 Description: Race in 'instant cancel' lock handling could lead to such locks
4202 never to be granted in case of SMP MDS
4203 Details : Do not destroy not yet granted but cbpending locks in
4207 Frequency : replay/resend of open
4209 Description: non null lock assetion failure in mds_intent_policy
4210 Details : Trying to replay/resend lockless open requests resulted in
4211 mds_open() returning 0 with no lock. Now it sets a flag if
4212 a lock is going to be returned.
4214 Severity : enhancement
4216 Description: Checksum enhancements
4217 Details : New checksum enhancements allow for resending RPCs that failed
4220 Severity : enhancement
4222 Description: Tunables on number of dirty pages in cacche
4223 Details : Allow to set limit on number of dirty pages cached.
4228 Description: client crash on unmount - lock still has references
4229 Details : In some error handling cases it was possible to leak a lock
4230 reference on a client while accessing a file. This was not
4231 harmful to the client during operation, but would cause the
4232 client to crash when the filesystem is unmounted.
4235 Frequency : specific case, rare
4237 Description: ETXTBSY on mds though file not in use by client
4238 Details : ETXTBSY is no longer incorrectly returned when attempting to
4239 chmod or chown a directory that the user previously tried to
4240 execute or a currently-executing binary.
4243 Frequency : extremely rare except on liblustre-based clients
4245 Description: Lustre space not freed when files are deleted
4246 Details : Clean up open-unlinked files after client eviction. Previously
4247 the unlink was skipped and the files remained as orphans.
4252 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4253 Details : In some cases it was possible to send a blocking callback to
4254 a client doing a glimpse, even though that client didn't get
4255 a lock granted. When the glimpse lock is cancelled on the OST
4256 the freed lock is left on the waiting list and corrupted the list.
4259 Frequency : all core dumps
4261 Description: Broke core dumps to lustre
4262 Details : Negative dentry may be unhashed if parent does not have UPDATE
4263 lock, but some callers, e.g. do_coredump, expect dentry to be
4264 hashed after successful create, hash it in ll_create_it.
4266 ------------------------------------------------------------------------------
4268 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4270 * Support for kernels:
4271 2.6.9-42.0.2.EL (RHEL 4)
4272 2.6.5-7.276 (SLES 9)
4273 2.4.21-40.EL (RHEL 3)
4274 2.6.12.6 vanilla (kernel.org)
4278 Frequency : always on RHEL 3
4280 Description: Number of open files grows over time
4281 Details : The number of open files grows over time, whether or not
4282 Lustre is started. This was due to a filp leak introduced
4283 by one of our kernel patches.
4285 ------------------------------------------------------------------------------
4287 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4289 * Support for kernels:
4290 2.6.9-42.EL (RHEL 4)
4291 2.6.5-7.276 (SLES 9)
4292 2.4.21-40.EL (RHEL 3)
4293 2.6.12.6 vanilla (kernel.org)
4298 Bugzilla : 5719, 9635, 9792, 9684
4299 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4300 Details : If a server is under heavy load and cannot reply to new
4301 connection requests before the client resends the (re)connect,
4302 the connection handling code can behave badly if two service
4303 threads are concurrently handing separate (re)connections from
4304 the same client. Add better locking to the connection handling
4305 code, and ensure that only a single connection will be processed
4306 for a given client UUID, even if the lock is dropped.
4308 Severity : enhancement
4310 Description: add TCP zero-copy support to kernel
4311 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4312 sends if the hardware supports scatter-gather and checksumming.
4313 This allows socklnd to do client-write and server-read more
4314 efficiently and reduce CPU utilization from skbuf copying.
4317 Frequency : only if NFS exporting from client
4319 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4320 Details : If setting attributes on a file created under NFS that had
4321 never been opened it would be possible to oops the client
4322 if the file had no objects.
4325 Frequency : always for liblustre
4327 Description: liblustre client does MDS+OSTs setattr RPC for each write
4328 Details : When doing a write from a liblustre client, the client
4329 incorrectly issued an RPC to the MDS and each OST the file was
4330 striped over in order to update the timestamps. When writing
4331 with small chunks and many clients this could overwhelm the MDS
4332 with RPCs. In all cases it would slow down the write because
4333 these RPCs are unnecessary.
4335 Severity : enhancement
4337 Description: allow number of MDS service threads to be changed at module load
4338 Details : It is now possible to change the number of MDS service threads
4339 running. Adding "options mds mds_num_threads={N}" to the MDS's
4340 /etc/modprobe.conf will set the number of threads for the next
4341 time Lustre is restarted (assuming the "mds" module is also
4342 reloaded at that time). The default number of threads will
4343 stay the same, 32 for most systems.
4348 Description: OST crash if filesystem is unformatted or corrupt
4349 Details : If an OST is started on a device that has never been formatted
4350 or if the filesystem is corrupt and cannot even mount then the
4351 error handling cleanup routines would dereference a NULL pointer.
4356 Description: NULL pointer deref in llap_from_page.
4357 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4358 incorrect) mapping if the page was truncated/reclaimed while it was
4359 searched for. Check for this condition and skip such pages when
4360 doing readahead. Introduce extra check to llap_from_page() to
4361 verify page->mapping->host is non-NULL (so page is not anonymous).
4364 Frequency : Sometimes when using sys_sendfile
4366 Description: "page not covered by a lock" warnings from ll_readpage
4367 Details : sendfile called ll_readpage without right page locks present.
4368 Now we introduced ll_file_sendfile that does necessary locking
4369 around call to generic_file_sendfile() much like we do in
4373 Frequency : with certain MDS communication failures at client mount time
4375 Description: NULL pointer deref after failed client mount
4376 Details : a client connection request may delayed by the network layer
4377 and not be sent until after the PTLRPC layer has timed out the
4378 request. If the client fails the mount immediately it will try
4379 to clean up before the network times out the request. Add a
4380 reference from the request import to the obd device and delay
4381 the cleanup until the network drops the request.
4384 Frequency : occasionally during client (re)connect
4386 Description: assertion failure during client (re)connect
4387 Details : processing a client connection request may be delayed by the
4388 client or server longer than the client connect timeout. This
4389 causes the client to resend the connection request. If the
4390 original connection request is replied in this interval, the
4391 client may trip an assertion failure in ptlrpc_connect_interpret()
4392 which thought it would be the only running connect process.
4395 Frequency : only with obd_echo servers and clients that are rebooted
4397 Description: kernel BUG accessing uninitialized data structure
4398 Details : When running an obd_echo server it did not start the ping_evictor
4399 thread, and when a client was evicted an uninitialized data
4400 structure was accessed. Start the ping_evictor in the RPC
4401 service startup instead of the OBD startup.
4403 Severity : enhancement
4404 Bugzilla : 10193 (patchless)
4405 Description: Remove dependency on various unexported kernel interfaces.
4406 Details : No longer need reparent_to_init, exit_mm, exit_files,
4407 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4410 Frequency : rare (only users of deprecated and unsupported LDAP config)
4412 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4413 Details : LDAP apparently contains 'lustreName' attributes instead of
4414 'name'. A simple remapping of the name is sufficient.
4417 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4419 Description: waiting_locks_callback trips kernel BUG if client is evicted
4420 Details : Running with the dump_on_timeout debug flag turned on makes
4421 it possible that the waiting_locks_callback() can try to dump
4422 the Lustre kernel debug logs from an interrupt handler. Defer
4423 this log dumping to the expired_lock_main() thread.
4425 Severity : enhancement
4427 Description: Support NFS exporting on 2.6 kernels.
4428 Details : Implement non-rawops metadata methods for NFS server to use without
4429 changing NFS server code.
4432 Frequency : very rare (synthetic metadata workload only)
4434 Description: two racing renames might cause an MDS thread to deadlock
4435 Details : Running the "racer" program may cause one MDS thread to rename
4436 a file from being the source of a rename to being the target of
4437 a rename at exactly the same time that another thread is doing
4438 so, and the second thread has already enqueued these locks after
4439 doing a lookup of the target and is trying to relock them in
4440 order. Ensure that we don't try to re-lock the same resource.
4443 Frequency : only very large systems with liblustre clients
4445 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4446 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4447 all OSTs in parallel. This allows the eviction of stale liblustre
4448 clients to proceed much faster than if they were done in series,
4449 and also offers similar improvements for other set_info RPCs.
4454 Description: excessive CPU usage during initial read phase on client
4455 Details : During the initial read phase on a client, it would agressively
4456 retry readahead on the file, consuming too much CPU and impacting
4457 performance (since 1.4.5.8). Improve the readahead algorithm
4458 to avoid this, and also improve some other common cases (read
4459 of small files in particular, where "small" is files smaller than
4460 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4465 Description: MDS crash when receiving packet with unknown intent.
4466 Details : Do not LBUG in unknown intent case, just return -EFAULT
4468 Severity : enhancement
4469 Bugzilla : 9293, 9385
4470 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4471 Details : Do not serialize getattr (non-intent version) and statfs.
4474 Frequency : occasional, when OST network is overloaded/intermittent
4476 Description: client evicted by OST after bulk IO timeout
4477 Details : If a client sends a bulk IO request (read or write) the OST
4478 may evict the client if it is unresposive to its data GET/PUT
4479 request. This is incorrect if the network is overloaded (takes
4480 too long to transfer the RPC data) or dropped the OST GET/PUT
4481 request. There is no need to evict the client at all, since
4482 the pinger and/or lock callbacks will handle this, and the
4483 client can restart the bulk request.
4486 Frequency : Always when mmapping file with no objects
4488 Description: client crashes when mmapping file with no objects
4489 Details : Check that we actually have objects in a file before doing any
4490 operations on objects in ll_vm_open, ll_vm_close and
4496 Description: Request leak when working with deleted CWD
4497 Details : Introduce advanced request refcount tracking for requests
4498 referenced from lustre intent.
4500 Severity : Enhancement
4502 Description: Cache open file handles on client.
4503 Details : MDS now will return special lock along with openhandle, if
4504 requested and client is allowed to hold openhandle, even if unused,
4505 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4506 closing files for every read/write openration.
4508 Severity : Enhancement
4510 Description: Cache open negative dentries on client when possible.
4511 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4512 negative dentries on lock revocation.
4517 Description: Remounting a client read-only wasn't possible with a zconf mount
4518 Details : It wasn't possible to remount a client read-only with llmount.
4520 Severity : enhancement
4521 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4522 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4523 MPICH is included to add Lustre-specific ADIO interfaces.
4524 This is based closely on the UFS ADIO layer and only differs
4525 in file creation, in order to allow the OST striping to be set.
4526 This is user-contributed code and not supported by CFS.
4531 Description: extended inode attributes (immutable, append-only) work improperly
4532 when 2.4 and 2.6 kernels are used on client/server or vice versa
4533 Details : Introduce kernel-independent values for these flags.
4535 Severity : enhancement
4538 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4539 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4540 granularity. Now, Fractional MB granularity is be supported,
4541 this is very useful for low memory system.
4543 Severity : enhancement
4545 Description: Getattr by fid
4546 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4547 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4548 have locks and avoids getattr rpc requests for it on every stat.
4551 Frequency : Always, for filesystems larger than 2TB
4553 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4554 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4555 filesystems larger than 2TB to be mounted with mballoc enabled.
4558 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4560 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4561 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4562 filesystems larger than 2TB is mounted then the kernel will
4563 silently corrupt the start of the filesystem. CONFIG_LBD is
4564 enabled for all CFS-supported kernels, but the possibility of
4565 this happening with a modified kernel config exists.
4567 Severity : enhancement
4569 Description: add client O_DIRECT support for 2.6 kernels
4570 Details : It is now possible to do O_DIRECT reads and writes to files
4571 in the Lustre client mountpoint on 2.6 kernel clients.
4573 Severity : enhancement
4575 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4576 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4577 parallel, not waiting for response from every OST before sending
4578 a rpc to the next OST.
4583 Description: setattr vs write race when updating file timestamps
4584 Details : Client processes that update a file timestamp into the past
4585 right after writing to the file (e.g. tar) it is possible that
4586 the updated file modification time can be reset to the current
4587 time due to a race between processing the setattr and write RPC.
4589 Severity : enhancement
4591 Description: Bring 'lfs find' closer in line with regular Linux find.
4592 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4593 -print0 options and obtains all the needed info through the lustre
4596 Severity : enhancement
4598 Description: support up to 1024 configured devices on one node
4599 Details : change obd_dev array from statically allocated to dynamically
4600 allocated structs as they are first used to reduce memory usage
4605 Description: Flush dirty partially truncated pages during truncate
4606 Details : Immediatelly flush partially truncated pages in filter_setattr,
4607 this way we completely avoid having any pages in page cache on OST
4608 and can retire ugly workarounds during writes to flush such pages.
4613 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4614 Details : protect inode from truncation within vfs_unlink() context
4615 just take a reference before calling vfs_unlink() and release it
4616 when parent's i_sem is free.
4621 Description: last_id value checked outside lock on OST caused LASSERT failure
4622 Details : If there were multiple MDS->OST object precreate requests in
4623 flight, it was possible that the OST's last object id was checked
4624 outside a lock and incorrectly tripped an assertion. Move checks
4625 inside locks, and discard old precreate requests.
4628 Frequency : always, if extents are used on OSTs
4630 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4631 Details : OSTs using the extents format would not zero the high 16 bits of
4632 the index physical block number. This is not a problem for any
4633 OST filesystems smaller than 16TB, and no kernels support ext3
4634 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4635 new/modified files) and can be fixed for existing filesystems
4636 with e2fsprogs-1.39-cfs1.
4641 Description: import connection selection may be incorrect if timer wraps
4642 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4643 import connections to be ignored if the 32-bit jiffies counter
4644 wraps. Use a 64-bit jiffies counter.
4647 Frequency : very large clusters immediately after boot
4649 Description: LNET request buffers exhausted under heavy short-term load
4650 Details : If a large number of client requests are generated on a service
4651 that has previously never seen so many requests it is possible
4652 that the request buffer growth cannot keep up with the spike in
4653 demand. Instead of dropping incoming requests, they are held in
4654 the LND until the RPC service can accept more requests.
4657 Frequency : Sometimes during replay
4659 Description: Assertion failure in ll_local_open after replay.
4660 Details : If replay happened on an open request reply before we were able
4661 to set replay handler, reply will become not swabbed tripping the
4662 assertion in ll_local_open. Now we set the handler right after
4663 recognising of open request
4666 Frequency : very rare
4668 Description: kernel reports "badness in vsnprintf"
4669 Details : Reading from the "recovery_status" /proc file in small chunks
4670 may cause a negative length in lprocfs_obd_rd_recovery_status()
4671 call to vsnprintf() (which is otherwise harmless). Exit early
4672 if there is no more space in the output buffer.
4674 Severity : enhancement
4676 Description: clear OBD RPC statistics by writing to them
4677 Details : It is now possible to clear the OBD RPC statistics by writing
4678 to the "stats" file.
4683 Description: Client mtime is not the same on different clients after utimes
4684 Details : In some cases, the client was using the utimes() syscall on
4685 a file cached on another node. The clients now validate the
4686 ctime from the MDS + OSTs to determine which one is right.
4691 Description: Inability to activate failout mode
4692 Details : lconf script incorrectly assumed that in pythong string's numeric
4693 value is used in comparisons.
4696 Frequency : always with multiple stripes per file
4698 Description: Inefficient object allocation for mutli-stripe files
4699 Details : When selecting which OSTs to stripe files over, for files with
4700 a stripe count that divides evenly into the number of OSTs,
4701 the MDS is always picking the same starting OST for each file.
4702 Return the OST selection heuristic to the original design.
4707 Description: mount failures may take full timeout to return an error
4708 Details : Under some heavy load conditions it is possible that a
4709 failed mount can wait for the full obd_timeout interval,
4710 possibly several minutes, before reporting an error.
4711 Instead return an error as soon as the status is known.
4713 Frequency : quota enabled and large files being deleted
4715 Description: releasing more than 4GB of quota at once hangs OST
4716 Details : If a user deletes more than 4GB of files on a single OST it
4717 will cause the OST to spin in an infinite loop. Release
4718 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4723 Description: statfs data retrieved from /proc may be stale or zero
4724 Details : When reading per-device statfs data from /proc, in the
4725 {kbytes,files}_{total,free,avail} files, it may appear
4726 as zero or be out of date.
4729 Frequency : systems with MD RAID1 external journal devices
4731 Description: lconf's call to blkid is confused by RAID1 journal devices
4732 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4733 of returning all block devices that match the journal UUID.
4736 Frequency : always, for aggregate stripe size over 4GB
4738 Description: assertion fails when trying to use 4GB stripe size
4739 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4740 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4743 Frequency : always on ppc64
4745 Description: the first write on an ext3 filesystem with mballoc got stuck
4746 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4747 perform endianness conversion.
4749 ------------------------------------------------------------------------------
4751 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4753 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4754 INTEROPERATE with older versions automatically. Please read the
4755 user documentation before upgrading any part of a live system.
4756 * WARNING: Lustre networking configuration changes are required with
4757 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4760 * Support for kernels:
4761 2.6.9-22.0.2.EL (RHEL 4)
4762 2.6.5-7.244 (SLES 9)
4763 2.6.12.6 vanilla (kernel.org)
4766 Severity : enhancement
4767 Bugzilla : 7981/8208
4768 Description: Introduced Lustre Networking (LNET)
4769 Details : LNET is new networking infrastructure for Lustre, it includes
4770 a reorganized network configuration mode (see the user
4771 documentation for full details) as well as support for routing
4772 between different network fabrics. Lustre Networking Devices
4773 (LNDs) for the supported network fabrics have also been
4774 created for this new infrastructure.
4776 Severity : enhancement
4777 Description: Introduced Access control lists
4778 Details : clients can set ACLs on files and directories in order to have
4779 more fine-grained permissions than the standard Unix UGO+RWX.
4780 The MDS must be started with the "-o acl" mount option.
4782 Severity : enhancement
4783 Description: Introduced filesystem quotas
4784 Details : Administrators may now establish per-user quotas on the
4787 Severity : enhancement
4789 Description: Configuration change for the XT3
4790 The PTLLND is now used to run Lustre over Portals on the XT3
4791 The configure option(s) --with-cray-portals are no longer used.
4792 Rather --with-portals=<path-to-portals-includes> is used to
4793 enable building on the XT3. In addition to enable XT3 specific
4794 features the option --enable-cray-xt3 must be used.
4799 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4800 Details : A race between allocating a new llcd and re-getting the llcd_lock
4801 allowed another thread to grab newly-allocated llcd.
4803 Severity : enhancement
4805 Description: 2.6 OST async journal commit and locking fix to improve performance
4806 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4807 2.6 kernels are now async as they already were in 2.4 kernels so
4808 that they can commit concurrently with the network bulk transfer.
4809 For block-allocated files the filter allocation semaphore is held
4810 to avoid filesystem fragmentation during allocation. BKL lock
4811 removed for 2.6 xattr operations where it is no longer needed.
4816 Description: lconf incorrectly determined whether two IP networks could talk
4817 Details : In some more complicated routing and multiple-network
4818 configurations, lconf will avoid trying to make a network
4819 connection to a disjoint part of the IP space. It was doing the
4820 math incorrectly for one set of cases.
4825 Description: Fix for potential infinite loop processing records in an llog.
4826 Details : If an llog record is corrupted/zeroed, it is possible to loop
4827 forever in llog_process(). Validate the llog record length
4828 and skip the remainder of the block on error.
4831 Frequency : occasional (liblustre only)
4833 Description: liblustre could not open files whose last component is a symlink
4834 Details : sysio_path_walk() would incorrectly pass the open intent to
4835 intermediate path components.
4838 Frequency : rare (liblustre only with non-standard tuning)
4839 Bugzilla : 7201 (7350)
4840 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4841 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4842 after it has been released, only do this for referenced locks
4844 Severity : enhancement
4846 Description: specifying an (invalid) directory default stripe_size of -1
4847 would reset the directory default striping
4848 Details : stripe_size -1 was used internally to signal directory stripe
4849 removal, now use "all default" to signal dir stripe removal
4850 as a directory striping of "all default" is not useful
4853 Frequency : common for large clusters running liblustre clients
4855 Description: doing an ls when liblustre clients are running is slow
4856 Details : sending a glimpse AST to a liblustre client waits for every AST
4857 to time out, as liblustre clients will not respond. Since they
4858 cannot cache data we refresh the OST lock LVB from disk instead.
4860 Severity : enhancement
4862 Description: doing an ls at the same time as file IO can be slow
4863 Details : enqueue and other "small" requests can be blocked behind many
4864 large IO requests. Create a new OST IO portal for non-IO
4865 requests so they can be processed faster.
4868 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4870 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4871 created file with mode 000
4872 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4873 allow an ftruncate() as a non-root user to a file with mode 000.
4874 Lustre now respects this flag to disable mode checks when
4875 truncating a file owned by the user
4878 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4881 Description: Revoking locks from clients that went dead or catatonic might take
4883 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4884 cancellation of such locks instant on servers without waiting for
4885 any reply from clients. Clients drops these locks when cancel
4886 notification from server is received without replying.
4889 Frequency : liblustre-only, when liblustre client dies or becomes busy
4891 Description: Doing ls on Linux clients can take a long time with active
4893 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4894 granting such locks to it in the first place if possible. Locks
4895 are taken by proxy on the OST during the read or write and
4896 dropped immediately afterward. Add connect flags handling, do
4897 not grant locks to liblustre clients for glimpse ASTs.
4899 Severity : enhancement
4901 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4902 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4903 sequential reads and behaves badly with applications doing random
4904 reads. Improve it by reading ahead at least read region, and
4905 avoiding excessive large RPC for small reads.
4907 Severity : enhancement
4909 Description: Creating more than 1000 files for a single job may cause a load
4910 imbalance on the OSTs if there are also a large number of OSTs.
4911 Details : qos_prep_create() uses an OST index reseed value that is an
4912 even multiple of the number of available OSTs so that if the
4913 reseed happens in the middle of the object allocation it will
4914 still utilize the OSTs as uniformly as possible.
4919 Description: OST or MDS may oops in ping_evictor_main()
4920 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4921 but doesn't restart at beginning of obd_exports_timed list
4924 Severity : enhancement
4926 Description: improve by-nid export eviction on the MDS and OST
4927 Details : allow multiple exports with the same NID to be evicted at one
4928 time without re-searching the exports list.
4931 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4933 Description: MDS may oops in groups_free()
4934 Details : in rare race conditions a newly allocated group_info struct is
4935 freed again, and this can be NULL. The 2.4 compatibility code
4936 for groups_free() checked for a NULL pointer, but 2.6 did not.
4939 Frequency : common for liblustre clients doing little filesystem IO
4940 Bugzilla : 9352, 7313
4941 Description: server may evict liblustre clients accessing contended locks
4942 Details : if a client is granted a lock or receives a completion AST
4943 with a blocking AST already set it would not reply to the AST
4944 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4947 Frequency : lfs setstripe, only systems with more than 160 OSTs
4949 Description: unable to set striping with a starting offset beyond OST 160
4950 Details : llapi_create_file() incorrectly limited the starting stripe
4951 index to the maximum single-file stripe count.
4954 Frequency : LDAP users only
4956 Description: lconf did not handle in-kernel recovery with LDAP properly
4957 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4959 Severity : enhancement
4961 Description: bind OST threads to NUMA nodes to improve performance
4962 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4963 node and do their allocations there to localize memory access
4965 Severity : enhancement
4967 Description: llmount can determine client NID directly from Myrinet (GM)
4968 Details : the client NID code from gmnalnid was moved directly into
4969 llmount, removing the need to use this or specifying the
4970 client NID explicitly when mounting GM clients with zeroconf
4973 Frequency : if client is started with down MDS
4975 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4976 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4977 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4978 ptlrpc_import_delayed_req() because we didn't check if the
4979 request was interrupted, and we also didn't break out of the
4980 event loop if there was no timeout
4985 Description: data loss during non-page-aligned writes to a single file from
4986 both multiple nodes and multiple threads on one node at same time
4987 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4988 inconsistency led to false short-reads, that were cached and later
4989 used by ->prepare_write() to fill in partially written page,
4990 leading to data loss.
4993 Frequency : always, if lconf --abort_recovery used
4995 Description: lconf --abort_recovery fails with 'Operation not supported'
4996 Details : lconf was attempting to abort recovery on the MDT device and not
4999 Severity : enhancement
5001 Description: remove cleanup logs
5002 Details : replace lconf-generated cleanup logs with lustre internal
5003 cleanup routines. Eliminates the need for client-cleanup and
5006 Severity : enhancement
5008 Description: add support for EAs (user and system) on lustre filesystems
5009 Details : it is now possible to store extended attributes in the Lustre
5010 client filesystem, and with the user_xattr mount option it
5011 is possible to allow users to store EAs on their files also
5013 Severity : enhancement
5015 Description: Add possibility (config option) to show minimal available OST free
5017 Details : When compiled with --enable-mindf configure option, statfs(2)
5018 (and so, df) will return least minimal free space available from
5019 all OSTs as amount of free space on FS, instead of summary of
5020 free spaces of all OSTs.
5022 Severity : enhancement
5024 Description: do not expand extent locks acquired on OST-side
5025 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5026 by server: they are not cached anyway.
5029 Frequency : when mmap is used/binaries executed from Lustre
5031 Description: Unmmap pages before throwing them away from read cache.
5032 Details : llap_shrink cache now attempts to unmap pages before discarding
5033 them (if unmapping failed - do not discard). SLES9 kernel has
5034 extra checks that trigger if this unmapping is not done first.
5039 Description: lconf didn't resolve symlinks before checking to see whether a
5040 given mountpoint was already in use
5043 Frequency : when migrating failover services
5044 Bugzilla : 6395, 9514
5045 Description: When migrating a subset of services from a node (e.g. failback
5046 from a failover service node) the remaining services would
5047 time out and evict clients.
5048 Details : lconf --force (implied by --failover) sets the global obd_timeout
5049 to 5 seconds in order to quickly disconnect, but this caused
5050 other RPCs to time out too quickly. Do not change the global
5051 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5053 Severity : enhancement
5054 Frequency : if MDS is started with down OST
5055 Bugzilla : 9439,5706
5056 Description: Allow startup/shutdown of an MDS without depending on the
5057 availability of the OSTs.
5058 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5059 Add appropriate locking and lov-osc refcounts for safe
5060 cleaning. Add osc abort_inflight calls in case the
5061 synchronize never started.
5064 Frequency : occasional (Cray XT3 only)
5066 Description: root not authorized to access files in CRAY_PORTALS environment
5067 Details : The client process capabilities were not honoured on the MDS in
5068 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5069 been accessed by an authorized user then root was able to access
5070 the file on the local client also. The root user capabilities
5071 are now allowed on the MDS, as this environment has secure UID.
5074 Frequency : occasional
5076 Description: ldiskfs "too long searching" message happens too often
5077 Details : A debugging message (otherwise harmless) prints too often on
5078 the OST console. This has been reduced to only happen when
5079 there are fragmentation problems on the filesystem.
5084 Description: Division by zero in statfs when all OSCs are inactive
5085 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5086 lov_statfs divides by value returned by lov_get_stripecnt().
5090 Bugzilla : 9489, 3273
5091 Description: First write from each client to each OST was only 4kB in size,
5092 to initialize client writeback cache, which caused sub-optimal
5093 RPCs and poor layout on disk for the first written file.
5094 Details : Clients now request an initial cache grant at (re)connect time
5095 and so that they can start streaming writes to the cache right
5096 away and always do full-sized RPCs if there is enough data.
5097 If the OST is rebooted the client also re-establishes its grant
5098 so that client cached writes will be honoured under the grant.
5103 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5104 Details : Now I/O RPCs go to different portal number and (presumably) fast
5105 lock requests (and glimses) and other RPCs get their own service
5106 threads pool that should be able to service those RPCs
5109 Severity : enhancement
5111 Description: Ability to exchange lustre version between client and servers and
5112 issue warnings at client side if client is too old. Also for
5113 liblustre clients there is ability to refuse connection of too old
5115 Details : New 'version' field is added to connect data structure that is
5116 filled with version info. That info is later checked by server and
5120 Frequency : rare, liblustre only.
5121 Bugzilla : 9296, 9581
5122 Description: Two simultaneous writes from liblustre at offset within same page
5123 might proceed at the same time overwriting eachother with stale
5125 Details : I/O lock withing llu_file_prwv was released too early, before data
5126 actually was hitting the wire. Extended lock-holding time until
5127 server acknowledges receiving data.
5130 Frequency : extremely rare. Never observed in practice.
5132 Description: avoid generating lustre_handle cookie of 0.
5133 Details : class_handle_hash() generates handle cookies by incrementing
5134 global counter, and can hit 0 occasionaly (this is unlikely, but
5135 not impossible, because initial value of cookie counter is
5136 selected randonly). Value of 0 is used as a sentinel meaning
5137 "unassigned handle" --- avoid it. Also coalesce two critical
5138 sections in this function into one.
5140 Severity : enhancement
5142 Description: allow liblustre clients to delegate truncate locking to OST
5143 Details : To avoid overhead of locking, liblustre client instructs OST to
5144 take extent lock in ost_punch() on client's behalf. New connection
5145 flag is added to handle backward compatibility.
5147 Severity : enhancement
5148 Bugzilla : 4928, 7341, 9758
5149 Description: allow number of OST service threads to be specified
5150 Details : a module parameter allows the number of OST service threads
5151 to be specified via "options ost ost_num_threads={N}" in the
5152 OSS's /etc/modules.conf or /etc/modprobe.conf.
5156 Bugzilla : 6146, 9635, 9895
5157 Description: servers crash with bad pointer in target_handle_connect()
5158 Details : In rare cases when a client is reconnecting it was possible that
5159 the connection request was the last reference for that export.
5160 We would temporarily drop the export reference and get a new
5161 one, but this may have been the last reference and the export
5162 was just destroyed. Get new reference before dropping old one.
5164 Severity : enhancement
5165 Frequency : if client is started with failover MDS
5167 Description: Allow multiple MDS hostnames in the mount command
5168 Details : Try to read the configuration from all specified MDS
5169 hostnames during a client mount in case the "primary"
5172 Severity : enhancement
5174 Description: Stop sending data to evicted clients as soon as possible.
5175 Details : Check if the client we are about to send or are sending data to
5176 was evicted already. (Check is done every second of waiting,
5177 for which l_wait_event interface was extended to allow checking
5178 of exit condition at specified intervals).
5181 Frequency : rare, normally only when NFS exporting is done from client
5183 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5185 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5188 Frequency : always, if srand() is called before liblustre initialization
5190 Description: Liblustre uses system PRNG disturbing its usage by user application
5191 Details : Introduce internal to lustre fast and high-quality PRNG for
5192 lustre usage and make liblustre and some other places in generic
5193 lustre code to use it.
5195 Severity : enhancement
5196 Bugzilla : 9477, 9557, 9870
5197 Description: Verify that the MDS configuration logs are updated when xml is
5198 Details : Check if the .xml configuration logs are newer than the config
5199 logs stored on the MDS and report an error if this is the case.
5200 Request --write-conf, or allow starting with --old_conf.
5202 Severity : enhancement
5204 Description: Handle symlinks in the path when checking if Lustre is mounted.
5205 Details : Resolve intermediate symlinks when checking if a client has
5206 mounted a filesystem to avoid duplicate client mounts.
5211 Description: lconf can hit an error exception but still return success.
5212 Details : The lconf command catches the Command error exception at the top
5213 level script context and will exit with the associated exit
5214 status, but doesn't ensure that this exit status is non-zero.
5219 Description: failure of ptlrpc thread startup can cause oops
5220 Details : Starting a ptlrpc service thread can fail if there are a large
5221 number of threads or the server memory is very fragmented.
5222 Handle this without oopsing.
5225 Frequency : always, only if liblustre and non-default acceptor port was used
5227 Description: liblustre cannot connect to servers with non-default acceptor port
5228 Details : tcpnal_set_default_params() was not called and was therefore
5229 ignoring the environment varaible TCPNAL_PORT, as well as other
5230 TCPNAL_ environment variables
5235 Description: two objects could be created on the same OST for a single file
5236 Details : If an OST is down, in some cases it was possible to create two
5237 objects on a single OST for a single file. No problems other
5238 than potential performance impact and spurious error messages.
5242 Bugzilla : 5681, 9562
5243 Description: Client may oops in ll_unhash_aliases
5244 Details : Client dcache may become inconsistent in race condition.
5245 In some cases "getcwd" can fail if the current directory is
5251 Description: Inode refcounting problems in NFS export code
5252 Details : link_raw functions used to call d_instantiate without obtaining
5253 extra inode reference first.
5257 Bugzilla : 9942, 9903
5258 Description: Referencing freed requests leading to crash, memleaks with NFS.
5259 Details : We used to require that call to ll_revalidate_it was always
5260 followed by ll_lookup_it. Also with revalidate_special() it is
5261 possible to call ll_revalidate_it() twice for the same dentry
5262 even if first occurence returned success. This fix changes semantic
5263 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5264 reference on a request referred from the intent.
5265 ll_intent_release() then releases such a request.
5268 Frequency : rare, normally benchmark loads only
5270 Description: unlinked inodes were kept in memory on the client
5271 Details : If a client is repeatedly creating and unlinking files it
5272 can accumulate a lot of stale inodes in the inode slab cache.
5273 If there is no other client load running this can cause the
5274 client node to run out of memory. Instead flush old inodes
5275 from client cache that have the same inode number as a new inode.
5278 Frequency : SLES9 2.6.5 kernel and long filenames only
5279 Bugzilla : 9969, 10379
5280 Description: utime reports stale NFS file handle
5281 Details : SLES9 uses out-of-dentry names in some cases, which confused
5282 the lustre dentry revalidation. Change it to always use the
5286 Frequency : rare, unless heavy write-truncate concurrency is continuous
5287 Bugzilla : 4180, 6984, 7171, 9963, 9331
5288 Description: OST becomes very slow and/or deadlocked during object unlink
5289 Details : filter_destroy() was holding onto the parent directory lock
5290 while truncating+unlinking objects. For very large objects this
5291 may block other threads for a long time and slow overall OST
5292 responsiveness. It may also be possible to get a lock ordering
5293 deadlock in this case, or run out of journal credits because of
5294 the combined truncate+unlink. Solution is to do object truncate
5295 first in one transaction without parent lock, and then do the
5296 final unlink in a new transaction with the parent lock. This
5297 reduces the lock hold time dramatically.
5300 Frequency : rare, 2.4 kernels only
5302 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5303 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5304 before unmounting their filesystem in order to avoid deadlock.
5305 The kernel_locked() function in 2.4 kernels only checks whether
5306 the kernel lock is held, not whether it is this process that is
5307 holding it as 2.6 kernels do.
5312 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5313 Details : The client ptlrpc code may be trying to reconnect to a down
5314 server before a previous connection attempt has timed out.
5315 Increase the reconnect interval to be longer than the connection
5316 timeout interval to avoid sending duplicate connections to
5320 Frequency : echo_client brw_test command
5322 Description: fix echo_client to work with OST preallocated code
5323 Details : OST preallocation code (5137) didn't take echo_client IO path
5324 into account: echo_client calls filter methods outside of any
5325 OST thread and, hence, there is no per-thread preallocated
5326 pages and buffers to use. Solution: hijack pga pages for IO. As
5327 a byproduct, this avoids unnecessary data copying.
5331 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5332 Description: Client can oops in mdc_commit_close() after open replay
5333 Details : It was possible for the MDS to return an open request with no
5334 transaction number in mds_finish_transno() if the client was
5335 evicted, but without actually returning an error. Clients
5336 would later try to replay that open and may trip an assertion
5337 Simplify the client close codepath, and always return an error
5338 from the MDS in case the open is not successful.
5341 Frequency : rare, 2.6 OSTs only
5343 Description: OST may deadlock under high load on fragmented files
5344 Details : If there was a heavy load and highly-fragmented OST filesystems
5345 it was possible to have all the OST threads deadlock waiting on
5346 allocation of biovecs, because the biovecs were not released
5347 until the entire RPC IO was completed. Instead, release biovecs
5348 as soon as they are complete to ensure forward IO progress.
5350 Severity : enhancement
5352 Description: Support for specifying external journal device at mount
5353 Details : If an OST or MDS device is formatted with an external journal
5354 device, this device major/minor is stored in the ext3 superblock
5355 and may not be valid for failover. Allow detecting and
5356 specifying the external journal at mount time.
5361 Description: Mounting an MDS with pending unlinked files may cause oops
5362 Details : target_finish_recovery() calls mds_postrecov() which returned
5363 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5364 considers this an error and immediately begins cleaning up the
5365 lov, just after starting the mds_lov process
5367 Severity : enhancement
5369 Description: Implement 'lfs df' to report actual free space on per-OST basis
5370 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5371 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5372 usage of objects; '-h' to report in human readable format.
5374 ------------------------------------------------------------------------------
5376 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5383 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5384 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5385 at startup during orphan unlinking. Instead, skip buddy update
5386 before setup, as it will be regenerated after recovery is complete.
5391 Description: If an OST is inactive, its locks might reference stale inodes.
5392 Details : lov_change_cbdata() must iterate over all namespaces, even if
5393 they are inactive to clear inode references from the lock.
5395 Severity : enhancement
5396 Frequency : occasional, if non-standard max_dirty_mb used
5398 Description: Client will block write RPCs if not enough grant
5399 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5400 then the client will block writes while waiting for another RPC
5401 to complete instead of consuming its dirty limit. With change
5402 we get improved performance when max_dirty_mb is small.
5404 Severity : enhancement
5405 Bugzilla : 3389, 6253
5406 Description: Add support for supplementary groups on the MDS.
5407 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5408 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5409 lookups for user supplementary groups into a cache.
5413 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5414 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5415 for liblustre O_EXCL is enforced by MDS.
5419 Description: readdir never returns NULL in liblustre.
5420 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5421 dentry in struct dirent.
5425 Frequency : liblustre only, depends on application IO pattern
5426 Description: liblustre clients evicted if not contacting servers
5427 Details : Don't put liblustre clients into the ping_evictor list, so
5428 they will not be evicted by the pinger ever.
5430 Severity : enhancement
5432 Description: Add ability to evict clients by NID from MDS.
5433 Details : By echoing "nid:$NID" string into
5434 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5435 $NID would be instantly evicted from this MDS and from all active
5436 OSTs connected to it.
5440 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5441 Details : lookup_it_finish() used to query file size from OSTs that was not
5446 Description: service threads change working directory to that of init
5447 Details : Starting lustre service threads may pin the working directory
5448 of the parent thread, making that filesystem busy. Threads
5449 now change to the working directory of init to avoid this.
5453 Frequency : during shutdown only
5454 Description: shutdown with a failed MDS or OST can cause unmount to hang
5455 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5460 Frequency : 2.6 only
5461 Description: chmod/chown may include an extra supplementary group
5462 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5463 supplementary group and if none is specified this is used.
5466 Bugzilla : 5479 (6816)
5468 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5469 Details : If lookup is in progress on a file that is unlinked we might try
5470 to revalidate the inode and fail in revalidate after lookup is
5471 complete and ll_file_open() enqueues the open again but
5472 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5475 Frequency : always, if lconf --abort_recovery used
5477 Description: lconf --abort_recovery fails with 'Operation not supported'
5478 Details : lconf was attempting to abort recovery on the MDT device and not
5481 ------------------------------------------------------------------------------
5483 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5488 Frequency : rare (only unsupported configurations with a node running as an
5490 Bugzilla : 6514, 5137
5491 Description: Mounting a Lustre file system on a node running as an OST could
5493 Details : OSTs now preallocates memory needed to write out data at
5494 startup, instead of when needed, to avoid having to
5495 allocate memory in possibly low memory situations.
5496 Specifically, if the file system is mounted on on OST,
5497 memory pressure could force it to try to write out data,
5498 which it needed to allocate memory to do. Due to the low
5499 memory, it would be unable to do so and the node would
5500 become unresponsive.
5502 Severity : enhancement
5504 Description: Addition of lconf --service command line option
5505 Details : lconf now accepts a '--service <arg>' option, which is
5506 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5508 Severity : enhancement
5510 Description: Failover mode is now the default for OSTs.
5511 Details : By default, OSTs will now run in failover mode. To return to
5512 the old behaviour, add '--failout' to the lmc line for OSTs.
5514 Severity : enhancement
5516 Description: Health checks are now provided for MDS and OSTs
5517 Details : Additional detailed health check information on MSD and OSTs
5518 is now provided through the procfs health_check value.
5521 Frequency : occasional, depends on IO load
5523 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5524 after numerous create/delete cycles
5525 Details : The ext3 inode allocation policy would not allocate new inodes
5526 very well on the OSTs because there are no new directories
5527 being created. Instead we look for groups with free space if
5528 the parent directories are nearly full.
5533 Description: Network or server problems during mount may cause partially
5534 mounted clients instead of returning an error.
5535 Details : The config llog parsing code may overwrite the error return
5536 code during mount error handling, returning success instead
5542 Description: MDS can fail to allocate large reply buffers
5543 Details : After long uptimes the MDS can fail to allocate large reply
5544 buffers (e.g. zconf client mount config records) due to memory
5545 fragmentation or consumption by the buffer cache. Preallocate
5546 some large reply buffers so that these replies can be sent even
5547 under memory pressure.
5551 Frequency : rare (liblustre)
5552 Description: fsx running with liblustre complained that using truncate() to
5553 extend the file doesn't work. This patch corrects that issue.
5554 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5555 ATTR_SIZE and lsm use in llu_setattr_raw.
5559 Frequency : rare, only 2.6 kernels
5560 Description: Unusual file access patterns on the MDS may result in inode
5561 data being lost in very rare circumstances.
5562 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5563 code in the 2.6 kernel caused the inode and/or EA data not to
5564 be read from disk, causing single-file corruption.
5568 Frequency : rare, only 2.6 filesystems using extents
5569 Description: Heavy concurrent write and delete load may cause data corruption.
5570 Details : It was possible under high-load situations to have an extent
5571 metadata block in the block device cache from a just-unlinked
5572 file overwrite a newly-allocated data block. We now unmap any
5573 metadata buffers that alias just-allocated data blocks.
5577 Frequency : filesystems with default stripe_count larger than 77
5578 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5579 Details : lconf specifies an inode size of 4096 bytes when the default
5580 stripe_count is larger than 77. This conflicts with the default
5581 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5582 this case to avoid pinning too much memory for large EAs.
5584 ------------------------------------------------------------------------------
5586 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5591 Frequency : rare (extremely heavy IO load with hundreds of clients)
5593 Description: Client is evicted, gets IO error writing to file
5594 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5595 caused clients to be evicted for AST timeouts. The fixes in
5596 bug 5192 mean we no longer need to have such short AST timeouts
5597 so ldlm_timeout has been increased.
5600 Frequency : occasional during --force or --failover shutdown under load
5601 Bugzilla : 5949, 4834
5602 Description: Server oops/LBUG if stopped with --force or --failover under load
5603 Details : a collection of import/export refcount and cleanup ordering
5604 issues fixed for safer force cleanup
5607 Frequency : only filesystems larger than 120 OSTs
5608 Bugzilla : 5990, 6223
5609 Description: lfs getstripe would oops on a very large filesystem
5610 Details : lov_getconfig used kfree on vmalloc'd memory
5613 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5614 Bugzilla : 6242, 6243
5615 Description: reading from files that had been truncated to a non-zero size
5616 but never opened returned no data
5617 Details : ll_file_read() reads zeros from no-object files to EOF
5622 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5623 Details : There's little harm in aborting MDS/OSS recovery and letting it
5624 try again, so I removed the LASSERT and return an error instead.
5626 Severity : enhancement
5628 Description: New debugging infrastructure for tracking down data corruption
5629 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5630 (b) cover more of the client-side code path, and (c) try to narrow
5631 down where problems occurred
5635 Bugzilla : 3819, 4364, 4397, 6313
5636 Description: Racing close and eviction MDS could cause assertion in mds_close
5637 Details : It was possible to get multiple mfd references during close and
5638 client eviction, leading to one thread referencing a freed mfd.
5640 Severity: : enhancement
5641 Bugzilla : 3262, 6359
5642 Description: Attempts to reconnect to servers are now more aggressive.
5643 Details : This builds on the enhanced upcall-less recovery that was added
5644 in 1.4.2. When trying to reconnect to servers, clients will
5645 now try each server in the failover group every 10 seconds. By
5646 default, clients would previously try one server every 25 seconds.
5651 Description: After recovery, certain operations trigger a failed
5652 assertion on a client.
5653 Details : Failing over an mds, using lconf -d --failover, while a
5654 client was doing a readdir() call would cause the client to
5655 LBUG after recovery completed and the readdir() was resent.
5657 Severity : enhancement
5659 Description: Default groups are now added by lconf
5660 Details : You can now run lconf --group <servicename> without having to
5661 manually add groups with lmc.
5664 Frequency : occasional
5666 Description: Nodes with an elan id of 0 trigger a failed assertion
5669 Frequency : always when accessing e.g. tty/console device nodes
5671 Description: tty and some other devices nodes cannot be used on lustre
5672 Details : file's private_data field is used by device data and lustre
5673 values in there got lost. New field was added to struct file to
5674 store fs-specific private data.
5677 Frequency : when exporting Lustre via NFS
5679 Description: NFSD failed occasionally when looking up a path component
5680 Details : NFSD is looking up ".." which was broken in ext3 directories
5681 that had grown large enough to become hashed.
5684 Frequency : Clusters with multiple interfaces not on the same subnet
5686 Description: Nodes will repeatedly try to reconnect to an interface which it
5687 cannot reach and report an error to the log.
5688 Details : Extra peer list entries will be created by lconf with some peers
5689 unreachable. lconf now validates the peer before adding it.
5692 Frequency : Only if a default stripe is set on the filesystem root.
5694 Description: Setting a default stripe on the filesystem root prevented the
5695 filesystem from being remounted.
5696 Details : The client was sending extra request flags in the root getattr
5697 request and did not allocate a reply buffer for the dir EA.
5700 Frequency : occasional, higher if lots of files are accessed by one client
5701 Bugzilla : 6159, 6097
5702 Description: Client trips assertion regarding lsm mismatch/magic
5703 Details : While revalidating inodes the VFS looks up inodes with ifind()
5704 and in rare cases can find an inode that is being freed.
5705 The ll_test_inode() code will free the lsm during ifind()
5706 when it finds an existing inode and then the VFS later attaches
5707 this free lsm to a new inode.
5711 Bugzilla : 6422, 7030
5712 Description: MDS deadlock between mkdir and client eviction
5713 Details : Creating a new file via mkdir or mknod (starting a transaction
5714 and getting the ns lock) can deadlock with client eviction
5715 (gets ns lock and trying to finish a synchronous transaction).
5718 Frequency : occasional
5719 Description: While starting a server, the fsfilt_ext3 module could not be
5721 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5722 kernels. Previously, lconf would still use the ext3 name
5723 when trying to load modules. Now, it will correctly use
5724 ext3 on 2.4 and ldiskfs on 2.6.
5726 Severity : enhancement
5727 Description: The default stripe count has been changed to 1
5728 Details : The interpretation of the default stripe count (0, to lfs
5729 or lmc) has been changed to mean striping across a single
5730 OST, rather than all available. For general usage we have
5731 found a stripe count of 1 or 2 works best.
5733 Severity : enhancement
5734 Description: Add support for compiling against Cray portals.
5735 Details : Conditional compiling for some areas that are different
5739 Frequency : occasional
5740 Bugzilla : 6409, 6834
5741 Description: Creating files with an explicit stripe count may lead to
5742 a failed assertion on the MDS
5743 Details : If some OSTs are full or unavailable, creating files may
5744 trigger a failed assertion on the MDS. Now, Lustre will
5745 try to use other servers or return an error to the
5749 Frequency : occasional
5751 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5752 caused lock timeout during readahead (since 1.4.2).
5753 Details : Processes doing readahead might match a lock that hasn't been
5754 granted yet if there are overlapping and conflicting lock
5755 requests. The readahead process waits on ungranted lock
5756 (original lock is CBPENDING), while OST waits for that process
5757 to cancel CBPENDING read lock and eventually evicts client.
5759 Severity : enhancement
5761 Description: Initial enabling of flock support for clients
5762 Details : Implements fcntl advisory locking and file status functions.
5763 This feature is provided as an optional mount flag (default
5764 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5765 locking are implemented yet, and those that are are not guaranteed
5766 to be completely correct in production environments.
5767 mount -t lustre -o [flock|noflock] ...
5770 Frequency : occasional
5772 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5773 assertion in the ext3 JBD (journaling) layer.
5774 Details : The b_committed_data struct is protected by the big kernel lock
5775 in 2.4 kernels, serializing journal_commit_transaction() and
5776 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5777 access to this struct. In 2.6 kernels there is finer grained
5778 locking to improve SMP performance of the JBD layer.
5782 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5787 Description: merge and cleanup kernel patches.
5790 -----------------------------------------------------------------------------
5792 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5794 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5795 versions of Lustre. Please update all servers and clients to
5796 version 1.4.2 or later at the same time. You must also run
5797 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5798 to update the configuration logs.
5800 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5801 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5802 - swab llog records of type '0' so we get proper header size/idx (5861)
5803 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5804 - fix rename of one directory over another leaking an inode (5953)
5805 - avoid SetPageDirty on 2.6 (5981)
5806 - don't re-add just-being-destroyed locks to the waiting list (5653)
5807 - when creating new directories, inherit the parent's custom
5808 striping settings if present parent (3048)
5809 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5810 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5811 - don't hold client locks on temporary worklist from l_lru (5666)
5812 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5813 - automatically evict dead clients (5921)
5814 - Update file size properly in create+truncate+fstat case (6196)
5815 - Do not unhash mountpoint dentries, do not allow removal of
5817 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5818 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5819 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5820 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5821 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5822 - Correctly handle reads of files with no objects (6243)
5823 - lctl recover will also mark a device active if deactivate used (5933)
5825 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5826 - allow --write-conf on an MDS with different nettype than client (5619)
5827 - don't write config llogs to MDS for mounts not from that MDS (5617)
5828 - lconf should create multiple TCP connections from a client (5201)
5829 - init scripts are now turned off by default; run chkconfig --on
5830 lustre and chkconfig --on lustrefs to use them
5831 - upcalls are no longer needed for clients to recover to failover
5833 - add --abort-recovery option to lconf to abort recovery on device
5835 - add support for an arbitrary number of OSTs (3026)
5836 - Quota support protocol changes.
5837 - forward compatibility changes to wire structs (6007)
5838 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5839 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5840 - improved "lustre status" script
5841 - initialize blocksize for non-regular files (6062)
5842 - added --disable-server and --disable-client configure options (5782)
5843 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5844 - Vanilla 2.4.29 support
5845 - increase maximum number of obd devices to 520 (6242)
5846 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5847 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5848 SLES 9 2.6.5 kernels
5849 - limit stripes per file to 160 (the maximum EA size) (6093)
5851 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5854 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5855 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5856 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5857 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5858 - small ext3 extents cleanups and fixes (5733)
5859 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5860 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5861 - avoid needless client->OST connect, fix handle mismatch (5317)
5862 - fix DLM error path that led to out-of-sync client, long delays (5779)
5863 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5864 - fix several locking issues related to i_size (5492,5624,5654,5672)
5865 - don't move pending lock onto export if it is already evicted (5683)
5866 - fix kernel oops when creating .foo in unlinked directory (5548)
5867 - fix deadlock in obdfilter statistics vs. object create (5811)
5868 - use time_{before,after} to avoid timer jiffies wrap (5882)
5869 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5870 - Do not leak request if server was not able to process it (5154)
5871 - If mds_open unable to find parent dir, make that negative lookup(5154)
5872 - don't create new directories with extent-mapping (5909, 5936)
5874 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5875 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5876 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5877 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5878 - support for RHEL4 kernel on i686 (5773)
5879 - provide error messages when incompatible logs are encountered (5898)
5881 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5882 * version 1.4.0.10 (1.4.1 release candidate 1)
5884 - don't keep a lock reference when lock is not granted (4238)
5885 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5886 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5887 - drop import inflight refcount on signal_completed_replay error (5255)
5888 - unlock page after async write error during send (3677)
5889 - handle missing objects in filter_preprw_read properly (5265)
5890 - no transno return for symlink open, don't save no-trasno open (3440)
5891 - don't try to complete elan receive that already failed (4012)
5892 - free RPC server reply state on error (5406)
5893 - clean up thread from ptlrpc_start_thread() on error (5160)
5894 - readahead could read extra page into cache that wasn't ejected (5388)
5895 - prevent races in class_attach/setup/cleanup/detach (5260)
5896 - don't dereference de->d_inode after l_dput of de (5458)
5897 - use "int" for stripe value returned from lock_to_stripe (5544)
5898 - mballoc allocation and error-checking fixes in 2.6 (5504)
5899 - block device patches to fix I/O request sizes in 2.6 (5482)
5900 - look up hostnames for IB nals (5602)
5901 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5902 - don't start multiple acceptors for the same port (5277)
5903 - fix incorrect LASSERT in mds_getattr_name (5635)
5904 - export a proc file for general "ping" checking (5628)
5905 - fix "lfs check" to not block when the MDS is down (5628)
5907 - service request history (4965)
5908 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5909 - create an "evict_client" /proc entry on OSTs, like the MDS has
5910 - fix mount usage message, return errors per mount(8) (5168)
5911 - change grep [] to grep "[]" in tests so they work in more UMLs
5912 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5913 - remove ancient LOV_MAGIC_V0 EA support (5047)
5914 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5915 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5916 - allow readahead allocations to fail when low on memory (5383)
5917 - mmap locking landed again, after considerable improvement (2828)
5918 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5920 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - send OST transaction number in read/write reply to free req (4966)
5924 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5925 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5926 - the watchdog thread now runs as interruptible (5246)
5927 - flock/lockf fixes (but it's still disabled, pending 5135)
5928 - don't use EXT3 constants in llite code (5094)
5929 - memory shortage at startup could cause assertion (5176)
5931 - reorganization of lov code
5932 - single portals codebase
5934 - add extents/mballoc support (5025)
5935 - direct I/O reads in the obdfilter (4048)
5936 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5938 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5941 - send OST transaction number in read/write reply to free req (4966)
5942 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5943 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5944 - the watchdog thread now runs as interruptible (5246)
5945 - handle missing objects in filter_preprw_read properly (5265)
5946 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5947 - drop import inflight refcount on signal_completed_replay error (5255)
5948 - unlock page after async write error during send (3677)
5949 - return original error code on reconstructed replies (3761)
5950 - no transno return for symlink open, don't save no-trasno open (3440)
5952 - add pid to ldlm debugging output (4922)
5953 - bump the watchdog timeouts -- we can't handle 30sec yet
5954 - extra debugging for orphan dentry/inode bug (5259)
5956 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5959 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5960 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5961 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5962 - fix NULL dereference in /proc/sys/portals/routes (4827)
5963 - allow failed mdc_close() operations to be interrupted (4561)
5964 - stop precreate on OST before MDS would time out on it (4778)
5965 - don't send partial-page writes before EOF from client (4410)
5966 - discard client grant for sub-page writes on large-page clients (4520)
5967 - don't free dentries not owned by NFS code, check generation (4806)
5968 - fix lsm leak if mds_create_objects() fails (4801)
5969 - limit debug_daemon file size, always print CERROR messages (4789)
5970 - use transno after validating reply (3892)
5971 - process timed out requests if import state changes (3754)
5972 - update mtime on OST during writes, return in glimpse (4829)
5973 - add mkfsoptions to LDAP (4679)
5974 - use ->max_readahead method instead of zapping global ra (5039)
5975 - don't interrupt __l_wait_event() during strace
5977 - add software watchdogs to catch hung threads quickly (4941)
5978 - make lustrefs init script start after nfs is mounted
5979 - fix CWARN/ERROR duplication (4930)
5980 - return async write errors to application if possible (2248)
5981 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5982 - print NAL number in %x format (4645)
5983 - update barely-supported suse-2.4.21-171 series (4842)
5984 - support for sles 9 %post scripts
5985 - support for building 2.6 kernel-source packages
5986 - support for sles km_* packages
5988 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5991 - ignore -ENOENT errors in osc_destroy (3639)
5992 - notify osc create thread that OSC is being cleaned up (4600)
5993 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5994 - reconstruct ost_handle() like mds_handle() (4657)
5995 - create a new thread to do import eviction to avoid deadlock (3969)
5996 - let lconf resolve symlinked-to devices (4629)
5997 - don't unlink "objects" from directory with default EA (4554)
5998 - hold socknal file ref over connect in case target is down (4394)
5999 - allow more than 32000 subdirectories in a single directory (3244)
6000 - fix blocks count for O_DIRECT writes (3751)
6001 - OST returns ENOSPC from object create when no space left (4539)
6002 - don't send truncate RPC if file size isn't changing (4410)
6003 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6004 - bind to privileged port in socknal and tcpnal (3689)
6006 - rate limit CERROR/CWARN console message to avoid overload (4519)
6007 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6008 - basic mmap support (3918)
6009 - kernel patch series update from b1_4 (4711)
6011 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6014 - avoid crash during MDS cleanup with OST shut down (2775)
6015 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6016 - don't use bad inodes on the MDS (3744)
6017 - dynamic object preallocation to improve recovery speed (4236)
6018 - don't hold spinlock over lock dumping or change debug flags (4401)
6019 - don't zero obd_dev when it is force cleaned (3651)
6020 - print grants to console if they go negative (4431)
6021 - "lctl deactivate" will stop automatic recovery attempts (3406)
6022 - look for existing locks in ldlm_handle_enqueue() (3764)
6023 - don't resolve lock handle twice in recovery avoiding race (4401)
6024 - revalidate should check working dir is a directory (4134)
6026 - don't always mark "slow" obdfilter messages as errors (4418)
6028 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6031 - don't close LustreDB during write_conf until it is done (3860)
6032 - fix typo in lconf for_each_profile (3821)
6033 - allow dumping logs from multiple threads at one time (3820)
6034 - don't allow multiple threads in OSC recovery (3812)
6035 - fix debug_size parameters (3864)
6036 - fix mds_postrecov to initialize import for llog ctxt (3121)
6037 - replace config semaphore with spinlock (3306)
6038 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6039 - don't allow enqueue to complete on a destroyed export (3822)
6040 - down write_lock before checking llog header bitmap (3825)
6041 - recover from lock replay timeout (3764)
6042 - up llog sem before sending rpc (3652)
6043 - reduce ns lock hold times when setting kms (3267)
6044 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6045 - fix NULL deref and obd_dev leak on setup error (3312)
6046 - replace some LBUG about llog ops with error handling (3841)
6047 - don't match INVALID dentries from d_lookup and spin (3784)
6048 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6049 - fix invalid assertion in ptlrpc_set_wait (3880)
6051 - add libwrap support for the TCP acceptor (3996)
6052 - add /proc/sys/portals/routes for non-root route listing (3994)
6053 - allow setting MDS UUID in .xml (2580)
6054 - print the stack of a process that LBUGs (4228)
6056 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6059 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6060 - make sure to unset replay flag from failed open requests (3440)
6061 - if default stripe count is 0, use OST count for inode size (3636)
6062 - update parent mtime/ctime on client for create/unlink (2611)
6063 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6064 - free recovery state on server during a forced cleanup (3571)
6065 - unregister_reply for resent reqs (3063)
6066 - loop back devices mounting and status check on 2.6 (3563)
6067 - fix resource-creation race that can provoke i_size == 0 (3513)
6068 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6069 - more debugging for page-accounting assertion (3746)
6070 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6071 - avoid deadlock after precreation failure (3758)
6072 - fix race and lock order deadlock in orphan handling (3450, 3750)
6073 - add validity checks when grabbing inodes from l_ast_data (3599)
6075 - add /proc/.../recovery_status to obdfilter (3428)
6076 - lightweight CDEBUG infrastructure, debug daemon (3668)
6077 - change default OSC RPC parameters to be better on small clusters
6078 - turn off OST read cache for files smaller than 32MB
6079 - install man pages and include them in rpms (3100)
6080 - add new init script for (un)mounting lustre filesystems (2593)
6081 - run chkconfig in %post for init scripts (3701)
6082 - drop scimac NAL (unmaintained)
6084 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6087 - clean kiobufs before and after use (3485)
6088 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6089 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6090 - change init script to fail more clearly if not run as root (1528)
6091 - allow clients to reconnect during replay (1742)
6092 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6093 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6094 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6095 - fix cond_resched (3554)
6096 - only evict unfinished clients after recovery (3515)
6097 - allow bulk resend, prevent data loss (3570)
6098 - dynamic ptlrpc request buffer allocation (2102)
6099 - don't allow unlinking open directory if it isn't empty (2904)
6100 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6101 - remove extraneous obd dereference causing LASSERT failure (3334)
6102 - don't use get_cycles() when creating temp. files on the mds (3156)
6103 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6104 - handle EEXIST for set-stripe, set proper directory name (3336)
6106 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6107 - fix ksocknal_fmb_callback() error messages (2918)
6109 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6112 - don't copy lvb into (possibly NULL) reply on error (2983)
6113 - don't deref dentry after dput, don't free lvb on error (2922)
6114 - use the kms to determine writeback rpc length (2947)
6115 - increment oti_logcookies when osc is inactive (2948)
6116 - update client's i_blocks count via lvb messages (2543)
6117 - handle intent open/close of special files properly (1557)
6118 - mount MDS with errors=remount-ro, like obdfilter (2009)
6119 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6120 - don't use cancelling-locks' kms values (2947)
6121 - use highest lock extent for kms, not last one (2925)
6122 - don't dereference ERR_PTR() dentry in error handling path (3107)
6123 - fix thread race in portals_debug_dumplog() (3122)
6124 - create lprocfs device entries at setup instead of at attach (1519)
6125 - common AST error handler, don't evict client on completion race (3145)
6126 - zero nameidata in detach_mnt in 2.6 (3118)
6127 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6128 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6129 - initialize RPC timeout timer earlier for 2.6 (3219)
6130 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6131 - print nal/nid for unknown nid (3258)
6132 - additional checks for oscc recovery before doing precreate (3284)
6133 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6134 - don't crash in mdc_close for bad permissions on open (3285)
6135 - zero i_rdev for non-device files (3147)
6136 - clear page->private before handing to FS, better assertion (3119)
6137 - tune the read pipeline (3236)
6138 - fix incorrect decref of invalidated dentry (2350)
6139 - provide read-ahead stats and refine rpc in flight stats (3328)
6140 - don't hold journal transaction open across create RPC (3313)
6141 - update atime on MDS at close time (3265)
6142 - close LDAP connection when recovering to avoid server load (3315)
6143 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6144 - don't leak open file on MDS after open resend (3325)
6145 - serialize filter_precreate and filter_destroy_precreated (3329)
6146 - loop device shouldn't call sync_dev() for nul device (3092)
6147 - clear page cache after eviction (2766)
6148 - resynchronize MDS->OST in background (2824)
6149 - refuse to mount the same filesystem twice on same mountpoint (3394)
6150 - allow llmount to create routes for mounting behind routers (3320)
6151 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6152 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6153 - fix rare problem with rename on htree directories (3417)
6155 - allow default OST striping configuration per directory (1414)
6156 - fix compilation for qswnal for 2.6 kernels (3125)
6157 - increase maximum number of MDS request buffers for large systems
6158 - change liblustreapi to be useful for external progs like lfsck (3098)
6159 - increase local configuration timeout for slow disks (3353)
6160 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6162 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6165 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6166 - don't overwrite extent policy data in reply if lock was blocked (2901)
6167 - drop filter export grants atomically with removal from device (2663)
6168 - del obd_self_export from work_list in class_disconnect_exports (2908)
6169 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6170 - swab reply message in mdc_close, other PPC fixes (2464)
6171 - fix destroying of named logs (2325)
6172 - overwrite old logs when running lconf --write_conf (2264)
6173 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6174 - fix race in target_handle_connect (2898)
6175 - mds_reint_create() should take same inode create lock (2926)
6176 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6177 - don't close files for self_export to avoid uninitialized obd (2936)
6178 - allow MDS with the same name as client node (2939)
6179 - hold dentry reference for closed log files for unlink (2325)
6180 - reserve space for all logs during transactions (2059)
6181 - don't evict page beyond end of stripe extent (2925)
6182 - don't oops on a deleted current working directory (2399)
6183 - handle hard links to targets without a parent properly (2517)
6184 - don't dereference NULL lock when racing during eviction (2867)
6185 - don't grow lock extents when lots of conflicting locks (2919)
6187 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6190 - account for cache space usage on clients to avoid data loss (974)
6191 - lfsck support in lustre kernel code (2349)
6192 - reduce journal credits needed for BRW writes (2370)
6193 - orphan handling to avoid losing space on client/server crashes
6194 - ptlrpcd can be blocked, stopping ALL progress (2477)
6195 - use lock value blocks to assist in proper KMS, faster stat (1021)
6196 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6197 - recovery for initial connections (2355)
6198 - fixes for mds_cleanup_orphans (1934)
6199 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6200 - block all file creations until orphan recovery completes (1901)
6201 - client remove rq_connection from request struct (2423)
6202 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6203 - recovery timer race (2670)
6204 - mdc_close recovey bug (2532)
6205 - ptlrpc cleanup bug (2710)
6206 - mds timeout on local locks (2588)
6207 - namespace lock held during RPCs (2431)
6208 - handle interrupted sync write properly (2503)
6209 - don't try to handle a message that hasn't been replied to (2699)
6210 - client assert failure during cleanup after abort recovery (2701)
6211 - leak mdc device after failed mount (2712)
6212 - ptlrpc_check_set allows timedout requests to complete (2714)
6213 - wait for inflight reqs when ptlrpcd finishes (2710)
6214 - make sure unregistered services are removed from the srv_list
6215 - reset bulk XID's when resending them (caught by 1138 test)
6216 - unregister_bulk after timeout
6217 - fix lconf error (2694)
6218 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6219 - readahead locks pages, leaves pending causing memory pressure (2673)
6220 - increase OST request buffers to 4096 on large machines (2729)
6221 - fix up permission of existing directories in simple_mkdir (2661)
6222 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6223 - don't assign transno to errored transactions (2742)
6224 - don't delete objects on OST if given a bogus objid from MDS (2751)
6225 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6226 - if rq_no_resend, then timeout request after recovery (2432)
6227 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6228 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6229 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6230 - MDS recovery shouldn't delete objects that it has given out (2730)
6231 - if enqueue arrives after completion, don't clobber LVB (2819)
6232 - don't unlock pages twice when trigger_group_io returns error (2814)
6233 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6234 - don't write pages to disk if there was an error (1450)
6235 - don't ping imports that have recovery disabled (2676)
6236 - take buffered bytes into account when balancing socknal conn (2817)
6237 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6238 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6239 - remove little-endian swabbing of llog records (1987)
6240 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6241 - retry reposting request buffers if they fail (1191)
6242 - grow extent at grant time to avoid granting a revoked lock (2809)
6243 - lock revoke doesn't evict page if covered by a second lock (2765)
6244 - disable VM readahead to avoid reading outside lock extents (2805)
6246 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6247 - updated kernel patches for hp-2.4.20 kernel (2681)
6249 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6252 - fix truncated write corruption (2366)
6253 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6255 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6256 - ensure len(uuid) < 37 in lmc (1171)
6257 - fix ia64 OOPS in llog_test (2255)
6258 - zero end of page at obdfilter for partial page writes (2648)
6259 - don't leave stale dentries around after renames (bug 2428)
6260 - fix timeouts when evicting a client with a single lock held (2642)
6261 - set deadline for the initial HELLO message to drain (2634)
6262 - print out dotted-quad IP addresses in the socknal (2302)
6264 - additional debugging for MDS client eviction problem (2443)
6265 - fix mkfsoptions support for osts (2603, 2604)
6267 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6270 - add series for the vanilla 2.6.0 kernel
6271 - add series for the vanilla 2.4.24 kernel
6272 - add series for a cray x86/64 UL kernel drop
6273 - fix xattr patches for the vanilla 2.4.19 series
6275 - generate true UUIDs in lmc (1171)
6276 - have portals stack dumping break in UML (2466)
6277 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6278 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6279 - wake callback waiting threads on client eviction (2460)
6280 - Add --ptldebug and --subsystem to lmc (1719)
6281 - update assertion to allow safe interrupt allocation
6282 - set rq_no_resend for cancel requests (2432)
6283 - recalculate ptlrpcd timeout after resend (2494)
6284 - call vfs_rmdir when removing pending directories (2368)
6285 - fix renaming a file to itself (2429)
6286 - lmc creates a default one-stripe lov (2454)
6287 - expand procfs space to handle large clusters (2326)
6288 - increase UML stack to avoid overflow
6289 - update lconf's list of debug and subsystem masks
6290 - fix lfs find --obd (2510)
6291 - /proc tunable for disabling filter read caching (2591)
6292 - stop rpm packages from altering slapd.conf (2301)
6293 - disable nagle in the socknal under 0conf (2578)
6294 - choose mds inode size based on stripe count (2572)
6295 - fix kernel-source rpm problems (2516)
6297 - add --disable-doc to avoid pdf generation (2421)
6298 - update documentation, tests, type-os, comments
6299 - avoid format warnings on ia64
6300 - remove the TOE NAL
6301 - tiny code cleanups by removing unused fields
6303 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6306 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6307 - remove the most blatant lies from BUILDING file (2371)
6308 - change default debug level to reasonable production setting
6309 - reduce client side cache size to reduce cache flush time
6310 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6311 - make TCP zerocopy and pinger support enabled by default (2476)
6312 - sync writes completed after process exits caused crashes (2319)
6313 - maintain correct mount count on the MDS (2356)
6314 - backout 1557, because 2316 wasn't really fixed
6315 - better file I/O statistics gathering in /proc
6316 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6317 - another kernel patch to fix zero-copy TCP function export
6318 - don't take duplicate lock when processing re-sent getattr (2420)
6319 - lctl uses obd_self_export instead of creating new conn (2353)
6320 - MDS/OST recovery case which requires object creation asserted (2425)
6321 - move lfs from /usr/sbin to /usr/bin in packages
6322 - fix race between mds_client_add and mds_client_free (2417)
6323 - use kmalloc instead of slabs in portals (2430)
6324 - don't create duplicate records when a failover MDS is present (2442)
6325 - remove unnecessary mount age check (2332)
6326 - don't remove directory inodes from locks prematurely (2451)
6327 - don't break if MDS service name is the same as hostname (2103)
6328 - fix races in client write RPC generation when cache full (2482)
6330 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6333 - remove now-unused request->rq_obd (278)
6334 - if an allocation fails, print out how much memory we've used (1933)
6335 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6336 - add the "configurable stack size" patch to most series files (1256)
6337 - ability to write large log records, for 100+ OST configs (2306)
6338 - fix NULL deref when filter_prep fails (2314)
6339 - fix operator precedence error in filter_sync
6340 - dynamic allocation of socknal TX descriptors (2315)
6341 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6342 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6343 - ensure that CWARN messages in Portals always get to the syslog
6344 - __init/__exit are not for prototype decls (ldlm_init/exit)
6345 - x86-64 compile warning fixes
6346 - fix gateway LMC keyword conflict (2318)
6347 - fix MDS lock inversions in getattr/reint paths (1844)
6348 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6349 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6350 - DLM race condition prevented some lock evictions (2328)
6351 - ENOMEM detection and retry on socknal sends (2230)
6352 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6353 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6354 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6355 - better error messages when a client is rejected during recovery (1505)
6356 - avoid cancelling locks which were never granted, after failure (2330)
6357 - fix i_sem/journal inversion in mds_client_add (2333)
6358 - fix truncate/getattr lock cycle deadlock (2334)
6359 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6360 - fix two rare exit paths which could leak an l_lock() ref (2321)
6361 - fencepost error in MDS/OST orphan recovery (2226)
6362 - make log record alignment 8 bytes (1988)
6363 - lstripe now fails when requested offset > ost_count (2237)
6364 - ensure that all kernel series have a complete list.h (1607)
6365 - fix crashes in special-file operations (2316)
6366 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6368 - allow configurable automake binary, for testing new versions
6369 - small update to the lfs documentation
6371 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6373 * fix negative export reference count in fsfilt_sync (2312)
6375 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6376 * release candidate 0.9.1
6378 - orphans are moved into the PENDING directory for possible recovery
6379 - replayed opens now open by fid for orphan/rename safety (1042)
6380 - last close of an orphan inode generates a transno (683)
6381 - chdir() and mount() now pin the directory entry (1020)
6382 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6383 - discard very old requests without processing them (1502)
6384 - remove some common, well-understood CERRORs (1505)
6385 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6386 - clear "grant" flags in OST replies until OST grant code lands (1644)
6387 - fix read performance by not clobbering i_blksize on client (1598)
6388 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6389 - make LRU size a /proc tunable, clears locks when reduced (707)
6390 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6391 - server locks take references on exports now (1558)
6392 - build fixes for 2.4.20-rh trees (1663)
6393 - return an error from lov_create if all OSCs are inactive (1751)
6394 - fix import levels when a reconnect happens without a timeout (1597)
6395 - exit early from mds_open if we get a lookup error (1749)
6396 - partial page read at EOF wouldn't wait for disk before sending (1642)
6397 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6398 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6399 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6400 - write/truncate lock inversion (1639)
6401 - Don't auto-load obdclass, portals modules during cleanup (1495)
6402 - fix timestamps from jumping to "now" (1763)
6403 - extra journal assertions (1648)
6404 - add an extra multiunlink test (1771)
6405 - fix read_record/write_record API (1776)
6406 - fix leak of offset_extent, possible incorrect i_size later (1772)
6407 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6408 - Debugging for the kqswnal_get_idle_tx problems (1820)
6409 - Allow recovery to be attempted multiple times (1536)
6410 - Write out MDS last_rcvd file after it is first created (1600)
6411 - Fix tx_descriptor leak in failed transmit situations (1827)
6412 - ext3 journaling fixes for assertion failure after IO error (1871)
6413 - class_export_put() on freed export after completion AST error (1896)
6414 - Fix revalidate looping in VFS (1322)
6415 - Don't access a freed export during MDS_REINT timeout (1521)
6416 - Add open-unlink recovery support on the MDS (1673,1764)
6417 - Return an error if no MDS data was read from last_rcvd (1946)
6418 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6419 - Don't setup a disk device that doesn't match exported UUID (317)
6420 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6421 - avoid committing NULL handle in force close
6422 - local.sh is now a one-stripe LOV configuration
6423 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6424 - fix ext3 htree duplicate directory entry corruption (1516)
6425 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6426 - update to new LOV EA format (2097)
6427 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6429 2003-06-15 Phil Schwan <phil@clusterfs.com>
6432 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6433 - per-import/export recovery handling (958, 931, 959)
6434 - multiple last-rcvd slots, for serving multiple FSes (949)
6435 - connections are again shared between multiple imp/exports (963, 964)
6436 - "umount -f" would hang if any requests needed to be sent (393, 978)
6437 - avoid pinning large req buffer by copying for queued messages (989)
6438 - add "uuid" to "lctl device" command to help upcalls (991)
6439 - "open" RPCs with transnos would confuse recovery counters (1037)
6440 - do proper endian conversion of all wire messages (288, 340, 891)
6441 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6442 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6443 - fix ptlrpc_connection leak in target_handle_connect (1174)
6444 - fix import refcounting bug in OST and MDS cleanup (1134)
6445 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6446 - fix very unlikely obd_types race condition (501)
6447 - remove osc_open hack for echo_client (1187)
6448 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6449 - a failure in read_inode2 leads to deadlock (1139)
6450 - cancel ack-locks as soon as transaction is committed (1072)
6451 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6452 - make sure to commitrw after any preprw to avoid deadlock (1162)
6453 - failing to execute a file in a lustre FS would lock inode (1203)
6454 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6455 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6456 - obd_uuid2tgt would walk off the end of the list (1255)
6457 - on IA64 the osc would give portals incorrect bulk size (1258)
6458 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6459 - fix lock inversion caused by new llite matching code (1282)
6460 - limit the number of dirty pages on a client to 10MB (1286)
6461 - timed out locks were not being corrected cancelled (1289)
6462 - fix O_DIRECT above 4GB on IA-32 (1292)
6463 * major user-visible changes
6464 - fail out/fail over policy now controlled by the upcall (993)
6466 - add OBD_PING to check server availability and failure (954)
6467 - lustre messages are now sent in sending host order (288, 340, 891)
6468 - add eadatalen to MDS getattr reply (340)
6469 - OST read replies may contain second buffer, with per-page status (593)
6471 2003-03-11 Phil Schwan <phil@clusterfs.com>
6474 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6475 - failed open()s could cause deadlock; fixed (867, 869)
6476 - stop cancelling OST locks when files are closed (481)
6477 - overlapping XID spaces caused network corruption (851, 853)
6478 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6479 - setattr_raw added, to do single-RPC, server-side setattrs
6480 - lmc/lconf syntax change for OST UUIDs
6481 - fix crashy race condition between ptlrpc_free_req and osc_close
6482 - don't use request in mdc_enqueue if we hit a timeout (889)
6483 - don't set the inode i_size for regular files from the MDS (896)
6484 - handle out of order completion AST (842)
6485 - don't LBUG if a lock request times out after receiving AST (913)
6486 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6487 - if a bad lock AST arrives, send an error instead of dropping entirely
6488 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6489 - fix leak in bulk IO when only partially completed (899, 900, 926)
6490 - fix O_DIRECT for ia64 (55)
6491 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6492 - C-z after timeout could hang a process forever; fixed (977)
6494 - client-side I/O cache (678, 924, 929, 941, 970)
6496 - READPAGE and SETATTRs which don't take server-side locks get
6499 2003-02-11 Phil Schwan <phil@clusterfs.com>
6502 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6503 - Fix fsfilt_extN_readpage() to read a full page of directory
6504 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6505 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6506 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6507 - Most intent operations were removed, in favour of a new RPC mode
6508 that does a single RPC to the server and bypasses most of the VFS
6509 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6510 - Aggressively cancel local locks on DLM servers
6511 - mds_reint_unlink sends EA to the client if it's the last nlink.
6512 client uses that EA to unlink OST objects.
6513 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6514 - recursive symlinks were fixed (439)
6515 - fixed NULL deref in DEBUG_REQ
6516 - filter_update_lastobjid no longer calls sync, which annoyed extN
6517 - fixed multi-client small-writes to a single file problem (445)
6518 - fixed mtime updates during file writes (607)
6519 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6520 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6521 - fixed Linux OST size reporting problem (444, 656)
6522 - OST now updates object mtime with writes or setattr (607, 619)
6523 - client verifies file size before zeroing page past EOF (445)
6524 - OST now writes last allocated objid to disk with allocation (108)
6525 - LOV on echo now works (409)
6527 - mds_reint_unlink sends a new buffer, with the EA included. this
6528 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6529 set if a regular file was being unlinked, and it was the last link
6530 - use PtlGet from the target for bulk writes (315)
6531 - OST now updates object mtime with writes or setattr (607, 619)
6532 - LDLM now has a grant-time callback to revalidate locked items, if
6534 - Many MDS operations were reorganized to combat race conditions
6536 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6537 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6539 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6542 - Fully reactivate OST imports after reconnection (512, others)
6543 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6544 - More graceful error handling for truncating on dead OST (515)
6545 - Don't error out unless we're actually accessing dead stripes (474)
6546 - Fix garbage sizes when stripes are missing (410)
6547 - LRU counters were broken, causing constant lock purge (433, 432)
6548 - garbage on read from stripes with failed OSTs (441)
6549 - mark OSCs as active before reconnecting during recovery (438)
6550 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6551 - lfind did not preserve OST order in output (443)
6552 - symlinks cause hung clients, incorrect data (439)
6553 - stop dereferencing request after dropping refcount (457)
6554 - don't LASSERT(spin_is_locked) on non-SMP (455)
6555 - fixes for many rename() bugs
6556 - fstat didn't correctly synchronize attributes (399)
6557 - server must handle lock cancellation during blocking AST prep (487)
6558 - bulk descriptors were free()d too soon (511)
6559 - fix paths in lconf, which would load incorrect modules (451, 507)
6560 - fix confusing lconf 'host not found' error message (386)
6561 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6562 - fix race condition in mdc_blocking_ast() for inode access (526)
6563 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6564 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6565 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6566 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6567 - fix lstripe in several ways (kernel side) (527)
6568 - fix request leak in ldlm_cli_enqueue (262)
6569 - incorrect OSC was marked inactive after OST failure
6570 - call mds_fs_cleanup before unmounting filesystem (524)
6571 - fix races between taking ns_lock and ldlm_lock_change_resource
6572 - fix races updating LOV export open file list
6573 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6574 - fix recovery NULL deref in ldlm_cli_cancel_unused
6575 - fix some DLM races by using new hash table for lock handles (419)
6576 - permit the client to specify desired inodes, at replay
6577 - duplicate requests when we queue them for replay reintegration
6578 - fix last_rcvd offset calculation
6579 - sync after each recovered transaction, so we always make progress
6580 - never, not always, ERESTART requests without transnos
6581 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6583 - skip replay if the MDS says that the client is already connected
6584 - don't check for a recovery-enabled export to match lctl's UUID
6585 - don't INC_USE_COUNT for phantom exports
6586 - don't crash when cleaning up phantom exports (567)
6587 - don't double-finish or set replay data for errored mdc_open requests
6588 - abort requests when they time out, so we don't get old replies
6589 - send/receive replies for AST messages again
6590 - if the client says that it doesn't have the lock, cancel it on the
6592 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6593 mark it as destroyed, it will all work out when decref is called
6594 - fix module use counts (22, 581)
6596 - ASTs now expect a reply (server cancels lock on error reply)
6598 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6601 - fix many simultaneous client startup (392)
6602 - fix dentry->d_it clobbering
6603 - credentials weren't being shipped for readdir/getattr operations
6604 - remove invalid assertions triggered during some concurrent MD
6606 - proper Lustre versions added (336, 389)
6607 - fix memory leak for create error case (398)
6608 - fix LOV locking bug that would get cli/srv out of sync
6609 - fix echo client over LOV (409)
6610 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6611 - fix double-O_EXCL intent crash (424)
6612 - avoid sending multiple lock CANCELs (352)
6614 - MDS can do multi-client recovery (modulo bugs in new code)
6616 - many updates, edits, cleanups
6618 2002-11-18 Phil Schwan <phil@clusterfs.com>
6621 - fix null d_it dereference (346)
6622 - fix full OST/dbench hang (333)
6623 - fix permission problem with file removal (286)
6624 - fix removal of OSCs from LOV when they fail
6625 - fix NULL deref during bulk timeout (214)
6626 - fix problems related to multiple filesystems on one MDS (241)
6627 - fixed serious subtle metadata locking bugs
6628 - free locks on clients when inodes are removed due to memory
6630 - fix inode pointer in lock data (285)
6631 - partial support for multiple MDS on a single host (241)
6632 - data locks weren't cancelled at clear_inode time (290, 311)
6633 - intent locks could lead to unbounded lock growth (205)
6634 - added a maximum lock count, an LRU list, and a flusher
6635 - fix multiple rename (365)
6636 - properly abstracted the echo client
6637 - OSC locked 1 byte too many; fixed
6638 - rewrote brw callback code:
6639 - fixed recovery bugs related to LOVs (306)
6640 - fixed too-many-pages-in-one-write crash (191)
6641 - fixed (again) crash in sync_io_timeout (214)
6642 - probably fixed callback-related race (385)
6644 - Add capability to MDS protocol
6645 - LDLM cancellations and callbacks on different portals
6647 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6650 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6651 - defer open object destruction to close time (601981, 138)
6652 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6653 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6654 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6655 - don't call obd_setattr in ll_file_release for destroyed objects
6656 * protocol change to lustre_msg: move |version| and add |flags|
6657 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6658 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6659 * added replay of create, unlink, link and rename operations during
6660 MDS failover; recovery should be much more robust now
6661 * remove failed OSCs from LOVs (only lov_create uses this so far)
6662 * the lustre-HOWTO was brought (more) up to date (582544)
6664 2002-10-23 Phil Schwan <phil@clusterfs.com>
6667 - in-use dentries weren't being reused properly (617851)
6668 - prevent multiple LDLM setup (599178)
6669 - fix LOV size calculations for truncate (617853)
6670 - fix client handling of MDS intent errors (POSIX)
6671 - fix permission bug in lovstripe.c test (624321)
6672 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6673 - truncate past end of file could corrupt data
6674 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6675 - a race in recovery could return ETIMEDOUT to apps (623947)
6676 - building outside the source directory was fixed
6677 * the lustre-HOWTO was brought (more) up to date (582544)
6678 * major progress was made on recovery functionality
6680 2002-10-10 Phil Schwan <phil@clusterfs.com>
6683 - recovery deadlock fix
6684 - rm -rf causes LBUG fix (617817)
6685 - file open by multiple tasks fix (618962)
6686 - directory permissions bugs (602707 and 620007)
6687 - journal_stop fixed with locking (611313)
6688 - O_APPEND failures resolved (618273, perhaps 614459)
6689 - lconf PATH fix (619770)
6690 - IA64 build fix (621450)
6691 - RPC buffer sizes scale with amount of memory
6693 2002-10-01 Phil Schwan <phil@clusterfs.com>
6696 - locks would be cancelled without throwing away data pages,
6697 resulting in inconsistent data (605627)
6698 - inode attributes were not always being refreshed (605627, 612449)
6699 - lconf now continues to cleanup after lctl reports an error
6700 - MDS now enforces user permissions (602707)
6701 - lprocfs cleanup fixed, but not yet enabled (614157)
6702 - fixed infinite server hang, should a client not respond to an AST
6703 - avoid going into recovery if user calls readlink() with a buffer
6704 that's too small (613941)
6705 - AST RPCs no longer require replies (614867) -- this may be changed
6706 - don't crash server if client sends an IOV that's too big (611336)
6707 - fixed lock conversion deadlock (611892)
6708 - fixed the following of symlinks (614622)
6709 * recovery: the server can remove locks from a client that dies, other
6710 clients can make progress
6711 * more extN patch fixes
6712 * compile-time configurable ptlrpc buffer allocations
6714 - collaborative read cache document
6715 - Lustre Lite Performance CDR document-in-progress
6717 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6720 - fix typo in patch-2.4.18
6722 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6725 - clear ptlrpc request each time in handle_incoming_request()
6726 - unlink of files now destroys the object on the OST
6728 2002-09-19 Peter Braam <braam@clusterfs.com>
6730 * add hard link support
6731 * change obdfile creation method
6732 * kernel patch changed
6734 2002-09-19 Peter Braam <braam@clusterfs.com>
6737 - stack overflow bug in extN fixed
6739 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6741 * documentation updates
6742 - add man pages for config tools
6743 - update tests/README to describe testing with new config tools
6744 - finish metadata API descriptions
6745 * bug fixes and cleanups
6746 - statfs workaround for 16TB limit
6747 - LOV stripe allocation improved, can stripe on subset of OSTs
6748 - LOV file size/IO offset was wrong for files > 4GB in size
6749 - object EA data was being dropped, caused files to be unreadable
6750 - memory overflow with non-LOV OST caused memory corruption
6751 - fixed regression tests to work with new config tools, obdfilter
6752 - fixed bug when directory size became larger than 1 block
6753 - fixed bug (for single client case) when PWD was deleted
6754 - invalidate local directory pages when doing intent-based ops
6755 - avoid LDLM oops when lock callback contained bad data
6757 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6759 * documentation updates
6760 * bug fixes and cleanups
6761 - configuration tools
6764 - 64-bit compile warnings
6765 - 64-bit internal statfs data
6767 * test_brw on persistent OST devices
6769 * lprocfs (disabled)
6771 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6773 * documentation updates
6774 * bug fixes and cleanups
6775 * configuration tools
6777 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6780 * many small fixes to 0_5_4
6781 * io/network handling
6782 * thinkos in MDS operations
6784 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6787 * crucial basic fixes to 0.5.3
6788 * IOR, Iozone work over Elan
6791 2002-08-07 Phil Schwan <phil@clusterfs.com>
6792 * version 0_5_3, our first alpha
6793 * we use the new Portals iovs
6794 * documentation updates
6795 * bug fixes and cleanups
6796 * small changes in the DLM wire protocol
6798 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6799 * version 0_5_1 with some initial stability,
6800 * locking on MD and file I/O.
6801 * documentation updates
6802 * several bug fixes since 0.5.0
6803 * small changes in wire protocol
6805 2002-07-18 Phil Schwan <phil@clusterfs.com>
6807 * delivered as Lustre Light Alpha
6808 * fixed a crash after handling invalid MDS requests
6809 * fixed directory pages for architectures with non-4k pages sizes
6811 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6812 * release version v0_4_4
6813 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6815 * Fixes a number of interruption problems with OST operations.
6816 * Update documentation for portals header changes
6817 * Move all wire protocol structs/defines to lustre_idl.h
6818 * Fixes symlink length bug.
6819 * Add tcpdump to repository.
6821 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6822 * release version v0_4_3
6823 * Fixes statfs for inodes on extN.
6824 * Fixes bug in runtests which would delete /etc/hosts.
6825 * Use 64-bit object IDs wherever possible (not into VFS though)
6826 Remove ost_get_info, which is unused by lustre, and out of date.
6828 2002-07-03 Peter Braam <braam@clusterfs.com>
6829 * release version v0_4_2 Fixes a lookup error (type not passed)
6830 * move forward to head of Portals
6831 * move forward to latest Lustre kernel
6833 2002-06-25 Peter Braam <braam@clusterfs.com>
6834 * release version v0_4_1. Hopefully stable on single node use.