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 4.4.120-92.70 (SLES12 SP2)
31 4.4.0-142 (Ubuntu 16.04)
32 vanilla linux 4.15.0 (ZFS only)
33 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
34 * Recommended ZFS / SPL version: 0.7.13
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
40 * Note for Ubuntu 14.04.5:
41 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
42 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
46 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
47 for currently supported client and server kernel versions.
48 * Server primary kernels built and tested during release cycle:
49 3.10.0-957.27.2.el7 (RHEL7.6)
50 3.10.0-1062.1.1.el7 (RHEL7.7)
51 * Server known to build on the following kernels (others may also work):
52 2.6.32-696.18.7.el6 (RHEL6.9)
53 3.10.0-862.14.4.el7 (RHEL7.5)
54 4.4.120-92.70 (SLES12 SP2)
55 4.4.180-94.97 (SLES12 SP3)
56 4.4.0-142 (Ubuntu 16.04)
57 4.15.0-45 (Ubuntu 18.04)
58 vanilla linux 4.6.7 (ZFS only)
59 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
60 * Client primary unpatched kernels built and tested during release cycle:
61 3.10.0-957.27.2.el7 (RHEL7.6)
62 3.10.0-1062.1.1.el7 (RHEL7.7)
63 4.18.0-80.7.1.el8 (RHEL8.0)
64 4.4.180-94.97 (SLES12 SP3)
65 4.12.14-95.29 (SLES12 SP4)
66 4.15.0-45 (Ubuntu 18.04)
67 * Client known to build on unpatched kernels (others may also work):
68 2.6.32-696.18.7.el6 (RHEL6.9)
69 2.6.32-754.6.3.el6 (RHEL6.10)
70 3.10.0-862.14.4.el7 (RHEL7.5)
71 4.14.0-49.13.1.el7a (RHEL7.5)
72 4.14.0-115.2.2.el7a (RHEL7.6)
73 4.4.120-92.70 (SLES12 SP2)
74 4.4.0-142 (Ubuntu 16.04)
75 vanilla linux 4.15.0 (ZFS only)
76 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
77 * Recommended ZFS / SPL version: 0.7.13
78 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
79 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
80 stack overflow. For more information, please refer to bugzilla 17630.
81 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
82 the re-exporting nodes to be mounted with "32bitapi" mount option
83 * Note for Ubuntu 14.04.5:
84 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
85 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
89 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
90 for currently supported client and server kernel versions.
91 * Server known to build on the following kernels (others may also work):
92 2.6.32-696.18.7.el6 (RHEL6.9)
93 3.10.0-862.14.4.el7 (RHEL7.5)
94 3.10.0-957.10.1.el7 (RHEL7.6)
95 4.4.120-92.70 (SLES12 SP2)
96 4.4.162-94.69 (SLES12 SP3)
97 4.4.0-142 (Ubuntu 16.04)
98 4.15.0-45 (Ubuntu 18.04)
99 vanilla linux 4.6.7 (ZFS only)
100 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
101 * Client known to build on unpatched kernels (others may also work):
102 2.6.32-696.18.7.el6 (RHEL6.9)
103 2.6.32-754.6.3.el6 (RHEL6.10)
104 3.10.0-862.14.4.el7 (RHEL7.5)
105 3.10.0-957.10.1.el7 (RHEL7.6)
106 4.14.0-49.13.1.el7a (RHEL7.5)
107 4.14.0-115.2.2.el7a (RHEL7.6)
108 4.4.120-92.70 (SLES12 SP2)
109 4.4.162-94.69 (SLES12 SP3)
110 4.12.14-95.13 (SLES12 SP4)
111 4.4.0-142 (Ubuntu 16.04)
112 4.15.0-45 (Ubuntu 18.04)
113 vanilla linux 4.15.0 (ZFS only)
114 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
115 * Recommended ZFS / SPL version: 0.7.13
116 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
117 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
118 stack overflow. For more information, please refer to bugzilla 17630.
119 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
120 the re-exporting nodes to be mounted with "32bitapi" mount option
121 * Note for Ubuntu 14.04.5:
122 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
123 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
128 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
129 for currently supported client and server kernel versions.
130 * Server known to build on the following kernels (others may also work):
131 2.6.32-696.18.7.el6 (RHEL6.9)
132 3.10.0-862.14.4.el7 (RHEL7.5)
133 3.10.0-957.10.1.el7 (RHEL7.6)
134 4.4.120-92.70 (SLES12 SP2)
135 4.4.162-94.69 (SLES12 SP3)
136 4.4.0-142 (Ubuntu 16.04)
137 4.15.0-45 (Ubuntu 18.04)
138 vanilla linux 4.6.7 (ZFS only)
139 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
140 * Client known to build on unpatched kernels (others may also work):
141 2.6.32-696.18.7.el6 (RHEL6.9)
142 2.6.32-754.6.3.el6 (RHEL6.10)
143 3.10.0-862.14.4.el7 (RHEL7.5)
144 3.10.0-957.10.1.el7 (RHEL7.6)
145 4.14.0-49.13.1.el7a (RHEL7.5)
146 4.14.0-115.2.2.el7a (RHEL7.6)
147 4.4.120-92.70 (SLES12 SP2)
148 4.4.162-94.69 (SLES12 SP3)
149 4.12.14-95.13 (SLES12 SP4)
150 4.4.0-142 (Ubuntu 16.04)
151 4.15.0-45 (Ubuntu 18.04)
152 vanilla linux 4.15.0 (ZFS only)
153 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
154 * Recommended ZFS / SPL version: 0.7.13
155 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
156 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
157 stack overflow. For more information, please refer to bugzilla 17630.
158 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
159 the re-exporting nodes to be mounted with "32bitapi" mount option
160 * Note for Ubuntu 14.04.5:
161 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
162 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
166 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
167 for currently supported client and server kernel versions.
168 * Server known to build on the following kernels (others may also work):
169 2.6.32-696.18.7.el6 (RHEL6.9)
170 3.10.0-862.14.4.el7 (RHEL7.5)
171 3.10.0-957.el7 (RHEL7.6)
172 4.4.120-92.70 (SLES12 SP2)
173 4.4.162-94.69 (SLES12 SP3)
174 4.4.0-131 (Ubuntu 16.04)
175 4.15.0-32 (Ubuntu 18.04)
176 vanilla linux 4.6.7 (ZFS only)
177 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
178 * Client known to build on unpatched kernels (others may also work):
179 2.6.32-696.18.7.el6 (RHEL6.9)
180 2.6.32-754.6.3.el6 (RHEL6.10)
181 3.10.0-862.14.4.el7 (RHEL7.5)
182 3.10.0-957.el7 (RHEL7.6)
183 4.14.0-49.13.1.el7a (RHEL7.5)
184 4.14.0-115.2.2.el7a (RHEL7.6)
185 4.4.120-92.70 (SLES12 SP2)
186 4.4.162-94.69 (SLES12 SP3)
187 4.4.0-131 (Ubuntu 16.04)
188 4.15.0-32 (Ubuntu 18.04)
189 vanilla linux 4.15.0 (ZFS only)
190 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
191 * Recommended ZFS / SPL version: 0.7.9
192 * Tested with ZFS / SPL version: 0.7.9
193 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
194 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
195 stack overflow. For more information, please refer to bugzilla 17630.
196 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
197 the re-exporting nodes to be mounted with "32bitapi" mount option
198 * Note for Ubuntu 14.04.5:
199 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
200 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
202 --------------------------------------------------------------------------------
205 2018-04-03 Intel Corporation
207 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
208 for currently supported client and server kernel versions.
209 * Server known to build on patched kernels:
210 2.6.32-431.29.2.el6 (RHEL6.5)
211 2.6.32-504.30.3.el6 (RHEL6.6)
212 2.6.32-573.26.1.el6 (RHEL6.7)
213 2.6.32-642.15.1.el6 (RHEL6.8)
214 2.6.32-696.18.7.el6 (RHEL6.9)
215 3.10.0-693.21.1.el7 (RHEL7.4)
216 3.0.101-0.47.71 (SLES11 SP3)
217 3.0.101-107 (SLES11 SP4)
218 3.12.74-60.64.40 (SLES12 SP1)
219 4.4.114-92.64 (SLES12 SP2)
220 4.4.114-94.11 (SLES12 SP3)
221 3.13.0-101 (Ubuntu 14.04, ZFS only)
222 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
223 4.4.0-116 (Ubuntu 16.04)
224 vanilla linux 4.6.7 (ZFS only)
225 * Client known to build on unpatched kernels:
226 2.6.32-431.29.2.el6 (RHEL6.5)
227 2.6.32-504.30.3.el6 (RHEL6.6)
228 2.6.32-573.26.1.el6 (RHEL6.7)
229 2.6.32-642.15.1.el6 (RHEL6.8)
230 2.6.32-696.18.7.el6 (RHEL6.9)
231 3.10.0-693.21.1.el7 (RHEL7.4)
232 3.0.101-0.47.71 (SLES11 SP3)
233 3.0.101-107 (SLES11 SP4)
234 3.12.74-60.64.40 (SLES12 SP1)
235 4.4.114-92.64 (SLES12 SP2)
236 4.4.114-94.11 (SLES12 SP3)
237 3.13.0-101 (Ubuntu 14.04)
238 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
239 4.4.0-116 (Ubuntu 16.04)
241 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
242 * Recommended ZFS / SPL version: 0.7.6
243 * Tested with ZFS / SPL version: 0.7.6
244 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
245 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
246 stack overflow. For more information, please refer to bugzilla 17630.
247 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
248 the re-exporting nodes to be mounted with "32bitapi" mount option
249 * Note for Ubuntu 14.04.5:
250 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
251 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
253 --------------------------------------------------------------------------------
255 2017-07-05 Intel Corporation
257 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
258 for currently supported client and server kernel versions.
259 * Server known to build on patched kernels:
260 2.6.32-431.29.2.el6 (RHEL6.5)
261 2.6.32-504.30.3.el6 (RHEL6.6)
262 2.6.32-573.26.1.el6 (RHEL6.7)
263 2.6.32-642.15.1.el6 (RHEL6.8)
264 2.6.32-696.3.1.el6 (RHEL6.9)
265 3.10.0-514.21.1.el7 (RHEL7.3)
266 3.0.101-0.47.71 (SLES11 SP3)
267 3.0.101-100 (SLES11 SP4)
268 3.12.74-60.64.40 (SLES12 SP1)
269 4.4.59-92.17 (SLES12 SP2)
270 vanilla linux 4.6.7 (ZFS only)
271 * Client known to build on unpatched kernels:
272 2.6.32-431.29.2.el6 (RHEL6.5)
273 2.6.32-504.30.3.el6 (RHEL6.6)
274 2.6.32-573.26.1.el6 (RHEL6.7)
275 2.6.32-642.15.1.el6 (RHEL6.8)
276 2.6.32-696.3.1.el6 (RHEL6.9)
277 3.10.0-514.21.1.el7 (RHEL7.3)
278 3.0.101-0.47.71 (SLES11 SP3)
279 3.0.101-100 (SLES11 SP4)
280 3.12.74-60.64.40 (SLES12 SP1)
281 4.4.59-92.17 (SLES12 SP2)
283 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
284 * Recommended ZFS / SPL version: 0.6.5.9
285 * Tested with ZFS / SPL version: 0.6.5.9
286 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
287 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
288 stack overflow. For more information, please refer to bugzilla 17630.
289 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
290 the re-exporting nodes to be mounted with "32bitapi" mount option
292 --------------------------------------------------------------------------------
294 12-07-2016 Intel Corporation
296 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
297 for currently supported client and server kernel versions.
298 * Server known to build on patched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.30.3.el6 (RHEL6.6)
301 2.6.32-573.26.1.el6 (RHEL6.7)
302 2.6.32-642.11.1.el6 (RHEL6.8)
303 3.10.0-514.el7 (RHEL7.3)
304 3.0.101-0.47.71 (SLES11 SP3)
305 3.0.101-80 (SLES11 SP4)
306 3.12.62-60.64.8 (SLES12 SP1)
307 vanilla linux 4.5.7 (ZFS only)
308 * Client known to build on unpatched kernels:
309 2.6.32-431.29.2.el6 (RHEL6.5)
310 2.6.32-504.30.3.el6 (RHEL6.6)
311 2.6.32-573.26.1.el6 (RHEL6.7)
312 2.6.32-642.11.1.el6 (RHEL6.8)
313 3.10.0-514.el7 (RHEL7.3)
314 3.0.101-0.47.71 (SLES11 SP3)
315 3.0.101-80 (SLES11 SP4)
316 3.12.62-60.64.8 (SLES12 SP1)
318 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
319 * Recommended ZFS / SPL version: 0.6.5.8
320 * Tested with ZFS / SPL version: 0.6.5.8
321 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
322 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
323 stack overflow. For more information, please refer to bugzilla 17630.
324 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
325 the re-exporting nodes to be mounted with "32bitapi" mount option
327 --------------------------------------------------------------------------------
329 02-29-2016 Intel Corporation
331 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
332 for currently supported client and server kernel versions.
333 * Server known to build on patched kernels:
334 2.6.32-431.29.2.el6 (RHEL6.5)
335 2.6.32-504.30.3.el6 (RHEL6.6)
336 2.6.32-573.12.1.el6 (RHEL6.7)
337 3.10.0-327.3.1.el7 (RHEL7.2)
338 3.0.101-0.47.71 (SLES11 SP3)
339 3.0.101-68 (SLES11 SP4)
340 vanilla linux 4.2.1 (ZFS only)
341 * Client known to build on unpatched kernels:
342 2.6.32-431.29.2.el6 (RHEL6.5)
343 2.6.32-504.30.3.el6 (RHEL6.6)
344 2.6.32-573.12.1.el6 (RHEL6.7)
345 3.10.0-327.3.1.el7 (RHEL7.2)
346 3.0.101-0.47.71 (SLES11 SP3)
347 3.0.101-68 (SLES11 SP4)
350 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
351 * Recommended ZFS / SPL version: 0.6.4.2
352 * Tested with ZFS / SPL version: 0.6.4.2
353 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
354 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
355 stack overflow. For more information, please refer to bugzilla 17630.
356 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
357 the re-exporting nodes to be mounted with "32bitapi" mount option
359 --------------------------------------------------------------------------------
361 03-10-2015 Intel Corporation
363 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
364 for currently supported client and server kernel versions.
365 * Server known to build on patched kernels:
366 2.6.32-431.29.2.el6 (RHEL6.5)
367 2.6.32-504.8.1.el6 (RHEL6.6)
368 3.0.101-0.46 (SLES11 SP3)
369 * Client known to build on unpatched kernels:
370 2.6.32-431.29.2.el6 (RHEL6.5)
371 2.6.32-504.8.1.el6 (RHEL6.6)
372 3.10.0-123.20.1.el7 (RHEL7)
373 3.0.101-0.46 (SLES11 SP3)
374 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
375 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
376 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
377 stack overflow. For more information, please refer to bugzilla 17630.
378 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
379 the re-exporting nodes to be mounted with "32bitapi" mount option
381 Severity : enhancement
383 Description: control OST-index in IDIF via ROCOMPAT flag.
384 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
385 last_rcvd file. For new formatted OST device, it will be auto set;
386 for the case of upgrading from old OST device, you can enable it
387 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
388 enabled, for new created OST-object, its IDIF-in-LMA will contain
389 the OST-index; for the existing OST-object, the OSD will convert
390 old format IDIF as new format IDIF with OST-index stored in the
391 LMA EA when accessing such OST-object or via OI scrub. Once such
392 flag is enabled, it cannot be reverted back, so the system cannot
393 be downgraded to the orignal incompatible version.
395 --------------------------------------------------------------------------------
397 07-30-2014 Intel Corporation
399 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
400 for currently supported client and server kernel versions.
401 * Server known to build on patched kernels:
402 2.6.32-431.20.3.el6 (RHEL6.5)
403 3.0.101-0.31 (SLES11 SP3)
404 * Client known to build on unpatched kernels:
405 2.6.32-431.20.3.el6 (RHEL6.5)
406 3.0.101-0.31 (SLES11 SP3)
409 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
410 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
411 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
412 stack overflow. For more information, please refer to bugzilla 17630.
413 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
414 the re-exporting nodes to be mounted with "32bitapi" mount option
416 --------------------------------------------------------------------------------
418 10-31-2013 Intel Corporation
420 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
421 * Server builds for kernels:
422 2.6.32-358.18.1.el6 (RHEL6.4)
423 * Client builds for unpatched kernels:
424 2.6.32-358.18.1.el6 (RHEL6.4)
425 3.0.92-0.8 (SLES11 SP3)
427 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
428 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
429 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
430 stack overflow. For more information, please refer to bugzilla 17630.
431 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
432 the re-exporting nodes to be mounted with "32bitapi" mount option
434 --------------------------------------------------------------------------------
436 05-30-2013 Intel Corporation
438 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
439 * Server builds for kernels:
440 2.6.32-358.6.2.el6 (RHEL6.4)
441 * Client builds for unpatched kernels:
442 2.6.32-358.6.2.el6 (RHEL6.4)
443 3.0.74-0.6.6 (SLES11 SP2)
445 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
446 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
447 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
448 stack overflow. For more information, please refer to bugzilla 17630.
449 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
450 the re-exporting nodes to be mounted with "32bitapi" mount option
452 --------------------------------------------------------------------------------
454 9-30-2012 Whamcloud, Inc.
456 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
457 * Server support for kernels:
458 2.6.32-279.5.1.el6 (RHEL6)
459 * Client support for unpatched kernels:
460 2.6.32-279.5.1.el6 (RHEL6)
462 2.6.18-238.19.1.el5 (RHEL5)
463 2.6.16.60-0.69.1 (SLES10)
464 * Recommended e2fsprogs version: 1.41.90.wc4
465 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
466 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
467 stack overflow. For more information, please refer to bugzilla 17630.
468 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
469 the re-exporting nodes to be mounted with "32bitapi" mount option
471 Severity : enhancement
473 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
475 Severity : enhancement
477 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
479 Severity : enhancement
481 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
483 Severity : enhancement
485 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
487 Severity : enhancement
489 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
491 --------------------------------------------------------------------------------
493 04-30-2012 Whamcloud, Inc.
495 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
496 * Server support for kernels:
497 2.6.32-220.4.2.el6 (RHEL6)
498 * Client support for unpatched kernels:
499 2.6.32-220.4.2.el6 (RHEL6)
501 2.6.18-238.19.1.el5 (RHEL5)
502 2.6.16.60-0.69.1 (SLES10)
503 * Recommended e2fsprogs version: 1.41.90.wc4
504 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
505 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
506 stack overflow. For more information, please refer to bugzilla 17630.
507 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
508 the re-exporting nodes to be mounted with "32bitapi" mount option
510 Severity : enhancement
512 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
514 Severity : enhancement
516 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
518 Severity : enhancement
520 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
522 Severity : enhancement
524 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
526 --------------------------------------------------------------------------------
528 09-30-2011 Whamcloud, Inc.
530 * Server support for kernels:
531 2.6.18-238.19.1.el5 (RHEL5)
532 2.6.32-131.6.1.el6 (RHEL6)
533 * Client support for unpatched kernels:
534 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
535 2.6.18-238.19.1.el5 (RHEL5)
536 2.6.32-131.6.1.el6 (RHEL6)
537 2.6.29.4-167.fc11 (FC11)
538 2.6.16.60-0.69.1 (SLES10)
540 * Recommended e2fsprogs version: 1.41.90.wc3
541 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
542 Lustre filesystem with 4K stack may cause a stack overflow. For more
543 information, please refer to bugzilla 17630.
544 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
545 mounted with "32bitapi" mount option
547 Severity : enhancement
548 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
550 Severity : enhancement
551 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
553 Severity : enhancement
554 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
557 Frequency : only with SLES10
559 Description: Use OFED "KMP" provided by Novell
560 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
561 Lustre is now built against this package. That means you need to
562 install the ofed-kmp package from Novell for the patchless client.
563 Note that the ofed-kmp that Novell ships may not exactly match the
564 kernel version but should still be compatible.
566 Severity : enhancement
568 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
569 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
571 --------------------------------------------------------------------------------
573 2010-07-15 Oracle, Inc.
575 * Server support for kernels:
576 2.6.18-164.11.1.el5 (RHEL 5)
577 2.6.18-164.11.1.0.1.el5 (OEL 5)
578 * Client support for unpatched kernels:
579 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
580 2.6.18-164.11.1.el5 (RHEL 5),
581 2.6.18-164.11.1.0.1.el5 (OEL 5)
582 2.6.16.60-0.42.8 (SLES 10),
584 2.6.29.4-167.fc11 (FC11)
585 * Recommended e2fsprogs version: 1.41.10-sun2
587 Severity : enhancement
589 Description: Distribute lfs_migrate script with lustre
590 Details : The lfs_migrate mentioned in the manual is now included in the
591 lustre tree in lustre/scripts and will distributed in lustre RPMs.
593 Severity : enhancement
595 Description: Failover nid list added to import proc file.
597 Severity : enhancement
599 Description: "som_preview" mount option required to use SOM
600 Details : The clients will need to use som_preview mount option
604 Description: Lustre does not do 1MB IOs to HW RAID
605 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
607 Severity : enhancement
609 Description: lfs quota output is not very convenient for awk/sed-parsing
610 Details : Some positions in lfs quota output table could be empty or
611 non-empty which made it hard to parse it with scripts, now
612 a dash is put instead of space where there is not supposed
617 Description: Update to OFED 1.5.1
618 Details : OFED 1.5.1 will be used in environments that don't use
619 vendor supplied OFED.
623 Description: Remove lnet. prefix from lctl params display
624 Details : Also renamed lnet.memused to lnet_memused to
629 Description: Use conf_param -d to remove permanant settings.
633 Description: Add list_param -R
634 Details : lctl list_param now supports recursive mode.
637 Bugzilla : 21829, 22312
638 Description: Fix llobdstat.
639 Details : 1. Now updating the write_bytes counter
640 2. Port script fixes from 1.8.
644 Description: decrease the usage of memory on clients.
645 Details : 1. On clients, recycle dentries and inodes unused.
646 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
651 Description: llobdstat fix and enhancement
652 Details : add a counter to set a limit to how many samples will be returned
653 fix a wildcard in the path to limit to obdfilter stats only
657 Description: lfs check only allowed for root
658 Details : Code cleanups and added a regresion test of non-root use of
663 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
664 Details : Fix hr->hr_index overload.
668 Description: Fix FIEMAP functionality.
672 Description: FC11 patchless client support added
676 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
677 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
681 Description: New acc-sm script and various test-framework enhancments.
685 Description: Previously landed patch was reverted.
689 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
693 Description: Add ldiskfs maxdirsize mount option.
697 Description: Fix mkfs.lustre support for 16TB LUNs.
701 Description: Print mmp_check_interval
702 Details : print mmp_check_interval and make it possible to abort
703 mount operation in case it takes too long.
707 Description: Add cascading_rw to lustre mpi tests.
711 Description: Store SOM data in EA.
712 Details : Also several SOM related cleanups.
716 Description: Add refcount to md_open_data
720 Description: Use refcount on lov_request_set to prevent memory corruption.
724 Description: Statahead: use parent dir's i_mutex for synchronization.
728 Description: loadgen improvements.
732 Description: Enable peer health detection in ksocklnd and ko2iblnd.
736 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
740 Description: Fix race on obd_devcie:md_stats
744 Description: Fix SLES 11 client builds.
745 Details : Resolved uninitalized pointer gcc warnings.
749 Description: Allow non-root access for "lfs check".
750 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
754 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
756 Severity : enhancement
758 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
759 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
763 Description: Sanity test_27p fails.
764 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
765 Sanity tests 27 were cleaned up and improved.
769 Description: Unify req format on client/servers
770 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
771 buffers and swabbers approach..
776 Description: drop join file stuff.
778 Severity : enhancement
780 Description: Add 2.6.27 and SLES11 (patchless client) support
784 Description: Fix in ptlrpc_expire_one_request() to print the signed time
787 Severity : enhancement
789 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
792 Severity : enhancement
794 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
796 Severity : enhancement
798 Description: correctly handle big reply message.
799 Details : send LNet event if reply is bigger than buffer and adjust this buffer
804 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
806 Severity : enhancement
808 Description: Add support for OEL5.
812 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
814 Severity : enhancement
816 Description: "weak-modules" support
817 Details : Implement "weak-modules" support which enables kernel modules
818 to be used with any kernel that implements the same kABI. In
819 order to achieve this modules are now installed in
820 /lib/modules/$(uname -r)/updates/kernel on all distributions.
824 Description: client could not reconnect to OST because of an active request.
825 Details : The client now retries to reconnect to the same server, if a
826 connect request failed with EBUSY or -EAGAIN. Test case 35b
827 conf-sanity.sh has been created.
831 Description: Don't allow make backward step on assignin osc next id.
832 Details : race between allocation next id and ll_sync thread can be cause
833 of set wrong osc next id and can be kill valid ost objects.
837 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
841 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
845 Description: Error handling in osc_statfs_interpret() has been improved.
846 Details : Check in osc_statfs_interpret() for EBADR.
850 Description: Do not send statfs() requests to OSTs disabled by administrator.
851 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
853 Severity : enhancement
855 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
857 Severity : enhancement
859 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
861 Severity : enhancement
863 Description: Enhancement for lfs(1) command to use numeric uid/gid.
867 Description: Improved cleanup procedure for conf-sanity test_46a.
871 Description: Do not start bulk transfer, if client tried to reconnect.
872 Details : Do not start bulk transfer, if client tried to reconnect,
873 but failed due to active rpcs. Check for exp_abort_active_req
874 has been implemented.
878 Description: Prevent inconsistences between linux and lustre mount structures.
879 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
880 Make the sleep interruptible.
882 Severity : enhancement
884 Description: provide server to client comms path
885 Details : server to client communications path via new LDLM_SET_INFO rpc
889 Description: truncate starts GFP_FS allocation under transaction causing deadlock
890 Details : ldiskfs_truncate calls grab_cache_page which may start page
891 allocation under an open transaction. This may lead to
892 calling prune_icache with consequent lustre reentrance.
896 Description: send correctly lsm on open replay
897 Details : MDS is trust to LSM size on replay open, but client can set wrong size
902 Description: enable client interface failover
903 Details : When a child reconnects from another NID, properly update export
904 nid hash position and ldlm reverse import.
906 Severity : enhancement
908 Description: Communicate OST degraded/readonly state via statfs to MDS
909 Details : Flags in the statfs returned from OSTs indicate whether the
910 OST is in a degraded RAID state, or if the filesystem has
911 turned read-only after a filesystem error is detected.
916 Description: don't panic if EPROTO was hit when reading symlink
917 Details : correctly handling request reference in error cases.
919 Severity : enhancement
921 Description: Add LustreNetLink, a kernel-userspace communcation path.
923 Severity : enhancement
925 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
930 Description: access to llog context before init.
931 Details : move handling CATALOGS file at osc layer and forbid access to llog
935 Frequency : rare, only when client is writing to OST on local node
937 Description: Avoid deadlock for local client writes
938 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
939 memory freeing context. This allows OST threads to set the
940 PF_MEMALLOC flag on task structures in order to allocate memory
941 from reserved pools and complete IO.
942 Use GFP_HIGHUSER for OST allocations for non-local client writes,
943 so that the OST threads generate memory pressure and allow
944 inactive pages to be reclaimed.
946 Severity : enhancement
948 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
953 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
954 Details : move ldlm namespace creation in setup phase to avoid grab
955 _lprocfs_lock with cli_sem held.
959 Description: Temporarily disable grant shrink.
960 Details : Disable the feature for debugging.
963 Frequency : only when formatting test filesystems
965 Description: Unable to run several mkfs.lustre on loop devices at the same time
966 Details : mkfs.lustre returns error 256 on the concurrent loop devices
967 formatting. The solution is to proper handle the error.
969 Severity : enhancement
971 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
973 Severity : enhancement
975 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
980 Description: resolve race between obd_disconnect and class_disconnect_exports
981 Details : if obd_disconnect will be called to already disconnected export he
982 forget release one reference and osc module can't unloaded.
984 Severity : enhancement
986 Description: Allow tuning service thread via /proc
987 Details : For each service a new
988 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
989 created that can be used to set min/max thread counts, and get the
990 current number of running threads.
995 Descriptoin: don't return error if have partially created objects for file.
996 Details : lov_update_create_set uses set->set_success as index for created
997 objects, so if some requests will be failed, they will have hole
998 at end of array and qos_shrink_lsm for allocate correct lsm.
1000 Severity : enhancement
1002 Description: Update OFED support to 1.4.1
1007 Description: Reduce small size read RPC
1008 Details : Set read-ahead limit for every file and only do read-ahead when
1009 available read-ahead pages are bigger than 1M to avoid small size
1012 Severity : enhancement
1014 Description: move AT tunable parameters for more consistent usage
1015 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1017 Severity : enhancement
1019 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1020 Details : allow skip disconnected ost for send statfs request and hide error
1026 Description: fix racy locking of mballoc block bitmaps causing BUG
1027 Details : The locking of the mballoc buddy bitmap and the in-memory
1028 block bitmap was using two different spin locks in some
1029 cases. This made it possible to incorrectly access the
1030 mballoc bitmap while another process was modifying it,
1031 causing a sanity assertion to fail. While no on-disk corruption
1032 was reported, there was some risk of this happening.
1035 Frequency : rare, on llog test 6
1037 Descriptoin: don't allow connect to already connected import
1038 Details : allowing connect to already connected import is hide connecting problem.
1043 Descriptoin: allow kill process which wait statahead result
1044 Details : for some reasons 'ls' can stick in waiting result from statahead,
1045 in this case need way for kill this process.
1047 Severity : enhancement
1049 Description: Add state history info file, enhance import info file
1050 Details : Track import connection state changes in a new osc/mdc proc file;
1051 add overview-type data to the osc/mdc import proc file.
1053 Severity : enhancement
1055 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1057 Severity : enhancement
1059 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1064 Descriptoin: OST grant too much space to client even there are not enough space.
1065 Details : Client will shrink its grant cache to OST if there are no write
1066 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1067 this grant cache if there are already not enough avaible space
1068 (left_space < total_clients * 32M).
1071 Frequency : start MDS on uncleanly shutdowned MDS device
1073 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1074 Details : stay in waiting mds<>ost recovery finished produce random bugs
1075 due race between two ll_sync thread for one lov target. send
1076 ACTIVATE event only if connect realy finished and import have
1080 Frequency : rare, connect and disconnect target at same time
1082 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1083 Details : don't call obd_disconnect under lov_lock. this long time
1084 operation and can block ptlrpcd which answer to connect request.
1089 Descriptoin: don't lose wakeup for imp_recovery_waitq
1090 Details : recover_import_no_retry or invalidate_import and import_close can
1091 both sleep on imp_recovery_waitq, but we was send only one wakeup
1095 Frequency : always with long access acl
1097 Descriptoin: mds can't pack reply with long acl.
1098 Details : mds don't control size of acl but they limited by reint/getattr
1101 Severity : enhancement
1103 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1105 Severity : enhancement
1107 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1110 Frequency : start MDS on uncleanly shutdowned MDS device
1112 Descriptoin: aborting recovery hang on MDS
1113 Details : don't throttle destroy RPCs for the MDT.
1116 Frequency : on remount
1118 Description: external journal device not working after the remount
1119 Details : clear dev_rdonly flag for external journal devices in
1125 Description: shutdown vs evict race
1126 Details : client_disconnect_export vs connect request race.
1127 if client will evicted at this time - we start invalidate
1128 thread without referece to import and import can be freed
1132 Frequency : rare, need acl's on inode.
1134 Description: client can't handle ost additional correctly
1135 Details : if ost was added after client connected to mds client can have
1136 hit lnet_try_match_md ... to big messages to wide striped files.
1137 in this case need teach client to handle config events about add
1138 lov target and update client max ea size at that event.
1140 Severity : enhancement
1142 Description: Changelogs
1143 Details : Changelogs are a lightweight mechanism to track filesystem
1144 metadata and namespace changes. The changelog is recorded
1145 permanently on the MDTs, and is periodically "consumed" / purged
1146 when records are no longer needed.
1148 Severity : enhancement
1150 Description: compact fld format with extents
1151 Details : Store range of seq rather than every seq in FLD. Seq
1152 controller update FLD rather than clients. In Case of CMD, mdt0
1153 has FLD, all other metadata server act as non persistent proxy
1154 for FLD queries and cache fld entries in fld cache.
1159 Description: don't skip ost target if they assigned to file
1160 Details : Drop slow OSCs if we can, but not for requested start idx.
1161 This means "if OSC is slow and it is not the requested
1162 start OST, then it can be skipped, otherwise skip it only
1163 if it is inactive/recovering/out-of-space.
1167 Description: more cleanup in mds_lov
1168 Details : not send LOV EA under replay, we can't know about they size at this
1169 time. Don't allow client connect to mds before any ost connected,
1170 for avoid problems with LOV EA size and returning EIO to client.
1172 Severity : enhancement
1174 Description: Interoperability at server side (Disk interoperability)
1176 Severity : enhancement
1178 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1180 Severity : enhancement
1182 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1184 Severity : enhancement
1186 Description: New client IO stack (CLIO).
1188 Severity : enhancement
1190 Description: Commit on sharing. Eliminate inter-client dependencies between
1191 uncommitted transactions by doing transaction commits.
1192 Thereby clients may recovery independently.
1195 Frequency : Create a symlink file with a very long name
1197 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1198 Details : If there is no extra space in the request for early cancels,
1199 ldlm_req_handles_avail() returns 0 instead of a negative value.
1201 Severity : enhancement
1203 Description: Add /proc entry for import status
1204 Details : The mdc, osc, and mgc import directories now have
1205 an import directory that contains useful import data for debugging
1206 connection problems.
1208 Severity : enhancement
1210 Description: Re-disable certain /proc logging
1211 Details : Enable and disable client's offset_stats, extents_stats and
1212 extents_stats_per_process stats logging on the fly.
1215 Frequency : Only on FC kernels 2.6.22+
1217 Description: oops in statahead
1218 Details : Do not drop reference count for the dentry from VFS when lookup,
1219 VFS will do that by itself.
1221 Severity : enhancement
1223 Description: Generic /proc file permissions
1224 Details : Set /Proc file permissions in a more generic way to enable non-
1225 root users operate on some /proc files.
1229 Description: Hitting mdc_commit_close() ASSERTION
1230 Details : Properly handle request reference release in
1231 ll_release_openhandle().
1235 Description: quota recovery deadlock during mds failover
1236 Details : This patch includes att18982, att18236, att18237 in bz14840.
1238 1. fix osts hang when mds does failover with quotaon
1239 2. prevent watchdog storm when osts threads wait for the
1244 Frequency : only patchless client
1245 Description: add workaround for race between add/remove dentry from hash
1247 Severity : enhancement
1249 Description: Allow OST glimpses to return PW locks
1253 Description: LBUG when llog conf file is full
1254 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1258 Description: Prevent import from entering FULL state when server in recovery
1262 Description: service mount cannot take device name with ":"
1263 Details : Only when device name contains ":/" will mount treat it as
1269 Description: replace ptlrpcd with the statahead thread to interpret the async
1270 statahead RPC callback
1274 Frequency : on recovery
1275 Description: I/O failures after umount during fail back
1276 Details : if client reconnected to restarted server we need join to recovery
1277 instead of find server handler is changed and process self
1278 eviction with cancel all locks.
1280 Severity : enhancement
1282 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1284 Severity : enhancement
1286 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1288 Severity : enhancement
1290 Description: Upcall on Lustre log has been dumped
1291 Details : Allow for a user mode script to be called once a Lustre log has
1292 been dumped. It passes the filename of the dumped log to the
1293 script, the location of the script can be specified via
1294 /proc/sys/lnet/debug_log_upcall.
1299 Description: avoid idr_remove called for id which is not allocated.
1300 Details : Move assigment s_dev for clustered nfs to end of initialization,
1301 to avoid problem with error handling.
1306 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1307 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1308 into CONN_USED_HASH and this prodice warning when put connection
1309 again in unused hash.
1312 Severity : enhancement
1314 Description: File striping can now be set to use an arbitrary pool of OSTs.
1316 Severity : enhancement
1318 Description: Export bytes_read/bytes_write count on OSC/OST.
1322 Description: Early reply size mismatch, MGC loses connection
1323 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1324 the connect flags are properly negotiated.
1329 Description: quotas are not honored with O_DIRECT
1330 Details : all writes with the flag O_DIRECT will use grants which leads to
1331 this problem. Now using OBD_BRW_SYNC to guard this.
1335 Description: add quota statistics
1336 Details : 1. sort out quota proc entries and proc code.
1337 2. add quota statistics
1339 Severity : enhancement
1341 Description: enable quota support for HEAD.
1345 Description: Properly propagate oinfo flags from lov to osc for statfs
1346 Details : restore missing copy oi_flags to lov requests.
1348 Severity : enhancement
1350 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1351 : l_getgroups(8), lst(8), routerstat(8)
1352 Details : included man pages for llobdstat(8), llstat(8),
1353 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1355 Severity : enhancement
1357 Description: configure's --enable-quota should check the
1358 : kernel .config for CONFIG_QUOTA
1359 Details : configure is terminated if --enable-quota is passed but
1360 : no quota support is in kernel
1364 Description: Remove portals compatibility
1365 Details : Remove portals compatibility, not interoperable with releases
1370 Description: Resolve device initialization race
1371 Details : Prevent proc handler from accessing devices added to the
1372 obd_devs array but yet be intialized.
1374 Severity : enhancement
1376 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1378 Severity : enhancement
1380 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1385 Description: Using wrong pointer in osc_brw_prep_request
1386 Details : Access to array[-1] can produce panic if kernel compiled with
1387 CONFIG_PAGE_ALLOC enabled
1391 Description: Client runs out of low memory
1392 Details : Consider only lowmem when counting initial number of llap pages
1396 Description: *optional* service tags registration
1397 Details : if the "service tags" package is installed on a Lustre node
1398 When the filesystem is mounted, a local-node service tag will
1399 be created. See http://inventory.sun.com/ for more information
1400 about the Service Tags asset management system.
1404 Description: Kernel BUG tries to release flock
1405 Details : Lustre does not destroy flock lock before last reference goes
1406 away. So always drop flock locks when client is evicted and
1407 perform unlock regardless of successfulness of speaking to MDS.
1411 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1415 Description: sanity test 65a fails if stripecount of -1 is set
1416 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1421 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1422 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1425 Severity : enhancement
1427 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1431 Description: do not process already freed flock
1432 Details : flock can possibly be freed by another thread before it reaches
1433 to ldlm_flock_completion_ast.
1437 Description: LBUG during stress test
1438 Details : Need properly lock accesses the flock deadlock detection list.
1442 Description: oops in page fault handler
1443 Details : kernel page fault handler can return two special 'pages' in error
1444 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1448 Description: timeout with invalidate import.
1449 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1450 should be handled by ptlrpcd. This produce long age waiting and
1451 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1453 Severity : enhancement
1455 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1459 Description: enable MGS and MDT services start separately
1460 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1461 a co-located MGS without starting the MGS, which is a complement
1462 to 'nosvc' mount option.
1466 Frequency : after recovery
1467 Description: precreate to many object's after del orphan.
1468 Details : del orphan st in oscc last_id == next_id and this triger growing
1469 count of precreated objects. Set flag LOW to skip increase count
1470 of precreated objects.
1474 Frequency : rare, on clear nid stats
1475 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1476 Details : when clean nid stats sometimes try destroy live entry,
1477 and this produce panic in free.
1481 Description: Stack overflow during MDS log replay
1482 ease stack pressure by using a thread dealing llog_process.
1486 Description: wait until IO finished before start new when do lock cancel.
1487 Details : VM protocol want old IO finished before start new, in this case
1488 need wait until PG_writeback is cleared until check dirty flag and
1489 call writepages in lock cancel callback.
1491 Severity : enhancement
1493 Description: using special macro for print time and cleanup in includes.
1497 Description: mds_mfd_close() ASSERTION(rc == 0)
1498 Details : In mds_mfd_close(), we need protect inode's writecount change
1499 within its orphan write semaphore to prevent possible races.
1503 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1507 Frequency : rare, on shutdown ost
1508 Description: don't hit live lock with umount ost.
1509 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1510 use shrink_dcache_sb instead.
1514 Description: don't panic with use echo client
1515 Details : echo client pass NULL as client nid pointer and this produce null
1516 pointer dereference.
1520 Description: fix build on ppc32
1521 Details : compile code with -m64 flag produce wrong object file for ppc32.
1525 Description: add message levels for liblustreapi
1529 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1530 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1531 will result in an -EAGAIN and in the exit call path we call
1532 alloc_rr() which will with increasing aggressiveness attempt to
1533 aquire precreated objects on the minimum number of required OSCs.
1537 Description: Use old size assignment to avoid deadlock
1538 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1539 the scheduling while holding a spinlock. We do not need locking
1540 for size in ll_update_inode() because size is only updated from
1541 the MDS for directories or files without objects, so there is no
1542 other place to do the update, and concurrent access to such inodes
1543 are protected by the inode lock.
1547 Description: resolve "_IOWR redefined" build error on SLES10
1551 Description: dump the memory debugging after all modules are unloaded to
1552 suppress false negative in conf_sanity test 39
1554 Severity : enhancement
1556 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1561 Description: fix SLES kernel versioning
1562 Details : the kernel version for our SLES 10 kernel did not include a "-"
1563 before the "smp" at the end. while this was not a problem in
1564 general, it did mean that software trying to use the kernel
1565 version to try to detect a vendor specific kernel would fail.
1566 this was most evident by the OFED build scripts.
1570 Description: Don't update lov_desc members until making sure they are valid
1571 Details : When updating lov_desc members via proc fs, need fix their
1572 validities before doing the real update.
1576 Description: don't put request into delay list while invalidate in flight.
1577 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1578 import in flight. this produce timeout for invalidate and sometimes
1579 can cause stale data.
1583 Frequency : on ppc only
1584 Description: not convert ost objects for directory because it's not exist.
1585 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1589 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1591 Severity : enhancement
1593 Description: Update to support 2.6.22.14 vanilla kernel.
1597 Frequency : rare, on recovery
1598 Description: read procfs can produce deadlock in some situation
1599 Details : Holding lprocfs lock which send rpc can produce block for destroy
1600 obd objects and this also block reconnect with -EALREADY. This
1601 isn't fix all lprocfs bugs - but make it rare.
1603 Severity : enhancement
1605 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1608 Frequency : frequent on X2 node
1610 Description: mdc_set_open_replay_data LBUG
1611 Details : Set replay data for requests that are eligible for replay.
1615 Description: lustre_mgs: operation 101 on unconnected MGS
1616 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1617 MGC, and late on MGC cannot successfully connect to MGS and a lot
1618 of the error messages complaining that MGS is not connected.
1622 Frequency : on network error
1623 Description: panic with double free request if network error
1624 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1625 but it's true only for synchronus enqueue, for async enqueue
1626 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1629 Severity : enhancement
1631 Description: client-side metadata stat-ahead during readdir(directory readahead)
1632 Details : perform client-side metadata stat-ahead when the client detects
1633 readdir and sequential stat of dir entries therein
1636 Frequency : on start mds
1638 Description: Implement get_info(last_id) in obdfilter.
1641 Frequency : occasional
1643 Description: Correctly check stale fid, not start epoch if ost not support SOM
1644 Details : open with flag O_CREATE need set old fid in op_fid3 because
1645 op_fid2 was overwritten with new generated fid, but mds can answer
1646 with one of these two fids and both is not stale. Setattr
1647 incorrectly started an epoch and assume will be called
1648 done_writing, but without SOM done_writing ever being called.
1651 Frequency : rare, depends on device drivers and load
1653 Description: MDS or OSS nodes crash due to stack overflow
1654 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1655 In some cases, in conjunction with device drivers that use a lot
1656 of stack the MDS (or possibly OSS) service threads could overflow
1657 the stack. One change which was identified to consume additional
1658 stack has been reworked to avoid the extra stack usage.
1661 Frequency : occasional
1663 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1664 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1665 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1666 oscc_has_objects and see OSC has no objests and call
1667 oscc_internal_create to resend create request.
1669 Severity : enhancement
1671 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1673 Severity : enhancement
1675 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1678 Frequency : very rare
1680 Description: Fix replay if there is an un-replied request and open
1681 Details : In some cases, older replay request will revert the
1682 mcd->mcd_last_xid on MDS which is used to record the client's
1683 latest sent request.
1685 Severity : enhancement
1687 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1689 Severity : enhancement
1691 Description: Add rhel5 support to HEAD.
1693 Serverity : enhancement
1695 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1700 Description: Don't allow skipping OSTs if index has been specified.
1701 Details : Don't allow skipping OSTs if index has been specified, make
1702 locking in internal create lots better.
1706 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1707 Details : access to bitfield in structure is always rounded to long
1708 and this produce problem with not atomic change any bit.
1712 Description: Lustre make rpms failed.
1713 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1714 builds Lustre rpms from tarball.
1717 Frequency : rare on shutdown ost
1719 Description: If llog cancel was not send before clean_exports phase, this can
1720 produce deadlock in llog code.
1721 Details : If llog thread has last reference to obd and call class_import_put
1722 this produce deadlock because llog_cleanup_commit_master wait when
1723 last llog_commit_thread exited, but this never success because was
1724 called from llog_commit_thread.
1728 Description: allow userland application know is lost one of stripes.
1729 Details : fill lvb_blocks with error code on ost and return it to
1730 application if error flag found.
1734 Description: NULL lov_tgts causing MDS oops
1735 Details : more safe checks for NULL lov_tgts for avoid oops.
1737 Severity : enhancement
1739 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1743 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1744 Details : If an OST is down the MDS will hang indefinitely in
1745 obd_statfs_rqset() waiting for the statfs data. While for
1746 MDS QOS usage of statfs, it should not stuck in waiting.
1748 Severity : enhancement
1750 Description: remote_acl support
1751 Details : Support ACL-based permission check for remote user.
1752 Support setfacl/getfacl for remote user with the utils
1753 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1754 the system "{s,g}etfacl" utils.
1756 Severity : enhancement
1758 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1760 Severity : enhancement
1762 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1766 Description: make mgs_setparam() handle fsname containing dash
1767 Details : fsname containing a dash does not work with lctl conf_param
1769 Severity : enhancement
1771 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1773 Severity : enhancement
1775 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1779 Frequency : rare, at shutdown
1780 Description: access already free / zero obd_namespace.
1781 Details : if client_disconnect_export was called without force flag set,
1782 and exist connect request in flight, this can produce access to
1783 NULL pointer (or already free pointer) when connect_interpret
1784 store ocd flags in obd_namespace.
1788 Frequency : only at startup
1789 Description: not alloc memory with spinlock held.
1790 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1791 if any spinlock held.
1793 Severity : enhancement
1795 Description: make lustre randomly fail allocating memory
1796 Details : Make lustre randomly failed allocating memory for testing purpose.
1798 Severity : enhancement
1800 Description: lost problems with lov objid file
1801 Details : Fixes some scability and access to not inited memory problems
1802 in work with lov objdid file.
1807 Description: lfs find does not continue on file error
1808 Details : Continue other files processing when a file/dir is absent.
1812 Description: Inconsistent usage of lustre_pack_reply()
1813 Details : Standardize the usage of lustre_pack_reply() such that it
1814 always generate a CERROR on failure.
1817 Frequency : occasional
1819 Description: MDS hang or stay in waiting lock
1820 Details : If client receive lock with CBPENDING flag ldlm need send lock
1821 cancel as separate rpc, to avoid situation when cancel request
1822 can't processed due all i/o threads stay in wait lock.
1826 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1830 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1834 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1838 Description: Update to RHEL4 latest kernel.
1840 Severity : enhancement
1842 Description: Build SLES10 patchless client fails
1843 Details : The configure was broken by run ./configure with
1844 --with-linux-obj=.... argument for patchless client. When the
1845 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1846 can't search header adequately. Use absolute path such as
1847 -I($LINUX)/include instead.
1851 Description: interrupt oig_wait produce painc on resend.
1852 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1853 private set, and this produce situation when rq_ptlrpcd_data not
1854 copyed to new allocated request and triggered LBUG on assert
1855 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1856 ptlrpcd set if request is changed and can be safety changed to use
1861 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1862 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1863 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1864 these macros and replace them with their difinitions instead.
1868 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1869 Details : Update which_patch & target file for SLES10 latest kernel.
1872 Description: Remove client patches from SLES 10 kernel.
1873 Details : This causes SLES 10 clients to behave as patchless clients
1874 even on a Lustre-patched (server) kernel.
1876 Severity : enhancement
1878 Description: self-adjustable client's lru lists
1879 Details : use adaptive algorithm for managing client cached locks lru
1880 lists according to current server load, other client's work
1881 pattern, memory activities, etc. Both, server and client
1882 side namespaces provide number of proc tunables for controlling
1885 Severity : enhancement
1887 Description: light-weight GSS support
1888 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1889 and reduce performance overhead.
1891 Severity : enhancement
1893 Description: Linux keyring support
1894 Details : Support using service of Linux keyring for Lustre GSS internal
1895 context refresh/cache mechanism.
1899 Description: Fix errors in lfs documentation
1900 Details : Fixes man pages
1904 Description: don't use GFP_* in generic Lustre code.
1905 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1909 Description: obdclass is limited by single OBD_ALLOC(idarray)
1910 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1914 Description: warnings with build patchless client with vanila 2.6.19 and up
1915 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1916 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1920 Frequency : Only for SLES
1921 Description: sanity_quota fail test_1
1922 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1923 file, which makes TSTID[2] inunique.
1927 Frequency : Only for SLES9
1928 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1929 Details : Update target/ChangeLog/which_patch .
1933 Frequency : Only for test_10 in sanity-quota.sh
1934 Description: a bug in quota test code
1935 Details : checking a test flag in a wrong place
1939 Frequency : Only for test_13 in sanity-quota.sh
1940 Description: a bug in quota test script
1941 Details : shouldn't check overfull allocated quota
1945 Description: jbd statistics
1946 Details : Port older jbd statistics patch for sles10
1948 Severity : enhancement
1950 Description: Add printing inode info into message about error in writepage.
1955 Description: reply_lock_interpret crash due to race with it and lock cancel.
1956 Details : Do not replay locks that are being cancelled. Do not reference
1957 locks by their address during replay, just by their handle.
1961 Frequency : When flocks are used.
1962 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1963 Details : Flock locks might destroy just granted lock if it could be merged
1964 with another existing flock, this is done in completion handler,
1965 so teach ldlm_cli_enquque_fini that this is a valid case for
1971 Description: Oops in read and write path when failing to allocate lock.
1972 Details : Check if lock allocation failed and return error back.
1976 Description: lstripe command fails for valid OST index
1977 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1978 instead of lov->desc.ld_active_tgt_count.
1982 Description: sanity.sh failed test 103
1983 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1988 Description: df doesn't work properly if diskfs blocksize != 4K
1989 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1993 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1994 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1996 Severity : enhancement
1998 Description: Adaptive timeouts
1999 Details : RPC timeouts adapt to changing server load and network
2000 conditions to reduce resend attempts and improve recovery time.
2004 Description: llapi_file_create() does not allow some changes
2005 Details : add llapi_file_open() that allows specifying the mode and
2006 open flags, and also returns an open file handle.
2010 Description: merge and cleanup kernel patches.
2011 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2015 Description: Add journal checksum support.(Kernel part)
2016 Details : The journal checksum feature adds two new flags i.e
2017 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2018 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2019 indicates that the commit block contains the checksum for
2020 the blocks described by the descriptor blocks. Now commit
2021 record can be sent to disk without waiting for descriptor
2022 blocks to be written to disk. This behavior is controlled
2023 using JBD2_FEATURE_ASYNC_COMMIT flag.
2027 Description: OSS needs mutliple precreate threads
2028 Details : Add ability to start more than one create thread per OSS.
2032 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2033 Details : execute lfs setstripe on client
2037 Description: mds_obd_create error creating tmp object
2038 Details : When the user sets quota on root, llog will be affected and can't
2039 create files and write files.
2042 Frequency : Always on ia64 patchless client, and possibly others.
2044 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2045 Details : This allows the patchless client to be loaded on architectures
2046 without this export.
2050 Description: RedHat Update kernel for RHEL5
2051 Details : Add the kernel config file for RHEL5.
2055 Description: RedHat Update kernel for RHEL5
2056 Details : Modify the kernel config file more closer RHEL5.
2060 Description: Build failure against Centos5 (RHEL5)
2061 Details : Define PAGE_SIZE when it isn't present.
2065 Description: test_7 on sanity-quota.sh can't be used on separate machines
2066 Details : add facet to handle it
2070 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2071 Details : with NFS, the anon dentry's parent was set to itself in
2072 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2073 corresponding dentry other than use rec->ur_name.
2075 Severity : enhancement
2077 Description: lfs setstripe enhancement
2078 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2082 Description: enable data checksumming by default
2083 Details : enable checksum by default, allow --disable-checksum
2084 configure option and "-o nochecksum" mount option. Checksums
2085 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2089 Description: lustre support for RHEL5
2090 Details : Add support for RHEL5.
2094 Description: Client eviction due to failover config
2095 Details : after a connection loss, the lustre client should attempt to
2096 reconnect to the last active server first before trying the
2097 other potential connections.
2101 Description: when mds and osts use different quota unit(32bit and 64bit),
2102 quota will be released repeatly.
2103 Details : void sending multiple quota reqs to mds, which will keep the
2104 status between the reqs.
2108 Description: osts not allocated evenly to files
2109 Details : change the condition to increase offset_idx
2112 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2114 Description: Data corruption for OSTs that are formatted larger than 2TB
2116 Details : When generating the bio request for lustre file writes the
2117 sector number would overflow a temporary variable before being
2118 used for the IO. The data reads correctly from Lustre (which
2119 will overflow in a similar manner) but other file data or
2120 filesystem metadata may be corrupted in some cases.
2124 Description: Tune the kernel for good SCSI performance.
2125 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2126 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2131 Description: rewrite ext2-derived code in obdclass/uuid.c
2132 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2133 from scratch preserving functionality.
2137 Description: "lfs find -obd UUID" prints directories
2138 Details : "lfs find -obd UUID" will return all directory names instead
2139 of just file names. It is incorrect because the directories
2140 do not reside on the OSTs.
2144 Description: use i_size_read and i_size_write in 2.6 port
2145 Details : replace inode->i_size access with i_size_read/write()
2148 Frequency : when removing large files
2150 Description: scheduling issue during removal of large Lustre files
2151 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2152 It causes scheduling issues when removing large files (17TB in the
2156 Frequency : only with liblustre clients on XT3
2158 Description: evictions taking too long
2159 Details : allow llrd to evict clients directly on OSTs
2162 Frequency : only on ppc
2164 Description: /proc/fs/lustre/devices broken on ppc
2165 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2166 We should make sure the type of 'index' is loff_t and then cast
2167 explicitly as needed below. Do not assign an explicitly cast
2171 Frequency : only for rhel5
2173 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2174 Details : Modify the target file & which_kernel.
2176 Severity : enhancement
2178 Description: omit set fsid for export NFS
2179 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2180 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2184 Frequency : Always, for kernels after 2.6.16
2185 Description: Fix warning idr_remove called for id=.. which is not allocated.
2186 Details : Last kernels save old s_dev before kill super and not allow
2187 to restore from callback - restore it before call kill_anon_super.
2191 Description: buffer overruns could theoretically occur
2192 Details : llapi_semantic_traverse() modifies the "path" argument by
2193 appending values to the end of the origin string, and a
2194 overrun may occur. Adding buffer overrun check in liblustreapi.
2198 Description: Fix error on 'ls .' at the top of the Lustre mount.
2199 Details : Don't revalidate dentry if it is a root dentry.
2203 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2204 Details : Modify vm-tunables-rhel4.patch.
2208 Description: Kernel config for 2.6.18-vanilla.
2209 Details : Modify targets/2.6-vanilla.target.in.
2210 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2211 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2212 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2213 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2217 Description: improve handling recoverable errors
2218 Details : if request processig with error which can be recoverable on server
2219 request should be resend, otherwise page released from cache and
2224 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2225 Details : Modify target file & which_patch.
2226 A flaw was found in the IA32 system call emulation provided
2227 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2228 value could be stored in the %RAX register, which could trigger an
2229 out-of-bounds system call table access. An untrusted local user
2230 could exploit this flaw to run code in the kernel
2231 (ie a root privilege escalation). (CVE-2007-4573).
2235 Description: O_DIRECT bypasses client statistics.
2236 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2237 read_ahead_stats were updated. copied the stats section from
2238 osc_send_oap_rpc() into async_internal().
2242 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2246 Description: change order of libsysio includes
2247 Details : '#include sysio.h' should always come before '#include xtio.h'
2251 Description: handle "serious error: objid * already exists" more gracefully
2252 Details : If LAST_ID value on disk is smaller than the objects existing in
2253 the O/0/d* directories, it indicates disk corruption and causes an
2254 LBUG(). If the object is 0-length, then we should use the existing
2255 object. This will help to avoid a full fsck in most cases.
2257 Severity : enhancement
2259 Description: adapt the lustre_config script to support the upgrade case
2260 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2261 server targets to 1.6 in parallel.
2265 Description: To avoid grant space > avaible space when the disk is almost
2266 full. Without this patch you might see the error "grant XXXX >
2267 available" or some LBUG about grant, when the disk is almost
2269 Details : In filter_check_grant, for non_grant cache write, we should
2270 check the left space by if (*left > ungranted + bytes), instead
2271 of (*left > ungranted), because only we are sure the left space
2272 is enough for another "bytes", then the ungrant space should be
2273 increase. In client, we should update cl_avail_grant only there
2274 is OBD_MD_FLGRANT in the reply.
2278 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2279 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2280 described in CVE-2007-4573. This problem could allow untrusted
2281 local users to gain root access.
2284 Frequency : when using O_DIRECT and quotas
2286 Description: Incorrect file ownership on O_DIRECT output files
2287 Details : block usage reported by 'lfs quota' does not take into account
2288 files that have been written with O_DIRECT.
2293 Description: touch file failed when fs is not full
2294 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2295 otherwise we can get ENOSP while fs is not full.
2299 Description: parallel lock callbacks
2300 Details : Instead of sending blocking and completion callbacks as separated
2301 requests, adding them to a set and sending in parallel.
2304 Frequency : only for Cray XT3
2305 Bugzilla : 12829/13455
2306 Description: Changing primary group doesn't change the group lustre assigns to
2308 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2309 overridden with the primary group provided by the group upcall,
2310 whereas the supplied fsgid can be trusted if it is in the list of
2311 supplementary groups returned by the group upcall.
2313 Severity : enhancement
2315 Description: Allow masking D_WARNING, D_ERROR messages from console
2316 Details : Console messages can now be disabled via lnet.printk.
2320 Description: User code with malformed file open parameter crashes client node
2321 Details : Before packing join_file req, all the related reference should be
2322 checked carefully in case some malformed flags cause fake
2323 join_file req on client.
2327 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2328 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2329 mode is equal to the granted mode. However, it should also check
2330 that LDLM_FL_CP_REQD is not set.
2334 Description: Heavy nfs access might result in deadlocks
2335 Details : After ELC code landed, it is now improper to enqueue any mds
2336 locks under och_sem, because enqueue might want to decide to
2337 cancel open locks for same inode we are holding och_sem for.
2341 Description: Client eviction while running blogbench
2342 Details : A lot of unlink operations with concurrent I/O can lead to a
2343 deadlock causing evictions. To address the problem, the number of
2344 oustanding OST_DESTROY requests is now throttled to
2345 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2346 callbacks are processed in priority.
2350 Description: enable ACLs on MDS by default
2351 Details : ACLs must be enabled on MDS by default.
2354 Frequency : PPC/PPC64 only
2356 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2357 Details : fix duplicated definitions between asm-ppc64/types.h and
2358 lustre_types.h on PPC.
2361 Frequency : PPC/PPC64 only
2363 Description: asm-ppc/segment.h does not exist
2364 Details : fix compile issue on PPC.
2368 Description: better handle error messages in extents code
2371 Frequency : RHEL4 only
2373 Description: mkfs is very slow on IA64/RHEL4
2374 Details : A performance regression has been discovered in the MPT Fusion
2375 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2376 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2377 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2379 Severity : enhancement
2381 Description: SNMP support enhancement
2382 Details : Adding total number of sampled request for an MDS node in snmp
2385 Severity : enhancement
2387 Description: Optimize ldlm waiting list processing for PR extent locks
2388 Details : When processing waiting list for read extent lock and meeting read
2389 lock that is same or wider to it that is not contended, skip
2390 processing rest of the list and immediatelly return current
2391 status of conflictness, since we are guaranteed there are no
2392 conflicting locks in the rest of the list.
2396 Description: Time out and refuse to reconnect
2397 Details : When the failover node is the primary node, it is possible
2398 to have two identical connections in imp_conn_list. We must
2399 compare not conn's pointers but NIDs, otherwise we can defeat
2400 connection throttling.
2404 Description: port llog fixes from b1_6 into HEAD
2405 Details : Port llog reference couting and some llog cleanups from b1_6
2406 (bug 10800) into HEAD, for protect from panic and access to already
2407 free llog structures.
2411 Description: Detect stride IO mode in read-ahead
2412 Details : When a client does stride read, read-ahead should detect that and
2413 read-ahead pages according to the detected stride pattern.
2417 Description: data checksumming impacts single node performance
2418 Details : add support for several checksum algorithm. Currently, only CRC32
2419 and Adler-32 are supported. The checksum type can be changed on
2420 the fly via /proc/fs/lustre/osc/*/checksum_type.
2424 Description: use adler32 for page checksums
2425 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2430 Description: build for x2 fails
2431 Details : fix compile issue on Cray systems.
2435 Description: Properly match for duplicate locks
2436 Details : Due to different lock order from skiplists code, we need to
2437 traverse entire list for now
2440 Frequency : only on PPC/SLES10
2442 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2443 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2444 be defined. Add a hack in mkfs_lustre.c to work around this compile
2449 Description: LASSERT on MDS when client holding flock lock dies
2450 Details : ldlm pool logic depends on number of granted locks equal to
2451 number of released locks which is not true for flock locks, so
2452 just exclude such locks from consideration.
2456 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2457 Details : Use fsfilt_sync() for both the whole filesystem sync and
2458 individual file sync to eliminate dangerous inode locking
2459 with I_LOCK that can lead to a deadlock.
2463 Description: performance in 1.6.3
2464 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2465 sure that 1MB requests are not fragmented by the block layer.
2467 Severity : enhancement
2469 Description: organize the server-side client stats on per-nid basis
2470 Details : Change the structure of stats under obdfilter and mds to
2478 The "uuid"s file would list the uuids of _active_ exports.
2479 And the clear entry is to clear all stats and stale nids.
2481 Severity : enhancement
2483 Description: eliminate client locks in face of contention
2484 Details : file contention detection and lockless i/o implementation
2485 for contended files.
2489 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2492 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2494 Description: MDS may lose file striping (and hence file data) in some cases
2495 Details : If there are additional extended attributes stored on the MDS,
2496 in particular ACLs, SELinux, or user attributes (if user_xattr
2497 is specified for the client mount options) then there is a risk
2498 of attribute loss. Additionally, the Lustre file striping
2499 needs to be larger than default (e.g. striped over all OSTs),
2500 and an additional attribute must be stored initially in the
2501 inode and then increase in size enough to be moved to the
2502 external attribute block (e.g. ACL growing in size) for file
2507 Description: skiplist implementation simplification
2508 Details : skiplists are used to group compatible locks on granted list
2509 that was implemented as tracking first and last lock of each
2510 lock group the patch changes that to using doubly linked lists
2514 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2515 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2516 because LASSERT on both the data supplied by a client, and the
2517 data on disk is dangerous and incorrect.
2519 Severity : enhancement
2521 Description: Slow truncate/writes to huge files at high offsets.
2522 Details : Directly associate cached pages to lock that protect those pages,
2523 this allows us to quickly find what pages to write and remove
2524 once lock callback is received.
2528 Description: more ldlm soft lockups
2529 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2530 starve other threads from the resource lock for a long time in
2531 case of long waiting queue, so change the debug level from
2532 D_OTHER to the less frequently used D_INFO.
2534 Severity : enhancement
2536 Description: add -gid, -group, -uid, -user options to lfs find
2540 Description: Hung threads in invalidate_inode_pages2_range
2541 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2542 once one is completed. As a result, some RPCs are stuck in the
2543 queue and are never sent.
2547 Description: filter threads hungs on waiting journal commit
2548 Details : Cleanup filter group llog code, then only filter group llog will
2549 be only created in the MDS/OST syncing process.
2553 Description: Procfs and llog threads access destoryed import sometimes.
2554 Details : Sync the import destoryed process with procfs and llog threads by
2555 the import refcount and semaphore.
2557 Severity : enhancement
2559 Description: openlock cache of b1_6 port to HEAD
2564 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2565 Details : If insertion of an extent fails, then discard the inode
2566 preallocation and free data blocks else it can lead to duplicate
2571 Description: don't always update ctime in ext3_xattr_set_handle()
2572 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2573 In some cases the ctime should not be updated, for example for
2574 2.0->1.8 compatibility it is necessary to delete an xattr and it
2575 should not update the ctime.
2579 Bugzilla : 15713/16362
2580 Description: Assertion in iopen_connect_dentry in 1.6.3
2581 Details : looking up an inode via iopen with the wrong generation number can
2582 populate the dcache with a disconneced dentry while the inode
2583 number is in the process of being reallocated. This causes an
2584 assertion failure in iopen since the inode's dentry list contains
2585 both a connected and disconnected dentry.
2589 Description: assertion failure in ldlm_handle2lock()
2590 Details : fix a race between class_handle_unhash() and class_handle2object()
2591 introduced in lustre 1.6.5 by bug 13622.
2596 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2597 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2602 Description: oops in sd_iostats_seq_show()
2603 Details : unloading/reloading the scsi low level driver triggers a kernel
2604 bug when trying to access the sd iostat file.
2609 Description: Kernel panics during QLogic driver reload
2610 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2611 patch, causing memory corruption.
2616 Description: journal_dev option does not work in b1_6
2617 Details : pass mount option during pre-mount.
2619 Severity : enhancement
2621 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2622 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2623 extent information of a file. It can be used to map logical blocks
2624 in a file to physical blocks in the block device.
2628 Description: LDLM soft lockups - improvement
2629 Details : It is be possible to send the lock handle along with each read
2630 or write request because the client is already doing a lock match
2631 itself so there isn't any reason the OST should have to re-do that
2635 Frequency : only on Cray X2
2637 Description: X2 build failures
2638 Details : fix build failures on Cray X2.
2642 Description: xid & resent requests
2643 Details : Initialize RPC XID from clock at startup (randomly if clock is
2646 Severity : enhancement
2648 Description: Add lustre_start utility to start or stop multiple Lustre servers
2653 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2654 Details : In case of memory pressure, list_del() can be called twice on
2655 req->rq_history_list, causing a kernel oops.
2659 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2660 Details : fix stack overflow in the distributed lock manager by defering
2661 export eviction after a failed AST to the elt thread instead of
2662 handling it in the dlm interpret routine.
2666 Description: Convert some comments to new format.
2667 Details : Update documenting comments to match doxygen conventions.
2671 Description: Grammar fixes.
2672 Details : A couple of trivial spelling fixes.
2676 Description: OSD_COUNTERS-mandatory
2677 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2678 to simplify the code.
2682 Description: simplify cmm_device freeing logic.
2683 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2687 Description: Add lockdep support to dt_object_operations locking interface.
2688 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2689 parameter indicating lock ordering. Update mdd code to use new
2694 Description: Introduce failloc constants for lockless IO tests.
2695 Details : Add two new failloc constants to test lockless IO. Only one of
2696 them in implemented---another is checked in yet to be landed
2701 Description: Add lockdep support for inode mutex.
2702 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2703 in the situations where Lustre has to lock more than one inode
2708 Description: Add optional invariants checking support.
2709 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2710 configure switch. This macro is to be used for consistency and
2711 sanity checks that are too expensive to be left in `production'
2716 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2717 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2722 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2723 Details : Use spin_lock_nested() in (the only) situation where more than
2724 one ldlm_lock is locked simultaneously. Also, fix possible
2725 dead-lock in ldlm_lock_change_resource() by enforcing particular
2730 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2731 Details : Instead of specifying each ldlm_lock call-back through separate
2732 parameter, wrap them into struct ldlm_callback_suite.
2736 Description: Kill join_lru obd method and its callers.
2737 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2738 in memory---a requirement implemented through ->o_join_lru() obd
2739 method in HEAD. Remove this method and its users.
2743 Description: Add asynchronous ldlm ENQUEUE completion handler.
2744 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2745 case that stock ldlm_completion_ast() cannot handle as it waits
2746 until lock is granted. Introduce new ldlm_completion_ast_async()
2747 for this. Also comment ldlm_completion_ast().
2751 Description: ldlm_error <-> errno conversion.
2752 Details : Add functions to map (rather arbitrary) between LDLM error codes
2753 and standard errno values. CLIO needs this to prevent LDLM specific
2754 constants from escaping ldlm and osc.
2758 Description: Kill unused ldlm_handle2lock_ns() function.
2759 Details : Kill unused ldlm_handle2lock_ns() function.
2763 Description: Add lu_ref support to ldlm_lock
2764 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2765 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2766 and ldlm_resource. LDLM interface has to be changed, because code that
2767 releases a reference on a lock, has to "know" what reference this is.
2768 In the most frequent case
2770 lock = ldlm_handle2lock(handle);
2772 LDLM_LOCK_PUT(lock);
2774 no changes are required. When any other reference (received _not_ from
2775 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2776 instead of LDLM_LOCK_PUT().
2778 Arguably, changes are pervasive, and interface requires some discipline
2779 for proper use. On the other hand, it was very instrumental in finding
2780 a few leaked lock references.
2784 Description: Add ldlm_lock_addref_try().
2785 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2786 attempts to addref a lock that might be being canceled
2791 Description: Add ldlm_weigh_callback().
2792 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2793 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2794 hard-coded `number of pages' logic.
2798 Description: Add lockdep annotations to llog code.
2799 Details : Use appropriately tagged _nested() locking calls in the places
2800 where llog takes more than one ->lgh_lock lock.
2804 Description: Add loi_kms_set().
2805 Details : Wrap kms updates into a helper function.
2809 Description: Constify instances of struct lsm_operations.
2810 Details : Constify instances of struct lsm_operations.
2814 Description: lu_conf support.
2815 Details : On a server, a file system object is uniquely identified
2816 by a fid, which is sufficient to locate and load all object
2817 state (inode). On a client, on the other hand, more data are
2818 necessary instantiate an object. Change lu_object_find() and
2819 friends to take additional `lu_conf' argument describing object.
2820 Typically this includes layout information.
2824 Description: lu_context fixes.
2825 Details : Introduce new lu_context functions that are needed on the client
2826 side, where some system threads (ptlrpcd) are shared by multiple
2827 modules, and so cannot be stopped during module shutdown.
2831 Description: Add start and stop methods to lu_device_type_operations.
2832 Details : Introduce two new methods in lu_device_type_operations, that are
2833 invoked when first instance of a given type is created and last one
2834 is destroyed respectively. This is need by CLIO.
2838 Description: Add lu_ref support to struct lu_device.
2839 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2840 track leaked references.
2844 Description: Introduce lu_kmem_descr.
2845 Details : lu_kmem_descr and its companion interface allow to create
2846 and destroy a number of kmem caches at once.
2850 Description: Fix lu_object finalization race.
2851 Details : Fix a race between lu_object_find() finding an object and its
2852 concurrent finalization. This race is (most likely) not possible
2853 on the server, but might happen on the client.
2857 Description: Introduce lu_ref interface.
2858 Details : lu_ref is a debugging module allowing to track references to
2859 a given object. It is quite cpu expensive, and has to be
2860 explicitly enabled with --enable-lu_ref. See usage description
2865 Description: Factor lu_site procfs stats into a separate function.
2866 Details : Separate lu_site stats printing code into a separate function
2867 that can be reused on a client.
2871 Description: Constify instances of struct {lu,dt,md}_device_operations.
2872 Details : Constify instances of struct {lu,dt,md}_device_operations.
2876 Description: Introduce struct md_site and move meta-data specific parts of
2877 struct lu_site here.
2878 Details : Move md-specific fields out of struct lu_site into special struct
2879 md_site, so that lu_site can be used on a client.
2883 Description: Kill mdd_lov_destroy().
2884 Details : Remove unused mdd code.
2888 Description: Add st_block checking to multistat.c.
2889 Details : Add st_block checking to multistat.c.
2893 Description: Add lu_ref support to struct obd_device.
2894 Details : Add lu_ref tracking to obd_device.
2898 Description: Kill obd_set_fail_loc().
2899 Details : Remove unused code.
2903 Description: Add special type for ptlrpc_request interpret functions.
2904 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2905 there. Actual usage will be in CLIO.
2909 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2910 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2911 doesn't work with lockdep.
2915 Description: Add rwv.c test program.
2916 Details : New testing program exercising readv(2) and writev(2) (Qian).
2920 Description: Add sendfile.c test program.
2921 Details : New testing program exercising sendfile(2) (Jay).
2925 Description: Ratelimit a message that can be very frequent.
2926 Details : Ratelimit a memory allocation failure message that can
2931 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2932 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2936 Description: Liblustre build fixes.
2937 Details : Liblustre build fixes.
2941 Description: libcfs: add cfs_{need,cond}_resched() interface.
2942 Details : libcfs: add cfs_{need,cond}_resched() definition and
2943 implementations for Linux, NT, and liblustre.
2945 Severity : enhancement
2947 Description: More exported tunables for mballoc
2948 Details : Add support for tunable preallocation window and new tunables for
2949 large/small requests
2953 Description: Detect corruption of block bitmap and checking for preallocations
2954 Details : Checks validity of on-disk block bitmap. Also it does better
2955 checking of number of applied preallocations. When corruption is
2956 found, it turns filesystem readonly to prevent further corruptions.
2960 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2961 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2962 window to 0. If it does hit cache miss, and read-pattern is still
2963 stride-io mode, does not reset the stride window, but also does not
2964 increase the stride window length in this case.
2968 Frequency : only for big-endian servers
2969 Description: Check if system is big-endian while mounting fs with extents feature
2970 Details : Mounting a filesystem with extents feature will fail on big-endian
2971 systems since ext3-based ldiskfs is not supported on big-endian
2972 systems. This can be over-riden with "bigendian_extents" mount option.
2974 Severity : enhancement
2976 Description: The root squash functionality
2977 Details : A security feature, which is to prevent users from being able
2978 to mount lustre on their desktop, run as root, and delete
2979 all of the files in the filesystem. The goal is accomplished by
2980 remapping user id (UID) and group id (GID) of the root user to
2981 a UID and GID specified by the system administartor via Lustre
2982 configuration management server (MGS). The functionality also
2983 allows to specify sets of clients for which the remapping does
2988 Description: Excessive recovery window
2989 Details : With AT enabled, the recovery window can be excessively long (6000+
2990 seconds). To address this problem, we no longer use
2991 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2992 timeout no longer depends on the service time, it is set to
2993 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2994 time via pb_service_time.
2998 Description: Watchdog triggered on MDS failover
2999 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3000 the OSTs know that the MDS "UUID" can be reused for the same export
3001 from a different NID, so we do not need to wait for the export to be
3005 Frequency : rare, only if using MMP with Linux RAID
3007 Description: MMP doesn't work with Linux RAID
3008 Details : While using HA for Lustre servers with Linux RAID, it is possible
3009 that MMP will not detect multiple mounts. To make this work we
3010 need to unplug the device queue in RAID when the MMP block is being
3011 written. Also while reading the MMP block, we should read it from
3012 disk and not the cached one.
3014 Severity : enhancement
3016 Description: open file using fid
3017 Details : A file can be opened using just its fid, like
3018 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3021 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3024 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3025 Details : add proper locking for creation of jbd_4k slab cache
3029 Description: MMP check in ext3_remount() fails without displaying any error
3030 Details : When multiple mount protection fails during remount, proper error
3033 Severity : enhancement
3035 Description: Allow stripe size to be up to 4G-64k
3036 Details : Fix math logic to allow large stripe sizes.
3040 Description: add check for >8TB ldiskfs filesystems
3041 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3042 Don't allow >8TB ldiskfs filesystems to be mounted without
3043 force_over_8tb mount option
3048 Descriptoin: OSS complains "too many blocks" and crashes
3049 Details : In filter_commitrw_write():
3050 1. lquota_chkquota();
3051 2. obdo_from_inode();
3052 3. lquota_pending_commit();
3053 "2." has changed uid and gid in this case, then caused
3054 lquota_chkquota() and lquota_pending_commit() to use
3055 different uid and gid. That is the root of the bug.
3059 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3060 Details : Detect and print out the 'stripe count = -1' case properly.
3064 Description: Out or order replies might be lost on replay
3065 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3066 request with tid smaller than one currently being added, add it
3067 to the start, not end of the list.
3069 --------------------------------------------------------------------------------
3071 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3073 * Support for kernels:
3074 2.6.5-7.283 (SLES 9),
3075 2.6.9-55.EL (RHEL 4),
3076 2.6.16.46-0.14 (SLES 10),
3077 2.6.18.8 vanilla (kernel.org)
3078 * Client support for unpatched kernels:
3079 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3080 2.6.16 - 2.6.22 vanilla (kernel.org)
3081 * Due to recently discovered recovery problems, we do not recommend
3082 using patchless RHEL 4 clients with this or any earlier release.
3083 * Recommended e2fsprogs version: 1.39.cfs8
3084 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3085 * Starting with this release, the ldiskfs backing filesystem required
3086 by Lustre is now in its own package, lustre-ldiskfs. This package
3087 should be installed. It is versioned separately from Lustre and
3088 may be released separately in future.
3092 Description: block reactivating mgc import until all deactivates complete
3093 Details : Fix race when failing back MDT/MGS to itself (testing)
3096 Frequency : at statup only
3098 Description: mds_lov_synchronize race leads to various problems
3099 Details : simultaneous MDT->OST connections at startup can cause the
3100 sync to abort, leaving the OSC in a bad state.
3102 Severity : enhancement
3104 Description: add optional extra BUILD_VERSION info
3105 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3106 to override the lustre version.
3109 Frequency : 2.6.18 servers only
3111 Description: ll_kern_mount() doesn't release the module reference
3112 Details : The ldiskfs module reference count never drops down to 0
3113 because ll_kern_mount() doesn't release the module reference.
3118 Description: server LBUG when using old ost_num_threads parameter
3119 Details : Accept the old ost_num_threads parameter but warn that it
3120 is deprecated, and fix an off-by-one error that caused an LBUG.
3125 Description: Transient SCSI error results in persistent IO issue
3126 Details : iobuf->dr_error is not reinitialized to 0 between two
3130 Frequency : sometimes when underlying device returns I/O errors
3132 Description: OSTs not going read-only during write failures
3133 Details : OSTs are not remounted read-only when the journal commit threads
3134 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3135 instead of the ext3 wrappers.
3139 Description: poor connect scaling with increasing client count
3140 Details : Don't run filter_grant_sanity_check for more than 100 exports
3141 to improve scaling for large numbers of clients.
3144 Frequency : SLES10 only
3146 Description: sanity-quota.sh quotacheck failed: rc = -22
3147 Details : Quotas cannot be enabled on SLES10.
3150 Frequency : liblustre clients only
3152 Description: getdirentries does not give error when run on compute nodes
3153 Details : getdirentries does not fail when the size specified as an argument
3154 is too small to contain at least one entry
3156 Severity : enhancement
3158 Description: Add LNET router traceability for debug purposes
3159 Details : If a checksum failure occurs with a router as part of the
3160 IO path, the NID of the last router that forwarded the bulk data
3161 is printed so it can be identified.
3166 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3167 Details : Due to a race condition, liblustre clients were occasionally
3168 evicted incorrectly.
3170 Severity : enhancement
3172 Description: lfs setstripe use optional parameters instead of postional
3175 Severity : enhancement
3177 Description: Nanosecond timestamp support for ldiskfs
3178 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3179 resolution timestamps. There is not yet support for this at
3180 the Lustre filesystem level.
3183 Frequency : during server recovery
3185 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3186 Details : request with rq_no_resend flag not awake l_wait_event if they get
3190 Frequency : nfs export on patchless client
3192 Description: connectathon hang when test nfs export over patchless client
3193 Details : Disconnected dentry cannot be found with lookup, so we do not need
3194 to unhash it or make it invalid
3197 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3198 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3199 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3204 Description: open req refcounting wrong on reconnect
3205 Details : If reconnect happened between getting open reply from server and
3206 call to mdc_set_replay_data in ll_file_open, we will schedule
3207 replay for unreferenced request that we are about to free.
3208 Subsequent close will crash in variety of ways.
3209 Check that request is still eligible for replay in
3210 mdc_set_replay_data().
3215 Description: disable writes to filesystem when reading health_check file
3216 Details : the default for reading the health_check proc file has changed
3217 to NOT do a journal transaction and write to disk, because this
3218 can cause reads of the /proc file to hang and block HA state
3219 checking on a healthy but otherwise heavily loaded system. It
3220 is possible to return to the previous behaviour during configure
3221 with --enable-health-write.
3223 Severity : enhancement
3225 Description: 64-bit inode version
3226 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3227 to the inode. This will be required for version-based recovery.
3232 Description: MDS fails to start if a duplicate client export is detected
3233 Details : in some rare cases it was possible for a client to connect to
3234 an MDS multiple times. Upon recovery the MDS would detect this
3235 and fail during startup. Handle this more gracefully.
3237 Severity : enhancement
3239 Description: Add -o localflock option to simulate old noflock
3241 Details : This will achieve local-only flock/fcntl locks
3247 Description: log_commit_thread vs filter_destroy race leads to crash
3248 Details : Take import reference before releasing llog record semaphore
3253 Description: Wrong request locking in request set processing
3254 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3255 imp_delayed_list, in this place should be used imp_lock.
3258 Frequency : when reconnection
3260 Description: Grant Leak when osc reconnect to OST
3261 Details : When osc reconnect ost, OST(filter) should check whether it
3262 should grant more space to client by comparing fed_grant and
3263 cl_avail_grant, and return the granted space to client instead
3264 of "new granted" space, because client will call osc_init_grant
3265 to update the client grant space info.
3268 Frequency : when client reconnect to OST
3270 Description: Grant Leak when osc do resend and replay bulk write
3271 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3272 bulk write request, because the grant info while be sync between
3273 OSC and OST when reconnect, and we should ignore the grant info
3274 these of resend/replay write req.
3279 Description: Grant space more than available left space sometimes.
3280 Details : When then OST is about to be full, if two bulk writing from
3281 different clients came to OST. Accord the avaliable space of the
3282 OST, the first req should be permitted, and the second one
3283 should be denied by ENOSPC. But if the seconde arrived before
3284 the first one is committed. The OST might wrongly permit second
3285 writing, which will cause grant space > available space.
3288 Frequency : when client is evicted
3290 Description: Grant might be wrongly erased when osc is evicted by OST
3291 Details : when the import is evicted by server, it will fork another
3292 thread ptlrpc_invalidate_import_thread to invalidate the
3293 import, where the grant will be set to 0. While the original
3294 thread will update the grant it got when connecting. So if
3295 the former happened latter, the grant will be wrongly errased
3296 because of this race.
3301 Description: Checking Stale with correct fid
3302 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3303 to check whether it is stale, because sometimes, we want the
3304 enqueue happened anyway, and op_data.fid2 will not be initialized.
3306 Severity : enhancement
3308 Description: update patchless client
3309 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3312 Frequency : only with 2.4 kernel
3314 Description: random memory corruption
3315 Details : size of struct ll_inode_info is to big for union inode.u and this
3316 can be cause of random memory corruption.
3321 Description: Memory leak in recovery
3322 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3323 It should also check obd_fail before fsfilt_start, otherwise if
3324 fsfilt_start return -EROFS,(failover mds during mds recovery).
3325 then the req will return with repmsg->transno = 0 and rc = EROFS.
3326 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3327 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3328 be freed no matter whether fsfilt_commit success or not.
3331 Frequency : only with huge count clients
3333 Description: Prevents from taking the superblock lock in llap_from_page for
3335 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3336 for llap_from_page call in ll_removepage prevents from taking the
3337 superblock lock for a soon died page.
3342 Description: Not check open intent error before release open handle
3343 Details : in some rare cases, the open intent error is not checked before
3344 release open handle, which may cause
3345 ASSERTION(open_req->rq_transno != 0), because it tries to release
3346 the failed open handle.
3351 Description: Set cat log bitmap only after create log success.
3352 Details : in some rare cases, the cat log bitmap is set too early. and it
3353 should be set only after create log success.
3357 Description: Accessing a block bevice can re-enable I/O when Lustre is
3358 tearing down a device.
3359 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3364 Description: service threads may hog cpus when there are a lot of requests
3365 Details : Insert cond_resched to give other threads a chance to use some CPU
3370 Description: the cat log was not initialized in recovery
3371 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3372 unlink log on mds will not be initialized until mds post
3373 recovery. And also in mds post recovery, the unlink log will
3374 initialization will be done asynchronausly, so there will be race
3375 between add unlink log and unlink log initialization.
3379 Description: brw_stats were being printed incorrectly
3380 Details : brw_stats were being printed as log2 but all of them were not
3381 recorded as log2. Also remove some code duplication arising from
3382 filter_tally_{read,write}.
3386 Frequency : rare, only in recovery.
3387 Description: ASSERTION(req->rq_type != LI_POISON) failed
3388 Details : imp_lock should be held while iterating over imp_sending_list for
3389 prevent destroy request after get timeout in ptlrpc_queue_wait.
3393 Description: replay-single.sh test 52 fails
3394 Details : A lock's skiplist need to be cleanup when it being unlinked
3395 from its resource list.
3399 Description: Short directio read returns full requested size rather than
3401 Details : Direct I/O operations should return actual amount of bytes
3402 transferred rather than requested size.
3404 Severity : enhancement
3406 Description: metadata RPC reduction (e.g. for rm performance)
3407 Details : decrease the amount of synchronous RPC between clients and servers
3408 by canceling conflicing lock before the operation on the client
3409 and packing thier handles into the main operation RPC to server.
3411 Severity : enhancement
3413 Description: Async OSC create to avoid the blocking unnecessarily.
3414 Details : If an OST has no remaining object, system will block on the
3415 creation when it needs to create a new object on this OST. Now,
3416 ways use pre-created objects when available, instead of blocking on
3417 an empty osc while others are not empty. If we must block, we
3418 block for the shortest possible period of time.
3422 Description: improve handling recoverable errors
3423 Details : if request processig with error which can be recoverable on server
3424 request should be resend, otherwise page released from cache and
3427 Severity : enhancement
3429 Description: refine locking for avoid write wrong info into lov_objid file
3430 Details : fix possible races with add new target and write/update data in
3433 --------------------------------------------------------------------------------
3435 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3440 Frequency : on some architectures
3442 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3443 Details : Uninitialized flags sometimes cause configuration commands to
3447 Frequency : patchless clients only
3449 Description: missing __iget() symbol export
3450 Details : The __iget() symbol export is missing. To avoid the need for
3451 this on patchless clients the deathrow inode reaper is turned
3452 off, and we depend on the VM to clean up old inodes. This
3453 dependency was during via the fix for bug 12181.
3455 --------------------------------------------------------------------------------
3457 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3459 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3460 INTEROPERATE with older versions automatically. In many cases a
3461 special upgrade step is needed. Please read the
3462 user documentation before upgrading any part of a live system.
3463 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3464 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3465 * WARNING: Lustre configuration and startup changes are required with
3466 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3470 Severity : enhancement
3472 Description: MountConf
3473 Details : Lustre configuration is now managed via mkfs and mount
3474 commands instead of lmc and lconf. New obd types (MGS, MGC)
3475 are added for dynamic configuration management. See
3476 https://mail.clusterfs.com/wikis/lustre/MountConf for
3479 Severity : enhancement
3481 Description: dynamic OST addition
3482 Details : OSTs can now be added to a live filesystem
3484 Severity : enhancement
3486 Description: startup order invariance
3487 Details : MDTs and OSTs can be started in any order. Clients only
3488 require the MDT to complete startup.
3490 Severity : enhancement
3492 Description: parallel, asynchronous orphan cleanup
3493 Details : orphan cleanup is now performed in separate threads for each
3494 OST, allowing parallel non-blocking operation.
3496 Severity : enhancement
3498 Description: optimized stripe assignment
3499 Details : stripe assignments are now made based on ost space available,
3500 ost previous usage, and OSS previous usage, in order to try
3501 to optimize storage space and networking resources.
3503 Severity : enhancement
3505 Description: Permanently set tunables
3506 Details : All writable /proc/fs/lustre tunables can now be permanently
3507 set on a per-server basis, at mkfs time or on a live system.
3509 Severity : enhancement
3511 Description: Lustre message v2
3512 Details : Add lustre message format v2.
3514 Severity : enhancement
3516 Description: client OST exclusion list
3517 Details : Clients can be started with a list of OSTs that should be
3518 declared "inactive" for known non-responsive OSTs.
3522 Description: SPEC SFS validation failure on NFS v2 over lustre.
3523 Details : Changes the blocksize for regular files to be 2x RPC size,
3524 and not depend on stripe size.
3526 Severity : enhancement
3528 Description: Multiple MD RPCs in flight.
3529 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3530 To avoid overly-overloading MDS, introduce a limit on number of
3531 MDS RPCs in flight for a single client and add /proc controls
3532 to adjust this limit.
3534 Severity : enhancement
3536 Description: client read/write statistics
3537 Details : Add client read/write call usage stats for performance
3538 analysis of user processes.
3539 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3540 file access. extents_stats shows chunk size distribution.
3541 extents_stats_per_process show chunk size distribution per
3544 Severity : enhancement
3546 Description: mds statistics
3547 Details : Add detailed mds operations statistics in
3548 /proc/fs/lustre/mds/*/stats.
3552 Description: Failure of copying files with lustre special EAs.
3553 Details : Client side always return success for setxattr call for lustre
3554 special xattr (currently only "trusted.lov").
3559 Description: Refcount LNET uuids
3560 Details : The global LNET uuid list grew linearly with every startup;
3561 refcount repeated list entries instead of always adding to
3564 Severity : enhancement
3566 Description: Dynamic service threads
3567 Details : Within a small range, start extra service threads
3568 automatically when the request queue builds up.
3571 Frequency : mixed-endian client/server environments
3573 Description: mixed-endian crashes
3574 Details : The new msg_v2 system had some failures in mixed-endian
3577 Severity : enhancement
3579 Description: Easy OST removal
3580 Details : OSTs can be permanently deactivated with e.g. 'lctl
3581 conf_param lustre-OST0001.osc.active=0'
3583 Severity : enhancement
3585 Description: MGS proc entries
3586 Details : Added basic proc entries for the MGS showing what filesystems
3589 Severity : enhancement
3591 Description: provide MGS failover
3592 Details : Added config lock reacquisition after MGS server failover.
3594 Severity : enhancement
3596 Description: add Linux 2.4 support
3597 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3601 Description: a large application tries to do I/O to the same resource and dies
3602 in the middle of it.
3603 Details : Check the req->rq_arrival time after the call to
3604 ost_brw_lock_get(), but before we do anything about
3605 processing it & sending the BULK transfer request. This
3606 should help move old stale pending locks off the queue as
3607 quickly as obd_timeout.
3610 Frequency : when an incorrect nid is specified during startup
3612 Description: ptlrpc connect to non-existant node causes kernel crash
3613 Details : LNET can't be re-entered from an event callback, which
3614 happened when we expire a message after the export has been
3615 cleaned up. Instead, hand the zombie cleanup off to another
3618 Severity : enhancement
3620 Description: plain/inodebits lock performance improvement
3621 Details : Grouping plain/inodebits in granted list by their request modes
3622 and bits policy, thus improving the performance of search through
3626 Frequency : only if OST filesystem is corrupted
3628 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3629 Details : for a short time RPCs with bulk IO are in the replay list,
3630 but replay of bulk IOs is unimplemented. If the OST filesystem
3631 is corrupted due to disk cache incoherency and then replay is
3632 started it is possible to trip an assertion. Avoid putting
3633 committed RPCs into the replay list at all to avoid this issue.
3636 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3638 Description: System hang on startup
3639 Details : This bug allowed the liblustre (e.g. catamount) client to
3640 return to the app before handling all startup RPCs. This
3641 could leave the node unresponsive to lustre network traffic
3642 and manifested as a server ptllnd timeout.
3644 Severity : enhancement
3646 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3647 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3648 boolean to control whether to print peer debug info when a
3649 client's RPC times out.
3652 Frequency : only for kernels with patches from Lustre below 1.4.3
3654 Description: Remove old rdonly API
3655 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3658 Frequency : only for devices with external journals
3660 Description: Set external device read-only also
3661 Details : During a commanded failover stop, we set the disk device
3662 read-only while the server shuts down. We now also set any
3663 external journal device read-only at the same time.
3666 Frequency : when upgrading from 1.4 while trying to change parameters
3668 Description: The wrong (new) MDC name was used when setting parameters for
3669 upgraded MDT's. Also allows changing of OSC (and MDC)
3670 parameters if --writeconf is specified at tunefs upgrade time.
3673 Frequency : when setting specific ost indicies
3675 Description: QOS code breaks on skipped indicies
3676 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3677 created with --index need not be sequential.
3679 Severity : enhancement
3681 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3682 Details : The uninit_groups feature works in conjunction with the kernel
3683 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3684 up the pass1 processing of e2fsck. This is a read-only feature
3685 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3686 mount filesystems that have had this feature enabled.
3688 Severity : enhancement
3690 Description: Improve multi-block allocation algorithm to avoid fragmentation
3691 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3692 allocation locality and avoid filesystem fragmentation.
3694 ------------------------------------------------------------------------------
3696 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3698 * Support for kernels:
3699 2.4.21-47.0.1.EL (RHEL 3)
3700 2.6.5-7.283 (SLES 9)
3701 2.6.9-42.0.10.EL (RHEL 4)
3702 2.6.12.6 vanilla (kernel.org)
3703 2.6.16.27-0.9 (SLES 10)
3704 * Recommended e2fsprogs version: 1.39.cfs5
3706 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3710 Frequency : occasional, depends on client load and configuration
3711 Bugzilla : 12181, 12203
3712 Description: data loss for recently-modified files
3714 Details : In some cases it is possible that recently written or created
3715 files may not be written to disk in a timely manner (this should
3716 normally be within 30s unless client IO load is very high).
3717 The problem appears as zero-length files or files that are a
3718 multiple of 1MB in size after a client crash or client eviction
3719 that are missing data at the end of the file.
3721 This problem is more likely to be hit on clients where files are
3722 repeatedly created and unlinked in the same directory, clients
3723 have a large amount of RAM, have many CPUs, the filesystem has
3724 many OSTs, the clients are rebooted frequently, and/or the files
3725 are not accessed by other nodes after being written.
3727 The presence of the problem can be detected by looking at
3728 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3729 smaller than the second (nr_unused) then dirty files will not
3730 be flushed automatically to disk. "sync; sleep 10" should be
3731 run several times on the node before unmounting it to update
3732 Lustre (this is also safe to run on nodes without this problem).
3734 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3735 that can cause this same problem, so customers using that kernel
3736 also need to update the kernel in addition to Lustre. In order
3737 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3739 It is normal that files written just before a client crash (less
3740 than 30s) may not yet have been flushed to disk, even for local
3744 Frequency : frequent on thin XT3 nodes
3746 Description: UUID collision on thin XT3 Linux nodes
3747 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3748 randomly, since we relied on an insufficiently-seeded PRNG.
3753 Description: OSS hangs after "All ost request buffers busy"
3754 Details : A deadlock between quota and journal operations caused OSS
3755 hangs after printing "All ost request buffers busy."
3758 Frequency : always on liblustre builds
3760 Description: Cleanup compiler warnings on liblustre
3763 Frequency : always on liblustre builds on XT3
3765 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3769 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3770 > 32000 subdirectories
3771 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3772 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3773 subdirectory count crosses 32000. This will aid e2fsck to
3774 correctly handle more than 32000 subdirectories.
3777 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3779 Description: System hang on startup
3780 Details : This bug allowed the liblustre (e.g. catamount) client to
3781 return to the app before handling all startup RPCs. This
3782 could leave the node unresponsive to lustre network traffic
3783 and manifested as a server ptllnd timeout.
3785 Severity : enhancement
3787 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3788 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3789 boolean to control whether to print peer debug info when a
3790 client's RPC times out.
3795 Description: make O_SYNC working on 2.6 kernels
3796 Details : 2.6 kernels use different method for mark pages for write,
3797 so need add a code to lustre for O_SYNC work.
3802 Description: Failure to close file and release space on NFS
3803 Details : Put inode details into lock acquired in ll_intent_file_open.
3804 Use mdc_intent_lock in ll_intent_open to properly
3805 detect all kind of errors unhandled by mdc_enqueue.
3810 Description: proc file read during shutdown sometimes raced obd removal,
3812 Details : Add lock to prevent obd access after proc file removal.
3815 Frequency : Only for files larger than 4GB on 32-bit clients.
3817 Description: improperly doing page alignment of locks
3818 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3819 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3824 Description: RPCs being resent when they shouldn't be
3825 Details : Some RPCs that should not be resent are being resent. This
3826 can cause inconsistencies in the RPC state machine. Do not
3827 resend such requests.
3830 Frequency : rare, only with NFS export
3832 Description: Crash on NFS re-export node
3833 Details : under very unusual load conditions an assertion is hit in
3834 ll_intent_file_open()
3837 Frequency : only if OST filesystem is corrupted
3839 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3840 Details : for a short time RPCs with bulk IO are in the replay list,
3841 but replay of bulk IOs is unimplemented. If the OST filesystem
3842 is corrupted due to disk cache incoherency and then replay is
3843 started it is possible to trip an assertion. Avoid putting
3844 committed RPCs into the replay list at all to avoid this issue.
3849 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3850 Details : Large single O_DIRECT read and write calls can fail to allocate
3851 a sufficiently large buffer to process the request. In case of
3852 allocation failure the allocation is retried with a smaller
3853 buffer and broken into smaller requests.
3855 Severity : enhancement
3857 Description: Add -o localflock option to simulate old noflock behaviour.
3858 Details : This will achieve local-only flock/fcntl locks coherentness.
3863 Description: versioning check is incomplete
3864 Details : Checking the version difference of client vs. server, report
3865 error if the gap is too big.
3870 Description: add support PG_writeback bit
3871 Details : add support for PG_writeback bit for Lustre, for more carefull
3872 work with page cache in 2.6 kernel. This also fix some deadlocks
3873 and remove hack for work O_SYNC with 2.6 kernel.
3875 Severity : enhancement
3877 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3878 Details : The uninit_groups feature works in conjunction with the kernel
3879 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3880 up the pass1 processing of e2fsck. This is a read-only feature
3881 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3882 mount filesystems that have had this feature enabled.
3884 Severity : enhancement
3886 Description: Improve multi-block allocation algorithm to avoid fragmentation
3887 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3888 allocation locality and avoid filesystem fragmentation.
3890 ------------------------------------------------------------------------------
3892 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3894 * Support for kernels:
3895 2.6.9-42.0.3.EL (RHEL 4)
3896 2.6.5-7.276 (SLES 9)
3897 2.4.21-47.0.1.EL (RHEL 3)
3898 2.6.12.6 vanilla (kernel.org)
3899 2.6.16.21-0.8 (SLES10)
3900 * Recommended e2fsprogs version: 1.39.cfs2-0
3902 * The backwards-compatible /proc/sys/portals symlink has been removed
3903 in this release. Before upgrading, please ensure that you change
3904 any configuration scripts or /etc/sysctl.conf files that access
3905 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3906 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3907 in advance of the upgrade on any system running Lustre 1.4.6 or
3908 newer, since /proc/sys/lnet was added in that version.
3909 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3913 Frequency : only when quota is used
3915 Description: avoid scanning export list for quota master
3916 Details : Change the algorithms to avoid scanning export list in order
3917 to improve the efficiency.
3920 Frequency : MDS failover only, very rarely
3922 Description: "went back in time" messages on mds failover
3923 Details : The greatest transno may be lost when the current operation
3924 finishes with an error (transno==0) and the client's last_rcvd
3925 record is over-written. Save the greatest transno in the
3926 mds_last_transno for this case.
3929 Frequency : always for specific kernels and striping counts
3931 Description: client may get "Matching packet too big" without ACL support
3932 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3933 when trying to access files in certain configurations. The
3934 clients should in fact be denied when mounting because they do
3935 not understand ACLs.
3938 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3940 Description: many clients connecting with IO in progress causes connect timeouts
3941 Details : Avoid synchronous journal commits to avoid delays caused by many
3942 clients connecting/disconnecting when bulk IO is in progress.
3943 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3944 OST_IO_PORTAL to avoid delays behind potentially many pending
3948 Frequency : occasionally with multiple writers to a single file
3950 Description: shared writes to file may result in wrong size reported by stat()
3951 Details : Allow growing of kms when extent lock is cancelled
3954 Frequency : always with random mmap IO to multi-striped file
3956 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3957 Details : Only if the hole is at the end of OST object so that kms is too
3958 small. Fix is to increase kms accordingly in ll_nopage.
3961 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3963 Description: writes to a missing object would leak memory on the OST
3964 Details : If there is an inconsistency between the MDS and OST filesystems,
3965 such that the MDS references an object that doesn't exist, writes
3966 to that object will leak memory due to incorrect cleanup in the
3967 error handling path, eventually running out of memory on the OST.
3972 Description: Creating too long symlink causes lustre errors
3973 Details : Check symlink and name lengths before sending requests to MDS.
3976 Frequency : only if flock is enabled (not on by default)
3978 Description: posix locks not released on fd closure on 2.6.9+
3979 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3980 kernels, this caused such locks not to be released on fd close and
3981 then assertions on fs unmount about still used locks.
3984 Frequency : MDS failover only, very rarely
3986 Description: clients may get ASSERTION(granted_lock != NULL)
3987 Details : When request was taking a long time, and a client was resending
3988 a getattr by name lock request. The were multiple lock requests
3989 with the same client lock handle and
3990 mds_getattr_name->fixup_handle_for_resent_request found one of the
3991 lock handles but later failed with ASSERTION(granted_lock != NULL).
3996 Description: handle->h_buffer_credits > 0, assertion failure
3997 Details : h_buffer_credits is zero after truncate, causing assertion
3998 failure. This patch extends the transaction or creates a new
4002 Frequency : NFS re-export or patchless client
4003 Bugzilla : 11179, 10796
4004 Description: Crash on NFS re-export node (__d_move)
4005 Details : We do not want to hash the dentry if we don't have a lock.
4006 But if this dentry is later used in d_move, we'd hit uninitialised
4007 list head d_hash, so we just do this to init d_hash field but
4008 leave dentry unhashed.
4011 Frequency : NFS re-export or patchless client
4013 Description: NFS exports has problem with symbolic link
4014 Details : lustre client didn't properly install dentry when re-exported
4015 to NFS or running patchless client.
4018 Frequency : NFS re-export or patchless client
4020 Description: Various nfs/patchless fixes.
4021 Details : fixes reuse disconected alias for lookup process - this fixes
4022 warning "find_exported_dentry: npd != pd",
4023 fix permission error with open files at nfs.
4024 fix apply umask when do revalidate.
4027 Frequency : occasional
4029 Description: Crash on NFS re-export node
4030 Details : calling clear_page() on the wrong pointer triggered oops in
4031 generic_mapping_read().
4034 Frequency : rarely, using O_DIRECT IO
4036 Description: unaligned directio crashes client with LASSERT
4037 Details : check for unaligned buffers before trying any requests.
4040 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4042 Description: stale data returned from RAID cache
4043 Details : If only a small amount of IO is done to the RAID device before
4044 reading it again it is possible to get stale data from the RAID
4045 cache instead of reading it from disk.
4048 Frequency : always for sles10 kernel
4050 Description: sles10 support
4051 Details : ll_follow_link: compile fixes and using of nd_set_link
4052 under newer kernels.
4055 Frequency : depends on arch, kernel and compiler version, always on sles10
4058 Description: recursive or deep enough symlinks cause stack overflow
4059 Details : getting rid of large stack-allocated variable in
4063 Frequency : depends on hardware
4065 Description: lustre write performance loss in the SLES10 kernel
4066 Details : the performance loss is caused by using of write barriers in the
4067 ext3 code. The SLES10 kernel turns barrier support on by
4068 default. The fix is to undo that change for ldiskfs.
4070 ------------------------------------------------------------------------------
4072 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4074 * Support for kernels:
4075 2.6.9-42.0.3EL (RHEL 4)
4076 2.6.5-7.276 (SLES 9)
4077 2.4.21-47.0.1.EL (RHEL 3)
4078 2.6.12.6 vanilla (kernel.org)
4082 Frequency : quota enabled and large files being deleted
4084 Description: releasing more than 4GB of quota at once hangs OST
4085 Details : If a user deletes more than 4GB of files on a single OST it
4086 will cause the OST to spin in an infinite loop. Release
4087 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4092 Description: statfs data retrieved from /proc may be stale or zero
4093 Details : When reading per-device statfs data from /proc, in the
4094 {kbytes,files}_{total,free,avail} files, it may appear
4095 as zero or be out of date.
4098 Frequency : systems with MD RAID1 external journal devices
4100 Description: lconf's call to blkid is confused by RAID1 journal devices
4101 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4102 of returning all block devices that match the journal UUID.
4105 Frequency : always, for aggregate stripe size over 4GB
4107 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4108 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4109 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4112 Frequency : always if "lfs find" used on a local file/directory
4114 Description: "lfs find" segfaults if used on a local file/directory
4115 Details : The case where a directory component was not specified wasn't
4116 handled correctly. Handle this properly.
4119 Frequency : always on ppc64
4121 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4122 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4123 perform endianness conversion.
4126 Frequency : rarely (truncate to non-zero file size after write under load)
4127 Bugzilla : 10730, 10687
4128 Description: Files padded with zeros to next 4K multiple
4129 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4130 it is possible that files that are truncated to a non-zero size
4131 immediately after being written are filled with zero bytes beyond
4132 the truncated size. No file data is lost.
4134 Severity : enhancement
4136 Description: Allow recovery/failover for liblustre clients.
4137 Details : liblustre clients were unaware of failover configurations until
4140 Severity : enhancement
4142 Description: user file locks should fail when not mounting with flock option
4143 Details : Set up an error-returning stub in ll_file_operations.lock field
4144 to prevent incorrect behaviour when client is mounted without
4145 flock option. Also, set up properly f_op->flock field for
4149 Frequency : always on ia64
4151 Description: "lfs df" loops on printing out MDS statfs information
4152 Details : The obd_ioctl_data was not initialized and in some systems
4153 this caused a failure during the ioctl that did not return
4154 an error. Initialize the struct and return an error on failure.
4157 Frequency : SLES 9 only
4159 Description: Error of copying files with lustre special EAs as root
4160 Details : Client side always return success for setxattr call for lustre
4161 special xattr (currently only "trusted.lov").
4164 Frequency : rarely on clusters with both ia64+i386 clients
4166 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4167 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4168 clients are doing concurrent writes to the same file it is
4169 possible that the ia64 clients may LASSERT because the OST
4170 extent locks are not PAGE_SIZE aligned. Ensure that grown
4171 locks are always aligned on the request boundary.
4174 Frequency : specific use, occasional
4176 Description: Overwriting in use executable truncates on-disk binary image
4177 Details : If one node attempts to overwrite an executable in use by
4178 another node, we now correctly return ETXTBSY instead of
4179 truncating the file.
4181 Severity : enhancement
4183 Description: Async OSC create to avoid the blocking unnecessarily.
4184 Details : If an OST has no remaining object, system will block on the creating
4185 when need to create a new object on this OST. Now, ways use
4186 pre-created objects when available, instead of blocking on an
4187 empty osc while others are not empty. If we must block, we block
4188 for the shortest possible period of time.
4193 Description: chmod on Lustre root is propagated to other clients
4194 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4195 invalid by the follow_mount time.
4198 Frequency : liblustre clients only
4200 Description: Race in 'instant cancel' lock handling could lead to such locks
4201 never to be granted in case of SMP MDS
4202 Details : Do not destroy not yet granted but cbpending locks in
4206 Frequency : replay/resend of open
4208 Description: non null lock assetion failure in mds_intent_policy
4209 Details : Trying to replay/resend lockless open requests resulted in
4210 mds_open() returning 0 with no lock. Now it sets a flag if
4211 a lock is going to be returned.
4213 Severity : enhancement
4215 Description: Checksum enhancements
4216 Details : New checksum enhancements allow for resending RPCs that failed
4219 Severity : enhancement
4221 Description: Tunables on number of dirty pages in cacche
4222 Details : Allow to set limit on number of dirty pages cached.
4227 Description: client crash on unmount - lock still has references
4228 Details : In some error handling cases it was possible to leak a lock
4229 reference on a client while accessing a file. This was not
4230 harmful to the client during operation, but would cause the
4231 client to crash when the filesystem is unmounted.
4234 Frequency : specific case, rare
4236 Description: ETXTBSY on mds though file not in use by client
4237 Details : ETXTBSY is no longer incorrectly returned when attempting to
4238 chmod or chown a directory that the user previously tried to
4239 execute or a currently-executing binary.
4242 Frequency : extremely rare except on liblustre-based clients
4244 Description: Lustre space not freed when files are deleted
4245 Details : Clean up open-unlinked files after client eviction. Previously
4246 the unlink was skipped and the files remained as orphans.
4251 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4252 Details : In some cases it was possible to send a blocking callback to
4253 a client doing a glimpse, even though that client didn't get
4254 a lock granted. When the glimpse lock is cancelled on the OST
4255 the freed lock is left on the waiting list and corrupted the list.
4258 Frequency : all core dumps
4260 Description: Broke core dumps to lustre
4261 Details : Negative dentry may be unhashed if parent does not have UPDATE
4262 lock, but some callers, e.g. do_coredump, expect dentry to be
4263 hashed after successful create, hash it in ll_create_it.
4265 ------------------------------------------------------------------------------
4267 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4269 * Support for kernels:
4270 2.6.9-42.0.2.EL (RHEL 4)
4271 2.6.5-7.276 (SLES 9)
4272 2.4.21-40.EL (RHEL 3)
4273 2.6.12.6 vanilla (kernel.org)
4277 Frequency : always on RHEL 3
4279 Description: Number of open files grows over time
4280 Details : The number of open files grows over time, whether or not
4281 Lustre is started. This was due to a filp leak introduced
4282 by one of our kernel patches.
4284 ------------------------------------------------------------------------------
4286 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4288 * Support for kernels:
4289 2.6.9-42.EL (RHEL 4)
4290 2.6.5-7.276 (SLES 9)
4291 2.4.21-40.EL (RHEL 3)
4292 2.6.12.6 vanilla (kernel.org)
4297 Bugzilla : 5719, 9635, 9792, 9684
4298 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4299 Details : If a server is under heavy load and cannot reply to new
4300 connection requests before the client resends the (re)connect,
4301 the connection handling code can behave badly if two service
4302 threads are concurrently handing separate (re)connections from
4303 the same client. Add better locking to the connection handling
4304 code, and ensure that only a single connection will be processed
4305 for a given client UUID, even if the lock is dropped.
4307 Severity : enhancement
4309 Description: add TCP zero-copy support to kernel
4310 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4311 sends if the hardware supports scatter-gather and checksumming.
4312 This allows socklnd to do client-write and server-read more
4313 efficiently and reduce CPU utilization from skbuf copying.
4316 Frequency : only if NFS exporting from client
4318 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4319 Details : If setting attributes on a file created under NFS that had
4320 never been opened it would be possible to oops the client
4321 if the file had no objects.
4324 Frequency : always for liblustre
4326 Description: liblustre client does MDS+OSTs setattr RPC for each write
4327 Details : When doing a write from a liblustre client, the client
4328 incorrectly issued an RPC to the MDS and each OST the file was
4329 striped over in order to update the timestamps. When writing
4330 with small chunks and many clients this could overwhelm the MDS
4331 with RPCs. In all cases it would slow down the write because
4332 these RPCs are unnecessary.
4334 Severity : enhancement
4336 Description: allow number of MDS service threads to be changed at module load
4337 Details : It is now possible to change the number of MDS service threads
4338 running. Adding "options mds mds_num_threads={N}" to the MDS's
4339 /etc/modprobe.conf will set the number of threads for the next
4340 time Lustre is restarted (assuming the "mds" module is also
4341 reloaded at that time). The default number of threads will
4342 stay the same, 32 for most systems.
4347 Description: OST crash if filesystem is unformatted or corrupt
4348 Details : If an OST is started on a device that has never been formatted
4349 or if the filesystem is corrupt and cannot even mount then the
4350 error handling cleanup routines would dereference a NULL pointer.
4355 Description: NULL pointer deref in llap_from_page.
4356 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4357 incorrect) mapping if the page was truncated/reclaimed while it was
4358 searched for. Check for this condition and skip such pages when
4359 doing readahead. Introduce extra check to llap_from_page() to
4360 verify page->mapping->host is non-NULL (so page is not anonymous).
4363 Frequency : Sometimes when using sys_sendfile
4365 Description: "page not covered by a lock" warnings from ll_readpage
4366 Details : sendfile called ll_readpage without right page locks present.
4367 Now we introduced ll_file_sendfile that does necessary locking
4368 around call to generic_file_sendfile() much like we do in
4372 Frequency : with certain MDS communication failures at client mount time
4374 Description: NULL pointer deref after failed client mount
4375 Details : a client connection request may delayed by the network layer
4376 and not be sent until after the PTLRPC layer has timed out the
4377 request. If the client fails the mount immediately it will try
4378 to clean up before the network times out the request. Add a
4379 reference from the request import to the obd device and delay
4380 the cleanup until the network drops the request.
4383 Frequency : occasionally during client (re)connect
4385 Description: assertion failure during client (re)connect
4386 Details : processing a client connection request may be delayed by the
4387 client or server longer than the client connect timeout. This
4388 causes the client to resend the connection request. If the
4389 original connection request is replied in this interval, the
4390 client may trip an assertion failure in ptlrpc_connect_interpret()
4391 which thought it would be the only running connect process.
4394 Frequency : only with obd_echo servers and clients that are rebooted
4396 Description: kernel BUG accessing uninitialized data structure
4397 Details : When running an obd_echo server it did not start the ping_evictor
4398 thread, and when a client was evicted an uninitialized data
4399 structure was accessed. Start the ping_evictor in the RPC
4400 service startup instead of the OBD startup.
4402 Severity : enhancement
4403 Bugzilla : 10193 (patchless)
4404 Description: Remove dependency on various unexported kernel interfaces.
4405 Details : No longer need reparent_to_init, exit_mm, exit_files,
4406 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4409 Frequency : rare (only users of deprecated and unsupported LDAP config)
4411 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4412 Details : LDAP apparently contains 'lustreName' attributes instead of
4413 'name'. A simple remapping of the name is sufficient.
4416 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4418 Description: waiting_locks_callback trips kernel BUG if client is evicted
4419 Details : Running with the dump_on_timeout debug flag turned on makes
4420 it possible that the waiting_locks_callback() can try to dump
4421 the Lustre kernel debug logs from an interrupt handler. Defer
4422 this log dumping to the expired_lock_main() thread.
4424 Severity : enhancement
4426 Description: Support NFS exporting on 2.6 kernels.
4427 Details : Implement non-rawops metadata methods for NFS server to use without
4428 changing NFS server code.
4431 Frequency : very rare (synthetic metadata workload only)
4433 Description: two racing renames might cause an MDS thread to deadlock
4434 Details : Running the "racer" program may cause one MDS thread to rename
4435 a file from being the source of a rename to being the target of
4436 a rename at exactly the same time that another thread is doing
4437 so, and the second thread has already enqueued these locks after
4438 doing a lookup of the target and is trying to relock them in
4439 order. Ensure that we don't try to re-lock the same resource.
4442 Frequency : only very large systems with liblustre clients
4444 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4445 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4446 all OSTs in parallel. This allows the eviction of stale liblustre
4447 clients to proceed much faster than if they were done in series,
4448 and also offers similar improvements for other set_info RPCs.
4453 Description: excessive CPU usage during initial read phase on client
4454 Details : During the initial read phase on a client, it would agressively
4455 retry readahead on the file, consuming too much CPU and impacting
4456 performance (since 1.4.5.8). Improve the readahead algorithm
4457 to avoid this, and also improve some other common cases (read
4458 of small files in particular, where "small" is files smaller than
4459 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4464 Description: MDS crash when receiving packet with unknown intent.
4465 Details : Do not LBUG in unknown intent case, just return -EFAULT
4467 Severity : enhancement
4468 Bugzilla : 9293, 9385
4469 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4470 Details : Do not serialize getattr (non-intent version) and statfs.
4473 Frequency : occasional, when OST network is overloaded/intermittent
4475 Description: client evicted by OST after bulk IO timeout
4476 Details : If a client sends a bulk IO request (read or write) the OST
4477 may evict the client if it is unresposive to its data GET/PUT
4478 request. This is incorrect if the network is overloaded (takes
4479 too long to transfer the RPC data) or dropped the OST GET/PUT
4480 request. There is no need to evict the client at all, since
4481 the pinger and/or lock callbacks will handle this, and the
4482 client can restart the bulk request.
4485 Frequency : Always when mmapping file with no objects
4487 Description: client crashes when mmapping file with no objects
4488 Details : Check that we actually have objects in a file before doing any
4489 operations on objects in ll_vm_open, ll_vm_close and
4495 Description: Request leak when working with deleted CWD
4496 Details : Introduce advanced request refcount tracking for requests
4497 referenced from lustre intent.
4499 Severity : Enhancement
4501 Description: Cache open file handles on client.
4502 Details : MDS now will return special lock along with openhandle, if
4503 requested and client is allowed to hold openhandle, even if unused,
4504 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4505 closing files for every read/write openration.
4507 Severity : Enhancement
4509 Description: Cache open negative dentries on client when possible.
4510 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4511 negative dentries on lock revocation.
4516 Description: Remounting a client read-only wasn't possible with a zconf mount
4517 Details : It wasn't possible to remount a client read-only with llmount.
4519 Severity : enhancement
4520 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4521 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4522 MPICH is included to add Lustre-specific ADIO interfaces.
4523 This is based closely on the UFS ADIO layer and only differs
4524 in file creation, in order to allow the OST striping to be set.
4525 This is user-contributed code and not supported by CFS.
4530 Description: extended inode attributes (immutable, append-only) work improperly
4531 when 2.4 and 2.6 kernels are used on client/server or vice versa
4532 Details : Introduce kernel-independent values for these flags.
4534 Severity : enhancement
4537 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4538 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4539 granularity. Now, Fractional MB granularity is be supported,
4540 this is very useful for low memory system.
4542 Severity : enhancement
4544 Description: Getattr by fid
4545 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4546 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4547 have locks and avoids getattr rpc requests for it on every stat.
4550 Frequency : Always, for filesystems larger than 2TB
4552 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4553 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4554 filesystems larger than 2TB to be mounted with mballoc enabled.
4557 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4559 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4560 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4561 filesystems larger than 2TB is mounted then the kernel will
4562 silently corrupt the start of the filesystem. CONFIG_LBD is
4563 enabled for all CFS-supported kernels, but the possibility of
4564 this happening with a modified kernel config exists.
4566 Severity : enhancement
4568 Description: add client O_DIRECT support for 2.6 kernels
4569 Details : It is now possible to do O_DIRECT reads and writes to files
4570 in the Lustre client mountpoint on 2.6 kernel clients.
4572 Severity : enhancement
4574 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4575 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4576 parallel, not waiting for response from every OST before sending
4577 a rpc to the next OST.
4582 Description: setattr vs write race when updating file timestamps
4583 Details : Client processes that update a file timestamp into the past
4584 right after writing to the file (e.g. tar) it is possible that
4585 the updated file modification time can be reset to the current
4586 time due to a race between processing the setattr and write RPC.
4588 Severity : enhancement
4590 Description: Bring 'lfs find' closer in line with regular Linux find.
4591 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4592 -print0 options and obtains all the needed info through the lustre
4595 Severity : enhancement
4597 Description: support up to 1024 configured devices on one node
4598 Details : change obd_dev array from statically allocated to dynamically
4599 allocated structs as they are first used to reduce memory usage
4604 Description: Flush dirty partially truncated pages during truncate
4605 Details : Immediatelly flush partially truncated pages in filter_setattr,
4606 this way we completely avoid having any pages in page cache on OST
4607 and can retire ugly workarounds during writes to flush such pages.
4612 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4613 Details : protect inode from truncation within vfs_unlink() context
4614 just take a reference before calling vfs_unlink() and release it
4615 when parent's i_sem is free.
4620 Description: last_id value checked outside lock on OST caused LASSERT failure
4621 Details : If there were multiple MDS->OST object precreate requests in
4622 flight, it was possible that the OST's last object id was checked
4623 outside a lock and incorrectly tripped an assertion. Move checks
4624 inside locks, and discard old precreate requests.
4627 Frequency : always, if extents are used on OSTs
4629 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4630 Details : OSTs using the extents format would not zero the high 16 bits of
4631 the index physical block number. This is not a problem for any
4632 OST filesystems smaller than 16TB, and no kernels support ext3
4633 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4634 new/modified files) and can be fixed for existing filesystems
4635 with e2fsprogs-1.39-cfs1.
4640 Description: import connection selection may be incorrect if timer wraps
4641 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4642 import connections to be ignored if the 32-bit jiffies counter
4643 wraps. Use a 64-bit jiffies counter.
4646 Frequency : very large clusters immediately after boot
4648 Description: LNET request buffers exhausted under heavy short-term load
4649 Details : If a large number of client requests are generated on a service
4650 that has previously never seen so many requests it is possible
4651 that the request buffer growth cannot keep up with the spike in
4652 demand. Instead of dropping incoming requests, they are held in
4653 the LND until the RPC service can accept more requests.
4656 Frequency : Sometimes during replay
4658 Description: Assertion failure in ll_local_open after replay.
4659 Details : If replay happened on an open request reply before we were able
4660 to set replay handler, reply will become not swabbed tripping the
4661 assertion in ll_local_open. Now we set the handler right after
4662 recognising of open request
4665 Frequency : very rare
4667 Description: kernel reports "badness in vsnprintf"
4668 Details : Reading from the "recovery_status" /proc file in small chunks
4669 may cause a negative length in lprocfs_obd_rd_recovery_status()
4670 call to vsnprintf() (which is otherwise harmless). Exit early
4671 if there is no more space in the output buffer.
4673 Severity : enhancement
4675 Description: clear OBD RPC statistics by writing to them
4676 Details : It is now possible to clear the OBD RPC statistics by writing
4677 to the "stats" file.
4682 Description: Client mtime is not the same on different clients after utimes
4683 Details : In some cases, the client was using the utimes() syscall on
4684 a file cached on another node. The clients now validate the
4685 ctime from the MDS + OSTs to determine which one is right.
4690 Description: Inability to activate failout mode
4691 Details : lconf script incorrectly assumed that in pythong string's numeric
4692 value is used in comparisons.
4695 Frequency : always with multiple stripes per file
4697 Description: Inefficient object allocation for mutli-stripe files
4698 Details : When selecting which OSTs to stripe files over, for files with
4699 a stripe count that divides evenly into the number of OSTs,
4700 the MDS is always picking the same starting OST for each file.
4701 Return the OST selection heuristic to the original design.
4706 Description: mount failures may take full timeout to return an error
4707 Details : Under some heavy load conditions it is possible that a
4708 failed mount can wait for the full obd_timeout interval,
4709 possibly several minutes, before reporting an error.
4710 Instead return an error as soon as the status is known.
4712 Frequency : quota enabled and large files being deleted
4714 Description: releasing more than 4GB of quota at once hangs OST
4715 Details : If a user deletes more than 4GB of files on a single OST it
4716 will cause the OST to spin in an infinite loop. Release
4717 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4722 Description: statfs data retrieved from /proc may be stale or zero
4723 Details : When reading per-device statfs data from /proc, in the
4724 {kbytes,files}_{total,free,avail} files, it may appear
4725 as zero or be out of date.
4728 Frequency : systems with MD RAID1 external journal devices
4730 Description: lconf's call to blkid is confused by RAID1 journal devices
4731 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4732 of returning all block devices that match the journal UUID.
4735 Frequency : always, for aggregate stripe size over 4GB
4737 Description: assertion fails when trying to use 4GB stripe size
4738 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4739 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4742 Frequency : always on ppc64
4744 Description: the first write on an ext3 filesystem with mballoc got stuck
4745 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4746 perform endianness conversion.
4748 ------------------------------------------------------------------------------
4750 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4752 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4753 INTEROPERATE with older versions automatically. Please read the
4754 user documentation before upgrading any part of a live system.
4755 * WARNING: Lustre networking configuration changes are required with
4756 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4759 * Support for kernels:
4760 2.6.9-22.0.2.EL (RHEL 4)
4761 2.6.5-7.244 (SLES 9)
4762 2.6.12.6 vanilla (kernel.org)
4765 Severity : enhancement
4766 Bugzilla : 7981/8208
4767 Description: Introduced Lustre Networking (LNET)
4768 Details : LNET is new networking infrastructure for Lustre, it includes
4769 a reorganized network configuration mode (see the user
4770 documentation for full details) as well as support for routing
4771 between different network fabrics. Lustre Networking Devices
4772 (LNDs) for the supported network fabrics have also been
4773 created for this new infrastructure.
4775 Severity : enhancement
4776 Description: Introduced Access control lists
4777 Details : clients can set ACLs on files and directories in order to have
4778 more fine-grained permissions than the standard Unix UGO+RWX.
4779 The MDS must be started with the "-o acl" mount option.
4781 Severity : enhancement
4782 Description: Introduced filesystem quotas
4783 Details : Administrators may now establish per-user quotas on the
4786 Severity : enhancement
4788 Description: Configuration change for the XT3
4789 The PTLLND is now used to run Lustre over Portals on the XT3
4790 The configure option(s) --with-cray-portals are no longer used.
4791 Rather --with-portals=<path-to-portals-includes> is used to
4792 enable building on the XT3. In addition to enable XT3 specific
4793 features the option --enable-cray-xt3 must be used.
4798 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4799 Details : A race between allocating a new llcd and re-getting the llcd_lock
4800 allowed another thread to grab newly-allocated llcd.
4802 Severity : enhancement
4804 Description: 2.6 OST async journal commit and locking fix to improve performance
4805 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4806 2.6 kernels are now async as they already were in 2.4 kernels so
4807 that they can commit concurrently with the network bulk transfer.
4808 For block-allocated files the filter allocation semaphore is held
4809 to avoid filesystem fragmentation during allocation. BKL lock
4810 removed for 2.6 xattr operations where it is no longer needed.
4815 Description: lconf incorrectly determined whether two IP networks could talk
4816 Details : In some more complicated routing and multiple-network
4817 configurations, lconf will avoid trying to make a network
4818 connection to a disjoint part of the IP space. It was doing the
4819 math incorrectly for one set of cases.
4824 Description: Fix for potential infinite loop processing records in an llog.
4825 Details : If an llog record is corrupted/zeroed, it is possible to loop
4826 forever in llog_process(). Validate the llog record length
4827 and skip the remainder of the block on error.
4830 Frequency : occasional (liblustre only)
4832 Description: liblustre could not open files whose last component is a symlink
4833 Details : sysio_path_walk() would incorrectly pass the open intent to
4834 intermediate path components.
4837 Frequency : rare (liblustre only with non-standard tuning)
4838 Bugzilla : 7201 (7350)
4839 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4840 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4841 after it has been released, only do this for referenced locks
4843 Severity : enhancement
4845 Description: specifying an (invalid) directory default stripe_size of -1
4846 would reset the directory default striping
4847 Details : stripe_size -1 was used internally to signal directory stripe
4848 removal, now use "all default" to signal dir stripe removal
4849 as a directory striping of "all default" is not useful
4852 Frequency : common for large clusters running liblustre clients
4854 Description: doing an ls when liblustre clients are running is slow
4855 Details : sending a glimpse AST to a liblustre client waits for every AST
4856 to time out, as liblustre clients will not respond. Since they
4857 cannot cache data we refresh the OST lock LVB from disk instead.
4859 Severity : enhancement
4861 Description: doing an ls at the same time as file IO can be slow
4862 Details : enqueue and other "small" requests can be blocked behind many
4863 large IO requests. Create a new OST IO portal for non-IO
4864 requests so they can be processed faster.
4867 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4869 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4870 created file with mode 000
4871 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4872 allow an ftruncate() as a non-root user to a file with mode 000.
4873 Lustre now respects this flag to disable mode checks when
4874 truncating a file owned by the user
4877 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4880 Description: Revoking locks from clients that went dead or catatonic might take
4882 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4883 cancellation of such locks instant on servers without waiting for
4884 any reply from clients. Clients drops these locks when cancel
4885 notification from server is received without replying.
4888 Frequency : liblustre-only, when liblustre client dies or becomes busy
4890 Description: Doing ls on Linux clients can take a long time with active
4892 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4893 granting such locks to it in the first place if possible. Locks
4894 are taken by proxy on the OST during the read or write and
4895 dropped immediately afterward. Add connect flags handling, do
4896 not grant locks to liblustre clients for glimpse ASTs.
4898 Severity : enhancement
4900 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4901 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4902 sequential reads and behaves badly with applications doing random
4903 reads. Improve it by reading ahead at least read region, and
4904 avoiding excessive large RPC for small reads.
4906 Severity : enhancement
4908 Description: Creating more than 1000 files for a single job may cause a load
4909 imbalance on the OSTs if there are also a large number of OSTs.
4910 Details : qos_prep_create() uses an OST index reseed value that is an
4911 even multiple of the number of available OSTs so that if the
4912 reseed happens in the middle of the object allocation it will
4913 still utilize the OSTs as uniformly as possible.
4918 Description: OST or MDS may oops in ping_evictor_main()
4919 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4920 but doesn't restart at beginning of obd_exports_timed list
4923 Severity : enhancement
4925 Description: improve by-nid export eviction on the MDS and OST
4926 Details : allow multiple exports with the same NID to be evicted at one
4927 time without re-searching the exports list.
4930 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4932 Description: MDS may oops in groups_free()
4933 Details : in rare race conditions a newly allocated group_info struct is
4934 freed again, and this can be NULL. The 2.4 compatibility code
4935 for groups_free() checked for a NULL pointer, but 2.6 did not.
4938 Frequency : common for liblustre clients doing little filesystem IO
4939 Bugzilla : 9352, 7313
4940 Description: server may evict liblustre clients accessing contended locks
4941 Details : if a client is granted a lock or receives a completion AST
4942 with a blocking AST already set it would not reply to the AST
4943 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4946 Frequency : lfs setstripe, only systems with more than 160 OSTs
4948 Description: unable to set striping with a starting offset beyond OST 160
4949 Details : llapi_create_file() incorrectly limited the starting stripe
4950 index to the maximum single-file stripe count.
4953 Frequency : LDAP users only
4955 Description: lconf did not handle in-kernel recovery with LDAP properly
4956 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4958 Severity : enhancement
4960 Description: bind OST threads to NUMA nodes to improve performance
4961 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4962 node and do their allocations there to localize memory access
4964 Severity : enhancement
4966 Description: llmount can determine client NID directly from Myrinet (GM)
4967 Details : the client NID code from gmnalnid was moved directly into
4968 llmount, removing the need to use this or specifying the
4969 client NID explicitly when mounting GM clients with zeroconf
4972 Frequency : if client is started with down MDS
4974 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4975 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4976 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4977 ptlrpc_import_delayed_req() because we didn't check if the
4978 request was interrupted, and we also didn't break out of the
4979 event loop if there was no timeout
4984 Description: data loss during non-page-aligned writes to a single file from
4985 both multiple nodes and multiple threads on one node at same time
4986 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4987 inconsistency led to false short-reads, that were cached and later
4988 used by ->prepare_write() to fill in partially written page,
4989 leading to data loss.
4992 Frequency : always, if lconf --abort_recovery used
4994 Description: lconf --abort_recovery fails with 'Operation not supported'
4995 Details : lconf was attempting to abort recovery on the MDT device and not
4998 Severity : enhancement
5000 Description: remove cleanup logs
5001 Details : replace lconf-generated cleanup logs with lustre internal
5002 cleanup routines. Eliminates the need for client-cleanup and
5005 Severity : enhancement
5007 Description: add support for EAs (user and system) on lustre filesystems
5008 Details : it is now possible to store extended attributes in the Lustre
5009 client filesystem, and with the user_xattr mount option it
5010 is possible to allow users to store EAs on their files also
5012 Severity : enhancement
5014 Description: Add possibility (config option) to show minimal available OST free
5016 Details : When compiled with --enable-mindf configure option, statfs(2)
5017 (and so, df) will return least minimal free space available from
5018 all OSTs as amount of free space on FS, instead of summary of
5019 free spaces of all OSTs.
5021 Severity : enhancement
5023 Description: do not expand extent locks acquired on OST-side
5024 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5025 by server: they are not cached anyway.
5028 Frequency : when mmap is used/binaries executed from Lustre
5030 Description: Unmmap pages before throwing them away from read cache.
5031 Details : llap_shrink cache now attempts to unmap pages before discarding
5032 them (if unmapping failed - do not discard). SLES9 kernel has
5033 extra checks that trigger if this unmapping is not done first.
5038 Description: lconf didn't resolve symlinks before checking to see whether a
5039 given mountpoint was already in use
5042 Frequency : when migrating failover services
5043 Bugzilla : 6395, 9514
5044 Description: When migrating a subset of services from a node (e.g. failback
5045 from a failover service node) the remaining services would
5046 time out and evict clients.
5047 Details : lconf --force (implied by --failover) sets the global obd_timeout
5048 to 5 seconds in order to quickly disconnect, but this caused
5049 other RPCs to time out too quickly. Do not change the global
5050 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5052 Severity : enhancement
5053 Frequency : if MDS is started with down OST
5054 Bugzilla : 9439,5706
5055 Description: Allow startup/shutdown of an MDS without depending on the
5056 availability of the OSTs.
5057 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5058 Add appropriate locking and lov-osc refcounts for safe
5059 cleaning. Add osc abort_inflight calls in case the
5060 synchronize never started.
5063 Frequency : occasional (Cray XT3 only)
5065 Description: root not authorized to access files in CRAY_PORTALS environment
5066 Details : The client process capabilities were not honoured on the MDS in
5067 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5068 been accessed by an authorized user then root was able to access
5069 the file on the local client also. The root user capabilities
5070 are now allowed on the MDS, as this environment has secure UID.
5073 Frequency : occasional
5075 Description: ldiskfs "too long searching" message happens too often
5076 Details : A debugging message (otherwise harmless) prints too often on
5077 the OST console. This has been reduced to only happen when
5078 there are fragmentation problems on the filesystem.
5083 Description: Division by zero in statfs when all OSCs are inactive
5084 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5085 lov_statfs divides by value returned by lov_get_stripecnt().
5089 Bugzilla : 9489, 3273
5090 Description: First write from each client to each OST was only 4kB in size,
5091 to initialize client writeback cache, which caused sub-optimal
5092 RPCs and poor layout on disk for the first written file.
5093 Details : Clients now request an initial cache grant at (re)connect time
5094 and so that they can start streaming writes to the cache right
5095 away and always do full-sized RPCs if there is enough data.
5096 If the OST is rebooted the client also re-establishes its grant
5097 so that client cached writes will be honoured under the grant.
5102 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5103 Details : Now I/O RPCs go to different portal number and (presumably) fast
5104 lock requests (and glimses) and other RPCs get their own service
5105 threads pool that should be able to service those RPCs
5108 Severity : enhancement
5110 Description: Ability to exchange lustre version between client and servers and
5111 issue warnings at client side if client is too old. Also for
5112 liblustre clients there is ability to refuse connection of too old
5114 Details : New 'version' field is added to connect data structure that is
5115 filled with version info. That info is later checked by server and
5119 Frequency : rare, liblustre only.
5120 Bugzilla : 9296, 9581
5121 Description: Two simultaneous writes from liblustre at offset within same page
5122 might proceed at the same time overwriting eachother with stale
5124 Details : I/O lock withing llu_file_prwv was released too early, before data
5125 actually was hitting the wire. Extended lock-holding time until
5126 server acknowledges receiving data.
5129 Frequency : extremely rare. Never observed in practice.
5131 Description: avoid generating lustre_handle cookie of 0.
5132 Details : class_handle_hash() generates handle cookies by incrementing
5133 global counter, and can hit 0 occasionaly (this is unlikely, but
5134 not impossible, because initial value of cookie counter is
5135 selected randonly). Value of 0 is used as a sentinel meaning
5136 "unassigned handle" --- avoid it. Also coalesce two critical
5137 sections in this function into one.
5139 Severity : enhancement
5141 Description: allow liblustre clients to delegate truncate locking to OST
5142 Details : To avoid overhead of locking, liblustre client instructs OST to
5143 take extent lock in ost_punch() on client's behalf. New connection
5144 flag is added to handle backward compatibility.
5146 Severity : enhancement
5147 Bugzilla : 4928, 7341, 9758
5148 Description: allow number of OST service threads to be specified
5149 Details : a module parameter allows the number of OST service threads
5150 to be specified via "options ost ost_num_threads={N}" in the
5151 OSS's /etc/modules.conf or /etc/modprobe.conf.
5155 Bugzilla : 6146, 9635, 9895
5156 Description: servers crash with bad pointer in target_handle_connect()
5157 Details : In rare cases when a client is reconnecting it was possible that
5158 the connection request was the last reference for that export.
5159 We would temporarily drop the export reference and get a new
5160 one, but this may have been the last reference and the export
5161 was just destroyed. Get new reference before dropping old one.
5163 Severity : enhancement
5164 Frequency : if client is started with failover MDS
5166 Description: Allow multiple MDS hostnames in the mount command
5167 Details : Try to read the configuration from all specified MDS
5168 hostnames during a client mount in case the "primary"
5171 Severity : enhancement
5173 Description: Stop sending data to evicted clients as soon as possible.
5174 Details : Check if the client we are about to send or are sending data to
5175 was evicted already. (Check is done every second of waiting,
5176 for which l_wait_event interface was extended to allow checking
5177 of exit condition at specified intervals).
5180 Frequency : rare, normally only when NFS exporting is done from client
5182 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5184 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5187 Frequency : always, if srand() is called before liblustre initialization
5189 Description: Liblustre uses system PRNG disturbing its usage by user application
5190 Details : Introduce internal to lustre fast and high-quality PRNG for
5191 lustre usage and make liblustre and some other places in generic
5192 lustre code to use it.
5194 Severity : enhancement
5195 Bugzilla : 9477, 9557, 9870
5196 Description: Verify that the MDS configuration logs are updated when xml is
5197 Details : Check if the .xml configuration logs are newer than the config
5198 logs stored on the MDS and report an error if this is the case.
5199 Request --write-conf, or allow starting with --old_conf.
5201 Severity : enhancement
5203 Description: Handle symlinks in the path when checking if Lustre is mounted.
5204 Details : Resolve intermediate symlinks when checking if a client has
5205 mounted a filesystem to avoid duplicate client mounts.
5210 Description: lconf can hit an error exception but still return success.
5211 Details : The lconf command catches the Command error exception at the top
5212 level script context and will exit with the associated exit
5213 status, but doesn't ensure that this exit status is non-zero.
5218 Description: failure of ptlrpc thread startup can cause oops
5219 Details : Starting a ptlrpc service thread can fail if there are a large
5220 number of threads or the server memory is very fragmented.
5221 Handle this without oopsing.
5224 Frequency : always, only if liblustre and non-default acceptor port was used
5226 Description: liblustre cannot connect to servers with non-default acceptor port
5227 Details : tcpnal_set_default_params() was not called and was therefore
5228 ignoring the environment varaible TCPNAL_PORT, as well as other
5229 TCPNAL_ environment variables
5234 Description: two objects could be created on the same OST for a single file
5235 Details : If an OST is down, in some cases it was possible to create two
5236 objects on a single OST for a single file. No problems other
5237 than potential performance impact and spurious error messages.
5241 Bugzilla : 5681, 9562
5242 Description: Client may oops in ll_unhash_aliases
5243 Details : Client dcache may become inconsistent in race condition.
5244 In some cases "getcwd" can fail if the current directory is
5250 Description: Inode refcounting problems in NFS export code
5251 Details : link_raw functions used to call d_instantiate without obtaining
5252 extra inode reference first.
5256 Bugzilla : 9942, 9903
5257 Description: Referencing freed requests leading to crash, memleaks with NFS.
5258 Details : We used to require that call to ll_revalidate_it was always
5259 followed by ll_lookup_it. Also with revalidate_special() it is
5260 possible to call ll_revalidate_it() twice for the same dentry
5261 even if first occurence returned success. This fix changes semantic
5262 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5263 reference on a request referred from the intent.
5264 ll_intent_release() then releases such a request.
5267 Frequency : rare, normally benchmark loads only
5269 Description: unlinked inodes were kept in memory on the client
5270 Details : If a client is repeatedly creating and unlinking files it
5271 can accumulate a lot of stale inodes in the inode slab cache.
5272 If there is no other client load running this can cause the
5273 client node to run out of memory. Instead flush old inodes
5274 from client cache that have the same inode number as a new inode.
5277 Frequency : SLES9 2.6.5 kernel and long filenames only
5278 Bugzilla : 9969, 10379
5279 Description: utime reports stale NFS file handle
5280 Details : SLES9 uses out-of-dentry names in some cases, which confused
5281 the lustre dentry revalidation. Change it to always use the
5285 Frequency : rare, unless heavy write-truncate concurrency is continuous
5286 Bugzilla : 4180, 6984, 7171, 9963, 9331
5287 Description: OST becomes very slow and/or deadlocked during object unlink
5288 Details : filter_destroy() was holding onto the parent directory lock
5289 while truncating+unlinking objects. For very large objects this
5290 may block other threads for a long time and slow overall OST
5291 responsiveness. It may also be possible to get a lock ordering
5292 deadlock in this case, or run out of journal credits because of
5293 the combined truncate+unlink. Solution is to do object truncate
5294 first in one transaction without parent lock, and then do the
5295 final unlink in a new transaction with the parent lock. This
5296 reduces the lock hold time dramatically.
5299 Frequency : rare, 2.4 kernels only
5301 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5302 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5303 before unmounting their filesystem in order to avoid deadlock.
5304 The kernel_locked() function in 2.4 kernels only checks whether
5305 the kernel lock is held, not whether it is this process that is
5306 holding it as 2.6 kernels do.
5311 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5312 Details : The client ptlrpc code may be trying to reconnect to a down
5313 server before a previous connection attempt has timed out.
5314 Increase the reconnect interval to be longer than the connection
5315 timeout interval to avoid sending duplicate connections to
5319 Frequency : echo_client brw_test command
5321 Description: fix echo_client to work with OST preallocated code
5322 Details : OST preallocation code (5137) didn't take echo_client IO path
5323 into account: echo_client calls filter methods outside of any
5324 OST thread and, hence, there is no per-thread preallocated
5325 pages and buffers to use. Solution: hijack pga pages for IO. As
5326 a byproduct, this avoids unnecessary data copying.
5330 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5331 Description: Client can oops in mdc_commit_close() after open replay
5332 Details : It was possible for the MDS to return an open request with no
5333 transaction number in mds_finish_transno() if the client was
5334 evicted, but without actually returning an error. Clients
5335 would later try to replay that open and may trip an assertion
5336 Simplify the client close codepath, and always return an error
5337 from the MDS in case the open is not successful.
5340 Frequency : rare, 2.6 OSTs only
5342 Description: OST may deadlock under high load on fragmented files
5343 Details : If there was a heavy load and highly-fragmented OST filesystems
5344 it was possible to have all the OST threads deadlock waiting on
5345 allocation of biovecs, because the biovecs were not released
5346 until the entire RPC IO was completed. Instead, release biovecs
5347 as soon as they are complete to ensure forward IO progress.
5349 Severity : enhancement
5351 Description: Support for specifying external journal device at mount
5352 Details : If an OST or MDS device is formatted with an external journal
5353 device, this device major/minor is stored in the ext3 superblock
5354 and may not be valid for failover. Allow detecting and
5355 specifying the external journal at mount time.
5360 Description: Mounting an MDS with pending unlinked files may cause oops
5361 Details : target_finish_recovery() calls mds_postrecov() which returned
5362 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5363 considers this an error and immediately begins cleaning up the
5364 lov, just after starting the mds_lov process
5366 Severity : enhancement
5368 Description: Implement 'lfs df' to report actual free space on per-OST basis
5369 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5370 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5371 usage of objects; '-h' to report in human readable format.
5373 ------------------------------------------------------------------------------
5375 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5382 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5383 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5384 at startup during orphan unlinking. Instead, skip buddy update
5385 before setup, as it will be regenerated after recovery is complete.
5390 Description: If an OST is inactive, its locks might reference stale inodes.
5391 Details : lov_change_cbdata() must iterate over all namespaces, even if
5392 they are inactive to clear inode references from the lock.
5394 Severity : enhancement
5395 Frequency : occasional, if non-standard max_dirty_mb used
5397 Description: Client will block write RPCs if not enough grant
5398 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5399 then the client will block writes while waiting for another RPC
5400 to complete instead of consuming its dirty limit. With change
5401 we get improved performance when max_dirty_mb is small.
5403 Severity : enhancement
5404 Bugzilla : 3389, 6253
5405 Description: Add support for supplementary groups on the MDS.
5406 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5407 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5408 lookups for user supplementary groups into a cache.
5412 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5413 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5414 for liblustre O_EXCL is enforced by MDS.
5418 Description: readdir never returns NULL in liblustre.
5419 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5420 dentry in struct dirent.
5424 Frequency : liblustre only, depends on application IO pattern
5425 Description: liblustre clients evicted if not contacting servers
5426 Details : Don't put liblustre clients into the ping_evictor list, so
5427 they will not be evicted by the pinger ever.
5429 Severity : enhancement
5431 Description: Add ability to evict clients by NID from MDS.
5432 Details : By echoing "nid:$NID" string into
5433 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5434 $NID would be instantly evicted from this MDS and from all active
5435 OSTs connected to it.
5439 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5440 Details : lookup_it_finish() used to query file size from OSTs that was not
5445 Description: service threads change working directory to that of init
5446 Details : Starting lustre service threads may pin the working directory
5447 of the parent thread, making that filesystem busy. Threads
5448 now change to the working directory of init to avoid this.
5452 Frequency : during shutdown only
5453 Description: shutdown with a failed MDS or OST can cause unmount to hang
5454 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5459 Frequency : 2.6 only
5460 Description: chmod/chown may include an extra supplementary group
5461 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5462 supplementary group and if none is specified this is used.
5465 Bugzilla : 5479 (6816)
5467 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5468 Details : If lookup is in progress on a file that is unlinked we might try
5469 to revalidate the inode and fail in revalidate after lookup is
5470 complete and ll_file_open() enqueues the open again but
5471 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5474 Frequency : always, if lconf --abort_recovery used
5476 Description: lconf --abort_recovery fails with 'Operation not supported'
5477 Details : lconf was attempting to abort recovery on the MDT device and not
5480 ------------------------------------------------------------------------------
5482 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5487 Frequency : rare (only unsupported configurations with a node running as an
5489 Bugzilla : 6514, 5137
5490 Description: Mounting a Lustre file system on a node running as an OST could
5492 Details : OSTs now preallocates memory needed to write out data at
5493 startup, instead of when needed, to avoid having to
5494 allocate memory in possibly low memory situations.
5495 Specifically, if the file system is mounted on on OST,
5496 memory pressure could force it to try to write out data,
5497 which it needed to allocate memory to do. Due to the low
5498 memory, it would be unable to do so and the node would
5499 become unresponsive.
5501 Severity : enhancement
5503 Description: Addition of lconf --service command line option
5504 Details : lconf now accepts a '--service <arg>' option, which is
5505 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5507 Severity : enhancement
5509 Description: Failover mode is now the default for OSTs.
5510 Details : By default, OSTs will now run in failover mode. To return to
5511 the old behaviour, add '--failout' to the lmc line for OSTs.
5513 Severity : enhancement
5515 Description: Health checks are now provided for MDS and OSTs
5516 Details : Additional detailed health check information on MSD and OSTs
5517 is now provided through the procfs health_check value.
5520 Frequency : occasional, depends on IO load
5522 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5523 after numerous create/delete cycles
5524 Details : The ext3 inode allocation policy would not allocate new inodes
5525 very well on the OSTs because there are no new directories
5526 being created. Instead we look for groups with free space if
5527 the parent directories are nearly full.
5532 Description: Network or server problems during mount may cause partially
5533 mounted clients instead of returning an error.
5534 Details : The config llog parsing code may overwrite the error return
5535 code during mount error handling, returning success instead
5541 Description: MDS can fail to allocate large reply buffers
5542 Details : After long uptimes the MDS can fail to allocate large reply
5543 buffers (e.g. zconf client mount config records) due to memory
5544 fragmentation or consumption by the buffer cache. Preallocate
5545 some large reply buffers so that these replies can be sent even
5546 under memory pressure.
5550 Frequency : rare (liblustre)
5551 Description: fsx running with liblustre complained that using truncate() to
5552 extend the file doesn't work. This patch corrects that issue.
5553 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5554 ATTR_SIZE and lsm use in llu_setattr_raw.
5558 Frequency : rare, only 2.6 kernels
5559 Description: Unusual file access patterns on the MDS may result in inode
5560 data being lost in very rare circumstances.
5561 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5562 code in the 2.6 kernel caused the inode and/or EA data not to
5563 be read from disk, causing single-file corruption.
5567 Frequency : rare, only 2.6 filesystems using extents
5568 Description: Heavy concurrent write and delete load may cause data corruption.
5569 Details : It was possible under high-load situations to have an extent
5570 metadata block in the block device cache from a just-unlinked
5571 file overwrite a newly-allocated data block. We now unmap any
5572 metadata buffers that alias just-allocated data blocks.
5576 Frequency : filesystems with default stripe_count larger than 77
5577 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5578 Details : lconf specifies an inode size of 4096 bytes when the default
5579 stripe_count is larger than 77. This conflicts with the default
5580 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5581 this case to avoid pinning too much memory for large EAs.
5583 ------------------------------------------------------------------------------
5585 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5590 Frequency : rare (extremely heavy IO load with hundreds of clients)
5592 Description: Client is evicted, gets IO error writing to file
5593 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5594 caused clients to be evicted for AST timeouts. The fixes in
5595 bug 5192 mean we no longer need to have such short AST timeouts
5596 so ldlm_timeout has been increased.
5599 Frequency : occasional during --force or --failover shutdown under load
5600 Bugzilla : 5949, 4834
5601 Description: Server oops/LBUG if stopped with --force or --failover under load
5602 Details : a collection of import/export refcount and cleanup ordering
5603 issues fixed for safer force cleanup
5606 Frequency : only filesystems larger than 120 OSTs
5607 Bugzilla : 5990, 6223
5608 Description: lfs getstripe would oops on a very large filesystem
5609 Details : lov_getconfig used kfree on vmalloc'd memory
5612 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5613 Bugzilla : 6242, 6243
5614 Description: reading from files that had been truncated to a non-zero size
5615 but never opened returned no data
5616 Details : ll_file_read() reads zeros from no-object files to EOF
5621 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5622 Details : There's little harm in aborting MDS/OSS recovery and letting it
5623 try again, so I removed the LASSERT and return an error instead.
5625 Severity : enhancement
5627 Description: New debugging infrastructure for tracking down data corruption
5628 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5629 (b) cover more of the client-side code path, and (c) try to narrow
5630 down where problems occurred
5634 Bugzilla : 3819, 4364, 4397, 6313
5635 Description: Racing close and eviction MDS could cause assertion in mds_close
5636 Details : It was possible to get multiple mfd references during close and
5637 client eviction, leading to one thread referencing a freed mfd.
5639 Severity: : enhancement
5640 Bugzilla : 3262, 6359
5641 Description: Attempts to reconnect to servers are now more aggressive.
5642 Details : This builds on the enhanced upcall-less recovery that was added
5643 in 1.4.2. When trying to reconnect to servers, clients will
5644 now try each server in the failover group every 10 seconds. By
5645 default, clients would previously try one server every 25 seconds.
5650 Description: After recovery, certain operations trigger a failed
5651 assertion on a client.
5652 Details : Failing over an mds, using lconf -d --failover, while a
5653 client was doing a readdir() call would cause the client to
5654 LBUG after recovery completed and the readdir() was resent.
5656 Severity : enhancement
5658 Description: Default groups are now added by lconf
5659 Details : You can now run lconf --group <servicename> without having to
5660 manually add groups with lmc.
5663 Frequency : occasional
5665 Description: Nodes with an elan id of 0 trigger a failed assertion
5668 Frequency : always when accessing e.g. tty/console device nodes
5670 Description: tty and some other devices nodes cannot be used on lustre
5671 Details : file's private_data field is used by device data and lustre
5672 values in there got lost. New field was added to struct file to
5673 store fs-specific private data.
5676 Frequency : when exporting Lustre via NFS
5678 Description: NFSD failed occasionally when looking up a path component
5679 Details : NFSD is looking up ".." which was broken in ext3 directories
5680 that had grown large enough to become hashed.
5683 Frequency : Clusters with multiple interfaces not on the same subnet
5685 Description: Nodes will repeatedly try to reconnect to an interface which it
5686 cannot reach and report an error to the log.
5687 Details : Extra peer list entries will be created by lconf with some peers
5688 unreachable. lconf now validates the peer before adding it.
5691 Frequency : Only if a default stripe is set on the filesystem root.
5693 Description: Setting a default stripe on the filesystem root prevented the
5694 filesystem from being remounted.
5695 Details : The client was sending extra request flags in the root getattr
5696 request and did not allocate a reply buffer for the dir EA.
5699 Frequency : occasional, higher if lots of files are accessed by one client
5700 Bugzilla : 6159, 6097
5701 Description: Client trips assertion regarding lsm mismatch/magic
5702 Details : While revalidating inodes the VFS looks up inodes with ifind()
5703 and in rare cases can find an inode that is being freed.
5704 The ll_test_inode() code will free the lsm during ifind()
5705 when it finds an existing inode and then the VFS later attaches
5706 this free lsm to a new inode.
5710 Bugzilla : 6422, 7030
5711 Description: MDS deadlock between mkdir and client eviction
5712 Details : Creating a new file via mkdir or mknod (starting a transaction
5713 and getting the ns lock) can deadlock with client eviction
5714 (gets ns lock and trying to finish a synchronous transaction).
5717 Frequency : occasional
5718 Description: While starting a server, the fsfilt_ext3 module could not be
5720 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5721 kernels. Previously, lconf would still use the ext3 name
5722 when trying to load modules. Now, it will correctly use
5723 ext3 on 2.4 and ldiskfs on 2.6.
5725 Severity : enhancement
5726 Description: The default stripe count has been changed to 1
5727 Details : The interpretation of the default stripe count (0, to lfs
5728 or lmc) has been changed to mean striping across a single
5729 OST, rather than all available. For general usage we have
5730 found a stripe count of 1 or 2 works best.
5732 Severity : enhancement
5733 Description: Add support for compiling against Cray portals.
5734 Details : Conditional compiling for some areas that are different
5738 Frequency : occasional
5739 Bugzilla : 6409, 6834
5740 Description: Creating files with an explicit stripe count may lead to
5741 a failed assertion on the MDS
5742 Details : If some OSTs are full or unavailable, creating files may
5743 trigger a failed assertion on the MDS. Now, Lustre will
5744 try to use other servers or return an error to the
5748 Frequency : occasional
5750 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5751 caused lock timeout during readahead (since 1.4.2).
5752 Details : Processes doing readahead might match a lock that hasn't been
5753 granted yet if there are overlapping and conflicting lock
5754 requests. The readahead process waits on ungranted lock
5755 (original lock is CBPENDING), while OST waits for that process
5756 to cancel CBPENDING read lock and eventually evicts client.
5758 Severity : enhancement
5760 Description: Initial enabling of flock support for clients
5761 Details : Implements fcntl advisory locking and file status functions.
5762 This feature is provided as an optional mount flag (default
5763 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5764 locking are implemented yet, and those that are are not guaranteed
5765 to be completely correct in production environments.
5766 mount -t lustre -o [flock|noflock] ...
5769 Frequency : occasional
5771 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5772 assertion in the ext3 JBD (journaling) layer.
5773 Details : The b_committed_data struct is protected by the big kernel lock
5774 in 2.4 kernels, serializing journal_commit_transaction() and
5775 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5776 access to this struct. In 2.6 kernels there is finer grained
5777 locking to improve SMP performance of the JBD layer.
5781 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5786 Description: merge and cleanup kernel patches.
5789 -----------------------------------------------------------------------------
5791 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5793 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5794 versions of Lustre. Please update all servers and clients to
5795 version 1.4.2 or later at the same time. You must also run
5796 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5797 to update the configuration logs.
5799 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5800 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5801 - swab llog records of type '0' so we get proper header size/idx (5861)
5802 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5803 - fix rename of one directory over another leaking an inode (5953)
5804 - avoid SetPageDirty on 2.6 (5981)
5805 - don't re-add just-being-destroyed locks to the waiting list (5653)
5806 - when creating new directories, inherit the parent's custom
5807 striping settings if present parent (3048)
5808 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5809 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5810 - don't hold client locks on temporary worklist from l_lru (5666)
5811 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5812 - automatically evict dead clients (5921)
5813 - Update file size properly in create+truncate+fstat case (6196)
5814 - Do not unhash mountpoint dentries, do not allow removal of
5816 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5817 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5818 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5819 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5820 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5821 - Correctly handle reads of files with no objects (6243)
5822 - lctl recover will also mark a device active if deactivate used (5933)
5824 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5825 - allow --write-conf on an MDS with different nettype than client (5619)
5826 - don't write config llogs to MDS for mounts not from that MDS (5617)
5827 - lconf should create multiple TCP connections from a client (5201)
5828 - init scripts are now turned off by default; run chkconfig --on
5829 lustre and chkconfig --on lustrefs to use them
5830 - upcalls are no longer needed for clients to recover to failover
5832 - add --abort-recovery option to lconf to abort recovery on device
5834 - add support for an arbitrary number of OSTs (3026)
5835 - Quota support protocol changes.
5836 - forward compatibility changes to wire structs (6007)
5837 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5838 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5839 - improved "lustre status" script
5840 - initialize blocksize for non-regular files (6062)
5841 - added --disable-server and --disable-client configure options (5782)
5842 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5843 - Vanilla 2.4.29 support
5844 - increase maximum number of obd devices to 520 (6242)
5845 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5846 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5847 SLES 9 2.6.5 kernels
5848 - limit stripes per file to 160 (the maximum EA size) (6093)
5850 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5853 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5854 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5855 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5856 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5857 - small ext3 extents cleanups and fixes (5733)
5858 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5859 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5860 - avoid needless client->OST connect, fix handle mismatch (5317)
5861 - fix DLM error path that led to out-of-sync client, long delays (5779)
5862 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5863 - fix several locking issues related to i_size (5492,5624,5654,5672)
5864 - don't move pending lock onto export if it is already evicted (5683)
5865 - fix kernel oops when creating .foo in unlinked directory (5548)
5866 - fix deadlock in obdfilter statistics vs. object create (5811)
5867 - use time_{before,after} to avoid timer jiffies wrap (5882)
5868 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5869 - Do not leak request if server was not able to process it (5154)
5870 - If mds_open unable to find parent dir, make that negative lookup(5154)
5871 - don't create new directories with extent-mapping (5909, 5936)
5873 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5874 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5875 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5876 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5877 - support for RHEL4 kernel on i686 (5773)
5878 - provide error messages when incompatible logs are encountered (5898)
5880 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5881 * version 1.4.0.10 (1.4.1 release candidate 1)
5883 - don't keep a lock reference when lock is not granted (4238)
5884 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5885 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5886 - drop import inflight refcount on signal_completed_replay error (5255)
5887 - unlock page after async write error during send (3677)
5888 - handle missing objects in filter_preprw_read properly (5265)
5889 - no transno return for symlink open, don't save no-trasno open (3440)
5890 - don't try to complete elan receive that already failed (4012)
5891 - free RPC server reply state on error (5406)
5892 - clean up thread from ptlrpc_start_thread() on error (5160)
5893 - readahead could read extra page into cache that wasn't ejected (5388)
5894 - prevent races in class_attach/setup/cleanup/detach (5260)
5895 - don't dereference de->d_inode after l_dput of de (5458)
5896 - use "int" for stripe value returned from lock_to_stripe (5544)
5897 - mballoc allocation and error-checking fixes in 2.6 (5504)
5898 - block device patches to fix I/O request sizes in 2.6 (5482)
5899 - look up hostnames for IB nals (5602)
5900 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5901 - don't start multiple acceptors for the same port (5277)
5902 - fix incorrect LASSERT in mds_getattr_name (5635)
5903 - export a proc file for general "ping" checking (5628)
5904 - fix "lfs check" to not block when the MDS is down (5628)
5906 - service request history (4965)
5907 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5908 - create an "evict_client" /proc entry on OSTs, like the MDS has
5909 - fix mount usage message, return errors per mount(8) (5168)
5910 - change grep [] to grep "[]" in tests so they work in more UMLs
5911 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5912 - remove ancient LOV_MAGIC_V0 EA support (5047)
5913 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5914 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5915 - allow readahead allocations to fail when low on memory (5383)
5916 - mmap locking landed again, after considerable improvement (2828)
5917 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5919 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5922 - send OST transaction number in read/write reply to free req (4966)
5923 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5924 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5925 - the watchdog thread now runs as interruptible (5246)
5926 - flock/lockf fixes (but it's still disabled, pending 5135)
5927 - don't use EXT3 constants in llite code (5094)
5928 - memory shortage at startup could cause assertion (5176)
5930 - reorganization of lov code
5931 - single portals codebase
5933 - add extents/mballoc support (5025)
5934 - direct I/O reads in the obdfilter (4048)
5935 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5937 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5940 - send OST transaction number in read/write reply to free req (4966)
5941 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5942 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5943 - the watchdog thread now runs as interruptible (5246)
5944 - handle missing objects in filter_preprw_read properly (5265)
5945 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5946 - drop import inflight refcount on signal_completed_replay error (5255)
5947 - unlock page after async write error during send (3677)
5948 - return original error code on reconstructed replies (3761)
5949 - no transno return for symlink open, don't save no-trasno open (3440)
5951 - add pid to ldlm debugging output (4922)
5952 - bump the watchdog timeouts -- we can't handle 30sec yet
5953 - extra debugging for orphan dentry/inode bug (5259)
5955 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5958 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5959 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5960 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5961 - fix NULL dereference in /proc/sys/portals/routes (4827)
5962 - allow failed mdc_close() operations to be interrupted (4561)
5963 - stop precreate on OST before MDS would time out on it (4778)
5964 - don't send partial-page writes before EOF from client (4410)
5965 - discard client grant for sub-page writes on large-page clients (4520)
5966 - don't free dentries not owned by NFS code, check generation (4806)
5967 - fix lsm leak if mds_create_objects() fails (4801)
5968 - limit debug_daemon file size, always print CERROR messages (4789)
5969 - use transno after validating reply (3892)
5970 - process timed out requests if import state changes (3754)
5971 - update mtime on OST during writes, return in glimpse (4829)
5972 - add mkfsoptions to LDAP (4679)
5973 - use ->max_readahead method instead of zapping global ra (5039)
5974 - don't interrupt __l_wait_event() during strace
5976 - add software watchdogs to catch hung threads quickly (4941)
5977 - make lustrefs init script start after nfs is mounted
5978 - fix CWARN/ERROR duplication (4930)
5979 - return async write errors to application if possible (2248)
5980 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5981 - print NAL number in %x format (4645)
5982 - update barely-supported suse-2.4.21-171 series (4842)
5983 - support for sles 9 %post scripts
5984 - support for building 2.6 kernel-source packages
5985 - support for sles km_* packages
5987 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5990 - ignore -ENOENT errors in osc_destroy (3639)
5991 - notify osc create thread that OSC is being cleaned up (4600)
5992 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5993 - reconstruct ost_handle() like mds_handle() (4657)
5994 - create a new thread to do import eviction to avoid deadlock (3969)
5995 - let lconf resolve symlinked-to devices (4629)
5996 - don't unlink "objects" from directory with default EA (4554)
5997 - hold socknal file ref over connect in case target is down (4394)
5998 - allow more than 32000 subdirectories in a single directory (3244)
5999 - fix blocks count for O_DIRECT writes (3751)
6000 - OST returns ENOSPC from object create when no space left (4539)
6001 - don't send truncate RPC if file size isn't changing (4410)
6002 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6003 - bind to privileged port in socknal and tcpnal (3689)
6005 - rate limit CERROR/CWARN console message to avoid overload (4519)
6006 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6007 - basic mmap support (3918)
6008 - kernel patch series update from b1_4 (4711)
6010 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6013 - avoid crash during MDS cleanup with OST shut down (2775)
6014 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6015 - don't use bad inodes on the MDS (3744)
6016 - dynamic object preallocation to improve recovery speed (4236)
6017 - don't hold spinlock over lock dumping or change debug flags (4401)
6018 - don't zero obd_dev when it is force cleaned (3651)
6019 - print grants to console if they go negative (4431)
6020 - "lctl deactivate" will stop automatic recovery attempts (3406)
6021 - look for existing locks in ldlm_handle_enqueue() (3764)
6022 - don't resolve lock handle twice in recovery avoiding race (4401)
6023 - revalidate should check working dir is a directory (4134)
6025 - don't always mark "slow" obdfilter messages as errors (4418)
6027 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6030 - don't close LustreDB during write_conf until it is done (3860)
6031 - fix typo in lconf for_each_profile (3821)
6032 - allow dumping logs from multiple threads at one time (3820)
6033 - don't allow multiple threads in OSC recovery (3812)
6034 - fix debug_size parameters (3864)
6035 - fix mds_postrecov to initialize import for llog ctxt (3121)
6036 - replace config semaphore with spinlock (3306)
6037 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6038 - don't allow enqueue to complete on a destroyed export (3822)
6039 - down write_lock before checking llog header bitmap (3825)
6040 - recover from lock replay timeout (3764)
6041 - up llog sem before sending rpc (3652)
6042 - reduce ns lock hold times when setting kms (3267)
6043 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6044 - fix NULL deref and obd_dev leak on setup error (3312)
6045 - replace some LBUG about llog ops with error handling (3841)
6046 - don't match INVALID dentries from d_lookup and spin (3784)
6047 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6048 - fix invalid assertion in ptlrpc_set_wait (3880)
6050 - add libwrap support for the TCP acceptor (3996)
6051 - add /proc/sys/portals/routes for non-root route listing (3994)
6052 - allow setting MDS UUID in .xml (2580)
6053 - print the stack of a process that LBUGs (4228)
6055 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6058 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6059 - make sure to unset replay flag from failed open requests (3440)
6060 - if default stripe count is 0, use OST count for inode size (3636)
6061 - update parent mtime/ctime on client for create/unlink (2611)
6062 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6063 - free recovery state on server during a forced cleanup (3571)
6064 - unregister_reply for resent reqs (3063)
6065 - loop back devices mounting and status check on 2.6 (3563)
6066 - fix resource-creation race that can provoke i_size == 0 (3513)
6067 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6068 - more debugging for page-accounting assertion (3746)
6069 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6070 - avoid deadlock after precreation failure (3758)
6071 - fix race and lock order deadlock in orphan handling (3450, 3750)
6072 - add validity checks when grabbing inodes from l_ast_data (3599)
6074 - add /proc/.../recovery_status to obdfilter (3428)
6075 - lightweight CDEBUG infrastructure, debug daemon (3668)
6076 - change default OSC RPC parameters to be better on small clusters
6077 - turn off OST read cache for files smaller than 32MB
6078 - install man pages and include them in rpms (3100)
6079 - add new init script for (un)mounting lustre filesystems (2593)
6080 - run chkconfig in %post for init scripts (3701)
6081 - drop scimac NAL (unmaintained)
6083 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6086 - clean kiobufs before and after use (3485)
6087 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6088 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6089 - change init script to fail more clearly if not run as root (1528)
6090 - allow clients to reconnect during replay (1742)
6091 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6092 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6093 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6094 - fix cond_resched (3554)
6095 - only evict unfinished clients after recovery (3515)
6096 - allow bulk resend, prevent data loss (3570)
6097 - dynamic ptlrpc request buffer allocation (2102)
6098 - don't allow unlinking open directory if it isn't empty (2904)
6099 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6100 - remove extraneous obd dereference causing LASSERT failure (3334)
6101 - don't use get_cycles() when creating temp. files on the mds (3156)
6102 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6103 - handle EEXIST for set-stripe, set proper directory name (3336)
6105 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6106 - fix ksocknal_fmb_callback() error messages (2918)
6108 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6111 - don't copy lvb into (possibly NULL) reply on error (2983)
6112 - don't deref dentry after dput, don't free lvb on error (2922)
6113 - use the kms to determine writeback rpc length (2947)
6114 - increment oti_logcookies when osc is inactive (2948)
6115 - update client's i_blocks count via lvb messages (2543)
6116 - handle intent open/close of special files properly (1557)
6117 - mount MDS with errors=remount-ro, like obdfilter (2009)
6118 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6119 - don't use cancelling-locks' kms values (2947)
6120 - use highest lock extent for kms, not last one (2925)
6121 - don't dereference ERR_PTR() dentry in error handling path (3107)
6122 - fix thread race in portals_debug_dumplog() (3122)
6123 - create lprocfs device entries at setup instead of at attach (1519)
6124 - common AST error handler, don't evict client on completion race (3145)
6125 - zero nameidata in detach_mnt in 2.6 (3118)
6126 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6127 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6128 - initialize RPC timeout timer earlier for 2.6 (3219)
6129 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6130 - print nal/nid for unknown nid (3258)
6131 - additional checks for oscc recovery before doing precreate (3284)
6132 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6133 - don't crash in mdc_close for bad permissions on open (3285)
6134 - zero i_rdev for non-device files (3147)
6135 - clear page->private before handing to FS, better assertion (3119)
6136 - tune the read pipeline (3236)
6137 - fix incorrect decref of invalidated dentry (2350)
6138 - provide read-ahead stats and refine rpc in flight stats (3328)
6139 - don't hold journal transaction open across create RPC (3313)
6140 - update atime on MDS at close time (3265)
6141 - close LDAP connection when recovering to avoid server load (3315)
6142 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6143 - don't leak open file on MDS after open resend (3325)
6144 - serialize filter_precreate and filter_destroy_precreated (3329)
6145 - loop device shouldn't call sync_dev() for nul device (3092)
6146 - clear page cache after eviction (2766)
6147 - resynchronize MDS->OST in background (2824)
6148 - refuse to mount the same filesystem twice on same mountpoint (3394)
6149 - allow llmount to create routes for mounting behind routers (3320)
6150 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6151 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6152 - fix rare problem with rename on htree directories (3417)
6154 - allow default OST striping configuration per directory (1414)
6155 - fix compilation for qswnal for 2.6 kernels (3125)
6156 - increase maximum number of MDS request buffers for large systems
6157 - change liblustreapi to be useful for external progs like lfsck (3098)
6158 - increase local configuration timeout for slow disks (3353)
6159 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6161 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6164 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6165 - don't overwrite extent policy data in reply if lock was blocked (2901)
6166 - drop filter export grants atomically with removal from device (2663)
6167 - del obd_self_export from work_list in class_disconnect_exports (2908)
6168 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6169 - swab reply message in mdc_close, other PPC fixes (2464)
6170 - fix destroying of named logs (2325)
6171 - overwrite old logs when running lconf --write_conf (2264)
6172 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6173 - fix race in target_handle_connect (2898)
6174 - mds_reint_create() should take same inode create lock (2926)
6175 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6176 - don't close files for self_export to avoid uninitialized obd (2936)
6177 - allow MDS with the same name as client node (2939)
6178 - hold dentry reference for closed log files for unlink (2325)
6179 - reserve space for all logs during transactions (2059)
6180 - don't evict page beyond end of stripe extent (2925)
6181 - don't oops on a deleted current working directory (2399)
6182 - handle hard links to targets without a parent properly (2517)
6183 - don't dereference NULL lock when racing during eviction (2867)
6184 - don't grow lock extents when lots of conflicting locks (2919)
6186 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6189 - account for cache space usage on clients to avoid data loss (974)
6190 - lfsck support in lustre kernel code (2349)
6191 - reduce journal credits needed for BRW writes (2370)
6192 - orphan handling to avoid losing space on client/server crashes
6193 - ptlrpcd can be blocked, stopping ALL progress (2477)
6194 - use lock value blocks to assist in proper KMS, faster stat (1021)
6195 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6196 - recovery for initial connections (2355)
6197 - fixes for mds_cleanup_orphans (1934)
6198 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6199 - block all file creations until orphan recovery completes (1901)
6200 - client remove rq_connection from request struct (2423)
6201 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6202 - recovery timer race (2670)
6203 - mdc_close recovey bug (2532)
6204 - ptlrpc cleanup bug (2710)
6205 - mds timeout on local locks (2588)
6206 - namespace lock held during RPCs (2431)
6207 - handle interrupted sync write properly (2503)
6208 - don't try to handle a message that hasn't been replied to (2699)
6209 - client assert failure during cleanup after abort recovery (2701)
6210 - leak mdc device after failed mount (2712)
6211 - ptlrpc_check_set allows timedout requests to complete (2714)
6212 - wait for inflight reqs when ptlrpcd finishes (2710)
6213 - make sure unregistered services are removed from the srv_list
6214 - reset bulk XID's when resending them (caught by 1138 test)
6215 - unregister_bulk after timeout
6216 - fix lconf error (2694)
6217 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6218 - readahead locks pages, leaves pending causing memory pressure (2673)
6219 - increase OST request buffers to 4096 on large machines (2729)
6220 - fix up permission of existing directories in simple_mkdir (2661)
6221 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6222 - don't assign transno to errored transactions (2742)
6223 - don't delete objects on OST if given a bogus objid from MDS (2751)
6224 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6225 - if rq_no_resend, then timeout request after recovery (2432)
6226 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6227 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6228 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6229 - MDS recovery shouldn't delete objects that it has given out (2730)
6230 - if enqueue arrives after completion, don't clobber LVB (2819)
6231 - don't unlock pages twice when trigger_group_io returns error (2814)
6232 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6233 - don't write pages to disk if there was an error (1450)
6234 - don't ping imports that have recovery disabled (2676)
6235 - take buffered bytes into account when balancing socknal conn (2817)
6236 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6237 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6238 - remove little-endian swabbing of llog records (1987)
6239 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6240 - retry reposting request buffers if they fail (1191)
6241 - grow extent at grant time to avoid granting a revoked lock (2809)
6242 - lock revoke doesn't evict page if covered by a second lock (2765)
6243 - disable VM readahead to avoid reading outside lock extents (2805)
6245 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6246 - updated kernel patches for hp-2.4.20 kernel (2681)
6248 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6251 - fix truncated write corruption (2366)
6252 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6254 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6255 - ensure len(uuid) < 37 in lmc (1171)
6256 - fix ia64 OOPS in llog_test (2255)
6257 - zero end of page at obdfilter for partial page writes (2648)
6258 - don't leave stale dentries around after renames (bug 2428)
6259 - fix timeouts when evicting a client with a single lock held (2642)
6260 - set deadline for the initial HELLO message to drain (2634)
6261 - print out dotted-quad IP addresses in the socknal (2302)
6263 - additional debugging for MDS client eviction problem (2443)
6264 - fix mkfsoptions support for osts (2603, 2604)
6266 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6269 - add series for the vanilla 2.6.0 kernel
6270 - add series for the vanilla 2.4.24 kernel
6271 - add series for a cray x86/64 UL kernel drop
6272 - fix xattr patches for the vanilla 2.4.19 series
6274 - generate true UUIDs in lmc (1171)
6275 - have portals stack dumping break in UML (2466)
6276 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6277 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6278 - wake callback waiting threads on client eviction (2460)
6279 - Add --ptldebug and --subsystem to lmc (1719)
6280 - update assertion to allow safe interrupt allocation
6281 - set rq_no_resend for cancel requests (2432)
6282 - recalculate ptlrpcd timeout after resend (2494)
6283 - call vfs_rmdir when removing pending directories (2368)
6284 - fix renaming a file to itself (2429)
6285 - lmc creates a default one-stripe lov (2454)
6286 - expand procfs space to handle large clusters (2326)
6287 - increase UML stack to avoid overflow
6288 - update lconf's list of debug and subsystem masks
6289 - fix lfs find --obd (2510)
6290 - /proc tunable for disabling filter read caching (2591)
6291 - stop rpm packages from altering slapd.conf (2301)
6292 - disable nagle in the socknal under 0conf (2578)
6293 - choose mds inode size based on stripe count (2572)
6294 - fix kernel-source rpm problems (2516)
6296 - add --disable-doc to avoid pdf generation (2421)
6297 - update documentation, tests, type-os, comments
6298 - avoid format warnings on ia64
6299 - remove the TOE NAL
6300 - tiny code cleanups by removing unused fields
6302 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6305 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6306 - remove the most blatant lies from BUILDING file (2371)
6307 - change default debug level to reasonable production setting
6308 - reduce client side cache size to reduce cache flush time
6309 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6310 - make TCP zerocopy and pinger support enabled by default (2476)
6311 - sync writes completed after process exits caused crashes (2319)
6312 - maintain correct mount count on the MDS (2356)
6313 - backout 1557, because 2316 wasn't really fixed
6314 - better file I/O statistics gathering in /proc
6315 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6316 - another kernel patch to fix zero-copy TCP function export
6317 - don't take duplicate lock when processing re-sent getattr (2420)
6318 - lctl uses obd_self_export instead of creating new conn (2353)
6319 - MDS/OST recovery case which requires object creation asserted (2425)
6320 - move lfs from /usr/sbin to /usr/bin in packages
6321 - fix race between mds_client_add and mds_client_free (2417)
6322 - use kmalloc instead of slabs in portals (2430)
6323 - don't create duplicate records when a failover MDS is present (2442)
6324 - remove unnecessary mount age check (2332)
6325 - don't remove directory inodes from locks prematurely (2451)
6326 - don't break if MDS service name is the same as hostname (2103)
6327 - fix races in client write RPC generation when cache full (2482)
6329 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6332 - remove now-unused request->rq_obd (278)
6333 - if an allocation fails, print out how much memory we've used (1933)
6334 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6335 - add the "configurable stack size" patch to most series files (1256)
6336 - ability to write large log records, for 100+ OST configs (2306)
6337 - fix NULL deref when filter_prep fails (2314)
6338 - fix operator precedence error in filter_sync
6339 - dynamic allocation of socknal TX descriptors (2315)
6340 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6341 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6342 - ensure that CWARN messages in Portals always get to the syslog
6343 - __init/__exit are not for prototype decls (ldlm_init/exit)
6344 - x86-64 compile warning fixes
6345 - fix gateway LMC keyword conflict (2318)
6346 - fix MDS lock inversions in getattr/reint paths (1844)
6347 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6348 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6349 - DLM race condition prevented some lock evictions (2328)
6350 - ENOMEM detection and retry on socknal sends (2230)
6351 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6352 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6353 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6354 - better error messages when a client is rejected during recovery (1505)
6355 - avoid cancelling locks which were never granted, after failure (2330)
6356 - fix i_sem/journal inversion in mds_client_add (2333)
6357 - fix truncate/getattr lock cycle deadlock (2334)
6358 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6359 - fix two rare exit paths which could leak an l_lock() ref (2321)
6360 - fencepost error in MDS/OST orphan recovery (2226)
6361 - make log record alignment 8 bytes (1988)
6362 - lstripe now fails when requested offset > ost_count (2237)
6363 - ensure that all kernel series have a complete list.h (1607)
6364 - fix crashes in special-file operations (2316)
6365 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6367 - allow configurable automake binary, for testing new versions
6368 - small update to the lfs documentation
6370 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6372 * fix negative export reference count in fsfilt_sync (2312)
6374 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6375 * release candidate 0.9.1
6377 - orphans are moved into the PENDING directory for possible recovery
6378 - replayed opens now open by fid for orphan/rename safety (1042)
6379 - last close of an orphan inode generates a transno (683)
6380 - chdir() and mount() now pin the directory entry (1020)
6381 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6382 - discard very old requests without processing them (1502)
6383 - remove some common, well-understood CERRORs (1505)
6384 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6385 - clear "grant" flags in OST replies until OST grant code lands (1644)
6386 - fix read performance by not clobbering i_blksize on client (1598)
6387 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6388 - make LRU size a /proc tunable, clears locks when reduced (707)
6389 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6390 - server locks take references on exports now (1558)
6391 - build fixes for 2.4.20-rh trees (1663)
6392 - return an error from lov_create if all OSCs are inactive (1751)
6393 - fix import levels when a reconnect happens without a timeout (1597)
6394 - exit early from mds_open if we get a lookup error (1749)
6395 - partial page read at EOF wouldn't wait for disk before sending (1642)
6396 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6397 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6398 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6399 - write/truncate lock inversion (1639)
6400 - Don't auto-load obdclass, portals modules during cleanup (1495)
6401 - fix timestamps from jumping to "now" (1763)
6402 - extra journal assertions (1648)
6403 - add an extra multiunlink test (1771)
6404 - fix read_record/write_record API (1776)
6405 - fix leak of offset_extent, possible incorrect i_size later (1772)
6406 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6407 - Debugging for the kqswnal_get_idle_tx problems (1820)
6408 - Allow recovery to be attempted multiple times (1536)
6409 - Write out MDS last_rcvd file after it is first created (1600)
6410 - Fix tx_descriptor leak in failed transmit situations (1827)
6411 - ext3 journaling fixes for assertion failure after IO error (1871)
6412 - class_export_put() on freed export after completion AST error (1896)
6413 - Fix revalidate looping in VFS (1322)
6414 - Don't access a freed export during MDS_REINT timeout (1521)
6415 - Add open-unlink recovery support on the MDS (1673,1764)
6416 - Return an error if no MDS data was read from last_rcvd (1946)
6417 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6418 - Don't setup a disk device that doesn't match exported UUID (317)
6419 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6420 - avoid committing NULL handle in force close
6421 - local.sh is now a one-stripe LOV configuration
6422 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6423 - fix ext3 htree duplicate directory entry corruption (1516)
6424 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6425 - update to new LOV EA format (2097)
6426 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6428 2003-06-15 Phil Schwan <phil@clusterfs.com>
6431 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6432 - per-import/export recovery handling (958, 931, 959)
6433 - multiple last-rcvd slots, for serving multiple FSes (949)
6434 - connections are again shared between multiple imp/exports (963, 964)
6435 - "umount -f" would hang if any requests needed to be sent (393, 978)
6436 - avoid pinning large req buffer by copying for queued messages (989)
6437 - add "uuid" to "lctl device" command to help upcalls (991)
6438 - "open" RPCs with transnos would confuse recovery counters (1037)
6439 - do proper endian conversion of all wire messages (288, 340, 891)
6440 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6441 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6442 - fix ptlrpc_connection leak in target_handle_connect (1174)
6443 - fix import refcounting bug in OST and MDS cleanup (1134)
6444 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6445 - fix very unlikely obd_types race condition (501)
6446 - remove osc_open hack for echo_client (1187)
6447 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6448 - a failure in read_inode2 leads to deadlock (1139)
6449 - cancel ack-locks as soon as transaction is committed (1072)
6450 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6451 - make sure to commitrw after any preprw to avoid deadlock (1162)
6452 - failing to execute a file in a lustre FS would lock inode (1203)
6453 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6454 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6455 - obd_uuid2tgt would walk off the end of the list (1255)
6456 - on IA64 the osc would give portals incorrect bulk size (1258)
6457 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6458 - fix lock inversion caused by new llite matching code (1282)
6459 - limit the number of dirty pages on a client to 10MB (1286)
6460 - timed out locks were not being corrected cancelled (1289)
6461 - fix O_DIRECT above 4GB on IA-32 (1292)
6462 * major user-visible changes
6463 - fail out/fail over policy now controlled by the upcall (993)
6465 - add OBD_PING to check server availability and failure (954)
6466 - lustre messages are now sent in sending host order (288, 340, 891)
6467 - add eadatalen to MDS getattr reply (340)
6468 - OST read replies may contain second buffer, with per-page status (593)
6470 2003-03-11 Phil Schwan <phil@clusterfs.com>
6473 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6474 - failed open()s could cause deadlock; fixed (867, 869)
6475 - stop cancelling OST locks when files are closed (481)
6476 - overlapping XID spaces caused network corruption (851, 853)
6477 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6478 - setattr_raw added, to do single-RPC, server-side setattrs
6479 - lmc/lconf syntax change for OST UUIDs
6480 - fix crashy race condition between ptlrpc_free_req and osc_close
6481 - don't use request in mdc_enqueue if we hit a timeout (889)
6482 - don't set the inode i_size for regular files from the MDS (896)
6483 - handle out of order completion AST (842)
6484 - don't LBUG if a lock request times out after receiving AST (913)
6485 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6486 - if a bad lock AST arrives, send an error instead of dropping entirely
6487 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6488 - fix leak in bulk IO when only partially completed (899, 900, 926)
6489 - fix O_DIRECT for ia64 (55)
6490 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6491 - C-z after timeout could hang a process forever; fixed (977)
6493 - client-side I/O cache (678, 924, 929, 941, 970)
6495 - READPAGE and SETATTRs which don't take server-side locks get
6498 2003-02-11 Phil Schwan <phil@clusterfs.com>
6501 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6502 - Fix fsfilt_extN_readpage() to read a full page of directory
6503 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6504 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6505 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6506 - Most intent operations were removed, in favour of a new RPC mode
6507 that does a single RPC to the server and bypasses most of the VFS
6508 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6509 - Aggressively cancel local locks on DLM servers
6510 - mds_reint_unlink sends EA to the client if it's the last nlink.
6511 client uses that EA to unlink OST objects.
6512 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6513 - recursive symlinks were fixed (439)
6514 - fixed NULL deref in DEBUG_REQ
6515 - filter_update_lastobjid no longer calls sync, which annoyed extN
6516 - fixed multi-client small-writes to a single file problem (445)
6517 - fixed mtime updates during file writes (607)
6518 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6519 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6520 - fixed Linux OST size reporting problem (444, 656)
6521 - OST now updates object mtime with writes or setattr (607, 619)
6522 - client verifies file size before zeroing page past EOF (445)
6523 - OST now writes last allocated objid to disk with allocation (108)
6524 - LOV on echo now works (409)
6526 - mds_reint_unlink sends a new buffer, with the EA included. this
6527 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6528 set if a regular file was being unlinked, and it was the last link
6529 - use PtlGet from the target for bulk writes (315)
6530 - OST now updates object mtime with writes or setattr (607, 619)
6531 - LDLM now has a grant-time callback to revalidate locked items, if
6533 - Many MDS operations were reorganized to combat race conditions
6535 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6536 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6538 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6541 - Fully reactivate OST imports after reconnection (512, others)
6542 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6543 - More graceful error handling for truncating on dead OST (515)
6544 - Don't error out unless we're actually accessing dead stripes (474)
6545 - Fix garbage sizes when stripes are missing (410)
6546 - LRU counters were broken, causing constant lock purge (433, 432)
6547 - garbage on read from stripes with failed OSTs (441)
6548 - mark OSCs as active before reconnecting during recovery (438)
6549 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6550 - lfind did not preserve OST order in output (443)
6551 - symlinks cause hung clients, incorrect data (439)
6552 - stop dereferencing request after dropping refcount (457)
6553 - don't LASSERT(spin_is_locked) on non-SMP (455)
6554 - fixes for many rename() bugs
6555 - fstat didn't correctly synchronize attributes (399)
6556 - server must handle lock cancellation during blocking AST prep (487)
6557 - bulk descriptors were free()d too soon (511)
6558 - fix paths in lconf, which would load incorrect modules (451, 507)
6559 - fix confusing lconf 'host not found' error message (386)
6560 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6561 - fix race condition in mdc_blocking_ast() for inode access (526)
6562 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6563 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6564 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6565 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6566 - fix lstripe in several ways (kernel side) (527)
6567 - fix request leak in ldlm_cli_enqueue (262)
6568 - incorrect OSC was marked inactive after OST failure
6569 - call mds_fs_cleanup before unmounting filesystem (524)
6570 - fix races between taking ns_lock and ldlm_lock_change_resource
6571 - fix races updating LOV export open file list
6572 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6573 - fix recovery NULL deref in ldlm_cli_cancel_unused
6574 - fix some DLM races by using new hash table for lock handles (419)
6575 - permit the client to specify desired inodes, at replay
6576 - duplicate requests when we queue them for replay reintegration
6577 - fix last_rcvd offset calculation
6578 - sync after each recovered transaction, so we always make progress
6579 - never, not always, ERESTART requests without transnos
6580 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6582 - skip replay if the MDS says that the client is already connected
6583 - don't check for a recovery-enabled export to match lctl's UUID
6584 - don't INC_USE_COUNT for phantom exports
6585 - don't crash when cleaning up phantom exports (567)
6586 - don't double-finish or set replay data for errored mdc_open requests
6587 - abort requests when they time out, so we don't get old replies
6588 - send/receive replies for AST messages again
6589 - if the client says that it doesn't have the lock, cancel it on the
6591 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6592 mark it as destroyed, it will all work out when decref is called
6593 - fix module use counts (22, 581)
6595 - ASTs now expect a reply (server cancels lock on error reply)
6597 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6600 - fix many simultaneous client startup (392)
6601 - fix dentry->d_it clobbering
6602 - credentials weren't being shipped for readdir/getattr operations
6603 - remove invalid assertions triggered during some concurrent MD
6605 - proper Lustre versions added (336, 389)
6606 - fix memory leak for create error case (398)
6607 - fix LOV locking bug that would get cli/srv out of sync
6608 - fix echo client over LOV (409)
6609 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6610 - fix double-O_EXCL intent crash (424)
6611 - avoid sending multiple lock CANCELs (352)
6613 - MDS can do multi-client recovery (modulo bugs in new code)
6615 - many updates, edits, cleanups
6617 2002-11-18 Phil Schwan <phil@clusterfs.com>
6620 - fix null d_it dereference (346)
6621 - fix full OST/dbench hang (333)
6622 - fix permission problem with file removal (286)
6623 - fix removal of OSCs from LOV when they fail
6624 - fix NULL deref during bulk timeout (214)
6625 - fix problems related to multiple filesystems on one MDS (241)
6626 - fixed serious subtle metadata locking bugs
6627 - free locks on clients when inodes are removed due to memory
6629 - fix inode pointer in lock data (285)
6630 - partial support for multiple MDS on a single host (241)
6631 - data locks weren't cancelled at clear_inode time (290, 311)
6632 - intent locks could lead to unbounded lock growth (205)
6633 - added a maximum lock count, an LRU list, and a flusher
6634 - fix multiple rename (365)
6635 - properly abstracted the echo client
6636 - OSC locked 1 byte too many; fixed
6637 - rewrote brw callback code:
6638 - fixed recovery bugs related to LOVs (306)
6639 - fixed too-many-pages-in-one-write crash (191)
6640 - fixed (again) crash in sync_io_timeout (214)
6641 - probably fixed callback-related race (385)
6643 - Add capability to MDS protocol
6644 - LDLM cancellations and callbacks on different portals
6646 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6649 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6650 - defer open object destruction to close time (601981, 138)
6651 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6652 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6653 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6654 - don't call obd_setattr in ll_file_release for destroyed objects
6655 * protocol change to lustre_msg: move |version| and add |flags|
6656 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6657 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6658 * added replay of create, unlink, link and rename operations during
6659 MDS failover; recovery should be much more robust now
6660 * remove failed OSCs from LOVs (only lov_create uses this so far)
6661 * the lustre-HOWTO was brought (more) up to date (582544)
6663 2002-10-23 Phil Schwan <phil@clusterfs.com>
6666 - in-use dentries weren't being reused properly (617851)
6667 - prevent multiple LDLM setup (599178)
6668 - fix LOV size calculations for truncate (617853)
6669 - fix client handling of MDS intent errors (POSIX)
6670 - fix permission bug in lovstripe.c test (624321)
6671 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6672 - truncate past end of file could corrupt data
6673 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6674 - a race in recovery could return ETIMEDOUT to apps (623947)
6675 - building outside the source directory was fixed
6676 * the lustre-HOWTO was brought (more) up to date (582544)
6677 * major progress was made on recovery functionality
6679 2002-10-10 Phil Schwan <phil@clusterfs.com>
6682 - recovery deadlock fix
6683 - rm -rf causes LBUG fix (617817)
6684 - file open by multiple tasks fix (618962)
6685 - directory permissions bugs (602707 and 620007)
6686 - journal_stop fixed with locking (611313)
6687 - O_APPEND failures resolved (618273, perhaps 614459)
6688 - lconf PATH fix (619770)
6689 - IA64 build fix (621450)
6690 - RPC buffer sizes scale with amount of memory
6692 2002-10-01 Phil Schwan <phil@clusterfs.com>
6695 - locks would be cancelled without throwing away data pages,
6696 resulting in inconsistent data (605627)
6697 - inode attributes were not always being refreshed (605627, 612449)
6698 - lconf now continues to cleanup after lctl reports an error
6699 - MDS now enforces user permissions (602707)
6700 - lprocfs cleanup fixed, but not yet enabled (614157)
6701 - fixed infinite server hang, should a client not respond to an AST
6702 - avoid going into recovery if user calls readlink() with a buffer
6703 that's too small (613941)
6704 - AST RPCs no longer require replies (614867) -- this may be changed
6705 - don't crash server if client sends an IOV that's too big (611336)
6706 - fixed lock conversion deadlock (611892)
6707 - fixed the following of symlinks (614622)
6708 * recovery: the server can remove locks from a client that dies, other
6709 clients can make progress
6710 * more extN patch fixes
6711 * compile-time configurable ptlrpc buffer allocations
6713 - collaborative read cache document
6714 - Lustre Lite Performance CDR document-in-progress
6716 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6719 - fix typo in patch-2.4.18
6721 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6724 - clear ptlrpc request each time in handle_incoming_request()
6725 - unlink of files now destroys the object on the OST
6727 2002-09-19 Peter Braam <braam@clusterfs.com>
6729 * add hard link support
6730 * change obdfile creation method
6731 * kernel patch changed
6733 2002-09-19 Peter Braam <braam@clusterfs.com>
6736 - stack overflow bug in extN fixed
6738 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6740 * documentation updates
6741 - add man pages for config tools
6742 - update tests/README to describe testing with new config tools
6743 - finish metadata API descriptions
6744 * bug fixes and cleanups
6745 - statfs workaround for 16TB limit
6746 - LOV stripe allocation improved, can stripe on subset of OSTs
6747 - LOV file size/IO offset was wrong for files > 4GB in size
6748 - object EA data was being dropped, caused files to be unreadable
6749 - memory overflow with non-LOV OST caused memory corruption
6750 - fixed regression tests to work with new config tools, obdfilter
6751 - fixed bug when directory size became larger than 1 block
6752 - fixed bug (for single client case) when PWD was deleted
6753 - invalidate local directory pages when doing intent-based ops
6754 - avoid LDLM oops when lock callback contained bad data
6756 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6758 * documentation updates
6759 * bug fixes and cleanups
6760 - configuration tools
6763 - 64-bit compile warnings
6764 - 64-bit internal statfs data
6766 * test_brw on persistent OST devices
6768 * lprocfs (disabled)
6770 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6772 * documentation updates
6773 * bug fixes and cleanups
6774 * configuration tools
6776 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6779 * many small fixes to 0_5_4
6780 * io/network handling
6781 * thinkos in MDS operations
6783 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6786 * crucial basic fixes to 0.5.3
6787 * IOR, Iozone work over Elan
6790 2002-08-07 Phil Schwan <phil@clusterfs.com>
6791 * version 0_5_3, our first alpha
6792 * we use the new Portals iovs
6793 * documentation updates
6794 * bug fixes and cleanups
6795 * small changes in the DLM wire protocol
6797 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6798 * version 0_5_1 with some initial stability,
6799 * locking on MD and file I/O.
6800 * documentation updates
6801 * several bug fixes since 0.5.0
6802 * small changes in wire protocol
6804 2002-07-18 Phil Schwan <phil@clusterfs.com>
6806 * delivered as Lustre Light Alpha
6807 * fixed a crash after handling invalid MDS requests
6808 * fixed directory pages for architectures with non-4k pages sizes
6810 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6811 * release version v0_4_4
6812 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6814 * Fixes a number of interruption problems with OST operations.
6815 * Update documentation for portals header changes
6816 * Move all wire protocol structs/defines to lustre_idl.h
6817 * Fixes symlink length bug.
6818 * Add tcpdump to repository.
6820 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6821 * release version v0_4_3
6822 * Fixes statfs for inodes on extN.
6823 * Fixes bug in runtests which would delete /etc/hosts.
6824 * Use 64-bit object IDs wherever possible (not into VFS though)
6825 Remove ost_get_info, which is unused by lustre, and out of date.
6827 2002-07-03 Peter Braam <braam@clusterfs.com>
6828 * release version v0_4_2 Fixes a lookup error (type not passed)
6829 * move forward to head of Portals
6830 * move forward to latest Lustre kernel
6832 2002-06-25 Peter Braam <braam@clusterfs.com>
6833 * release version v0_4_1. Hopefully stable on single node use.