3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.1.1.el7 (RHEL7.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 4.18.0-80.7.1.el8 (RHEL8.0)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.100 (SLES12 SP3)
13 4.4.0-131 (Ubuntu 16.04)
14 4.15.0-32 (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 3.10.0-1062.1.1.el7 (RHEL7.7)
19 4.12.14-95.29 (SLES12 SP4)
20 4.15.0-48 (Ubuntu 18.04)
21 * Other clients known to build on unpatched kernels at some point (others may also work):
22 3.10.0-862.14.4.el7 (RHEL7.5)
23 3.10.0-957.27.2.el7 (RHEL7.6)
24 4.14.0-49.13.1.el7a (RHEL7.5)
25 4.14.0-115.2.2.el7a (RHEL7.6)
26 4.18.0-80.7.1.el8 (RHEL8.0)
27 4.4.120-92.70 (SLES12 SP2)
28 4.4.180-94.100 (SLES12 SP3)
29 4.4.0-131 (Ubuntu 16.04)
30 vanilla linux 4.15.0 (ZFS only)
31 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
32 * Recommended ZFS / SPL version: 0.7.13
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
38 * Note for Ubuntu 14.04.5:
39 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
40 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
42 --------------------------------------------------------------------------------
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 3.10.0-1062.1.1.el7 (RHEL7.7)
50 * Other server kernels known to build and work at some point (others may also work):
51 3.10.0-862.14.4.el7 (RHEL7.5)
52 3.10.0-957.27.2.el7 (RHEL7.6)
53 4.18.0-80.7.1.el8 (RHEL8.0)
54 4.4.120-92.70 (SLES12 SP2)
55 4.4.180-94.100 (SLES12 SP3)
56 4.4.0-131 (Ubuntu 16.04)
57 4.15.0-32 (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 3.10.0-1062.1.1.el7 (RHEL7.7)
62 4.12.14-95.29 (SLES12 SP4)
63 4.15.0-48 (Ubuntu 18.04)
64 * Other clients known to build on unpatched kernels at some point (others may also work):
65 3.10.0-862.14.4.el7 (RHEL7.5)
66 3.10.0-957.27.2.el7 (RHEL7.6)
67 4.14.0-49.13.1.el7a (RHEL7.5)
68 4.14.0-115.2.2.el7a (RHEL7.6)
69 4.18.0-80.7.1.el8 (RHEL8.0)
70 4.4.120-92.70 (SLES12 SP2)
71 4.4.180-94.100 (SLES12 SP3)
72 4.4.0-131 (Ubuntu 16.04)
73 vanilla linux 4.15.0 (ZFS only)
74 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
75 * Recommended ZFS / SPL version: 0.7.13
76 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
77 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
78 stack overflow. For more information, please refer to bugzilla 17630.
79 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
80 the re-exporting nodes to be mounted with "32bitapi" mount option
81 * Note for Ubuntu 14.04.5:
82 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
83 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
85 --------------------------------------------------------------------------------
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.el7 (RHEL7.6)
95 4.4.120-92.70 (SLES12 SP2)
96 4.4.162-94.69 (SLES12 SP3)
97 4.4.0-131 (Ubuntu 16.04)
98 4.15.0-32 (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.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 4.4.0-131 (Ubuntu 16.04)
111 4.15.0-32 (Ubuntu 18.04)
112 vanilla linux 4.15.0 (ZFS only)
113 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
114 * Recommended ZFS / SPL version: 0.7.9
115 * Tested with ZFS / SPL version: 0.7.9
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
125 --------------------------------------------------------------------------------
127 2018-04-03 Intel Corporation
129 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
130 for currently supported client and server kernel versions.
131 * Server known to build on patched kernels:
132 2.6.32-431.29.2.el6 (RHEL6.5)
133 2.6.32-504.30.3.el6 (RHEL6.6)
134 2.6.32-573.26.1.el6 (RHEL6.7)
135 2.6.32-642.15.1.el6 (RHEL6.8)
136 2.6.32-696.18.7.el6 (RHEL6.9)
137 3.10.0-693.21.1.el7 (RHEL7.4)
138 3.0.101-0.47.71 (SLES11 SP3)
139 3.0.101-107 (SLES11 SP4)
140 3.12.74-60.64.40 (SLES12 SP1)
141 4.4.114-92.64 (SLES12 SP2)
142 4.4.114-94.11 (SLES12 SP3)
143 3.13.0-101 (Ubuntu 14.04, ZFS only)
144 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
145 4.4.0-116 (Ubuntu 16.04)
146 vanilla linux 4.6.7 (ZFS only)
147 * Client known to build on unpatched kernels:
148 2.6.32-431.29.2.el6 (RHEL6.5)
149 2.6.32-504.30.3.el6 (RHEL6.6)
150 2.6.32-573.26.1.el6 (RHEL6.7)
151 2.6.32-642.15.1.el6 (RHEL6.8)
152 2.6.32-696.18.7.el6 (RHEL6.9)
153 3.10.0-693.21.1.el7 (RHEL7.4)
154 3.0.101-0.47.71 (SLES11 SP3)
155 3.0.101-107 (SLES11 SP4)
156 3.12.74-60.64.40 (SLES12 SP1)
157 4.4.114-92.64 (SLES12 SP2)
158 4.4.114-94.11 (SLES12 SP3)
159 3.13.0-101 (Ubuntu 14.04)
160 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
161 4.4.0-116 (Ubuntu 16.04)
163 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
164 * Recommended ZFS / SPL version: 0.7.6
165 * Tested with ZFS / SPL version: 0.7.6
166 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
167 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
168 stack overflow. For more information, please refer to bugzilla 17630.
169 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
170 the re-exporting nodes to be mounted with "32bitapi" mount option
171 * Note for Ubuntu 14.04.5:
172 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
173 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
175 --------------------------------------------------------------------------------
177 2017-07-05 Intel Corporation
179 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
180 for currently supported client and server kernel versions.
181 * Server known to build on patched kernels:
182 2.6.32-431.29.2.el6 (RHEL6.5)
183 2.6.32-504.30.3.el6 (RHEL6.6)
184 2.6.32-573.26.1.el6 (RHEL6.7)
185 2.6.32-642.15.1.el6 (RHEL6.8)
186 2.6.32-696.3.1.el6 (RHEL6.9)
187 3.10.0-514.21.1.el7 (RHEL7.3)
188 3.0.101-0.47.71 (SLES11 SP3)
189 3.0.101-100 (SLES11 SP4)
190 3.12.74-60.64.40 (SLES12 SP1)
191 4.4.59-92.17 (SLES12 SP2)
192 vanilla linux 4.6.7 (ZFS only)
193 * Client known to build on unpatched kernels:
194 2.6.32-431.29.2.el6 (RHEL6.5)
195 2.6.32-504.30.3.el6 (RHEL6.6)
196 2.6.32-573.26.1.el6 (RHEL6.7)
197 2.6.32-642.15.1.el6 (RHEL6.8)
198 2.6.32-696.3.1.el6 (RHEL6.9)
199 3.10.0-514.21.1.el7 (RHEL7.3)
200 3.0.101-0.47.71 (SLES11 SP3)
201 3.0.101-100 (SLES11 SP4)
202 3.12.74-60.64.40 (SLES12 SP1)
203 4.4.59-92.17 (SLES12 SP2)
205 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
206 * Recommended ZFS / SPL version: 0.6.5.9
207 * Tested with ZFS / SPL version: 0.6.5.9
208 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
209 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
210 stack overflow. For more information, please refer to bugzilla 17630.
211 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
212 the re-exporting nodes to be mounted with "32bitapi" mount option
214 --------------------------------------------------------------------------------
216 12-07-2016 Intel Corporation
218 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
219 for currently supported client and server kernel versions.
220 * Server known to build on patched kernels:
221 2.6.32-431.29.2.el6 (RHEL6.5)
222 2.6.32-504.30.3.el6 (RHEL6.6)
223 2.6.32-573.26.1.el6 (RHEL6.7)
224 2.6.32-642.11.1.el6 (RHEL6.8)
225 3.10.0-514.el7 (RHEL7.3)
226 3.0.101-0.47.71 (SLES11 SP3)
227 3.0.101-80 (SLES11 SP4)
228 3.12.62-60.64.8 (SLES12 SP1)
229 vanilla linux 4.5.7 (ZFS only)
230 * Client known to build on unpatched kernels:
231 2.6.32-431.29.2.el6 (RHEL6.5)
232 2.6.32-504.30.3.el6 (RHEL6.6)
233 2.6.32-573.26.1.el6 (RHEL6.7)
234 2.6.32-642.11.1.el6 (RHEL6.8)
235 3.10.0-514.el7 (RHEL7.3)
236 3.0.101-0.47.71 (SLES11 SP3)
237 3.0.101-80 (SLES11 SP4)
238 3.12.62-60.64.8 (SLES12 SP1)
240 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
241 * Recommended ZFS / SPL version: 0.6.5.8
242 * Tested with ZFS / SPL version: 0.6.5.8
243 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
244 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
245 stack overflow. For more information, please refer to bugzilla 17630.
246 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
247 the re-exporting nodes to be mounted with "32bitapi" mount option
249 --------------------------------------------------------------------------------
251 02-29-2016 Intel Corporation
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on patched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.12.1.el6 (RHEL6.7)
259 3.10.0-327.3.1.el7 (RHEL7.2)
260 3.0.101-0.47.71 (SLES11 SP3)
261 3.0.101-68 (SLES11 SP4)
262 vanilla linux 4.2.1 (ZFS only)
263 * Client known to build on unpatched kernels:
264 2.6.32-431.29.2.el6 (RHEL6.5)
265 2.6.32-504.30.3.el6 (RHEL6.6)
266 2.6.32-573.12.1.el6 (RHEL6.7)
267 3.10.0-327.3.1.el7 (RHEL7.2)
268 3.0.101-0.47.71 (SLES11 SP3)
269 3.0.101-68 (SLES11 SP4)
272 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
273 * Recommended ZFS / SPL version: 0.6.4.2
274 * Tested with ZFS / SPL version: 0.6.4.2
275 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
276 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
277 stack overflow. For more information, please refer to bugzilla 17630.
278 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
279 the re-exporting nodes to be mounted with "32bitapi" mount option
281 --------------------------------------------------------------------------------
283 03-10-2015 Intel Corporation
285 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
286 for currently supported client and server kernel versions.
287 * Server known to build on patched kernels:
288 2.6.32-431.29.2.el6 (RHEL6.5)
289 2.6.32-504.8.1.el6 (RHEL6.6)
290 3.0.101-0.46 (SLES11 SP3)
291 * Client known to build on unpatched kernels:
292 2.6.32-431.29.2.el6 (RHEL6.5)
293 2.6.32-504.8.1.el6 (RHEL6.6)
294 3.10.0-123.20.1.el7 (RHEL7)
295 3.0.101-0.46 (SLES11 SP3)
296 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
297 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
298 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
299 stack overflow. For more information, please refer to bugzilla 17630.
300 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
301 the re-exporting nodes to be mounted with "32bitapi" mount option
303 Severity : enhancement
305 Description: control OST-index in IDIF via ROCOMPAT flag.
306 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
307 last_rcvd file. For new formatted OST device, it will be auto set;
308 for the case of upgrading from old OST device, you can enable it
309 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
310 enabled, for new created OST-object, its IDIF-in-LMA will contain
311 the OST-index; for the existing OST-object, the OSD will convert
312 old format IDIF as new format IDIF with OST-index stored in the
313 LMA EA when accessing such OST-object or via OI scrub. Once such
314 flag is enabled, it cannot be reverted back, so the system cannot
315 be downgraded to the orignal incompatible version.
317 --------------------------------------------------------------------------------
319 07-30-2014 Intel Corporation
321 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
322 for currently supported client and server kernel versions.
323 * Server known to build on patched kernels:
324 2.6.32-431.20.3.el6 (RHEL6.5)
325 3.0.101-0.31 (SLES11 SP3)
326 * Client known to build on unpatched kernels:
327 2.6.32-431.20.3.el6 (RHEL6.5)
328 3.0.101-0.31 (SLES11 SP3)
331 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
332 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
333 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
334 stack overflow. For more information, please refer to bugzilla 17630.
335 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
336 the re-exporting nodes to be mounted with "32bitapi" mount option
338 --------------------------------------------------------------------------------
340 10-31-2013 Intel Corporation
342 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
343 * Server builds for kernels:
344 2.6.32-358.18.1.el6 (RHEL6.4)
345 * Client builds for unpatched kernels:
346 2.6.32-358.18.1.el6 (RHEL6.4)
347 3.0.92-0.8 (SLES11 SP3)
349 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
350 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
351 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
352 stack overflow. For more information, please refer to bugzilla 17630.
353 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
354 the re-exporting nodes to be mounted with "32bitapi" mount option
356 --------------------------------------------------------------------------------
358 05-30-2013 Intel Corporation
360 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
361 * Server builds for kernels:
362 2.6.32-358.6.2.el6 (RHEL6.4)
363 * Client builds for unpatched kernels:
364 2.6.32-358.6.2.el6 (RHEL6.4)
365 3.0.74-0.6.6 (SLES11 SP2)
367 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
368 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
369 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
370 stack overflow. For more information, please refer to bugzilla 17630.
371 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
372 the re-exporting nodes to be mounted with "32bitapi" mount option
374 --------------------------------------------------------------------------------
376 9-30-2012 Whamcloud, Inc.
378 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
379 * Server support for kernels:
380 2.6.32-279.5.1.el6 (RHEL6)
381 * Client support for unpatched kernels:
382 2.6.32-279.5.1.el6 (RHEL6)
384 2.6.18-238.19.1.el5 (RHEL5)
385 2.6.16.60-0.69.1 (SLES10)
386 * Recommended e2fsprogs version: 1.41.90.wc4
387 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
388 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
389 stack overflow. For more information, please refer to bugzilla 17630.
390 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
391 the re-exporting nodes to be mounted with "32bitapi" mount option
393 Severity : enhancement
395 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
397 Severity : enhancement
399 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
401 Severity : enhancement
403 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
405 Severity : enhancement
407 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
409 Severity : enhancement
411 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
413 --------------------------------------------------------------------------------
415 04-30-2012 Whamcloud, Inc.
417 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
418 * Server support for kernels:
419 2.6.32-220.4.2.el6 (RHEL6)
420 * Client support for unpatched kernels:
421 2.6.32-220.4.2.el6 (RHEL6)
423 2.6.18-238.19.1.el5 (RHEL5)
424 2.6.16.60-0.69.1 (SLES10)
425 * Recommended e2fsprogs version: 1.41.90.wc4
426 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
427 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
428 stack overflow. For more information, please refer to bugzilla 17630.
429 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
430 the re-exporting nodes to be mounted with "32bitapi" mount option
432 Severity : enhancement
434 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
436 Severity : enhancement
438 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
440 Severity : enhancement
442 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
444 Severity : enhancement
446 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
448 --------------------------------------------------------------------------------
450 09-30-2011 Whamcloud, Inc.
452 * Server support for kernels:
453 2.6.18-238.19.1.el5 (RHEL5)
454 2.6.32-131.6.1.el6 (RHEL6)
455 * Client support for unpatched kernels:
456 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
457 2.6.18-238.19.1.el5 (RHEL5)
458 2.6.32-131.6.1.el6 (RHEL6)
459 2.6.29.4-167.fc11 (FC11)
460 2.6.16.60-0.69.1 (SLES10)
462 * Recommended e2fsprogs version: 1.41.90.wc3
463 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
464 Lustre filesystem with 4K stack may cause a stack overflow. For more
465 information, please refer to bugzilla 17630.
466 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
467 mounted with "32bitapi" mount option
469 Severity : enhancement
470 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
472 Severity : enhancement
473 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
475 Severity : enhancement
476 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
479 Frequency : only with SLES10
481 Description: Use OFED "KMP" provided by Novell
482 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
483 Lustre is now built against this package. That means you need to
484 install the ofed-kmp package from Novell for the patchless client.
485 Note that the ofed-kmp that Novell ships may not exactly match the
486 kernel version but should still be compatible.
488 Severity : enhancement
490 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
491 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
493 --------------------------------------------------------------------------------
495 2010-07-15 Oracle, Inc.
497 * Server support for kernels:
498 2.6.18-164.11.1.el5 (RHEL 5)
499 2.6.18-164.11.1.0.1.el5 (OEL 5)
500 * Client support for unpatched kernels:
501 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
502 2.6.18-164.11.1.el5 (RHEL 5),
503 2.6.18-164.11.1.0.1.el5 (OEL 5)
504 2.6.16.60-0.42.8 (SLES 10),
506 2.6.29.4-167.fc11 (FC11)
507 * Recommended e2fsprogs version: 1.41.10-sun2
509 Severity : enhancement
511 Description: Distribute lfs_migrate script with lustre
512 Details : The lfs_migrate mentioned in the manual is now included in the
513 lustre tree in lustre/scripts and will distributed in lustre RPMs.
515 Severity : enhancement
517 Description: Failover nid list added to import proc file.
519 Severity : enhancement
521 Description: "som_preview" mount option required to use SOM
522 Details : The clients will need to use som_preview mount option
526 Description: Lustre does not do 1MB IOs to HW RAID
527 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
529 Severity : enhancement
531 Description: lfs quota output is not very convenient for awk/sed-parsing
532 Details : Some positions in lfs quota output table could be empty or
533 non-empty which made it hard to parse it with scripts, now
534 a dash is put instead of space where there is not supposed
539 Description: Update to OFED 1.5.1
540 Details : OFED 1.5.1 will be used in environments that don't use
541 vendor supplied OFED.
545 Description: Remove lnet. prefix from lctl params display
546 Details : Also renamed lnet.memused to lnet_memused to
551 Description: Use conf_param -d to remove permanant settings.
555 Description: Add list_param -R
556 Details : lctl list_param now supports recursive mode.
559 Bugzilla : 21829, 22312
560 Description: Fix llobdstat.
561 Details : 1. Now updating the write_bytes counter
562 2. Port script fixes from 1.8.
566 Description: decrease the usage of memory on clients.
567 Details : 1. On clients, recycle dentries and inodes unused.
568 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
573 Description: llobdstat fix and enhancement
574 Details : add a counter to set a limit to how many samples will be returned
575 fix a wildcard in the path to limit to obdfilter stats only
579 Description: lfs check only allowed for root
580 Details : Code cleanups and added a regresion test of non-root use of
585 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
586 Details : Fix hr->hr_index overload.
590 Description: Fix FIEMAP functionality.
594 Description: FC11 patchless client support added
598 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
599 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
603 Description: New acc-sm script and various test-framework enhancments.
607 Description: Previously landed patch was reverted.
611 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
615 Description: Add ldiskfs maxdirsize mount option.
619 Description: Fix mkfs.lustre support for 16TB LUNs.
623 Description: Print mmp_check_interval
624 Details : print mmp_check_interval and make it possible to abort
625 mount operation in case it takes too long.
629 Description: Add cascading_rw to lustre mpi tests.
633 Description: Store SOM data in EA.
634 Details : Also several SOM related cleanups.
638 Description: Add refcount to md_open_data
642 Description: Use refcount on lov_request_set to prevent memory corruption.
646 Description: Statahead: use parent dir's i_mutex for synchronization.
650 Description: loadgen improvements.
654 Description: Enable peer health detection in ksocklnd and ko2iblnd.
658 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
662 Description: Fix race on obd_devcie:md_stats
666 Description: Fix SLES 11 client builds.
667 Details : Resolved uninitalized pointer gcc warnings.
671 Description: Allow non-root access for "lfs check".
672 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
676 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
678 Severity : enhancement
680 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
681 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
685 Description: Sanity test_27p fails.
686 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
687 Sanity tests 27 were cleaned up and improved.
691 Description: Unify req format on client/servers
692 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
693 buffers and swabbers approach..
698 Description: drop join file stuff.
700 Severity : enhancement
702 Description: Add 2.6.27 and SLES11 (patchless client) support
706 Description: Fix in ptlrpc_expire_one_request() to print the signed time
709 Severity : enhancement
711 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
714 Severity : enhancement
716 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
718 Severity : enhancement
720 Description: correctly handle big reply message.
721 Details : send LNet event if reply is bigger than buffer and adjust this buffer
726 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
728 Severity : enhancement
730 Description: Add support for OEL5.
734 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
736 Severity : enhancement
738 Description: "weak-modules" support
739 Details : Implement "weak-modules" support which enables kernel modules
740 to be used with any kernel that implements the same kABI. In
741 order to achieve this modules are now installed in
742 /lib/modules/$(uname -r)/updates/kernel on all distributions.
746 Description: client could not reconnect to OST because of an active request.
747 Details : The client now retries to reconnect to the same server, if a
748 connect request failed with EBUSY or -EAGAIN. Test case 35b
749 conf-sanity.sh has been created.
753 Description: Don't allow make backward step on assignin osc next id.
754 Details : race between allocation next id and ll_sync thread can be cause
755 of set wrong osc next id and can be kill valid ost objects.
759 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
763 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
767 Description: Error handling in osc_statfs_interpret() has been improved.
768 Details : Check in osc_statfs_interpret() for EBADR.
772 Description: Do not send statfs() requests to OSTs disabled by administrator.
773 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
775 Severity : enhancement
777 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
779 Severity : enhancement
781 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
783 Severity : enhancement
785 Description: Enhancement for lfs(1) command to use numeric uid/gid.
789 Description: Improved cleanup procedure for conf-sanity test_46a.
793 Description: Do not start bulk transfer, if client tried to reconnect.
794 Details : Do not start bulk transfer, if client tried to reconnect,
795 but failed due to active rpcs. Check for exp_abort_active_req
796 has been implemented.
800 Description: Prevent inconsistences between linux and lustre mount structures.
801 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
802 Make the sleep interruptible.
804 Severity : enhancement
806 Description: provide server to client comms path
807 Details : server to client communications path via new LDLM_SET_INFO rpc
811 Description: truncate starts GFP_FS allocation under transaction causing deadlock
812 Details : ldiskfs_truncate calls grab_cache_page which may start page
813 allocation under an open transaction. This may lead to
814 calling prune_icache with consequent lustre reentrance.
818 Description: send correctly lsm on open replay
819 Details : MDS is trust to LSM size on replay open, but client can set wrong size
824 Description: enable client interface failover
825 Details : When a child reconnects from another NID, properly update export
826 nid hash position and ldlm reverse import.
828 Severity : enhancement
830 Description: Communicate OST degraded/readonly state via statfs to MDS
831 Details : Flags in the statfs returned from OSTs indicate whether the
832 OST is in a degraded RAID state, or if the filesystem has
833 turned read-only after a filesystem error is detected.
838 Description: don't panic if EPROTO was hit when reading symlink
839 Details : correctly handling request reference in error cases.
841 Severity : enhancement
843 Description: Add LustreNetLink, a kernel-userspace communcation path.
845 Severity : enhancement
847 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
852 Description: access to llog context before init.
853 Details : move handling CATALOGS file at osc layer and forbid access to llog
857 Frequency : rare, only when client is writing to OST on local node
859 Description: Avoid deadlock for local client writes
860 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
861 memory freeing context. This allows OST threads to set the
862 PF_MEMALLOC flag on task structures in order to allocate memory
863 from reserved pools and complete IO.
864 Use GFP_HIGHUSER for OST allocations for non-local client writes,
865 so that the OST threads generate memory pressure and allow
866 inactive pages to be reclaimed.
868 Severity : enhancement
870 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
875 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
876 Details : move ldlm namespace creation in setup phase to avoid grab
877 _lprocfs_lock with cli_sem held.
881 Description: Temporarily disable grant shrink.
882 Details : Disable the feature for debugging.
885 Frequency : only when formatting test filesystems
887 Description: Unable to run several mkfs.lustre on loop devices at the same time
888 Details : mkfs.lustre returns error 256 on the concurrent loop devices
889 formatting. The solution is to proper handle the error.
891 Severity : enhancement
893 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
895 Severity : enhancement
897 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
902 Description: resolve race between obd_disconnect and class_disconnect_exports
903 Details : if obd_disconnect will be called to already disconnected export he
904 forget release one reference and osc module can't unloaded.
906 Severity : enhancement
908 Description: Allow tuning service thread via /proc
909 Details : For each service a new
910 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
911 created that can be used to set min/max thread counts, and get the
912 current number of running threads.
917 Descriptoin: don't return error if have partially created objects for file.
918 Details : lov_update_create_set uses set->set_success as index for created
919 objects, so if some requests will be failed, they will have hole
920 at end of array and qos_shrink_lsm for allocate correct lsm.
922 Severity : enhancement
924 Description: Update OFED support to 1.4.1
929 Description: Reduce small size read RPC
930 Details : Set read-ahead limit for every file and only do read-ahead when
931 available read-ahead pages are bigger than 1M to avoid small size
934 Severity : enhancement
936 Description: move AT tunable parameters for more consistent usage
937 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
939 Severity : enhancement
941 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
942 Details : allow skip disconnected ost for send statfs request and hide error
948 Description: fix racy locking of mballoc block bitmaps causing BUG
949 Details : The locking of the mballoc buddy bitmap and the in-memory
950 block bitmap was using two different spin locks in some
951 cases. This made it possible to incorrectly access the
952 mballoc bitmap while another process was modifying it,
953 causing a sanity assertion to fail. While no on-disk corruption
954 was reported, there was some risk of this happening.
957 Frequency : rare, on llog test 6
959 Descriptoin: don't allow connect to already connected import
960 Details : allowing connect to already connected import is hide connecting problem.
965 Descriptoin: allow kill process which wait statahead result
966 Details : for some reasons 'ls' can stick in waiting result from statahead,
967 in this case need way for kill this process.
969 Severity : enhancement
971 Description: Add state history info file, enhance import info file
972 Details : Track import connection state changes in a new osc/mdc proc file;
973 add overview-type data to the osc/mdc import proc file.
975 Severity : enhancement
977 Description: MDS create should not wait for statfs RPC while holding DLM lock.
979 Severity : enhancement
981 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
986 Descriptoin: OST grant too much space to client even there are not enough space.
987 Details : Client will shrink its grant cache to OST if there are no write
988 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
989 this grant cache if there are already not enough avaible space
990 (left_space < total_clients * 32M).
993 Frequency : start MDS on uncleanly shutdowned MDS device
995 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
996 Details : stay in waiting mds<>ost recovery finished produce random bugs
997 due race between two ll_sync thread for one lov target. send
998 ACTIVATE event only if connect realy finished and import have
1002 Frequency : rare, connect and disconnect target at same time
1004 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1005 Details : don't call obd_disconnect under lov_lock. this long time
1006 operation and can block ptlrpcd which answer to connect request.
1011 Descriptoin: don't lose wakeup for imp_recovery_waitq
1012 Details : recover_import_no_retry or invalidate_import and import_close can
1013 both sleep on imp_recovery_waitq, but we was send only one wakeup
1017 Frequency : always with long access acl
1019 Descriptoin: mds can't pack reply with long acl.
1020 Details : mds don't control size of acl but they limited by reint/getattr
1023 Severity : enhancement
1025 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1027 Severity : enhancement
1029 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1032 Frequency : start MDS on uncleanly shutdowned MDS device
1034 Descriptoin: aborting recovery hang on MDS
1035 Details : don't throttle destroy RPCs for the MDT.
1038 Frequency : on remount
1040 Description: external journal device not working after the remount
1041 Details : clear dev_rdonly flag for external journal devices in
1047 Description: shutdown vs evict race
1048 Details : client_disconnect_export vs connect request race.
1049 if client will evicted at this time - we start invalidate
1050 thread without referece to import and import can be freed
1054 Frequency : rare, need acl's on inode.
1056 Description: client can't handle ost additional correctly
1057 Details : if ost was added after client connected to mds client can have
1058 hit lnet_try_match_md ... to big messages to wide striped files.
1059 in this case need teach client to handle config events about add
1060 lov target and update client max ea size at that event.
1062 Severity : enhancement
1064 Description: Changelogs
1065 Details : Changelogs are a lightweight mechanism to track filesystem
1066 metadata and namespace changes. The changelog is recorded
1067 permanently on the MDTs, and is periodically "consumed" / purged
1068 when records are no longer needed.
1070 Severity : enhancement
1072 Description: compact fld format with extents
1073 Details : Store range of seq rather than every seq in FLD. Seq
1074 controller update FLD rather than clients. In Case of CMD, mdt0
1075 has FLD, all other metadata server act as non persistent proxy
1076 for FLD queries and cache fld entries in fld cache.
1081 Description: don't skip ost target if they assigned to file
1082 Details : Drop slow OSCs if we can, but not for requested start idx.
1083 This means "if OSC is slow and it is not the requested
1084 start OST, then it can be skipped, otherwise skip it only
1085 if it is inactive/recovering/out-of-space.
1089 Description: more cleanup in mds_lov
1090 Details : not send LOV EA under replay, we can't know about they size at this
1091 time. Don't allow client connect to mds before any ost connected,
1092 for avoid problems with LOV EA size and returning EIO to client.
1094 Severity : enhancement
1096 Description: Interoperability at server side (Disk interoperability)
1098 Severity : enhancement
1100 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1102 Severity : enhancement
1104 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1106 Severity : enhancement
1108 Description: New client IO stack (CLIO).
1110 Severity : enhancement
1112 Description: Commit on sharing. Eliminate inter-client dependencies between
1113 uncommitted transactions by doing transaction commits.
1114 Thereby clients may recovery independently.
1117 Frequency : Create a symlink file with a very long name
1119 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1120 Details : If there is no extra space in the request for early cancels,
1121 ldlm_req_handles_avail() returns 0 instead of a negative value.
1123 Severity : enhancement
1125 Description: Add /proc entry for import status
1126 Details : The mdc, osc, and mgc import directories now have
1127 an import directory that contains useful import data for debugging
1128 connection problems.
1130 Severity : enhancement
1132 Description: Re-disable certain /proc logging
1133 Details : Enable and disable client's offset_stats, extents_stats and
1134 extents_stats_per_process stats logging on the fly.
1137 Frequency : Only on FC kernels 2.6.22+
1139 Description: oops in statahead
1140 Details : Do not drop reference count for the dentry from VFS when lookup,
1141 VFS will do that by itself.
1143 Severity : enhancement
1145 Description: Generic /proc file permissions
1146 Details : Set /Proc file permissions in a more generic way to enable non-
1147 root users operate on some /proc files.
1151 Description: Hitting mdc_commit_close() ASSERTION
1152 Details : Properly handle request reference release in
1153 ll_release_openhandle().
1157 Description: quota recovery deadlock during mds failover
1158 Details : This patch includes att18982, att18236, att18237 in bz14840.
1160 1. fix osts hang when mds does failover with quotaon
1161 2. prevent watchdog storm when osts threads wait for the
1166 Frequency : only patchless client
1167 Description: add workaround for race between add/remove dentry from hash
1169 Severity : enhancement
1171 Description: Allow OST glimpses to return PW locks
1175 Description: LBUG when llog conf file is full
1176 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1180 Description: Prevent import from entering FULL state when server in recovery
1184 Description: service mount cannot take device name with ":"
1185 Details : Only when device name contains ":/" will mount treat it as
1191 Description: replace ptlrpcd with the statahead thread to interpret the async
1192 statahead RPC callback
1196 Frequency : on recovery
1197 Description: I/O failures after umount during fail back
1198 Details : if client reconnected to restarted server we need join to recovery
1199 instead of find server handler is changed and process self
1200 eviction with cancel all locks.
1202 Severity : enhancement
1204 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1206 Severity : enhancement
1208 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1210 Severity : enhancement
1212 Description: Upcall on Lustre log has been dumped
1213 Details : Allow for a user mode script to be called once a Lustre log has
1214 been dumped. It passes the filename of the dumped log to the
1215 script, the location of the script can be specified via
1216 /proc/sys/lnet/debug_log_upcall.
1221 Description: avoid idr_remove called for id which is not allocated.
1222 Details : Move assigment s_dev for clustered nfs to end of initialization,
1223 to avoid problem with error handling.
1228 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1229 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1230 into CONN_USED_HASH and this prodice warning when put connection
1231 again in unused hash.
1234 Severity : enhancement
1236 Description: File striping can now be set to use an arbitrary pool of OSTs.
1238 Severity : enhancement
1240 Description: Export bytes_read/bytes_write count on OSC/OST.
1244 Description: Early reply size mismatch, MGC loses connection
1245 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1246 the connect flags are properly negotiated.
1251 Description: quotas are not honored with O_DIRECT
1252 Details : all writes with the flag O_DIRECT will use grants which leads to
1253 this problem. Now using OBD_BRW_SYNC to guard this.
1257 Description: add quota statistics
1258 Details : 1. sort out quota proc entries and proc code.
1259 2. add quota statistics
1261 Severity : enhancement
1263 Description: enable quota support for HEAD.
1267 Description: Properly propagate oinfo flags from lov to osc for statfs
1268 Details : restore missing copy oi_flags to lov requests.
1270 Severity : enhancement
1272 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1273 : l_getgroups(8), lst(8), routerstat(8)
1274 Details : included man pages for llobdstat(8), llstat(8),
1275 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1277 Severity : enhancement
1279 Description: configure's --enable-quota should check the
1280 : kernel .config for CONFIG_QUOTA
1281 Details : configure is terminated if --enable-quota is passed but
1282 : no quota support is in kernel
1286 Description: Remove portals compatibility
1287 Details : Remove portals compatibility, not interoperable with releases
1292 Description: Resolve device initialization race
1293 Details : Prevent proc handler from accessing devices added to the
1294 obd_devs array but yet be intialized.
1296 Severity : enhancement
1298 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1300 Severity : enhancement
1302 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1307 Description: Using wrong pointer in osc_brw_prep_request
1308 Details : Access to array[-1] can produce panic if kernel compiled with
1309 CONFIG_PAGE_ALLOC enabled
1313 Description: Client runs out of low memory
1314 Details : Consider only lowmem when counting initial number of llap pages
1318 Description: *optional* service tags registration
1319 Details : if the "service tags" package is installed on a Lustre node
1320 When the filesystem is mounted, a local-node service tag will
1321 be created. See http://inventory.sun.com/ for more information
1322 about the Service Tags asset management system.
1326 Description: Kernel BUG tries to release flock
1327 Details : Lustre does not destroy flock lock before last reference goes
1328 away. So always drop flock locks when client is evicted and
1329 perform unlock regardless of successfulness of speaking to MDS.
1333 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1337 Description: sanity test 65a fails if stripecount of -1 is set
1338 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1343 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1344 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1347 Severity : enhancement
1349 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1353 Description: do not process already freed flock
1354 Details : flock can possibly be freed by another thread before it reaches
1355 to ldlm_flock_completion_ast.
1359 Description: LBUG during stress test
1360 Details : Need properly lock accesses the flock deadlock detection list.
1364 Description: oops in page fault handler
1365 Details : kernel page fault handler can return two special 'pages' in error
1366 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1370 Description: timeout with invalidate import.
1371 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1372 should be handled by ptlrpcd. This produce long age waiting and
1373 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1375 Severity : enhancement
1377 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1381 Description: enable MGS and MDT services start separately
1382 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1383 a co-located MGS without starting the MGS, which is a complement
1384 to 'nosvc' mount option.
1388 Frequency : after recovery
1389 Description: precreate to many object's after del orphan.
1390 Details : del orphan st in oscc last_id == next_id and this triger growing
1391 count of precreated objects. Set flag LOW to skip increase count
1392 of precreated objects.
1396 Frequency : rare, on clear nid stats
1397 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1398 Details : when clean nid stats sometimes try destroy live entry,
1399 and this produce panic in free.
1403 Description: Stack overflow during MDS log replay
1404 ease stack pressure by using a thread dealing llog_process.
1408 Description: wait until IO finished before start new when do lock cancel.
1409 Details : VM protocol want old IO finished before start new, in this case
1410 need wait until PG_writeback is cleared until check dirty flag and
1411 call writepages in lock cancel callback.
1413 Severity : enhancement
1415 Description: using special macro for print time and cleanup in includes.
1419 Description: mds_mfd_close() ASSERTION(rc == 0)
1420 Details : In mds_mfd_close(), we need protect inode's writecount change
1421 within its orphan write semaphore to prevent possible races.
1425 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1429 Frequency : rare, on shutdown ost
1430 Description: don't hit live lock with umount ost.
1431 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1432 use shrink_dcache_sb instead.
1436 Description: don't panic with use echo client
1437 Details : echo client pass NULL as client nid pointer and this produce null
1438 pointer dereference.
1442 Description: fix build on ppc32
1443 Details : compile code with -m64 flag produce wrong object file for ppc32.
1447 Description: add message levels for liblustreapi
1451 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1452 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1453 will result in an -EAGAIN and in the exit call path we call
1454 alloc_rr() which will with increasing aggressiveness attempt to
1455 aquire precreated objects on the minimum number of required OSCs.
1459 Description: Use old size assignment to avoid deadlock
1460 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1461 the scheduling while holding a spinlock. We do not need locking
1462 for size in ll_update_inode() because size is only updated from
1463 the MDS for directories or files without objects, so there is no
1464 other place to do the update, and concurrent access to such inodes
1465 are protected by the inode lock.
1469 Description: resolve "_IOWR redefined" build error on SLES10
1473 Description: dump the memory debugging after all modules are unloaded to
1474 suppress false negative in conf_sanity test 39
1476 Severity : enhancement
1478 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1483 Description: fix SLES kernel versioning
1484 Details : the kernel version for our SLES 10 kernel did not include a "-"
1485 before the "smp" at the end. while this was not a problem in
1486 general, it did mean that software trying to use the kernel
1487 version to try to detect a vendor specific kernel would fail.
1488 this was most evident by the OFED build scripts.
1492 Description: Don't update lov_desc members until making sure they are valid
1493 Details : When updating lov_desc members via proc fs, need fix their
1494 validities before doing the real update.
1498 Description: don't put request into delay list while invalidate in flight.
1499 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1500 import in flight. this produce timeout for invalidate and sometimes
1501 can cause stale data.
1505 Frequency : on ppc only
1506 Description: not convert ost objects for directory because it's not exist.
1507 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1511 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1513 Severity : enhancement
1515 Description: Update to support 2.6.22.14 vanilla kernel.
1519 Frequency : rare, on recovery
1520 Description: read procfs can produce deadlock in some situation
1521 Details : Holding lprocfs lock which send rpc can produce block for destroy
1522 obd objects and this also block reconnect with -EALREADY. This
1523 isn't fix all lprocfs bugs - but make it rare.
1525 Severity : enhancement
1527 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1530 Frequency : frequent on X2 node
1532 Description: mdc_set_open_replay_data LBUG
1533 Details : Set replay data for requests that are eligible for replay.
1537 Description: lustre_mgs: operation 101 on unconnected MGS
1538 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1539 MGC, and late on MGC cannot successfully connect to MGS and a lot
1540 of the error messages complaining that MGS is not connected.
1544 Frequency : on network error
1545 Description: panic with double free request if network error
1546 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1547 but it's true only for synchronus enqueue, for async enqueue
1548 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1551 Severity : enhancement
1553 Description: client-side metadata stat-ahead during readdir(directory readahead)
1554 Details : perform client-side metadata stat-ahead when the client detects
1555 readdir and sequential stat of dir entries therein
1558 Frequency : on start mds
1560 Description: Implement get_info(last_id) in obdfilter.
1563 Frequency : occasional
1565 Description: Correctly check stale fid, not start epoch if ost not support SOM
1566 Details : open with flag O_CREATE need set old fid in op_fid3 because
1567 op_fid2 was overwritten with new generated fid, but mds can answer
1568 with one of these two fids and both is not stale. Setattr
1569 incorrectly started an epoch and assume will be called
1570 done_writing, but without SOM done_writing ever being called.
1573 Frequency : rare, depends on device drivers and load
1575 Description: MDS or OSS nodes crash due to stack overflow
1576 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1577 In some cases, in conjunction with device drivers that use a lot
1578 of stack the MDS (or possibly OSS) service threads could overflow
1579 the stack. One change which was identified to consume additional
1580 stack has been reworked to avoid the extra stack usage.
1583 Frequency : occasional
1585 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1586 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1587 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1588 oscc_has_objects and see OSC has no objests and call
1589 oscc_internal_create to resend create request.
1591 Severity : enhancement
1593 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1595 Severity : enhancement
1597 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1600 Frequency : very rare
1602 Description: Fix replay if there is an un-replied request and open
1603 Details : In some cases, older replay request will revert the
1604 mcd->mcd_last_xid on MDS which is used to record the client's
1605 latest sent request.
1607 Severity : enhancement
1609 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1611 Severity : enhancement
1613 Description: Add rhel5 support to HEAD.
1615 Serverity : enhancement
1617 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1622 Description: Don't allow skipping OSTs if index has been specified.
1623 Details : Don't allow skipping OSTs if index has been specified, make
1624 locking in internal create lots better.
1628 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1629 Details : access to bitfield in structure is always rounded to long
1630 and this produce problem with not atomic change any bit.
1634 Description: Lustre make rpms failed.
1635 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1636 builds Lustre rpms from tarball.
1639 Frequency : rare on shutdown ost
1641 Description: If llog cancel was not send before clean_exports phase, this can
1642 produce deadlock in llog code.
1643 Details : If llog thread has last reference to obd and call class_import_put
1644 this produce deadlock because llog_cleanup_commit_master wait when
1645 last llog_commit_thread exited, but this never success because was
1646 called from llog_commit_thread.
1650 Description: allow userland application know is lost one of stripes.
1651 Details : fill lvb_blocks with error code on ost and return it to
1652 application if error flag found.
1656 Description: NULL lov_tgts causing MDS oops
1657 Details : more safe checks for NULL lov_tgts for avoid oops.
1659 Severity : enhancement
1661 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1665 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1666 Details : If an OST is down the MDS will hang indefinitely in
1667 obd_statfs_rqset() waiting for the statfs data. While for
1668 MDS QOS usage of statfs, it should not stuck in waiting.
1670 Severity : enhancement
1672 Description: remote_acl support
1673 Details : Support ACL-based permission check for remote user.
1674 Support setfacl/getfacl for remote user with the utils
1675 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1676 the system "{s,g}etfacl" utils.
1678 Severity : enhancement
1680 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1682 Severity : enhancement
1684 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1688 Description: make mgs_setparam() handle fsname containing dash
1689 Details : fsname containing a dash does not work with lctl conf_param
1691 Severity : enhancement
1693 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1695 Severity : enhancement
1697 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1701 Frequency : rare, at shutdown
1702 Description: access already free / zero obd_namespace.
1703 Details : if client_disconnect_export was called without force flag set,
1704 and exist connect request in flight, this can produce access to
1705 NULL pointer (or already free pointer) when connect_interpret
1706 store ocd flags in obd_namespace.
1710 Frequency : only at startup
1711 Description: not alloc memory with spinlock held.
1712 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1713 if any spinlock held.
1715 Severity : enhancement
1717 Description: make lustre randomly fail allocating memory
1718 Details : Make lustre randomly failed allocating memory for testing purpose.
1720 Severity : enhancement
1722 Description: lost problems with lov objid file
1723 Details : Fixes some scability and access to not inited memory problems
1724 in work with lov objdid file.
1729 Description: lfs find does not continue on file error
1730 Details : Continue other files processing when a file/dir is absent.
1734 Description: Inconsistent usage of lustre_pack_reply()
1735 Details : Standardize the usage of lustre_pack_reply() such that it
1736 always generate a CERROR on failure.
1739 Frequency : occasional
1741 Description: MDS hang or stay in waiting lock
1742 Details : If client receive lock with CBPENDING flag ldlm need send lock
1743 cancel as separate rpc, to avoid situation when cancel request
1744 can't processed due all i/o threads stay in wait lock.
1748 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1752 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1756 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1760 Description: Update to RHEL4 latest kernel.
1762 Severity : enhancement
1764 Description: Build SLES10 patchless client fails
1765 Details : The configure was broken by run ./configure with
1766 --with-linux-obj=.... argument for patchless client. When the
1767 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1768 can't search header adequately. Use absolute path such as
1769 -I($LINUX)/include instead.
1773 Description: interrupt oig_wait produce painc on resend.
1774 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1775 private set, and this produce situation when rq_ptlrpcd_data not
1776 copyed to new allocated request and triggered LBUG on assert
1777 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1778 ptlrpcd set if request is changed and can be safety changed to use
1783 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1784 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1785 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1786 these macros and replace them with their difinitions instead.
1790 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1791 Details : Update which_patch & target file for SLES10 latest kernel.
1794 Description: Remove client patches from SLES 10 kernel.
1795 Details : This causes SLES 10 clients to behave as patchless clients
1796 even on a Lustre-patched (server) kernel.
1798 Severity : enhancement
1800 Description: self-adjustable client's lru lists
1801 Details : use adaptive algorithm for managing client cached locks lru
1802 lists according to current server load, other client's work
1803 pattern, memory activities, etc. Both, server and client
1804 side namespaces provide number of proc tunables for controlling
1807 Severity : enhancement
1809 Description: light-weight GSS support
1810 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1811 and reduce performance overhead.
1813 Severity : enhancement
1815 Description: Linux keyring support
1816 Details : Support using service of Linux keyring for Lustre GSS internal
1817 context refresh/cache mechanism.
1821 Description: Fix errors in lfs documentation
1822 Details : Fixes man pages
1826 Description: don't use GFP_* in generic Lustre code.
1827 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1831 Description: obdclass is limited by single OBD_ALLOC(idarray)
1832 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1836 Description: warnings with build patchless client with vanila 2.6.19 and up
1837 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1838 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1842 Frequency : Only for SLES
1843 Description: sanity_quota fail test_1
1844 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1845 file, which makes TSTID[2] inunique.
1849 Frequency : Only for SLES9
1850 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1851 Details : Update target/ChangeLog/which_patch .
1855 Frequency : Only for test_10 in sanity-quota.sh
1856 Description: a bug in quota test code
1857 Details : checking a test flag in a wrong place
1861 Frequency : Only for test_13 in sanity-quota.sh
1862 Description: a bug in quota test script
1863 Details : shouldn't check overfull allocated quota
1867 Description: jbd statistics
1868 Details : Port older jbd statistics patch for sles10
1870 Severity : enhancement
1872 Description: Add printing inode info into message about error in writepage.
1877 Description: reply_lock_interpret crash due to race with it and lock cancel.
1878 Details : Do not replay locks that are being cancelled. Do not reference
1879 locks by their address during replay, just by their handle.
1883 Frequency : When flocks are used.
1884 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1885 Details : Flock locks might destroy just granted lock if it could be merged
1886 with another existing flock, this is done in completion handler,
1887 so teach ldlm_cli_enquque_fini that this is a valid case for
1893 Description: Oops in read and write path when failing to allocate lock.
1894 Details : Check if lock allocation failed and return error back.
1898 Description: lstripe command fails for valid OST index
1899 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1900 instead of lov->desc.ld_active_tgt_count.
1904 Description: sanity.sh failed test 103
1905 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1910 Description: df doesn't work properly if diskfs blocksize != 4K
1911 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1915 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1916 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1918 Severity : enhancement
1920 Description: Adaptive timeouts
1921 Details : RPC timeouts adapt to changing server load and network
1922 conditions to reduce resend attempts and improve recovery time.
1926 Description: llapi_file_create() does not allow some changes
1927 Details : add llapi_file_open() that allows specifying the mode and
1928 open flags, and also returns an open file handle.
1932 Description: merge and cleanup kernel patches.
1933 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1937 Description: Add journal checksum support.(Kernel part)
1938 Details : The journal checksum feature adds two new flags i.e
1939 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1940 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1941 indicates that the commit block contains the checksum for
1942 the blocks described by the descriptor blocks. Now commit
1943 record can be sent to disk without waiting for descriptor
1944 blocks to be written to disk. This behavior is controlled
1945 using JBD2_FEATURE_ASYNC_COMMIT flag.
1949 Description: OSS needs mutliple precreate threads
1950 Details : Add ability to start more than one create thread per OSS.
1954 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1955 Details : execute lfs setstripe on client
1959 Description: mds_obd_create error creating tmp object
1960 Details : When the user sets quota on root, llog will be affected and can't
1961 create files and write files.
1964 Frequency : Always on ia64 patchless client, and possibly others.
1966 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1967 Details : This allows the patchless client to be loaded on architectures
1968 without this export.
1972 Description: RedHat Update kernel for RHEL5
1973 Details : Add the kernel config file for RHEL5.
1977 Description: RedHat Update kernel for RHEL5
1978 Details : Modify the kernel config file more closer RHEL5.
1982 Description: Build failure against Centos5 (RHEL5)
1983 Details : Define PAGE_SIZE when it isn't present.
1987 Description: test_7 on sanity-quota.sh can't be used on separate machines
1988 Details : add facet to handle it
1992 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1993 Details : with NFS, the anon dentry's parent was set to itself in
1994 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1995 corresponding dentry other than use rec->ur_name.
1997 Severity : enhancement
1999 Description: lfs setstripe enhancement
2000 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2004 Description: enable data checksumming by default
2005 Details : enable checksum by default, allow --disable-checksum
2006 configure option and "-o nochecksum" mount option. Checksums
2007 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2011 Description: lustre support for RHEL5
2012 Details : Add support for RHEL5.
2016 Description: Client eviction due to failover config
2017 Details : after a connection loss, the lustre client should attempt to
2018 reconnect to the last active server first before trying the
2019 other potential connections.
2023 Description: when mds and osts use different quota unit(32bit and 64bit),
2024 quota will be released repeatly.
2025 Details : void sending multiple quota reqs to mds, which will keep the
2026 status between the reqs.
2030 Description: osts not allocated evenly to files
2031 Details : change the condition to increase offset_idx
2034 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2036 Description: Data corruption for OSTs that are formatted larger than 2TB
2038 Details : When generating the bio request for lustre file writes the
2039 sector number would overflow a temporary variable before being
2040 used for the IO. The data reads correctly from Lustre (which
2041 will overflow in a similar manner) but other file data or
2042 filesystem metadata may be corrupted in some cases.
2046 Description: Tune the kernel for good SCSI performance.
2047 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2048 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2053 Description: rewrite ext2-derived code in obdclass/uuid.c
2054 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2055 from scratch preserving functionality.
2059 Description: "lfs find -obd UUID" prints directories
2060 Details : "lfs find -obd UUID" will return all directory names instead
2061 of just file names. It is incorrect because the directories
2062 do not reside on the OSTs.
2066 Description: use i_size_read and i_size_write in 2.6 port
2067 Details : replace inode->i_size access with i_size_read/write()
2070 Frequency : when removing large files
2072 Description: scheduling issue during removal of large Lustre files
2073 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2074 It causes scheduling issues when removing large files (17TB in the
2078 Frequency : only with liblustre clients on XT3
2080 Description: evictions taking too long
2081 Details : allow llrd to evict clients directly on OSTs
2084 Frequency : only on ppc
2086 Description: /proc/fs/lustre/devices broken on ppc
2087 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2088 We should make sure the type of 'index' is loff_t and then cast
2089 explicitly as needed below. Do not assign an explicitly cast
2093 Frequency : only for rhel5
2095 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2096 Details : Modify the target file & which_kernel.
2098 Severity : enhancement
2100 Description: omit set fsid for export NFS
2101 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2102 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2106 Frequency : Always, for kernels after 2.6.16
2107 Description: Fix warning idr_remove called for id=.. which is not allocated.
2108 Details : Last kernels save old s_dev before kill super and not allow
2109 to restore from callback - restore it before call kill_anon_super.
2113 Description: buffer overruns could theoretically occur
2114 Details : llapi_semantic_traverse() modifies the "path" argument by
2115 appending values to the end of the origin string, and a
2116 overrun may occur. Adding buffer overrun check in liblustreapi.
2120 Description: Fix error on 'ls .' at the top of the Lustre mount.
2121 Details : Don't revalidate dentry if it is a root dentry.
2125 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2126 Details : Modify vm-tunables-rhel4.patch.
2130 Description: Kernel config for 2.6.18-vanilla.
2131 Details : Modify targets/2.6-vanilla.target.in.
2132 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2133 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2134 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2135 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2139 Description: improve handling recoverable errors
2140 Details : if request processig with error which can be recoverable on server
2141 request should be resend, otherwise page released from cache and
2146 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2147 Details : Modify target file & which_patch.
2148 A flaw was found in the IA32 system call emulation provided
2149 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2150 value could be stored in the %RAX register, which could trigger an
2151 out-of-bounds system call table access. An untrusted local user
2152 could exploit this flaw to run code in the kernel
2153 (ie a root privilege escalation). (CVE-2007-4573).
2157 Description: O_DIRECT bypasses client statistics.
2158 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2159 read_ahead_stats were updated. copied the stats section from
2160 osc_send_oap_rpc() into async_internal().
2164 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2168 Description: change order of libsysio includes
2169 Details : '#include sysio.h' should always come before '#include xtio.h'
2173 Description: handle "serious error: objid * already exists" more gracefully
2174 Details : If LAST_ID value on disk is smaller than the objects existing in
2175 the O/0/d* directories, it indicates disk corruption and causes an
2176 LBUG(). If the object is 0-length, then we should use the existing
2177 object. This will help to avoid a full fsck in most cases.
2179 Severity : enhancement
2181 Description: adapt the lustre_config script to support the upgrade case
2182 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2183 server targets to 1.6 in parallel.
2187 Description: To avoid grant space > avaible space when the disk is almost
2188 full. Without this patch you might see the error "grant XXXX >
2189 available" or some LBUG about grant, when the disk is almost
2191 Details : In filter_check_grant, for non_grant cache write, we should
2192 check the left space by if (*left > ungranted + bytes), instead
2193 of (*left > ungranted), because only we are sure the left space
2194 is enough for another "bytes", then the ungrant space should be
2195 increase. In client, we should update cl_avail_grant only there
2196 is OBD_MD_FLGRANT in the reply.
2200 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2201 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2202 described in CVE-2007-4573. This problem could allow untrusted
2203 local users to gain root access.
2206 Frequency : when using O_DIRECT and quotas
2208 Description: Incorrect file ownership on O_DIRECT output files
2209 Details : block usage reported by 'lfs quota' does not take into account
2210 files that have been written with O_DIRECT.
2215 Description: touch file failed when fs is not full
2216 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2217 otherwise we can get ENOSP while fs is not full.
2221 Description: parallel lock callbacks
2222 Details : Instead of sending blocking and completion callbacks as separated
2223 requests, adding them to a set and sending in parallel.
2226 Frequency : only for Cray XT3
2227 Bugzilla : 12829/13455
2228 Description: Changing primary group doesn't change the group lustre assigns to
2230 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2231 overridden with the primary group provided by the group upcall,
2232 whereas the supplied fsgid can be trusted if it is in the list of
2233 supplementary groups returned by the group upcall.
2235 Severity : enhancement
2237 Description: Allow masking D_WARNING, D_ERROR messages from console
2238 Details : Console messages can now be disabled via lnet.printk.
2242 Description: User code with malformed file open parameter crashes client node
2243 Details : Before packing join_file req, all the related reference should be
2244 checked carefully in case some malformed flags cause fake
2245 join_file req on client.
2249 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2250 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2251 mode is equal to the granted mode. However, it should also check
2252 that LDLM_FL_CP_REQD is not set.
2256 Description: Heavy nfs access might result in deadlocks
2257 Details : After ELC code landed, it is now improper to enqueue any mds
2258 locks under och_sem, because enqueue might want to decide to
2259 cancel open locks for same inode we are holding och_sem for.
2263 Description: Client eviction while running blogbench
2264 Details : A lot of unlink operations with concurrent I/O can lead to a
2265 deadlock causing evictions. To address the problem, the number of
2266 oustanding OST_DESTROY requests is now throttled to
2267 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2268 callbacks are processed in priority.
2272 Description: enable ACLs on MDS by default
2273 Details : ACLs must be enabled on MDS by default.
2276 Frequency : PPC/PPC64 only
2278 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2279 Details : fix duplicated definitions between asm-ppc64/types.h and
2280 lustre_types.h on PPC.
2283 Frequency : PPC/PPC64 only
2285 Description: asm-ppc/segment.h does not exist
2286 Details : fix compile issue on PPC.
2290 Description: better handle error messages in extents code
2293 Frequency : RHEL4 only
2295 Description: mkfs is very slow on IA64/RHEL4
2296 Details : A performance regression has been discovered in the MPT Fusion
2297 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2298 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2299 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2301 Severity : enhancement
2303 Description: SNMP support enhancement
2304 Details : Adding total number of sampled request for an MDS node in snmp
2307 Severity : enhancement
2309 Description: Optimize ldlm waiting list processing for PR extent locks
2310 Details : When processing waiting list for read extent lock and meeting read
2311 lock that is same or wider to it that is not contended, skip
2312 processing rest of the list and immediatelly return current
2313 status of conflictness, since we are guaranteed there are no
2314 conflicting locks in the rest of the list.
2318 Description: Time out and refuse to reconnect
2319 Details : When the failover node is the primary node, it is possible
2320 to have two identical connections in imp_conn_list. We must
2321 compare not conn's pointers but NIDs, otherwise we can defeat
2322 connection throttling.
2326 Description: port llog fixes from b1_6 into HEAD
2327 Details : Port llog reference couting and some llog cleanups from b1_6
2328 (bug 10800) into HEAD, for protect from panic and access to already
2329 free llog structures.
2333 Description: Detect stride IO mode in read-ahead
2334 Details : When a client does stride read, read-ahead should detect that and
2335 read-ahead pages according to the detected stride pattern.
2339 Description: data checksumming impacts single node performance
2340 Details : add support for several checksum algorithm. Currently, only CRC32
2341 and Adler-32 are supported. The checksum type can be changed on
2342 the fly via /proc/fs/lustre/osc/*/checksum_type.
2346 Description: use adler32 for page checksums
2347 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2352 Description: build for x2 fails
2353 Details : fix compile issue on Cray systems.
2357 Description: Properly match for duplicate locks
2358 Details : Due to different lock order from skiplists code, we need to
2359 traverse entire list for now
2362 Frequency : only on PPC/SLES10
2364 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2365 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2366 be defined. Add a hack in mkfs_lustre.c to work around this compile
2371 Description: LASSERT on MDS when client holding flock lock dies
2372 Details : ldlm pool logic depends on number of granted locks equal to
2373 number of released locks which is not true for flock locks, so
2374 just exclude such locks from consideration.
2378 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2379 Details : Use fsfilt_sync() for both the whole filesystem sync and
2380 individual file sync to eliminate dangerous inode locking
2381 with I_LOCK that can lead to a deadlock.
2385 Description: performance in 1.6.3
2386 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2387 sure that 1MB requests are not fragmented by the block layer.
2389 Severity : enhancement
2391 Description: organize the server-side client stats on per-nid basis
2392 Details : Change the structure of stats under obdfilter and mds to
2400 The "uuid"s file would list the uuids of _active_ exports.
2401 And the clear entry is to clear all stats and stale nids.
2403 Severity : enhancement
2405 Description: eliminate client locks in face of contention
2406 Details : file contention detection and lockless i/o implementation
2407 for contended files.
2411 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2414 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2416 Description: MDS may lose file striping (and hence file data) in some cases
2417 Details : If there are additional extended attributes stored on the MDS,
2418 in particular ACLs, SELinux, or user attributes (if user_xattr
2419 is specified for the client mount options) then there is a risk
2420 of attribute loss. Additionally, the Lustre file striping
2421 needs to be larger than default (e.g. striped over all OSTs),
2422 and an additional attribute must be stored initially in the
2423 inode and then increase in size enough to be moved to the
2424 external attribute block (e.g. ACL growing in size) for file
2429 Description: skiplist implementation simplification
2430 Details : skiplists are used to group compatible locks on granted list
2431 that was implemented as tracking first and last lock of each
2432 lock group the patch changes that to using doubly linked lists
2436 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2437 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2438 because LASSERT on both the data supplied by a client, and the
2439 data on disk is dangerous and incorrect.
2441 Severity : enhancement
2443 Description: Slow truncate/writes to huge files at high offsets.
2444 Details : Directly associate cached pages to lock that protect those pages,
2445 this allows us to quickly find what pages to write and remove
2446 once lock callback is received.
2450 Description: more ldlm soft lockups
2451 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2452 starve other threads from the resource lock for a long time in
2453 case of long waiting queue, so change the debug level from
2454 D_OTHER to the less frequently used D_INFO.
2456 Severity : enhancement
2458 Description: add -gid, -group, -uid, -user options to lfs find
2462 Description: Hung threads in invalidate_inode_pages2_range
2463 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2464 once one is completed. As a result, some RPCs are stuck in the
2465 queue and are never sent.
2469 Description: filter threads hungs on waiting journal commit
2470 Details : Cleanup filter group llog code, then only filter group llog will
2471 be only created in the MDS/OST syncing process.
2475 Description: Procfs and llog threads access destoryed import sometimes.
2476 Details : Sync the import destoryed process with procfs and llog threads by
2477 the import refcount and semaphore.
2479 Severity : enhancement
2481 Description: openlock cache of b1_6 port to HEAD
2486 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2487 Details : If insertion of an extent fails, then discard the inode
2488 preallocation and free data blocks else it can lead to duplicate
2493 Description: don't always update ctime in ext3_xattr_set_handle()
2494 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2495 In some cases the ctime should not be updated, for example for
2496 2.0->1.8 compatibility it is necessary to delete an xattr and it
2497 should not update the ctime.
2501 Bugzilla : 15713/16362
2502 Description: Assertion in iopen_connect_dentry in 1.6.3
2503 Details : looking up an inode via iopen with the wrong generation number can
2504 populate the dcache with a disconneced dentry while the inode
2505 number is in the process of being reallocated. This causes an
2506 assertion failure in iopen since the inode's dentry list contains
2507 both a connected and disconnected dentry.
2511 Description: assertion failure in ldlm_handle2lock()
2512 Details : fix a race between class_handle_unhash() and class_handle2object()
2513 introduced in lustre 1.6.5 by bug 13622.
2518 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2519 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2524 Description: oops in sd_iostats_seq_show()
2525 Details : unloading/reloading the scsi low level driver triggers a kernel
2526 bug when trying to access the sd iostat file.
2531 Description: Kernel panics during QLogic driver reload
2532 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2533 patch, causing memory corruption.
2538 Description: journal_dev option does not work in b1_6
2539 Details : pass mount option during pre-mount.
2541 Severity : enhancement
2543 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2544 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2545 extent information of a file. It can be used to map logical blocks
2546 in a file to physical blocks in the block device.
2550 Description: LDLM soft lockups - improvement
2551 Details : It is be possible to send the lock handle along with each read
2552 or write request because the client is already doing a lock match
2553 itself so there isn't any reason the OST should have to re-do that
2557 Frequency : only on Cray X2
2559 Description: X2 build failures
2560 Details : fix build failures on Cray X2.
2564 Description: xid & resent requests
2565 Details : Initialize RPC XID from clock at startup (randomly if clock is
2568 Severity : enhancement
2570 Description: Add lustre_start utility to start or stop multiple Lustre servers
2575 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2576 Details : In case of memory pressure, list_del() can be called twice on
2577 req->rq_history_list, causing a kernel oops.
2581 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2582 Details : fix stack overflow in the distributed lock manager by defering
2583 export eviction after a failed AST to the elt thread instead of
2584 handling it in the dlm interpret routine.
2588 Description: Convert some comments to new format.
2589 Details : Update documenting comments to match doxygen conventions.
2593 Description: Grammar fixes.
2594 Details : A couple of trivial spelling fixes.
2598 Description: OSD_COUNTERS-mandatory
2599 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2600 to simplify the code.
2604 Description: simplify cmm_device freeing logic.
2605 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2609 Description: Add lockdep support to dt_object_operations locking interface.
2610 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2611 parameter indicating lock ordering. Update mdd code to use new
2616 Description: Introduce failloc constants for lockless IO tests.
2617 Details : Add two new failloc constants to test lockless IO. Only one of
2618 them in implemented---another is checked in yet to be landed
2623 Description: Add lockdep support for inode mutex.
2624 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2625 in the situations where Lustre has to lock more than one inode
2630 Description: Add optional invariants checking support.
2631 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2632 configure switch. This macro is to be used for consistency and
2633 sanity checks that are too expensive to be left in `production'
2638 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2639 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2644 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2645 Details : Use spin_lock_nested() in (the only) situation where more than
2646 one ldlm_lock is locked simultaneously. Also, fix possible
2647 dead-lock in ldlm_lock_change_resource() by enforcing particular
2652 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2653 Details : Instead of specifying each ldlm_lock call-back through separate
2654 parameter, wrap them into struct ldlm_callback_suite.
2658 Description: Kill join_lru obd method and its callers.
2659 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2660 in memory---a requirement implemented through ->o_join_lru() obd
2661 method in HEAD. Remove this method and its users.
2665 Description: Add asynchronous ldlm ENQUEUE completion handler.
2666 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2667 case that stock ldlm_completion_ast() cannot handle as it waits
2668 until lock is granted. Introduce new ldlm_completion_ast_async()
2669 for this. Also comment ldlm_completion_ast().
2673 Description: ldlm_error <-> errno conversion.
2674 Details : Add functions to map (rather arbitrary) between LDLM error codes
2675 and standard errno values. CLIO needs this to prevent LDLM specific
2676 constants from escaping ldlm and osc.
2680 Description: Kill unused ldlm_handle2lock_ns() function.
2681 Details : Kill unused ldlm_handle2lock_ns() function.
2685 Description: Add lu_ref support to ldlm_lock
2686 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2687 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2688 and ldlm_resource. LDLM interface has to be changed, because code that
2689 releases a reference on a lock, has to "know" what reference this is.
2690 In the most frequent case
2692 lock = ldlm_handle2lock(handle);
2694 LDLM_LOCK_PUT(lock);
2696 no changes are required. When any other reference (received _not_ from
2697 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2698 instead of LDLM_LOCK_PUT().
2700 Arguably, changes are pervasive, and interface requires some discipline
2701 for proper use. On the other hand, it was very instrumental in finding
2702 a few leaked lock references.
2706 Description: Add ldlm_lock_addref_try().
2707 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2708 attempts to addref a lock that might be being canceled
2713 Description: Add ldlm_weigh_callback().
2714 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2715 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2716 hard-coded `number of pages' logic.
2720 Description: Add lockdep annotations to llog code.
2721 Details : Use appropriately tagged _nested() locking calls in the places
2722 where llog takes more than one ->lgh_lock lock.
2726 Description: Add loi_kms_set().
2727 Details : Wrap kms updates into a helper function.
2731 Description: Constify instances of struct lsm_operations.
2732 Details : Constify instances of struct lsm_operations.
2736 Description: lu_conf support.
2737 Details : On a server, a file system object is uniquely identified
2738 by a fid, which is sufficient to locate and load all object
2739 state (inode). On a client, on the other hand, more data are
2740 necessary instantiate an object. Change lu_object_find() and
2741 friends to take additional `lu_conf' argument describing object.
2742 Typically this includes layout information.
2746 Description: lu_context fixes.
2747 Details : Introduce new lu_context functions that are needed on the client
2748 side, where some system threads (ptlrpcd) are shared by multiple
2749 modules, and so cannot be stopped during module shutdown.
2753 Description: Add start and stop methods to lu_device_type_operations.
2754 Details : Introduce two new methods in lu_device_type_operations, that are
2755 invoked when first instance of a given type is created and last one
2756 is destroyed respectively. This is need by CLIO.
2760 Description: Add lu_ref support to struct lu_device.
2761 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2762 track leaked references.
2766 Description: Introduce lu_kmem_descr.
2767 Details : lu_kmem_descr and its companion interface allow to create
2768 and destroy a number of kmem caches at once.
2772 Description: Fix lu_object finalization race.
2773 Details : Fix a race between lu_object_find() finding an object and its
2774 concurrent finalization. This race is (most likely) not possible
2775 on the server, but might happen on the client.
2779 Description: Introduce lu_ref interface.
2780 Details : lu_ref is a debugging module allowing to track references to
2781 a given object. It is quite cpu expensive, and has to be
2782 explicitly enabled with --enable-lu_ref. See usage description
2787 Description: Factor lu_site procfs stats into a separate function.
2788 Details : Separate lu_site stats printing code into a separate function
2789 that can be reused on a client.
2793 Description: Constify instances of struct {lu,dt,md}_device_operations.
2794 Details : Constify instances of struct {lu,dt,md}_device_operations.
2798 Description: Introduce struct md_site and move meta-data specific parts of
2799 struct lu_site here.
2800 Details : Move md-specific fields out of struct lu_site into special struct
2801 md_site, so that lu_site can be used on a client.
2805 Description: Kill mdd_lov_destroy().
2806 Details : Remove unused mdd code.
2810 Description: Add st_block checking to multistat.c.
2811 Details : Add st_block checking to multistat.c.
2815 Description: Add lu_ref support to struct obd_device.
2816 Details : Add lu_ref tracking to obd_device.
2820 Description: Kill obd_set_fail_loc().
2821 Details : Remove unused code.
2825 Description: Add special type for ptlrpc_request interpret functions.
2826 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2827 there. Actual usage will be in CLIO.
2831 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2832 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2833 doesn't work with lockdep.
2837 Description: Add rwv.c test program.
2838 Details : New testing program exercising readv(2) and writev(2) (Qian).
2842 Description: Add sendfile.c test program.
2843 Details : New testing program exercising sendfile(2) (Jay).
2847 Description: Ratelimit a message that can be very frequent.
2848 Details : Ratelimit a memory allocation failure message that can
2853 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2854 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2858 Description: Liblustre build fixes.
2859 Details : Liblustre build fixes.
2863 Description: libcfs: add cfs_{need,cond}_resched() interface.
2864 Details : libcfs: add cfs_{need,cond}_resched() definition and
2865 implementations for Linux, NT, and liblustre.
2867 Severity : enhancement
2869 Description: More exported tunables for mballoc
2870 Details : Add support for tunable preallocation window and new tunables for
2871 large/small requests
2875 Description: Detect corruption of block bitmap and checking for preallocations
2876 Details : Checks validity of on-disk block bitmap. Also it does better
2877 checking of number of applied preallocations. When corruption is
2878 found, it turns filesystem readonly to prevent further corruptions.
2882 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2883 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2884 window to 0. If it does hit cache miss, and read-pattern is still
2885 stride-io mode, does not reset the stride window, but also does not
2886 increase the stride window length in this case.
2890 Frequency : only for big-endian servers
2891 Description: Check if system is big-endian while mounting fs with extents feature
2892 Details : Mounting a filesystem with extents feature will fail on big-endian
2893 systems since ext3-based ldiskfs is not supported on big-endian
2894 systems. This can be over-riden with "bigendian_extents" mount option.
2896 Severity : enhancement
2898 Description: The root squash functionality
2899 Details : A security feature, which is to prevent users from being able
2900 to mount lustre on their desktop, run as root, and delete
2901 all of the files in the filesystem. The goal is accomplished by
2902 remapping user id (UID) and group id (GID) of the root user to
2903 a UID and GID specified by the system administartor via Lustre
2904 configuration management server (MGS). The functionality also
2905 allows to specify sets of clients for which the remapping does
2910 Description: Excessive recovery window
2911 Details : With AT enabled, the recovery window can be excessively long (6000+
2912 seconds). To address this problem, we no longer use
2913 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2914 timeout no longer depends on the service time, it is set to
2915 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2916 time via pb_service_time.
2920 Description: Watchdog triggered on MDS failover
2921 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2922 the OSTs know that the MDS "UUID" can be reused for the same export
2923 from a different NID, so we do not need to wait for the export to be
2927 Frequency : rare, only if using MMP with Linux RAID
2929 Description: MMP doesn't work with Linux RAID
2930 Details : While using HA for Lustre servers with Linux RAID, it is possible
2931 that MMP will not detect multiple mounts. To make this work we
2932 need to unplug the device queue in RAID when the MMP block is being
2933 written. Also while reading the MMP block, we should read it from
2934 disk and not the cached one.
2936 Severity : enhancement
2938 Description: open file using fid
2939 Details : A file can be opened using just its fid, like
2940 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2943 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2946 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2947 Details : add proper locking for creation of jbd_4k slab cache
2951 Description: MMP check in ext3_remount() fails without displaying any error
2952 Details : When multiple mount protection fails during remount, proper error
2955 Severity : enhancement
2957 Description: Allow stripe size to be up to 4G-64k
2958 Details : Fix math logic to allow large stripe sizes.
2962 Description: add check for >8TB ldiskfs filesystems
2963 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2964 Don't allow >8TB ldiskfs filesystems to be mounted without
2965 force_over_8tb mount option
2970 Descriptoin: OSS complains "too many blocks" and crashes
2971 Details : In filter_commitrw_write():
2972 1. lquota_chkquota();
2973 2. obdo_from_inode();
2974 3. lquota_pending_commit();
2975 "2." has changed uid and gid in this case, then caused
2976 lquota_chkquota() and lquota_pending_commit() to use
2977 different uid and gid. That is the root of the bug.
2981 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2982 Details : Detect and print out the 'stripe count = -1' case properly.
2986 Description: Out or order replies might be lost on replay
2987 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2988 request with tid smaller than one currently being added, add it
2989 to the start, not end of the list.
2991 --------------------------------------------------------------------------------
2993 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2995 * Support for kernels:
2996 2.6.5-7.283 (SLES 9),
2997 2.6.9-55.EL (RHEL 4),
2998 2.6.16.46-0.14 (SLES 10),
2999 2.6.18.8 vanilla (kernel.org)
3000 * Client support for unpatched kernels:
3001 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3002 2.6.16 - 2.6.22 vanilla (kernel.org)
3003 * Due to recently discovered recovery problems, we do not recommend
3004 using patchless RHEL 4 clients with this or any earlier release.
3005 * Recommended e2fsprogs version: 1.39.cfs8
3006 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3007 * Starting with this release, the ldiskfs backing filesystem required
3008 by Lustre is now in its own package, lustre-ldiskfs. This package
3009 should be installed. It is versioned separately from Lustre and
3010 may be released separately in future.
3014 Description: block reactivating mgc import until all deactivates complete
3015 Details : Fix race when failing back MDT/MGS to itself (testing)
3018 Frequency : at statup only
3020 Description: mds_lov_synchronize race leads to various problems
3021 Details : simultaneous MDT->OST connections at startup can cause the
3022 sync to abort, leaving the OSC in a bad state.
3024 Severity : enhancement
3026 Description: add optional extra BUILD_VERSION info
3027 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3028 to override the lustre version.
3031 Frequency : 2.6.18 servers only
3033 Description: ll_kern_mount() doesn't release the module reference
3034 Details : The ldiskfs module reference count never drops down to 0
3035 because ll_kern_mount() doesn't release the module reference.
3040 Description: server LBUG when using old ost_num_threads parameter
3041 Details : Accept the old ost_num_threads parameter but warn that it
3042 is deprecated, and fix an off-by-one error that caused an LBUG.
3047 Description: Transient SCSI error results in persistent IO issue
3048 Details : iobuf->dr_error is not reinitialized to 0 between two
3052 Frequency : sometimes when underlying device returns I/O errors
3054 Description: OSTs not going read-only during write failures
3055 Details : OSTs are not remounted read-only when the journal commit threads
3056 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3057 instead of the ext3 wrappers.
3061 Description: poor connect scaling with increasing client count
3062 Details : Don't run filter_grant_sanity_check for more than 100 exports
3063 to improve scaling for large numbers of clients.
3066 Frequency : SLES10 only
3068 Description: sanity-quota.sh quotacheck failed: rc = -22
3069 Details : Quotas cannot be enabled on SLES10.
3072 Frequency : liblustre clients only
3074 Description: getdirentries does not give error when run on compute nodes
3075 Details : getdirentries does not fail when the size specified as an argument
3076 is too small to contain at least one entry
3078 Severity : enhancement
3080 Description: Add LNET router traceability for debug purposes
3081 Details : If a checksum failure occurs with a router as part of the
3082 IO path, the NID of the last router that forwarded the bulk data
3083 is printed so it can be identified.
3088 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3089 Details : Due to a race condition, liblustre clients were occasionally
3090 evicted incorrectly.
3092 Severity : enhancement
3094 Description: lfs setstripe use optional parameters instead of postional
3097 Severity : enhancement
3099 Description: Nanosecond timestamp support for ldiskfs
3100 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3101 resolution timestamps. There is not yet support for this at
3102 the Lustre filesystem level.
3105 Frequency : during server recovery
3107 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3108 Details : request with rq_no_resend flag not awake l_wait_event if they get
3112 Frequency : nfs export on patchless client
3114 Description: connectathon hang when test nfs export over patchless client
3115 Details : Disconnected dentry cannot be found with lookup, so we do not need
3116 to unhash it or make it invalid
3119 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3120 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3121 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3126 Description: open req refcounting wrong on reconnect
3127 Details : If reconnect happened between getting open reply from server and
3128 call to mdc_set_replay_data in ll_file_open, we will schedule
3129 replay for unreferenced request that we are about to free.
3130 Subsequent close will crash in variety of ways.
3131 Check that request is still eligible for replay in
3132 mdc_set_replay_data().
3137 Description: disable writes to filesystem when reading health_check file
3138 Details : the default for reading the health_check proc file has changed
3139 to NOT do a journal transaction and write to disk, because this
3140 can cause reads of the /proc file to hang and block HA state
3141 checking on a healthy but otherwise heavily loaded system. It
3142 is possible to return to the previous behaviour during configure
3143 with --enable-health-write.
3145 Severity : enhancement
3147 Description: 64-bit inode version
3148 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3149 to the inode. This will be required for version-based recovery.
3154 Description: MDS fails to start if a duplicate client export is detected
3155 Details : in some rare cases it was possible for a client to connect to
3156 an MDS multiple times. Upon recovery the MDS would detect this
3157 and fail during startup. Handle this more gracefully.
3159 Severity : enhancement
3161 Description: Add -o localflock option to simulate old noflock
3163 Details : This will achieve local-only flock/fcntl locks
3169 Description: log_commit_thread vs filter_destroy race leads to crash
3170 Details : Take import reference before releasing llog record semaphore
3175 Description: Wrong request locking in request set processing
3176 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3177 imp_delayed_list, in this place should be used imp_lock.
3180 Frequency : when reconnection
3182 Description: Grant Leak when osc reconnect to OST
3183 Details : When osc reconnect ost, OST(filter) should check whether it
3184 should grant more space to client by comparing fed_grant and
3185 cl_avail_grant, and return the granted space to client instead
3186 of "new granted" space, because client will call osc_init_grant
3187 to update the client grant space info.
3190 Frequency : when client reconnect to OST
3192 Description: Grant Leak when osc do resend and replay bulk write
3193 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3194 bulk write request, because the grant info while be sync between
3195 OSC and OST when reconnect, and we should ignore the grant info
3196 these of resend/replay write req.
3201 Description: Grant space more than available left space sometimes.
3202 Details : When then OST is about to be full, if two bulk writing from
3203 different clients came to OST. Accord the avaliable space of the
3204 OST, the first req should be permitted, and the second one
3205 should be denied by ENOSPC. But if the seconde arrived before
3206 the first one is committed. The OST might wrongly permit second
3207 writing, which will cause grant space > available space.
3210 Frequency : when client is evicted
3212 Description: Grant might be wrongly erased when osc is evicted by OST
3213 Details : when the import is evicted by server, it will fork another
3214 thread ptlrpc_invalidate_import_thread to invalidate the
3215 import, where the grant will be set to 0. While the original
3216 thread will update the grant it got when connecting. So if
3217 the former happened latter, the grant will be wrongly errased
3218 because of this race.
3223 Description: Checking Stale with correct fid
3224 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3225 to check whether it is stale, because sometimes, we want the
3226 enqueue happened anyway, and op_data.fid2 will not be initialized.
3228 Severity : enhancement
3230 Description: update patchless client
3231 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3234 Frequency : only with 2.4 kernel
3236 Description: random memory corruption
3237 Details : size of struct ll_inode_info is to big for union inode.u and this
3238 can be cause of random memory corruption.
3243 Description: Memory leak in recovery
3244 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3245 It should also check obd_fail before fsfilt_start, otherwise if
3246 fsfilt_start return -EROFS,(failover mds during mds recovery).
3247 then the req will return with repmsg->transno = 0 and rc = EROFS.
3248 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3249 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3250 be freed no matter whether fsfilt_commit success or not.
3253 Frequency : only with huge count clients
3255 Description: Prevents from taking the superblock lock in llap_from_page for
3257 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3258 for llap_from_page call in ll_removepage prevents from taking the
3259 superblock lock for a soon died page.
3264 Description: Not check open intent error before release open handle
3265 Details : in some rare cases, the open intent error is not checked before
3266 release open handle, which may cause
3267 ASSERTION(open_req->rq_transno != 0), because it tries to release
3268 the failed open handle.
3273 Description: Set cat log bitmap only after create log success.
3274 Details : in some rare cases, the cat log bitmap is set too early. and it
3275 should be set only after create log success.
3279 Description: Accessing a block bevice can re-enable I/O when Lustre is
3280 tearing down a device.
3281 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3286 Description: service threads may hog cpus when there are a lot of requests
3287 Details : Insert cond_resched to give other threads a chance to use some CPU
3292 Description: the cat log was not initialized in recovery
3293 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3294 unlink log on mds will not be initialized until mds post
3295 recovery. And also in mds post recovery, the unlink log will
3296 initialization will be done asynchronausly, so there will be race
3297 between add unlink log and unlink log initialization.
3301 Description: brw_stats were being printed incorrectly
3302 Details : brw_stats were being printed as log2 but all of them were not
3303 recorded as log2. Also remove some code duplication arising from
3304 filter_tally_{read,write}.
3308 Frequency : rare, only in recovery.
3309 Description: ASSERTION(req->rq_type != LI_POISON) failed
3310 Details : imp_lock should be held while iterating over imp_sending_list for
3311 prevent destroy request after get timeout in ptlrpc_queue_wait.
3315 Description: replay-single.sh test 52 fails
3316 Details : A lock's skiplist need to be cleanup when it being unlinked
3317 from its resource list.
3321 Description: Short directio read returns full requested size rather than
3323 Details : Direct I/O operations should return actual amount of bytes
3324 transferred rather than requested size.
3326 Severity : enhancement
3328 Description: metadata RPC reduction (e.g. for rm performance)
3329 Details : decrease the amount of synchronous RPC between clients and servers
3330 by canceling conflicing lock before the operation on the client
3331 and packing thier handles into the main operation RPC to server.
3333 Severity : enhancement
3335 Description: Async OSC create to avoid the blocking unnecessarily.
3336 Details : If an OST has no remaining object, system will block on the
3337 creation when it needs to create a new object on this OST. Now,
3338 ways use pre-created objects when available, instead of blocking on
3339 an empty osc while others are not empty. If we must block, we
3340 block for the shortest possible period of time.
3344 Description: improve handling recoverable errors
3345 Details : if request processig with error which can be recoverable on server
3346 request should be resend, otherwise page released from cache and
3349 Severity : enhancement
3351 Description: refine locking for avoid write wrong info into lov_objid file
3352 Details : fix possible races with add new target and write/update data in
3355 --------------------------------------------------------------------------------
3357 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3362 Frequency : on some architectures
3364 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3365 Details : Uninitialized flags sometimes cause configuration commands to
3369 Frequency : patchless clients only
3371 Description: missing __iget() symbol export
3372 Details : The __iget() symbol export is missing. To avoid the need for
3373 this on patchless clients the deathrow inode reaper is turned
3374 off, and we depend on the VM to clean up old inodes. This
3375 dependency was during via the fix for bug 12181.
3377 --------------------------------------------------------------------------------
3379 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3381 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3382 INTEROPERATE with older versions automatically. In many cases a
3383 special upgrade step is needed. Please read the
3384 user documentation before upgrading any part of a live system.
3385 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3386 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3387 * WARNING: Lustre configuration and startup changes are required with
3388 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3392 Severity : enhancement
3394 Description: MountConf
3395 Details : Lustre configuration is now managed via mkfs and mount
3396 commands instead of lmc and lconf. New obd types (MGS, MGC)
3397 are added for dynamic configuration management. See
3398 https://mail.clusterfs.com/wikis/lustre/MountConf for
3401 Severity : enhancement
3403 Description: dynamic OST addition
3404 Details : OSTs can now be added to a live filesystem
3406 Severity : enhancement
3408 Description: startup order invariance
3409 Details : MDTs and OSTs can be started in any order. Clients only
3410 require the MDT to complete startup.
3412 Severity : enhancement
3414 Description: parallel, asynchronous orphan cleanup
3415 Details : orphan cleanup is now performed in separate threads for each
3416 OST, allowing parallel non-blocking operation.
3418 Severity : enhancement
3420 Description: optimized stripe assignment
3421 Details : stripe assignments are now made based on ost space available,
3422 ost previous usage, and OSS previous usage, in order to try
3423 to optimize storage space and networking resources.
3425 Severity : enhancement
3427 Description: Permanently set tunables
3428 Details : All writable /proc/fs/lustre tunables can now be permanently
3429 set on a per-server basis, at mkfs time or on a live system.
3431 Severity : enhancement
3433 Description: Lustre message v2
3434 Details : Add lustre message format v2.
3436 Severity : enhancement
3438 Description: client OST exclusion list
3439 Details : Clients can be started with a list of OSTs that should be
3440 declared "inactive" for known non-responsive OSTs.
3444 Description: SPEC SFS validation failure on NFS v2 over lustre.
3445 Details : Changes the blocksize for regular files to be 2x RPC size,
3446 and not depend on stripe size.
3448 Severity : enhancement
3450 Description: Multiple MD RPCs in flight.
3451 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3452 To avoid overly-overloading MDS, introduce a limit on number of
3453 MDS RPCs in flight for a single client and add /proc controls
3454 to adjust this limit.
3456 Severity : enhancement
3458 Description: client read/write statistics
3459 Details : Add client read/write call usage stats for performance
3460 analysis of user processes.
3461 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3462 file access. extents_stats shows chunk size distribution.
3463 extents_stats_per_process show chunk size distribution per
3466 Severity : enhancement
3468 Description: mds statistics
3469 Details : Add detailed mds operations statistics in
3470 /proc/fs/lustre/mds/*/stats.
3474 Description: Failure of copying files with lustre special EAs.
3475 Details : Client side always return success for setxattr call for lustre
3476 special xattr (currently only "trusted.lov").
3481 Description: Refcount LNET uuids
3482 Details : The global LNET uuid list grew linearly with every startup;
3483 refcount repeated list entries instead of always adding to
3486 Severity : enhancement
3488 Description: Dynamic service threads
3489 Details : Within a small range, start extra service threads
3490 automatically when the request queue builds up.
3493 Frequency : mixed-endian client/server environments
3495 Description: mixed-endian crashes
3496 Details : The new msg_v2 system had some failures in mixed-endian
3499 Severity : enhancement
3501 Description: Easy OST removal
3502 Details : OSTs can be permanently deactivated with e.g. 'lctl
3503 conf_param lustre-OST0001.osc.active=0'
3505 Severity : enhancement
3507 Description: MGS proc entries
3508 Details : Added basic proc entries for the MGS showing what filesystems
3511 Severity : enhancement
3513 Description: provide MGS failover
3514 Details : Added config lock reacquisition after MGS server failover.
3516 Severity : enhancement
3518 Description: add Linux 2.4 support
3519 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3523 Description: a large application tries to do I/O to the same resource and dies
3524 in the middle of it.
3525 Details : Check the req->rq_arrival time after the call to
3526 ost_brw_lock_get(), but before we do anything about
3527 processing it & sending the BULK transfer request. This
3528 should help move old stale pending locks off the queue as
3529 quickly as obd_timeout.
3532 Frequency : when an incorrect nid is specified during startup
3534 Description: ptlrpc connect to non-existant node causes kernel crash
3535 Details : LNET can't be re-entered from an event callback, which
3536 happened when we expire a message after the export has been
3537 cleaned up. Instead, hand the zombie cleanup off to another
3540 Severity : enhancement
3542 Description: plain/inodebits lock performance improvement
3543 Details : Grouping plain/inodebits in granted list by their request modes
3544 and bits policy, thus improving the performance of search through
3548 Frequency : only if OST filesystem is corrupted
3550 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3551 Details : for a short time RPCs with bulk IO are in the replay list,
3552 but replay of bulk IOs is unimplemented. If the OST filesystem
3553 is corrupted due to disk cache incoherency and then replay is
3554 started it is possible to trip an assertion. Avoid putting
3555 committed RPCs into the replay list at all to avoid this issue.
3558 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3560 Description: System hang on startup
3561 Details : This bug allowed the liblustre (e.g. catamount) client to
3562 return to the app before handling all startup RPCs. This
3563 could leave the node unresponsive to lustre network traffic
3564 and manifested as a server ptllnd timeout.
3566 Severity : enhancement
3568 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3569 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3570 boolean to control whether to print peer debug info when a
3571 client's RPC times out.
3574 Frequency : only for kernels with patches from Lustre below 1.4.3
3576 Description: Remove old rdonly API
3577 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3580 Frequency : only for devices with external journals
3582 Description: Set external device read-only also
3583 Details : During a commanded failover stop, we set the disk device
3584 read-only while the server shuts down. We now also set any
3585 external journal device read-only at the same time.
3588 Frequency : when upgrading from 1.4 while trying to change parameters
3590 Description: The wrong (new) MDC name was used when setting parameters for
3591 upgraded MDT's. Also allows changing of OSC (and MDC)
3592 parameters if --writeconf is specified at tunefs upgrade time.
3595 Frequency : when setting specific ost indicies
3597 Description: QOS code breaks on skipped indicies
3598 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3599 created with --index need not be sequential.
3601 Severity : enhancement
3603 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3604 Details : The uninit_groups feature works in conjunction with the kernel
3605 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3606 up the pass1 processing of e2fsck. This is a read-only feature
3607 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3608 mount filesystems that have had this feature enabled.
3610 Severity : enhancement
3612 Description: Improve multi-block allocation algorithm to avoid fragmentation
3613 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3614 allocation locality and avoid filesystem fragmentation.
3616 ------------------------------------------------------------------------------
3618 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3620 * Support for kernels:
3621 2.4.21-47.0.1.EL (RHEL 3)
3622 2.6.5-7.283 (SLES 9)
3623 2.6.9-42.0.10.EL (RHEL 4)
3624 2.6.12.6 vanilla (kernel.org)
3625 2.6.16.27-0.9 (SLES 10)
3626 * Recommended e2fsprogs version: 1.39.cfs5
3628 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3632 Frequency : occasional, depends on client load and configuration
3633 Bugzilla : 12181, 12203
3634 Description: data loss for recently-modified files
3636 Details : In some cases it is possible that recently written or created
3637 files may not be written to disk in a timely manner (this should
3638 normally be within 30s unless client IO load is very high).
3639 The problem appears as zero-length files or files that are a
3640 multiple of 1MB in size after a client crash or client eviction
3641 that are missing data at the end of the file.
3643 This problem is more likely to be hit on clients where files are
3644 repeatedly created and unlinked in the same directory, clients
3645 have a large amount of RAM, have many CPUs, the filesystem has
3646 many OSTs, the clients are rebooted frequently, and/or the files
3647 are not accessed by other nodes after being written.
3649 The presence of the problem can be detected by looking at
3650 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3651 smaller than the second (nr_unused) then dirty files will not
3652 be flushed automatically to disk. "sync; sleep 10" should be
3653 run several times on the node before unmounting it to update
3654 Lustre (this is also safe to run on nodes without this problem).
3656 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3657 that can cause this same problem, so customers using that kernel
3658 also need to update the kernel in addition to Lustre. In order
3659 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3661 It is normal that files written just before a client crash (less
3662 than 30s) may not yet have been flushed to disk, even for local
3666 Frequency : frequent on thin XT3 nodes
3668 Description: UUID collision on thin XT3 Linux nodes
3669 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3670 randomly, since we relied on an insufficiently-seeded PRNG.
3675 Description: OSS hangs after "All ost request buffers busy"
3676 Details : A deadlock between quota and journal operations caused OSS
3677 hangs after printing "All ost request buffers busy."
3680 Frequency : always on liblustre builds
3682 Description: Cleanup compiler warnings on liblustre
3685 Frequency : always on liblustre builds on XT3
3687 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3691 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3692 > 32000 subdirectories
3693 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3694 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3695 subdirectory count crosses 32000. This will aid e2fsck to
3696 correctly handle more than 32000 subdirectories.
3699 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3701 Description: System hang on startup
3702 Details : This bug allowed the liblustre (e.g. catamount) client to
3703 return to the app before handling all startup RPCs. This
3704 could leave the node unresponsive to lustre network traffic
3705 and manifested as a server ptllnd timeout.
3707 Severity : enhancement
3709 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3710 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3711 boolean to control whether to print peer debug info when a
3712 client's RPC times out.
3717 Description: make O_SYNC working on 2.6 kernels
3718 Details : 2.6 kernels use different method for mark pages for write,
3719 so need add a code to lustre for O_SYNC work.
3724 Description: Failure to close file and release space on NFS
3725 Details : Put inode details into lock acquired in ll_intent_file_open.
3726 Use mdc_intent_lock in ll_intent_open to properly
3727 detect all kind of errors unhandled by mdc_enqueue.
3732 Description: proc file read during shutdown sometimes raced obd removal,
3734 Details : Add lock to prevent obd access after proc file removal.
3737 Frequency : Only for files larger than 4GB on 32-bit clients.
3739 Description: improperly doing page alignment of locks
3740 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3741 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3746 Description: RPCs being resent when they shouldn't be
3747 Details : Some RPCs that should not be resent are being resent. This
3748 can cause inconsistencies in the RPC state machine. Do not
3749 resend such requests.
3752 Frequency : rare, only with NFS export
3754 Description: Crash on NFS re-export node
3755 Details : under very unusual load conditions an assertion is hit in
3756 ll_intent_file_open()
3759 Frequency : only if OST filesystem is corrupted
3761 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3762 Details : for a short time RPCs with bulk IO are in the replay list,
3763 but replay of bulk IOs is unimplemented. If the OST filesystem
3764 is corrupted due to disk cache incoherency and then replay is
3765 started it is possible to trip an assertion. Avoid putting
3766 committed RPCs into the replay list at all to avoid this issue.
3771 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3772 Details : Large single O_DIRECT read and write calls can fail to allocate
3773 a sufficiently large buffer to process the request. In case of
3774 allocation failure the allocation is retried with a smaller
3775 buffer and broken into smaller requests.
3777 Severity : enhancement
3779 Description: Add -o localflock option to simulate old noflock behaviour.
3780 Details : This will achieve local-only flock/fcntl locks coherentness.
3785 Description: versioning check is incomplete
3786 Details : Checking the version difference of client vs. server, report
3787 error if the gap is too big.
3792 Description: add support PG_writeback bit
3793 Details : add support for PG_writeback bit for Lustre, for more carefull
3794 work with page cache in 2.6 kernel. This also fix some deadlocks
3795 and remove hack for work O_SYNC with 2.6 kernel.
3797 Severity : enhancement
3799 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3800 Details : The uninit_groups feature works in conjunction with the kernel
3801 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3802 up the pass1 processing of e2fsck. This is a read-only feature
3803 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3804 mount filesystems that have had this feature enabled.
3806 Severity : enhancement
3808 Description: Improve multi-block allocation algorithm to avoid fragmentation
3809 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3810 allocation locality and avoid filesystem fragmentation.
3812 ------------------------------------------------------------------------------
3814 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3816 * Support for kernels:
3817 2.6.9-42.0.3.EL (RHEL 4)
3818 2.6.5-7.276 (SLES 9)
3819 2.4.21-47.0.1.EL (RHEL 3)
3820 2.6.12.6 vanilla (kernel.org)
3821 2.6.16.21-0.8 (SLES10)
3822 * Recommended e2fsprogs version: 1.39.cfs2-0
3824 * The backwards-compatible /proc/sys/portals symlink has been removed
3825 in this release. Before upgrading, please ensure that you change
3826 any configuration scripts or /etc/sysctl.conf files that access
3827 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3828 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3829 in advance of the upgrade on any system running Lustre 1.4.6 or
3830 newer, since /proc/sys/lnet was added in that version.
3831 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3835 Frequency : only when quota is used
3837 Description: avoid scanning export list for quota master
3838 Details : Change the algorithms to avoid scanning export list in order
3839 to improve the efficiency.
3842 Frequency : MDS failover only, very rarely
3844 Description: "went back in time" messages on mds failover
3845 Details : The greatest transno may be lost when the current operation
3846 finishes with an error (transno==0) and the client's last_rcvd
3847 record is over-written. Save the greatest transno in the
3848 mds_last_transno for this case.
3851 Frequency : always for specific kernels and striping counts
3853 Description: client may get "Matching packet too big" without ACL support
3854 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3855 when trying to access files in certain configurations. The
3856 clients should in fact be denied when mounting because they do
3857 not understand ACLs.
3860 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3862 Description: many clients connecting with IO in progress causes connect timeouts
3863 Details : Avoid synchronous journal commits to avoid delays caused by many
3864 clients connecting/disconnecting when bulk IO is in progress.
3865 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3866 OST_IO_PORTAL to avoid delays behind potentially many pending
3870 Frequency : occasionally with multiple writers to a single file
3872 Description: shared writes to file may result in wrong size reported by stat()
3873 Details : Allow growing of kms when extent lock is cancelled
3876 Frequency : always with random mmap IO to multi-striped file
3878 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3879 Details : Only if the hole is at the end of OST object so that kms is too
3880 small. Fix is to increase kms accordingly in ll_nopage.
3883 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3885 Description: writes to a missing object would leak memory on the OST
3886 Details : If there is an inconsistency between the MDS and OST filesystems,
3887 such that the MDS references an object that doesn't exist, writes
3888 to that object will leak memory due to incorrect cleanup in the
3889 error handling path, eventually running out of memory on the OST.
3894 Description: Creating too long symlink causes lustre errors
3895 Details : Check symlink and name lengths before sending requests to MDS.
3898 Frequency : only if flock is enabled (not on by default)
3900 Description: posix locks not released on fd closure on 2.6.9+
3901 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3902 kernels, this caused such locks not to be released on fd close and
3903 then assertions on fs unmount about still used locks.
3906 Frequency : MDS failover only, very rarely
3908 Description: clients may get ASSERTION(granted_lock != NULL)
3909 Details : When request was taking a long time, and a client was resending
3910 a getattr by name lock request. The were multiple lock requests
3911 with the same client lock handle and
3912 mds_getattr_name->fixup_handle_for_resent_request found one of the
3913 lock handles but later failed with ASSERTION(granted_lock != NULL).
3918 Description: handle->h_buffer_credits > 0, assertion failure
3919 Details : h_buffer_credits is zero after truncate, causing assertion
3920 failure. This patch extends the transaction or creates a new
3924 Frequency : NFS re-export or patchless client
3925 Bugzilla : 11179, 10796
3926 Description: Crash on NFS re-export node (__d_move)
3927 Details : We do not want to hash the dentry if we don't have a lock.
3928 But if this dentry is later used in d_move, we'd hit uninitialised
3929 list head d_hash, so we just do this to init d_hash field but
3930 leave dentry unhashed.
3933 Frequency : NFS re-export or patchless client
3935 Description: NFS exports has problem with symbolic link
3936 Details : lustre client didn't properly install dentry when re-exported
3937 to NFS or running patchless client.
3940 Frequency : NFS re-export or patchless client
3942 Description: Various nfs/patchless fixes.
3943 Details : fixes reuse disconected alias for lookup process - this fixes
3944 warning "find_exported_dentry: npd != pd",
3945 fix permission error with open files at nfs.
3946 fix apply umask when do revalidate.
3949 Frequency : occasional
3951 Description: Crash on NFS re-export node
3952 Details : calling clear_page() on the wrong pointer triggered oops in
3953 generic_mapping_read().
3956 Frequency : rarely, using O_DIRECT IO
3958 Description: unaligned directio crashes client with LASSERT
3959 Details : check for unaligned buffers before trying any requests.
3962 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3964 Description: stale data returned from RAID cache
3965 Details : If only a small amount of IO is done to the RAID device before
3966 reading it again it is possible to get stale data from the RAID
3967 cache instead of reading it from disk.
3970 Frequency : always for sles10 kernel
3972 Description: sles10 support
3973 Details : ll_follow_link: compile fixes and using of nd_set_link
3974 under newer kernels.
3977 Frequency : depends on arch, kernel and compiler version, always on sles10
3980 Description: recursive or deep enough symlinks cause stack overflow
3981 Details : getting rid of large stack-allocated variable in
3985 Frequency : depends on hardware
3987 Description: lustre write performance loss in the SLES10 kernel
3988 Details : the performance loss is caused by using of write barriers in the
3989 ext3 code. The SLES10 kernel turns barrier support on by
3990 default. The fix is to undo that change for ldiskfs.
3992 ------------------------------------------------------------------------------
3994 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3996 * Support for kernels:
3997 2.6.9-42.0.3EL (RHEL 4)
3998 2.6.5-7.276 (SLES 9)
3999 2.4.21-47.0.1.EL (RHEL 3)
4000 2.6.12.6 vanilla (kernel.org)
4004 Frequency : quota enabled and large files being deleted
4006 Description: releasing more than 4GB of quota at once hangs OST
4007 Details : If a user deletes more than 4GB of files on a single OST it
4008 will cause the OST to spin in an infinite loop. Release
4009 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4014 Description: statfs data retrieved from /proc may be stale or zero
4015 Details : When reading per-device statfs data from /proc, in the
4016 {kbytes,files}_{total,free,avail} files, it may appear
4017 as zero or be out of date.
4020 Frequency : systems with MD RAID1 external journal devices
4022 Description: lconf's call to blkid is confused by RAID1 journal devices
4023 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4024 of returning all block devices that match the journal UUID.
4027 Frequency : always, for aggregate stripe size over 4GB
4029 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4030 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4031 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4034 Frequency : always if "lfs find" used on a local file/directory
4036 Description: "lfs find" segfaults if used on a local file/directory
4037 Details : The case where a directory component was not specified wasn't
4038 handled correctly. Handle this properly.
4041 Frequency : always on ppc64
4043 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4044 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4045 perform endianness conversion.
4048 Frequency : rarely (truncate to non-zero file size after write under load)
4049 Bugzilla : 10730, 10687
4050 Description: Files padded with zeros to next 4K multiple
4051 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4052 it is possible that files that are truncated to a non-zero size
4053 immediately after being written are filled with zero bytes beyond
4054 the truncated size. No file data is lost.
4056 Severity : enhancement
4058 Description: Allow recovery/failover for liblustre clients.
4059 Details : liblustre clients were unaware of failover configurations until
4062 Severity : enhancement
4064 Description: user file locks should fail when not mounting with flock option
4065 Details : Set up an error-returning stub in ll_file_operations.lock field
4066 to prevent incorrect behaviour when client is mounted without
4067 flock option. Also, set up properly f_op->flock field for
4071 Frequency : always on ia64
4073 Description: "lfs df" loops on printing out MDS statfs information
4074 Details : The obd_ioctl_data was not initialized and in some systems
4075 this caused a failure during the ioctl that did not return
4076 an error. Initialize the struct and return an error on failure.
4079 Frequency : SLES 9 only
4081 Description: Error of copying files with lustre special EAs as root
4082 Details : Client side always return success for setxattr call for lustre
4083 special xattr (currently only "trusted.lov").
4086 Frequency : rarely on clusters with both ia64+i386 clients
4088 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4089 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4090 clients are doing concurrent writes to the same file it is
4091 possible that the ia64 clients may LASSERT because the OST
4092 extent locks are not PAGE_SIZE aligned. Ensure that grown
4093 locks are always aligned on the request boundary.
4096 Frequency : specific use, occasional
4098 Description: Overwriting in use executable truncates on-disk binary image
4099 Details : If one node attempts to overwrite an executable in use by
4100 another node, we now correctly return ETXTBSY instead of
4101 truncating the file.
4103 Severity : enhancement
4105 Description: Async OSC create to avoid the blocking unnecessarily.
4106 Details : If an OST has no remaining object, system will block on the creating
4107 when need to create a new object on this OST. Now, ways use
4108 pre-created objects when available, instead of blocking on an
4109 empty osc while others are not empty. If we must block, we block
4110 for the shortest possible period of time.
4115 Description: chmod on Lustre root is propagated to other clients
4116 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4117 invalid by the follow_mount time.
4120 Frequency : liblustre clients only
4122 Description: Race in 'instant cancel' lock handling could lead to such locks
4123 never to be granted in case of SMP MDS
4124 Details : Do not destroy not yet granted but cbpending locks in
4128 Frequency : replay/resend of open
4130 Description: non null lock assetion failure in mds_intent_policy
4131 Details : Trying to replay/resend lockless open requests resulted in
4132 mds_open() returning 0 with no lock. Now it sets a flag if
4133 a lock is going to be returned.
4135 Severity : enhancement
4137 Description: Checksum enhancements
4138 Details : New checksum enhancements allow for resending RPCs that failed
4141 Severity : enhancement
4143 Description: Tunables on number of dirty pages in cacche
4144 Details : Allow to set limit on number of dirty pages cached.
4149 Description: client crash on unmount - lock still has references
4150 Details : In some error handling cases it was possible to leak a lock
4151 reference on a client while accessing a file. This was not
4152 harmful to the client during operation, but would cause the
4153 client to crash when the filesystem is unmounted.
4156 Frequency : specific case, rare
4158 Description: ETXTBSY on mds though file not in use by client
4159 Details : ETXTBSY is no longer incorrectly returned when attempting to
4160 chmod or chown a directory that the user previously tried to
4161 execute or a currently-executing binary.
4164 Frequency : extremely rare except on liblustre-based clients
4166 Description: Lustre space not freed when files are deleted
4167 Details : Clean up open-unlinked files after client eviction. Previously
4168 the unlink was skipped and the files remained as orphans.
4173 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4174 Details : In some cases it was possible to send a blocking callback to
4175 a client doing a glimpse, even though that client didn't get
4176 a lock granted. When the glimpse lock is cancelled on the OST
4177 the freed lock is left on the waiting list and corrupted the list.
4180 Frequency : all core dumps
4182 Description: Broke core dumps to lustre
4183 Details : Negative dentry may be unhashed if parent does not have UPDATE
4184 lock, but some callers, e.g. do_coredump, expect dentry to be
4185 hashed after successful create, hash it in ll_create_it.
4187 ------------------------------------------------------------------------------
4189 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4191 * Support for kernels:
4192 2.6.9-42.0.2.EL (RHEL 4)
4193 2.6.5-7.276 (SLES 9)
4194 2.4.21-40.EL (RHEL 3)
4195 2.6.12.6 vanilla (kernel.org)
4199 Frequency : always on RHEL 3
4201 Description: Number of open files grows over time
4202 Details : The number of open files grows over time, whether or not
4203 Lustre is started. This was due to a filp leak introduced
4204 by one of our kernel patches.
4206 ------------------------------------------------------------------------------
4208 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4210 * Support for kernels:
4211 2.6.9-42.EL (RHEL 4)
4212 2.6.5-7.276 (SLES 9)
4213 2.4.21-40.EL (RHEL 3)
4214 2.6.12.6 vanilla (kernel.org)
4219 Bugzilla : 5719, 9635, 9792, 9684
4220 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4221 Details : If a server is under heavy load and cannot reply to new
4222 connection requests before the client resends the (re)connect,
4223 the connection handling code can behave badly if two service
4224 threads are concurrently handing separate (re)connections from
4225 the same client. Add better locking to the connection handling
4226 code, and ensure that only a single connection will be processed
4227 for a given client UUID, even if the lock is dropped.
4229 Severity : enhancement
4231 Description: add TCP zero-copy support to kernel
4232 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4233 sends if the hardware supports scatter-gather and checksumming.
4234 This allows socklnd to do client-write and server-read more
4235 efficiently and reduce CPU utilization from skbuf copying.
4238 Frequency : only if NFS exporting from client
4240 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4241 Details : If setting attributes on a file created under NFS that had
4242 never been opened it would be possible to oops the client
4243 if the file had no objects.
4246 Frequency : always for liblustre
4248 Description: liblustre client does MDS+OSTs setattr RPC for each write
4249 Details : When doing a write from a liblustre client, the client
4250 incorrectly issued an RPC to the MDS and each OST the file was
4251 striped over in order to update the timestamps. When writing
4252 with small chunks and many clients this could overwhelm the MDS
4253 with RPCs. In all cases it would slow down the write because
4254 these RPCs are unnecessary.
4256 Severity : enhancement
4258 Description: allow number of MDS service threads to be changed at module load
4259 Details : It is now possible to change the number of MDS service threads
4260 running. Adding "options mds mds_num_threads={N}" to the MDS's
4261 /etc/modprobe.conf will set the number of threads for the next
4262 time Lustre is restarted (assuming the "mds" module is also
4263 reloaded at that time). The default number of threads will
4264 stay the same, 32 for most systems.
4269 Description: OST crash if filesystem is unformatted or corrupt
4270 Details : If an OST is started on a device that has never been formatted
4271 or if the filesystem is corrupt and cannot even mount then the
4272 error handling cleanup routines would dereference a NULL pointer.
4277 Description: NULL pointer deref in llap_from_page.
4278 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4279 incorrect) mapping if the page was truncated/reclaimed while it was
4280 searched for. Check for this condition and skip such pages when
4281 doing readahead. Introduce extra check to llap_from_page() to
4282 verify page->mapping->host is non-NULL (so page is not anonymous).
4285 Frequency : Sometimes when using sys_sendfile
4287 Description: "page not covered by a lock" warnings from ll_readpage
4288 Details : sendfile called ll_readpage without right page locks present.
4289 Now we introduced ll_file_sendfile that does necessary locking
4290 around call to generic_file_sendfile() much like we do in
4294 Frequency : with certain MDS communication failures at client mount time
4296 Description: NULL pointer deref after failed client mount
4297 Details : a client connection request may delayed by the network layer
4298 and not be sent until after the PTLRPC layer has timed out the
4299 request. If the client fails the mount immediately it will try
4300 to clean up before the network times out the request. Add a
4301 reference from the request import to the obd device and delay
4302 the cleanup until the network drops the request.
4305 Frequency : occasionally during client (re)connect
4307 Description: assertion failure during client (re)connect
4308 Details : processing a client connection request may be delayed by the
4309 client or server longer than the client connect timeout. This
4310 causes the client to resend the connection request. If the
4311 original connection request is replied in this interval, the
4312 client may trip an assertion failure in ptlrpc_connect_interpret()
4313 which thought it would be the only running connect process.
4316 Frequency : only with obd_echo servers and clients that are rebooted
4318 Description: kernel BUG accessing uninitialized data structure
4319 Details : When running an obd_echo server it did not start the ping_evictor
4320 thread, and when a client was evicted an uninitialized data
4321 structure was accessed. Start the ping_evictor in the RPC
4322 service startup instead of the OBD startup.
4324 Severity : enhancement
4325 Bugzilla : 10193 (patchless)
4326 Description: Remove dependency on various unexported kernel interfaces.
4327 Details : No longer need reparent_to_init, exit_mm, exit_files,
4328 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4331 Frequency : rare (only users of deprecated and unsupported LDAP config)
4333 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4334 Details : LDAP apparently contains 'lustreName' attributes instead of
4335 'name'. A simple remapping of the name is sufficient.
4338 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4340 Description: waiting_locks_callback trips kernel BUG if client is evicted
4341 Details : Running with the dump_on_timeout debug flag turned on makes
4342 it possible that the waiting_locks_callback() can try to dump
4343 the Lustre kernel debug logs from an interrupt handler. Defer
4344 this log dumping to the expired_lock_main() thread.
4346 Severity : enhancement
4348 Description: Support NFS exporting on 2.6 kernels.
4349 Details : Implement non-rawops metadata methods for NFS server to use without
4350 changing NFS server code.
4353 Frequency : very rare (synthetic metadata workload only)
4355 Description: two racing renames might cause an MDS thread to deadlock
4356 Details : Running the "racer" program may cause one MDS thread to rename
4357 a file from being the source of a rename to being the target of
4358 a rename at exactly the same time that another thread is doing
4359 so, and the second thread has already enqueued these locks after
4360 doing a lookup of the target and is trying to relock them in
4361 order. Ensure that we don't try to re-lock the same resource.
4364 Frequency : only very large systems with liblustre clients
4366 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4367 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4368 all OSTs in parallel. This allows the eviction of stale liblustre
4369 clients to proceed much faster than if they were done in series,
4370 and also offers similar improvements for other set_info RPCs.
4375 Description: excessive CPU usage during initial read phase on client
4376 Details : During the initial read phase on a client, it would agressively
4377 retry readahead on the file, consuming too much CPU and impacting
4378 performance (since 1.4.5.8). Improve the readahead algorithm
4379 to avoid this, and also improve some other common cases (read
4380 of small files in particular, where "small" is files smaller than
4381 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4386 Description: MDS crash when receiving packet with unknown intent.
4387 Details : Do not LBUG in unknown intent case, just return -EFAULT
4389 Severity : enhancement
4390 Bugzilla : 9293, 9385
4391 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4392 Details : Do not serialize getattr (non-intent version) and statfs.
4395 Frequency : occasional, when OST network is overloaded/intermittent
4397 Description: client evicted by OST after bulk IO timeout
4398 Details : If a client sends a bulk IO request (read or write) the OST
4399 may evict the client if it is unresposive to its data GET/PUT
4400 request. This is incorrect if the network is overloaded (takes
4401 too long to transfer the RPC data) or dropped the OST GET/PUT
4402 request. There is no need to evict the client at all, since
4403 the pinger and/or lock callbacks will handle this, and the
4404 client can restart the bulk request.
4407 Frequency : Always when mmapping file with no objects
4409 Description: client crashes when mmapping file with no objects
4410 Details : Check that we actually have objects in a file before doing any
4411 operations on objects in ll_vm_open, ll_vm_close and
4417 Description: Request leak when working with deleted CWD
4418 Details : Introduce advanced request refcount tracking for requests
4419 referenced from lustre intent.
4421 Severity : Enhancement
4423 Description: Cache open file handles on client.
4424 Details : MDS now will return special lock along with openhandle, if
4425 requested and client is allowed to hold openhandle, even if unused,
4426 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4427 closing files for every read/write openration.
4429 Severity : Enhancement
4431 Description: Cache open negative dentries on client when possible.
4432 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4433 negative dentries on lock revocation.
4438 Description: Remounting a client read-only wasn't possible with a zconf mount
4439 Details : It wasn't possible to remount a client read-only with llmount.
4441 Severity : enhancement
4442 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4443 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4444 MPICH is included to add Lustre-specific ADIO interfaces.
4445 This is based closely on the UFS ADIO layer and only differs
4446 in file creation, in order to allow the OST striping to be set.
4447 This is user-contributed code and not supported by CFS.
4452 Description: extended inode attributes (immutable, append-only) work improperly
4453 when 2.4 and 2.6 kernels are used on client/server or vice versa
4454 Details : Introduce kernel-independent values for these flags.
4456 Severity : enhancement
4459 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4460 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4461 granularity. Now, Fractional MB granularity is be supported,
4462 this is very useful for low memory system.
4464 Severity : enhancement
4466 Description: Getattr by fid
4467 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4468 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4469 have locks and avoids getattr rpc requests for it on every stat.
4472 Frequency : Always, for filesystems larger than 2TB
4474 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4475 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4476 filesystems larger than 2TB to be mounted with mballoc enabled.
4479 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4481 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4482 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4483 filesystems larger than 2TB is mounted then the kernel will
4484 silently corrupt the start of the filesystem. CONFIG_LBD is
4485 enabled for all CFS-supported kernels, but the possibility of
4486 this happening with a modified kernel config exists.
4488 Severity : enhancement
4490 Description: add client O_DIRECT support for 2.6 kernels
4491 Details : It is now possible to do O_DIRECT reads and writes to files
4492 in the Lustre client mountpoint on 2.6 kernel clients.
4494 Severity : enhancement
4496 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4497 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4498 parallel, not waiting for response from every OST before sending
4499 a rpc to the next OST.
4504 Description: setattr vs write race when updating file timestamps
4505 Details : Client processes that update a file timestamp into the past
4506 right after writing to the file (e.g. tar) it is possible that
4507 the updated file modification time can be reset to the current
4508 time due to a race between processing the setattr and write RPC.
4510 Severity : enhancement
4512 Description: Bring 'lfs find' closer in line with regular Linux find.
4513 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4514 -print0 options and obtains all the needed info through the lustre
4517 Severity : enhancement
4519 Description: support up to 1024 configured devices on one node
4520 Details : change obd_dev array from statically allocated to dynamically
4521 allocated structs as they are first used to reduce memory usage
4526 Description: Flush dirty partially truncated pages during truncate
4527 Details : Immediatelly flush partially truncated pages in filter_setattr,
4528 this way we completely avoid having any pages in page cache on OST
4529 and can retire ugly workarounds during writes to flush such pages.
4534 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4535 Details : protect inode from truncation within vfs_unlink() context
4536 just take a reference before calling vfs_unlink() and release it
4537 when parent's i_sem is free.
4542 Description: last_id value checked outside lock on OST caused LASSERT failure
4543 Details : If there were multiple MDS->OST object precreate requests in
4544 flight, it was possible that the OST's last object id was checked
4545 outside a lock and incorrectly tripped an assertion. Move checks
4546 inside locks, and discard old precreate requests.
4549 Frequency : always, if extents are used on OSTs
4551 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4552 Details : OSTs using the extents format would not zero the high 16 bits of
4553 the index physical block number. This is not a problem for any
4554 OST filesystems smaller than 16TB, and no kernels support ext3
4555 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4556 new/modified files) and can be fixed for existing filesystems
4557 with e2fsprogs-1.39-cfs1.
4562 Description: import connection selection may be incorrect if timer wraps
4563 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4564 import connections to be ignored if the 32-bit jiffies counter
4565 wraps. Use a 64-bit jiffies counter.
4568 Frequency : very large clusters immediately after boot
4570 Description: LNET request buffers exhausted under heavy short-term load
4571 Details : If a large number of client requests are generated on a service
4572 that has previously never seen so many requests it is possible
4573 that the request buffer growth cannot keep up with the spike in
4574 demand. Instead of dropping incoming requests, they are held in
4575 the LND until the RPC service can accept more requests.
4578 Frequency : Sometimes during replay
4580 Description: Assertion failure in ll_local_open after replay.
4581 Details : If replay happened on an open request reply before we were able
4582 to set replay handler, reply will become not swabbed tripping the
4583 assertion in ll_local_open. Now we set the handler right after
4584 recognising of open request
4587 Frequency : very rare
4589 Description: kernel reports "badness in vsnprintf"
4590 Details : Reading from the "recovery_status" /proc file in small chunks
4591 may cause a negative length in lprocfs_obd_rd_recovery_status()
4592 call to vsnprintf() (which is otherwise harmless). Exit early
4593 if there is no more space in the output buffer.
4595 Severity : enhancement
4597 Description: clear OBD RPC statistics by writing to them
4598 Details : It is now possible to clear the OBD RPC statistics by writing
4599 to the "stats" file.
4604 Description: Client mtime is not the same on different clients after utimes
4605 Details : In some cases, the client was using the utimes() syscall on
4606 a file cached on another node. The clients now validate the
4607 ctime from the MDS + OSTs to determine which one is right.
4612 Description: Inability to activate failout mode
4613 Details : lconf script incorrectly assumed that in pythong string's numeric
4614 value is used in comparisons.
4617 Frequency : always with multiple stripes per file
4619 Description: Inefficient object allocation for mutli-stripe files
4620 Details : When selecting which OSTs to stripe files over, for files with
4621 a stripe count that divides evenly into the number of OSTs,
4622 the MDS is always picking the same starting OST for each file.
4623 Return the OST selection heuristic to the original design.
4628 Description: mount failures may take full timeout to return an error
4629 Details : Under some heavy load conditions it is possible that a
4630 failed mount can wait for the full obd_timeout interval,
4631 possibly several minutes, before reporting an error.
4632 Instead return an error as soon as the status is known.
4634 Frequency : quota enabled and large files being deleted
4636 Description: releasing more than 4GB of quota at once hangs OST
4637 Details : If a user deletes more than 4GB of files on a single OST it
4638 will cause the OST to spin in an infinite loop. Release
4639 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4644 Description: statfs data retrieved from /proc may be stale or zero
4645 Details : When reading per-device statfs data from /proc, in the
4646 {kbytes,files}_{total,free,avail} files, it may appear
4647 as zero or be out of date.
4650 Frequency : systems with MD RAID1 external journal devices
4652 Description: lconf's call to blkid is confused by RAID1 journal devices
4653 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4654 of returning all block devices that match the journal UUID.
4657 Frequency : always, for aggregate stripe size over 4GB
4659 Description: assertion fails when trying to use 4GB stripe size
4660 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4661 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4664 Frequency : always on ppc64
4666 Description: the first write on an ext3 filesystem with mballoc got stuck
4667 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4668 perform endianness conversion.
4670 ------------------------------------------------------------------------------
4672 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4674 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4675 INTEROPERATE with older versions automatically. Please read the
4676 user documentation before upgrading any part of a live system.
4677 * WARNING: Lustre networking configuration changes are required with
4678 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4681 * Support for kernels:
4682 2.6.9-22.0.2.EL (RHEL 4)
4683 2.6.5-7.244 (SLES 9)
4684 2.6.12.6 vanilla (kernel.org)
4687 Severity : enhancement
4688 Bugzilla : 7981/8208
4689 Description: Introduced Lustre Networking (LNET)
4690 Details : LNET is new networking infrastructure for Lustre, it includes
4691 a reorganized network configuration mode (see the user
4692 documentation for full details) as well as support for routing
4693 between different network fabrics. Lustre Networking Devices
4694 (LNDs) for the supported network fabrics have also been
4695 created for this new infrastructure.
4697 Severity : enhancement
4698 Description: Introduced Access control lists
4699 Details : clients can set ACLs on files and directories in order to have
4700 more fine-grained permissions than the standard Unix UGO+RWX.
4701 The MDS must be started with the "-o acl" mount option.
4703 Severity : enhancement
4704 Description: Introduced filesystem quotas
4705 Details : Administrators may now establish per-user quotas on the
4708 Severity : enhancement
4710 Description: Configuration change for the XT3
4711 The PTLLND is now used to run Lustre over Portals on the XT3
4712 The configure option(s) --with-cray-portals are no longer used.
4713 Rather --with-portals=<path-to-portals-includes> is used to
4714 enable building on the XT3. In addition to enable XT3 specific
4715 features the option --enable-cray-xt3 must be used.
4720 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4721 Details : A race between allocating a new llcd and re-getting the llcd_lock
4722 allowed another thread to grab newly-allocated llcd.
4724 Severity : enhancement
4726 Description: 2.6 OST async journal commit and locking fix to improve performance
4727 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4728 2.6 kernels are now async as they already were in 2.4 kernels so
4729 that they can commit concurrently with the network bulk transfer.
4730 For block-allocated files the filter allocation semaphore is held
4731 to avoid filesystem fragmentation during allocation. BKL lock
4732 removed for 2.6 xattr operations where it is no longer needed.
4737 Description: lconf incorrectly determined whether two IP networks could talk
4738 Details : In some more complicated routing and multiple-network
4739 configurations, lconf will avoid trying to make a network
4740 connection to a disjoint part of the IP space. It was doing the
4741 math incorrectly for one set of cases.
4746 Description: Fix for potential infinite loop processing records in an llog.
4747 Details : If an llog record is corrupted/zeroed, it is possible to loop
4748 forever in llog_process(). Validate the llog record length
4749 and skip the remainder of the block on error.
4752 Frequency : occasional (liblustre only)
4754 Description: liblustre could not open files whose last component is a symlink
4755 Details : sysio_path_walk() would incorrectly pass the open intent to
4756 intermediate path components.
4759 Frequency : rare (liblustre only with non-standard tuning)
4760 Bugzilla : 7201 (7350)
4761 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4762 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4763 after it has been released, only do this for referenced locks
4765 Severity : enhancement
4767 Description: specifying an (invalid) directory default stripe_size of -1
4768 would reset the directory default striping
4769 Details : stripe_size -1 was used internally to signal directory stripe
4770 removal, now use "all default" to signal dir stripe removal
4771 as a directory striping of "all default" is not useful
4774 Frequency : common for large clusters running liblustre clients
4776 Description: doing an ls when liblustre clients are running is slow
4777 Details : sending a glimpse AST to a liblustre client waits for every AST
4778 to time out, as liblustre clients will not respond. Since they
4779 cannot cache data we refresh the OST lock LVB from disk instead.
4781 Severity : enhancement
4783 Description: doing an ls at the same time as file IO can be slow
4784 Details : enqueue and other "small" requests can be blocked behind many
4785 large IO requests. Create a new OST IO portal for non-IO
4786 requests so they can be processed faster.
4789 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4791 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4792 created file with mode 000
4793 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4794 allow an ftruncate() as a non-root user to a file with mode 000.
4795 Lustre now respects this flag to disable mode checks when
4796 truncating a file owned by the user
4799 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4802 Description: Revoking locks from clients that went dead or catatonic might take
4804 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4805 cancellation of such locks instant on servers without waiting for
4806 any reply from clients. Clients drops these locks when cancel
4807 notification from server is received without replying.
4810 Frequency : liblustre-only, when liblustre client dies or becomes busy
4812 Description: Doing ls on Linux clients can take a long time with active
4814 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4815 granting such locks to it in the first place if possible. Locks
4816 are taken by proxy on the OST during the read or write and
4817 dropped immediately afterward. Add connect flags handling, do
4818 not grant locks to liblustre clients for glimpse ASTs.
4820 Severity : enhancement
4822 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4823 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4824 sequential reads and behaves badly with applications doing random
4825 reads. Improve it by reading ahead at least read region, and
4826 avoiding excessive large RPC for small reads.
4828 Severity : enhancement
4830 Description: Creating more than 1000 files for a single job may cause a load
4831 imbalance on the OSTs if there are also a large number of OSTs.
4832 Details : qos_prep_create() uses an OST index reseed value that is an
4833 even multiple of the number of available OSTs so that if the
4834 reseed happens in the middle of the object allocation it will
4835 still utilize the OSTs as uniformly as possible.
4840 Description: OST or MDS may oops in ping_evictor_main()
4841 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4842 but doesn't restart at beginning of obd_exports_timed list
4845 Severity : enhancement
4847 Description: improve by-nid export eviction on the MDS and OST
4848 Details : allow multiple exports with the same NID to be evicted at one
4849 time without re-searching the exports list.
4852 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4854 Description: MDS may oops in groups_free()
4855 Details : in rare race conditions a newly allocated group_info struct is
4856 freed again, and this can be NULL. The 2.4 compatibility code
4857 for groups_free() checked for a NULL pointer, but 2.6 did not.
4860 Frequency : common for liblustre clients doing little filesystem IO
4861 Bugzilla : 9352, 7313
4862 Description: server may evict liblustre clients accessing contended locks
4863 Details : if a client is granted a lock or receives a completion AST
4864 with a blocking AST already set it would not reply to the AST
4865 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4868 Frequency : lfs setstripe, only systems with more than 160 OSTs
4870 Description: unable to set striping with a starting offset beyond OST 160
4871 Details : llapi_create_file() incorrectly limited the starting stripe
4872 index to the maximum single-file stripe count.
4875 Frequency : LDAP users only
4877 Description: lconf did not handle in-kernel recovery with LDAP properly
4878 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4880 Severity : enhancement
4882 Description: bind OST threads to NUMA nodes to improve performance
4883 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4884 node and do their allocations there to localize memory access
4886 Severity : enhancement
4888 Description: llmount can determine client NID directly from Myrinet (GM)
4889 Details : the client NID code from gmnalnid was moved directly into
4890 llmount, removing the need to use this or specifying the
4891 client NID explicitly when mounting GM clients with zeroconf
4894 Frequency : if client is started with down MDS
4896 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4897 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4898 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4899 ptlrpc_import_delayed_req() because we didn't check if the
4900 request was interrupted, and we also didn't break out of the
4901 event loop if there was no timeout
4906 Description: data loss during non-page-aligned writes to a single file from
4907 both multiple nodes and multiple threads on one node at same time
4908 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4909 inconsistency led to false short-reads, that were cached and later
4910 used by ->prepare_write() to fill in partially written page,
4911 leading to data loss.
4914 Frequency : always, if lconf --abort_recovery used
4916 Description: lconf --abort_recovery fails with 'Operation not supported'
4917 Details : lconf was attempting to abort recovery on the MDT device and not
4920 Severity : enhancement
4922 Description: remove cleanup logs
4923 Details : replace lconf-generated cleanup logs with lustre internal
4924 cleanup routines. Eliminates the need for client-cleanup and
4927 Severity : enhancement
4929 Description: add support for EAs (user and system) on lustre filesystems
4930 Details : it is now possible to store extended attributes in the Lustre
4931 client filesystem, and with the user_xattr mount option it
4932 is possible to allow users to store EAs on their files also
4934 Severity : enhancement
4936 Description: Add possibility (config option) to show minimal available OST free
4938 Details : When compiled with --enable-mindf configure option, statfs(2)
4939 (and so, df) will return least minimal free space available from
4940 all OSTs as amount of free space on FS, instead of summary of
4941 free spaces of all OSTs.
4943 Severity : enhancement
4945 Description: do not expand extent locks acquired on OST-side
4946 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4947 by server: they are not cached anyway.
4950 Frequency : when mmap is used/binaries executed from Lustre
4952 Description: Unmmap pages before throwing them away from read cache.
4953 Details : llap_shrink cache now attempts to unmap pages before discarding
4954 them (if unmapping failed - do not discard). SLES9 kernel has
4955 extra checks that trigger if this unmapping is not done first.
4960 Description: lconf didn't resolve symlinks before checking to see whether a
4961 given mountpoint was already in use
4964 Frequency : when migrating failover services
4965 Bugzilla : 6395, 9514
4966 Description: When migrating a subset of services from a node (e.g. failback
4967 from a failover service node) the remaining services would
4968 time out and evict clients.
4969 Details : lconf --force (implied by --failover) sets the global obd_timeout
4970 to 5 seconds in order to quickly disconnect, but this caused
4971 other RPCs to time out too quickly. Do not change the global
4972 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4974 Severity : enhancement
4975 Frequency : if MDS is started with down OST
4976 Bugzilla : 9439,5706
4977 Description: Allow startup/shutdown of an MDS without depending on the
4978 availability of the OSTs.
4979 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4980 Add appropriate locking and lov-osc refcounts for safe
4981 cleaning. Add osc abort_inflight calls in case the
4982 synchronize never started.
4985 Frequency : occasional (Cray XT3 only)
4987 Description: root not authorized to access files in CRAY_PORTALS environment
4988 Details : The client process capabilities were not honoured on the MDS in
4989 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4990 been accessed by an authorized user then root was able to access
4991 the file on the local client also. The root user capabilities
4992 are now allowed on the MDS, as this environment has secure UID.
4995 Frequency : occasional
4997 Description: ldiskfs "too long searching" message happens too often
4998 Details : A debugging message (otherwise harmless) prints too often on
4999 the OST console. This has been reduced to only happen when
5000 there are fragmentation problems on the filesystem.
5005 Description: Division by zero in statfs when all OSCs are inactive
5006 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5007 lov_statfs divides by value returned by lov_get_stripecnt().
5011 Bugzilla : 9489, 3273
5012 Description: First write from each client to each OST was only 4kB in size,
5013 to initialize client writeback cache, which caused sub-optimal
5014 RPCs and poor layout on disk for the first written file.
5015 Details : Clients now request an initial cache grant at (re)connect time
5016 and so that they can start streaming writes to the cache right
5017 away and always do full-sized RPCs if there is enough data.
5018 If the OST is rebooted the client also re-establishes its grant
5019 so that client cached writes will be honoured under the grant.
5024 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5025 Details : Now I/O RPCs go to different portal number and (presumably) fast
5026 lock requests (and glimses) and other RPCs get their own service
5027 threads pool that should be able to service those RPCs
5030 Severity : enhancement
5032 Description: Ability to exchange lustre version between client and servers and
5033 issue warnings at client side if client is too old. Also for
5034 liblustre clients there is ability to refuse connection of too old
5036 Details : New 'version' field is added to connect data structure that is
5037 filled with version info. That info is later checked by server and
5041 Frequency : rare, liblustre only.
5042 Bugzilla : 9296, 9581
5043 Description: Two simultaneous writes from liblustre at offset within same page
5044 might proceed at the same time overwriting eachother with stale
5046 Details : I/O lock withing llu_file_prwv was released too early, before data
5047 actually was hitting the wire. Extended lock-holding time until
5048 server acknowledges receiving data.
5051 Frequency : extremely rare. Never observed in practice.
5053 Description: avoid generating lustre_handle cookie of 0.
5054 Details : class_handle_hash() generates handle cookies by incrementing
5055 global counter, and can hit 0 occasionaly (this is unlikely, but
5056 not impossible, because initial value of cookie counter is
5057 selected randonly). Value of 0 is used as a sentinel meaning
5058 "unassigned handle" --- avoid it. Also coalesce two critical
5059 sections in this function into one.
5061 Severity : enhancement
5063 Description: allow liblustre clients to delegate truncate locking to OST
5064 Details : To avoid overhead of locking, liblustre client instructs OST to
5065 take extent lock in ost_punch() on client's behalf. New connection
5066 flag is added to handle backward compatibility.
5068 Severity : enhancement
5069 Bugzilla : 4928, 7341, 9758
5070 Description: allow number of OST service threads to be specified
5071 Details : a module parameter allows the number of OST service threads
5072 to be specified via "options ost ost_num_threads={N}" in the
5073 OSS's /etc/modules.conf or /etc/modprobe.conf.
5077 Bugzilla : 6146, 9635, 9895
5078 Description: servers crash with bad pointer in target_handle_connect()
5079 Details : In rare cases when a client is reconnecting it was possible that
5080 the connection request was the last reference for that export.
5081 We would temporarily drop the export reference and get a new
5082 one, but this may have been the last reference and the export
5083 was just destroyed. Get new reference before dropping old one.
5085 Severity : enhancement
5086 Frequency : if client is started with failover MDS
5088 Description: Allow multiple MDS hostnames in the mount command
5089 Details : Try to read the configuration from all specified MDS
5090 hostnames during a client mount in case the "primary"
5093 Severity : enhancement
5095 Description: Stop sending data to evicted clients as soon as possible.
5096 Details : Check if the client we are about to send or are sending data to
5097 was evicted already. (Check is done every second of waiting,
5098 for which l_wait_event interface was extended to allow checking
5099 of exit condition at specified intervals).
5102 Frequency : rare, normally only when NFS exporting is done from client
5104 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5106 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5109 Frequency : always, if srand() is called before liblustre initialization
5111 Description: Liblustre uses system PRNG disturbing its usage by user application
5112 Details : Introduce internal to lustre fast and high-quality PRNG for
5113 lustre usage and make liblustre and some other places in generic
5114 lustre code to use it.
5116 Severity : enhancement
5117 Bugzilla : 9477, 9557, 9870
5118 Description: Verify that the MDS configuration logs are updated when xml is
5119 Details : Check if the .xml configuration logs are newer than the config
5120 logs stored on the MDS and report an error if this is the case.
5121 Request --write-conf, or allow starting with --old_conf.
5123 Severity : enhancement
5125 Description: Handle symlinks in the path when checking if Lustre is mounted.
5126 Details : Resolve intermediate symlinks when checking if a client has
5127 mounted a filesystem to avoid duplicate client mounts.
5132 Description: lconf can hit an error exception but still return success.
5133 Details : The lconf command catches the Command error exception at the top
5134 level script context and will exit with the associated exit
5135 status, but doesn't ensure that this exit status is non-zero.
5140 Description: failure of ptlrpc thread startup can cause oops
5141 Details : Starting a ptlrpc service thread can fail if there are a large
5142 number of threads or the server memory is very fragmented.
5143 Handle this without oopsing.
5146 Frequency : always, only if liblustre and non-default acceptor port was used
5148 Description: liblustre cannot connect to servers with non-default acceptor port
5149 Details : tcpnal_set_default_params() was not called and was therefore
5150 ignoring the environment varaible TCPNAL_PORT, as well as other
5151 TCPNAL_ environment variables
5156 Description: two objects could be created on the same OST for a single file
5157 Details : If an OST is down, in some cases it was possible to create two
5158 objects on a single OST for a single file. No problems other
5159 than potential performance impact and spurious error messages.
5163 Bugzilla : 5681, 9562
5164 Description: Client may oops in ll_unhash_aliases
5165 Details : Client dcache may become inconsistent in race condition.
5166 In some cases "getcwd" can fail if the current directory is
5172 Description: Inode refcounting problems in NFS export code
5173 Details : link_raw functions used to call d_instantiate without obtaining
5174 extra inode reference first.
5178 Bugzilla : 9942, 9903
5179 Description: Referencing freed requests leading to crash, memleaks with NFS.
5180 Details : We used to require that call to ll_revalidate_it was always
5181 followed by ll_lookup_it. Also with revalidate_special() it is
5182 possible to call ll_revalidate_it() twice for the same dentry
5183 even if first occurence returned success. This fix changes semantic
5184 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5185 reference on a request referred from the intent.
5186 ll_intent_release() then releases such a request.
5189 Frequency : rare, normally benchmark loads only
5191 Description: unlinked inodes were kept in memory on the client
5192 Details : If a client is repeatedly creating and unlinking files it
5193 can accumulate a lot of stale inodes in the inode slab cache.
5194 If there is no other client load running this can cause the
5195 client node to run out of memory. Instead flush old inodes
5196 from client cache that have the same inode number as a new inode.
5199 Frequency : SLES9 2.6.5 kernel and long filenames only
5200 Bugzilla : 9969, 10379
5201 Description: utime reports stale NFS file handle
5202 Details : SLES9 uses out-of-dentry names in some cases, which confused
5203 the lustre dentry revalidation. Change it to always use the
5207 Frequency : rare, unless heavy write-truncate concurrency is continuous
5208 Bugzilla : 4180, 6984, 7171, 9963, 9331
5209 Description: OST becomes very slow and/or deadlocked during object unlink
5210 Details : filter_destroy() was holding onto the parent directory lock
5211 while truncating+unlinking objects. For very large objects this
5212 may block other threads for a long time and slow overall OST
5213 responsiveness. It may also be possible to get a lock ordering
5214 deadlock in this case, or run out of journal credits because of
5215 the combined truncate+unlink. Solution is to do object truncate
5216 first in one transaction without parent lock, and then do the
5217 final unlink in a new transaction with the parent lock. This
5218 reduces the lock hold time dramatically.
5221 Frequency : rare, 2.4 kernels only
5223 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5224 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5225 before unmounting their filesystem in order to avoid deadlock.
5226 The kernel_locked() function in 2.4 kernels only checks whether
5227 the kernel lock is held, not whether it is this process that is
5228 holding it as 2.6 kernels do.
5233 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5234 Details : The client ptlrpc code may be trying to reconnect to a down
5235 server before a previous connection attempt has timed out.
5236 Increase the reconnect interval to be longer than the connection
5237 timeout interval to avoid sending duplicate connections to
5241 Frequency : echo_client brw_test command
5243 Description: fix echo_client to work with OST preallocated code
5244 Details : OST preallocation code (5137) didn't take echo_client IO path
5245 into account: echo_client calls filter methods outside of any
5246 OST thread and, hence, there is no per-thread preallocated
5247 pages and buffers to use. Solution: hijack pga pages for IO. As
5248 a byproduct, this avoids unnecessary data copying.
5252 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5253 Description: Client can oops in mdc_commit_close() after open replay
5254 Details : It was possible for the MDS to return an open request with no
5255 transaction number in mds_finish_transno() if the client was
5256 evicted, but without actually returning an error. Clients
5257 would later try to replay that open and may trip an assertion
5258 Simplify the client close codepath, and always return an error
5259 from the MDS in case the open is not successful.
5262 Frequency : rare, 2.6 OSTs only
5264 Description: OST may deadlock under high load on fragmented files
5265 Details : If there was a heavy load and highly-fragmented OST filesystems
5266 it was possible to have all the OST threads deadlock waiting on
5267 allocation of biovecs, because the biovecs were not released
5268 until the entire RPC IO was completed. Instead, release biovecs
5269 as soon as they are complete to ensure forward IO progress.
5271 Severity : enhancement
5273 Description: Support for specifying external journal device at mount
5274 Details : If an OST or MDS device is formatted with an external journal
5275 device, this device major/minor is stored in the ext3 superblock
5276 and may not be valid for failover. Allow detecting and
5277 specifying the external journal at mount time.
5282 Description: Mounting an MDS with pending unlinked files may cause oops
5283 Details : target_finish_recovery() calls mds_postrecov() which returned
5284 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5285 considers this an error and immediately begins cleaning up the
5286 lov, just after starting the mds_lov process
5288 Severity : enhancement
5290 Description: Implement 'lfs df' to report actual free space on per-OST basis
5291 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5292 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5293 usage of objects; '-h' to report in human readable format.
5295 ------------------------------------------------------------------------------
5297 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5304 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5305 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5306 at startup during orphan unlinking. Instead, skip buddy update
5307 before setup, as it will be regenerated after recovery is complete.
5312 Description: If an OST is inactive, its locks might reference stale inodes.
5313 Details : lov_change_cbdata() must iterate over all namespaces, even if
5314 they are inactive to clear inode references from the lock.
5316 Severity : enhancement
5317 Frequency : occasional, if non-standard max_dirty_mb used
5319 Description: Client will block write RPCs if not enough grant
5320 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5321 then the client will block writes while waiting for another RPC
5322 to complete instead of consuming its dirty limit. With change
5323 we get improved performance when max_dirty_mb is small.
5325 Severity : enhancement
5326 Bugzilla : 3389, 6253
5327 Description: Add support for supplementary groups on the MDS.
5328 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5329 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5330 lookups for user supplementary groups into a cache.
5334 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5335 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5336 for liblustre O_EXCL is enforced by MDS.
5340 Description: readdir never returns NULL in liblustre.
5341 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5342 dentry in struct dirent.
5346 Frequency : liblustre only, depends on application IO pattern
5347 Description: liblustre clients evicted if not contacting servers
5348 Details : Don't put liblustre clients into the ping_evictor list, so
5349 they will not be evicted by the pinger ever.
5351 Severity : enhancement
5353 Description: Add ability to evict clients by NID from MDS.
5354 Details : By echoing "nid:$NID" string into
5355 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5356 $NID would be instantly evicted from this MDS and from all active
5357 OSTs connected to it.
5361 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5362 Details : lookup_it_finish() used to query file size from OSTs that was not
5367 Description: service threads change working directory to that of init
5368 Details : Starting lustre service threads may pin the working directory
5369 of the parent thread, making that filesystem busy. Threads
5370 now change to the working directory of init to avoid this.
5374 Frequency : during shutdown only
5375 Description: shutdown with a failed MDS or OST can cause unmount to hang
5376 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5381 Frequency : 2.6 only
5382 Description: chmod/chown may include an extra supplementary group
5383 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5384 supplementary group and if none is specified this is used.
5387 Bugzilla : 5479 (6816)
5389 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5390 Details : If lookup is in progress on a file that is unlinked we might try
5391 to revalidate the inode and fail in revalidate after lookup is
5392 complete and ll_file_open() enqueues the open again but
5393 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5396 Frequency : always, if lconf --abort_recovery used
5398 Description: lconf --abort_recovery fails with 'Operation not supported'
5399 Details : lconf was attempting to abort recovery on the MDT device and not
5402 ------------------------------------------------------------------------------
5404 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5409 Frequency : rare (only unsupported configurations with a node running as an
5411 Bugzilla : 6514, 5137
5412 Description: Mounting a Lustre file system on a node running as an OST could
5414 Details : OSTs now preallocates memory needed to write out data at
5415 startup, instead of when needed, to avoid having to
5416 allocate memory in possibly low memory situations.
5417 Specifically, if the file system is mounted on on OST,
5418 memory pressure could force it to try to write out data,
5419 which it needed to allocate memory to do. Due to the low
5420 memory, it would be unable to do so and the node would
5421 become unresponsive.
5423 Severity : enhancement
5425 Description: Addition of lconf --service command line option
5426 Details : lconf now accepts a '--service <arg>' option, which is
5427 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5429 Severity : enhancement
5431 Description: Failover mode is now the default for OSTs.
5432 Details : By default, OSTs will now run in failover mode. To return to
5433 the old behaviour, add '--failout' to the lmc line for OSTs.
5435 Severity : enhancement
5437 Description: Health checks are now provided for MDS and OSTs
5438 Details : Additional detailed health check information on MSD and OSTs
5439 is now provided through the procfs health_check value.
5442 Frequency : occasional, depends on IO load
5444 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5445 after numerous create/delete cycles
5446 Details : The ext3 inode allocation policy would not allocate new inodes
5447 very well on the OSTs because there are no new directories
5448 being created. Instead we look for groups with free space if
5449 the parent directories are nearly full.
5454 Description: Network or server problems during mount may cause partially
5455 mounted clients instead of returning an error.
5456 Details : The config llog parsing code may overwrite the error return
5457 code during mount error handling, returning success instead
5463 Description: MDS can fail to allocate large reply buffers
5464 Details : After long uptimes the MDS can fail to allocate large reply
5465 buffers (e.g. zconf client mount config records) due to memory
5466 fragmentation or consumption by the buffer cache. Preallocate
5467 some large reply buffers so that these replies can be sent even
5468 under memory pressure.
5472 Frequency : rare (liblustre)
5473 Description: fsx running with liblustre complained that using truncate() to
5474 extend the file doesn't work. This patch corrects that issue.
5475 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5476 ATTR_SIZE and lsm use in llu_setattr_raw.
5480 Frequency : rare, only 2.6 kernels
5481 Description: Unusual file access patterns on the MDS may result in inode
5482 data being lost in very rare circumstances.
5483 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5484 code in the 2.6 kernel caused the inode and/or EA data not to
5485 be read from disk, causing single-file corruption.
5489 Frequency : rare, only 2.6 filesystems using extents
5490 Description: Heavy concurrent write and delete load may cause data corruption.
5491 Details : It was possible under high-load situations to have an extent
5492 metadata block in the block device cache from a just-unlinked
5493 file overwrite a newly-allocated data block. We now unmap any
5494 metadata buffers that alias just-allocated data blocks.
5498 Frequency : filesystems with default stripe_count larger than 77
5499 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5500 Details : lconf specifies an inode size of 4096 bytes when the default
5501 stripe_count is larger than 77. This conflicts with the default
5502 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5503 this case to avoid pinning too much memory for large EAs.
5505 ------------------------------------------------------------------------------
5507 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5512 Frequency : rare (extremely heavy IO load with hundreds of clients)
5514 Description: Client is evicted, gets IO error writing to file
5515 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5516 caused clients to be evicted for AST timeouts. The fixes in
5517 bug 5192 mean we no longer need to have such short AST timeouts
5518 so ldlm_timeout has been increased.
5521 Frequency : occasional during --force or --failover shutdown under load
5522 Bugzilla : 5949, 4834
5523 Description: Server oops/LBUG if stopped with --force or --failover under load
5524 Details : a collection of import/export refcount and cleanup ordering
5525 issues fixed for safer force cleanup
5528 Frequency : only filesystems larger than 120 OSTs
5529 Bugzilla : 5990, 6223
5530 Description: lfs getstripe would oops on a very large filesystem
5531 Details : lov_getconfig used kfree on vmalloc'd memory
5534 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5535 Bugzilla : 6242, 6243
5536 Description: reading from files that had been truncated to a non-zero size
5537 but never opened returned no data
5538 Details : ll_file_read() reads zeros from no-object files to EOF
5543 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5544 Details : There's little harm in aborting MDS/OSS recovery and letting it
5545 try again, so I removed the LASSERT and return an error instead.
5547 Severity : enhancement
5549 Description: New debugging infrastructure for tracking down data corruption
5550 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5551 (b) cover more of the client-side code path, and (c) try to narrow
5552 down where problems occurred
5556 Bugzilla : 3819, 4364, 4397, 6313
5557 Description: Racing close and eviction MDS could cause assertion in mds_close
5558 Details : It was possible to get multiple mfd references during close and
5559 client eviction, leading to one thread referencing a freed mfd.
5561 Severity: : enhancement
5562 Bugzilla : 3262, 6359
5563 Description: Attempts to reconnect to servers are now more aggressive.
5564 Details : This builds on the enhanced upcall-less recovery that was added
5565 in 1.4.2. When trying to reconnect to servers, clients will
5566 now try each server in the failover group every 10 seconds. By
5567 default, clients would previously try one server every 25 seconds.
5572 Description: After recovery, certain operations trigger a failed
5573 assertion on a client.
5574 Details : Failing over an mds, using lconf -d --failover, while a
5575 client was doing a readdir() call would cause the client to
5576 LBUG after recovery completed and the readdir() was resent.
5578 Severity : enhancement
5580 Description: Default groups are now added by lconf
5581 Details : You can now run lconf --group <servicename> without having to
5582 manually add groups with lmc.
5585 Frequency : occasional
5587 Description: Nodes with an elan id of 0 trigger a failed assertion
5590 Frequency : always when accessing e.g. tty/console device nodes
5592 Description: tty and some other devices nodes cannot be used on lustre
5593 Details : file's private_data field is used by device data and lustre
5594 values in there got lost. New field was added to struct file to
5595 store fs-specific private data.
5598 Frequency : when exporting Lustre via NFS
5600 Description: NFSD failed occasionally when looking up a path component
5601 Details : NFSD is looking up ".." which was broken in ext3 directories
5602 that had grown large enough to become hashed.
5605 Frequency : Clusters with multiple interfaces not on the same subnet
5607 Description: Nodes will repeatedly try to reconnect to an interface which it
5608 cannot reach and report an error to the log.
5609 Details : Extra peer list entries will be created by lconf with some peers
5610 unreachable. lconf now validates the peer before adding it.
5613 Frequency : Only if a default stripe is set on the filesystem root.
5615 Description: Setting a default stripe on the filesystem root prevented the
5616 filesystem from being remounted.
5617 Details : The client was sending extra request flags in the root getattr
5618 request and did not allocate a reply buffer for the dir EA.
5621 Frequency : occasional, higher if lots of files are accessed by one client
5622 Bugzilla : 6159, 6097
5623 Description: Client trips assertion regarding lsm mismatch/magic
5624 Details : While revalidating inodes the VFS looks up inodes with ifind()
5625 and in rare cases can find an inode that is being freed.
5626 The ll_test_inode() code will free the lsm during ifind()
5627 when it finds an existing inode and then the VFS later attaches
5628 this free lsm to a new inode.
5632 Bugzilla : 6422, 7030
5633 Description: MDS deadlock between mkdir and client eviction
5634 Details : Creating a new file via mkdir or mknod (starting a transaction
5635 and getting the ns lock) can deadlock with client eviction
5636 (gets ns lock and trying to finish a synchronous transaction).
5639 Frequency : occasional
5640 Description: While starting a server, the fsfilt_ext3 module could not be
5642 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5643 kernels. Previously, lconf would still use the ext3 name
5644 when trying to load modules. Now, it will correctly use
5645 ext3 on 2.4 and ldiskfs on 2.6.
5647 Severity : enhancement
5648 Description: The default stripe count has been changed to 1
5649 Details : The interpretation of the default stripe count (0, to lfs
5650 or lmc) has been changed to mean striping across a single
5651 OST, rather than all available. For general usage we have
5652 found a stripe count of 1 or 2 works best.
5654 Severity : enhancement
5655 Description: Add support for compiling against Cray portals.
5656 Details : Conditional compiling for some areas that are different
5660 Frequency : occasional
5661 Bugzilla : 6409, 6834
5662 Description: Creating files with an explicit stripe count may lead to
5663 a failed assertion on the MDS
5664 Details : If some OSTs are full or unavailable, creating files may
5665 trigger a failed assertion on the MDS. Now, Lustre will
5666 try to use other servers or return an error to the
5670 Frequency : occasional
5672 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5673 caused lock timeout during readahead (since 1.4.2).
5674 Details : Processes doing readahead might match a lock that hasn't been
5675 granted yet if there are overlapping and conflicting lock
5676 requests. The readahead process waits on ungranted lock
5677 (original lock is CBPENDING), while OST waits for that process
5678 to cancel CBPENDING read lock and eventually evicts client.
5680 Severity : enhancement
5682 Description: Initial enabling of flock support for clients
5683 Details : Implements fcntl advisory locking and file status functions.
5684 This feature is provided as an optional mount flag (default
5685 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5686 locking are implemented yet, and those that are are not guaranteed
5687 to be completely correct in production environments.
5688 mount -t lustre -o [flock|noflock] ...
5691 Frequency : occasional
5693 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5694 assertion in the ext3 JBD (journaling) layer.
5695 Details : The b_committed_data struct is protected by the big kernel lock
5696 in 2.4 kernels, serializing journal_commit_transaction() and
5697 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5698 access to this struct. In 2.6 kernels there is finer grained
5699 locking to improve SMP performance of the JBD layer.
5703 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5708 Description: merge and cleanup kernel patches.
5711 -----------------------------------------------------------------------------
5713 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5715 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5716 versions of Lustre. Please update all servers and clients to
5717 version 1.4.2 or later at the same time. You must also run
5718 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5719 to update the configuration logs.
5721 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5722 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5723 - swab llog records of type '0' so we get proper header size/idx (5861)
5724 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5725 - fix rename of one directory over another leaking an inode (5953)
5726 - avoid SetPageDirty on 2.6 (5981)
5727 - don't re-add just-being-destroyed locks to the waiting list (5653)
5728 - when creating new directories, inherit the parent's custom
5729 striping settings if present parent (3048)
5730 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5731 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5732 - don't hold client locks on temporary worklist from l_lru (5666)
5733 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5734 - automatically evict dead clients (5921)
5735 - Update file size properly in create+truncate+fstat case (6196)
5736 - Do not unhash mountpoint dentries, do not allow removal of
5738 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5739 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5740 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5741 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5742 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5743 - Correctly handle reads of files with no objects (6243)
5744 - lctl recover will also mark a device active if deactivate used (5933)
5746 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5747 - allow --write-conf on an MDS with different nettype than client (5619)
5748 - don't write config llogs to MDS for mounts not from that MDS (5617)
5749 - lconf should create multiple TCP connections from a client (5201)
5750 - init scripts are now turned off by default; run chkconfig --on
5751 lustre and chkconfig --on lustrefs to use them
5752 - upcalls are no longer needed for clients to recover to failover
5754 - add --abort-recovery option to lconf to abort recovery on device
5756 - add support for an arbitrary number of OSTs (3026)
5757 - Quota support protocol changes.
5758 - forward compatibility changes to wire structs (6007)
5759 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5760 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5761 - improved "lustre status" script
5762 - initialize blocksize for non-regular files (6062)
5763 - added --disable-server and --disable-client configure options (5782)
5764 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5765 - Vanilla 2.4.29 support
5766 - increase maximum number of obd devices to 520 (6242)
5767 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5768 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5769 SLES 9 2.6.5 kernels
5770 - limit stripes per file to 160 (the maximum EA size) (6093)
5772 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5775 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5776 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5777 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5778 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5779 - small ext3 extents cleanups and fixes (5733)
5780 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5781 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5782 - avoid needless client->OST connect, fix handle mismatch (5317)
5783 - fix DLM error path that led to out-of-sync client, long delays (5779)
5784 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5785 - fix several locking issues related to i_size (5492,5624,5654,5672)
5786 - don't move pending lock onto export if it is already evicted (5683)
5787 - fix kernel oops when creating .foo in unlinked directory (5548)
5788 - fix deadlock in obdfilter statistics vs. object create (5811)
5789 - use time_{before,after} to avoid timer jiffies wrap (5882)
5790 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5791 - Do not leak request if server was not able to process it (5154)
5792 - If mds_open unable to find parent dir, make that negative lookup(5154)
5793 - don't create new directories with extent-mapping (5909, 5936)
5795 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5796 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5797 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5798 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5799 - support for RHEL4 kernel on i686 (5773)
5800 - provide error messages when incompatible logs are encountered (5898)
5802 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5803 * version 1.4.0.10 (1.4.1 release candidate 1)
5805 - don't keep a lock reference when lock is not granted (4238)
5806 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5807 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5808 - drop import inflight refcount on signal_completed_replay error (5255)
5809 - unlock page after async write error during send (3677)
5810 - handle missing objects in filter_preprw_read properly (5265)
5811 - no transno return for symlink open, don't save no-trasno open (3440)
5812 - don't try to complete elan receive that already failed (4012)
5813 - free RPC server reply state on error (5406)
5814 - clean up thread from ptlrpc_start_thread() on error (5160)
5815 - readahead could read extra page into cache that wasn't ejected (5388)
5816 - prevent races in class_attach/setup/cleanup/detach (5260)
5817 - don't dereference de->d_inode after l_dput of de (5458)
5818 - use "int" for stripe value returned from lock_to_stripe (5544)
5819 - mballoc allocation and error-checking fixes in 2.6 (5504)
5820 - block device patches to fix I/O request sizes in 2.6 (5482)
5821 - look up hostnames for IB nals (5602)
5822 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5823 - don't start multiple acceptors for the same port (5277)
5824 - fix incorrect LASSERT in mds_getattr_name (5635)
5825 - export a proc file for general "ping" checking (5628)
5826 - fix "lfs check" to not block when the MDS is down (5628)
5828 - service request history (4965)
5829 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5830 - create an "evict_client" /proc entry on OSTs, like the MDS has
5831 - fix mount usage message, return errors per mount(8) (5168)
5832 - change grep [] to grep "[]" in tests so they work in more UMLs
5833 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5834 - remove ancient LOV_MAGIC_V0 EA support (5047)
5835 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5836 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5837 - allow readahead allocations to fail when low on memory (5383)
5838 - mmap locking landed again, after considerable improvement (2828)
5839 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5841 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5844 - send OST transaction number in read/write reply to free req (4966)
5845 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5846 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5847 - the watchdog thread now runs as interruptible (5246)
5848 - flock/lockf fixes (but it's still disabled, pending 5135)
5849 - don't use EXT3 constants in llite code (5094)
5850 - memory shortage at startup could cause assertion (5176)
5852 - reorganization of lov code
5853 - single portals codebase
5855 - add extents/mballoc support (5025)
5856 - direct I/O reads in the obdfilter (4048)
5857 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5859 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5862 - send OST transaction number in read/write reply to free req (4966)
5863 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5864 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5865 - the watchdog thread now runs as interruptible (5246)
5866 - handle missing objects in filter_preprw_read properly (5265)
5867 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5868 - drop import inflight refcount on signal_completed_replay error (5255)
5869 - unlock page after async write error during send (3677)
5870 - return original error code on reconstructed replies (3761)
5871 - no transno return for symlink open, don't save no-trasno open (3440)
5873 - add pid to ldlm debugging output (4922)
5874 - bump the watchdog timeouts -- we can't handle 30sec yet
5875 - extra debugging for orphan dentry/inode bug (5259)
5877 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5880 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5881 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5882 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5883 - fix NULL dereference in /proc/sys/portals/routes (4827)
5884 - allow failed mdc_close() operations to be interrupted (4561)
5885 - stop precreate on OST before MDS would time out on it (4778)
5886 - don't send partial-page writes before EOF from client (4410)
5887 - discard client grant for sub-page writes on large-page clients (4520)
5888 - don't free dentries not owned by NFS code, check generation (4806)
5889 - fix lsm leak if mds_create_objects() fails (4801)
5890 - limit debug_daemon file size, always print CERROR messages (4789)
5891 - use transno after validating reply (3892)
5892 - process timed out requests if import state changes (3754)
5893 - update mtime on OST during writes, return in glimpse (4829)
5894 - add mkfsoptions to LDAP (4679)
5895 - use ->max_readahead method instead of zapping global ra (5039)
5896 - don't interrupt __l_wait_event() during strace
5898 - add software watchdogs to catch hung threads quickly (4941)
5899 - make lustrefs init script start after nfs is mounted
5900 - fix CWARN/ERROR duplication (4930)
5901 - return async write errors to application if possible (2248)
5902 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5903 - print NAL number in %x format (4645)
5904 - update barely-supported suse-2.4.21-171 series (4842)
5905 - support for sles 9 %post scripts
5906 - support for building 2.6 kernel-source packages
5907 - support for sles km_* packages
5909 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - ignore -ENOENT errors in osc_destroy (3639)
5913 - notify osc create thread that OSC is being cleaned up (4600)
5914 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5915 - reconstruct ost_handle() like mds_handle() (4657)
5916 - create a new thread to do import eviction to avoid deadlock (3969)
5917 - let lconf resolve symlinked-to devices (4629)
5918 - don't unlink "objects" from directory with default EA (4554)
5919 - hold socknal file ref over connect in case target is down (4394)
5920 - allow more than 32000 subdirectories in a single directory (3244)
5921 - fix blocks count for O_DIRECT writes (3751)
5922 - OST returns ENOSPC from object create when no space left (4539)
5923 - don't send truncate RPC if file size isn't changing (4410)
5924 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5925 - bind to privileged port in socknal and tcpnal (3689)
5927 - rate limit CERROR/CWARN console message to avoid overload (4519)
5928 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5929 - basic mmap support (3918)
5930 - kernel patch series update from b1_4 (4711)
5932 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5935 - avoid crash during MDS cleanup with OST shut down (2775)
5936 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5937 - don't use bad inodes on the MDS (3744)
5938 - dynamic object preallocation to improve recovery speed (4236)
5939 - don't hold spinlock over lock dumping or change debug flags (4401)
5940 - don't zero obd_dev when it is force cleaned (3651)
5941 - print grants to console if they go negative (4431)
5942 - "lctl deactivate" will stop automatic recovery attempts (3406)
5943 - look for existing locks in ldlm_handle_enqueue() (3764)
5944 - don't resolve lock handle twice in recovery avoiding race (4401)
5945 - revalidate should check working dir is a directory (4134)
5947 - don't always mark "slow" obdfilter messages as errors (4418)
5949 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5952 - don't close LustreDB during write_conf until it is done (3860)
5953 - fix typo in lconf for_each_profile (3821)
5954 - allow dumping logs from multiple threads at one time (3820)
5955 - don't allow multiple threads in OSC recovery (3812)
5956 - fix debug_size parameters (3864)
5957 - fix mds_postrecov to initialize import for llog ctxt (3121)
5958 - replace config semaphore with spinlock (3306)
5959 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5960 - don't allow enqueue to complete on a destroyed export (3822)
5961 - down write_lock before checking llog header bitmap (3825)
5962 - recover from lock replay timeout (3764)
5963 - up llog sem before sending rpc (3652)
5964 - reduce ns lock hold times when setting kms (3267)
5965 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5966 - fix NULL deref and obd_dev leak on setup error (3312)
5967 - replace some LBUG about llog ops with error handling (3841)
5968 - don't match INVALID dentries from d_lookup and spin (3784)
5969 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5970 - fix invalid assertion in ptlrpc_set_wait (3880)
5972 - add libwrap support for the TCP acceptor (3996)
5973 - add /proc/sys/portals/routes for non-root route listing (3994)
5974 - allow setting MDS UUID in .xml (2580)
5975 - print the stack of a process that LBUGs (4228)
5977 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5980 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5981 - make sure to unset replay flag from failed open requests (3440)
5982 - if default stripe count is 0, use OST count for inode size (3636)
5983 - update parent mtime/ctime on client for create/unlink (2611)
5984 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5985 - free recovery state on server during a forced cleanup (3571)
5986 - unregister_reply for resent reqs (3063)
5987 - loop back devices mounting and status check on 2.6 (3563)
5988 - fix resource-creation race that can provoke i_size == 0 (3513)
5989 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5990 - more debugging for page-accounting assertion (3746)
5991 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5992 - avoid deadlock after precreation failure (3758)
5993 - fix race and lock order deadlock in orphan handling (3450, 3750)
5994 - add validity checks when grabbing inodes from l_ast_data (3599)
5996 - add /proc/.../recovery_status to obdfilter (3428)
5997 - lightweight CDEBUG infrastructure, debug daemon (3668)
5998 - change default OSC RPC parameters to be better on small clusters
5999 - turn off OST read cache for files smaller than 32MB
6000 - install man pages and include them in rpms (3100)
6001 - add new init script for (un)mounting lustre filesystems (2593)
6002 - run chkconfig in %post for init scripts (3701)
6003 - drop scimac NAL (unmaintained)
6005 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6008 - clean kiobufs before and after use (3485)
6009 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6010 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6011 - change init script to fail more clearly if not run as root (1528)
6012 - allow clients to reconnect during replay (1742)
6013 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6014 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6015 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6016 - fix cond_resched (3554)
6017 - only evict unfinished clients after recovery (3515)
6018 - allow bulk resend, prevent data loss (3570)
6019 - dynamic ptlrpc request buffer allocation (2102)
6020 - don't allow unlinking open directory if it isn't empty (2904)
6021 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6022 - remove extraneous obd dereference causing LASSERT failure (3334)
6023 - don't use get_cycles() when creating temp. files on the mds (3156)
6024 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6025 - handle EEXIST for set-stripe, set proper directory name (3336)
6027 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6028 - fix ksocknal_fmb_callback() error messages (2918)
6030 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6033 - don't copy lvb into (possibly NULL) reply on error (2983)
6034 - don't deref dentry after dput, don't free lvb on error (2922)
6035 - use the kms to determine writeback rpc length (2947)
6036 - increment oti_logcookies when osc is inactive (2948)
6037 - update client's i_blocks count via lvb messages (2543)
6038 - handle intent open/close of special files properly (1557)
6039 - mount MDS with errors=remount-ro, like obdfilter (2009)
6040 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6041 - don't use cancelling-locks' kms values (2947)
6042 - use highest lock extent for kms, not last one (2925)
6043 - don't dereference ERR_PTR() dentry in error handling path (3107)
6044 - fix thread race in portals_debug_dumplog() (3122)
6045 - create lprocfs device entries at setup instead of at attach (1519)
6046 - common AST error handler, don't evict client on completion race (3145)
6047 - zero nameidata in detach_mnt in 2.6 (3118)
6048 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6049 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6050 - initialize RPC timeout timer earlier for 2.6 (3219)
6051 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6052 - print nal/nid for unknown nid (3258)
6053 - additional checks for oscc recovery before doing precreate (3284)
6054 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6055 - don't crash in mdc_close for bad permissions on open (3285)
6056 - zero i_rdev for non-device files (3147)
6057 - clear page->private before handing to FS, better assertion (3119)
6058 - tune the read pipeline (3236)
6059 - fix incorrect decref of invalidated dentry (2350)
6060 - provide read-ahead stats and refine rpc in flight stats (3328)
6061 - don't hold journal transaction open across create RPC (3313)
6062 - update atime on MDS at close time (3265)
6063 - close LDAP connection when recovering to avoid server load (3315)
6064 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6065 - don't leak open file on MDS after open resend (3325)
6066 - serialize filter_precreate and filter_destroy_precreated (3329)
6067 - loop device shouldn't call sync_dev() for nul device (3092)
6068 - clear page cache after eviction (2766)
6069 - resynchronize MDS->OST in background (2824)
6070 - refuse to mount the same filesystem twice on same mountpoint (3394)
6071 - allow llmount to create routes for mounting behind routers (3320)
6072 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6073 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6074 - fix rare problem with rename on htree directories (3417)
6076 - allow default OST striping configuration per directory (1414)
6077 - fix compilation for qswnal for 2.6 kernels (3125)
6078 - increase maximum number of MDS request buffers for large systems
6079 - change liblustreapi to be useful for external progs like lfsck (3098)
6080 - increase local configuration timeout for slow disks (3353)
6081 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6083 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6086 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6087 - don't overwrite extent policy data in reply if lock was blocked (2901)
6088 - drop filter export grants atomically with removal from device (2663)
6089 - del obd_self_export from work_list in class_disconnect_exports (2908)
6090 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6091 - swab reply message in mdc_close, other PPC fixes (2464)
6092 - fix destroying of named logs (2325)
6093 - overwrite old logs when running lconf --write_conf (2264)
6094 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6095 - fix race in target_handle_connect (2898)
6096 - mds_reint_create() should take same inode create lock (2926)
6097 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6098 - don't close files for self_export to avoid uninitialized obd (2936)
6099 - allow MDS with the same name as client node (2939)
6100 - hold dentry reference for closed log files for unlink (2325)
6101 - reserve space for all logs during transactions (2059)
6102 - don't evict page beyond end of stripe extent (2925)
6103 - don't oops on a deleted current working directory (2399)
6104 - handle hard links to targets without a parent properly (2517)
6105 - don't dereference NULL lock when racing during eviction (2867)
6106 - don't grow lock extents when lots of conflicting locks (2919)
6108 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6111 - account for cache space usage on clients to avoid data loss (974)
6112 - lfsck support in lustre kernel code (2349)
6113 - reduce journal credits needed for BRW writes (2370)
6114 - orphan handling to avoid losing space on client/server crashes
6115 - ptlrpcd can be blocked, stopping ALL progress (2477)
6116 - use lock value blocks to assist in proper KMS, faster stat (1021)
6117 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6118 - recovery for initial connections (2355)
6119 - fixes for mds_cleanup_orphans (1934)
6120 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6121 - block all file creations until orphan recovery completes (1901)
6122 - client remove rq_connection from request struct (2423)
6123 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6124 - recovery timer race (2670)
6125 - mdc_close recovey bug (2532)
6126 - ptlrpc cleanup bug (2710)
6127 - mds timeout on local locks (2588)
6128 - namespace lock held during RPCs (2431)
6129 - handle interrupted sync write properly (2503)
6130 - don't try to handle a message that hasn't been replied to (2699)
6131 - client assert failure during cleanup after abort recovery (2701)
6132 - leak mdc device after failed mount (2712)
6133 - ptlrpc_check_set allows timedout requests to complete (2714)
6134 - wait for inflight reqs when ptlrpcd finishes (2710)
6135 - make sure unregistered services are removed from the srv_list
6136 - reset bulk XID's when resending them (caught by 1138 test)
6137 - unregister_bulk after timeout
6138 - fix lconf error (2694)
6139 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6140 - readahead locks pages, leaves pending causing memory pressure (2673)
6141 - increase OST request buffers to 4096 on large machines (2729)
6142 - fix up permission of existing directories in simple_mkdir (2661)
6143 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6144 - don't assign transno to errored transactions (2742)
6145 - don't delete objects on OST if given a bogus objid from MDS (2751)
6146 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6147 - if rq_no_resend, then timeout request after recovery (2432)
6148 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6149 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6150 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6151 - MDS recovery shouldn't delete objects that it has given out (2730)
6152 - if enqueue arrives after completion, don't clobber LVB (2819)
6153 - don't unlock pages twice when trigger_group_io returns error (2814)
6154 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6155 - don't write pages to disk if there was an error (1450)
6156 - don't ping imports that have recovery disabled (2676)
6157 - take buffered bytes into account when balancing socknal conn (2817)
6158 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6159 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6160 - remove little-endian swabbing of llog records (1987)
6161 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6162 - retry reposting request buffers if they fail (1191)
6163 - grow extent at grant time to avoid granting a revoked lock (2809)
6164 - lock revoke doesn't evict page if covered by a second lock (2765)
6165 - disable VM readahead to avoid reading outside lock extents (2805)
6167 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6168 - updated kernel patches for hp-2.4.20 kernel (2681)
6170 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6173 - fix truncated write corruption (2366)
6174 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6176 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6177 - ensure len(uuid) < 37 in lmc (1171)
6178 - fix ia64 OOPS in llog_test (2255)
6179 - zero end of page at obdfilter for partial page writes (2648)
6180 - don't leave stale dentries around after renames (bug 2428)
6181 - fix timeouts when evicting a client with a single lock held (2642)
6182 - set deadline for the initial HELLO message to drain (2634)
6183 - print out dotted-quad IP addresses in the socknal (2302)
6185 - additional debugging for MDS client eviction problem (2443)
6186 - fix mkfsoptions support for osts (2603, 2604)
6188 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6191 - add series for the vanilla 2.6.0 kernel
6192 - add series for the vanilla 2.4.24 kernel
6193 - add series for a cray x86/64 UL kernel drop
6194 - fix xattr patches for the vanilla 2.4.19 series
6196 - generate true UUIDs in lmc (1171)
6197 - have portals stack dumping break in UML (2466)
6198 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6199 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6200 - wake callback waiting threads on client eviction (2460)
6201 - Add --ptldebug and --subsystem to lmc (1719)
6202 - update assertion to allow safe interrupt allocation
6203 - set rq_no_resend for cancel requests (2432)
6204 - recalculate ptlrpcd timeout after resend (2494)
6205 - call vfs_rmdir when removing pending directories (2368)
6206 - fix renaming a file to itself (2429)
6207 - lmc creates a default one-stripe lov (2454)
6208 - expand procfs space to handle large clusters (2326)
6209 - increase UML stack to avoid overflow
6210 - update lconf's list of debug and subsystem masks
6211 - fix lfs find --obd (2510)
6212 - /proc tunable for disabling filter read caching (2591)
6213 - stop rpm packages from altering slapd.conf (2301)
6214 - disable nagle in the socknal under 0conf (2578)
6215 - choose mds inode size based on stripe count (2572)
6216 - fix kernel-source rpm problems (2516)
6218 - add --disable-doc to avoid pdf generation (2421)
6219 - update documentation, tests, type-os, comments
6220 - avoid format warnings on ia64
6221 - remove the TOE NAL
6222 - tiny code cleanups by removing unused fields
6224 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6227 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6228 - remove the most blatant lies from BUILDING file (2371)
6229 - change default debug level to reasonable production setting
6230 - reduce client side cache size to reduce cache flush time
6231 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6232 - make TCP zerocopy and pinger support enabled by default (2476)
6233 - sync writes completed after process exits caused crashes (2319)
6234 - maintain correct mount count on the MDS (2356)
6235 - backout 1557, because 2316 wasn't really fixed
6236 - better file I/O statistics gathering in /proc
6237 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6238 - another kernel patch to fix zero-copy TCP function export
6239 - don't take duplicate lock when processing re-sent getattr (2420)
6240 - lctl uses obd_self_export instead of creating new conn (2353)
6241 - MDS/OST recovery case which requires object creation asserted (2425)
6242 - move lfs from /usr/sbin to /usr/bin in packages
6243 - fix race between mds_client_add and mds_client_free (2417)
6244 - use kmalloc instead of slabs in portals (2430)
6245 - don't create duplicate records when a failover MDS is present (2442)
6246 - remove unnecessary mount age check (2332)
6247 - don't remove directory inodes from locks prematurely (2451)
6248 - don't break if MDS service name is the same as hostname (2103)
6249 - fix races in client write RPC generation when cache full (2482)
6251 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6254 - remove now-unused request->rq_obd (278)
6255 - if an allocation fails, print out how much memory we've used (1933)
6256 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6257 - add the "configurable stack size" patch to most series files (1256)
6258 - ability to write large log records, for 100+ OST configs (2306)
6259 - fix NULL deref when filter_prep fails (2314)
6260 - fix operator precedence error in filter_sync
6261 - dynamic allocation of socknal TX descriptors (2315)
6262 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6263 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6264 - ensure that CWARN messages in Portals always get to the syslog
6265 - __init/__exit are not for prototype decls (ldlm_init/exit)
6266 - x86-64 compile warning fixes
6267 - fix gateway LMC keyword conflict (2318)
6268 - fix MDS lock inversions in getattr/reint paths (1844)
6269 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6270 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6271 - DLM race condition prevented some lock evictions (2328)
6272 - ENOMEM detection and retry on socknal sends (2230)
6273 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6274 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6275 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6276 - better error messages when a client is rejected during recovery (1505)
6277 - avoid cancelling locks which were never granted, after failure (2330)
6278 - fix i_sem/journal inversion in mds_client_add (2333)
6279 - fix truncate/getattr lock cycle deadlock (2334)
6280 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6281 - fix two rare exit paths which could leak an l_lock() ref (2321)
6282 - fencepost error in MDS/OST orphan recovery (2226)
6283 - make log record alignment 8 bytes (1988)
6284 - lstripe now fails when requested offset > ost_count (2237)
6285 - ensure that all kernel series have a complete list.h (1607)
6286 - fix crashes in special-file operations (2316)
6287 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6289 - allow configurable automake binary, for testing new versions
6290 - small update to the lfs documentation
6292 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6294 * fix negative export reference count in fsfilt_sync (2312)
6296 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6297 * release candidate 0.9.1
6299 - orphans are moved into the PENDING directory for possible recovery
6300 - replayed opens now open by fid for orphan/rename safety (1042)
6301 - last close of an orphan inode generates a transno (683)
6302 - chdir() and mount() now pin the directory entry (1020)
6303 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6304 - discard very old requests without processing them (1502)
6305 - remove some common, well-understood CERRORs (1505)
6306 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6307 - clear "grant" flags in OST replies until OST grant code lands (1644)
6308 - fix read performance by not clobbering i_blksize on client (1598)
6309 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6310 - make LRU size a /proc tunable, clears locks when reduced (707)
6311 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6312 - server locks take references on exports now (1558)
6313 - build fixes for 2.4.20-rh trees (1663)
6314 - return an error from lov_create if all OSCs are inactive (1751)
6315 - fix import levels when a reconnect happens without a timeout (1597)
6316 - exit early from mds_open if we get a lookup error (1749)
6317 - partial page read at EOF wouldn't wait for disk before sending (1642)
6318 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6319 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6320 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6321 - write/truncate lock inversion (1639)
6322 - Don't auto-load obdclass, portals modules during cleanup (1495)
6323 - fix timestamps from jumping to "now" (1763)
6324 - extra journal assertions (1648)
6325 - add an extra multiunlink test (1771)
6326 - fix read_record/write_record API (1776)
6327 - fix leak of offset_extent, possible incorrect i_size later (1772)
6328 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6329 - Debugging for the kqswnal_get_idle_tx problems (1820)
6330 - Allow recovery to be attempted multiple times (1536)
6331 - Write out MDS last_rcvd file after it is first created (1600)
6332 - Fix tx_descriptor leak in failed transmit situations (1827)
6333 - ext3 journaling fixes for assertion failure after IO error (1871)
6334 - class_export_put() on freed export after completion AST error (1896)
6335 - Fix revalidate looping in VFS (1322)
6336 - Don't access a freed export during MDS_REINT timeout (1521)
6337 - Add open-unlink recovery support on the MDS (1673,1764)
6338 - Return an error if no MDS data was read from last_rcvd (1946)
6339 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6340 - Don't setup a disk device that doesn't match exported UUID (317)
6341 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6342 - avoid committing NULL handle in force close
6343 - local.sh is now a one-stripe LOV configuration
6344 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6345 - fix ext3 htree duplicate directory entry corruption (1516)
6346 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6347 - update to new LOV EA format (2097)
6348 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6350 2003-06-15 Phil Schwan <phil@clusterfs.com>
6353 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6354 - per-import/export recovery handling (958, 931, 959)
6355 - multiple last-rcvd slots, for serving multiple FSes (949)
6356 - connections are again shared between multiple imp/exports (963, 964)
6357 - "umount -f" would hang if any requests needed to be sent (393, 978)
6358 - avoid pinning large req buffer by copying for queued messages (989)
6359 - add "uuid" to "lctl device" command to help upcalls (991)
6360 - "open" RPCs with transnos would confuse recovery counters (1037)
6361 - do proper endian conversion of all wire messages (288, 340, 891)
6362 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6363 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6364 - fix ptlrpc_connection leak in target_handle_connect (1174)
6365 - fix import refcounting bug in OST and MDS cleanup (1134)
6366 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6367 - fix very unlikely obd_types race condition (501)
6368 - remove osc_open hack for echo_client (1187)
6369 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6370 - a failure in read_inode2 leads to deadlock (1139)
6371 - cancel ack-locks as soon as transaction is committed (1072)
6372 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6373 - make sure to commitrw after any preprw to avoid deadlock (1162)
6374 - failing to execute a file in a lustre FS would lock inode (1203)
6375 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6376 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6377 - obd_uuid2tgt would walk off the end of the list (1255)
6378 - on IA64 the osc would give portals incorrect bulk size (1258)
6379 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6380 - fix lock inversion caused by new llite matching code (1282)
6381 - limit the number of dirty pages on a client to 10MB (1286)
6382 - timed out locks were not being corrected cancelled (1289)
6383 - fix O_DIRECT above 4GB on IA-32 (1292)
6384 * major user-visible changes
6385 - fail out/fail over policy now controlled by the upcall (993)
6387 - add OBD_PING to check server availability and failure (954)
6388 - lustre messages are now sent in sending host order (288, 340, 891)
6389 - add eadatalen to MDS getattr reply (340)
6390 - OST read replies may contain second buffer, with per-page status (593)
6392 2003-03-11 Phil Schwan <phil@clusterfs.com>
6395 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6396 - failed open()s could cause deadlock; fixed (867, 869)
6397 - stop cancelling OST locks when files are closed (481)
6398 - overlapping XID spaces caused network corruption (851, 853)
6399 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6400 - setattr_raw added, to do single-RPC, server-side setattrs
6401 - lmc/lconf syntax change for OST UUIDs
6402 - fix crashy race condition between ptlrpc_free_req and osc_close
6403 - don't use request in mdc_enqueue if we hit a timeout (889)
6404 - don't set the inode i_size for regular files from the MDS (896)
6405 - handle out of order completion AST (842)
6406 - don't LBUG if a lock request times out after receiving AST (913)
6407 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6408 - if a bad lock AST arrives, send an error instead of dropping entirely
6409 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6410 - fix leak in bulk IO when only partially completed (899, 900, 926)
6411 - fix O_DIRECT for ia64 (55)
6412 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6413 - C-z after timeout could hang a process forever; fixed (977)
6415 - client-side I/O cache (678, 924, 929, 941, 970)
6417 - READPAGE and SETATTRs which don't take server-side locks get
6420 2003-02-11 Phil Schwan <phil@clusterfs.com>
6423 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6424 - Fix fsfilt_extN_readpage() to read a full page of directory
6425 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6426 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6427 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6428 - Most intent operations were removed, in favour of a new RPC mode
6429 that does a single RPC to the server and bypasses most of the VFS
6430 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6431 - Aggressively cancel local locks on DLM servers
6432 - mds_reint_unlink sends EA to the client if it's the last nlink.
6433 client uses that EA to unlink OST objects.
6434 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6435 - recursive symlinks were fixed (439)
6436 - fixed NULL deref in DEBUG_REQ
6437 - filter_update_lastobjid no longer calls sync, which annoyed extN
6438 - fixed multi-client small-writes to a single file problem (445)
6439 - fixed mtime updates during file writes (607)
6440 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6441 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6442 - fixed Linux OST size reporting problem (444, 656)
6443 - OST now updates object mtime with writes or setattr (607, 619)
6444 - client verifies file size before zeroing page past EOF (445)
6445 - OST now writes last allocated objid to disk with allocation (108)
6446 - LOV on echo now works (409)
6448 - mds_reint_unlink sends a new buffer, with the EA included. this
6449 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6450 set if a regular file was being unlinked, and it was the last link
6451 - use PtlGet from the target for bulk writes (315)
6452 - OST now updates object mtime with writes or setattr (607, 619)
6453 - LDLM now has a grant-time callback to revalidate locked items, if
6455 - Many MDS operations were reorganized to combat race conditions
6457 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6458 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6460 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6463 - Fully reactivate OST imports after reconnection (512, others)
6464 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6465 - More graceful error handling for truncating on dead OST (515)
6466 - Don't error out unless we're actually accessing dead stripes (474)
6467 - Fix garbage sizes when stripes are missing (410)
6468 - LRU counters were broken, causing constant lock purge (433, 432)
6469 - garbage on read from stripes with failed OSTs (441)
6470 - mark OSCs as active before reconnecting during recovery (438)
6471 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6472 - lfind did not preserve OST order in output (443)
6473 - symlinks cause hung clients, incorrect data (439)
6474 - stop dereferencing request after dropping refcount (457)
6475 - don't LASSERT(spin_is_locked) on non-SMP (455)
6476 - fixes for many rename() bugs
6477 - fstat didn't correctly synchronize attributes (399)
6478 - server must handle lock cancellation during blocking AST prep (487)
6479 - bulk descriptors were free()d too soon (511)
6480 - fix paths in lconf, which would load incorrect modules (451, 507)
6481 - fix confusing lconf 'host not found' error message (386)
6482 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6483 - fix race condition in mdc_blocking_ast() for inode access (526)
6484 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6485 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6486 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6487 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6488 - fix lstripe in several ways (kernel side) (527)
6489 - fix request leak in ldlm_cli_enqueue (262)
6490 - incorrect OSC was marked inactive after OST failure
6491 - call mds_fs_cleanup before unmounting filesystem (524)
6492 - fix races between taking ns_lock and ldlm_lock_change_resource
6493 - fix races updating LOV export open file list
6494 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6495 - fix recovery NULL deref in ldlm_cli_cancel_unused
6496 - fix some DLM races by using new hash table for lock handles (419)
6497 - permit the client to specify desired inodes, at replay
6498 - duplicate requests when we queue them for replay reintegration
6499 - fix last_rcvd offset calculation
6500 - sync after each recovered transaction, so we always make progress
6501 - never, not always, ERESTART requests without transnos
6502 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6504 - skip replay if the MDS says that the client is already connected
6505 - don't check for a recovery-enabled export to match lctl's UUID
6506 - don't INC_USE_COUNT for phantom exports
6507 - don't crash when cleaning up phantom exports (567)
6508 - don't double-finish or set replay data for errored mdc_open requests
6509 - abort requests when they time out, so we don't get old replies
6510 - send/receive replies for AST messages again
6511 - if the client says that it doesn't have the lock, cancel it on the
6513 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6514 mark it as destroyed, it will all work out when decref is called
6515 - fix module use counts (22, 581)
6517 - ASTs now expect a reply (server cancels lock on error reply)
6519 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6522 - fix many simultaneous client startup (392)
6523 - fix dentry->d_it clobbering
6524 - credentials weren't being shipped for readdir/getattr operations
6525 - remove invalid assertions triggered during some concurrent MD
6527 - proper Lustre versions added (336, 389)
6528 - fix memory leak for create error case (398)
6529 - fix LOV locking bug that would get cli/srv out of sync
6530 - fix echo client over LOV (409)
6531 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6532 - fix double-O_EXCL intent crash (424)
6533 - avoid sending multiple lock CANCELs (352)
6535 - MDS can do multi-client recovery (modulo bugs in new code)
6537 - many updates, edits, cleanups
6539 2002-11-18 Phil Schwan <phil@clusterfs.com>
6542 - fix null d_it dereference (346)
6543 - fix full OST/dbench hang (333)
6544 - fix permission problem with file removal (286)
6545 - fix removal of OSCs from LOV when they fail
6546 - fix NULL deref during bulk timeout (214)
6547 - fix problems related to multiple filesystems on one MDS (241)
6548 - fixed serious subtle metadata locking bugs
6549 - free locks on clients when inodes are removed due to memory
6551 - fix inode pointer in lock data (285)
6552 - partial support for multiple MDS on a single host (241)
6553 - data locks weren't cancelled at clear_inode time (290, 311)
6554 - intent locks could lead to unbounded lock growth (205)
6555 - added a maximum lock count, an LRU list, and a flusher
6556 - fix multiple rename (365)
6557 - properly abstracted the echo client
6558 - OSC locked 1 byte too many; fixed
6559 - rewrote brw callback code:
6560 - fixed recovery bugs related to LOVs (306)
6561 - fixed too-many-pages-in-one-write crash (191)
6562 - fixed (again) crash in sync_io_timeout (214)
6563 - probably fixed callback-related race (385)
6565 - Add capability to MDS protocol
6566 - LDLM cancellations and callbacks on different portals
6568 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6571 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6572 - defer open object destruction to close time (601981, 138)
6573 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6574 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6575 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6576 - don't call obd_setattr in ll_file_release for destroyed objects
6577 * protocol change to lustre_msg: move |version| and add |flags|
6578 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6579 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6580 * added replay of create, unlink, link and rename operations during
6581 MDS failover; recovery should be much more robust now
6582 * remove failed OSCs from LOVs (only lov_create uses this so far)
6583 * the lustre-HOWTO was brought (more) up to date (582544)
6585 2002-10-23 Phil Schwan <phil@clusterfs.com>
6588 - in-use dentries weren't being reused properly (617851)
6589 - prevent multiple LDLM setup (599178)
6590 - fix LOV size calculations for truncate (617853)
6591 - fix client handling of MDS intent errors (POSIX)
6592 - fix permission bug in lovstripe.c test (624321)
6593 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6594 - truncate past end of file could corrupt data
6595 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6596 - a race in recovery could return ETIMEDOUT to apps (623947)
6597 - building outside the source directory was fixed
6598 * the lustre-HOWTO was brought (more) up to date (582544)
6599 * major progress was made on recovery functionality
6601 2002-10-10 Phil Schwan <phil@clusterfs.com>
6604 - recovery deadlock fix
6605 - rm -rf causes LBUG fix (617817)
6606 - file open by multiple tasks fix (618962)
6607 - directory permissions bugs (602707 and 620007)
6608 - journal_stop fixed with locking (611313)
6609 - O_APPEND failures resolved (618273, perhaps 614459)
6610 - lconf PATH fix (619770)
6611 - IA64 build fix (621450)
6612 - RPC buffer sizes scale with amount of memory
6614 2002-10-01 Phil Schwan <phil@clusterfs.com>
6617 - locks would be cancelled without throwing away data pages,
6618 resulting in inconsistent data (605627)
6619 - inode attributes were not always being refreshed (605627, 612449)
6620 - lconf now continues to cleanup after lctl reports an error
6621 - MDS now enforces user permissions (602707)
6622 - lprocfs cleanup fixed, but not yet enabled (614157)
6623 - fixed infinite server hang, should a client not respond to an AST
6624 - avoid going into recovery if user calls readlink() with a buffer
6625 that's too small (613941)
6626 - AST RPCs no longer require replies (614867) -- this may be changed
6627 - don't crash server if client sends an IOV that's too big (611336)
6628 - fixed lock conversion deadlock (611892)
6629 - fixed the following of symlinks (614622)
6630 * recovery: the server can remove locks from a client that dies, other
6631 clients can make progress
6632 * more extN patch fixes
6633 * compile-time configurable ptlrpc buffer allocations
6635 - collaborative read cache document
6636 - Lustre Lite Performance CDR document-in-progress
6638 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6641 - fix typo in patch-2.4.18
6643 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6646 - clear ptlrpc request each time in handle_incoming_request()
6647 - unlink of files now destroys the object on the OST
6649 2002-09-19 Peter Braam <braam@clusterfs.com>
6651 * add hard link support
6652 * change obdfile creation method
6653 * kernel patch changed
6655 2002-09-19 Peter Braam <braam@clusterfs.com>
6658 - stack overflow bug in extN fixed
6660 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6662 * documentation updates
6663 - add man pages for config tools
6664 - update tests/README to describe testing with new config tools
6665 - finish metadata API descriptions
6666 * bug fixes and cleanups
6667 - statfs workaround for 16TB limit
6668 - LOV stripe allocation improved, can stripe on subset of OSTs
6669 - LOV file size/IO offset was wrong for files > 4GB in size
6670 - object EA data was being dropped, caused files to be unreadable
6671 - memory overflow with non-LOV OST caused memory corruption
6672 - fixed regression tests to work with new config tools, obdfilter
6673 - fixed bug when directory size became larger than 1 block
6674 - fixed bug (for single client case) when PWD was deleted
6675 - invalidate local directory pages when doing intent-based ops
6676 - avoid LDLM oops when lock callback contained bad data
6678 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6680 * documentation updates
6681 * bug fixes and cleanups
6682 - configuration tools
6685 - 64-bit compile warnings
6686 - 64-bit internal statfs data
6688 * test_brw on persistent OST devices
6690 * lprocfs (disabled)
6692 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6694 * documentation updates
6695 * bug fixes and cleanups
6696 * configuration tools
6698 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6701 * many small fixes to 0_5_4
6702 * io/network handling
6703 * thinkos in MDS operations
6705 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6708 * crucial basic fixes to 0.5.3
6709 * IOR, Iozone work over Elan
6712 2002-08-07 Phil Schwan <phil@clusterfs.com>
6713 * version 0_5_3, our first alpha
6714 * we use the new Portals iovs
6715 * documentation updates
6716 * bug fixes and cleanups
6717 * small changes in the DLM wire protocol
6719 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6720 * version 0_5_1 with some initial stability,
6721 * locking on MD and file I/O.
6722 * documentation updates
6723 * several bug fixes since 0.5.0
6724 * small changes in wire protocol
6726 2002-07-18 Phil Schwan <phil@clusterfs.com>
6728 * delivered as Lustre Light Alpha
6729 * fixed a crash after handling invalid MDS requests
6730 * fixed directory pages for architectures with non-4k pages sizes
6732 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6733 * release version v0_4_4
6734 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6736 * Fixes a number of interruption problems with OST operations.
6737 * Update documentation for portals header changes
6738 * Move all wire protocol structs/defines to lustre_idl.h
6739 * Fixes symlink length bug.
6740 * Add tcpdump to repository.
6742 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6743 * release version v0_4_3
6744 * Fixes statfs for inodes on extN.
6745 * Fixes bug in runtests which would delete /etc/hosts.
6746 * Use 64-bit object IDs wherever possible (not into VFS though)
6747 Remove ost_get_info, which is unused by lustre, and out of date.
6749 2002-07-03 Peter Braam <braam@clusterfs.com>
6750 * release version v0_4_2 Fixes a lookup error (type not passed)
6751 * move forward to head of Portals
6752 * move forward to latest Lustre kernel
6754 2002-06-25 Peter Braam <braam@clusterfs.com>
6755 * release version v0_4_1. Hopefully stable on single node use.