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 3.10.0-1062.9.1.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 4.18.0-80.11.2.el8 (RHEL8.0)
11 4.18.0-147.3.1.el8 (RHEL8.1)
12 4.4.120-92.70 (SLES12 SP2)
13 4.4.180-94.100 (SLES12 SP3)
14 4.4.0-131 (Ubuntu 16.04)
15 4.15.0-32 (Ubuntu 18.04)
16 vanilla linux 4.6.7 (ZFS only)
17 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
18 * Client primary unpatched kernels built and tested during release cycle:
19 3.10.0-1062.9.1.el7 (RHEL7.7)
20 3.10.0-1127.el7 (RHEL7.8)
21 4.12.14-95.45 (SLES12 SP4)
22 4.12.14-197.29 (SLES15 SP1)
23 4.15.0-48 (Ubuntu 18.04)
24 * Other clients known to build on unpatched kernels at some point (others may also work):
25 3.10.0-862.14.4.el7 (RHEL7.5)
26 3.10.0-957.27.2.el7 (RHEL7.6)
27 4.14.0-49.13.1.el7a (RHEL7.5)
28 4.14.0-115.2.2.el7a (RHEL7.6)
29 4.18.0-80.11.2.el8 (RHEL8.0)
30 4.18.0-147.3.1.el8 (RHEL8.1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.180-94.100 (SLES12 SP3)
33 4.4.0-131 (Ubuntu 16.04)
34 vanilla linux 4.15.0 (ZFS only)
35 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
36 * Recommended ZFS version: 0.8.3
37 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
38 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
39 stack overflow. For more information, please refer to bugzilla 17630.
40 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
41 the re-exporting nodes to be mounted with "32bitapi" mount option
42 * Note for Ubuntu 14.04.5:
43 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
44 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
46 --------------------------------------------------------------------------------
50 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
51 for currently supported client and server kernel versions.
52 * Server primary kernels built and tested during release cycle:
53 3.10.0-1062.1.1.el7 (RHEL7.7)
54 * Other server kernels known to build and work at some point (others may also work):
55 3.10.0-862.14.4.el7 (RHEL7.5)
56 3.10.0-957.27.2.el7 (RHEL7.6)
57 4.18.0-80.7.1.el8 (RHEL8.0)
58 4.4.120-92.70 (SLES12 SP2)
59 4.4.180-94.100 (SLES12 SP3)
60 4.4.0-131 (Ubuntu 16.04)
61 4.15.0-32 (Ubuntu 18.04)
62 vanilla linux 4.6.7 (ZFS only)
63 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
64 * Client primary unpatched kernels built and tested during release cycle:
65 3.10.0-1062.1.1.el7 (RHEL7.7)
66 4.12.14-95.29 (SLES12 SP4)
67 4.15.0-48 (Ubuntu 18.04)
68 * Other clients known to build on unpatched kernels at some point (others may also work):
69 3.10.0-862.14.4.el7 (RHEL7.5)
70 3.10.0-957.27.2.el7 (RHEL7.6)
71 4.14.0-49.13.1.el7a (RHEL7.5)
72 4.14.0-115.2.2.el7a (RHEL7.6)
73 4.18.0-80.7.1.el8 (RHEL8.0)
74 4.4.120-92.70 (SLES12 SP2)
75 4.4.180-94.100 (SLES12 SP3)
76 4.4.0-131 (Ubuntu 16.04)
77 vanilla linux 4.15.0 (ZFS only)
78 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
79 * Recommended ZFS / SPL version: 0.7.13
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
85 * Note for Ubuntu 14.04.5:
86 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
87 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
89 --------------------------------------------------------------------------------
93 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
94 for currently supported client and server kernel versions.
95 * Server known to build on the following kernels (others may also work):
96 2.6.32-696.18.7.el6 (RHEL6.9)
97 3.10.0-862.14.4.el7 (RHEL7.5)
98 3.10.0-957.el7 (RHEL7.6)
99 4.4.120-92.70 (SLES12 SP2)
100 4.4.162-94.69 (SLES12 SP3)
101 4.4.0-131 (Ubuntu 16.04)
102 4.15.0-32 (Ubuntu 18.04)
103 vanilla linux 4.6.7 (ZFS only)
104 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
105 * Client known to build on unpatched kernels (others may also work):
106 2.6.32-696.18.7.el6 (RHEL6.9)
107 2.6.32-754.6.3.el6 (RHEL6.10)
108 3.10.0-862.14.4.el7 (RHEL7.5)
109 3.10.0-957.el7 (RHEL7.6)
110 4.14.0-49.13.1.el7a (RHEL7.5)
111 4.14.0-115.2.2.el7a (RHEL7.6)
112 4.4.120-92.70 (SLES12 SP2)
113 4.4.162-94.69 (SLES12 SP3)
114 4.4.0-131 (Ubuntu 16.04)
115 4.15.0-32 (Ubuntu 18.04)
116 vanilla linux 4.15.0 (ZFS only)
117 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
118 * Recommended ZFS / SPL version: 0.7.9
119 * Tested with ZFS / SPL version: 0.7.9
120 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
121 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
122 stack overflow. For more information, please refer to bugzilla 17630.
123 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
124 the re-exporting nodes to be mounted with "32bitapi" mount option
125 * Note for Ubuntu 14.04.5:
126 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
127 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
129 --------------------------------------------------------------------------------
131 2018-04-03 Intel Corporation
133 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
134 for currently supported client and server kernel versions.
135 * Server known to build on patched kernels:
136 2.6.32-431.29.2.el6 (RHEL6.5)
137 2.6.32-504.30.3.el6 (RHEL6.6)
138 2.6.32-573.26.1.el6 (RHEL6.7)
139 2.6.32-642.15.1.el6 (RHEL6.8)
140 2.6.32-696.18.7.el6 (RHEL6.9)
141 3.10.0-693.21.1.el7 (RHEL7.4)
142 3.0.101-0.47.71 (SLES11 SP3)
143 3.0.101-107 (SLES11 SP4)
144 3.12.74-60.64.40 (SLES12 SP1)
145 4.4.114-92.64 (SLES12 SP2)
146 4.4.114-94.11 (SLES12 SP3)
147 3.13.0-101 (Ubuntu 14.04, ZFS only)
148 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
149 4.4.0-116 (Ubuntu 16.04)
150 vanilla linux 4.6.7 (ZFS only)
151 * Client known to build on unpatched kernels:
152 2.6.32-431.29.2.el6 (RHEL6.5)
153 2.6.32-504.30.3.el6 (RHEL6.6)
154 2.6.32-573.26.1.el6 (RHEL6.7)
155 2.6.32-642.15.1.el6 (RHEL6.8)
156 2.6.32-696.18.7.el6 (RHEL6.9)
157 3.10.0-693.21.1.el7 (RHEL7.4)
158 3.0.101-0.47.71 (SLES11 SP3)
159 3.0.101-107 (SLES11 SP4)
160 3.12.74-60.64.40 (SLES12 SP1)
161 4.4.114-92.64 (SLES12 SP2)
162 4.4.114-94.11 (SLES12 SP3)
163 3.13.0-101 (Ubuntu 14.04)
164 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
165 4.4.0-116 (Ubuntu 16.04)
167 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
168 * Recommended ZFS / SPL version: 0.7.6
169 * Tested with ZFS / SPL version: 0.7.6
170 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
171 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
172 stack overflow. For more information, please refer to bugzilla 17630.
173 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
174 the re-exporting nodes to be mounted with "32bitapi" mount option
175 * Note for Ubuntu 14.04.5:
176 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
177 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
179 --------------------------------------------------------------------------------
181 2017-07-05 Intel Corporation
183 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
184 for currently supported client and server kernel versions.
185 * Server known to build on patched kernels:
186 2.6.32-431.29.2.el6 (RHEL6.5)
187 2.6.32-504.30.3.el6 (RHEL6.6)
188 2.6.32-573.26.1.el6 (RHEL6.7)
189 2.6.32-642.15.1.el6 (RHEL6.8)
190 2.6.32-696.3.1.el6 (RHEL6.9)
191 3.10.0-514.21.1.el7 (RHEL7.3)
192 3.0.101-0.47.71 (SLES11 SP3)
193 3.0.101-100 (SLES11 SP4)
194 3.12.74-60.64.40 (SLES12 SP1)
195 4.4.59-92.17 (SLES12 SP2)
196 vanilla linux 4.6.7 (ZFS only)
197 * Client known to build on unpatched kernels:
198 2.6.32-431.29.2.el6 (RHEL6.5)
199 2.6.32-504.30.3.el6 (RHEL6.6)
200 2.6.32-573.26.1.el6 (RHEL6.7)
201 2.6.32-642.15.1.el6 (RHEL6.8)
202 2.6.32-696.3.1.el6 (RHEL6.9)
203 3.10.0-514.21.1.el7 (RHEL7.3)
204 3.0.101-0.47.71 (SLES11 SP3)
205 3.0.101-100 (SLES11 SP4)
206 3.12.74-60.64.40 (SLES12 SP1)
207 4.4.59-92.17 (SLES12 SP2)
209 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
210 * Recommended ZFS / SPL version: 0.6.5.9
211 * Tested with ZFS / SPL version: 0.6.5.9
212 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
213 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
214 stack overflow. For more information, please refer to bugzilla 17630.
215 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
216 the re-exporting nodes to be mounted with "32bitapi" mount option
218 --------------------------------------------------------------------------------
220 12-07-2016 Intel Corporation
222 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
223 for currently supported client and server kernel versions.
224 * Server known to build on patched kernels:
225 2.6.32-431.29.2.el6 (RHEL6.5)
226 2.6.32-504.30.3.el6 (RHEL6.6)
227 2.6.32-573.26.1.el6 (RHEL6.7)
228 2.6.32-642.11.1.el6 (RHEL6.8)
229 3.10.0-514.el7 (RHEL7.3)
230 3.0.101-0.47.71 (SLES11 SP3)
231 3.0.101-80 (SLES11 SP4)
232 3.12.62-60.64.8 (SLES12 SP1)
233 vanilla linux 4.5.7 (ZFS only)
234 * Client known to build on unpatched kernels:
235 2.6.32-431.29.2.el6 (RHEL6.5)
236 2.6.32-504.30.3.el6 (RHEL6.6)
237 2.6.32-573.26.1.el6 (RHEL6.7)
238 2.6.32-642.11.1.el6 (RHEL6.8)
239 3.10.0-514.el7 (RHEL7.3)
240 3.0.101-0.47.71 (SLES11 SP3)
241 3.0.101-80 (SLES11 SP4)
242 3.12.62-60.64.8 (SLES12 SP1)
244 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
245 * Recommended ZFS / SPL version: 0.6.5.8
246 * Tested with ZFS / SPL version: 0.6.5.8
247 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
248 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
249 stack overflow. For more information, please refer to bugzilla 17630.
250 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
251 the re-exporting nodes to be mounted with "32bitapi" mount option
253 --------------------------------------------------------------------------------
255 02-29-2016 Intel Corporation
257 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
258 for currently supported client and server kernel versions.
259 * Server known to build on patched kernels:
260 2.6.32-431.29.2.el6 (RHEL6.5)
261 2.6.32-504.30.3.el6 (RHEL6.6)
262 2.6.32-573.12.1.el6 (RHEL6.7)
263 3.10.0-327.3.1.el7 (RHEL7.2)
264 3.0.101-0.47.71 (SLES11 SP3)
265 3.0.101-68 (SLES11 SP4)
266 vanilla linux 4.2.1 (ZFS only)
267 * Client known to build on unpatched kernels:
268 2.6.32-431.29.2.el6 (RHEL6.5)
269 2.6.32-504.30.3.el6 (RHEL6.6)
270 2.6.32-573.12.1.el6 (RHEL6.7)
271 3.10.0-327.3.1.el7 (RHEL7.2)
272 3.0.101-0.47.71 (SLES11 SP3)
273 3.0.101-68 (SLES11 SP4)
276 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
277 * Recommended ZFS / SPL version: 0.6.4.2
278 * Tested with ZFS / SPL version: 0.6.4.2
279 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
280 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
281 stack overflow. For more information, please refer to bugzilla 17630.
282 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
283 the re-exporting nodes to be mounted with "32bitapi" mount option
285 --------------------------------------------------------------------------------
287 03-10-2015 Intel Corporation
289 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
290 for currently supported client and server kernel versions.
291 * Server known to build on patched kernels:
292 2.6.32-431.29.2.el6 (RHEL6.5)
293 2.6.32-504.8.1.el6 (RHEL6.6)
294 3.0.101-0.46 (SLES11 SP3)
295 * Client known to build on unpatched kernels:
296 2.6.32-431.29.2.el6 (RHEL6.5)
297 2.6.32-504.8.1.el6 (RHEL6.6)
298 3.10.0-123.20.1.el7 (RHEL7)
299 3.0.101-0.46 (SLES11 SP3)
300 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
301 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
302 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
303 stack overflow. For more information, please refer to bugzilla 17630.
304 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
305 the re-exporting nodes to be mounted with "32bitapi" mount option
307 Severity : enhancement
309 Description: control OST-index in IDIF via ROCOMPAT flag.
310 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
311 last_rcvd file. For new formatted OST device, it will be auto set;
312 for the case of upgrading from old OST device, you can enable it
313 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
314 enabled, for new created OST-object, its IDIF-in-LMA will contain
315 the OST-index; for the existing OST-object, the OSD will convert
316 old format IDIF as new format IDIF with OST-index stored in the
317 LMA EA when accessing such OST-object or via OI scrub. Once such
318 flag is enabled, it cannot be reverted back, so the system cannot
319 be downgraded to the orignal incompatible version.
321 --------------------------------------------------------------------------------
323 07-30-2014 Intel Corporation
325 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
326 for currently supported client and server kernel versions.
327 * Server known to build on patched kernels:
328 2.6.32-431.20.3.el6 (RHEL6.5)
329 3.0.101-0.31 (SLES11 SP3)
330 * Client known to build on unpatched kernels:
331 2.6.32-431.20.3.el6 (RHEL6.5)
332 3.0.101-0.31 (SLES11 SP3)
335 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
336 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
337 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
338 stack overflow. For more information, please refer to bugzilla 17630.
339 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
340 the re-exporting nodes to be mounted with "32bitapi" mount option
342 --------------------------------------------------------------------------------
344 10-31-2013 Intel Corporation
346 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
347 * Server builds for kernels:
348 2.6.32-358.18.1.el6 (RHEL6.4)
349 * Client builds for unpatched kernels:
350 2.6.32-358.18.1.el6 (RHEL6.4)
351 3.0.92-0.8 (SLES11 SP3)
353 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
354 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
355 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
356 stack overflow. For more information, please refer to bugzilla 17630.
357 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
358 the re-exporting nodes to be mounted with "32bitapi" mount option
360 --------------------------------------------------------------------------------
362 05-30-2013 Intel Corporation
364 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
365 * Server builds for kernels:
366 2.6.32-358.6.2.el6 (RHEL6.4)
367 * Client builds for unpatched kernels:
368 2.6.32-358.6.2.el6 (RHEL6.4)
369 3.0.74-0.6.6 (SLES11 SP2)
371 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
372 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
373 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
374 stack overflow. For more information, please refer to bugzilla 17630.
375 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
376 the re-exporting nodes to be mounted with "32bitapi" mount option
378 --------------------------------------------------------------------------------
380 9-30-2012 Whamcloud, Inc.
382 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
383 * Server support for kernels:
384 2.6.32-279.5.1.el6 (RHEL6)
385 * Client support for unpatched kernels:
386 2.6.32-279.5.1.el6 (RHEL6)
388 2.6.18-238.19.1.el5 (RHEL5)
389 2.6.16.60-0.69.1 (SLES10)
390 * Recommended e2fsprogs version: 1.41.90.wc4
391 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
392 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
393 stack overflow. For more information, please refer to bugzilla 17630.
394 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
395 the re-exporting nodes to be mounted with "32bitapi" mount option
397 Severity : enhancement
399 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
401 Severity : enhancement
403 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
405 Severity : enhancement
407 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
409 Severity : enhancement
411 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
413 Severity : enhancement
415 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
417 --------------------------------------------------------------------------------
419 04-30-2012 Whamcloud, Inc.
421 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
422 * Server support for kernels:
423 2.6.32-220.4.2.el6 (RHEL6)
424 * Client support for unpatched kernels:
425 2.6.32-220.4.2.el6 (RHEL6)
427 2.6.18-238.19.1.el5 (RHEL5)
428 2.6.16.60-0.69.1 (SLES10)
429 * Recommended e2fsprogs version: 1.41.90.wc4
430 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
431 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
432 stack overflow. For more information, please refer to bugzilla 17630.
433 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
434 the re-exporting nodes to be mounted with "32bitapi" mount option
436 Severity : enhancement
438 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
440 Severity : enhancement
442 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
444 Severity : enhancement
446 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
448 Severity : enhancement
450 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
452 --------------------------------------------------------------------------------
454 09-30-2011 Whamcloud, Inc.
456 * Server support for kernels:
457 2.6.18-238.19.1.el5 (RHEL5)
458 2.6.32-131.6.1.el6 (RHEL6)
459 * Client support for unpatched kernels:
460 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
461 2.6.18-238.19.1.el5 (RHEL5)
462 2.6.32-131.6.1.el6 (RHEL6)
463 2.6.29.4-167.fc11 (FC11)
464 2.6.16.60-0.69.1 (SLES10)
466 * Recommended e2fsprogs version: 1.41.90.wc3
467 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
468 Lustre filesystem with 4K stack may cause a stack overflow. For more
469 information, please refer to bugzilla 17630.
470 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
471 mounted with "32bitapi" mount option
473 Severity : enhancement
474 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
476 Severity : enhancement
477 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
479 Severity : enhancement
480 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
483 Frequency : only with SLES10
485 Description: Use OFED "KMP" provided by Novell
486 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
487 Lustre is now built against this package. That means you need to
488 install the ofed-kmp package from Novell for the patchless client.
489 Note that the ofed-kmp that Novell ships may not exactly match the
490 kernel version but should still be compatible.
492 Severity : enhancement
494 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
495 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
497 --------------------------------------------------------------------------------
499 2010-07-15 Oracle, Inc.
501 * Server support for kernels:
502 2.6.18-164.11.1.el5 (RHEL 5)
503 2.6.18-164.11.1.0.1.el5 (OEL 5)
504 * Client support for unpatched kernels:
505 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
506 2.6.18-164.11.1.el5 (RHEL 5),
507 2.6.18-164.11.1.0.1.el5 (OEL 5)
508 2.6.16.60-0.42.8 (SLES 10),
510 2.6.29.4-167.fc11 (FC11)
511 * Recommended e2fsprogs version: 1.41.10-sun2
513 Severity : enhancement
515 Description: Distribute lfs_migrate script with lustre
516 Details : The lfs_migrate mentioned in the manual is now included in the
517 lustre tree in lustre/scripts and will distributed in lustre RPMs.
519 Severity : enhancement
521 Description: Failover nid list added to import proc file.
523 Severity : enhancement
525 Description: "som_preview" mount option required to use SOM
526 Details : The clients will need to use som_preview mount option
530 Description: Lustre does not do 1MB IOs to HW RAID
531 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
533 Severity : enhancement
535 Description: lfs quota output is not very convenient for awk/sed-parsing
536 Details : Some positions in lfs quota output table could be empty or
537 non-empty which made it hard to parse it with scripts, now
538 a dash is put instead of space where there is not supposed
543 Description: Update to OFED 1.5.1
544 Details : OFED 1.5.1 will be used in environments that don't use
545 vendor supplied OFED.
549 Description: Remove lnet. prefix from lctl params display
550 Details : Also renamed lnet.memused to lnet_memused to
555 Description: Use conf_param -d to remove permanant settings.
559 Description: Add list_param -R
560 Details : lctl list_param now supports recursive mode.
563 Bugzilla : 21829, 22312
564 Description: Fix llobdstat.
565 Details : 1. Now updating the write_bytes counter
566 2. Port script fixes from 1.8.
570 Description: decrease the usage of memory on clients.
571 Details : 1. On clients, recycle dentries and inodes unused.
572 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
577 Description: llobdstat fix and enhancement
578 Details : add a counter to set a limit to how many samples will be returned
579 fix a wildcard in the path to limit to obdfilter stats only
583 Description: lfs check only allowed for root
584 Details : Code cleanups and added a regresion test of non-root use of
589 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
590 Details : Fix hr->hr_index overload.
594 Description: Fix FIEMAP functionality.
598 Description: FC11 patchless client support added
602 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
603 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
607 Description: New acc-sm script and various test-framework enhancments.
611 Description: Previously landed patch was reverted.
615 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
619 Description: Add ldiskfs maxdirsize mount option.
623 Description: Fix mkfs.lustre support for 16TB LUNs.
627 Description: Print mmp_check_interval
628 Details : print mmp_check_interval and make it possible to abort
629 mount operation in case it takes too long.
633 Description: Add cascading_rw to lustre mpi tests.
637 Description: Store SOM data in EA.
638 Details : Also several SOM related cleanups.
642 Description: Add refcount to md_open_data
646 Description: Use refcount on lov_request_set to prevent memory corruption.
650 Description: Statahead: use parent dir's i_mutex for synchronization.
654 Description: loadgen improvements.
658 Description: Enable peer health detection in ksocklnd and ko2iblnd.
662 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
666 Description: Fix race on obd_devcie:md_stats
670 Description: Fix SLES 11 client builds.
671 Details : Resolved uninitalized pointer gcc warnings.
675 Description: Allow non-root access for "lfs check".
676 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
680 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
682 Severity : enhancement
684 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
685 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
689 Description: Sanity test_27p fails.
690 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
691 Sanity tests 27 were cleaned up and improved.
695 Description: Unify req format on client/servers
696 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
697 buffers and swabbers approach..
702 Description: drop join file stuff.
704 Severity : enhancement
706 Description: Add 2.6.27 and SLES11 (patchless client) support
710 Description: Fix in ptlrpc_expire_one_request() to print the signed time
713 Severity : enhancement
715 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
718 Severity : enhancement
720 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
722 Severity : enhancement
724 Description: correctly handle big reply message.
725 Details : send LNet event if reply is bigger than buffer and adjust this buffer
730 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
732 Severity : enhancement
734 Description: Add support for OEL5.
738 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
740 Severity : enhancement
742 Description: "weak-modules" support
743 Details : Implement "weak-modules" support which enables kernel modules
744 to be used with any kernel that implements the same kABI. In
745 order to achieve this modules are now installed in
746 /lib/modules/$(uname -r)/updates/kernel on all distributions.
750 Description: client could not reconnect to OST because of an active request.
751 Details : The client now retries to reconnect to the same server, if a
752 connect request failed with EBUSY or -EAGAIN. Test case 35b
753 conf-sanity.sh has been created.
757 Description: Don't allow make backward step on assignin osc next id.
758 Details : race between allocation next id and ll_sync thread can be cause
759 of set wrong osc next id and can be kill valid ost objects.
763 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
767 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
771 Description: Error handling in osc_statfs_interpret() has been improved.
772 Details : Check in osc_statfs_interpret() for EBADR.
776 Description: Do not send statfs() requests to OSTs disabled by administrator.
777 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
779 Severity : enhancement
781 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
783 Severity : enhancement
785 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
787 Severity : enhancement
789 Description: Enhancement for lfs(1) command to use numeric uid/gid.
793 Description: Improved cleanup procedure for conf-sanity test_46a.
797 Description: Do not start bulk transfer, if client tried to reconnect.
798 Details : Do not start bulk transfer, if client tried to reconnect,
799 but failed due to active rpcs. Check for exp_abort_active_req
800 has been implemented.
804 Description: Prevent inconsistences between linux and lustre mount structures.
805 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
806 Make the sleep interruptible.
808 Severity : enhancement
810 Description: provide server to client comms path
811 Details : server to client communications path via new LDLM_SET_INFO rpc
815 Description: truncate starts GFP_FS allocation under transaction causing deadlock
816 Details : ldiskfs_truncate calls grab_cache_page which may start page
817 allocation under an open transaction. This may lead to
818 calling prune_icache with consequent lustre reentrance.
822 Description: send correctly lsm on open replay
823 Details : MDS is trust to LSM size on replay open, but client can set wrong size
828 Description: enable client interface failover
829 Details : When a child reconnects from another NID, properly update export
830 nid hash position and ldlm reverse import.
832 Severity : enhancement
834 Description: Communicate OST degraded/readonly state via statfs to MDS
835 Details : Flags in the statfs returned from OSTs indicate whether the
836 OST is in a degraded RAID state, or if the filesystem has
837 turned read-only after a filesystem error is detected.
842 Description: don't panic if EPROTO was hit when reading symlink
843 Details : correctly handling request reference in error cases.
845 Severity : enhancement
847 Description: Add LustreNetLink, a kernel-userspace communcation path.
849 Severity : enhancement
851 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
856 Description: access to llog context before init.
857 Details : move handling CATALOGS file at osc layer and forbid access to llog
861 Frequency : rare, only when client is writing to OST on local node
863 Description: Avoid deadlock for local client writes
864 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
865 memory freeing context. This allows OST threads to set the
866 PF_MEMALLOC flag on task structures in order to allocate memory
867 from reserved pools and complete IO.
868 Use GFP_HIGHUSER for OST allocations for non-local client writes,
869 so that the OST threads generate memory pressure and allow
870 inactive pages to be reclaimed.
872 Severity : enhancement
874 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
879 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
880 Details : move ldlm namespace creation in setup phase to avoid grab
881 _lprocfs_lock with cli_sem held.
885 Description: Temporarily disable grant shrink.
886 Details : Disable the feature for debugging.
889 Frequency : only when formatting test filesystems
891 Description: Unable to run several mkfs.lustre on loop devices at the same time
892 Details : mkfs.lustre returns error 256 on the concurrent loop devices
893 formatting. The solution is to proper handle the error.
895 Severity : enhancement
897 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
899 Severity : enhancement
901 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
906 Description: resolve race between obd_disconnect and class_disconnect_exports
907 Details : if obd_disconnect will be called to already disconnected export he
908 forget release one reference and osc module can't unloaded.
910 Severity : enhancement
912 Description: Allow tuning service thread via /proc
913 Details : For each service a new
914 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
915 created that can be used to set min/max thread counts, and get the
916 current number of running threads.
921 Descriptoin: don't return error if have partially created objects for file.
922 Details : lov_update_create_set uses set->set_success as index for created
923 objects, so if some requests will be failed, they will have hole
924 at end of array and qos_shrink_lsm for allocate correct lsm.
926 Severity : enhancement
928 Description: Update OFED support to 1.4.1
933 Description: Reduce small size read RPC
934 Details : Set read-ahead limit for every file and only do read-ahead when
935 available read-ahead pages are bigger than 1M to avoid small size
938 Severity : enhancement
940 Description: move AT tunable parameters for more consistent usage
941 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
943 Severity : enhancement
945 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
946 Details : allow skip disconnected ost for send statfs request and hide error
952 Description: fix racy locking of mballoc block bitmaps causing BUG
953 Details : The locking of the mballoc buddy bitmap and the in-memory
954 block bitmap was using two different spin locks in some
955 cases. This made it possible to incorrectly access the
956 mballoc bitmap while another process was modifying it,
957 causing a sanity assertion to fail. While no on-disk corruption
958 was reported, there was some risk of this happening.
961 Frequency : rare, on llog test 6
963 Descriptoin: don't allow connect to already connected import
964 Details : allowing connect to already connected import is hide connecting problem.
969 Descriptoin: allow kill process which wait statahead result
970 Details : for some reasons 'ls' can stick in waiting result from statahead,
971 in this case need way for kill this process.
973 Severity : enhancement
975 Description: Add state history info file, enhance import info file
976 Details : Track import connection state changes in a new osc/mdc proc file;
977 add overview-type data to the osc/mdc import proc file.
979 Severity : enhancement
981 Description: MDS create should not wait for statfs RPC while holding DLM lock.
983 Severity : enhancement
985 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
990 Descriptoin: OST grant too much space to client even there are not enough space.
991 Details : Client will shrink its grant cache to OST if there are no write
992 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
993 this grant cache if there are already not enough avaible space
994 (left_space < total_clients * 32M).
997 Frequency : start MDS on uncleanly shutdowned MDS device
999 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1000 Details : stay in waiting mds<>ost recovery finished produce random bugs
1001 due race between two ll_sync thread for one lov target. send
1002 ACTIVATE event only if connect realy finished and import have
1006 Frequency : rare, connect and disconnect target at same time
1008 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1009 Details : don't call obd_disconnect under lov_lock. this long time
1010 operation and can block ptlrpcd which answer to connect request.
1015 Descriptoin: don't lose wakeup for imp_recovery_waitq
1016 Details : recover_import_no_retry or invalidate_import and import_close can
1017 both sleep on imp_recovery_waitq, but we was send only one wakeup
1021 Frequency : always with long access acl
1023 Descriptoin: mds can't pack reply with long acl.
1024 Details : mds don't control size of acl but they limited by reint/getattr
1027 Severity : enhancement
1029 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1031 Severity : enhancement
1033 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1036 Frequency : start MDS on uncleanly shutdowned MDS device
1038 Descriptoin: aborting recovery hang on MDS
1039 Details : don't throttle destroy RPCs for the MDT.
1042 Frequency : on remount
1044 Description: external journal device not working after the remount
1045 Details : clear dev_rdonly flag for external journal devices in
1051 Description: shutdown vs evict race
1052 Details : client_disconnect_export vs connect request race.
1053 if client will evicted at this time - we start invalidate
1054 thread without referece to import and import can be freed
1058 Frequency : rare, need acl's on inode.
1060 Description: client can't handle ost additional correctly
1061 Details : if ost was added after client connected to mds client can have
1062 hit lnet_try_match_md ... to big messages to wide striped files.
1063 in this case need teach client to handle config events about add
1064 lov target and update client max ea size at that event.
1066 Severity : enhancement
1068 Description: Changelogs
1069 Details : Changelogs are a lightweight mechanism to track filesystem
1070 metadata and namespace changes. The changelog is recorded
1071 permanently on the MDTs, and is periodically "consumed" / purged
1072 when records are no longer needed.
1074 Severity : enhancement
1076 Description: compact fld format with extents
1077 Details : Store range of seq rather than every seq in FLD. Seq
1078 controller update FLD rather than clients. In Case of CMD, mdt0
1079 has FLD, all other metadata server act as non persistent proxy
1080 for FLD queries and cache fld entries in fld cache.
1085 Description: don't skip ost target if they assigned to file
1086 Details : Drop slow OSCs if we can, but not for requested start idx.
1087 This means "if OSC is slow and it is not the requested
1088 start OST, then it can be skipped, otherwise skip it only
1089 if it is inactive/recovering/out-of-space.
1093 Description: more cleanup in mds_lov
1094 Details : not send LOV EA under replay, we can't know about they size at this
1095 time. Don't allow client connect to mds before any ost connected,
1096 for avoid problems with LOV EA size and returning EIO to client.
1098 Severity : enhancement
1100 Description: Interoperability at server side (Disk interoperability)
1102 Severity : enhancement
1104 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1106 Severity : enhancement
1108 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1110 Severity : enhancement
1112 Description: New client IO stack (CLIO).
1114 Severity : enhancement
1116 Description: Commit on sharing. Eliminate inter-client dependencies between
1117 uncommitted transactions by doing transaction commits.
1118 Thereby clients may recovery independently.
1121 Frequency : Create a symlink file with a very long name
1123 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1124 Details : If there is no extra space in the request for early cancels,
1125 ldlm_req_handles_avail() returns 0 instead of a negative value.
1127 Severity : enhancement
1129 Description: Add /proc entry for import status
1130 Details : The mdc, osc, and mgc import directories now have
1131 an import directory that contains useful import data for debugging
1132 connection problems.
1134 Severity : enhancement
1136 Description: Re-disable certain /proc logging
1137 Details : Enable and disable client's offset_stats, extents_stats and
1138 extents_stats_per_process stats logging on the fly.
1141 Frequency : Only on FC kernels 2.6.22+
1143 Description: oops in statahead
1144 Details : Do not drop reference count for the dentry from VFS when lookup,
1145 VFS will do that by itself.
1147 Severity : enhancement
1149 Description: Generic /proc file permissions
1150 Details : Set /Proc file permissions in a more generic way to enable non-
1151 root users operate on some /proc files.
1155 Description: Hitting mdc_commit_close() ASSERTION
1156 Details : Properly handle request reference release in
1157 ll_release_openhandle().
1161 Description: quota recovery deadlock during mds failover
1162 Details : This patch includes att18982, att18236, att18237 in bz14840.
1164 1. fix osts hang when mds does failover with quotaon
1165 2. prevent watchdog storm when osts threads wait for the
1170 Frequency : only patchless client
1171 Description: add workaround for race between add/remove dentry from hash
1173 Severity : enhancement
1175 Description: Allow OST glimpses to return PW locks
1179 Description: LBUG when llog conf file is full
1180 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1184 Description: Prevent import from entering FULL state when server in recovery
1188 Description: service mount cannot take device name with ":"
1189 Details : Only when device name contains ":/" will mount treat it as
1195 Description: replace ptlrpcd with the statahead thread to interpret the async
1196 statahead RPC callback
1200 Frequency : on recovery
1201 Description: I/O failures after umount during fail back
1202 Details : if client reconnected to restarted server we need join to recovery
1203 instead of find server handler is changed and process self
1204 eviction with cancel all locks.
1206 Severity : enhancement
1208 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1210 Severity : enhancement
1212 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1214 Severity : enhancement
1216 Description: Upcall on Lustre log has been dumped
1217 Details : Allow for a user mode script to be called once a Lustre log has
1218 been dumped. It passes the filename of the dumped log to the
1219 script, the location of the script can be specified via
1220 /proc/sys/lnet/debug_log_upcall.
1225 Description: avoid idr_remove called for id which is not allocated.
1226 Details : Move assigment s_dev for clustered nfs to end of initialization,
1227 to avoid problem with error handling.
1232 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1233 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1234 into CONN_USED_HASH and this prodice warning when put connection
1235 again in unused hash.
1238 Severity : enhancement
1240 Description: File striping can now be set to use an arbitrary pool of OSTs.
1242 Severity : enhancement
1244 Description: Export bytes_read/bytes_write count on OSC/OST.
1248 Description: Early reply size mismatch, MGC loses connection
1249 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1250 the connect flags are properly negotiated.
1255 Description: quotas are not honored with O_DIRECT
1256 Details : all writes with the flag O_DIRECT will use grants which leads to
1257 this problem. Now using OBD_BRW_SYNC to guard this.
1261 Description: add quota statistics
1262 Details : 1. sort out quota proc entries and proc code.
1263 2. add quota statistics
1265 Severity : enhancement
1267 Description: enable quota support for HEAD.
1271 Description: Properly propagate oinfo flags from lov to osc for statfs
1272 Details : restore missing copy oi_flags to lov requests.
1274 Severity : enhancement
1276 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1277 : l_getgroups(8), lst(8), routerstat(8)
1278 Details : included man pages for llobdstat(8), llstat(8),
1279 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1281 Severity : enhancement
1283 Description: configure's --enable-quota should check the
1284 : kernel .config for CONFIG_QUOTA
1285 Details : configure is terminated if --enable-quota is passed but
1286 : no quota support is in kernel
1290 Description: Remove portals compatibility
1291 Details : Remove portals compatibility, not interoperable with releases
1296 Description: Resolve device initialization race
1297 Details : Prevent proc handler from accessing devices added to the
1298 obd_devs array but yet be intialized.
1300 Severity : enhancement
1302 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1304 Severity : enhancement
1306 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1311 Description: Using wrong pointer in osc_brw_prep_request
1312 Details : Access to array[-1] can produce panic if kernel compiled with
1313 CONFIG_PAGE_ALLOC enabled
1317 Description: Client runs out of low memory
1318 Details : Consider only lowmem when counting initial number of llap pages
1322 Description: *optional* service tags registration
1323 Details : if the "service tags" package is installed on a Lustre node
1324 When the filesystem is mounted, a local-node service tag will
1325 be created. See http://inventory.sun.com/ for more information
1326 about the Service Tags asset management system.
1330 Description: Kernel BUG tries to release flock
1331 Details : Lustre does not destroy flock lock before last reference goes
1332 away. So always drop flock locks when client is evicted and
1333 perform unlock regardless of successfulness of speaking to MDS.
1337 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1341 Description: sanity test 65a fails if stripecount of -1 is set
1342 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1347 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1348 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1351 Severity : enhancement
1353 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1357 Description: do not process already freed flock
1358 Details : flock can possibly be freed by another thread before it reaches
1359 to ldlm_flock_completion_ast.
1363 Description: LBUG during stress test
1364 Details : Need properly lock accesses the flock deadlock detection list.
1368 Description: oops in page fault handler
1369 Details : kernel page fault handler can return two special 'pages' in error
1370 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1374 Description: timeout with invalidate import.
1375 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1376 should be handled by ptlrpcd. This produce long age waiting and
1377 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1379 Severity : enhancement
1381 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1385 Description: enable MGS and MDT services start separately
1386 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1387 a co-located MGS without starting the MGS, which is a complement
1388 to 'nosvc' mount option.
1392 Frequency : after recovery
1393 Description: precreate to many object's after del orphan.
1394 Details : del orphan st in oscc last_id == next_id and this triger growing
1395 count of precreated objects. Set flag LOW to skip increase count
1396 of precreated objects.
1400 Frequency : rare, on clear nid stats
1401 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1402 Details : when clean nid stats sometimes try destroy live entry,
1403 and this produce panic in free.
1407 Description: Stack overflow during MDS log replay
1408 ease stack pressure by using a thread dealing llog_process.
1412 Description: wait until IO finished before start new when do lock cancel.
1413 Details : VM protocol want old IO finished before start new, in this case
1414 need wait until PG_writeback is cleared until check dirty flag and
1415 call writepages in lock cancel callback.
1417 Severity : enhancement
1419 Description: using special macro for print time and cleanup in includes.
1423 Description: mds_mfd_close() ASSERTION(rc == 0)
1424 Details : In mds_mfd_close(), we need protect inode's writecount change
1425 within its orphan write semaphore to prevent possible races.
1429 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1433 Frequency : rare, on shutdown ost
1434 Description: don't hit live lock with umount ost.
1435 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1436 use shrink_dcache_sb instead.
1440 Description: don't panic with use echo client
1441 Details : echo client pass NULL as client nid pointer and this produce null
1442 pointer dereference.
1446 Description: fix build on ppc32
1447 Details : compile code with -m64 flag produce wrong object file for ppc32.
1451 Description: add message levels for liblustreapi
1455 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1456 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1457 will result in an -EAGAIN and in the exit call path we call
1458 alloc_rr() which will with increasing aggressiveness attempt to
1459 aquire precreated objects on the minimum number of required OSCs.
1463 Description: Use old size assignment to avoid deadlock
1464 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1465 the scheduling while holding a spinlock. We do not need locking
1466 for size in ll_update_inode() because size is only updated from
1467 the MDS for directories or files without objects, so there is no
1468 other place to do the update, and concurrent access to such inodes
1469 are protected by the inode lock.
1473 Description: resolve "_IOWR redefined" build error on SLES10
1477 Description: dump the memory debugging after all modules are unloaded to
1478 suppress false negative in conf_sanity test 39
1480 Severity : enhancement
1482 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1487 Description: fix SLES kernel versioning
1488 Details : the kernel version for our SLES 10 kernel did not include a "-"
1489 before the "smp" at the end. while this was not a problem in
1490 general, it did mean that software trying to use the kernel
1491 version to try to detect a vendor specific kernel would fail.
1492 this was most evident by the OFED build scripts.
1496 Description: Don't update lov_desc members until making sure they are valid
1497 Details : When updating lov_desc members via proc fs, need fix their
1498 validities before doing the real update.
1502 Description: don't put request into delay list while invalidate in flight.
1503 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1504 import in flight. this produce timeout for invalidate and sometimes
1505 can cause stale data.
1509 Frequency : on ppc only
1510 Description: not convert ost objects for directory because it's not exist.
1511 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1515 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1517 Severity : enhancement
1519 Description: Update to support 2.6.22.14 vanilla kernel.
1523 Frequency : rare, on recovery
1524 Description: read procfs can produce deadlock in some situation
1525 Details : Holding lprocfs lock which send rpc can produce block for destroy
1526 obd objects and this also block reconnect with -EALREADY. This
1527 isn't fix all lprocfs bugs - but make it rare.
1529 Severity : enhancement
1531 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1534 Frequency : frequent on X2 node
1536 Description: mdc_set_open_replay_data LBUG
1537 Details : Set replay data for requests that are eligible for replay.
1541 Description: lustre_mgs: operation 101 on unconnected MGS
1542 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1543 MGC, and late on MGC cannot successfully connect to MGS and a lot
1544 of the error messages complaining that MGS is not connected.
1548 Frequency : on network error
1549 Description: panic with double free request if network error
1550 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1551 but it's true only for synchronus enqueue, for async enqueue
1552 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1555 Severity : enhancement
1557 Description: client-side metadata stat-ahead during readdir(directory readahead)
1558 Details : perform client-side metadata stat-ahead when the client detects
1559 readdir and sequential stat of dir entries therein
1562 Frequency : on start mds
1564 Description: Implement get_info(last_id) in obdfilter.
1567 Frequency : occasional
1569 Description: Correctly check stale fid, not start epoch if ost not support SOM
1570 Details : open with flag O_CREATE need set old fid in op_fid3 because
1571 op_fid2 was overwritten with new generated fid, but mds can answer
1572 with one of these two fids and both is not stale. Setattr
1573 incorrectly started an epoch and assume will be called
1574 done_writing, but without SOM done_writing ever being called.
1577 Frequency : rare, depends on device drivers and load
1579 Description: MDS or OSS nodes crash due to stack overflow
1580 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1581 In some cases, in conjunction with device drivers that use a lot
1582 of stack the MDS (or possibly OSS) service threads could overflow
1583 the stack. One change which was identified to consume additional
1584 stack has been reworked to avoid the extra stack usage.
1587 Frequency : occasional
1589 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1590 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1591 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1592 oscc_has_objects and see OSC has no objests and call
1593 oscc_internal_create to resend create request.
1595 Severity : enhancement
1597 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1599 Severity : enhancement
1601 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1604 Frequency : very rare
1606 Description: Fix replay if there is an un-replied request and open
1607 Details : In some cases, older replay request will revert the
1608 mcd->mcd_last_xid on MDS which is used to record the client's
1609 latest sent request.
1611 Severity : enhancement
1613 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1615 Severity : enhancement
1617 Description: Add rhel5 support to HEAD.
1619 Serverity : enhancement
1621 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1626 Description: Don't allow skipping OSTs if index has been specified.
1627 Details : Don't allow skipping OSTs if index has been specified, make
1628 locking in internal create lots better.
1632 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1633 Details : access to bitfield in structure is always rounded to long
1634 and this produce problem with not atomic change any bit.
1638 Description: Lustre make rpms failed.
1639 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1640 builds Lustre rpms from tarball.
1643 Frequency : rare on shutdown ost
1645 Description: If llog cancel was not send before clean_exports phase, this can
1646 produce deadlock in llog code.
1647 Details : If llog thread has last reference to obd and call class_import_put
1648 this produce deadlock because llog_cleanup_commit_master wait when
1649 last llog_commit_thread exited, but this never success because was
1650 called from llog_commit_thread.
1654 Description: allow userland application know is lost one of stripes.
1655 Details : fill lvb_blocks with error code on ost and return it to
1656 application if error flag found.
1660 Description: NULL lov_tgts causing MDS oops
1661 Details : more safe checks for NULL lov_tgts for avoid oops.
1663 Severity : enhancement
1665 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1669 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1670 Details : If an OST is down the MDS will hang indefinitely in
1671 obd_statfs_rqset() waiting for the statfs data. While for
1672 MDS QOS usage of statfs, it should not stuck in waiting.
1674 Severity : enhancement
1676 Description: remote_acl support
1677 Details : Support ACL-based permission check for remote user.
1678 Support setfacl/getfacl for remote user with the utils
1679 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1680 the system "{s,g}etfacl" utils.
1682 Severity : enhancement
1684 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1686 Severity : enhancement
1688 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1692 Description: make mgs_setparam() handle fsname containing dash
1693 Details : fsname containing a dash does not work with lctl conf_param
1695 Severity : enhancement
1697 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1699 Severity : enhancement
1701 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1705 Frequency : rare, at shutdown
1706 Description: access already free / zero obd_namespace.
1707 Details : if client_disconnect_export was called without force flag set,
1708 and exist connect request in flight, this can produce access to
1709 NULL pointer (or already free pointer) when connect_interpret
1710 store ocd flags in obd_namespace.
1714 Frequency : only at startup
1715 Description: not alloc memory with spinlock held.
1716 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1717 if any spinlock held.
1719 Severity : enhancement
1721 Description: make lustre randomly fail allocating memory
1722 Details : Make lustre randomly failed allocating memory for testing purpose.
1724 Severity : enhancement
1726 Description: lost problems with lov objid file
1727 Details : Fixes some scability and access to not inited memory problems
1728 in work with lov objdid file.
1733 Description: lfs find does not continue on file error
1734 Details : Continue other files processing when a file/dir is absent.
1738 Description: Inconsistent usage of lustre_pack_reply()
1739 Details : Standardize the usage of lustre_pack_reply() such that it
1740 always generate a CERROR on failure.
1743 Frequency : occasional
1745 Description: MDS hang or stay in waiting lock
1746 Details : If client receive lock with CBPENDING flag ldlm need send lock
1747 cancel as separate rpc, to avoid situation when cancel request
1748 can't processed due all i/o threads stay in wait lock.
1752 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1756 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1760 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1764 Description: Update to RHEL4 latest kernel.
1766 Severity : enhancement
1768 Description: Build SLES10 patchless client fails
1769 Details : The configure was broken by run ./configure with
1770 --with-linux-obj=.... argument for patchless client. When the
1771 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1772 can't search header adequately. Use absolute path such as
1773 -I($LINUX)/include instead.
1777 Description: interrupt oig_wait produce painc on resend.
1778 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1779 private set, and this produce situation when rq_ptlrpcd_data not
1780 copyed to new allocated request and triggered LBUG on assert
1781 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1782 ptlrpcd set if request is changed and can be safety changed to use
1787 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1788 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1789 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1790 these macros and replace them with their difinitions instead.
1794 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1795 Details : Update which_patch & target file for SLES10 latest kernel.
1798 Description: Remove client patches from SLES 10 kernel.
1799 Details : This causes SLES 10 clients to behave as patchless clients
1800 even on a Lustre-patched (server) kernel.
1802 Severity : enhancement
1804 Description: self-adjustable client's lru lists
1805 Details : use adaptive algorithm for managing client cached locks lru
1806 lists according to current server load, other client's work
1807 pattern, memory activities, etc. Both, server and client
1808 side namespaces provide number of proc tunables for controlling
1811 Severity : enhancement
1813 Description: light-weight GSS support
1814 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1815 and reduce performance overhead.
1817 Severity : enhancement
1819 Description: Linux keyring support
1820 Details : Support using service of Linux keyring for Lustre GSS internal
1821 context refresh/cache mechanism.
1825 Description: Fix errors in lfs documentation
1826 Details : Fixes man pages
1830 Description: don't use GFP_* in generic Lustre code.
1831 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1835 Description: obdclass is limited by single OBD_ALLOC(idarray)
1836 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1840 Description: warnings with build patchless client with vanila 2.6.19 and up
1841 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1842 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1846 Frequency : Only for SLES
1847 Description: sanity_quota fail test_1
1848 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1849 file, which makes TSTID[2] inunique.
1853 Frequency : Only for SLES9
1854 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1855 Details : Update target/ChangeLog/which_patch .
1859 Frequency : Only for test_10 in sanity-quota.sh
1860 Description: a bug in quota test code
1861 Details : checking a test flag in a wrong place
1865 Frequency : Only for test_13 in sanity-quota.sh
1866 Description: a bug in quota test script
1867 Details : shouldn't check overfull allocated quota
1871 Description: jbd statistics
1872 Details : Port older jbd statistics patch for sles10
1874 Severity : enhancement
1876 Description: Add printing inode info into message about error in writepage.
1881 Description: reply_lock_interpret crash due to race with it and lock cancel.
1882 Details : Do not replay locks that are being cancelled. Do not reference
1883 locks by their address during replay, just by their handle.
1887 Frequency : When flocks are used.
1888 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1889 Details : Flock locks might destroy just granted lock if it could be merged
1890 with another existing flock, this is done in completion handler,
1891 so teach ldlm_cli_enquque_fini that this is a valid case for
1897 Description: Oops in read and write path when failing to allocate lock.
1898 Details : Check if lock allocation failed and return error back.
1902 Description: lstripe command fails for valid OST index
1903 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1904 instead of lov->desc.ld_active_tgt_count.
1908 Description: sanity.sh failed test 103
1909 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1914 Description: df doesn't work properly if diskfs blocksize != 4K
1915 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1919 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1920 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1922 Severity : enhancement
1924 Description: Adaptive timeouts
1925 Details : RPC timeouts adapt to changing server load and network
1926 conditions to reduce resend attempts and improve recovery time.
1930 Description: llapi_file_create() does not allow some changes
1931 Details : add llapi_file_open() that allows specifying the mode and
1932 open flags, and also returns an open file handle.
1936 Description: merge and cleanup kernel patches.
1937 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1941 Description: Add journal checksum support.(Kernel part)
1942 Details : The journal checksum feature adds two new flags i.e
1943 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1944 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1945 indicates that the commit block contains the checksum for
1946 the blocks described by the descriptor blocks. Now commit
1947 record can be sent to disk without waiting for descriptor
1948 blocks to be written to disk. This behavior is controlled
1949 using JBD2_FEATURE_ASYNC_COMMIT flag.
1953 Description: OSS needs mutliple precreate threads
1954 Details : Add ability to start more than one create thread per OSS.
1958 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1959 Details : execute lfs setstripe on client
1963 Description: mds_obd_create error creating tmp object
1964 Details : When the user sets quota on root, llog will be affected and can't
1965 create files and write files.
1968 Frequency : Always on ia64 patchless client, and possibly others.
1970 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1971 Details : This allows the patchless client to be loaded on architectures
1972 without this export.
1976 Description: RedHat Update kernel for RHEL5
1977 Details : Add the kernel config file for RHEL5.
1981 Description: RedHat Update kernel for RHEL5
1982 Details : Modify the kernel config file more closer RHEL5.
1986 Description: Build failure against Centos5 (RHEL5)
1987 Details : Define PAGE_SIZE when it isn't present.
1991 Description: test_7 on sanity-quota.sh can't be used on separate machines
1992 Details : add facet to handle it
1996 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1997 Details : with NFS, the anon dentry's parent was set to itself in
1998 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1999 corresponding dentry other than use rec->ur_name.
2001 Severity : enhancement
2003 Description: lfs setstripe enhancement
2004 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2008 Description: enable data checksumming by default
2009 Details : enable checksum by default, allow --disable-checksum
2010 configure option and "-o nochecksum" mount option. Checksums
2011 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2015 Description: lustre support for RHEL5
2016 Details : Add support for RHEL5.
2020 Description: Client eviction due to failover config
2021 Details : after a connection loss, the lustre client should attempt to
2022 reconnect to the last active server first before trying the
2023 other potential connections.
2027 Description: when mds and osts use different quota unit(32bit and 64bit),
2028 quota will be released repeatly.
2029 Details : void sending multiple quota reqs to mds, which will keep the
2030 status between the reqs.
2034 Description: osts not allocated evenly to files
2035 Details : change the condition to increase offset_idx
2038 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2040 Description: Data corruption for OSTs that are formatted larger than 2TB
2042 Details : When generating the bio request for lustre file writes the
2043 sector number would overflow a temporary variable before being
2044 used for the IO. The data reads correctly from Lustre (which
2045 will overflow in a similar manner) but other file data or
2046 filesystem metadata may be corrupted in some cases.
2050 Description: Tune the kernel for good SCSI performance.
2051 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2052 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2057 Description: rewrite ext2-derived code in obdclass/uuid.c
2058 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2059 from scratch preserving functionality.
2063 Description: "lfs find -obd UUID" prints directories
2064 Details : "lfs find -obd UUID" will return all directory names instead
2065 of just file names. It is incorrect because the directories
2066 do not reside on the OSTs.
2070 Description: use i_size_read and i_size_write in 2.6 port
2071 Details : replace inode->i_size access with i_size_read/write()
2074 Frequency : when removing large files
2076 Description: scheduling issue during removal of large Lustre files
2077 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2078 It causes scheduling issues when removing large files (17TB in the
2082 Frequency : only with liblustre clients on XT3
2084 Description: evictions taking too long
2085 Details : allow llrd to evict clients directly on OSTs
2088 Frequency : only on ppc
2090 Description: /proc/fs/lustre/devices broken on ppc
2091 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2092 We should make sure the type of 'index' is loff_t and then cast
2093 explicitly as needed below. Do not assign an explicitly cast
2097 Frequency : only for rhel5
2099 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2100 Details : Modify the target file & which_kernel.
2102 Severity : enhancement
2104 Description: omit set fsid for export NFS
2105 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2106 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2110 Frequency : Always, for kernels after 2.6.16
2111 Description: Fix warning idr_remove called for id=.. which is not allocated.
2112 Details : Last kernels save old s_dev before kill super and not allow
2113 to restore from callback - restore it before call kill_anon_super.
2117 Description: buffer overruns could theoretically occur
2118 Details : llapi_semantic_traverse() modifies the "path" argument by
2119 appending values to the end of the origin string, and a
2120 overrun may occur. Adding buffer overrun check in liblustreapi.
2124 Description: Fix error on 'ls .' at the top of the Lustre mount.
2125 Details : Don't revalidate dentry if it is a root dentry.
2129 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2130 Details : Modify vm-tunables-rhel4.patch.
2134 Description: Kernel config for 2.6.18-vanilla.
2135 Details : Modify targets/2.6-vanilla.target.in.
2136 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2137 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2138 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2139 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2143 Description: improve handling recoverable errors
2144 Details : if request processig with error which can be recoverable on server
2145 request should be resend, otherwise page released from cache and
2150 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2151 Details : Modify target file & which_patch.
2152 A flaw was found in the IA32 system call emulation provided
2153 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2154 value could be stored in the %RAX register, which could trigger an
2155 out-of-bounds system call table access. An untrusted local user
2156 could exploit this flaw to run code in the kernel
2157 (ie a root privilege escalation). (CVE-2007-4573).
2161 Description: O_DIRECT bypasses client statistics.
2162 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2163 read_ahead_stats were updated. copied the stats section from
2164 osc_send_oap_rpc() into async_internal().
2168 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2172 Description: change order of libsysio includes
2173 Details : '#include sysio.h' should always come before '#include xtio.h'
2177 Description: handle "serious error: objid * already exists" more gracefully
2178 Details : If LAST_ID value on disk is smaller than the objects existing in
2179 the O/0/d* directories, it indicates disk corruption and causes an
2180 LBUG(). If the object is 0-length, then we should use the existing
2181 object. This will help to avoid a full fsck in most cases.
2183 Severity : enhancement
2185 Description: adapt the lustre_config script to support the upgrade case
2186 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2187 server targets to 1.6 in parallel.
2191 Description: To avoid grant space > avaible space when the disk is almost
2192 full. Without this patch you might see the error "grant XXXX >
2193 available" or some LBUG about grant, when the disk is almost
2195 Details : In filter_check_grant, for non_grant cache write, we should
2196 check the left space by if (*left > ungranted + bytes), instead
2197 of (*left > ungranted), because only we are sure the left space
2198 is enough for another "bytes", then the ungrant space should be
2199 increase. In client, we should update cl_avail_grant only there
2200 is OBD_MD_FLGRANT in the reply.
2204 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2205 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2206 described in CVE-2007-4573. This problem could allow untrusted
2207 local users to gain root access.
2210 Frequency : when using O_DIRECT and quotas
2212 Description: Incorrect file ownership on O_DIRECT output files
2213 Details : block usage reported by 'lfs quota' does not take into account
2214 files that have been written with O_DIRECT.
2219 Description: touch file failed when fs is not full
2220 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2221 otherwise we can get ENOSP while fs is not full.
2225 Description: parallel lock callbacks
2226 Details : Instead of sending blocking and completion callbacks as separated
2227 requests, adding them to a set and sending in parallel.
2230 Frequency : only for Cray XT3
2231 Bugzilla : 12829/13455
2232 Description: Changing primary group doesn't change the group lustre assigns to
2234 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2235 overridden with the primary group provided by the group upcall,
2236 whereas the supplied fsgid can be trusted if it is in the list of
2237 supplementary groups returned by the group upcall.
2239 Severity : enhancement
2241 Description: Allow masking D_WARNING, D_ERROR messages from console
2242 Details : Console messages can now be disabled via lnet.printk.
2246 Description: User code with malformed file open parameter crashes client node
2247 Details : Before packing join_file req, all the related reference should be
2248 checked carefully in case some malformed flags cause fake
2249 join_file req on client.
2253 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2254 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2255 mode is equal to the granted mode. However, it should also check
2256 that LDLM_FL_CP_REQD is not set.
2260 Description: Heavy nfs access might result in deadlocks
2261 Details : After ELC code landed, it is now improper to enqueue any mds
2262 locks under och_sem, because enqueue might want to decide to
2263 cancel open locks for same inode we are holding och_sem for.
2267 Description: Client eviction while running blogbench
2268 Details : A lot of unlink operations with concurrent I/O can lead to a
2269 deadlock causing evictions. To address the problem, the number of
2270 oustanding OST_DESTROY requests is now throttled to
2271 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2272 callbacks are processed in priority.
2276 Description: enable ACLs on MDS by default
2277 Details : ACLs must be enabled on MDS by default.
2280 Frequency : PPC/PPC64 only
2282 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2283 Details : fix duplicated definitions between asm-ppc64/types.h and
2284 lustre_types.h on PPC.
2287 Frequency : PPC/PPC64 only
2289 Description: asm-ppc/segment.h does not exist
2290 Details : fix compile issue on PPC.
2294 Description: better handle error messages in extents code
2297 Frequency : RHEL4 only
2299 Description: mkfs is very slow on IA64/RHEL4
2300 Details : A performance regression has been discovered in the MPT Fusion
2301 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2302 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2303 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2305 Severity : enhancement
2307 Description: SNMP support enhancement
2308 Details : Adding total number of sampled request for an MDS node in snmp
2311 Severity : enhancement
2313 Description: Optimize ldlm waiting list processing for PR extent locks
2314 Details : When processing waiting list for read extent lock and meeting read
2315 lock that is same or wider to it that is not contended, skip
2316 processing rest of the list and immediatelly return current
2317 status of conflictness, since we are guaranteed there are no
2318 conflicting locks in the rest of the list.
2322 Description: Time out and refuse to reconnect
2323 Details : When the failover node is the primary node, it is possible
2324 to have two identical connections in imp_conn_list. We must
2325 compare not conn's pointers but NIDs, otherwise we can defeat
2326 connection throttling.
2330 Description: port llog fixes from b1_6 into HEAD
2331 Details : Port llog reference couting and some llog cleanups from b1_6
2332 (bug 10800) into HEAD, for protect from panic and access to already
2333 free llog structures.
2337 Description: Detect stride IO mode in read-ahead
2338 Details : When a client does stride read, read-ahead should detect that and
2339 read-ahead pages according to the detected stride pattern.
2343 Description: data checksumming impacts single node performance
2344 Details : add support for several checksum algorithm. Currently, only CRC32
2345 and Adler-32 are supported. The checksum type can be changed on
2346 the fly via /proc/fs/lustre/osc/*/checksum_type.
2350 Description: use adler32 for page checksums
2351 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2356 Description: build for x2 fails
2357 Details : fix compile issue on Cray systems.
2361 Description: Properly match for duplicate locks
2362 Details : Due to different lock order from skiplists code, we need to
2363 traverse entire list for now
2366 Frequency : only on PPC/SLES10
2368 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2369 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2370 be defined. Add a hack in mkfs_lustre.c to work around this compile
2375 Description: LASSERT on MDS when client holding flock lock dies
2376 Details : ldlm pool logic depends on number of granted locks equal to
2377 number of released locks which is not true for flock locks, so
2378 just exclude such locks from consideration.
2382 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2383 Details : Use fsfilt_sync() for both the whole filesystem sync and
2384 individual file sync to eliminate dangerous inode locking
2385 with I_LOCK that can lead to a deadlock.
2389 Description: performance in 1.6.3
2390 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2391 sure that 1MB requests are not fragmented by the block layer.
2393 Severity : enhancement
2395 Description: organize the server-side client stats on per-nid basis
2396 Details : Change the structure of stats under obdfilter and mds to
2404 The "uuid"s file would list the uuids of _active_ exports.
2405 And the clear entry is to clear all stats and stale nids.
2407 Severity : enhancement
2409 Description: eliminate client locks in face of contention
2410 Details : file contention detection and lockless i/o implementation
2411 for contended files.
2415 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2418 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2420 Description: MDS may lose file striping (and hence file data) in some cases
2421 Details : If there are additional extended attributes stored on the MDS,
2422 in particular ACLs, SELinux, or user attributes (if user_xattr
2423 is specified for the client mount options) then there is a risk
2424 of attribute loss. Additionally, the Lustre file striping
2425 needs to be larger than default (e.g. striped over all OSTs),
2426 and an additional attribute must be stored initially in the
2427 inode and then increase in size enough to be moved to the
2428 external attribute block (e.g. ACL growing in size) for file
2433 Description: skiplist implementation simplification
2434 Details : skiplists are used to group compatible locks on granted list
2435 that was implemented as tracking first and last lock of each
2436 lock group the patch changes that to using doubly linked lists
2440 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2441 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2442 because LASSERT on both the data supplied by a client, and the
2443 data on disk is dangerous and incorrect.
2445 Severity : enhancement
2447 Description: Slow truncate/writes to huge files at high offsets.
2448 Details : Directly associate cached pages to lock that protect those pages,
2449 this allows us to quickly find what pages to write and remove
2450 once lock callback is received.
2454 Description: more ldlm soft lockups
2455 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2456 starve other threads from the resource lock for a long time in
2457 case of long waiting queue, so change the debug level from
2458 D_OTHER to the less frequently used D_INFO.
2460 Severity : enhancement
2462 Description: add -gid, -group, -uid, -user options to lfs find
2466 Description: Hung threads in invalidate_inode_pages2_range
2467 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2468 once one is completed. As a result, some RPCs are stuck in the
2469 queue and are never sent.
2473 Description: filter threads hungs on waiting journal commit
2474 Details : Cleanup filter group llog code, then only filter group llog will
2475 be only created in the MDS/OST syncing process.
2479 Description: Procfs and llog threads access destoryed import sometimes.
2480 Details : Sync the import destoryed process with procfs and llog threads by
2481 the import refcount and semaphore.
2483 Severity : enhancement
2485 Description: openlock cache of b1_6 port to HEAD
2490 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2491 Details : If insertion of an extent fails, then discard the inode
2492 preallocation and free data blocks else it can lead to duplicate
2497 Description: don't always update ctime in ext3_xattr_set_handle()
2498 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2499 In some cases the ctime should not be updated, for example for
2500 2.0->1.8 compatibility it is necessary to delete an xattr and it
2501 should not update the ctime.
2505 Bugzilla : 15713/16362
2506 Description: Assertion in iopen_connect_dentry in 1.6.3
2507 Details : looking up an inode via iopen with the wrong generation number can
2508 populate the dcache with a disconneced dentry while the inode
2509 number is in the process of being reallocated. This causes an
2510 assertion failure in iopen since the inode's dentry list contains
2511 both a connected and disconnected dentry.
2515 Description: assertion failure in ldlm_handle2lock()
2516 Details : fix a race between class_handle_unhash() and class_handle2object()
2517 introduced in lustre 1.6.5 by bug 13622.
2522 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2523 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2528 Description: oops in sd_iostats_seq_show()
2529 Details : unloading/reloading the scsi low level driver triggers a kernel
2530 bug when trying to access the sd iostat file.
2535 Description: Kernel panics during QLogic driver reload
2536 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2537 patch, causing memory corruption.
2542 Description: journal_dev option does not work in b1_6
2543 Details : pass mount option during pre-mount.
2545 Severity : enhancement
2547 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2548 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2549 extent information of a file. It can be used to map logical blocks
2550 in a file to physical blocks in the block device.
2554 Description: LDLM soft lockups - improvement
2555 Details : It is be possible to send the lock handle along with each read
2556 or write request because the client is already doing a lock match
2557 itself so there isn't any reason the OST should have to re-do that
2561 Frequency : only on Cray X2
2563 Description: X2 build failures
2564 Details : fix build failures on Cray X2.
2568 Description: xid & resent requests
2569 Details : Initialize RPC XID from clock at startup (randomly if clock is
2572 Severity : enhancement
2574 Description: Add lustre_start utility to start or stop multiple Lustre servers
2579 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2580 Details : In case of memory pressure, list_del() can be called twice on
2581 req->rq_history_list, causing a kernel oops.
2585 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2586 Details : fix stack overflow in the distributed lock manager by defering
2587 export eviction after a failed AST to the elt thread instead of
2588 handling it in the dlm interpret routine.
2592 Description: Convert some comments to new format.
2593 Details : Update documenting comments to match doxygen conventions.
2597 Description: Grammar fixes.
2598 Details : A couple of trivial spelling fixes.
2602 Description: OSD_COUNTERS-mandatory
2603 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2604 to simplify the code.
2608 Description: simplify cmm_device freeing logic.
2609 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2613 Description: Add lockdep support to dt_object_operations locking interface.
2614 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2615 parameter indicating lock ordering. Update mdd code to use new
2620 Description: Introduce failloc constants for lockless IO tests.
2621 Details : Add two new failloc constants to test lockless IO. Only one of
2622 them in implemented---another is checked in yet to be landed
2627 Description: Add lockdep support for inode mutex.
2628 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2629 in the situations where Lustre has to lock more than one inode
2634 Description: Add optional invariants checking support.
2635 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2636 configure switch. This macro is to be used for consistency and
2637 sanity checks that are too expensive to be left in `production'
2642 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2643 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2648 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2649 Details : Use spin_lock_nested() in (the only) situation where more than
2650 one ldlm_lock is locked simultaneously. Also, fix possible
2651 dead-lock in ldlm_lock_change_resource() by enforcing particular
2656 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2657 Details : Instead of specifying each ldlm_lock call-back through separate
2658 parameter, wrap them into struct ldlm_callback_suite.
2662 Description: Kill join_lru obd method and its callers.
2663 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2664 in memory---a requirement implemented through ->o_join_lru() obd
2665 method in HEAD. Remove this method and its users.
2669 Description: Add asynchronous ldlm ENQUEUE completion handler.
2670 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2671 case that stock ldlm_completion_ast() cannot handle as it waits
2672 until lock is granted. Introduce new ldlm_completion_ast_async()
2673 for this. Also comment ldlm_completion_ast().
2677 Description: ldlm_error <-> errno conversion.
2678 Details : Add functions to map (rather arbitrary) between LDLM error codes
2679 and standard errno values. CLIO needs this to prevent LDLM specific
2680 constants from escaping ldlm and osc.
2684 Description: Kill unused ldlm_handle2lock_ns() function.
2685 Details : Kill unused ldlm_handle2lock_ns() function.
2689 Description: Add lu_ref support to ldlm_lock
2690 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2691 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2692 and ldlm_resource. LDLM interface has to be changed, because code that
2693 releases a reference on a lock, has to "know" what reference this is.
2694 In the most frequent case
2696 lock = ldlm_handle2lock(handle);
2698 LDLM_LOCK_PUT(lock);
2700 no changes are required. When any other reference (received _not_ from
2701 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2702 instead of LDLM_LOCK_PUT().
2704 Arguably, changes are pervasive, and interface requires some discipline
2705 for proper use. On the other hand, it was very instrumental in finding
2706 a few leaked lock references.
2710 Description: Add ldlm_lock_addref_try().
2711 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2712 attempts to addref a lock that might be being canceled
2717 Description: Add ldlm_weigh_callback().
2718 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2719 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2720 hard-coded `number of pages' logic.
2724 Description: Add lockdep annotations to llog code.
2725 Details : Use appropriately tagged _nested() locking calls in the places
2726 where llog takes more than one ->lgh_lock lock.
2730 Description: Add loi_kms_set().
2731 Details : Wrap kms updates into a helper function.
2735 Description: Constify instances of struct lsm_operations.
2736 Details : Constify instances of struct lsm_operations.
2740 Description: lu_conf support.
2741 Details : On a server, a file system object is uniquely identified
2742 by a fid, which is sufficient to locate and load all object
2743 state (inode). On a client, on the other hand, more data are
2744 necessary instantiate an object. Change lu_object_find() and
2745 friends to take additional `lu_conf' argument describing object.
2746 Typically this includes layout information.
2750 Description: lu_context fixes.
2751 Details : Introduce new lu_context functions that are needed on the client
2752 side, where some system threads (ptlrpcd) are shared by multiple
2753 modules, and so cannot be stopped during module shutdown.
2757 Description: Add start and stop methods to lu_device_type_operations.
2758 Details : Introduce two new methods in lu_device_type_operations, that are
2759 invoked when first instance of a given type is created and last one
2760 is destroyed respectively. This is need by CLIO.
2764 Description: Add lu_ref support to struct lu_device.
2765 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2766 track leaked references.
2770 Description: Introduce lu_kmem_descr.
2771 Details : lu_kmem_descr and its companion interface allow to create
2772 and destroy a number of kmem caches at once.
2776 Description: Fix lu_object finalization race.
2777 Details : Fix a race between lu_object_find() finding an object and its
2778 concurrent finalization. This race is (most likely) not possible
2779 on the server, but might happen on the client.
2783 Description: Introduce lu_ref interface.
2784 Details : lu_ref is a debugging module allowing to track references to
2785 a given object. It is quite cpu expensive, and has to be
2786 explicitly enabled with --enable-lu_ref. See usage description
2791 Description: Factor lu_site procfs stats into a separate function.
2792 Details : Separate lu_site stats printing code into a separate function
2793 that can be reused on a client.
2797 Description: Constify instances of struct {lu,dt,md}_device_operations.
2798 Details : Constify instances of struct {lu,dt,md}_device_operations.
2802 Description: Introduce struct md_site and move meta-data specific parts of
2803 struct lu_site here.
2804 Details : Move md-specific fields out of struct lu_site into special struct
2805 md_site, so that lu_site can be used on a client.
2809 Description: Kill mdd_lov_destroy().
2810 Details : Remove unused mdd code.
2814 Description: Add st_block checking to multistat.c.
2815 Details : Add st_block checking to multistat.c.
2819 Description: Add lu_ref support to struct obd_device.
2820 Details : Add lu_ref tracking to obd_device.
2824 Description: Kill obd_set_fail_loc().
2825 Details : Remove unused code.
2829 Description: Add special type for ptlrpc_request interpret functions.
2830 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2831 there. Actual usage will be in CLIO.
2835 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2836 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2837 doesn't work with lockdep.
2841 Description: Add rwv.c test program.
2842 Details : New testing program exercising readv(2) and writev(2) (Qian).
2846 Description: Add sendfile.c test program.
2847 Details : New testing program exercising sendfile(2) (Jay).
2851 Description: Ratelimit a message that can be very frequent.
2852 Details : Ratelimit a memory allocation failure message that can
2857 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2858 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2862 Description: Liblustre build fixes.
2863 Details : Liblustre build fixes.
2867 Description: libcfs: add cfs_{need,cond}_resched() interface.
2868 Details : libcfs: add cfs_{need,cond}_resched() definition and
2869 implementations for Linux, NT, and liblustre.
2871 Severity : enhancement
2873 Description: More exported tunables for mballoc
2874 Details : Add support for tunable preallocation window and new tunables for
2875 large/small requests
2879 Description: Detect corruption of block bitmap and checking for preallocations
2880 Details : Checks validity of on-disk block bitmap. Also it does better
2881 checking of number of applied preallocations. When corruption is
2882 found, it turns filesystem readonly to prevent further corruptions.
2886 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2887 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2888 window to 0. If it does hit cache miss, and read-pattern is still
2889 stride-io mode, does not reset the stride window, but also does not
2890 increase the stride window length in this case.
2894 Frequency : only for big-endian servers
2895 Description: Check if system is big-endian while mounting fs with extents feature
2896 Details : Mounting a filesystem with extents feature will fail on big-endian
2897 systems since ext3-based ldiskfs is not supported on big-endian
2898 systems. This can be over-riden with "bigendian_extents" mount option.
2900 Severity : enhancement
2902 Description: The root squash functionality
2903 Details : A security feature, which is to prevent users from being able
2904 to mount lustre on their desktop, run as root, and delete
2905 all of the files in the filesystem. The goal is accomplished by
2906 remapping user id (UID) and group id (GID) of the root user to
2907 a UID and GID specified by the system administartor via Lustre
2908 configuration management server (MGS). The functionality also
2909 allows to specify sets of clients for which the remapping does
2914 Description: Excessive recovery window
2915 Details : With AT enabled, the recovery window can be excessively long (6000+
2916 seconds). To address this problem, we no longer use
2917 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2918 timeout no longer depends on the service time, it is set to
2919 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2920 time via pb_service_time.
2924 Description: Watchdog triggered on MDS failover
2925 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2926 the OSTs know that the MDS "UUID" can be reused for the same export
2927 from a different NID, so we do not need to wait for the export to be
2931 Frequency : rare, only if using MMP with Linux RAID
2933 Description: MMP doesn't work with Linux RAID
2934 Details : While using HA for Lustre servers with Linux RAID, it is possible
2935 that MMP will not detect multiple mounts. To make this work we
2936 need to unplug the device queue in RAID when the MMP block is being
2937 written. Also while reading the MMP block, we should read it from
2938 disk and not the cached one.
2940 Severity : enhancement
2942 Description: open file using fid
2943 Details : A file can be opened using just its fid, like
2944 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2947 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2950 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2951 Details : add proper locking for creation of jbd_4k slab cache
2955 Description: MMP check in ext3_remount() fails without displaying any error
2956 Details : When multiple mount protection fails during remount, proper error
2959 Severity : enhancement
2961 Description: Allow stripe size to be up to 4G-64k
2962 Details : Fix math logic to allow large stripe sizes.
2966 Description: add check for >8TB ldiskfs filesystems
2967 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2968 Don't allow >8TB ldiskfs filesystems to be mounted without
2969 force_over_8tb mount option
2974 Descriptoin: OSS complains "too many blocks" and crashes
2975 Details : In filter_commitrw_write():
2976 1. lquota_chkquota();
2977 2. obdo_from_inode();
2978 3. lquota_pending_commit();
2979 "2." has changed uid and gid in this case, then caused
2980 lquota_chkquota() and lquota_pending_commit() to use
2981 different uid and gid. That is the root of the bug.
2985 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2986 Details : Detect and print out the 'stripe count = -1' case properly.
2990 Description: Out or order replies might be lost on replay
2991 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2992 request with tid smaller than one currently being added, add it
2993 to the start, not end of the list.
2995 --------------------------------------------------------------------------------
2997 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2999 * Support for kernels:
3000 2.6.5-7.283 (SLES 9),
3001 2.6.9-55.EL (RHEL 4),
3002 2.6.16.46-0.14 (SLES 10),
3003 2.6.18.8 vanilla (kernel.org)
3004 * Client support for unpatched kernels:
3005 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3006 2.6.16 - 2.6.22 vanilla (kernel.org)
3007 * Due to recently discovered recovery problems, we do not recommend
3008 using patchless RHEL 4 clients with this or any earlier release.
3009 * Recommended e2fsprogs version: 1.39.cfs8
3010 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3011 * Starting with this release, the ldiskfs backing filesystem required
3012 by Lustre is now in its own package, lustre-ldiskfs. This package
3013 should be installed. It is versioned separately from Lustre and
3014 may be released separately in future.
3018 Description: block reactivating mgc import until all deactivates complete
3019 Details : Fix race when failing back MDT/MGS to itself (testing)
3022 Frequency : at statup only
3024 Description: mds_lov_synchronize race leads to various problems
3025 Details : simultaneous MDT->OST connections at startup can cause the
3026 sync to abort, leaving the OSC in a bad state.
3028 Severity : enhancement
3030 Description: add optional extra BUILD_VERSION info
3031 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3032 to override the lustre version.
3035 Frequency : 2.6.18 servers only
3037 Description: ll_kern_mount() doesn't release the module reference
3038 Details : The ldiskfs module reference count never drops down to 0
3039 because ll_kern_mount() doesn't release the module reference.
3044 Description: server LBUG when using old ost_num_threads parameter
3045 Details : Accept the old ost_num_threads parameter but warn that it
3046 is deprecated, and fix an off-by-one error that caused an LBUG.
3051 Description: Transient SCSI error results in persistent IO issue
3052 Details : iobuf->dr_error is not reinitialized to 0 between two
3056 Frequency : sometimes when underlying device returns I/O errors
3058 Description: OSTs not going read-only during write failures
3059 Details : OSTs are not remounted read-only when the journal commit threads
3060 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3061 instead of the ext3 wrappers.
3065 Description: poor connect scaling with increasing client count
3066 Details : Don't run filter_grant_sanity_check for more than 100 exports
3067 to improve scaling for large numbers of clients.
3070 Frequency : SLES10 only
3072 Description: sanity-quota.sh quotacheck failed: rc = -22
3073 Details : Quotas cannot be enabled on SLES10.
3076 Frequency : liblustre clients only
3078 Description: getdirentries does not give error when run on compute nodes
3079 Details : getdirentries does not fail when the size specified as an argument
3080 is too small to contain at least one entry
3082 Severity : enhancement
3084 Description: Add LNET router traceability for debug purposes
3085 Details : If a checksum failure occurs with a router as part of the
3086 IO path, the NID of the last router that forwarded the bulk data
3087 is printed so it can be identified.
3092 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3093 Details : Due to a race condition, liblustre clients were occasionally
3094 evicted incorrectly.
3096 Severity : enhancement
3098 Description: lfs setstripe use optional parameters instead of postional
3101 Severity : enhancement
3103 Description: Nanosecond timestamp support for ldiskfs
3104 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3105 resolution timestamps. There is not yet support for this at
3106 the Lustre filesystem level.
3109 Frequency : during server recovery
3111 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3112 Details : request with rq_no_resend flag not awake l_wait_event if they get
3116 Frequency : nfs export on patchless client
3118 Description: connectathon hang when test nfs export over patchless client
3119 Details : Disconnected dentry cannot be found with lookup, so we do not need
3120 to unhash it or make it invalid
3123 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3124 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3125 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3130 Description: open req refcounting wrong on reconnect
3131 Details : If reconnect happened between getting open reply from server and
3132 call to mdc_set_replay_data in ll_file_open, we will schedule
3133 replay for unreferenced request that we are about to free.
3134 Subsequent close will crash in variety of ways.
3135 Check that request is still eligible for replay in
3136 mdc_set_replay_data().
3141 Description: disable writes to filesystem when reading health_check file
3142 Details : the default for reading the health_check proc file has changed
3143 to NOT do a journal transaction and write to disk, because this
3144 can cause reads of the /proc file to hang and block HA state
3145 checking on a healthy but otherwise heavily loaded system. It
3146 is possible to return to the previous behaviour during configure
3147 with --enable-health-write.
3149 Severity : enhancement
3151 Description: 64-bit inode version
3152 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3153 to the inode. This will be required for version-based recovery.
3158 Description: MDS fails to start if a duplicate client export is detected
3159 Details : in some rare cases it was possible for a client to connect to
3160 an MDS multiple times. Upon recovery the MDS would detect this
3161 and fail during startup. Handle this more gracefully.
3163 Severity : enhancement
3165 Description: Add -o localflock option to simulate old noflock
3167 Details : This will achieve local-only flock/fcntl locks
3173 Description: log_commit_thread vs filter_destroy race leads to crash
3174 Details : Take import reference before releasing llog record semaphore
3179 Description: Wrong request locking in request set processing
3180 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3181 imp_delayed_list, in this place should be used imp_lock.
3184 Frequency : when reconnection
3186 Description: Grant Leak when osc reconnect to OST
3187 Details : When osc reconnect ost, OST(filter) should check whether it
3188 should grant more space to client by comparing fed_grant and
3189 cl_avail_grant, and return the granted space to client instead
3190 of "new granted" space, because client will call osc_init_grant
3191 to update the client grant space info.
3194 Frequency : when client reconnect to OST
3196 Description: Grant Leak when osc do resend and replay bulk write
3197 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3198 bulk write request, because the grant info while be sync between
3199 OSC and OST when reconnect, and we should ignore the grant info
3200 these of resend/replay write req.
3205 Description: Grant space more than available left space sometimes.
3206 Details : When then OST is about to be full, if two bulk writing from
3207 different clients came to OST. Accord the avaliable space of the
3208 OST, the first req should be permitted, and the second one
3209 should be denied by ENOSPC. But if the seconde arrived before
3210 the first one is committed. The OST might wrongly permit second
3211 writing, which will cause grant space > available space.
3214 Frequency : when client is evicted
3216 Description: Grant might be wrongly erased when osc is evicted by OST
3217 Details : when the import is evicted by server, it will fork another
3218 thread ptlrpc_invalidate_import_thread to invalidate the
3219 import, where the grant will be set to 0. While the original
3220 thread will update the grant it got when connecting. So if
3221 the former happened latter, the grant will be wrongly errased
3222 because of this race.
3227 Description: Checking Stale with correct fid
3228 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3229 to check whether it is stale, because sometimes, we want the
3230 enqueue happened anyway, and op_data.fid2 will not be initialized.
3232 Severity : enhancement
3234 Description: update patchless client
3235 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3238 Frequency : only with 2.4 kernel
3240 Description: random memory corruption
3241 Details : size of struct ll_inode_info is to big for union inode.u and this
3242 can be cause of random memory corruption.
3247 Description: Memory leak in recovery
3248 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3249 It should also check obd_fail before fsfilt_start, otherwise if
3250 fsfilt_start return -EROFS,(failover mds during mds recovery).
3251 then the req will return with repmsg->transno = 0 and rc = EROFS.
3252 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3253 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3254 be freed no matter whether fsfilt_commit success or not.
3257 Frequency : only with huge count clients
3259 Description: Prevents from taking the superblock lock in llap_from_page for
3261 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3262 for llap_from_page call in ll_removepage prevents from taking the
3263 superblock lock for a soon died page.
3268 Description: Not check open intent error before release open handle
3269 Details : in some rare cases, the open intent error is not checked before
3270 release open handle, which may cause
3271 ASSERTION(open_req->rq_transno != 0), because it tries to release
3272 the failed open handle.
3277 Description: Set cat log bitmap only after create log success.
3278 Details : in some rare cases, the cat log bitmap is set too early. and it
3279 should be set only after create log success.
3283 Description: Accessing a block bevice can re-enable I/O when Lustre is
3284 tearing down a device.
3285 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3290 Description: service threads may hog cpus when there are a lot of requests
3291 Details : Insert cond_resched to give other threads a chance to use some CPU
3296 Description: the cat log was not initialized in recovery
3297 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3298 unlink log on mds will not be initialized until mds post
3299 recovery. And also in mds post recovery, the unlink log will
3300 initialization will be done asynchronausly, so there will be race
3301 between add unlink log and unlink log initialization.
3305 Description: brw_stats were being printed incorrectly
3306 Details : brw_stats were being printed as log2 but all of them were not
3307 recorded as log2. Also remove some code duplication arising from
3308 filter_tally_{read,write}.
3312 Frequency : rare, only in recovery.
3313 Description: ASSERTION(req->rq_type != LI_POISON) failed
3314 Details : imp_lock should be held while iterating over imp_sending_list for
3315 prevent destroy request after get timeout in ptlrpc_queue_wait.
3319 Description: replay-single.sh test 52 fails
3320 Details : A lock's skiplist need to be cleanup when it being unlinked
3321 from its resource list.
3325 Description: Short directio read returns full requested size rather than
3327 Details : Direct I/O operations should return actual amount of bytes
3328 transferred rather than requested size.
3330 Severity : enhancement
3332 Description: metadata RPC reduction (e.g. for rm performance)
3333 Details : decrease the amount of synchronous RPC between clients and servers
3334 by canceling conflicing lock before the operation on the client
3335 and packing thier handles into the main operation RPC to server.
3337 Severity : enhancement
3339 Description: Async OSC create to avoid the blocking unnecessarily.
3340 Details : If an OST has no remaining object, system will block on the
3341 creation when it needs to create a new object on this OST. Now,
3342 ways use pre-created objects when available, instead of blocking on
3343 an empty osc while others are not empty. If we must block, we
3344 block for the shortest possible period of time.
3348 Description: improve handling recoverable errors
3349 Details : if request processig with error which can be recoverable on server
3350 request should be resend, otherwise page released from cache and
3353 Severity : enhancement
3355 Description: refine locking for avoid write wrong info into lov_objid file
3356 Details : fix possible races with add new target and write/update data in
3359 --------------------------------------------------------------------------------
3361 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3366 Frequency : on some architectures
3368 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3369 Details : Uninitialized flags sometimes cause configuration commands to
3373 Frequency : patchless clients only
3375 Description: missing __iget() symbol export
3376 Details : The __iget() symbol export is missing. To avoid the need for
3377 this on patchless clients the deathrow inode reaper is turned
3378 off, and we depend on the VM to clean up old inodes. This
3379 dependency was during via the fix for bug 12181.
3381 --------------------------------------------------------------------------------
3383 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3385 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3386 INTEROPERATE with older versions automatically. In many cases a
3387 special upgrade step is needed. Please read the
3388 user documentation before upgrading any part of a live system.
3389 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3390 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3391 * WARNING: Lustre configuration and startup changes are required with
3392 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3396 Severity : enhancement
3398 Description: MountConf
3399 Details : Lustre configuration is now managed via mkfs and mount
3400 commands instead of lmc and lconf. New obd types (MGS, MGC)
3401 are added for dynamic configuration management. See
3402 https://mail.clusterfs.com/wikis/lustre/MountConf for
3405 Severity : enhancement
3407 Description: dynamic OST addition
3408 Details : OSTs can now be added to a live filesystem
3410 Severity : enhancement
3412 Description: startup order invariance
3413 Details : MDTs and OSTs can be started in any order. Clients only
3414 require the MDT to complete startup.
3416 Severity : enhancement
3418 Description: parallel, asynchronous orphan cleanup
3419 Details : orphan cleanup is now performed in separate threads for each
3420 OST, allowing parallel non-blocking operation.
3422 Severity : enhancement
3424 Description: optimized stripe assignment
3425 Details : stripe assignments are now made based on ost space available,
3426 ost previous usage, and OSS previous usage, in order to try
3427 to optimize storage space and networking resources.
3429 Severity : enhancement
3431 Description: Permanently set tunables
3432 Details : All writable /proc/fs/lustre tunables can now be permanently
3433 set on a per-server basis, at mkfs time or on a live system.
3435 Severity : enhancement
3437 Description: Lustre message v2
3438 Details : Add lustre message format v2.
3440 Severity : enhancement
3442 Description: client OST exclusion list
3443 Details : Clients can be started with a list of OSTs that should be
3444 declared "inactive" for known non-responsive OSTs.
3448 Description: SPEC SFS validation failure on NFS v2 over lustre.
3449 Details : Changes the blocksize for regular files to be 2x RPC size,
3450 and not depend on stripe size.
3452 Severity : enhancement
3454 Description: Multiple MD RPCs in flight.
3455 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3456 To avoid overly-overloading MDS, introduce a limit on number of
3457 MDS RPCs in flight for a single client and add /proc controls
3458 to adjust this limit.
3460 Severity : enhancement
3462 Description: client read/write statistics
3463 Details : Add client read/write call usage stats for performance
3464 analysis of user processes.
3465 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3466 file access. extents_stats shows chunk size distribution.
3467 extents_stats_per_process show chunk size distribution per
3470 Severity : enhancement
3472 Description: mds statistics
3473 Details : Add detailed mds operations statistics in
3474 /proc/fs/lustre/mds/*/stats.
3478 Description: Failure of copying files with lustre special EAs.
3479 Details : Client side always return success for setxattr call for lustre
3480 special xattr (currently only "trusted.lov").
3485 Description: Refcount LNET uuids
3486 Details : The global LNET uuid list grew linearly with every startup;
3487 refcount repeated list entries instead of always adding to
3490 Severity : enhancement
3492 Description: Dynamic service threads
3493 Details : Within a small range, start extra service threads
3494 automatically when the request queue builds up.
3497 Frequency : mixed-endian client/server environments
3499 Description: mixed-endian crashes
3500 Details : The new msg_v2 system had some failures in mixed-endian
3503 Severity : enhancement
3505 Description: Easy OST removal
3506 Details : OSTs can be permanently deactivated with e.g. 'lctl
3507 conf_param lustre-OST0001.osc.active=0'
3509 Severity : enhancement
3511 Description: MGS proc entries
3512 Details : Added basic proc entries for the MGS showing what filesystems
3515 Severity : enhancement
3517 Description: provide MGS failover
3518 Details : Added config lock reacquisition after MGS server failover.
3520 Severity : enhancement
3522 Description: add Linux 2.4 support
3523 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3527 Description: a large application tries to do I/O to the same resource and dies
3528 in the middle of it.
3529 Details : Check the req->rq_arrival time after the call to
3530 ost_brw_lock_get(), but before we do anything about
3531 processing it & sending the BULK transfer request. This
3532 should help move old stale pending locks off the queue as
3533 quickly as obd_timeout.
3536 Frequency : when an incorrect nid is specified during startup
3538 Description: ptlrpc connect to non-existant node causes kernel crash
3539 Details : LNET can't be re-entered from an event callback, which
3540 happened when we expire a message after the export has been
3541 cleaned up. Instead, hand the zombie cleanup off to another
3544 Severity : enhancement
3546 Description: plain/inodebits lock performance improvement
3547 Details : Grouping plain/inodebits in granted list by their request modes
3548 and bits policy, thus improving the performance of search through
3552 Frequency : only if OST filesystem is corrupted
3554 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3555 Details : for a short time RPCs with bulk IO are in the replay list,
3556 but replay of bulk IOs is unimplemented. If the OST filesystem
3557 is corrupted due to disk cache incoherency and then replay is
3558 started it is possible to trip an assertion. Avoid putting
3559 committed RPCs into the replay list at all to avoid this issue.
3562 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3564 Description: System hang on startup
3565 Details : This bug allowed the liblustre (e.g. catamount) client to
3566 return to the app before handling all startup RPCs. This
3567 could leave the node unresponsive to lustre network traffic
3568 and manifested as a server ptllnd timeout.
3570 Severity : enhancement
3572 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3573 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3574 boolean to control whether to print peer debug info when a
3575 client's RPC times out.
3578 Frequency : only for kernels with patches from Lustre below 1.4.3
3580 Description: Remove old rdonly API
3581 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3584 Frequency : only for devices with external journals
3586 Description: Set external device read-only also
3587 Details : During a commanded failover stop, we set the disk device
3588 read-only while the server shuts down. We now also set any
3589 external journal device read-only at the same time.
3592 Frequency : when upgrading from 1.4 while trying to change parameters
3594 Description: The wrong (new) MDC name was used when setting parameters for
3595 upgraded MDT's. Also allows changing of OSC (and MDC)
3596 parameters if --writeconf is specified at tunefs upgrade time.
3599 Frequency : when setting specific ost indicies
3601 Description: QOS code breaks on skipped indicies
3602 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3603 created with --index need not be sequential.
3605 Severity : enhancement
3607 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3608 Details : The uninit_groups feature works in conjunction with the kernel
3609 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3610 up the pass1 processing of e2fsck. This is a read-only feature
3611 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3612 mount filesystems that have had this feature enabled.
3614 Severity : enhancement
3616 Description: Improve multi-block allocation algorithm to avoid fragmentation
3617 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3618 allocation locality and avoid filesystem fragmentation.
3620 ------------------------------------------------------------------------------
3622 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3624 * Support for kernels:
3625 2.4.21-47.0.1.EL (RHEL 3)
3626 2.6.5-7.283 (SLES 9)
3627 2.6.9-42.0.10.EL (RHEL 4)
3628 2.6.12.6 vanilla (kernel.org)
3629 2.6.16.27-0.9 (SLES 10)
3630 * Recommended e2fsprogs version: 1.39.cfs5
3632 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3636 Frequency : occasional, depends on client load and configuration
3637 Bugzilla : 12181, 12203
3638 Description: data loss for recently-modified files
3640 Details : In some cases it is possible that recently written or created
3641 files may not be written to disk in a timely manner (this should
3642 normally be within 30s unless client IO load is very high).
3643 The problem appears as zero-length files or files that are a
3644 multiple of 1MB in size after a client crash or client eviction
3645 that are missing data at the end of the file.
3647 This problem is more likely to be hit on clients where files are
3648 repeatedly created and unlinked in the same directory, clients
3649 have a large amount of RAM, have many CPUs, the filesystem has
3650 many OSTs, the clients are rebooted frequently, and/or the files
3651 are not accessed by other nodes after being written.
3653 The presence of the problem can be detected by looking at
3654 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3655 smaller than the second (nr_unused) then dirty files will not
3656 be flushed automatically to disk. "sync; sleep 10" should be
3657 run several times on the node before unmounting it to update
3658 Lustre (this is also safe to run on nodes without this problem).
3660 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3661 that can cause this same problem, so customers using that kernel
3662 also need to update the kernel in addition to Lustre. In order
3663 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3665 It is normal that files written just before a client crash (less
3666 than 30s) may not yet have been flushed to disk, even for local
3670 Frequency : frequent on thin XT3 nodes
3672 Description: UUID collision on thin XT3 Linux nodes
3673 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3674 randomly, since we relied on an insufficiently-seeded PRNG.
3679 Description: OSS hangs after "All ost request buffers busy"
3680 Details : A deadlock between quota and journal operations caused OSS
3681 hangs after printing "All ost request buffers busy."
3684 Frequency : always on liblustre builds
3686 Description: Cleanup compiler warnings on liblustre
3689 Frequency : always on liblustre builds on XT3
3691 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3695 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3696 > 32000 subdirectories
3697 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3698 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3699 subdirectory count crosses 32000. This will aid e2fsck to
3700 correctly handle more than 32000 subdirectories.
3703 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3705 Description: System hang on startup
3706 Details : This bug allowed the liblustre (e.g. catamount) client to
3707 return to the app before handling all startup RPCs. This
3708 could leave the node unresponsive to lustre network traffic
3709 and manifested as a server ptllnd timeout.
3711 Severity : enhancement
3713 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3714 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3715 boolean to control whether to print peer debug info when a
3716 client's RPC times out.
3721 Description: make O_SYNC working on 2.6 kernels
3722 Details : 2.6 kernels use different method for mark pages for write,
3723 so need add a code to lustre for O_SYNC work.
3728 Description: Failure to close file and release space on NFS
3729 Details : Put inode details into lock acquired in ll_intent_file_open.
3730 Use mdc_intent_lock in ll_intent_open to properly
3731 detect all kind of errors unhandled by mdc_enqueue.
3736 Description: proc file read during shutdown sometimes raced obd removal,
3738 Details : Add lock to prevent obd access after proc file removal.
3741 Frequency : Only for files larger than 4GB on 32-bit clients.
3743 Description: improperly doing page alignment of locks
3744 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3745 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3750 Description: RPCs being resent when they shouldn't be
3751 Details : Some RPCs that should not be resent are being resent. This
3752 can cause inconsistencies in the RPC state machine. Do not
3753 resend such requests.
3756 Frequency : rare, only with NFS export
3758 Description: Crash on NFS re-export node
3759 Details : under very unusual load conditions an assertion is hit in
3760 ll_intent_file_open()
3763 Frequency : only if OST filesystem is corrupted
3765 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3766 Details : for a short time RPCs with bulk IO are in the replay list,
3767 but replay of bulk IOs is unimplemented. If the OST filesystem
3768 is corrupted due to disk cache incoherency and then replay is
3769 started it is possible to trip an assertion. Avoid putting
3770 committed RPCs into the replay list at all to avoid this issue.
3775 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3776 Details : Large single O_DIRECT read and write calls can fail to allocate
3777 a sufficiently large buffer to process the request. In case of
3778 allocation failure the allocation is retried with a smaller
3779 buffer and broken into smaller requests.
3781 Severity : enhancement
3783 Description: Add -o localflock option to simulate old noflock behaviour.
3784 Details : This will achieve local-only flock/fcntl locks coherentness.
3789 Description: versioning check is incomplete
3790 Details : Checking the version difference of client vs. server, report
3791 error if the gap is too big.
3796 Description: add support PG_writeback bit
3797 Details : add support for PG_writeback bit for Lustre, for more carefull
3798 work with page cache in 2.6 kernel. This also fix some deadlocks
3799 and remove hack for work O_SYNC with 2.6 kernel.
3801 Severity : enhancement
3803 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3804 Details : The uninit_groups feature works in conjunction with the kernel
3805 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3806 up the pass1 processing of e2fsck. This is a read-only feature
3807 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3808 mount filesystems that have had this feature enabled.
3810 Severity : enhancement
3812 Description: Improve multi-block allocation algorithm to avoid fragmentation
3813 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3814 allocation locality and avoid filesystem fragmentation.
3816 ------------------------------------------------------------------------------
3818 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3820 * Support for kernels:
3821 2.6.9-42.0.3.EL (RHEL 4)
3822 2.6.5-7.276 (SLES 9)
3823 2.4.21-47.0.1.EL (RHEL 3)
3824 2.6.12.6 vanilla (kernel.org)
3825 2.6.16.21-0.8 (SLES10)
3826 * Recommended e2fsprogs version: 1.39.cfs2-0
3828 * The backwards-compatible /proc/sys/portals symlink has been removed
3829 in this release. Before upgrading, please ensure that you change
3830 any configuration scripts or /etc/sysctl.conf files that access
3831 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3832 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3833 in advance of the upgrade on any system running Lustre 1.4.6 or
3834 newer, since /proc/sys/lnet was added in that version.
3835 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3839 Frequency : only when quota is used
3841 Description: avoid scanning export list for quota master
3842 Details : Change the algorithms to avoid scanning export list in order
3843 to improve the efficiency.
3846 Frequency : MDS failover only, very rarely
3848 Description: "went back in time" messages on mds failover
3849 Details : The greatest transno may be lost when the current operation
3850 finishes with an error (transno==0) and the client's last_rcvd
3851 record is over-written. Save the greatest transno in the
3852 mds_last_transno for this case.
3855 Frequency : always for specific kernels and striping counts
3857 Description: client may get "Matching packet too big" without ACL support
3858 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3859 when trying to access files in certain configurations. The
3860 clients should in fact be denied when mounting because they do
3861 not understand ACLs.
3864 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3866 Description: many clients connecting with IO in progress causes connect timeouts
3867 Details : Avoid synchronous journal commits to avoid delays caused by many
3868 clients connecting/disconnecting when bulk IO is in progress.
3869 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3870 OST_IO_PORTAL to avoid delays behind potentially many pending
3874 Frequency : occasionally with multiple writers to a single file
3876 Description: shared writes to file may result in wrong size reported by stat()
3877 Details : Allow growing of kms when extent lock is cancelled
3880 Frequency : always with random mmap IO to multi-striped file
3882 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3883 Details : Only if the hole is at the end of OST object so that kms is too
3884 small. Fix is to increase kms accordingly in ll_nopage.
3887 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3889 Description: writes to a missing object would leak memory on the OST
3890 Details : If there is an inconsistency between the MDS and OST filesystems,
3891 such that the MDS references an object that doesn't exist, writes
3892 to that object will leak memory due to incorrect cleanup in the
3893 error handling path, eventually running out of memory on the OST.
3898 Description: Creating too long symlink causes lustre errors
3899 Details : Check symlink and name lengths before sending requests to MDS.
3902 Frequency : only if flock is enabled (not on by default)
3904 Description: posix locks not released on fd closure on 2.6.9+
3905 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3906 kernels, this caused such locks not to be released on fd close and
3907 then assertions on fs unmount about still used locks.
3910 Frequency : MDS failover only, very rarely
3912 Description: clients may get ASSERTION(granted_lock != NULL)
3913 Details : When request was taking a long time, and a client was resending
3914 a getattr by name lock request. The were multiple lock requests
3915 with the same client lock handle and
3916 mds_getattr_name->fixup_handle_for_resent_request found one of the
3917 lock handles but later failed with ASSERTION(granted_lock != NULL).
3922 Description: handle->h_buffer_credits > 0, assertion failure
3923 Details : h_buffer_credits is zero after truncate, causing assertion
3924 failure. This patch extends the transaction or creates a new
3928 Frequency : NFS re-export or patchless client
3929 Bugzilla : 11179, 10796
3930 Description: Crash on NFS re-export node (__d_move)
3931 Details : We do not want to hash the dentry if we don't have a lock.
3932 But if this dentry is later used in d_move, we'd hit uninitialised
3933 list head d_hash, so we just do this to init d_hash field but
3934 leave dentry unhashed.
3937 Frequency : NFS re-export or patchless client
3939 Description: NFS exports has problem with symbolic link
3940 Details : lustre client didn't properly install dentry when re-exported
3941 to NFS or running patchless client.
3944 Frequency : NFS re-export or patchless client
3946 Description: Various nfs/patchless fixes.
3947 Details : fixes reuse disconected alias for lookup process - this fixes
3948 warning "find_exported_dentry: npd != pd",
3949 fix permission error with open files at nfs.
3950 fix apply umask when do revalidate.
3953 Frequency : occasional
3955 Description: Crash on NFS re-export node
3956 Details : calling clear_page() on the wrong pointer triggered oops in
3957 generic_mapping_read().
3960 Frequency : rarely, using O_DIRECT IO
3962 Description: unaligned directio crashes client with LASSERT
3963 Details : check for unaligned buffers before trying any requests.
3966 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3968 Description: stale data returned from RAID cache
3969 Details : If only a small amount of IO is done to the RAID device before
3970 reading it again it is possible to get stale data from the RAID
3971 cache instead of reading it from disk.
3974 Frequency : always for sles10 kernel
3976 Description: sles10 support
3977 Details : ll_follow_link: compile fixes and using of nd_set_link
3978 under newer kernels.
3981 Frequency : depends on arch, kernel and compiler version, always on sles10
3984 Description: recursive or deep enough symlinks cause stack overflow
3985 Details : getting rid of large stack-allocated variable in
3989 Frequency : depends on hardware
3991 Description: lustre write performance loss in the SLES10 kernel
3992 Details : the performance loss is caused by using of write barriers in the
3993 ext3 code. The SLES10 kernel turns barrier support on by
3994 default. The fix is to undo that change for ldiskfs.
3996 ------------------------------------------------------------------------------
3998 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4000 * Support for kernels:
4001 2.6.9-42.0.3EL (RHEL 4)
4002 2.6.5-7.276 (SLES 9)
4003 2.4.21-47.0.1.EL (RHEL 3)
4004 2.6.12.6 vanilla (kernel.org)
4008 Frequency : quota enabled and large files being deleted
4010 Description: releasing more than 4GB of quota at once hangs OST
4011 Details : If a user deletes more than 4GB of files on a single OST it
4012 will cause the OST to spin in an infinite loop. Release
4013 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4018 Description: statfs data retrieved from /proc may be stale or zero
4019 Details : When reading per-device statfs data from /proc, in the
4020 {kbytes,files}_{total,free,avail} files, it may appear
4021 as zero or be out of date.
4024 Frequency : systems with MD RAID1 external journal devices
4026 Description: lconf's call to blkid is confused by RAID1 journal devices
4027 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4028 of returning all block devices that match the journal UUID.
4031 Frequency : always, for aggregate stripe size over 4GB
4033 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4034 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4035 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4038 Frequency : always if "lfs find" used on a local file/directory
4040 Description: "lfs find" segfaults if used on a local file/directory
4041 Details : The case where a directory component was not specified wasn't
4042 handled correctly. Handle this properly.
4045 Frequency : always on ppc64
4047 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4048 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4049 perform endianness conversion.
4052 Frequency : rarely (truncate to non-zero file size after write under load)
4053 Bugzilla : 10730, 10687
4054 Description: Files padded with zeros to next 4K multiple
4055 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4056 it is possible that files that are truncated to a non-zero size
4057 immediately after being written are filled with zero bytes beyond
4058 the truncated size. No file data is lost.
4060 Severity : enhancement
4062 Description: Allow recovery/failover for liblustre clients.
4063 Details : liblustre clients were unaware of failover configurations until
4066 Severity : enhancement
4068 Description: user file locks should fail when not mounting with flock option
4069 Details : Set up an error-returning stub in ll_file_operations.lock field
4070 to prevent incorrect behaviour when client is mounted without
4071 flock option. Also, set up properly f_op->flock field for
4075 Frequency : always on ia64
4077 Description: "lfs df" loops on printing out MDS statfs information
4078 Details : The obd_ioctl_data was not initialized and in some systems
4079 this caused a failure during the ioctl that did not return
4080 an error. Initialize the struct and return an error on failure.
4083 Frequency : SLES 9 only
4085 Description: Error of copying files with lustre special EAs as root
4086 Details : Client side always return success for setxattr call for lustre
4087 special xattr (currently only "trusted.lov").
4090 Frequency : rarely on clusters with both ia64+i386 clients
4092 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4093 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4094 clients are doing concurrent writes to the same file it is
4095 possible that the ia64 clients may LASSERT because the OST
4096 extent locks are not PAGE_SIZE aligned. Ensure that grown
4097 locks are always aligned on the request boundary.
4100 Frequency : specific use, occasional
4102 Description: Overwriting in use executable truncates on-disk binary image
4103 Details : If one node attempts to overwrite an executable in use by
4104 another node, we now correctly return ETXTBSY instead of
4105 truncating the file.
4107 Severity : enhancement
4109 Description: Async OSC create to avoid the blocking unnecessarily.
4110 Details : If an OST has no remaining object, system will block on the creating
4111 when need to create a new object on this OST. Now, ways use
4112 pre-created objects when available, instead of blocking on an
4113 empty osc while others are not empty. If we must block, we block
4114 for the shortest possible period of time.
4119 Description: chmod on Lustre root is propagated to other clients
4120 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4121 invalid by the follow_mount time.
4124 Frequency : liblustre clients only
4126 Description: Race in 'instant cancel' lock handling could lead to such locks
4127 never to be granted in case of SMP MDS
4128 Details : Do not destroy not yet granted but cbpending locks in
4132 Frequency : replay/resend of open
4134 Description: non null lock assetion failure in mds_intent_policy
4135 Details : Trying to replay/resend lockless open requests resulted in
4136 mds_open() returning 0 with no lock. Now it sets a flag if
4137 a lock is going to be returned.
4139 Severity : enhancement
4141 Description: Checksum enhancements
4142 Details : New checksum enhancements allow for resending RPCs that failed
4145 Severity : enhancement
4147 Description: Tunables on number of dirty pages in cacche
4148 Details : Allow to set limit on number of dirty pages cached.
4153 Description: client crash on unmount - lock still has references
4154 Details : In some error handling cases it was possible to leak a lock
4155 reference on a client while accessing a file. This was not
4156 harmful to the client during operation, but would cause the
4157 client to crash when the filesystem is unmounted.
4160 Frequency : specific case, rare
4162 Description: ETXTBSY on mds though file not in use by client
4163 Details : ETXTBSY is no longer incorrectly returned when attempting to
4164 chmod or chown a directory that the user previously tried to
4165 execute or a currently-executing binary.
4168 Frequency : extremely rare except on liblustre-based clients
4170 Description: Lustre space not freed when files are deleted
4171 Details : Clean up open-unlinked files after client eviction. Previously
4172 the unlink was skipped and the files remained as orphans.
4177 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4178 Details : In some cases it was possible to send a blocking callback to
4179 a client doing a glimpse, even though that client didn't get
4180 a lock granted. When the glimpse lock is cancelled on the OST
4181 the freed lock is left on the waiting list and corrupted the list.
4184 Frequency : all core dumps
4186 Description: Broke core dumps to lustre
4187 Details : Negative dentry may be unhashed if parent does not have UPDATE
4188 lock, but some callers, e.g. do_coredump, expect dentry to be
4189 hashed after successful create, hash it in ll_create_it.
4191 ------------------------------------------------------------------------------
4193 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4195 * Support for kernels:
4196 2.6.9-42.0.2.EL (RHEL 4)
4197 2.6.5-7.276 (SLES 9)
4198 2.4.21-40.EL (RHEL 3)
4199 2.6.12.6 vanilla (kernel.org)
4203 Frequency : always on RHEL 3
4205 Description: Number of open files grows over time
4206 Details : The number of open files grows over time, whether or not
4207 Lustre is started. This was due to a filp leak introduced
4208 by one of our kernel patches.
4210 ------------------------------------------------------------------------------
4212 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4214 * Support for kernels:
4215 2.6.9-42.EL (RHEL 4)
4216 2.6.5-7.276 (SLES 9)
4217 2.4.21-40.EL (RHEL 3)
4218 2.6.12.6 vanilla (kernel.org)
4223 Bugzilla : 5719, 9635, 9792, 9684
4224 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4225 Details : If a server is under heavy load and cannot reply to new
4226 connection requests before the client resends the (re)connect,
4227 the connection handling code can behave badly if two service
4228 threads are concurrently handing separate (re)connections from
4229 the same client. Add better locking to the connection handling
4230 code, and ensure that only a single connection will be processed
4231 for a given client UUID, even if the lock is dropped.
4233 Severity : enhancement
4235 Description: add TCP zero-copy support to kernel
4236 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4237 sends if the hardware supports scatter-gather and checksumming.
4238 This allows socklnd to do client-write and server-read more
4239 efficiently and reduce CPU utilization from skbuf copying.
4242 Frequency : only if NFS exporting from client
4244 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4245 Details : If setting attributes on a file created under NFS that had
4246 never been opened it would be possible to oops the client
4247 if the file had no objects.
4250 Frequency : always for liblustre
4252 Description: liblustre client does MDS+OSTs setattr RPC for each write
4253 Details : When doing a write from a liblustre client, the client
4254 incorrectly issued an RPC to the MDS and each OST the file was
4255 striped over in order to update the timestamps. When writing
4256 with small chunks and many clients this could overwhelm the MDS
4257 with RPCs. In all cases it would slow down the write because
4258 these RPCs are unnecessary.
4260 Severity : enhancement
4262 Description: allow number of MDS service threads to be changed at module load
4263 Details : It is now possible to change the number of MDS service threads
4264 running. Adding "options mds mds_num_threads={N}" to the MDS's
4265 /etc/modprobe.conf will set the number of threads for the next
4266 time Lustre is restarted (assuming the "mds" module is also
4267 reloaded at that time). The default number of threads will
4268 stay the same, 32 for most systems.
4273 Description: OST crash if filesystem is unformatted or corrupt
4274 Details : If an OST is started on a device that has never been formatted
4275 or if the filesystem is corrupt and cannot even mount then the
4276 error handling cleanup routines would dereference a NULL pointer.
4281 Description: NULL pointer deref in llap_from_page.
4282 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4283 incorrect) mapping if the page was truncated/reclaimed while it was
4284 searched for. Check for this condition and skip such pages when
4285 doing readahead. Introduce extra check to llap_from_page() to
4286 verify page->mapping->host is non-NULL (so page is not anonymous).
4289 Frequency : Sometimes when using sys_sendfile
4291 Description: "page not covered by a lock" warnings from ll_readpage
4292 Details : sendfile called ll_readpage without right page locks present.
4293 Now we introduced ll_file_sendfile that does necessary locking
4294 around call to generic_file_sendfile() much like we do in
4298 Frequency : with certain MDS communication failures at client mount time
4300 Description: NULL pointer deref after failed client mount
4301 Details : a client connection request may delayed by the network layer
4302 and not be sent until after the PTLRPC layer has timed out the
4303 request. If the client fails the mount immediately it will try
4304 to clean up before the network times out the request. Add a
4305 reference from the request import to the obd device and delay
4306 the cleanup until the network drops the request.
4309 Frequency : occasionally during client (re)connect
4311 Description: assertion failure during client (re)connect
4312 Details : processing a client connection request may be delayed by the
4313 client or server longer than the client connect timeout. This
4314 causes the client to resend the connection request. If the
4315 original connection request is replied in this interval, the
4316 client may trip an assertion failure in ptlrpc_connect_interpret()
4317 which thought it would be the only running connect process.
4320 Frequency : only with obd_echo servers and clients that are rebooted
4322 Description: kernel BUG accessing uninitialized data structure
4323 Details : When running an obd_echo server it did not start the ping_evictor
4324 thread, and when a client was evicted an uninitialized data
4325 structure was accessed. Start the ping_evictor in the RPC
4326 service startup instead of the OBD startup.
4328 Severity : enhancement
4329 Bugzilla : 10193 (patchless)
4330 Description: Remove dependency on various unexported kernel interfaces.
4331 Details : No longer need reparent_to_init, exit_mm, exit_files,
4332 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4335 Frequency : rare (only users of deprecated and unsupported LDAP config)
4337 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4338 Details : LDAP apparently contains 'lustreName' attributes instead of
4339 'name'. A simple remapping of the name is sufficient.
4342 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4344 Description: waiting_locks_callback trips kernel BUG if client is evicted
4345 Details : Running with the dump_on_timeout debug flag turned on makes
4346 it possible that the waiting_locks_callback() can try to dump
4347 the Lustre kernel debug logs from an interrupt handler. Defer
4348 this log dumping to the expired_lock_main() thread.
4350 Severity : enhancement
4352 Description: Support NFS exporting on 2.6 kernels.
4353 Details : Implement non-rawops metadata methods for NFS server to use without
4354 changing NFS server code.
4357 Frequency : very rare (synthetic metadata workload only)
4359 Description: two racing renames might cause an MDS thread to deadlock
4360 Details : Running the "racer" program may cause one MDS thread to rename
4361 a file from being the source of a rename to being the target of
4362 a rename at exactly the same time that another thread is doing
4363 so, and the second thread has already enqueued these locks after
4364 doing a lookup of the target and is trying to relock them in
4365 order. Ensure that we don't try to re-lock the same resource.
4368 Frequency : only very large systems with liblustre clients
4370 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4371 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4372 all OSTs in parallel. This allows the eviction of stale liblustre
4373 clients to proceed much faster than if they were done in series,
4374 and also offers similar improvements for other set_info RPCs.
4379 Description: excessive CPU usage during initial read phase on client
4380 Details : During the initial read phase on a client, it would agressively
4381 retry readahead on the file, consuming too much CPU and impacting
4382 performance (since 1.4.5.8). Improve the readahead algorithm
4383 to avoid this, and also improve some other common cases (read
4384 of small files in particular, where "small" is files smaller than
4385 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4390 Description: MDS crash when receiving packet with unknown intent.
4391 Details : Do not LBUG in unknown intent case, just return -EFAULT
4393 Severity : enhancement
4394 Bugzilla : 9293, 9385
4395 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4396 Details : Do not serialize getattr (non-intent version) and statfs.
4399 Frequency : occasional, when OST network is overloaded/intermittent
4401 Description: client evicted by OST after bulk IO timeout
4402 Details : If a client sends a bulk IO request (read or write) the OST
4403 may evict the client if it is unresposive to its data GET/PUT
4404 request. This is incorrect if the network is overloaded (takes
4405 too long to transfer the RPC data) or dropped the OST GET/PUT
4406 request. There is no need to evict the client at all, since
4407 the pinger and/or lock callbacks will handle this, and the
4408 client can restart the bulk request.
4411 Frequency : Always when mmapping file with no objects
4413 Description: client crashes when mmapping file with no objects
4414 Details : Check that we actually have objects in a file before doing any
4415 operations on objects in ll_vm_open, ll_vm_close and
4421 Description: Request leak when working with deleted CWD
4422 Details : Introduce advanced request refcount tracking for requests
4423 referenced from lustre intent.
4425 Severity : Enhancement
4427 Description: Cache open file handles on client.
4428 Details : MDS now will return special lock along with openhandle, if
4429 requested and client is allowed to hold openhandle, even if unused,
4430 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4431 closing files for every read/write openration.
4433 Severity : Enhancement
4435 Description: Cache open negative dentries on client when possible.
4436 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4437 negative dentries on lock revocation.
4442 Description: Remounting a client read-only wasn't possible with a zconf mount
4443 Details : It wasn't possible to remount a client read-only with llmount.
4445 Severity : enhancement
4446 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4447 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4448 MPICH is included to add Lustre-specific ADIO interfaces.
4449 This is based closely on the UFS ADIO layer and only differs
4450 in file creation, in order to allow the OST striping to be set.
4451 This is user-contributed code and not supported by CFS.
4456 Description: extended inode attributes (immutable, append-only) work improperly
4457 when 2.4 and 2.6 kernels are used on client/server or vice versa
4458 Details : Introduce kernel-independent values for these flags.
4460 Severity : enhancement
4463 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4464 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4465 granularity. Now, Fractional MB granularity is be supported,
4466 this is very useful for low memory system.
4468 Severity : enhancement
4470 Description: Getattr by fid
4471 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4472 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4473 have locks and avoids getattr rpc requests for it on every stat.
4476 Frequency : Always, for filesystems larger than 2TB
4478 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4479 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4480 filesystems larger than 2TB to be mounted with mballoc enabled.
4483 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4485 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4486 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4487 filesystems larger than 2TB is mounted then the kernel will
4488 silently corrupt the start of the filesystem. CONFIG_LBD is
4489 enabled for all CFS-supported kernels, but the possibility of
4490 this happening with a modified kernel config exists.
4492 Severity : enhancement
4494 Description: add client O_DIRECT support for 2.6 kernels
4495 Details : It is now possible to do O_DIRECT reads and writes to files
4496 in the Lustre client mountpoint on 2.6 kernel clients.
4498 Severity : enhancement
4500 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4501 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4502 parallel, not waiting for response from every OST before sending
4503 a rpc to the next OST.
4508 Description: setattr vs write race when updating file timestamps
4509 Details : Client processes that update a file timestamp into the past
4510 right after writing to the file (e.g. tar) it is possible that
4511 the updated file modification time can be reset to the current
4512 time due to a race between processing the setattr and write RPC.
4514 Severity : enhancement
4516 Description: Bring 'lfs find' closer in line with regular Linux find.
4517 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4518 -print0 options and obtains all the needed info through the lustre
4521 Severity : enhancement
4523 Description: support up to 1024 configured devices on one node
4524 Details : change obd_dev array from statically allocated to dynamically
4525 allocated structs as they are first used to reduce memory usage
4530 Description: Flush dirty partially truncated pages during truncate
4531 Details : Immediatelly flush partially truncated pages in filter_setattr,
4532 this way we completely avoid having any pages in page cache on OST
4533 and can retire ugly workarounds during writes to flush such pages.
4538 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4539 Details : protect inode from truncation within vfs_unlink() context
4540 just take a reference before calling vfs_unlink() and release it
4541 when parent's i_sem is free.
4546 Description: last_id value checked outside lock on OST caused LASSERT failure
4547 Details : If there were multiple MDS->OST object precreate requests in
4548 flight, it was possible that the OST's last object id was checked
4549 outside a lock and incorrectly tripped an assertion. Move checks
4550 inside locks, and discard old precreate requests.
4553 Frequency : always, if extents are used on OSTs
4555 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4556 Details : OSTs using the extents format would not zero the high 16 bits of
4557 the index physical block number. This is not a problem for any
4558 OST filesystems smaller than 16TB, and no kernels support ext3
4559 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4560 new/modified files) and can be fixed for existing filesystems
4561 with e2fsprogs-1.39-cfs1.
4566 Description: import connection selection may be incorrect if timer wraps
4567 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4568 import connections to be ignored if the 32-bit jiffies counter
4569 wraps. Use a 64-bit jiffies counter.
4572 Frequency : very large clusters immediately after boot
4574 Description: LNET request buffers exhausted under heavy short-term load
4575 Details : If a large number of client requests are generated on a service
4576 that has previously never seen so many requests it is possible
4577 that the request buffer growth cannot keep up with the spike in
4578 demand. Instead of dropping incoming requests, they are held in
4579 the LND until the RPC service can accept more requests.
4582 Frequency : Sometimes during replay
4584 Description: Assertion failure in ll_local_open after replay.
4585 Details : If replay happened on an open request reply before we were able
4586 to set replay handler, reply will become not swabbed tripping the
4587 assertion in ll_local_open. Now we set the handler right after
4588 recognising of open request
4591 Frequency : very rare
4593 Description: kernel reports "badness in vsnprintf"
4594 Details : Reading from the "recovery_status" /proc file in small chunks
4595 may cause a negative length in lprocfs_obd_rd_recovery_status()
4596 call to vsnprintf() (which is otherwise harmless). Exit early
4597 if there is no more space in the output buffer.
4599 Severity : enhancement
4601 Description: clear OBD RPC statistics by writing to them
4602 Details : It is now possible to clear the OBD RPC statistics by writing
4603 to the "stats" file.
4608 Description: Client mtime is not the same on different clients after utimes
4609 Details : In some cases, the client was using the utimes() syscall on
4610 a file cached on another node. The clients now validate the
4611 ctime from the MDS + OSTs to determine which one is right.
4616 Description: Inability to activate failout mode
4617 Details : lconf script incorrectly assumed that in pythong string's numeric
4618 value is used in comparisons.
4621 Frequency : always with multiple stripes per file
4623 Description: Inefficient object allocation for mutli-stripe files
4624 Details : When selecting which OSTs to stripe files over, for files with
4625 a stripe count that divides evenly into the number of OSTs,
4626 the MDS is always picking the same starting OST for each file.
4627 Return the OST selection heuristic to the original design.
4632 Description: mount failures may take full timeout to return an error
4633 Details : Under some heavy load conditions it is possible that a
4634 failed mount can wait for the full obd_timeout interval,
4635 possibly several minutes, before reporting an error.
4636 Instead return an error as soon as the status is known.
4638 Frequency : quota enabled and large files being deleted
4640 Description: releasing more than 4GB of quota at once hangs OST
4641 Details : If a user deletes more than 4GB of files on a single OST it
4642 will cause the OST to spin in an infinite loop. Release
4643 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4648 Description: statfs data retrieved from /proc may be stale or zero
4649 Details : When reading per-device statfs data from /proc, in the
4650 {kbytes,files}_{total,free,avail} files, it may appear
4651 as zero or be out of date.
4654 Frequency : systems with MD RAID1 external journal devices
4656 Description: lconf's call to blkid is confused by RAID1 journal devices
4657 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4658 of returning all block devices that match the journal UUID.
4661 Frequency : always, for aggregate stripe size over 4GB
4663 Description: assertion fails when trying to use 4GB stripe size
4664 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4665 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4668 Frequency : always on ppc64
4670 Description: the first write on an ext3 filesystem with mballoc got stuck
4671 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4672 perform endianness conversion.
4674 ------------------------------------------------------------------------------
4676 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4678 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4679 INTEROPERATE with older versions automatically. Please read the
4680 user documentation before upgrading any part of a live system.
4681 * WARNING: Lustre networking configuration changes are required with
4682 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4685 * Support for kernels:
4686 2.6.9-22.0.2.EL (RHEL 4)
4687 2.6.5-7.244 (SLES 9)
4688 2.6.12.6 vanilla (kernel.org)
4691 Severity : enhancement
4692 Bugzilla : 7981/8208
4693 Description: Introduced Lustre Networking (LNET)
4694 Details : LNET is new networking infrastructure for Lustre, it includes
4695 a reorganized network configuration mode (see the user
4696 documentation for full details) as well as support for routing
4697 between different network fabrics. Lustre Networking Devices
4698 (LNDs) for the supported network fabrics have also been
4699 created for this new infrastructure.
4701 Severity : enhancement
4702 Description: Introduced Access control lists
4703 Details : clients can set ACLs on files and directories in order to have
4704 more fine-grained permissions than the standard Unix UGO+RWX.
4705 The MDS must be started with the "-o acl" mount option.
4707 Severity : enhancement
4708 Description: Introduced filesystem quotas
4709 Details : Administrators may now establish per-user quotas on the
4712 Severity : enhancement
4714 Description: Configuration change for the XT3
4715 The PTLLND is now used to run Lustre over Portals on the XT3
4716 The configure option(s) --with-cray-portals are no longer used.
4717 Rather --with-portals=<path-to-portals-includes> is used to
4718 enable building on the XT3. In addition to enable XT3 specific
4719 features the option --enable-cray-xt3 must be used.
4724 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4725 Details : A race between allocating a new llcd and re-getting the llcd_lock
4726 allowed another thread to grab newly-allocated llcd.
4728 Severity : enhancement
4730 Description: 2.6 OST async journal commit and locking fix to improve performance
4731 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4732 2.6 kernels are now async as they already were in 2.4 kernels so
4733 that they can commit concurrently with the network bulk transfer.
4734 For block-allocated files the filter allocation semaphore is held
4735 to avoid filesystem fragmentation during allocation. BKL lock
4736 removed for 2.6 xattr operations where it is no longer needed.
4741 Description: lconf incorrectly determined whether two IP networks could talk
4742 Details : In some more complicated routing and multiple-network
4743 configurations, lconf will avoid trying to make a network
4744 connection to a disjoint part of the IP space. It was doing the
4745 math incorrectly for one set of cases.
4750 Description: Fix for potential infinite loop processing records in an llog.
4751 Details : If an llog record is corrupted/zeroed, it is possible to loop
4752 forever in llog_process(). Validate the llog record length
4753 and skip the remainder of the block on error.
4756 Frequency : occasional (liblustre only)
4758 Description: liblustre could not open files whose last component is a symlink
4759 Details : sysio_path_walk() would incorrectly pass the open intent to
4760 intermediate path components.
4763 Frequency : rare (liblustre only with non-standard tuning)
4764 Bugzilla : 7201 (7350)
4765 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4766 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4767 after it has been released, only do this for referenced locks
4769 Severity : enhancement
4771 Description: specifying an (invalid) directory default stripe_size of -1
4772 would reset the directory default striping
4773 Details : stripe_size -1 was used internally to signal directory stripe
4774 removal, now use "all default" to signal dir stripe removal
4775 as a directory striping of "all default" is not useful
4778 Frequency : common for large clusters running liblustre clients
4780 Description: doing an ls when liblustre clients are running is slow
4781 Details : sending a glimpse AST to a liblustre client waits for every AST
4782 to time out, as liblustre clients will not respond. Since they
4783 cannot cache data we refresh the OST lock LVB from disk instead.
4785 Severity : enhancement
4787 Description: doing an ls at the same time as file IO can be slow
4788 Details : enqueue and other "small" requests can be blocked behind many
4789 large IO requests. Create a new OST IO portal for non-IO
4790 requests so they can be processed faster.
4793 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4795 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4796 created file with mode 000
4797 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4798 allow an ftruncate() as a non-root user to a file with mode 000.
4799 Lustre now respects this flag to disable mode checks when
4800 truncating a file owned by the user
4803 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4806 Description: Revoking locks from clients that went dead or catatonic might take
4808 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4809 cancellation of such locks instant on servers without waiting for
4810 any reply from clients. Clients drops these locks when cancel
4811 notification from server is received without replying.
4814 Frequency : liblustre-only, when liblustre client dies or becomes busy
4816 Description: Doing ls on Linux clients can take a long time with active
4818 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4819 granting such locks to it in the first place if possible. Locks
4820 are taken by proxy on the OST during the read or write and
4821 dropped immediately afterward. Add connect flags handling, do
4822 not grant locks to liblustre clients for glimpse ASTs.
4824 Severity : enhancement
4826 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4827 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4828 sequential reads and behaves badly with applications doing random
4829 reads. Improve it by reading ahead at least read region, and
4830 avoiding excessive large RPC for small reads.
4832 Severity : enhancement
4834 Description: Creating more than 1000 files for a single job may cause a load
4835 imbalance on the OSTs if there are also a large number of OSTs.
4836 Details : qos_prep_create() uses an OST index reseed value that is an
4837 even multiple of the number of available OSTs so that if the
4838 reseed happens in the middle of the object allocation it will
4839 still utilize the OSTs as uniformly as possible.
4844 Description: OST or MDS may oops in ping_evictor_main()
4845 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4846 but doesn't restart at beginning of obd_exports_timed list
4849 Severity : enhancement
4851 Description: improve by-nid export eviction on the MDS and OST
4852 Details : allow multiple exports with the same NID to be evicted at one
4853 time without re-searching the exports list.
4856 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4858 Description: MDS may oops in groups_free()
4859 Details : in rare race conditions a newly allocated group_info struct is
4860 freed again, and this can be NULL. The 2.4 compatibility code
4861 for groups_free() checked for a NULL pointer, but 2.6 did not.
4864 Frequency : common for liblustre clients doing little filesystem IO
4865 Bugzilla : 9352, 7313
4866 Description: server may evict liblustre clients accessing contended locks
4867 Details : if a client is granted a lock or receives a completion AST
4868 with a blocking AST already set it would not reply to the AST
4869 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4872 Frequency : lfs setstripe, only systems with more than 160 OSTs
4874 Description: unable to set striping with a starting offset beyond OST 160
4875 Details : llapi_create_file() incorrectly limited the starting stripe
4876 index to the maximum single-file stripe count.
4879 Frequency : LDAP users only
4881 Description: lconf did not handle in-kernel recovery with LDAP properly
4882 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4884 Severity : enhancement
4886 Description: bind OST threads to NUMA nodes to improve performance
4887 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4888 node and do their allocations there to localize memory access
4890 Severity : enhancement
4892 Description: llmount can determine client NID directly from Myrinet (GM)
4893 Details : the client NID code from gmnalnid was moved directly into
4894 llmount, removing the need to use this or specifying the
4895 client NID explicitly when mounting GM clients with zeroconf
4898 Frequency : if client is started with down MDS
4900 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4901 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4902 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4903 ptlrpc_import_delayed_req() because we didn't check if the
4904 request was interrupted, and we also didn't break out of the
4905 event loop if there was no timeout
4910 Description: data loss during non-page-aligned writes to a single file from
4911 both multiple nodes and multiple threads on one node at same time
4912 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4913 inconsistency led to false short-reads, that were cached and later
4914 used by ->prepare_write() to fill in partially written page,
4915 leading to data loss.
4918 Frequency : always, if lconf --abort_recovery used
4920 Description: lconf --abort_recovery fails with 'Operation not supported'
4921 Details : lconf was attempting to abort recovery on the MDT device and not
4924 Severity : enhancement
4926 Description: remove cleanup logs
4927 Details : replace lconf-generated cleanup logs with lustre internal
4928 cleanup routines. Eliminates the need for client-cleanup and
4931 Severity : enhancement
4933 Description: add support for EAs (user and system) on lustre filesystems
4934 Details : it is now possible to store extended attributes in the Lustre
4935 client filesystem, and with the user_xattr mount option it
4936 is possible to allow users to store EAs on their files also
4938 Severity : enhancement
4940 Description: Add possibility (config option) to show minimal available OST free
4942 Details : When compiled with --enable-mindf configure option, statfs(2)
4943 (and so, df) will return least minimal free space available from
4944 all OSTs as amount of free space on FS, instead of summary of
4945 free spaces of all OSTs.
4947 Severity : enhancement
4949 Description: do not expand extent locks acquired on OST-side
4950 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4951 by server: they are not cached anyway.
4954 Frequency : when mmap is used/binaries executed from Lustre
4956 Description: Unmmap pages before throwing them away from read cache.
4957 Details : llap_shrink cache now attempts to unmap pages before discarding
4958 them (if unmapping failed - do not discard). SLES9 kernel has
4959 extra checks that trigger if this unmapping is not done first.
4964 Description: lconf didn't resolve symlinks before checking to see whether a
4965 given mountpoint was already in use
4968 Frequency : when migrating failover services
4969 Bugzilla : 6395, 9514
4970 Description: When migrating a subset of services from a node (e.g. failback
4971 from a failover service node) the remaining services would
4972 time out and evict clients.
4973 Details : lconf --force (implied by --failover) sets the global obd_timeout
4974 to 5 seconds in order to quickly disconnect, but this caused
4975 other RPCs to time out too quickly. Do not change the global
4976 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4978 Severity : enhancement
4979 Frequency : if MDS is started with down OST
4980 Bugzilla : 9439,5706
4981 Description: Allow startup/shutdown of an MDS without depending on the
4982 availability of the OSTs.
4983 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4984 Add appropriate locking and lov-osc refcounts for safe
4985 cleaning. Add osc abort_inflight calls in case the
4986 synchronize never started.
4989 Frequency : occasional (Cray XT3 only)
4991 Description: root not authorized to access files in CRAY_PORTALS environment
4992 Details : The client process capabilities were not honoured on the MDS in
4993 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4994 been accessed by an authorized user then root was able to access
4995 the file on the local client also. The root user capabilities
4996 are now allowed on the MDS, as this environment has secure UID.
4999 Frequency : occasional
5001 Description: ldiskfs "too long searching" message happens too often
5002 Details : A debugging message (otherwise harmless) prints too often on
5003 the OST console. This has been reduced to only happen when
5004 there are fragmentation problems on the filesystem.
5009 Description: Division by zero in statfs when all OSCs are inactive
5010 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5011 lov_statfs divides by value returned by lov_get_stripecnt().
5015 Bugzilla : 9489, 3273
5016 Description: First write from each client to each OST was only 4kB in size,
5017 to initialize client writeback cache, which caused sub-optimal
5018 RPCs and poor layout on disk for the first written file.
5019 Details : Clients now request an initial cache grant at (re)connect time
5020 and so that they can start streaming writes to the cache right
5021 away and always do full-sized RPCs if there is enough data.
5022 If the OST is rebooted the client also re-establishes its grant
5023 so that client cached writes will be honoured under the grant.
5028 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5029 Details : Now I/O RPCs go to different portal number and (presumably) fast
5030 lock requests (and glimses) and other RPCs get their own service
5031 threads pool that should be able to service those RPCs
5034 Severity : enhancement
5036 Description: Ability to exchange lustre version between client and servers and
5037 issue warnings at client side if client is too old. Also for
5038 liblustre clients there is ability to refuse connection of too old
5040 Details : New 'version' field is added to connect data structure that is
5041 filled with version info. That info is later checked by server and
5045 Frequency : rare, liblustre only.
5046 Bugzilla : 9296, 9581
5047 Description: Two simultaneous writes from liblustre at offset within same page
5048 might proceed at the same time overwriting eachother with stale
5050 Details : I/O lock withing llu_file_prwv was released too early, before data
5051 actually was hitting the wire. Extended lock-holding time until
5052 server acknowledges receiving data.
5055 Frequency : extremely rare. Never observed in practice.
5057 Description: avoid generating lustre_handle cookie of 0.
5058 Details : class_handle_hash() generates handle cookies by incrementing
5059 global counter, and can hit 0 occasionaly (this is unlikely, but
5060 not impossible, because initial value of cookie counter is
5061 selected randonly). Value of 0 is used as a sentinel meaning
5062 "unassigned handle" --- avoid it. Also coalesce two critical
5063 sections in this function into one.
5065 Severity : enhancement
5067 Description: allow liblustre clients to delegate truncate locking to OST
5068 Details : To avoid overhead of locking, liblustre client instructs OST to
5069 take extent lock in ost_punch() on client's behalf. New connection
5070 flag is added to handle backward compatibility.
5072 Severity : enhancement
5073 Bugzilla : 4928, 7341, 9758
5074 Description: allow number of OST service threads to be specified
5075 Details : a module parameter allows the number of OST service threads
5076 to be specified via "options ost ost_num_threads={N}" in the
5077 OSS's /etc/modules.conf or /etc/modprobe.conf.
5081 Bugzilla : 6146, 9635, 9895
5082 Description: servers crash with bad pointer in target_handle_connect()
5083 Details : In rare cases when a client is reconnecting it was possible that
5084 the connection request was the last reference for that export.
5085 We would temporarily drop the export reference and get a new
5086 one, but this may have been the last reference and the export
5087 was just destroyed. Get new reference before dropping old one.
5089 Severity : enhancement
5090 Frequency : if client is started with failover MDS
5092 Description: Allow multiple MDS hostnames in the mount command
5093 Details : Try to read the configuration from all specified MDS
5094 hostnames during a client mount in case the "primary"
5097 Severity : enhancement
5099 Description: Stop sending data to evicted clients as soon as possible.
5100 Details : Check if the client we are about to send or are sending data to
5101 was evicted already. (Check is done every second of waiting,
5102 for which l_wait_event interface was extended to allow checking
5103 of exit condition at specified intervals).
5106 Frequency : rare, normally only when NFS exporting is done from client
5108 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5110 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5113 Frequency : always, if srand() is called before liblustre initialization
5115 Description: Liblustre uses system PRNG disturbing its usage by user application
5116 Details : Introduce internal to lustre fast and high-quality PRNG for
5117 lustre usage and make liblustre and some other places in generic
5118 lustre code to use it.
5120 Severity : enhancement
5121 Bugzilla : 9477, 9557, 9870
5122 Description: Verify that the MDS configuration logs are updated when xml is
5123 Details : Check if the .xml configuration logs are newer than the config
5124 logs stored on the MDS and report an error if this is the case.
5125 Request --write-conf, or allow starting with --old_conf.
5127 Severity : enhancement
5129 Description: Handle symlinks in the path when checking if Lustre is mounted.
5130 Details : Resolve intermediate symlinks when checking if a client has
5131 mounted a filesystem to avoid duplicate client mounts.
5136 Description: lconf can hit an error exception but still return success.
5137 Details : The lconf command catches the Command error exception at the top
5138 level script context and will exit with the associated exit
5139 status, but doesn't ensure that this exit status is non-zero.
5144 Description: failure of ptlrpc thread startup can cause oops
5145 Details : Starting a ptlrpc service thread can fail if there are a large
5146 number of threads or the server memory is very fragmented.
5147 Handle this without oopsing.
5150 Frequency : always, only if liblustre and non-default acceptor port was used
5152 Description: liblustre cannot connect to servers with non-default acceptor port
5153 Details : tcpnal_set_default_params() was not called and was therefore
5154 ignoring the environment varaible TCPNAL_PORT, as well as other
5155 TCPNAL_ environment variables
5160 Description: two objects could be created on the same OST for a single file
5161 Details : If an OST is down, in some cases it was possible to create two
5162 objects on a single OST for a single file. No problems other
5163 than potential performance impact and spurious error messages.
5167 Bugzilla : 5681, 9562
5168 Description: Client may oops in ll_unhash_aliases
5169 Details : Client dcache may become inconsistent in race condition.
5170 In some cases "getcwd" can fail if the current directory is
5176 Description: Inode refcounting problems in NFS export code
5177 Details : link_raw functions used to call d_instantiate without obtaining
5178 extra inode reference first.
5182 Bugzilla : 9942, 9903
5183 Description: Referencing freed requests leading to crash, memleaks with NFS.
5184 Details : We used to require that call to ll_revalidate_it was always
5185 followed by ll_lookup_it. Also with revalidate_special() it is
5186 possible to call ll_revalidate_it() twice for the same dentry
5187 even if first occurence returned success. This fix changes semantic
5188 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5189 reference on a request referred from the intent.
5190 ll_intent_release() then releases such a request.
5193 Frequency : rare, normally benchmark loads only
5195 Description: unlinked inodes were kept in memory on the client
5196 Details : If a client is repeatedly creating and unlinking files it
5197 can accumulate a lot of stale inodes in the inode slab cache.
5198 If there is no other client load running this can cause the
5199 client node to run out of memory. Instead flush old inodes
5200 from client cache that have the same inode number as a new inode.
5203 Frequency : SLES9 2.6.5 kernel and long filenames only
5204 Bugzilla : 9969, 10379
5205 Description: utime reports stale NFS file handle
5206 Details : SLES9 uses out-of-dentry names in some cases, which confused
5207 the lustre dentry revalidation. Change it to always use the
5211 Frequency : rare, unless heavy write-truncate concurrency is continuous
5212 Bugzilla : 4180, 6984, 7171, 9963, 9331
5213 Description: OST becomes very slow and/or deadlocked during object unlink
5214 Details : filter_destroy() was holding onto the parent directory lock
5215 while truncating+unlinking objects. For very large objects this
5216 may block other threads for a long time and slow overall OST
5217 responsiveness. It may also be possible to get a lock ordering
5218 deadlock in this case, or run out of journal credits because of
5219 the combined truncate+unlink. Solution is to do object truncate
5220 first in one transaction without parent lock, and then do the
5221 final unlink in a new transaction with the parent lock. This
5222 reduces the lock hold time dramatically.
5225 Frequency : rare, 2.4 kernels only
5227 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5228 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5229 before unmounting their filesystem in order to avoid deadlock.
5230 The kernel_locked() function in 2.4 kernels only checks whether
5231 the kernel lock is held, not whether it is this process that is
5232 holding it as 2.6 kernels do.
5237 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5238 Details : The client ptlrpc code may be trying to reconnect to a down
5239 server before a previous connection attempt has timed out.
5240 Increase the reconnect interval to be longer than the connection
5241 timeout interval to avoid sending duplicate connections to
5245 Frequency : echo_client brw_test command
5247 Description: fix echo_client to work with OST preallocated code
5248 Details : OST preallocation code (5137) didn't take echo_client IO path
5249 into account: echo_client calls filter methods outside of any
5250 OST thread and, hence, there is no per-thread preallocated
5251 pages and buffers to use. Solution: hijack pga pages for IO. As
5252 a byproduct, this avoids unnecessary data copying.
5256 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5257 Description: Client can oops in mdc_commit_close() after open replay
5258 Details : It was possible for the MDS to return an open request with no
5259 transaction number in mds_finish_transno() if the client was
5260 evicted, but without actually returning an error. Clients
5261 would later try to replay that open and may trip an assertion
5262 Simplify the client close codepath, and always return an error
5263 from the MDS in case the open is not successful.
5266 Frequency : rare, 2.6 OSTs only
5268 Description: OST may deadlock under high load on fragmented files
5269 Details : If there was a heavy load and highly-fragmented OST filesystems
5270 it was possible to have all the OST threads deadlock waiting on
5271 allocation of biovecs, because the biovecs were not released
5272 until the entire RPC IO was completed. Instead, release biovecs
5273 as soon as they are complete to ensure forward IO progress.
5275 Severity : enhancement
5277 Description: Support for specifying external journal device at mount
5278 Details : If an OST or MDS device is formatted with an external journal
5279 device, this device major/minor is stored in the ext3 superblock
5280 and may not be valid for failover. Allow detecting and
5281 specifying the external journal at mount time.
5286 Description: Mounting an MDS with pending unlinked files may cause oops
5287 Details : target_finish_recovery() calls mds_postrecov() which returned
5288 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5289 considers this an error and immediately begins cleaning up the
5290 lov, just after starting the mds_lov process
5292 Severity : enhancement
5294 Description: Implement 'lfs df' to report actual free space on per-OST basis
5295 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5296 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5297 usage of objects; '-h' to report in human readable format.
5299 ------------------------------------------------------------------------------
5301 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5308 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5309 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5310 at startup during orphan unlinking. Instead, skip buddy update
5311 before setup, as it will be regenerated after recovery is complete.
5316 Description: If an OST is inactive, its locks might reference stale inodes.
5317 Details : lov_change_cbdata() must iterate over all namespaces, even if
5318 they are inactive to clear inode references from the lock.
5320 Severity : enhancement
5321 Frequency : occasional, if non-standard max_dirty_mb used
5323 Description: Client will block write RPCs if not enough grant
5324 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5325 then the client will block writes while waiting for another RPC
5326 to complete instead of consuming its dirty limit. With change
5327 we get improved performance when max_dirty_mb is small.
5329 Severity : enhancement
5330 Bugzilla : 3389, 6253
5331 Description: Add support for supplementary groups on the MDS.
5332 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5333 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5334 lookups for user supplementary groups into a cache.
5338 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5339 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5340 for liblustre O_EXCL is enforced by MDS.
5344 Description: readdir never returns NULL in liblustre.
5345 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5346 dentry in struct dirent.
5350 Frequency : liblustre only, depends on application IO pattern
5351 Description: liblustre clients evicted if not contacting servers
5352 Details : Don't put liblustre clients into the ping_evictor list, so
5353 they will not be evicted by the pinger ever.
5355 Severity : enhancement
5357 Description: Add ability to evict clients by NID from MDS.
5358 Details : By echoing "nid:$NID" string into
5359 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5360 $NID would be instantly evicted from this MDS and from all active
5361 OSTs connected to it.
5365 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5366 Details : lookup_it_finish() used to query file size from OSTs that was not
5371 Description: service threads change working directory to that of init
5372 Details : Starting lustre service threads may pin the working directory
5373 of the parent thread, making that filesystem busy. Threads
5374 now change to the working directory of init to avoid this.
5378 Frequency : during shutdown only
5379 Description: shutdown with a failed MDS or OST can cause unmount to hang
5380 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5385 Frequency : 2.6 only
5386 Description: chmod/chown may include an extra supplementary group
5387 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5388 supplementary group and if none is specified this is used.
5391 Bugzilla : 5479 (6816)
5393 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5394 Details : If lookup is in progress on a file that is unlinked we might try
5395 to revalidate the inode and fail in revalidate after lookup is
5396 complete and ll_file_open() enqueues the open again but
5397 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5400 Frequency : always, if lconf --abort_recovery used
5402 Description: lconf --abort_recovery fails with 'Operation not supported'
5403 Details : lconf was attempting to abort recovery on the MDT device and not
5406 ------------------------------------------------------------------------------
5408 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5413 Frequency : rare (only unsupported configurations with a node running as an
5415 Bugzilla : 6514, 5137
5416 Description: Mounting a Lustre file system on a node running as an OST could
5418 Details : OSTs now preallocates memory needed to write out data at
5419 startup, instead of when needed, to avoid having to
5420 allocate memory in possibly low memory situations.
5421 Specifically, if the file system is mounted on on OST,
5422 memory pressure could force it to try to write out data,
5423 which it needed to allocate memory to do. Due to the low
5424 memory, it would be unable to do so and the node would
5425 become unresponsive.
5427 Severity : enhancement
5429 Description: Addition of lconf --service command line option
5430 Details : lconf now accepts a '--service <arg>' option, which is
5431 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5433 Severity : enhancement
5435 Description: Failover mode is now the default for OSTs.
5436 Details : By default, OSTs will now run in failover mode. To return to
5437 the old behaviour, add '--failout' to the lmc line for OSTs.
5439 Severity : enhancement
5441 Description: Health checks are now provided for MDS and OSTs
5442 Details : Additional detailed health check information on MSD and OSTs
5443 is now provided through the procfs health_check value.
5446 Frequency : occasional, depends on IO load
5448 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5449 after numerous create/delete cycles
5450 Details : The ext3 inode allocation policy would not allocate new inodes
5451 very well on the OSTs because there are no new directories
5452 being created. Instead we look for groups with free space if
5453 the parent directories are nearly full.
5458 Description: Network or server problems during mount may cause partially
5459 mounted clients instead of returning an error.
5460 Details : The config llog parsing code may overwrite the error return
5461 code during mount error handling, returning success instead
5467 Description: MDS can fail to allocate large reply buffers
5468 Details : After long uptimes the MDS can fail to allocate large reply
5469 buffers (e.g. zconf client mount config records) due to memory
5470 fragmentation or consumption by the buffer cache. Preallocate
5471 some large reply buffers so that these replies can be sent even
5472 under memory pressure.
5476 Frequency : rare (liblustre)
5477 Description: fsx running with liblustre complained that using truncate() to
5478 extend the file doesn't work. This patch corrects that issue.
5479 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5480 ATTR_SIZE and lsm use in llu_setattr_raw.
5484 Frequency : rare, only 2.6 kernels
5485 Description: Unusual file access patterns on the MDS may result in inode
5486 data being lost in very rare circumstances.
5487 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5488 code in the 2.6 kernel caused the inode and/or EA data not to
5489 be read from disk, causing single-file corruption.
5493 Frequency : rare, only 2.6 filesystems using extents
5494 Description: Heavy concurrent write and delete load may cause data corruption.
5495 Details : It was possible under high-load situations to have an extent
5496 metadata block in the block device cache from a just-unlinked
5497 file overwrite a newly-allocated data block. We now unmap any
5498 metadata buffers that alias just-allocated data blocks.
5502 Frequency : filesystems with default stripe_count larger than 77
5503 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5504 Details : lconf specifies an inode size of 4096 bytes when the default
5505 stripe_count is larger than 77. This conflicts with the default
5506 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5507 this case to avoid pinning too much memory for large EAs.
5509 ------------------------------------------------------------------------------
5511 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5516 Frequency : rare (extremely heavy IO load with hundreds of clients)
5518 Description: Client is evicted, gets IO error writing to file
5519 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5520 caused clients to be evicted for AST timeouts. The fixes in
5521 bug 5192 mean we no longer need to have such short AST timeouts
5522 so ldlm_timeout has been increased.
5525 Frequency : occasional during --force or --failover shutdown under load
5526 Bugzilla : 5949, 4834
5527 Description: Server oops/LBUG if stopped with --force or --failover under load
5528 Details : a collection of import/export refcount and cleanup ordering
5529 issues fixed for safer force cleanup
5532 Frequency : only filesystems larger than 120 OSTs
5533 Bugzilla : 5990, 6223
5534 Description: lfs getstripe would oops on a very large filesystem
5535 Details : lov_getconfig used kfree on vmalloc'd memory
5538 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5539 Bugzilla : 6242, 6243
5540 Description: reading from files that had been truncated to a non-zero size
5541 but never opened returned no data
5542 Details : ll_file_read() reads zeros from no-object files to EOF
5547 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5548 Details : There's little harm in aborting MDS/OSS recovery and letting it
5549 try again, so I removed the LASSERT and return an error instead.
5551 Severity : enhancement
5553 Description: New debugging infrastructure for tracking down data corruption
5554 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5555 (b) cover more of the client-side code path, and (c) try to narrow
5556 down where problems occurred
5560 Bugzilla : 3819, 4364, 4397, 6313
5561 Description: Racing close and eviction MDS could cause assertion in mds_close
5562 Details : It was possible to get multiple mfd references during close and
5563 client eviction, leading to one thread referencing a freed mfd.
5565 Severity: : enhancement
5566 Bugzilla : 3262, 6359
5567 Description: Attempts to reconnect to servers are now more aggressive.
5568 Details : This builds on the enhanced upcall-less recovery that was added
5569 in 1.4.2. When trying to reconnect to servers, clients will
5570 now try each server in the failover group every 10 seconds. By
5571 default, clients would previously try one server every 25 seconds.
5576 Description: After recovery, certain operations trigger a failed
5577 assertion on a client.
5578 Details : Failing over an mds, using lconf -d --failover, while a
5579 client was doing a readdir() call would cause the client to
5580 LBUG after recovery completed and the readdir() was resent.
5582 Severity : enhancement
5584 Description: Default groups are now added by lconf
5585 Details : You can now run lconf --group <servicename> without having to
5586 manually add groups with lmc.
5589 Frequency : occasional
5591 Description: Nodes with an elan id of 0 trigger a failed assertion
5594 Frequency : always when accessing e.g. tty/console device nodes
5596 Description: tty and some other devices nodes cannot be used on lustre
5597 Details : file's private_data field is used by device data and lustre
5598 values in there got lost. New field was added to struct file to
5599 store fs-specific private data.
5602 Frequency : when exporting Lustre via NFS
5604 Description: NFSD failed occasionally when looking up a path component
5605 Details : NFSD is looking up ".." which was broken in ext3 directories
5606 that had grown large enough to become hashed.
5609 Frequency : Clusters with multiple interfaces not on the same subnet
5611 Description: Nodes will repeatedly try to reconnect to an interface which it
5612 cannot reach and report an error to the log.
5613 Details : Extra peer list entries will be created by lconf with some peers
5614 unreachable. lconf now validates the peer before adding it.
5617 Frequency : Only if a default stripe is set on the filesystem root.
5619 Description: Setting a default stripe on the filesystem root prevented the
5620 filesystem from being remounted.
5621 Details : The client was sending extra request flags in the root getattr
5622 request and did not allocate a reply buffer for the dir EA.
5625 Frequency : occasional, higher if lots of files are accessed by one client
5626 Bugzilla : 6159, 6097
5627 Description: Client trips assertion regarding lsm mismatch/magic
5628 Details : While revalidating inodes the VFS looks up inodes with ifind()
5629 and in rare cases can find an inode that is being freed.
5630 The ll_test_inode() code will free the lsm during ifind()
5631 when it finds an existing inode and then the VFS later attaches
5632 this free lsm to a new inode.
5636 Bugzilla : 6422, 7030
5637 Description: MDS deadlock between mkdir and client eviction
5638 Details : Creating a new file via mkdir or mknod (starting a transaction
5639 and getting the ns lock) can deadlock with client eviction
5640 (gets ns lock and trying to finish a synchronous transaction).
5643 Frequency : occasional
5644 Description: While starting a server, the fsfilt_ext3 module could not be
5646 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5647 kernels. Previously, lconf would still use the ext3 name
5648 when trying to load modules. Now, it will correctly use
5649 ext3 on 2.4 and ldiskfs on 2.6.
5651 Severity : enhancement
5652 Description: The default stripe count has been changed to 1
5653 Details : The interpretation of the default stripe count (0, to lfs
5654 or lmc) has been changed to mean striping across a single
5655 OST, rather than all available. For general usage we have
5656 found a stripe count of 1 or 2 works best.
5658 Severity : enhancement
5659 Description: Add support for compiling against Cray portals.
5660 Details : Conditional compiling for some areas that are different
5664 Frequency : occasional
5665 Bugzilla : 6409, 6834
5666 Description: Creating files with an explicit stripe count may lead to
5667 a failed assertion on the MDS
5668 Details : If some OSTs are full or unavailable, creating files may
5669 trigger a failed assertion on the MDS. Now, Lustre will
5670 try to use other servers or return an error to the
5674 Frequency : occasional
5676 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5677 caused lock timeout during readahead (since 1.4.2).
5678 Details : Processes doing readahead might match a lock that hasn't been
5679 granted yet if there are overlapping and conflicting lock
5680 requests. The readahead process waits on ungranted lock
5681 (original lock is CBPENDING), while OST waits for that process
5682 to cancel CBPENDING read lock and eventually evicts client.
5684 Severity : enhancement
5686 Description: Initial enabling of flock support for clients
5687 Details : Implements fcntl advisory locking and file status functions.
5688 This feature is provided as an optional mount flag (default
5689 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5690 locking are implemented yet, and those that are are not guaranteed
5691 to be completely correct in production environments.
5692 mount -t lustre -o [flock|noflock] ...
5695 Frequency : occasional
5697 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5698 assertion in the ext3 JBD (journaling) layer.
5699 Details : The b_committed_data struct is protected by the big kernel lock
5700 in 2.4 kernels, serializing journal_commit_transaction() and
5701 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5702 access to this struct. In 2.6 kernels there is finer grained
5703 locking to improve SMP performance of the JBD layer.
5707 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5712 Description: merge and cleanup kernel patches.
5715 -----------------------------------------------------------------------------
5717 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5719 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5720 versions of Lustre. Please update all servers and clients to
5721 version 1.4.2 or later at the same time. You must also run
5722 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5723 to update the configuration logs.
5725 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5726 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5727 - swab llog records of type '0' so we get proper header size/idx (5861)
5728 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5729 - fix rename of one directory over another leaking an inode (5953)
5730 - avoid SetPageDirty on 2.6 (5981)
5731 - don't re-add just-being-destroyed locks to the waiting list (5653)
5732 - when creating new directories, inherit the parent's custom
5733 striping settings if present parent (3048)
5734 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5735 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5736 - don't hold client locks on temporary worklist from l_lru (5666)
5737 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5738 - automatically evict dead clients (5921)
5739 - Update file size properly in create+truncate+fstat case (6196)
5740 - Do not unhash mountpoint dentries, do not allow removal of
5742 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5743 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5744 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5745 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5746 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5747 - Correctly handle reads of files with no objects (6243)
5748 - lctl recover will also mark a device active if deactivate used (5933)
5750 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5751 - allow --write-conf on an MDS with different nettype than client (5619)
5752 - don't write config llogs to MDS for mounts not from that MDS (5617)
5753 - lconf should create multiple TCP connections from a client (5201)
5754 - init scripts are now turned off by default; run chkconfig --on
5755 lustre and chkconfig --on lustrefs to use them
5756 - upcalls are no longer needed for clients to recover to failover
5758 - add --abort-recovery option to lconf to abort recovery on device
5760 - add support for an arbitrary number of OSTs (3026)
5761 - Quota support protocol changes.
5762 - forward compatibility changes to wire structs (6007)
5763 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5764 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5765 - improved "lustre status" script
5766 - initialize blocksize for non-regular files (6062)
5767 - added --disable-server and --disable-client configure options (5782)
5768 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5769 - Vanilla 2.4.29 support
5770 - increase maximum number of obd devices to 520 (6242)
5771 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5772 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5773 SLES 9 2.6.5 kernels
5774 - limit stripes per file to 160 (the maximum EA size) (6093)
5776 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5779 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5780 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5781 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5782 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5783 - small ext3 extents cleanups and fixes (5733)
5784 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5785 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5786 - avoid needless client->OST connect, fix handle mismatch (5317)
5787 - fix DLM error path that led to out-of-sync client, long delays (5779)
5788 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5789 - fix several locking issues related to i_size (5492,5624,5654,5672)
5790 - don't move pending lock onto export if it is already evicted (5683)
5791 - fix kernel oops when creating .foo in unlinked directory (5548)
5792 - fix deadlock in obdfilter statistics vs. object create (5811)
5793 - use time_{before,after} to avoid timer jiffies wrap (5882)
5794 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5795 - Do not leak request if server was not able to process it (5154)
5796 - If mds_open unable to find parent dir, make that negative lookup(5154)
5797 - don't create new directories with extent-mapping (5909, 5936)
5799 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5800 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5801 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5802 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5803 - support for RHEL4 kernel on i686 (5773)
5804 - provide error messages when incompatible logs are encountered (5898)
5806 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5807 * version 1.4.0.10 (1.4.1 release candidate 1)
5809 - don't keep a lock reference when lock is not granted (4238)
5810 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5811 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5812 - drop import inflight refcount on signal_completed_replay error (5255)
5813 - unlock page after async write error during send (3677)
5814 - handle missing objects in filter_preprw_read properly (5265)
5815 - no transno return for symlink open, don't save no-trasno open (3440)
5816 - don't try to complete elan receive that already failed (4012)
5817 - free RPC server reply state on error (5406)
5818 - clean up thread from ptlrpc_start_thread() on error (5160)
5819 - readahead could read extra page into cache that wasn't ejected (5388)
5820 - prevent races in class_attach/setup/cleanup/detach (5260)
5821 - don't dereference de->d_inode after l_dput of de (5458)
5822 - use "int" for stripe value returned from lock_to_stripe (5544)
5823 - mballoc allocation and error-checking fixes in 2.6 (5504)
5824 - block device patches to fix I/O request sizes in 2.6 (5482)
5825 - look up hostnames for IB nals (5602)
5826 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5827 - don't start multiple acceptors for the same port (5277)
5828 - fix incorrect LASSERT in mds_getattr_name (5635)
5829 - export a proc file for general "ping" checking (5628)
5830 - fix "lfs check" to not block when the MDS is down (5628)
5832 - service request history (4965)
5833 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5834 - create an "evict_client" /proc entry on OSTs, like the MDS has
5835 - fix mount usage message, return errors per mount(8) (5168)
5836 - change grep [] to grep "[]" in tests so they work in more UMLs
5837 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5838 - remove ancient LOV_MAGIC_V0 EA support (5047)
5839 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5840 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5841 - allow readahead allocations to fail when low on memory (5383)
5842 - mmap locking landed again, after considerable improvement (2828)
5843 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5845 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5848 - send OST transaction number in read/write reply to free req (4966)
5849 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5850 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5851 - the watchdog thread now runs as interruptible (5246)
5852 - flock/lockf fixes (but it's still disabled, pending 5135)
5853 - don't use EXT3 constants in llite code (5094)
5854 - memory shortage at startup could cause assertion (5176)
5856 - reorganization of lov code
5857 - single portals codebase
5859 - add extents/mballoc support (5025)
5860 - direct I/O reads in the obdfilter (4048)
5861 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5863 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5866 - send OST transaction number in read/write reply to free req (4966)
5867 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5868 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5869 - the watchdog thread now runs as interruptible (5246)
5870 - handle missing objects in filter_preprw_read properly (5265)
5871 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5872 - drop import inflight refcount on signal_completed_replay error (5255)
5873 - unlock page after async write error during send (3677)
5874 - return original error code on reconstructed replies (3761)
5875 - no transno return for symlink open, don't save no-trasno open (3440)
5877 - add pid to ldlm debugging output (4922)
5878 - bump the watchdog timeouts -- we can't handle 30sec yet
5879 - extra debugging for orphan dentry/inode bug (5259)
5881 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5884 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5885 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5886 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5887 - fix NULL dereference in /proc/sys/portals/routes (4827)
5888 - allow failed mdc_close() operations to be interrupted (4561)
5889 - stop precreate on OST before MDS would time out on it (4778)
5890 - don't send partial-page writes before EOF from client (4410)
5891 - discard client grant for sub-page writes on large-page clients (4520)
5892 - don't free dentries not owned by NFS code, check generation (4806)
5893 - fix lsm leak if mds_create_objects() fails (4801)
5894 - limit debug_daemon file size, always print CERROR messages (4789)
5895 - use transno after validating reply (3892)
5896 - process timed out requests if import state changes (3754)
5897 - update mtime on OST during writes, return in glimpse (4829)
5898 - add mkfsoptions to LDAP (4679)
5899 - use ->max_readahead method instead of zapping global ra (5039)
5900 - don't interrupt __l_wait_event() during strace
5902 - add software watchdogs to catch hung threads quickly (4941)
5903 - make lustrefs init script start after nfs is mounted
5904 - fix CWARN/ERROR duplication (4930)
5905 - return async write errors to application if possible (2248)
5906 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5907 - print NAL number in %x format (4645)
5908 - update barely-supported suse-2.4.21-171 series (4842)
5909 - support for sles 9 %post scripts
5910 - support for building 2.6 kernel-source packages
5911 - support for sles km_* packages
5913 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5916 - ignore -ENOENT errors in osc_destroy (3639)
5917 - notify osc create thread that OSC is being cleaned up (4600)
5918 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5919 - reconstruct ost_handle() like mds_handle() (4657)
5920 - create a new thread to do import eviction to avoid deadlock (3969)
5921 - let lconf resolve symlinked-to devices (4629)
5922 - don't unlink "objects" from directory with default EA (4554)
5923 - hold socknal file ref over connect in case target is down (4394)
5924 - allow more than 32000 subdirectories in a single directory (3244)
5925 - fix blocks count for O_DIRECT writes (3751)
5926 - OST returns ENOSPC from object create when no space left (4539)
5927 - don't send truncate RPC if file size isn't changing (4410)
5928 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5929 - bind to privileged port in socknal and tcpnal (3689)
5931 - rate limit CERROR/CWARN console message to avoid overload (4519)
5932 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5933 - basic mmap support (3918)
5934 - kernel patch series update from b1_4 (4711)
5936 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5939 - avoid crash during MDS cleanup with OST shut down (2775)
5940 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5941 - don't use bad inodes on the MDS (3744)
5942 - dynamic object preallocation to improve recovery speed (4236)
5943 - don't hold spinlock over lock dumping or change debug flags (4401)
5944 - don't zero obd_dev when it is force cleaned (3651)
5945 - print grants to console if they go negative (4431)
5946 - "lctl deactivate" will stop automatic recovery attempts (3406)
5947 - look for existing locks in ldlm_handle_enqueue() (3764)
5948 - don't resolve lock handle twice in recovery avoiding race (4401)
5949 - revalidate should check working dir is a directory (4134)
5951 - don't always mark "slow" obdfilter messages as errors (4418)
5953 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5956 - don't close LustreDB during write_conf until it is done (3860)
5957 - fix typo in lconf for_each_profile (3821)
5958 - allow dumping logs from multiple threads at one time (3820)
5959 - don't allow multiple threads in OSC recovery (3812)
5960 - fix debug_size parameters (3864)
5961 - fix mds_postrecov to initialize import for llog ctxt (3121)
5962 - replace config semaphore with spinlock (3306)
5963 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5964 - don't allow enqueue to complete on a destroyed export (3822)
5965 - down write_lock before checking llog header bitmap (3825)
5966 - recover from lock replay timeout (3764)
5967 - up llog sem before sending rpc (3652)
5968 - reduce ns lock hold times when setting kms (3267)
5969 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5970 - fix NULL deref and obd_dev leak on setup error (3312)
5971 - replace some LBUG about llog ops with error handling (3841)
5972 - don't match INVALID dentries from d_lookup and spin (3784)
5973 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5974 - fix invalid assertion in ptlrpc_set_wait (3880)
5976 - add libwrap support for the TCP acceptor (3996)
5977 - add /proc/sys/portals/routes for non-root route listing (3994)
5978 - allow setting MDS UUID in .xml (2580)
5979 - print the stack of a process that LBUGs (4228)
5981 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5984 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5985 - make sure to unset replay flag from failed open requests (3440)
5986 - if default stripe count is 0, use OST count for inode size (3636)
5987 - update parent mtime/ctime on client for create/unlink (2611)
5988 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5989 - free recovery state on server during a forced cleanup (3571)
5990 - unregister_reply for resent reqs (3063)
5991 - loop back devices mounting and status check on 2.6 (3563)
5992 - fix resource-creation race that can provoke i_size == 0 (3513)
5993 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5994 - more debugging for page-accounting assertion (3746)
5995 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5996 - avoid deadlock after precreation failure (3758)
5997 - fix race and lock order deadlock in orphan handling (3450, 3750)
5998 - add validity checks when grabbing inodes from l_ast_data (3599)
6000 - add /proc/.../recovery_status to obdfilter (3428)
6001 - lightweight CDEBUG infrastructure, debug daemon (3668)
6002 - change default OSC RPC parameters to be better on small clusters
6003 - turn off OST read cache for files smaller than 32MB
6004 - install man pages and include them in rpms (3100)
6005 - add new init script for (un)mounting lustre filesystems (2593)
6006 - run chkconfig in %post for init scripts (3701)
6007 - drop scimac NAL (unmaintained)
6009 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6012 - clean kiobufs before and after use (3485)
6013 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6014 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6015 - change init script to fail more clearly if not run as root (1528)
6016 - allow clients to reconnect during replay (1742)
6017 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6018 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6019 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6020 - fix cond_resched (3554)
6021 - only evict unfinished clients after recovery (3515)
6022 - allow bulk resend, prevent data loss (3570)
6023 - dynamic ptlrpc request buffer allocation (2102)
6024 - don't allow unlinking open directory if it isn't empty (2904)
6025 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6026 - remove extraneous obd dereference causing LASSERT failure (3334)
6027 - don't use get_cycles() when creating temp. files on the mds (3156)
6028 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6029 - handle EEXIST for set-stripe, set proper directory name (3336)
6031 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6032 - fix ksocknal_fmb_callback() error messages (2918)
6034 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6037 - don't copy lvb into (possibly NULL) reply on error (2983)
6038 - don't deref dentry after dput, don't free lvb on error (2922)
6039 - use the kms to determine writeback rpc length (2947)
6040 - increment oti_logcookies when osc is inactive (2948)
6041 - update client's i_blocks count via lvb messages (2543)
6042 - handle intent open/close of special files properly (1557)
6043 - mount MDS with errors=remount-ro, like obdfilter (2009)
6044 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6045 - don't use cancelling-locks' kms values (2947)
6046 - use highest lock extent for kms, not last one (2925)
6047 - don't dereference ERR_PTR() dentry in error handling path (3107)
6048 - fix thread race in portals_debug_dumplog() (3122)
6049 - create lprocfs device entries at setup instead of at attach (1519)
6050 - common AST error handler, don't evict client on completion race (3145)
6051 - zero nameidata in detach_mnt in 2.6 (3118)
6052 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6053 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6054 - initialize RPC timeout timer earlier for 2.6 (3219)
6055 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6056 - print nal/nid for unknown nid (3258)
6057 - additional checks for oscc recovery before doing precreate (3284)
6058 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6059 - don't crash in mdc_close for bad permissions on open (3285)
6060 - zero i_rdev for non-device files (3147)
6061 - clear page->private before handing to FS, better assertion (3119)
6062 - tune the read pipeline (3236)
6063 - fix incorrect decref of invalidated dentry (2350)
6064 - provide read-ahead stats and refine rpc in flight stats (3328)
6065 - don't hold journal transaction open across create RPC (3313)
6066 - update atime on MDS at close time (3265)
6067 - close LDAP connection when recovering to avoid server load (3315)
6068 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6069 - don't leak open file on MDS after open resend (3325)
6070 - serialize filter_precreate and filter_destroy_precreated (3329)
6071 - loop device shouldn't call sync_dev() for nul device (3092)
6072 - clear page cache after eviction (2766)
6073 - resynchronize MDS->OST in background (2824)
6074 - refuse to mount the same filesystem twice on same mountpoint (3394)
6075 - allow llmount to create routes for mounting behind routers (3320)
6076 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6077 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6078 - fix rare problem with rename on htree directories (3417)
6080 - allow default OST striping configuration per directory (1414)
6081 - fix compilation for qswnal for 2.6 kernels (3125)
6082 - increase maximum number of MDS request buffers for large systems
6083 - change liblustreapi to be useful for external progs like lfsck (3098)
6084 - increase local configuration timeout for slow disks (3353)
6085 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6087 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6090 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6091 - don't overwrite extent policy data in reply if lock was blocked (2901)
6092 - drop filter export grants atomically with removal from device (2663)
6093 - del obd_self_export from work_list in class_disconnect_exports (2908)
6094 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6095 - swab reply message in mdc_close, other PPC fixes (2464)
6096 - fix destroying of named logs (2325)
6097 - overwrite old logs when running lconf --write_conf (2264)
6098 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6099 - fix race in target_handle_connect (2898)
6100 - mds_reint_create() should take same inode create lock (2926)
6101 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6102 - don't close files for self_export to avoid uninitialized obd (2936)
6103 - allow MDS with the same name as client node (2939)
6104 - hold dentry reference for closed log files for unlink (2325)
6105 - reserve space for all logs during transactions (2059)
6106 - don't evict page beyond end of stripe extent (2925)
6107 - don't oops on a deleted current working directory (2399)
6108 - handle hard links to targets without a parent properly (2517)
6109 - don't dereference NULL lock when racing during eviction (2867)
6110 - don't grow lock extents when lots of conflicting locks (2919)
6112 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6115 - account for cache space usage on clients to avoid data loss (974)
6116 - lfsck support in lustre kernel code (2349)
6117 - reduce journal credits needed for BRW writes (2370)
6118 - orphan handling to avoid losing space on client/server crashes
6119 - ptlrpcd can be blocked, stopping ALL progress (2477)
6120 - use lock value blocks to assist in proper KMS, faster stat (1021)
6121 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6122 - recovery for initial connections (2355)
6123 - fixes for mds_cleanup_orphans (1934)
6124 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6125 - block all file creations until orphan recovery completes (1901)
6126 - client remove rq_connection from request struct (2423)
6127 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6128 - recovery timer race (2670)
6129 - mdc_close recovey bug (2532)
6130 - ptlrpc cleanup bug (2710)
6131 - mds timeout on local locks (2588)
6132 - namespace lock held during RPCs (2431)
6133 - handle interrupted sync write properly (2503)
6134 - don't try to handle a message that hasn't been replied to (2699)
6135 - client assert failure during cleanup after abort recovery (2701)
6136 - leak mdc device after failed mount (2712)
6137 - ptlrpc_check_set allows timedout requests to complete (2714)
6138 - wait for inflight reqs when ptlrpcd finishes (2710)
6139 - make sure unregistered services are removed from the srv_list
6140 - reset bulk XID's when resending them (caught by 1138 test)
6141 - unregister_bulk after timeout
6142 - fix lconf error (2694)
6143 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6144 - readahead locks pages, leaves pending causing memory pressure (2673)
6145 - increase OST request buffers to 4096 on large machines (2729)
6146 - fix up permission of existing directories in simple_mkdir (2661)
6147 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6148 - don't assign transno to errored transactions (2742)
6149 - don't delete objects on OST if given a bogus objid from MDS (2751)
6150 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6151 - if rq_no_resend, then timeout request after recovery (2432)
6152 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6153 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6154 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6155 - MDS recovery shouldn't delete objects that it has given out (2730)
6156 - if enqueue arrives after completion, don't clobber LVB (2819)
6157 - don't unlock pages twice when trigger_group_io returns error (2814)
6158 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6159 - don't write pages to disk if there was an error (1450)
6160 - don't ping imports that have recovery disabled (2676)
6161 - take buffered bytes into account when balancing socknal conn (2817)
6162 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6163 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6164 - remove little-endian swabbing of llog records (1987)
6165 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6166 - retry reposting request buffers if they fail (1191)
6167 - grow extent at grant time to avoid granting a revoked lock (2809)
6168 - lock revoke doesn't evict page if covered by a second lock (2765)
6169 - disable VM readahead to avoid reading outside lock extents (2805)
6171 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6172 - updated kernel patches for hp-2.4.20 kernel (2681)
6174 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6177 - fix truncated write corruption (2366)
6178 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6180 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6181 - ensure len(uuid) < 37 in lmc (1171)
6182 - fix ia64 OOPS in llog_test (2255)
6183 - zero end of page at obdfilter for partial page writes (2648)
6184 - don't leave stale dentries around after renames (bug 2428)
6185 - fix timeouts when evicting a client with a single lock held (2642)
6186 - set deadline for the initial HELLO message to drain (2634)
6187 - print out dotted-quad IP addresses in the socknal (2302)
6189 - additional debugging for MDS client eviction problem (2443)
6190 - fix mkfsoptions support for osts (2603, 2604)
6192 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6195 - add series for the vanilla 2.6.0 kernel
6196 - add series for the vanilla 2.4.24 kernel
6197 - add series for a cray x86/64 UL kernel drop
6198 - fix xattr patches for the vanilla 2.4.19 series
6200 - generate true UUIDs in lmc (1171)
6201 - have portals stack dumping break in UML (2466)
6202 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6203 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6204 - wake callback waiting threads on client eviction (2460)
6205 - Add --ptldebug and --subsystem to lmc (1719)
6206 - update assertion to allow safe interrupt allocation
6207 - set rq_no_resend for cancel requests (2432)
6208 - recalculate ptlrpcd timeout after resend (2494)
6209 - call vfs_rmdir when removing pending directories (2368)
6210 - fix renaming a file to itself (2429)
6211 - lmc creates a default one-stripe lov (2454)
6212 - expand procfs space to handle large clusters (2326)
6213 - increase UML stack to avoid overflow
6214 - update lconf's list of debug and subsystem masks
6215 - fix lfs find --obd (2510)
6216 - /proc tunable for disabling filter read caching (2591)
6217 - stop rpm packages from altering slapd.conf (2301)
6218 - disable nagle in the socknal under 0conf (2578)
6219 - choose mds inode size based on stripe count (2572)
6220 - fix kernel-source rpm problems (2516)
6222 - add --disable-doc to avoid pdf generation (2421)
6223 - update documentation, tests, type-os, comments
6224 - avoid format warnings on ia64
6225 - remove the TOE NAL
6226 - tiny code cleanups by removing unused fields
6228 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6231 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6232 - remove the most blatant lies from BUILDING file (2371)
6233 - change default debug level to reasonable production setting
6234 - reduce client side cache size to reduce cache flush time
6235 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6236 - make TCP zerocopy and pinger support enabled by default (2476)
6237 - sync writes completed after process exits caused crashes (2319)
6238 - maintain correct mount count on the MDS (2356)
6239 - backout 1557, because 2316 wasn't really fixed
6240 - better file I/O statistics gathering in /proc
6241 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6242 - another kernel patch to fix zero-copy TCP function export
6243 - don't take duplicate lock when processing re-sent getattr (2420)
6244 - lctl uses obd_self_export instead of creating new conn (2353)
6245 - MDS/OST recovery case which requires object creation asserted (2425)
6246 - move lfs from /usr/sbin to /usr/bin in packages
6247 - fix race between mds_client_add and mds_client_free (2417)
6248 - use kmalloc instead of slabs in portals (2430)
6249 - don't create duplicate records when a failover MDS is present (2442)
6250 - remove unnecessary mount age check (2332)
6251 - don't remove directory inodes from locks prematurely (2451)
6252 - don't break if MDS service name is the same as hostname (2103)
6253 - fix races in client write RPC generation when cache full (2482)
6255 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6258 - remove now-unused request->rq_obd (278)
6259 - if an allocation fails, print out how much memory we've used (1933)
6260 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6261 - add the "configurable stack size" patch to most series files (1256)
6262 - ability to write large log records, for 100+ OST configs (2306)
6263 - fix NULL deref when filter_prep fails (2314)
6264 - fix operator precedence error in filter_sync
6265 - dynamic allocation of socknal TX descriptors (2315)
6266 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6267 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6268 - ensure that CWARN messages in Portals always get to the syslog
6269 - __init/__exit are not for prototype decls (ldlm_init/exit)
6270 - x86-64 compile warning fixes
6271 - fix gateway LMC keyword conflict (2318)
6272 - fix MDS lock inversions in getattr/reint paths (1844)
6273 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6274 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6275 - DLM race condition prevented some lock evictions (2328)
6276 - ENOMEM detection and retry on socknal sends (2230)
6277 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6278 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6279 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6280 - better error messages when a client is rejected during recovery (1505)
6281 - avoid cancelling locks which were never granted, after failure (2330)
6282 - fix i_sem/journal inversion in mds_client_add (2333)
6283 - fix truncate/getattr lock cycle deadlock (2334)
6284 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6285 - fix two rare exit paths which could leak an l_lock() ref (2321)
6286 - fencepost error in MDS/OST orphan recovery (2226)
6287 - make log record alignment 8 bytes (1988)
6288 - lstripe now fails when requested offset > ost_count (2237)
6289 - ensure that all kernel series have a complete list.h (1607)
6290 - fix crashes in special-file operations (2316)
6291 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6293 - allow configurable automake binary, for testing new versions
6294 - small update to the lfs documentation
6296 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6298 * fix negative export reference count in fsfilt_sync (2312)
6300 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6301 * release candidate 0.9.1
6303 - orphans are moved into the PENDING directory for possible recovery
6304 - replayed opens now open by fid for orphan/rename safety (1042)
6305 - last close of an orphan inode generates a transno (683)
6306 - chdir() and mount() now pin the directory entry (1020)
6307 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6308 - discard very old requests without processing them (1502)
6309 - remove some common, well-understood CERRORs (1505)
6310 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6311 - clear "grant" flags in OST replies until OST grant code lands (1644)
6312 - fix read performance by not clobbering i_blksize on client (1598)
6313 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6314 - make LRU size a /proc tunable, clears locks when reduced (707)
6315 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6316 - server locks take references on exports now (1558)
6317 - build fixes for 2.4.20-rh trees (1663)
6318 - return an error from lov_create if all OSCs are inactive (1751)
6319 - fix import levels when a reconnect happens without a timeout (1597)
6320 - exit early from mds_open if we get a lookup error (1749)
6321 - partial page read at EOF wouldn't wait for disk before sending (1642)
6322 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6323 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6324 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6325 - write/truncate lock inversion (1639)
6326 - Don't auto-load obdclass, portals modules during cleanup (1495)
6327 - fix timestamps from jumping to "now" (1763)
6328 - extra journal assertions (1648)
6329 - add an extra multiunlink test (1771)
6330 - fix read_record/write_record API (1776)
6331 - fix leak of offset_extent, possible incorrect i_size later (1772)
6332 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6333 - Debugging for the kqswnal_get_idle_tx problems (1820)
6334 - Allow recovery to be attempted multiple times (1536)
6335 - Write out MDS last_rcvd file after it is first created (1600)
6336 - Fix tx_descriptor leak in failed transmit situations (1827)
6337 - ext3 journaling fixes for assertion failure after IO error (1871)
6338 - class_export_put() on freed export after completion AST error (1896)
6339 - Fix revalidate looping in VFS (1322)
6340 - Don't access a freed export during MDS_REINT timeout (1521)
6341 - Add open-unlink recovery support on the MDS (1673,1764)
6342 - Return an error if no MDS data was read from last_rcvd (1946)
6343 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6344 - Don't setup a disk device that doesn't match exported UUID (317)
6345 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6346 - avoid committing NULL handle in force close
6347 - local.sh is now a one-stripe LOV configuration
6348 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6349 - fix ext3 htree duplicate directory entry corruption (1516)
6350 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6351 - update to new LOV EA format (2097)
6352 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6354 2003-06-15 Phil Schwan <phil@clusterfs.com>
6357 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6358 - per-import/export recovery handling (958, 931, 959)
6359 - multiple last-rcvd slots, for serving multiple FSes (949)
6360 - connections are again shared between multiple imp/exports (963, 964)
6361 - "umount -f" would hang if any requests needed to be sent (393, 978)
6362 - avoid pinning large req buffer by copying for queued messages (989)
6363 - add "uuid" to "lctl device" command to help upcalls (991)
6364 - "open" RPCs with transnos would confuse recovery counters (1037)
6365 - do proper endian conversion of all wire messages (288, 340, 891)
6366 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6367 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6368 - fix ptlrpc_connection leak in target_handle_connect (1174)
6369 - fix import refcounting bug in OST and MDS cleanup (1134)
6370 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6371 - fix very unlikely obd_types race condition (501)
6372 - remove osc_open hack for echo_client (1187)
6373 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6374 - a failure in read_inode2 leads to deadlock (1139)
6375 - cancel ack-locks as soon as transaction is committed (1072)
6376 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6377 - make sure to commitrw after any preprw to avoid deadlock (1162)
6378 - failing to execute a file in a lustre FS would lock inode (1203)
6379 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6380 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6381 - obd_uuid2tgt would walk off the end of the list (1255)
6382 - on IA64 the osc would give portals incorrect bulk size (1258)
6383 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6384 - fix lock inversion caused by new llite matching code (1282)
6385 - limit the number of dirty pages on a client to 10MB (1286)
6386 - timed out locks were not being corrected cancelled (1289)
6387 - fix O_DIRECT above 4GB on IA-32 (1292)
6388 * major user-visible changes
6389 - fail out/fail over policy now controlled by the upcall (993)
6391 - add OBD_PING to check server availability and failure (954)
6392 - lustre messages are now sent in sending host order (288, 340, 891)
6393 - add eadatalen to MDS getattr reply (340)
6394 - OST read replies may contain second buffer, with per-page status (593)
6396 2003-03-11 Phil Schwan <phil@clusterfs.com>
6399 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6400 - failed open()s could cause deadlock; fixed (867, 869)
6401 - stop cancelling OST locks when files are closed (481)
6402 - overlapping XID spaces caused network corruption (851, 853)
6403 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6404 - setattr_raw added, to do single-RPC, server-side setattrs
6405 - lmc/lconf syntax change for OST UUIDs
6406 - fix crashy race condition between ptlrpc_free_req and osc_close
6407 - don't use request in mdc_enqueue if we hit a timeout (889)
6408 - don't set the inode i_size for regular files from the MDS (896)
6409 - handle out of order completion AST (842)
6410 - don't LBUG if a lock request times out after receiving AST (913)
6411 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6412 - if a bad lock AST arrives, send an error instead of dropping entirely
6413 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6414 - fix leak in bulk IO when only partially completed (899, 900, 926)
6415 - fix O_DIRECT for ia64 (55)
6416 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6417 - C-z after timeout could hang a process forever; fixed (977)
6419 - client-side I/O cache (678, 924, 929, 941, 970)
6421 - READPAGE and SETATTRs which don't take server-side locks get
6424 2003-02-11 Phil Schwan <phil@clusterfs.com>
6427 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6428 - Fix fsfilt_extN_readpage() to read a full page of directory
6429 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6430 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6431 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6432 - Most intent operations were removed, in favour of a new RPC mode
6433 that does a single RPC to the server and bypasses most of the VFS
6434 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6435 - Aggressively cancel local locks on DLM servers
6436 - mds_reint_unlink sends EA to the client if it's the last nlink.
6437 client uses that EA to unlink OST objects.
6438 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6439 - recursive symlinks were fixed (439)
6440 - fixed NULL deref in DEBUG_REQ
6441 - filter_update_lastobjid no longer calls sync, which annoyed extN
6442 - fixed multi-client small-writes to a single file problem (445)
6443 - fixed mtime updates during file writes (607)
6444 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6445 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6446 - fixed Linux OST size reporting problem (444, 656)
6447 - OST now updates object mtime with writes or setattr (607, 619)
6448 - client verifies file size before zeroing page past EOF (445)
6449 - OST now writes last allocated objid to disk with allocation (108)
6450 - LOV on echo now works (409)
6452 - mds_reint_unlink sends a new buffer, with the EA included. this
6453 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6454 set if a regular file was being unlinked, and it was the last link
6455 - use PtlGet from the target for bulk writes (315)
6456 - OST now updates object mtime with writes or setattr (607, 619)
6457 - LDLM now has a grant-time callback to revalidate locked items, if
6459 - Many MDS operations were reorganized to combat race conditions
6461 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6462 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6464 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6467 - Fully reactivate OST imports after reconnection (512, others)
6468 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6469 - More graceful error handling for truncating on dead OST (515)
6470 - Don't error out unless we're actually accessing dead stripes (474)
6471 - Fix garbage sizes when stripes are missing (410)
6472 - LRU counters were broken, causing constant lock purge (433, 432)
6473 - garbage on read from stripes with failed OSTs (441)
6474 - mark OSCs as active before reconnecting during recovery (438)
6475 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6476 - lfind did not preserve OST order in output (443)
6477 - symlinks cause hung clients, incorrect data (439)
6478 - stop dereferencing request after dropping refcount (457)
6479 - don't LASSERT(spin_is_locked) on non-SMP (455)
6480 - fixes for many rename() bugs
6481 - fstat didn't correctly synchronize attributes (399)
6482 - server must handle lock cancellation during blocking AST prep (487)
6483 - bulk descriptors were free()d too soon (511)
6484 - fix paths in lconf, which would load incorrect modules (451, 507)
6485 - fix confusing lconf 'host not found' error message (386)
6486 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6487 - fix race condition in mdc_blocking_ast() for inode access (526)
6488 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6489 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6490 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6491 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6492 - fix lstripe in several ways (kernel side) (527)
6493 - fix request leak in ldlm_cli_enqueue (262)
6494 - incorrect OSC was marked inactive after OST failure
6495 - call mds_fs_cleanup before unmounting filesystem (524)
6496 - fix races between taking ns_lock and ldlm_lock_change_resource
6497 - fix races updating LOV export open file list
6498 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6499 - fix recovery NULL deref in ldlm_cli_cancel_unused
6500 - fix some DLM races by using new hash table for lock handles (419)
6501 - permit the client to specify desired inodes, at replay
6502 - duplicate requests when we queue them for replay reintegration
6503 - fix last_rcvd offset calculation
6504 - sync after each recovered transaction, so we always make progress
6505 - never, not always, ERESTART requests without transnos
6506 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6508 - skip replay if the MDS says that the client is already connected
6509 - don't check for a recovery-enabled export to match lctl's UUID
6510 - don't INC_USE_COUNT for phantom exports
6511 - don't crash when cleaning up phantom exports (567)
6512 - don't double-finish or set replay data for errored mdc_open requests
6513 - abort requests when they time out, so we don't get old replies
6514 - send/receive replies for AST messages again
6515 - if the client says that it doesn't have the lock, cancel it on the
6517 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6518 mark it as destroyed, it will all work out when decref is called
6519 - fix module use counts (22, 581)
6521 - ASTs now expect a reply (server cancels lock on error reply)
6523 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6526 - fix many simultaneous client startup (392)
6527 - fix dentry->d_it clobbering
6528 - credentials weren't being shipped for readdir/getattr operations
6529 - remove invalid assertions triggered during some concurrent MD
6531 - proper Lustre versions added (336, 389)
6532 - fix memory leak for create error case (398)
6533 - fix LOV locking bug that would get cli/srv out of sync
6534 - fix echo client over LOV (409)
6535 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6536 - fix double-O_EXCL intent crash (424)
6537 - avoid sending multiple lock CANCELs (352)
6539 - MDS can do multi-client recovery (modulo bugs in new code)
6541 - many updates, edits, cleanups
6543 2002-11-18 Phil Schwan <phil@clusterfs.com>
6546 - fix null d_it dereference (346)
6547 - fix full OST/dbench hang (333)
6548 - fix permission problem with file removal (286)
6549 - fix removal of OSCs from LOV when they fail
6550 - fix NULL deref during bulk timeout (214)
6551 - fix problems related to multiple filesystems on one MDS (241)
6552 - fixed serious subtle metadata locking bugs
6553 - free locks on clients when inodes are removed due to memory
6555 - fix inode pointer in lock data (285)
6556 - partial support for multiple MDS on a single host (241)
6557 - data locks weren't cancelled at clear_inode time (290, 311)
6558 - intent locks could lead to unbounded lock growth (205)
6559 - added a maximum lock count, an LRU list, and a flusher
6560 - fix multiple rename (365)
6561 - properly abstracted the echo client
6562 - OSC locked 1 byte too many; fixed
6563 - rewrote brw callback code:
6564 - fixed recovery bugs related to LOVs (306)
6565 - fixed too-many-pages-in-one-write crash (191)
6566 - fixed (again) crash in sync_io_timeout (214)
6567 - probably fixed callback-related race (385)
6569 - Add capability to MDS protocol
6570 - LDLM cancellations and callbacks on different portals
6572 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6575 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6576 - defer open object destruction to close time (601981, 138)
6577 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6578 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6579 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6580 - don't call obd_setattr in ll_file_release for destroyed objects
6581 * protocol change to lustre_msg: move |version| and add |flags|
6582 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6583 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6584 * added replay of create, unlink, link and rename operations during
6585 MDS failover; recovery should be much more robust now
6586 * remove failed OSCs from LOVs (only lov_create uses this so far)
6587 * the lustre-HOWTO was brought (more) up to date (582544)
6589 2002-10-23 Phil Schwan <phil@clusterfs.com>
6592 - in-use dentries weren't being reused properly (617851)
6593 - prevent multiple LDLM setup (599178)
6594 - fix LOV size calculations for truncate (617853)
6595 - fix client handling of MDS intent errors (POSIX)
6596 - fix permission bug in lovstripe.c test (624321)
6597 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6598 - truncate past end of file could corrupt data
6599 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6600 - a race in recovery could return ETIMEDOUT to apps (623947)
6601 - building outside the source directory was fixed
6602 * the lustre-HOWTO was brought (more) up to date (582544)
6603 * major progress was made on recovery functionality
6605 2002-10-10 Phil Schwan <phil@clusterfs.com>
6608 - recovery deadlock fix
6609 - rm -rf causes LBUG fix (617817)
6610 - file open by multiple tasks fix (618962)
6611 - directory permissions bugs (602707 and 620007)
6612 - journal_stop fixed with locking (611313)
6613 - O_APPEND failures resolved (618273, perhaps 614459)
6614 - lconf PATH fix (619770)
6615 - IA64 build fix (621450)
6616 - RPC buffer sizes scale with amount of memory
6618 2002-10-01 Phil Schwan <phil@clusterfs.com>
6621 - locks would be cancelled without throwing away data pages,
6622 resulting in inconsistent data (605627)
6623 - inode attributes were not always being refreshed (605627, 612449)
6624 - lconf now continues to cleanup after lctl reports an error
6625 - MDS now enforces user permissions (602707)
6626 - lprocfs cleanup fixed, but not yet enabled (614157)
6627 - fixed infinite server hang, should a client not respond to an AST
6628 - avoid going into recovery if user calls readlink() with a buffer
6629 that's too small (613941)
6630 - AST RPCs no longer require replies (614867) -- this may be changed
6631 - don't crash server if client sends an IOV that's too big (611336)
6632 - fixed lock conversion deadlock (611892)
6633 - fixed the following of symlinks (614622)
6634 * recovery: the server can remove locks from a client that dies, other
6635 clients can make progress
6636 * more extN patch fixes
6637 * compile-time configurable ptlrpc buffer allocations
6639 - collaborative read cache document
6640 - Lustre Lite Performance CDR document-in-progress
6642 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6645 - fix typo in patch-2.4.18
6647 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6650 - clear ptlrpc request each time in handle_incoming_request()
6651 - unlink of files now destroys the object on the OST
6653 2002-09-19 Peter Braam <braam@clusterfs.com>
6655 * add hard link support
6656 * change obdfile creation method
6657 * kernel patch changed
6659 2002-09-19 Peter Braam <braam@clusterfs.com>
6662 - stack overflow bug in extN fixed
6664 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6666 * documentation updates
6667 - add man pages for config tools
6668 - update tests/README to describe testing with new config tools
6669 - finish metadata API descriptions
6670 * bug fixes and cleanups
6671 - statfs workaround for 16TB limit
6672 - LOV stripe allocation improved, can stripe on subset of OSTs
6673 - LOV file size/IO offset was wrong for files > 4GB in size
6674 - object EA data was being dropped, caused files to be unreadable
6675 - memory overflow with non-LOV OST caused memory corruption
6676 - fixed regression tests to work with new config tools, obdfilter
6677 - fixed bug when directory size became larger than 1 block
6678 - fixed bug (for single client case) when PWD was deleted
6679 - invalidate local directory pages when doing intent-based ops
6680 - avoid LDLM oops when lock callback contained bad data
6682 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6684 * documentation updates
6685 * bug fixes and cleanups
6686 - configuration tools
6689 - 64-bit compile warnings
6690 - 64-bit internal statfs data
6692 * test_brw on persistent OST devices
6694 * lprocfs (disabled)
6696 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6698 * documentation updates
6699 * bug fixes and cleanups
6700 * configuration tools
6702 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6705 * many small fixes to 0_5_4
6706 * io/network handling
6707 * thinkos in MDS operations
6709 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6712 * crucial basic fixes to 0.5.3
6713 * IOR, Iozone work over Elan
6716 2002-08-07 Phil Schwan <phil@clusterfs.com>
6717 * version 0_5_3, our first alpha
6718 * we use the new Portals iovs
6719 * documentation updates
6720 * bug fixes and cleanups
6721 * small changes in the DLM wire protocol
6723 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6724 * version 0_5_1 with some initial stability,
6725 * locking on MD and file I/O.
6726 * documentation updates
6727 * several bug fixes since 0.5.0
6728 * small changes in wire protocol
6730 2002-07-18 Phil Schwan <phil@clusterfs.com>
6732 * delivered as Lustre Light Alpha
6733 * fixed a crash after handling invalid MDS requests
6734 * fixed directory pages for architectures with non-4k pages sizes
6736 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6737 * release version v0_4_4
6738 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6740 * Fixes a number of interruption problems with OST operations.
6741 * Update documentation for portals header changes
6742 * Move all wire protocol structs/defines to lustre_idl.h
6743 * Fixes symlink length bug.
6744 * Add tcpdump to repository.
6746 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6747 * release version v0_4_3
6748 * Fixes statfs for inodes on extN.
6749 * Fixes bug in runtests which would delete /etc/hosts.
6750 * Use 64-bit object IDs wherever possible (not into VFS though)
6751 Remove ost_get_info, which is unused by lustre, and out of date.
6753 2002-07-03 Peter Braam <braam@clusterfs.com>
6754 * release version v0_4_2 Fixes a lookup error (type not passed)
6755 * move forward to head of Portals
6756 * move forward to latest Lustre kernel
6758 2002-06-25 Peter Braam <braam@clusterfs.com>
6759 * release version v0_4_1. Hopefully stable on single node use.