3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 3.10.0-1160.6.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.1.1.el8 (RHEL8.3)
16 4.4.120-92.70 (SLES12 SP2)
17 4.4.180-94.100 (SLES12 SP3)
18 4.4.0-131 (Ubuntu 16.04)
19 4.15.0-32 (Ubuntu 18.04)
20 5.4.0-48 (Ubuntu 20.04)
21 vanilla linux 5.4.0 (ZFS + ldiskfs)
22 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
23 * Client primary kernels built and tested during release cycle:
24 3.10.0-1062.18.1.el7 (RHEL7.7)
25 3.10.0-1127.19.1.el7 (RHEL7.8)
26 3.10.0-1160.6.1.el7 (RHEL7.9)
27 4.12.14-95.48 (SLES12 SP4)
28 4.12.14-122.20 (SLES12 SP5)
29 4.12.14-197.64 (SLES15 SP1)
30 5.3.18-24.24 (SLES15 SP2)
31 4.15.0-48 (Ubuntu 18.04)
32 * Other clients known to build on these kernels at some point (others may also work):
33 3.10.0-862.14.4.el7 (RHEL7.5)
34 3.10.0-957.27.2.el7 (RHEL7.6)
35 4.14.0-49.13.1.el7a (RHEL7.5)
36 4.14.0-115.2.2.el7a (RHEL7.6)
37 4.18.0-80.11.2.el8 (RHEL8.0)
38 4.18.0-147.8.1.el8 (RHEL8.1)
39 4.18.0-193.28.1.el8 (RHEL8.2)
40 4.18.0-240.1.1.el8 (RHEL8.3)
41 4.4.120-92.70 (SLES12 SP2)
42 4.4.180-94.100 (SLES12 SP3)
43 4.4.0-131 (Ubuntu 16.04)
44 5.4.0-37 (Ubuntu 20.04)
45 5.4.0 (vanilla kernel.org)
46 * Recommended e2fsprogs version: 1.45.6.wc3 or newer
47 * Recommended ZFS version: 2.0.0
48 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
49 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
50 stack overflow. For more information, please refer to bugzilla 17630.
51 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
52 the re-exporting nodes to be mounted with "32bitapi" mount option
54 --------------------------------------------------------------------------------
58 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
59 for currently supported client and server kernel versions.
60 * Server primary kernels built and tested during release cycle:
61 3.10.0-1062.1.1.el7 (RHEL7.7)
62 * Other server kernels known to build and work at some point (others may also work):
63 3.10.0-862.14.4.el7 (RHEL7.5)
64 3.10.0-957.27.2.el7 (RHEL7.6)
65 4.18.0-80.7.1.el8 (RHEL8.0)
66 4.4.120-92.70 (SLES12 SP2)
67 4.4.180-94.100 (SLES12 SP3)
68 4.4.0-131 (Ubuntu 16.04)
69 4.15.0-32 (Ubuntu 18.04)
70 vanilla linux 4.6.7 (ZFS only)
71 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
72 * Client primary unpatched kernels built and tested during release cycle:
73 3.10.0-1062.1.1.el7 (RHEL7.7)
74 4.12.14-95.29 (SLES12 SP4)
75 4.15.0-48 (Ubuntu 18.04)
76 * Other clients known to build on unpatched kernels at some point (others may also work):
77 3.10.0-862.14.4.el7 (RHEL7.5)
78 3.10.0-957.27.2.el7 (RHEL7.6)
79 4.14.0-49.13.1.el7a (RHEL7.5)
80 4.14.0-115.2.2.el7a (RHEL7.6)
81 4.18.0-80.7.1.el8 (RHEL8.0)
82 4.4.120-92.70 (SLES12 SP2)
83 4.4.180-94.100 (SLES12 SP3)
84 4.4.0-131 (Ubuntu 16.04)
85 vanilla linux 4.15.0 (ZFS only)
86 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
87 * Recommended ZFS / SPL version: 0.7.13
88 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
89 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
90 stack overflow. For more information, please refer to bugzilla 17630.
91 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
92 the re-exporting nodes to be mounted with "32bitapi" mount option
93 * Note for Ubuntu 14.04.5:
94 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
95 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
97 --------------------------------------------------------------------------------
101 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
102 for currently supported client and server kernel versions.
103 * Server known to build on the following kernels (others may also work):
104 2.6.32-696.18.7.el6 (RHEL6.9)
105 3.10.0-862.14.4.el7 (RHEL7.5)
106 3.10.0-957.el7 (RHEL7.6)
107 4.4.120-92.70 (SLES12 SP2)
108 4.4.162-94.69 (SLES12 SP3)
109 4.4.0-131 (Ubuntu 16.04)
110 4.15.0-32 (Ubuntu 18.04)
111 vanilla linux 4.6.7 (ZFS only)
112 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
113 * Client known to build on unpatched kernels (others may also work):
114 2.6.32-696.18.7.el6 (RHEL6.9)
115 2.6.32-754.6.3.el6 (RHEL6.10)
116 3.10.0-862.14.4.el7 (RHEL7.5)
117 3.10.0-957.el7 (RHEL7.6)
118 4.14.0-49.13.1.el7a (RHEL7.5)
119 4.14.0-115.2.2.el7a (RHEL7.6)
120 4.4.120-92.70 (SLES12 SP2)
121 4.4.162-94.69 (SLES12 SP3)
122 4.4.0-131 (Ubuntu 16.04)
123 4.15.0-32 (Ubuntu 18.04)
124 vanilla linux 4.15.0 (ZFS only)
125 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
126 * Recommended ZFS / SPL version: 0.7.9
127 * Tested with ZFS / SPL version: 0.7.9
128 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
129 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
130 stack overflow. For more information, please refer to bugzilla 17630.
131 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
132 the re-exporting nodes to be mounted with "32bitapi" mount option
133 * Note for Ubuntu 14.04.5:
134 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
135 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
137 --------------------------------------------------------------------------------
139 2018-04-03 Intel Corporation
141 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
142 for currently supported client and server kernel versions.
143 * Server known to build on patched kernels:
144 2.6.32-431.29.2.el6 (RHEL6.5)
145 2.6.32-504.30.3.el6 (RHEL6.6)
146 2.6.32-573.26.1.el6 (RHEL6.7)
147 2.6.32-642.15.1.el6 (RHEL6.8)
148 2.6.32-696.18.7.el6 (RHEL6.9)
149 3.10.0-693.21.1.el7 (RHEL7.4)
150 3.0.101-0.47.71 (SLES11 SP3)
151 3.0.101-107 (SLES11 SP4)
152 3.12.74-60.64.40 (SLES12 SP1)
153 4.4.114-92.64 (SLES12 SP2)
154 4.4.114-94.11 (SLES12 SP3)
155 3.13.0-101 (Ubuntu 14.04, ZFS only)
156 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
157 4.4.0-116 (Ubuntu 16.04)
158 vanilla linux 4.6.7 (ZFS only)
159 * Client known to build on unpatched kernels:
160 2.6.32-431.29.2.el6 (RHEL6.5)
161 2.6.32-504.30.3.el6 (RHEL6.6)
162 2.6.32-573.26.1.el6 (RHEL6.7)
163 2.6.32-642.15.1.el6 (RHEL6.8)
164 2.6.32-696.18.7.el6 (RHEL6.9)
165 3.10.0-693.21.1.el7 (RHEL7.4)
166 3.0.101-0.47.71 (SLES11 SP3)
167 3.0.101-107 (SLES11 SP4)
168 3.12.74-60.64.40 (SLES12 SP1)
169 4.4.114-92.64 (SLES12 SP2)
170 4.4.114-94.11 (SLES12 SP3)
171 3.13.0-101 (Ubuntu 14.04)
172 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
173 4.4.0-116 (Ubuntu 16.04)
175 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
176 * Recommended ZFS / SPL version: 0.7.6
177 * Tested with ZFS / SPL version: 0.7.6
178 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
179 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
180 stack overflow. For more information, please refer to bugzilla 17630.
181 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
182 the re-exporting nodes to be mounted with "32bitapi" mount option
183 * Note for Ubuntu 14.04.5:
184 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
185 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
187 --------------------------------------------------------------------------------
189 2017-07-05 Intel Corporation
191 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
192 for currently supported client and server kernel versions.
193 * Server known to build on patched kernels:
194 2.6.32-431.29.2.el6 (RHEL6.5)
195 2.6.32-504.30.3.el6 (RHEL6.6)
196 2.6.32-573.26.1.el6 (RHEL6.7)
197 2.6.32-642.15.1.el6 (RHEL6.8)
198 2.6.32-696.3.1.el6 (RHEL6.9)
199 3.10.0-514.21.1.el7 (RHEL7.3)
200 3.0.101-0.47.71 (SLES11 SP3)
201 3.0.101-100 (SLES11 SP4)
202 3.12.74-60.64.40 (SLES12 SP1)
203 4.4.59-92.17 (SLES12 SP2)
204 vanilla linux 4.6.7 (ZFS only)
205 * Client known to build on unpatched kernels:
206 2.6.32-431.29.2.el6 (RHEL6.5)
207 2.6.32-504.30.3.el6 (RHEL6.6)
208 2.6.32-573.26.1.el6 (RHEL6.7)
209 2.6.32-642.15.1.el6 (RHEL6.8)
210 2.6.32-696.3.1.el6 (RHEL6.9)
211 3.10.0-514.21.1.el7 (RHEL7.3)
212 3.0.101-0.47.71 (SLES11 SP3)
213 3.0.101-100 (SLES11 SP4)
214 3.12.74-60.64.40 (SLES12 SP1)
215 4.4.59-92.17 (SLES12 SP2)
217 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
218 * Recommended ZFS / SPL version: 0.6.5.9
219 * Tested with ZFS / SPL version: 0.6.5.9
220 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
221 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
222 stack overflow. For more information, please refer to bugzilla 17630.
223 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
224 the re-exporting nodes to be mounted with "32bitapi" mount option
226 --------------------------------------------------------------------------------
228 12-07-2016 Intel Corporation
230 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
231 for currently supported client and server kernel versions.
232 * Server known to build on patched kernels:
233 2.6.32-431.29.2.el6 (RHEL6.5)
234 2.6.32-504.30.3.el6 (RHEL6.6)
235 2.6.32-573.26.1.el6 (RHEL6.7)
236 2.6.32-642.11.1.el6 (RHEL6.8)
237 3.10.0-514.el7 (RHEL7.3)
238 3.0.101-0.47.71 (SLES11 SP3)
239 3.0.101-80 (SLES11 SP4)
240 3.12.62-60.64.8 (SLES12 SP1)
241 vanilla linux 4.5.7 (ZFS only)
242 * Client known to build on unpatched kernels:
243 2.6.32-431.29.2.el6 (RHEL6.5)
244 2.6.32-504.30.3.el6 (RHEL6.6)
245 2.6.32-573.26.1.el6 (RHEL6.7)
246 2.6.32-642.11.1.el6 (RHEL6.8)
247 3.10.0-514.el7 (RHEL7.3)
248 3.0.101-0.47.71 (SLES11 SP3)
249 3.0.101-80 (SLES11 SP4)
250 3.12.62-60.64.8 (SLES12 SP1)
252 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
253 * Recommended ZFS / SPL version: 0.6.5.8
254 * Tested with ZFS / SPL version: 0.6.5.8
255 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
256 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
257 stack overflow. For more information, please refer to bugzilla 17630.
258 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
259 the re-exporting nodes to be mounted with "32bitapi" mount option
261 --------------------------------------------------------------------------------
263 02-29-2016 Intel Corporation
265 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
266 for currently supported client and server kernel versions.
267 * Server known to build on patched 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)
274 vanilla linux 4.2.1 (ZFS only)
275 * Client known to build on unpatched kernels:
276 2.6.32-431.29.2.el6 (RHEL6.5)
277 2.6.32-504.30.3.el6 (RHEL6.6)
278 2.6.32-573.12.1.el6 (RHEL6.7)
279 3.10.0-327.3.1.el7 (RHEL7.2)
280 3.0.101-0.47.71 (SLES11 SP3)
281 3.0.101-68 (SLES11 SP4)
284 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
285 * Recommended ZFS / SPL version: 0.6.4.2
286 * Tested with ZFS / SPL version: 0.6.4.2
287 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
288 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
289 stack overflow. For more information, please refer to bugzilla 17630.
290 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
291 the re-exporting nodes to be mounted with "32bitapi" mount option
293 --------------------------------------------------------------------------------
295 03-10-2015 Intel Corporation
297 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
298 for currently supported client and server kernel versions.
299 * Server known to build on patched kernels:
300 2.6.32-431.29.2.el6 (RHEL6.5)
301 2.6.32-504.8.1.el6 (RHEL6.6)
302 3.0.101-0.46 (SLES11 SP3)
303 * Client known to build on unpatched kernels:
304 2.6.32-431.29.2.el6 (RHEL6.5)
305 2.6.32-504.8.1.el6 (RHEL6.6)
306 3.10.0-123.20.1.el7 (RHEL7)
307 3.0.101-0.46 (SLES11 SP3)
308 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
309 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
310 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
311 stack overflow. For more information, please refer to bugzilla 17630.
312 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
313 the re-exporting nodes to be mounted with "32bitapi" mount option
315 Severity : enhancement
317 Description: control OST-index in IDIF via ROCOMPAT flag.
318 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
319 last_rcvd file. For new formatted OST device, it will be auto set;
320 for the case of upgrading from old OST device, you can enable it
321 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
322 enabled, for new created OST-object, its IDIF-in-LMA will contain
323 the OST-index; for the existing OST-object, the OSD will convert
324 old format IDIF as new format IDIF with OST-index stored in the
325 LMA EA when accessing such OST-object or via OI scrub. Once such
326 flag is enabled, it cannot be reverted back, so the system cannot
327 be downgraded to the orignal incompatible version.
329 --------------------------------------------------------------------------------
331 07-30-2014 Intel Corporation
333 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
334 for currently supported client and server kernel versions.
335 * Server known to build on patched kernels:
336 2.6.32-431.20.3.el6 (RHEL6.5)
337 3.0.101-0.31 (SLES11 SP3)
338 * Client known to build on unpatched kernels:
339 2.6.32-431.20.3.el6 (RHEL6.5)
340 3.0.101-0.31 (SLES11 SP3)
343 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
344 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
345 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
346 stack overflow. For more information, please refer to bugzilla 17630.
347 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
348 the re-exporting nodes to be mounted with "32bitapi" mount option
350 --------------------------------------------------------------------------------
352 10-31-2013 Intel Corporation
354 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
355 * Server builds for kernels:
356 2.6.32-358.18.1.el6 (RHEL6.4)
357 * Client builds for unpatched kernels:
358 2.6.32-358.18.1.el6 (RHEL6.4)
359 3.0.92-0.8 (SLES11 SP3)
361 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
362 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
363 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
364 stack overflow. For more information, please refer to bugzilla 17630.
365 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
366 the re-exporting nodes to be mounted with "32bitapi" mount option
368 --------------------------------------------------------------------------------
370 05-30-2013 Intel Corporation
372 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
373 * Server builds for kernels:
374 2.6.32-358.6.2.el6 (RHEL6.4)
375 * Client builds for unpatched kernels:
376 2.6.32-358.6.2.el6 (RHEL6.4)
377 3.0.74-0.6.6 (SLES11 SP2)
379 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
380 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
381 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
382 stack overflow. For more information, please refer to bugzilla 17630.
383 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
384 the re-exporting nodes to be mounted with "32bitapi" mount option
386 --------------------------------------------------------------------------------
388 9-30-2012 Whamcloud, Inc.
390 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
391 * Server support for kernels:
392 2.6.32-279.5.1.el6 (RHEL6)
393 * Client support for unpatched kernels:
394 2.6.32-279.5.1.el6 (RHEL6)
396 2.6.18-238.19.1.el5 (RHEL5)
397 2.6.16.60-0.69.1 (SLES10)
398 * Recommended e2fsprogs version: 1.41.90.wc4
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
405 Severity : enhancement
407 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
409 Severity : enhancement
411 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
413 Severity : enhancement
415 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
417 Severity : enhancement
419 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
421 Severity : enhancement
423 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
425 --------------------------------------------------------------------------------
427 04-30-2012 Whamcloud, Inc.
429 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
430 * Server support for kernels:
431 2.6.32-220.4.2.el6 (RHEL6)
432 * Client support for unpatched kernels:
433 2.6.32-220.4.2.el6 (RHEL6)
435 2.6.18-238.19.1.el5 (RHEL5)
436 2.6.16.60-0.69.1 (SLES10)
437 * Recommended e2fsprogs version: 1.41.90.wc4
438 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
439 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
440 stack overflow. For more information, please refer to bugzilla 17630.
441 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
442 the re-exporting nodes to be mounted with "32bitapi" mount option
444 Severity : enhancement
446 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
448 Severity : enhancement
450 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
452 Severity : enhancement
454 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
456 Severity : enhancement
458 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
460 --------------------------------------------------------------------------------
462 09-30-2011 Whamcloud, Inc.
464 * Server support for kernels:
465 2.6.18-238.19.1.el5 (RHEL5)
466 2.6.32-131.6.1.el6 (RHEL6)
467 * Client support for unpatched kernels:
468 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
469 2.6.18-238.19.1.el5 (RHEL5)
470 2.6.32-131.6.1.el6 (RHEL6)
471 2.6.29.4-167.fc11 (FC11)
472 2.6.16.60-0.69.1 (SLES10)
474 * Recommended e2fsprogs version: 1.41.90.wc3
475 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
476 Lustre filesystem with 4K stack may cause a stack overflow. For more
477 information, please refer to bugzilla 17630.
478 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
479 mounted with "32bitapi" mount option
481 Severity : enhancement
482 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
484 Severity : enhancement
485 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
487 Severity : enhancement
488 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
491 Frequency : only with SLES10
493 Description: Use OFED "KMP" provided by Novell
494 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
495 Lustre is now built against this package. That means you need to
496 install the ofed-kmp package from Novell for the patchless client.
497 Note that the ofed-kmp that Novell ships may not exactly match the
498 kernel version but should still be compatible.
500 Severity : enhancement
502 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
503 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
505 --------------------------------------------------------------------------------
507 2010-07-15 Oracle, Inc.
509 * Server support for kernels:
510 2.6.18-164.11.1.el5 (RHEL 5)
511 2.6.18-164.11.1.0.1.el5 (OEL 5)
512 * Client support for unpatched kernels:
513 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
514 2.6.18-164.11.1.el5 (RHEL 5),
515 2.6.18-164.11.1.0.1.el5 (OEL 5)
516 2.6.16.60-0.42.8 (SLES 10),
518 2.6.29.4-167.fc11 (FC11)
519 * Recommended e2fsprogs version: 1.41.10-sun2
521 Severity : enhancement
523 Description: Distribute lfs_migrate script with lustre
524 Details : The lfs_migrate mentioned in the manual is now included in the
525 lustre tree in lustre/scripts and will distributed in lustre RPMs.
527 Severity : enhancement
529 Description: Failover nid list added to import proc file.
531 Severity : enhancement
533 Description: "som_preview" mount option required to use SOM
534 Details : The clients will need to use som_preview mount option
538 Description: Lustre does not do 1MB IOs to HW RAID
539 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
541 Severity : enhancement
543 Description: lfs quota output is not very convenient for awk/sed-parsing
544 Details : Some positions in lfs quota output table could be empty or
545 non-empty which made it hard to parse it with scripts, now
546 a dash is put instead of space where there is not supposed
551 Description: Update to OFED 1.5.1
552 Details : OFED 1.5.1 will be used in environments that don't use
553 vendor supplied OFED.
557 Description: Remove lnet. prefix from lctl params display
558 Details : Also renamed lnet.memused to lnet_memused to
563 Description: Use conf_param -d to remove permanant settings.
567 Description: Add list_param -R
568 Details : lctl list_param now supports recursive mode.
571 Bugzilla : 21829, 22312
572 Description: Fix llobdstat.
573 Details : 1. Now updating the write_bytes counter
574 2. Port script fixes from 1.8.
578 Description: decrease the usage of memory on clients.
579 Details : 1. On clients, recycle dentries and inodes unused.
580 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
585 Description: llobdstat fix and enhancement
586 Details : add a counter to set a limit to how many samples will be returned
587 fix a wildcard in the path to limit to obdfilter stats only
591 Description: lfs check only allowed for root
592 Details : Code cleanups and added a regresion test of non-root use of
597 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
598 Details : Fix hr->hr_index overload.
602 Description: Fix FIEMAP functionality.
606 Description: FC11 patchless client support added
610 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
611 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
615 Description: New acc-sm script and various test-framework enhancments.
619 Description: Previously landed patch was reverted.
623 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
627 Description: Add ldiskfs maxdirsize mount option.
631 Description: Fix mkfs.lustre support for 16TB LUNs.
635 Description: Print mmp_check_interval
636 Details : print mmp_check_interval and make it possible to abort
637 mount operation in case it takes too long.
641 Description: Add cascading_rw to lustre mpi tests.
645 Description: Store SOM data in EA.
646 Details : Also several SOM related cleanups.
650 Description: Add refcount to md_open_data
654 Description: Use refcount on lov_request_set to prevent memory corruption.
658 Description: Statahead: use parent dir's i_mutex for synchronization.
662 Description: loadgen improvements.
666 Description: Enable peer health detection in ksocklnd and ko2iblnd.
670 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
674 Description: Fix race on obd_devcie:md_stats
678 Description: Fix SLES 11 client builds.
679 Details : Resolved uninitalized pointer gcc warnings.
683 Description: Allow non-root access for "lfs check".
684 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
688 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
690 Severity : enhancement
692 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
693 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
697 Description: Sanity test_27p fails.
698 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
699 Sanity tests 27 were cleaned up and improved.
703 Description: Unify req format on client/servers
704 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
705 buffers and swabbers approach..
710 Description: drop join file stuff.
712 Severity : enhancement
714 Description: Add 2.6.27 and SLES11 (patchless client) support
718 Description: Fix in ptlrpc_expire_one_request() to print the signed time
721 Severity : enhancement
723 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
726 Severity : enhancement
728 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
730 Severity : enhancement
732 Description: correctly handle big reply message.
733 Details : send LNet event if reply is bigger than buffer and adjust this buffer
738 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
740 Severity : enhancement
742 Description: Add support for OEL5.
746 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
748 Severity : enhancement
750 Description: "weak-modules" support
751 Details : Implement "weak-modules" support which enables kernel modules
752 to be used with any kernel that implements the same kABI. In
753 order to achieve this modules are now installed in
754 /lib/modules/$(uname -r)/updates/kernel on all distributions.
758 Description: client could not reconnect to OST because of an active request.
759 Details : The client now retries to reconnect to the same server, if a
760 connect request failed with EBUSY or -EAGAIN. Test case 35b
761 conf-sanity.sh has been created.
765 Description: Don't allow make backward step on assignin osc next id.
766 Details : race between allocation next id and ll_sync thread can be cause
767 of set wrong osc next id and can be kill valid ost objects.
771 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
775 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
779 Description: Error handling in osc_statfs_interpret() has been improved.
780 Details : Check in osc_statfs_interpret() for EBADR.
784 Description: Do not send statfs() requests to OSTs disabled by administrator.
785 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
787 Severity : enhancement
789 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
791 Severity : enhancement
793 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
795 Severity : enhancement
797 Description: Enhancement for lfs(1) command to use numeric uid/gid.
801 Description: Improved cleanup procedure for conf-sanity test_46a.
805 Description: Do not start bulk transfer, if client tried to reconnect.
806 Details : Do not start bulk transfer, if client tried to reconnect,
807 but failed due to active rpcs. Check for exp_abort_active_req
808 has been implemented.
812 Description: Prevent inconsistences between linux and lustre mount structures.
813 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
814 Make the sleep interruptible.
816 Severity : enhancement
818 Description: provide server to client comms path
819 Details : server to client communications path via new LDLM_SET_INFO rpc
823 Description: truncate starts GFP_FS allocation under transaction causing deadlock
824 Details : ldiskfs_truncate calls grab_cache_page which may start page
825 allocation under an open transaction. This may lead to
826 calling prune_icache with consequent lustre reentrance.
830 Description: send correctly lsm on open replay
831 Details : MDS is trust to LSM size on replay open, but client can set wrong size
836 Description: enable client interface failover
837 Details : When a child reconnects from another NID, properly update export
838 nid hash position and ldlm reverse import.
840 Severity : enhancement
842 Description: Communicate OST degraded/readonly state via statfs to MDS
843 Details : Flags in the statfs returned from OSTs indicate whether the
844 OST is in a degraded RAID state, or if the filesystem has
845 turned read-only after a filesystem error is detected.
850 Description: don't panic if EPROTO was hit when reading symlink
851 Details : correctly handling request reference in error cases.
853 Severity : enhancement
855 Description: Add LustreNetLink, a kernel-userspace communcation path.
857 Severity : enhancement
859 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
864 Description: access to llog context before init.
865 Details : move handling CATALOGS file at osc layer and forbid access to llog
869 Frequency : rare, only when client is writing to OST on local node
871 Description: Avoid deadlock for local client writes
872 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
873 memory freeing context. This allows OST threads to set the
874 PF_MEMALLOC flag on task structures in order to allocate memory
875 from reserved pools and complete IO.
876 Use GFP_HIGHUSER for OST allocations for non-local client writes,
877 so that the OST threads generate memory pressure and allow
878 inactive pages to be reclaimed.
880 Severity : enhancement
882 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
887 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
888 Details : move ldlm namespace creation in setup phase to avoid grab
889 _lprocfs_lock with cli_sem held.
893 Description: Temporarily disable grant shrink.
894 Details : Disable the feature for debugging.
897 Frequency : only when formatting test filesystems
899 Description: Unable to run several mkfs.lustre on loop devices at the same time
900 Details : mkfs.lustre returns error 256 on the concurrent loop devices
901 formatting. The solution is to proper handle the error.
903 Severity : enhancement
905 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
907 Severity : enhancement
909 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
914 Description: resolve race between obd_disconnect and class_disconnect_exports
915 Details : if obd_disconnect will be called to already disconnected export he
916 forget release one reference and osc module can't unloaded.
918 Severity : enhancement
920 Description: Allow tuning service thread via /proc
921 Details : For each service a new
922 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
923 created that can be used to set min/max thread counts, and get the
924 current number of running threads.
929 Descriptoin: don't return error if have partially created objects for file.
930 Details : lov_update_create_set uses set->set_success as index for created
931 objects, so if some requests will be failed, they will have hole
932 at end of array and qos_shrink_lsm for allocate correct lsm.
934 Severity : enhancement
936 Description: Update OFED support to 1.4.1
941 Description: Reduce small size read RPC
942 Details : Set read-ahead limit for every file and only do read-ahead when
943 available read-ahead pages are bigger than 1M to avoid small size
946 Severity : enhancement
948 Description: move AT tunable parameters for more consistent usage
949 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
951 Severity : enhancement
953 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
954 Details : allow skip disconnected ost for send statfs request and hide error
960 Description: fix racy locking of mballoc block bitmaps causing BUG
961 Details : The locking of the mballoc buddy bitmap and the in-memory
962 block bitmap was using two different spin locks in some
963 cases. This made it possible to incorrectly access the
964 mballoc bitmap while another process was modifying it,
965 causing a sanity assertion to fail. While no on-disk corruption
966 was reported, there was some risk of this happening.
969 Frequency : rare, on llog test 6
971 Descriptoin: don't allow connect to already connected import
972 Details : allowing connect to already connected import is hide connecting problem.
977 Descriptoin: allow kill process which wait statahead result
978 Details : for some reasons 'ls' can stick in waiting result from statahead,
979 in this case need way for kill this process.
981 Severity : enhancement
983 Description: Add state history info file, enhance import info file
984 Details : Track import connection state changes in a new osc/mdc proc file;
985 add overview-type data to the osc/mdc import proc file.
987 Severity : enhancement
989 Description: MDS create should not wait for statfs RPC while holding DLM lock.
991 Severity : enhancement
993 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
998 Descriptoin: OST grant too much space to client even there are not enough space.
999 Details : Client will shrink its grant cache to OST if there are no write
1000 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1001 this grant cache if there are already not enough avaible space
1002 (left_space < total_clients * 32M).
1005 Frequency : start MDS on uncleanly shutdowned MDS device
1007 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1008 Details : stay in waiting mds<>ost recovery finished produce random bugs
1009 due race between two ll_sync thread for one lov target. send
1010 ACTIVATE event only if connect realy finished and import have
1014 Frequency : rare, connect and disconnect target at same time
1016 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1017 Details : don't call obd_disconnect under lov_lock. this long time
1018 operation and can block ptlrpcd which answer to connect request.
1023 Descriptoin: don't lose wakeup for imp_recovery_waitq
1024 Details : recover_import_no_retry or invalidate_import and import_close can
1025 both sleep on imp_recovery_waitq, but we was send only one wakeup
1029 Frequency : always with long access acl
1031 Descriptoin: mds can't pack reply with long acl.
1032 Details : mds don't control size of acl but they limited by reint/getattr
1035 Severity : enhancement
1037 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1039 Severity : enhancement
1041 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1044 Frequency : start MDS on uncleanly shutdowned MDS device
1046 Descriptoin: aborting recovery hang on MDS
1047 Details : don't throttle destroy RPCs for the MDT.
1050 Frequency : on remount
1052 Description: external journal device not working after the remount
1053 Details : clear dev_rdonly flag for external journal devices in
1059 Description: shutdown vs evict race
1060 Details : client_disconnect_export vs connect request race.
1061 if client will evicted at this time - we start invalidate
1062 thread without referece to import and import can be freed
1066 Frequency : rare, need acl's on inode.
1068 Description: client can't handle ost additional correctly
1069 Details : if ost was added after client connected to mds client can have
1070 hit lnet_try_match_md ... to big messages to wide striped files.
1071 in this case need teach client to handle config events about add
1072 lov target and update client max ea size at that event.
1074 Severity : enhancement
1076 Description: Changelogs
1077 Details : Changelogs are a lightweight mechanism to track filesystem
1078 metadata and namespace changes. The changelog is recorded
1079 permanently on the MDTs, and is periodically "consumed" / purged
1080 when records are no longer needed.
1082 Severity : enhancement
1084 Description: compact fld format with extents
1085 Details : Store range of seq rather than every seq in FLD. Seq
1086 controller update FLD rather than clients. In Case of CMD, mdt0
1087 has FLD, all other metadata server act as non persistent proxy
1088 for FLD queries and cache fld entries in fld cache.
1093 Description: don't skip ost target if they assigned to file
1094 Details : Drop slow OSCs if we can, but not for requested start idx.
1095 This means "if OSC is slow and it is not the requested
1096 start OST, then it can be skipped, otherwise skip it only
1097 if it is inactive/recovering/out-of-space.
1101 Description: more cleanup in mds_lov
1102 Details : not send LOV EA under replay, we can't know about they size at this
1103 time. Don't allow client connect to mds before any ost connected,
1104 for avoid problems with LOV EA size and returning EIO to client.
1106 Severity : enhancement
1108 Description: Interoperability at server side (Disk interoperability)
1110 Severity : enhancement
1112 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1114 Severity : enhancement
1116 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1118 Severity : enhancement
1120 Description: New client IO stack (CLIO).
1122 Severity : enhancement
1124 Description: Commit on sharing. Eliminate inter-client dependencies between
1125 uncommitted transactions by doing transaction commits.
1126 Thereby clients may recovery independently.
1129 Frequency : Create a symlink file with a very long name
1131 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1132 Details : If there is no extra space in the request for early cancels,
1133 ldlm_req_handles_avail() returns 0 instead of a negative value.
1135 Severity : enhancement
1137 Description: Add /proc entry for import status
1138 Details : The mdc, osc, and mgc import directories now have
1139 an import directory that contains useful import data for debugging
1140 connection problems.
1142 Severity : enhancement
1144 Description: Re-disable certain /proc logging
1145 Details : Enable and disable client's offset_stats, extents_stats and
1146 extents_stats_per_process stats logging on the fly.
1149 Frequency : Only on FC kernels 2.6.22+
1151 Description: oops in statahead
1152 Details : Do not drop reference count for the dentry from VFS when lookup,
1153 VFS will do that by itself.
1155 Severity : enhancement
1157 Description: Generic /proc file permissions
1158 Details : Set /Proc file permissions in a more generic way to enable non-
1159 root users operate on some /proc files.
1163 Description: Hitting mdc_commit_close() ASSERTION
1164 Details : Properly handle request reference release in
1165 ll_release_openhandle().
1169 Description: quota recovery deadlock during mds failover
1170 Details : This patch includes att18982, att18236, att18237 in bz14840.
1172 1. fix osts hang when mds does failover with quotaon
1173 2. prevent watchdog storm when osts threads wait for the
1178 Frequency : only patchless client
1179 Description: add workaround for race between add/remove dentry from hash
1181 Severity : enhancement
1183 Description: Allow OST glimpses to return PW locks
1187 Description: LBUG when llog conf file is full
1188 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1192 Description: Prevent import from entering FULL state when server in recovery
1196 Description: service mount cannot take device name with ":"
1197 Details : Only when device name contains ":/" will mount treat it as
1203 Description: replace ptlrpcd with the statahead thread to interpret the async
1204 statahead RPC callback
1208 Frequency : on recovery
1209 Description: I/O failures after umount during fail back
1210 Details : if client reconnected to restarted server we need join to recovery
1211 instead of find server handler is changed and process self
1212 eviction with cancel all locks.
1214 Severity : enhancement
1216 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1218 Severity : enhancement
1220 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1222 Severity : enhancement
1224 Description: Upcall on Lustre log has been dumped
1225 Details : Allow for a user mode script to be called once a Lustre log has
1226 been dumped. It passes the filename of the dumped log to the
1227 script, the location of the script can be specified via
1228 /proc/sys/lnet/debug_log_upcall.
1233 Description: avoid idr_remove called for id which is not allocated.
1234 Details : Move assigment s_dev for clustered nfs to end of initialization,
1235 to avoid problem with error handling.
1240 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1241 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1242 into CONN_USED_HASH and this prodice warning when put connection
1243 again in unused hash.
1246 Severity : enhancement
1248 Description: File striping can now be set to use an arbitrary pool of OSTs.
1250 Severity : enhancement
1252 Description: Export bytes_read/bytes_write count on OSC/OST.
1256 Description: Early reply size mismatch, MGC loses connection
1257 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1258 the connect flags are properly negotiated.
1263 Description: quotas are not honored with O_DIRECT
1264 Details : all writes with the flag O_DIRECT will use grants which leads to
1265 this problem. Now using OBD_BRW_SYNC to guard this.
1269 Description: add quota statistics
1270 Details : 1. sort out quota proc entries and proc code.
1271 2. add quota statistics
1273 Severity : enhancement
1275 Description: enable quota support for HEAD.
1279 Description: Properly propagate oinfo flags from lov to osc for statfs
1280 Details : restore missing copy oi_flags to lov requests.
1282 Severity : enhancement
1284 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1285 : l_getgroups(8), lst(8), routerstat(8)
1286 Details : included man pages for llobdstat(8), llstat(8),
1287 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1289 Severity : enhancement
1291 Description: configure's --enable-quota should check the
1292 : kernel .config for CONFIG_QUOTA
1293 Details : configure is terminated if --enable-quota is passed but
1294 : no quota support is in kernel
1298 Description: Remove portals compatibility
1299 Details : Remove portals compatibility, not interoperable with releases
1304 Description: Resolve device initialization race
1305 Details : Prevent proc handler from accessing devices added to the
1306 obd_devs array but yet be intialized.
1308 Severity : enhancement
1310 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1312 Severity : enhancement
1314 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1319 Description: Using wrong pointer in osc_brw_prep_request
1320 Details : Access to array[-1] can produce panic if kernel compiled with
1321 CONFIG_PAGE_ALLOC enabled
1325 Description: Client runs out of low memory
1326 Details : Consider only lowmem when counting initial number of llap pages
1330 Description: *optional* service tags registration
1331 Details : if the "service tags" package is installed on a Lustre node
1332 When the filesystem is mounted, a local-node service tag will
1333 be created. See http://inventory.sun.com/ for more information
1334 about the Service Tags asset management system.
1338 Description: Kernel BUG tries to release flock
1339 Details : Lustre does not destroy flock lock before last reference goes
1340 away. So always drop flock locks when client is evicted and
1341 perform unlock regardless of successfulness of speaking to MDS.
1345 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1349 Description: sanity test 65a fails if stripecount of -1 is set
1350 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1355 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1356 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1359 Severity : enhancement
1361 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1365 Description: do not process already freed flock
1366 Details : flock can possibly be freed by another thread before it reaches
1367 to ldlm_flock_completion_ast.
1371 Description: LBUG during stress test
1372 Details : Need properly lock accesses the flock deadlock detection list.
1376 Description: oops in page fault handler
1377 Details : kernel page fault handler can return two special 'pages' in error
1378 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1382 Description: timeout with invalidate import.
1383 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1384 should be handled by ptlrpcd. This produce long age waiting and
1385 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1387 Severity : enhancement
1389 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1393 Description: enable MGS and MDT services start separately
1394 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1395 a co-located MGS without starting the MGS, which is a complement
1396 to 'nosvc' mount option.
1400 Frequency : after recovery
1401 Description: precreate to many object's after del orphan.
1402 Details : del orphan st in oscc last_id == next_id and this triger growing
1403 count of precreated objects. Set flag LOW to skip increase count
1404 of precreated objects.
1408 Frequency : rare, on clear nid stats
1409 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1410 Details : when clean nid stats sometimes try destroy live entry,
1411 and this produce panic in free.
1415 Description: Stack overflow during MDS log replay
1416 ease stack pressure by using a thread dealing llog_process.
1420 Description: wait until IO finished before start new when do lock cancel.
1421 Details : VM protocol want old IO finished before start new, in this case
1422 need wait until PG_writeback is cleared until check dirty flag and
1423 call writepages in lock cancel callback.
1425 Severity : enhancement
1427 Description: using special macro for print time and cleanup in includes.
1431 Description: mds_mfd_close() ASSERTION(rc == 0)
1432 Details : In mds_mfd_close(), we need protect inode's writecount change
1433 within its orphan write semaphore to prevent possible races.
1437 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1441 Frequency : rare, on shutdown ost
1442 Description: don't hit live lock with umount ost.
1443 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1444 use shrink_dcache_sb instead.
1448 Description: don't panic with use echo client
1449 Details : echo client pass NULL as client nid pointer and this produce null
1450 pointer dereference.
1454 Description: fix build on ppc32
1455 Details : compile code with -m64 flag produce wrong object file for ppc32.
1459 Description: add message levels for liblustreapi
1463 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1464 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1465 will result in an -EAGAIN and in the exit call path we call
1466 alloc_rr() which will with increasing aggressiveness attempt to
1467 aquire precreated objects on the minimum number of required OSCs.
1471 Description: Use old size assignment to avoid deadlock
1472 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1473 the scheduling while holding a spinlock. We do not need locking
1474 for size in ll_update_inode() because size is only updated from
1475 the MDS for directories or files without objects, so there is no
1476 other place to do the update, and concurrent access to such inodes
1477 are protected by the inode lock.
1481 Description: resolve "_IOWR redefined" build error on SLES10
1485 Description: dump the memory debugging after all modules are unloaded to
1486 suppress false negative in conf_sanity test 39
1488 Severity : enhancement
1490 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1495 Description: fix SLES kernel versioning
1496 Details : the kernel version for our SLES 10 kernel did not include a "-"
1497 before the "smp" at the end. while this was not a problem in
1498 general, it did mean that software trying to use the kernel
1499 version to try to detect a vendor specific kernel would fail.
1500 this was most evident by the OFED build scripts.
1504 Description: Don't update lov_desc members until making sure they are valid
1505 Details : When updating lov_desc members via proc fs, need fix their
1506 validities before doing the real update.
1510 Description: don't put request into delay list while invalidate in flight.
1511 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1512 import in flight. this produce timeout for invalidate and sometimes
1513 can cause stale data.
1517 Frequency : on ppc only
1518 Description: not convert ost objects for directory because it's not exist.
1519 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1523 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1525 Severity : enhancement
1527 Description: Update to support 2.6.22.14 vanilla kernel.
1531 Frequency : rare, on recovery
1532 Description: read procfs can produce deadlock in some situation
1533 Details : Holding lprocfs lock which send rpc can produce block for destroy
1534 obd objects and this also block reconnect with -EALREADY. This
1535 isn't fix all lprocfs bugs - but make it rare.
1537 Severity : enhancement
1539 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1542 Frequency : frequent on X2 node
1544 Description: mdc_set_open_replay_data LBUG
1545 Details : Set replay data for requests that are eligible for replay.
1549 Description: lustre_mgs: operation 101 on unconnected MGS
1550 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1551 MGC, and late on MGC cannot successfully connect to MGS and a lot
1552 of the error messages complaining that MGS is not connected.
1556 Frequency : on network error
1557 Description: panic with double free request if network error
1558 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1559 but it's true only for synchronus enqueue, for async enqueue
1560 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1563 Severity : enhancement
1565 Description: client-side metadata stat-ahead during readdir(directory readahead)
1566 Details : perform client-side metadata stat-ahead when the client detects
1567 readdir and sequential stat of dir entries therein
1570 Frequency : on start mds
1572 Description: Implement get_info(last_id) in obdfilter.
1575 Frequency : occasional
1577 Description: Correctly check stale fid, not start epoch if ost not support SOM
1578 Details : open with flag O_CREATE need set old fid in op_fid3 because
1579 op_fid2 was overwritten with new generated fid, but mds can answer
1580 with one of these two fids and both is not stale. Setattr
1581 incorrectly started an epoch and assume will be called
1582 done_writing, but without SOM done_writing ever being called.
1585 Frequency : rare, depends on device drivers and load
1587 Description: MDS or OSS nodes crash due to stack overflow
1588 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1589 In some cases, in conjunction with device drivers that use a lot
1590 of stack the MDS (or possibly OSS) service threads could overflow
1591 the stack. One change which was identified to consume additional
1592 stack has been reworked to avoid the extra stack usage.
1595 Frequency : occasional
1597 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1598 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1599 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1600 oscc_has_objects and see OSC has no objests and call
1601 oscc_internal_create to resend create request.
1603 Severity : enhancement
1605 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1607 Severity : enhancement
1609 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1612 Frequency : very rare
1614 Description: Fix replay if there is an un-replied request and open
1615 Details : In some cases, older replay request will revert the
1616 mcd->mcd_last_xid on MDS which is used to record the client's
1617 latest sent request.
1619 Severity : enhancement
1621 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1623 Severity : enhancement
1625 Description: Add rhel5 support to HEAD.
1627 Serverity : enhancement
1629 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1634 Description: Don't allow skipping OSTs if index has been specified.
1635 Details : Don't allow skipping OSTs if index has been specified, make
1636 locking in internal create lots better.
1640 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1641 Details : access to bitfield in structure is always rounded to long
1642 and this produce problem with not atomic change any bit.
1646 Description: Lustre make rpms failed.
1647 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1648 builds Lustre rpms from tarball.
1651 Frequency : rare on shutdown ost
1653 Description: If llog cancel was not send before clean_exports phase, this can
1654 produce deadlock in llog code.
1655 Details : If llog thread has last reference to obd and call class_import_put
1656 this produce deadlock because llog_cleanup_commit_master wait when
1657 last llog_commit_thread exited, but this never success because was
1658 called from llog_commit_thread.
1662 Description: allow userland application know is lost one of stripes.
1663 Details : fill lvb_blocks with error code on ost and return it to
1664 application if error flag found.
1668 Description: NULL lov_tgts causing MDS oops
1669 Details : more safe checks for NULL lov_tgts for avoid oops.
1671 Severity : enhancement
1673 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1677 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1678 Details : If an OST is down the MDS will hang indefinitely in
1679 obd_statfs_rqset() waiting for the statfs data. While for
1680 MDS QOS usage of statfs, it should not stuck in waiting.
1682 Severity : enhancement
1684 Description: remote_acl support
1685 Details : Support ACL-based permission check for remote user.
1686 Support setfacl/getfacl for remote user with the utils
1687 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1688 the system "{s,g}etfacl" utils.
1690 Severity : enhancement
1692 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1694 Severity : enhancement
1696 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1700 Description: make mgs_setparam() handle fsname containing dash
1701 Details : fsname containing a dash does not work with lctl conf_param
1703 Severity : enhancement
1705 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1707 Severity : enhancement
1709 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1713 Frequency : rare, at shutdown
1714 Description: access already free / zero obd_namespace.
1715 Details : if client_disconnect_export was called without force flag set,
1716 and exist connect request in flight, this can produce access to
1717 NULL pointer (or already free pointer) when connect_interpret
1718 store ocd flags in obd_namespace.
1722 Frequency : only at startup
1723 Description: not alloc memory with spinlock held.
1724 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1725 if any spinlock held.
1727 Severity : enhancement
1729 Description: make lustre randomly fail allocating memory
1730 Details : Make lustre randomly failed allocating memory for testing purpose.
1732 Severity : enhancement
1734 Description: lost problems with lov objid file
1735 Details : Fixes some scability and access to not inited memory problems
1736 in work with lov objdid file.
1741 Description: lfs find does not continue on file error
1742 Details : Continue other files processing when a file/dir is absent.
1746 Description: Inconsistent usage of lustre_pack_reply()
1747 Details : Standardize the usage of lustre_pack_reply() such that it
1748 always generate a CERROR on failure.
1751 Frequency : occasional
1753 Description: MDS hang or stay in waiting lock
1754 Details : If client receive lock with CBPENDING flag ldlm need send lock
1755 cancel as separate rpc, to avoid situation when cancel request
1756 can't processed due all i/o threads stay in wait lock.
1760 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1764 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1768 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1772 Description: Update to RHEL4 latest kernel.
1774 Severity : enhancement
1776 Description: Build SLES10 patchless client fails
1777 Details : The configure was broken by run ./configure with
1778 --with-linux-obj=.... argument for patchless client. When the
1779 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1780 can't search header adequately. Use absolute path such as
1781 -I($LINUX)/include instead.
1785 Description: interrupt oig_wait produce painc on resend.
1786 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1787 private set, and this produce situation when rq_ptlrpcd_data not
1788 copyed to new allocated request and triggered LBUG on assert
1789 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1790 ptlrpcd set if request is changed and can be safety changed to use
1795 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1796 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1797 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1798 these macros and replace them with their difinitions instead.
1802 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1803 Details : Update which_patch & target file for SLES10 latest kernel.
1806 Description: Remove client patches from SLES 10 kernel.
1807 Details : This causes SLES 10 clients to behave as patchless clients
1808 even on a Lustre-patched (server) kernel.
1810 Severity : enhancement
1812 Description: self-adjustable client's lru lists
1813 Details : use adaptive algorithm for managing client cached locks lru
1814 lists according to current server load, other client's work
1815 pattern, memory activities, etc. Both, server and client
1816 side namespaces provide number of proc tunables for controlling
1819 Severity : enhancement
1821 Description: light-weight GSS support
1822 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1823 and reduce performance overhead.
1825 Severity : enhancement
1827 Description: Linux keyring support
1828 Details : Support using service of Linux keyring for Lustre GSS internal
1829 context refresh/cache mechanism.
1833 Description: Fix errors in lfs documentation
1834 Details : Fixes man pages
1838 Description: don't use GFP_* in generic Lustre code.
1839 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1843 Description: obdclass is limited by single OBD_ALLOC(idarray)
1844 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1848 Description: warnings with build patchless client with vanila 2.6.19 and up
1849 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1850 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1854 Frequency : Only for SLES
1855 Description: sanity_quota fail test_1
1856 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1857 file, which makes TSTID[2] inunique.
1861 Frequency : Only for SLES9
1862 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1863 Details : Update target/ChangeLog/which_patch .
1867 Frequency : Only for test_10 in sanity-quota.sh
1868 Description: a bug in quota test code
1869 Details : checking a test flag in a wrong place
1873 Frequency : Only for test_13 in sanity-quota.sh
1874 Description: a bug in quota test script
1875 Details : shouldn't check overfull allocated quota
1879 Description: jbd statistics
1880 Details : Port older jbd statistics patch for sles10
1882 Severity : enhancement
1884 Description: Add printing inode info into message about error in writepage.
1889 Description: reply_lock_interpret crash due to race with it and lock cancel.
1890 Details : Do not replay locks that are being cancelled. Do not reference
1891 locks by their address during replay, just by their handle.
1895 Frequency : When flocks are used.
1896 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1897 Details : Flock locks might destroy just granted lock if it could be merged
1898 with another existing flock, this is done in completion handler,
1899 so teach ldlm_cli_enquque_fini that this is a valid case for
1905 Description: Oops in read and write path when failing to allocate lock.
1906 Details : Check if lock allocation failed and return error back.
1910 Description: lstripe command fails for valid OST index
1911 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1912 instead of lov->desc.ld_active_tgt_count.
1916 Description: sanity.sh failed test 103
1917 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1922 Description: df doesn't work properly if diskfs blocksize != 4K
1923 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1927 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1928 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1930 Severity : enhancement
1932 Description: Adaptive timeouts
1933 Details : RPC timeouts adapt to changing server load and network
1934 conditions to reduce resend attempts and improve recovery time.
1938 Description: llapi_file_create() does not allow some changes
1939 Details : add llapi_file_open() that allows specifying the mode and
1940 open flags, and also returns an open file handle.
1944 Description: merge and cleanup kernel patches.
1945 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1949 Description: Add journal checksum support.(Kernel part)
1950 Details : The journal checksum feature adds two new flags i.e
1951 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1952 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1953 indicates that the commit block contains the checksum for
1954 the blocks described by the descriptor blocks. Now commit
1955 record can be sent to disk without waiting for descriptor
1956 blocks to be written to disk. This behavior is controlled
1957 using JBD2_FEATURE_ASYNC_COMMIT flag.
1961 Description: OSS needs mutliple precreate threads
1962 Details : Add ability to start more than one create thread per OSS.
1966 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1967 Details : execute lfs setstripe on client
1971 Description: mds_obd_create error creating tmp object
1972 Details : When the user sets quota on root, llog will be affected and can't
1973 create files and write files.
1976 Frequency : Always on ia64 patchless client, and possibly others.
1978 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1979 Details : This allows the patchless client to be loaded on architectures
1980 without this export.
1984 Description: RedHat Update kernel for RHEL5
1985 Details : Add the kernel config file for RHEL5.
1989 Description: RedHat Update kernel for RHEL5
1990 Details : Modify the kernel config file more closer RHEL5.
1994 Description: Build failure against Centos5 (RHEL5)
1995 Details : Define PAGE_SIZE when it isn't present.
1999 Description: test_7 on sanity-quota.sh can't be used on separate machines
2000 Details : add facet to handle it
2004 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2005 Details : with NFS, the anon dentry's parent was set to itself in
2006 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2007 corresponding dentry other than use rec->ur_name.
2009 Severity : enhancement
2011 Description: lfs setstripe enhancement
2012 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2016 Description: enable data checksumming by default
2017 Details : enable checksum by default, allow --disable-checksum
2018 configure option and "-o nochecksum" mount option. Checksums
2019 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2023 Description: lustre support for RHEL5
2024 Details : Add support for RHEL5.
2028 Description: Client eviction due to failover config
2029 Details : after a connection loss, the lustre client should attempt to
2030 reconnect to the last active server first before trying the
2031 other potential connections.
2035 Description: when mds and osts use different quota unit(32bit and 64bit),
2036 quota will be released repeatly.
2037 Details : void sending multiple quota reqs to mds, which will keep the
2038 status between the reqs.
2042 Description: osts not allocated evenly to files
2043 Details : change the condition to increase offset_idx
2046 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2048 Description: Data corruption for OSTs that are formatted larger than 2TB
2050 Details : When generating the bio request for lustre file writes the
2051 sector number would overflow a temporary variable before being
2052 used for the IO. The data reads correctly from Lustre (which
2053 will overflow in a similar manner) but other file data or
2054 filesystem metadata may be corrupted in some cases.
2058 Description: Tune the kernel for good SCSI performance.
2059 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2060 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2065 Description: rewrite ext2-derived code in obdclass/uuid.c
2066 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2067 from scratch preserving functionality.
2071 Description: "lfs find -obd UUID" prints directories
2072 Details : "lfs find -obd UUID" will return all directory names instead
2073 of just file names. It is incorrect because the directories
2074 do not reside on the OSTs.
2078 Description: use i_size_read and i_size_write in 2.6 port
2079 Details : replace inode->i_size access with i_size_read/write()
2082 Frequency : when removing large files
2084 Description: scheduling issue during removal of large Lustre files
2085 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2086 It causes scheduling issues when removing large files (17TB in the
2090 Frequency : only with liblustre clients on XT3
2092 Description: evictions taking too long
2093 Details : allow llrd to evict clients directly on OSTs
2096 Frequency : only on ppc
2098 Description: /proc/fs/lustre/devices broken on ppc
2099 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2100 We should make sure the type of 'index' is loff_t and then cast
2101 explicitly as needed below. Do not assign an explicitly cast
2105 Frequency : only for rhel5
2107 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2108 Details : Modify the target file & which_kernel.
2110 Severity : enhancement
2112 Description: omit set fsid for export NFS
2113 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2114 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2118 Frequency : Always, for kernels after 2.6.16
2119 Description: Fix warning idr_remove called for id=.. which is not allocated.
2120 Details : Last kernels save old s_dev before kill super and not allow
2121 to restore from callback - restore it before call kill_anon_super.
2125 Description: buffer overruns could theoretically occur
2126 Details : llapi_semantic_traverse() modifies the "path" argument by
2127 appending values to the end of the origin string, and a
2128 overrun may occur. Adding buffer overrun check in liblustreapi.
2132 Description: Fix error on 'ls .' at the top of the Lustre mount.
2133 Details : Don't revalidate dentry if it is a root dentry.
2137 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2138 Details : Modify vm-tunables-rhel4.patch.
2142 Description: Kernel config for 2.6.18-vanilla.
2143 Details : Modify targets/2.6-vanilla.target.in.
2144 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2145 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2146 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2147 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2151 Description: improve handling recoverable errors
2152 Details : if request processig with error which can be recoverable on server
2153 request should be resend, otherwise page released from cache and
2158 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2159 Details : Modify target file & which_patch.
2160 A flaw was found in the IA32 system call emulation provided
2161 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2162 value could be stored in the %RAX register, which could trigger an
2163 out-of-bounds system call table access. An untrusted local user
2164 could exploit this flaw to run code in the kernel
2165 (ie a root privilege escalation). (CVE-2007-4573).
2169 Description: O_DIRECT bypasses client statistics.
2170 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2171 read_ahead_stats were updated. copied the stats section from
2172 osc_send_oap_rpc() into async_internal().
2176 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2180 Description: change order of libsysio includes
2181 Details : '#include sysio.h' should always come before '#include xtio.h'
2185 Description: handle "serious error: objid * already exists" more gracefully
2186 Details : If LAST_ID value on disk is smaller than the objects existing in
2187 the O/0/d* directories, it indicates disk corruption and causes an
2188 LBUG(). If the object is 0-length, then we should use the existing
2189 object. This will help to avoid a full fsck in most cases.
2191 Severity : enhancement
2193 Description: adapt the lustre_config script to support the upgrade case
2194 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2195 server targets to 1.6 in parallel.
2199 Description: To avoid grant space > avaible space when the disk is almost
2200 full. Without this patch you might see the error "grant XXXX >
2201 available" or some LBUG about grant, when the disk is almost
2203 Details : In filter_check_grant, for non_grant cache write, we should
2204 check the left space by if (*left > ungranted + bytes), instead
2205 of (*left > ungranted), because only we are sure the left space
2206 is enough for another "bytes", then the ungrant space should be
2207 increase. In client, we should update cl_avail_grant only there
2208 is OBD_MD_FLGRANT in the reply.
2212 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2213 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2214 described in CVE-2007-4573. This problem could allow untrusted
2215 local users to gain root access.
2218 Frequency : when using O_DIRECT and quotas
2220 Description: Incorrect file ownership on O_DIRECT output files
2221 Details : block usage reported by 'lfs quota' does not take into account
2222 files that have been written with O_DIRECT.
2227 Description: touch file failed when fs is not full
2228 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2229 otherwise we can get ENOSP while fs is not full.
2233 Description: parallel lock callbacks
2234 Details : Instead of sending blocking and completion callbacks as separated
2235 requests, adding them to a set and sending in parallel.
2238 Frequency : only for Cray XT3
2239 Bugzilla : 12829/13455
2240 Description: Changing primary group doesn't change the group lustre assigns to
2242 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2243 overridden with the primary group provided by the group upcall,
2244 whereas the supplied fsgid can be trusted if it is in the list of
2245 supplementary groups returned by the group upcall.
2247 Severity : enhancement
2249 Description: Allow masking D_WARNING, D_ERROR messages from console
2250 Details : Console messages can now be disabled via lnet.printk.
2254 Description: User code with malformed file open parameter crashes client node
2255 Details : Before packing join_file req, all the related reference should be
2256 checked carefully in case some malformed flags cause fake
2257 join_file req on client.
2261 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2262 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2263 mode is equal to the granted mode. However, it should also check
2264 that LDLM_FL_CP_REQD is not set.
2268 Description: Heavy nfs access might result in deadlocks
2269 Details : After ELC code landed, it is now improper to enqueue any mds
2270 locks under och_sem, because enqueue might want to decide to
2271 cancel open locks for same inode we are holding och_sem for.
2275 Description: Client eviction while running blogbench
2276 Details : A lot of unlink operations with concurrent I/O can lead to a
2277 deadlock causing evictions. To address the problem, the number of
2278 oustanding OST_DESTROY requests is now throttled to
2279 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2280 callbacks are processed in priority.
2284 Description: enable ACLs on MDS by default
2285 Details : ACLs must be enabled on MDS by default.
2288 Frequency : PPC/PPC64 only
2290 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2291 Details : fix duplicated definitions between asm-ppc64/types.h and
2292 lustre_types.h on PPC.
2295 Frequency : PPC/PPC64 only
2297 Description: asm-ppc/segment.h does not exist
2298 Details : fix compile issue on PPC.
2302 Description: better handle error messages in extents code
2305 Frequency : RHEL4 only
2307 Description: mkfs is very slow on IA64/RHEL4
2308 Details : A performance regression has been discovered in the MPT Fusion
2309 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2310 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2311 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2313 Severity : enhancement
2315 Description: SNMP support enhancement
2316 Details : Adding total number of sampled request for an MDS node in snmp
2319 Severity : enhancement
2321 Description: Optimize ldlm waiting list processing for PR extent locks
2322 Details : When processing waiting list for read extent lock and meeting read
2323 lock that is same or wider to it that is not contended, skip
2324 processing rest of the list and immediatelly return current
2325 status of conflictness, since we are guaranteed there are no
2326 conflicting locks in the rest of the list.
2330 Description: Time out and refuse to reconnect
2331 Details : When the failover node is the primary node, it is possible
2332 to have two identical connections in imp_conn_list. We must
2333 compare not conn's pointers but NIDs, otherwise we can defeat
2334 connection throttling.
2338 Description: port llog fixes from b1_6 into HEAD
2339 Details : Port llog reference couting and some llog cleanups from b1_6
2340 (bug 10800) into HEAD, for protect from panic and access to already
2341 free llog structures.
2345 Description: Detect stride IO mode in read-ahead
2346 Details : When a client does stride read, read-ahead should detect that and
2347 read-ahead pages according to the detected stride pattern.
2351 Description: data checksumming impacts single node performance
2352 Details : add support for several checksum algorithm. Currently, only CRC32
2353 and Adler-32 are supported. The checksum type can be changed on
2354 the fly via /proc/fs/lustre/osc/*/checksum_type.
2358 Description: use adler32 for page checksums
2359 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2364 Description: build for x2 fails
2365 Details : fix compile issue on Cray systems.
2369 Description: Properly match for duplicate locks
2370 Details : Due to different lock order from skiplists code, we need to
2371 traverse entire list for now
2374 Frequency : only on PPC/SLES10
2376 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2377 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2378 be defined. Add a hack in mkfs_lustre.c to work around this compile
2383 Description: LASSERT on MDS when client holding flock lock dies
2384 Details : ldlm pool logic depends on number of granted locks equal to
2385 number of released locks which is not true for flock locks, so
2386 just exclude such locks from consideration.
2390 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2391 Details : Use fsfilt_sync() for both the whole filesystem sync and
2392 individual file sync to eliminate dangerous inode locking
2393 with I_LOCK that can lead to a deadlock.
2397 Description: performance in 1.6.3
2398 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2399 sure that 1MB requests are not fragmented by the block layer.
2401 Severity : enhancement
2403 Description: organize the server-side client stats on per-nid basis
2404 Details : Change the structure of stats under obdfilter and mds to
2412 The "uuid"s file would list the uuids of _active_ exports.
2413 And the clear entry is to clear all stats and stale nids.
2415 Severity : enhancement
2417 Description: eliminate client locks in face of contention
2418 Details : file contention detection and lockless i/o implementation
2419 for contended files.
2423 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2426 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2428 Description: MDS may lose file striping (and hence file data) in some cases
2429 Details : If there are additional extended attributes stored on the MDS,
2430 in particular ACLs, SELinux, or user attributes (if user_xattr
2431 is specified for the client mount options) then there is a risk
2432 of attribute loss. Additionally, the Lustre file striping
2433 needs to be larger than default (e.g. striped over all OSTs),
2434 and an additional attribute must be stored initially in the
2435 inode and then increase in size enough to be moved to the
2436 external attribute block (e.g. ACL growing in size) for file
2441 Description: skiplist implementation simplification
2442 Details : skiplists are used to group compatible locks on granted list
2443 that was implemented as tracking first and last lock of each
2444 lock group the patch changes that to using doubly linked lists
2448 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2449 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2450 because LASSERT on both the data supplied by a client, and the
2451 data on disk is dangerous and incorrect.
2453 Severity : enhancement
2455 Description: Slow truncate/writes to huge files at high offsets.
2456 Details : Directly associate cached pages to lock that protect those pages,
2457 this allows us to quickly find what pages to write and remove
2458 once lock callback is received.
2462 Description: more ldlm soft lockups
2463 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2464 starve other threads from the resource lock for a long time in
2465 case of long waiting queue, so change the debug level from
2466 D_OTHER to the less frequently used D_INFO.
2468 Severity : enhancement
2470 Description: add -gid, -group, -uid, -user options to lfs find
2474 Description: Hung threads in invalidate_inode_pages2_range
2475 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2476 once one is completed. As a result, some RPCs are stuck in the
2477 queue and are never sent.
2481 Description: filter threads hungs on waiting journal commit
2482 Details : Cleanup filter group llog code, then only filter group llog will
2483 be only created in the MDS/OST syncing process.
2487 Description: Procfs and llog threads access destoryed import sometimes.
2488 Details : Sync the import destoryed process with procfs and llog threads by
2489 the import refcount and semaphore.
2491 Severity : enhancement
2493 Description: openlock cache of b1_6 port to HEAD
2498 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2499 Details : If insertion of an extent fails, then discard the inode
2500 preallocation and free data blocks else it can lead to duplicate
2505 Description: don't always update ctime in ext3_xattr_set_handle()
2506 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2507 In some cases the ctime should not be updated, for example for
2508 2.0->1.8 compatibility it is necessary to delete an xattr and it
2509 should not update the ctime.
2513 Bugzilla : 15713/16362
2514 Description: Assertion in iopen_connect_dentry in 1.6.3
2515 Details : looking up an inode via iopen with the wrong generation number can
2516 populate the dcache with a disconneced dentry while the inode
2517 number is in the process of being reallocated. This causes an
2518 assertion failure in iopen since the inode's dentry list contains
2519 both a connected and disconnected dentry.
2523 Description: assertion failure in ldlm_handle2lock()
2524 Details : fix a race between class_handle_unhash() and class_handle2object()
2525 introduced in lustre 1.6.5 by bug 13622.
2530 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2531 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2536 Description: oops in sd_iostats_seq_show()
2537 Details : unloading/reloading the scsi low level driver triggers a kernel
2538 bug when trying to access the sd iostat file.
2543 Description: Kernel panics during QLogic driver reload
2544 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2545 patch, causing memory corruption.
2550 Description: journal_dev option does not work in b1_6
2551 Details : pass mount option during pre-mount.
2553 Severity : enhancement
2555 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2556 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2557 extent information of a file. It can be used to map logical blocks
2558 in a file to physical blocks in the block device.
2562 Description: LDLM soft lockups - improvement
2563 Details : It is be possible to send the lock handle along with each read
2564 or write request because the client is already doing a lock match
2565 itself so there isn't any reason the OST should have to re-do that
2569 Frequency : only on Cray X2
2571 Description: X2 build failures
2572 Details : fix build failures on Cray X2.
2576 Description: xid & resent requests
2577 Details : Initialize RPC XID from clock at startup (randomly if clock is
2580 Severity : enhancement
2582 Description: Add lustre_start utility to start or stop multiple Lustre servers
2587 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2588 Details : In case of memory pressure, list_del() can be called twice on
2589 req->rq_history_list, causing a kernel oops.
2593 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2594 Details : fix stack overflow in the distributed lock manager by defering
2595 export eviction after a failed AST to the elt thread instead of
2596 handling it in the dlm interpret routine.
2600 Description: Convert some comments to new format.
2601 Details : Update documenting comments to match doxygen conventions.
2605 Description: Grammar fixes.
2606 Details : A couple of trivial spelling fixes.
2610 Description: OSD_COUNTERS-mandatory
2611 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2612 to simplify the code.
2616 Description: simplify cmm_device freeing logic.
2617 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2621 Description: Add lockdep support to dt_object_operations locking interface.
2622 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2623 parameter indicating lock ordering. Update mdd code to use new
2628 Description: Introduce failloc constants for lockless IO tests.
2629 Details : Add two new failloc constants to test lockless IO. Only one of
2630 them in implemented---another is checked in yet to be landed
2635 Description: Add lockdep support for inode mutex.
2636 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2637 in the situations where Lustre has to lock more than one inode
2642 Description: Add optional invariants checking support.
2643 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2644 configure switch. This macro is to be used for consistency and
2645 sanity checks that are too expensive to be left in `production'
2650 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2651 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2656 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2657 Details : Use spin_lock_nested() in (the only) situation where more than
2658 one ldlm_lock is locked simultaneously. Also, fix possible
2659 dead-lock in ldlm_lock_change_resource() by enforcing particular
2664 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2665 Details : Instead of specifying each ldlm_lock call-back through separate
2666 parameter, wrap them into struct ldlm_callback_suite.
2670 Description: Kill join_lru obd method and its callers.
2671 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2672 in memory---a requirement implemented through ->o_join_lru() obd
2673 method in HEAD. Remove this method and its users.
2677 Description: Add asynchronous ldlm ENQUEUE completion handler.
2678 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2679 case that stock ldlm_completion_ast() cannot handle as it waits
2680 until lock is granted. Introduce new ldlm_completion_ast_async()
2681 for this. Also comment ldlm_completion_ast().
2685 Description: ldlm_error <-> errno conversion.
2686 Details : Add functions to map (rather arbitrary) between LDLM error codes
2687 and standard errno values. CLIO needs this to prevent LDLM specific
2688 constants from escaping ldlm and osc.
2692 Description: Kill unused ldlm_handle2lock_ns() function.
2693 Details : Kill unused ldlm_handle2lock_ns() function.
2697 Description: Add lu_ref support to ldlm_lock
2698 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2699 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2700 and ldlm_resource. LDLM interface has to be changed, because code that
2701 releases a reference on a lock, has to "know" what reference this is.
2702 In the most frequent case
2704 lock = ldlm_handle2lock(handle);
2706 LDLM_LOCK_PUT(lock);
2708 no changes are required. When any other reference (received _not_ from
2709 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2710 instead of LDLM_LOCK_PUT().
2712 Arguably, changes are pervasive, and interface requires some discipline
2713 for proper use. On the other hand, it was very instrumental in finding
2714 a few leaked lock references.
2718 Description: Add ldlm_lock_addref_try().
2719 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2720 attempts to addref a lock that might be being canceled
2725 Description: Add ldlm_weigh_callback().
2726 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2727 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2728 hard-coded `number of pages' logic.
2732 Description: Add lockdep annotations to llog code.
2733 Details : Use appropriately tagged _nested() locking calls in the places
2734 where llog takes more than one ->lgh_lock lock.
2738 Description: Add loi_kms_set().
2739 Details : Wrap kms updates into a helper function.
2743 Description: Constify instances of struct lsm_operations.
2744 Details : Constify instances of struct lsm_operations.
2748 Description: lu_conf support.
2749 Details : On a server, a file system object is uniquely identified
2750 by a fid, which is sufficient to locate and load all object
2751 state (inode). On a client, on the other hand, more data are
2752 necessary instantiate an object. Change lu_object_find() and
2753 friends to take additional `lu_conf' argument describing object.
2754 Typically this includes layout information.
2758 Description: lu_context fixes.
2759 Details : Introduce new lu_context functions that are needed on the client
2760 side, where some system threads (ptlrpcd) are shared by multiple
2761 modules, and so cannot be stopped during module shutdown.
2765 Description: Add start and stop methods to lu_device_type_operations.
2766 Details : Introduce two new methods in lu_device_type_operations, that are
2767 invoked when first instance of a given type is created and last one
2768 is destroyed respectively. This is need by CLIO.
2772 Description: Add lu_ref support to struct lu_device.
2773 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2774 track leaked references.
2778 Description: Introduce lu_kmem_descr.
2779 Details : lu_kmem_descr and its companion interface allow to create
2780 and destroy a number of kmem caches at once.
2784 Description: Fix lu_object finalization race.
2785 Details : Fix a race between lu_object_find() finding an object and its
2786 concurrent finalization. This race is (most likely) not possible
2787 on the server, but might happen on the client.
2791 Description: Introduce lu_ref interface.
2792 Details : lu_ref is a debugging module allowing to track references to
2793 a given object. It is quite cpu expensive, and has to be
2794 explicitly enabled with --enable-lu_ref. See usage description
2799 Description: Factor lu_site procfs stats into a separate function.
2800 Details : Separate lu_site stats printing code into a separate function
2801 that can be reused on a client.
2805 Description: Constify instances of struct {lu,dt,md}_device_operations.
2806 Details : Constify instances of struct {lu,dt,md}_device_operations.
2810 Description: Introduce struct md_site and move meta-data specific parts of
2811 struct lu_site here.
2812 Details : Move md-specific fields out of struct lu_site into special struct
2813 md_site, so that lu_site can be used on a client.
2817 Description: Kill mdd_lov_destroy().
2818 Details : Remove unused mdd code.
2822 Description: Add st_block checking to multistat.c.
2823 Details : Add st_block checking to multistat.c.
2827 Description: Add lu_ref support to struct obd_device.
2828 Details : Add lu_ref tracking to obd_device.
2832 Description: Kill obd_set_fail_loc().
2833 Details : Remove unused code.
2837 Description: Add special type for ptlrpc_request interpret functions.
2838 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2839 there. Actual usage will be in CLIO.
2843 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2844 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2845 doesn't work with lockdep.
2849 Description: Add rwv.c test program.
2850 Details : New testing program exercising readv(2) and writev(2) (Qian).
2854 Description: Add sendfile.c test program.
2855 Details : New testing program exercising sendfile(2) (Jay).
2859 Description: Ratelimit a message that can be very frequent.
2860 Details : Ratelimit a memory allocation failure message that can
2865 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2866 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2870 Description: Liblustre build fixes.
2871 Details : Liblustre build fixes.
2875 Description: libcfs: add cfs_{need,cond}_resched() interface.
2876 Details : libcfs: add cfs_{need,cond}_resched() definition and
2877 implementations for Linux, NT, and liblustre.
2879 Severity : enhancement
2881 Description: More exported tunables for mballoc
2882 Details : Add support for tunable preallocation window and new tunables for
2883 large/small requests
2887 Description: Detect corruption of block bitmap and checking for preallocations
2888 Details : Checks validity of on-disk block bitmap. Also it does better
2889 checking of number of applied preallocations. When corruption is
2890 found, it turns filesystem readonly to prevent further corruptions.
2894 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2895 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2896 window to 0. If it does hit cache miss, and read-pattern is still
2897 stride-io mode, does not reset the stride window, but also does not
2898 increase the stride window length in this case.
2902 Frequency : only for big-endian servers
2903 Description: Check if system is big-endian while mounting fs with extents feature
2904 Details : Mounting a filesystem with extents feature will fail on big-endian
2905 systems since ext3-based ldiskfs is not supported on big-endian
2906 systems. This can be over-riden with "bigendian_extents" mount option.
2908 Severity : enhancement
2910 Description: The root squash functionality
2911 Details : A security feature, which is to prevent users from being able
2912 to mount lustre on their desktop, run as root, and delete
2913 all of the files in the filesystem. The goal is accomplished by
2914 remapping user id (UID) and group id (GID) of the root user to
2915 a UID and GID specified by the system administartor via Lustre
2916 configuration management server (MGS). The functionality also
2917 allows to specify sets of clients for which the remapping does
2922 Description: Excessive recovery window
2923 Details : With AT enabled, the recovery window can be excessively long (6000+
2924 seconds). To address this problem, we no longer use
2925 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2926 timeout no longer depends on the service time, it is set to
2927 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2928 time via pb_service_time.
2932 Description: Watchdog triggered on MDS failover
2933 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2934 the OSTs know that the MDS "UUID" can be reused for the same export
2935 from a different NID, so we do not need to wait for the export to be
2939 Frequency : rare, only if using MMP with Linux RAID
2941 Description: MMP doesn't work with Linux RAID
2942 Details : While using HA for Lustre servers with Linux RAID, it is possible
2943 that MMP will not detect multiple mounts. To make this work we
2944 need to unplug the device queue in RAID when the MMP block is being
2945 written. Also while reading the MMP block, we should read it from
2946 disk and not the cached one.
2948 Severity : enhancement
2950 Description: open file using fid
2951 Details : A file can be opened using just its fid, like
2952 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2955 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2958 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2959 Details : add proper locking for creation of jbd_4k slab cache
2963 Description: MMP check in ext3_remount() fails without displaying any error
2964 Details : When multiple mount protection fails during remount, proper error
2967 Severity : enhancement
2969 Description: Allow stripe size to be up to 4G-64k
2970 Details : Fix math logic to allow large stripe sizes.
2974 Description: add check for >8TB ldiskfs filesystems
2975 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2976 Don't allow >8TB ldiskfs filesystems to be mounted without
2977 force_over_8tb mount option
2982 Descriptoin: OSS complains "too many blocks" and crashes
2983 Details : In filter_commitrw_write():
2984 1. lquota_chkquota();
2985 2. obdo_from_inode();
2986 3. lquota_pending_commit();
2987 "2." has changed uid and gid in this case, then caused
2988 lquota_chkquota() and lquota_pending_commit() to use
2989 different uid and gid. That is the root of the bug.
2993 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2994 Details : Detect and print out the 'stripe count = -1' case properly.
2998 Description: Out or order replies might be lost on replay
2999 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3000 request with tid smaller than one currently being added, add it
3001 to the start, not end of the list.
3003 --------------------------------------------------------------------------------
3005 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3007 * Support for kernels:
3008 2.6.5-7.283 (SLES 9),
3009 2.6.9-55.EL (RHEL 4),
3010 2.6.16.46-0.14 (SLES 10),
3011 2.6.18.8 vanilla (kernel.org)
3012 * Client support for unpatched kernels:
3013 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3014 2.6.16 - 2.6.22 vanilla (kernel.org)
3015 * Due to recently discovered recovery problems, we do not recommend
3016 using patchless RHEL 4 clients with this or any earlier release.
3017 * Recommended e2fsprogs version: 1.39.cfs8
3018 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3019 * Starting with this release, the ldiskfs backing filesystem required
3020 by Lustre is now in its own package, lustre-ldiskfs. This package
3021 should be installed. It is versioned separately from Lustre and
3022 may be released separately in future.
3026 Description: block reactivating mgc import until all deactivates complete
3027 Details : Fix race when failing back MDT/MGS to itself (testing)
3030 Frequency : at statup only
3032 Description: mds_lov_synchronize race leads to various problems
3033 Details : simultaneous MDT->OST connections at startup can cause the
3034 sync to abort, leaving the OSC in a bad state.
3036 Severity : enhancement
3038 Description: add optional extra BUILD_VERSION info
3039 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3040 to override the lustre version.
3043 Frequency : 2.6.18 servers only
3045 Description: ll_kern_mount() doesn't release the module reference
3046 Details : The ldiskfs module reference count never drops down to 0
3047 because ll_kern_mount() doesn't release the module reference.
3052 Description: server LBUG when using old ost_num_threads parameter
3053 Details : Accept the old ost_num_threads parameter but warn that it
3054 is deprecated, and fix an off-by-one error that caused an LBUG.
3059 Description: Transient SCSI error results in persistent IO issue
3060 Details : iobuf->dr_error is not reinitialized to 0 between two
3064 Frequency : sometimes when underlying device returns I/O errors
3066 Description: OSTs not going read-only during write failures
3067 Details : OSTs are not remounted read-only when the journal commit threads
3068 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3069 instead of the ext3 wrappers.
3073 Description: poor connect scaling with increasing client count
3074 Details : Don't run filter_grant_sanity_check for more than 100 exports
3075 to improve scaling for large numbers of clients.
3078 Frequency : SLES10 only
3080 Description: sanity-quota.sh quotacheck failed: rc = -22
3081 Details : Quotas cannot be enabled on SLES10.
3084 Frequency : liblustre clients only
3086 Description: getdirentries does not give error when run on compute nodes
3087 Details : getdirentries does not fail when the size specified as an argument
3088 is too small to contain at least one entry
3090 Severity : enhancement
3092 Description: Add LNET router traceability for debug purposes
3093 Details : If a checksum failure occurs with a router as part of the
3094 IO path, the NID of the last router that forwarded the bulk data
3095 is printed so it can be identified.
3100 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3101 Details : Due to a race condition, liblustre clients were occasionally
3102 evicted incorrectly.
3104 Severity : enhancement
3106 Description: lfs setstripe use optional parameters instead of postional
3109 Severity : enhancement
3111 Description: Nanosecond timestamp support for ldiskfs
3112 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3113 resolution timestamps. There is not yet support for this at
3114 the Lustre filesystem level.
3117 Frequency : during server recovery
3119 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3120 Details : request with rq_no_resend flag not awake l_wait_event if they get
3124 Frequency : nfs export on patchless client
3126 Description: connectathon hang when test nfs export over patchless client
3127 Details : Disconnected dentry cannot be found with lookup, so we do not need
3128 to unhash it or make it invalid
3131 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3132 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3133 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3138 Description: open req refcounting wrong on reconnect
3139 Details : If reconnect happened between getting open reply from server and
3140 call to mdc_set_replay_data in ll_file_open, we will schedule
3141 replay for unreferenced request that we are about to free.
3142 Subsequent close will crash in variety of ways.
3143 Check that request is still eligible for replay in
3144 mdc_set_replay_data().
3149 Description: disable writes to filesystem when reading health_check file
3150 Details : the default for reading the health_check proc file has changed
3151 to NOT do a journal transaction and write to disk, because this
3152 can cause reads of the /proc file to hang and block HA state
3153 checking on a healthy but otherwise heavily loaded system. It
3154 is possible to return to the previous behaviour during configure
3155 with --enable-health-write.
3157 Severity : enhancement
3159 Description: 64-bit inode version
3160 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3161 to the inode. This will be required for version-based recovery.
3166 Description: MDS fails to start if a duplicate client export is detected
3167 Details : in some rare cases it was possible for a client to connect to
3168 an MDS multiple times. Upon recovery the MDS would detect this
3169 and fail during startup. Handle this more gracefully.
3171 Severity : enhancement
3173 Description: Add -o localflock option to simulate old noflock
3175 Details : This will achieve local-only flock/fcntl locks
3181 Description: log_commit_thread vs filter_destroy race leads to crash
3182 Details : Take import reference before releasing llog record semaphore
3187 Description: Wrong request locking in request set processing
3188 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3189 imp_delayed_list, in this place should be used imp_lock.
3192 Frequency : when reconnection
3194 Description: Grant Leak when osc reconnect to OST
3195 Details : When osc reconnect ost, OST(filter) should check whether it
3196 should grant more space to client by comparing fed_grant and
3197 cl_avail_grant, and return the granted space to client instead
3198 of "new granted" space, because client will call osc_init_grant
3199 to update the client grant space info.
3202 Frequency : when client reconnect to OST
3204 Description: Grant Leak when osc do resend and replay bulk write
3205 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3206 bulk write request, because the grant info while be sync between
3207 OSC and OST when reconnect, and we should ignore the grant info
3208 these of resend/replay write req.
3213 Description: Grant space more than available left space sometimes.
3214 Details : When then OST is about to be full, if two bulk writing from
3215 different clients came to OST. Accord the avaliable space of the
3216 OST, the first req should be permitted, and the second one
3217 should be denied by ENOSPC. But if the seconde arrived before
3218 the first one is committed. The OST might wrongly permit second
3219 writing, which will cause grant space > available space.
3222 Frequency : when client is evicted
3224 Description: Grant might be wrongly erased when osc is evicted by OST
3225 Details : when the import is evicted by server, it will fork another
3226 thread ptlrpc_invalidate_import_thread to invalidate the
3227 import, where the grant will be set to 0. While the original
3228 thread will update the grant it got when connecting. So if
3229 the former happened latter, the grant will be wrongly errased
3230 because of this race.
3235 Description: Checking Stale with correct fid
3236 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3237 to check whether it is stale, because sometimes, we want the
3238 enqueue happened anyway, and op_data.fid2 will not be initialized.
3240 Severity : enhancement
3242 Description: update patchless client
3243 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3246 Frequency : only with 2.4 kernel
3248 Description: random memory corruption
3249 Details : size of struct ll_inode_info is to big for union inode.u and this
3250 can be cause of random memory corruption.
3255 Description: Memory leak in recovery
3256 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3257 It should also check obd_fail before fsfilt_start, otherwise if
3258 fsfilt_start return -EROFS,(failover mds during mds recovery).
3259 then the req will return with repmsg->transno = 0 and rc = EROFS.
3260 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3261 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3262 be freed no matter whether fsfilt_commit success or not.
3265 Frequency : only with huge count clients
3267 Description: Prevents from taking the superblock lock in llap_from_page for
3269 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3270 for llap_from_page call in ll_removepage prevents from taking the
3271 superblock lock for a soon died page.
3276 Description: Not check open intent error before release open handle
3277 Details : in some rare cases, the open intent error is not checked before
3278 release open handle, which may cause
3279 ASSERTION(open_req->rq_transno != 0), because it tries to release
3280 the failed open handle.
3285 Description: Set cat log bitmap only after create log success.
3286 Details : in some rare cases, the cat log bitmap is set too early. and it
3287 should be set only after create log success.
3291 Description: Accessing a block bevice can re-enable I/O when Lustre is
3292 tearing down a device.
3293 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3298 Description: service threads may hog cpus when there are a lot of requests
3299 Details : Insert cond_resched to give other threads a chance to use some CPU
3304 Description: the cat log was not initialized in recovery
3305 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3306 unlink log on mds will not be initialized until mds post
3307 recovery. And also in mds post recovery, the unlink log will
3308 initialization will be done asynchronausly, so there will be race
3309 between add unlink log and unlink log initialization.
3313 Description: brw_stats were being printed incorrectly
3314 Details : brw_stats were being printed as log2 but all of them were not
3315 recorded as log2. Also remove some code duplication arising from
3316 filter_tally_{read,write}.
3320 Frequency : rare, only in recovery.
3321 Description: ASSERTION(req->rq_type != LI_POISON) failed
3322 Details : imp_lock should be held while iterating over imp_sending_list for
3323 prevent destroy request after get timeout in ptlrpc_queue_wait.
3327 Description: replay-single.sh test 52 fails
3328 Details : A lock's skiplist need to be cleanup when it being unlinked
3329 from its resource list.
3333 Description: Short directio read returns full requested size rather than
3335 Details : Direct I/O operations should return actual amount of bytes
3336 transferred rather than requested size.
3338 Severity : enhancement
3340 Description: metadata RPC reduction (e.g. for rm performance)
3341 Details : decrease the amount of synchronous RPC between clients and servers
3342 by canceling conflicing lock before the operation on the client
3343 and packing thier handles into the main operation RPC to server.
3345 Severity : enhancement
3347 Description: Async OSC create to avoid the blocking unnecessarily.
3348 Details : If an OST has no remaining object, system will block on the
3349 creation when it needs to create a new object on this OST. Now,
3350 ways use pre-created objects when available, instead of blocking on
3351 an empty osc while others are not empty. If we must block, we
3352 block for the shortest possible period of time.
3356 Description: improve handling recoverable errors
3357 Details : if request processig with error which can be recoverable on server
3358 request should be resend, otherwise page released from cache and
3361 Severity : enhancement
3363 Description: refine locking for avoid write wrong info into lov_objid file
3364 Details : fix possible races with add new target and write/update data in
3367 --------------------------------------------------------------------------------
3369 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3374 Frequency : on some architectures
3376 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3377 Details : Uninitialized flags sometimes cause configuration commands to
3381 Frequency : patchless clients only
3383 Description: missing __iget() symbol export
3384 Details : The __iget() symbol export is missing. To avoid the need for
3385 this on patchless clients the deathrow inode reaper is turned
3386 off, and we depend on the VM to clean up old inodes. This
3387 dependency was during via the fix for bug 12181.
3389 --------------------------------------------------------------------------------
3391 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3393 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3394 INTEROPERATE with older versions automatically. In many cases a
3395 special upgrade step is needed. Please read the
3396 user documentation before upgrading any part of a live system.
3397 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3398 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3399 * WARNING: Lustre configuration and startup changes are required with
3400 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3404 Severity : enhancement
3406 Description: MountConf
3407 Details : Lustre configuration is now managed via mkfs and mount
3408 commands instead of lmc and lconf. New obd types (MGS, MGC)
3409 are added for dynamic configuration management. See
3410 https://mail.clusterfs.com/wikis/lustre/MountConf for
3413 Severity : enhancement
3415 Description: dynamic OST addition
3416 Details : OSTs can now be added to a live filesystem
3418 Severity : enhancement
3420 Description: startup order invariance
3421 Details : MDTs and OSTs can be started in any order. Clients only
3422 require the MDT to complete startup.
3424 Severity : enhancement
3426 Description: parallel, asynchronous orphan cleanup
3427 Details : orphan cleanup is now performed in separate threads for each
3428 OST, allowing parallel non-blocking operation.
3430 Severity : enhancement
3432 Description: optimized stripe assignment
3433 Details : stripe assignments are now made based on ost space available,
3434 ost previous usage, and OSS previous usage, in order to try
3435 to optimize storage space and networking resources.
3437 Severity : enhancement
3439 Description: Permanently set tunables
3440 Details : All writable /proc/fs/lustre tunables can now be permanently
3441 set on a per-server basis, at mkfs time or on a live system.
3443 Severity : enhancement
3445 Description: Lustre message v2
3446 Details : Add lustre message format v2.
3448 Severity : enhancement
3450 Description: client OST exclusion list
3451 Details : Clients can be started with a list of OSTs that should be
3452 declared "inactive" for known non-responsive OSTs.
3456 Description: SPEC SFS validation failure on NFS v2 over lustre.
3457 Details : Changes the blocksize for regular files to be 2x RPC size,
3458 and not depend on stripe size.
3460 Severity : enhancement
3462 Description: Multiple MD RPCs in flight.
3463 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3464 To avoid overly-overloading MDS, introduce a limit on number of
3465 MDS RPCs in flight for a single client and add /proc controls
3466 to adjust this limit.
3468 Severity : enhancement
3470 Description: client read/write statistics
3471 Details : Add client read/write call usage stats for performance
3472 analysis of user processes.
3473 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3474 file access. extents_stats shows chunk size distribution.
3475 extents_stats_per_process show chunk size distribution per
3478 Severity : enhancement
3480 Description: mds statistics
3481 Details : Add detailed mds operations statistics in
3482 /proc/fs/lustre/mds/*/stats.
3486 Description: Failure of copying files with lustre special EAs.
3487 Details : Client side always return success for setxattr call for lustre
3488 special xattr (currently only "trusted.lov").
3493 Description: Refcount LNET uuids
3494 Details : The global LNET uuid list grew linearly with every startup;
3495 refcount repeated list entries instead of always adding to
3498 Severity : enhancement
3500 Description: Dynamic service threads
3501 Details : Within a small range, start extra service threads
3502 automatically when the request queue builds up.
3505 Frequency : mixed-endian client/server environments
3507 Description: mixed-endian crashes
3508 Details : The new msg_v2 system had some failures in mixed-endian
3511 Severity : enhancement
3513 Description: Easy OST removal
3514 Details : OSTs can be permanently deactivated with e.g. 'lctl
3515 conf_param lustre-OST0001.osc.active=0'
3517 Severity : enhancement
3519 Description: MGS proc entries
3520 Details : Added basic proc entries for the MGS showing what filesystems
3523 Severity : enhancement
3525 Description: provide MGS failover
3526 Details : Added config lock reacquisition after MGS server failover.
3528 Severity : enhancement
3530 Description: add Linux 2.4 support
3531 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3535 Description: a large application tries to do I/O to the same resource and dies
3536 in the middle of it.
3537 Details : Check the req->rq_arrival time after the call to
3538 ost_brw_lock_get(), but before we do anything about
3539 processing it & sending the BULK transfer request. This
3540 should help move old stale pending locks off the queue as
3541 quickly as obd_timeout.
3544 Frequency : when an incorrect nid is specified during startup
3546 Description: ptlrpc connect to non-existant node causes kernel crash
3547 Details : LNET can't be re-entered from an event callback, which
3548 happened when we expire a message after the export has been
3549 cleaned up. Instead, hand the zombie cleanup off to another
3552 Severity : enhancement
3554 Description: plain/inodebits lock performance improvement
3555 Details : Grouping plain/inodebits in granted list by their request modes
3556 and bits policy, thus improving the performance of search through
3560 Frequency : only if OST filesystem is corrupted
3562 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3563 Details : for a short time RPCs with bulk IO are in the replay list,
3564 but replay of bulk IOs is unimplemented. If the OST filesystem
3565 is corrupted due to disk cache incoherency and then replay is
3566 started it is possible to trip an assertion. Avoid putting
3567 committed RPCs into the replay list at all to avoid this issue.
3570 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3572 Description: System hang on startup
3573 Details : This bug allowed the liblustre (e.g. catamount) client to
3574 return to the app before handling all startup RPCs. This
3575 could leave the node unresponsive to lustre network traffic
3576 and manifested as a server ptllnd timeout.
3578 Severity : enhancement
3580 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3581 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3582 boolean to control whether to print peer debug info when a
3583 client's RPC times out.
3586 Frequency : only for kernels with patches from Lustre below 1.4.3
3588 Description: Remove old rdonly API
3589 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3592 Frequency : only for devices with external journals
3594 Description: Set external device read-only also
3595 Details : During a commanded failover stop, we set the disk device
3596 read-only while the server shuts down. We now also set any
3597 external journal device read-only at the same time.
3600 Frequency : when upgrading from 1.4 while trying to change parameters
3602 Description: The wrong (new) MDC name was used when setting parameters for
3603 upgraded MDT's. Also allows changing of OSC (and MDC)
3604 parameters if --writeconf is specified at tunefs upgrade time.
3607 Frequency : when setting specific ost indicies
3609 Description: QOS code breaks on skipped indicies
3610 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3611 created with --index need not be sequential.
3613 Severity : enhancement
3615 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3616 Details : The uninit_groups feature works in conjunction with the kernel
3617 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3618 up the pass1 processing of e2fsck. This is a read-only feature
3619 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3620 mount filesystems that have had this feature enabled.
3622 Severity : enhancement
3624 Description: Improve multi-block allocation algorithm to avoid fragmentation
3625 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3626 allocation locality and avoid filesystem fragmentation.
3628 ------------------------------------------------------------------------------
3630 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3632 * Support for kernels:
3633 2.4.21-47.0.1.EL (RHEL 3)
3634 2.6.5-7.283 (SLES 9)
3635 2.6.9-42.0.10.EL (RHEL 4)
3636 2.6.12.6 vanilla (kernel.org)
3637 2.6.16.27-0.9 (SLES 10)
3638 * Recommended e2fsprogs version: 1.39.cfs5
3640 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3644 Frequency : occasional, depends on client load and configuration
3645 Bugzilla : 12181, 12203
3646 Description: data loss for recently-modified files
3648 Details : In some cases it is possible that recently written or created
3649 files may not be written to disk in a timely manner (this should
3650 normally be within 30s unless client IO load is very high).
3651 The problem appears as zero-length files or files that are a
3652 multiple of 1MB in size after a client crash or client eviction
3653 that are missing data at the end of the file.
3655 This problem is more likely to be hit on clients where files are
3656 repeatedly created and unlinked in the same directory, clients
3657 have a large amount of RAM, have many CPUs, the filesystem has
3658 many OSTs, the clients are rebooted frequently, and/or the files
3659 are not accessed by other nodes after being written.
3661 The presence of the problem can be detected by looking at
3662 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3663 smaller than the second (nr_unused) then dirty files will not
3664 be flushed automatically to disk. "sync; sleep 10" should be
3665 run several times on the node before unmounting it to update
3666 Lustre (this is also safe to run on nodes without this problem).
3668 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3669 that can cause this same problem, so customers using that kernel
3670 also need to update the kernel in addition to Lustre. In order
3671 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3673 It is normal that files written just before a client crash (less
3674 than 30s) may not yet have been flushed to disk, even for local
3678 Frequency : frequent on thin XT3 nodes
3680 Description: UUID collision on thin XT3 Linux nodes
3681 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3682 randomly, since we relied on an insufficiently-seeded PRNG.
3687 Description: OSS hangs after "All ost request buffers busy"
3688 Details : A deadlock between quota and journal operations caused OSS
3689 hangs after printing "All ost request buffers busy."
3692 Frequency : always on liblustre builds
3694 Description: Cleanup compiler warnings on liblustre
3697 Frequency : always on liblustre builds on XT3
3699 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3703 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3704 > 32000 subdirectories
3705 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3706 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3707 subdirectory count crosses 32000. This will aid e2fsck to
3708 correctly handle more than 32000 subdirectories.
3711 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3713 Description: System hang on startup
3714 Details : This bug allowed the liblustre (e.g. catamount) client to
3715 return to the app before handling all startup RPCs. This
3716 could leave the node unresponsive to lustre network traffic
3717 and manifested as a server ptllnd timeout.
3719 Severity : enhancement
3721 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3722 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3723 boolean to control whether to print peer debug info when a
3724 client's RPC times out.
3729 Description: make O_SYNC working on 2.6 kernels
3730 Details : 2.6 kernels use different method for mark pages for write,
3731 so need add a code to lustre for O_SYNC work.
3736 Description: Failure to close file and release space on NFS
3737 Details : Put inode details into lock acquired in ll_intent_file_open.
3738 Use mdc_intent_lock in ll_intent_open to properly
3739 detect all kind of errors unhandled by mdc_enqueue.
3744 Description: proc file read during shutdown sometimes raced obd removal,
3746 Details : Add lock to prevent obd access after proc file removal.
3749 Frequency : Only for files larger than 4GB on 32-bit clients.
3751 Description: improperly doing page alignment of locks
3752 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3753 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3758 Description: RPCs being resent when they shouldn't be
3759 Details : Some RPCs that should not be resent are being resent. This
3760 can cause inconsistencies in the RPC state machine. Do not
3761 resend such requests.
3764 Frequency : rare, only with NFS export
3766 Description: Crash on NFS re-export node
3767 Details : under very unusual load conditions an assertion is hit in
3768 ll_intent_file_open()
3771 Frequency : only if OST filesystem is corrupted
3773 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3774 Details : for a short time RPCs with bulk IO are in the replay list,
3775 but replay of bulk IOs is unimplemented. If the OST filesystem
3776 is corrupted due to disk cache incoherency and then replay is
3777 started it is possible to trip an assertion. Avoid putting
3778 committed RPCs into the replay list at all to avoid this issue.
3783 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3784 Details : Large single O_DIRECT read and write calls can fail to allocate
3785 a sufficiently large buffer to process the request. In case of
3786 allocation failure the allocation is retried with a smaller
3787 buffer and broken into smaller requests.
3789 Severity : enhancement
3791 Description: Add -o localflock option to simulate old noflock behaviour.
3792 Details : This will achieve local-only flock/fcntl locks coherentness.
3797 Description: versioning check is incomplete
3798 Details : Checking the version difference of client vs. server, report
3799 error if the gap is too big.
3804 Description: add support PG_writeback bit
3805 Details : add support for PG_writeback bit for Lustre, for more carefull
3806 work with page cache in 2.6 kernel. This also fix some deadlocks
3807 and remove hack for work O_SYNC with 2.6 kernel.
3809 Severity : enhancement
3811 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3812 Details : The uninit_groups feature works in conjunction with the kernel
3813 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3814 up the pass1 processing of e2fsck. This is a read-only feature
3815 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3816 mount filesystems that have had this feature enabled.
3818 Severity : enhancement
3820 Description: Improve multi-block allocation algorithm to avoid fragmentation
3821 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3822 allocation locality and avoid filesystem fragmentation.
3824 ------------------------------------------------------------------------------
3826 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3828 * Support for kernels:
3829 2.6.9-42.0.3.EL (RHEL 4)
3830 2.6.5-7.276 (SLES 9)
3831 2.4.21-47.0.1.EL (RHEL 3)
3832 2.6.12.6 vanilla (kernel.org)
3833 2.6.16.21-0.8 (SLES10)
3834 * Recommended e2fsprogs version: 1.39.cfs2-0
3836 * The backwards-compatible /proc/sys/portals symlink has been removed
3837 in this release. Before upgrading, please ensure that you change
3838 any configuration scripts or /etc/sysctl.conf files that access
3839 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3840 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3841 in advance of the upgrade on any system running Lustre 1.4.6 or
3842 newer, since /proc/sys/lnet was added in that version.
3843 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3847 Frequency : only when quota is used
3849 Description: avoid scanning export list for quota master
3850 Details : Change the algorithms to avoid scanning export list in order
3851 to improve the efficiency.
3854 Frequency : MDS failover only, very rarely
3856 Description: "went back in time" messages on mds failover
3857 Details : The greatest transno may be lost when the current operation
3858 finishes with an error (transno==0) and the client's last_rcvd
3859 record is over-written. Save the greatest transno in the
3860 mds_last_transno for this case.
3863 Frequency : always for specific kernels and striping counts
3865 Description: client may get "Matching packet too big" without ACL support
3866 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3867 when trying to access files in certain configurations. The
3868 clients should in fact be denied when mounting because they do
3869 not understand ACLs.
3872 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3874 Description: many clients connecting with IO in progress causes connect timeouts
3875 Details : Avoid synchronous journal commits to avoid delays caused by many
3876 clients connecting/disconnecting when bulk IO is in progress.
3877 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3878 OST_IO_PORTAL to avoid delays behind potentially many pending
3882 Frequency : occasionally with multiple writers to a single file
3884 Description: shared writes to file may result in wrong size reported by stat()
3885 Details : Allow growing of kms when extent lock is cancelled
3888 Frequency : always with random mmap IO to multi-striped file
3890 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3891 Details : Only if the hole is at the end of OST object so that kms is too
3892 small. Fix is to increase kms accordingly in ll_nopage.
3895 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3897 Description: writes to a missing object would leak memory on the OST
3898 Details : If there is an inconsistency between the MDS and OST filesystems,
3899 such that the MDS references an object that doesn't exist, writes
3900 to that object will leak memory due to incorrect cleanup in the
3901 error handling path, eventually running out of memory on the OST.
3906 Description: Creating too long symlink causes lustre errors
3907 Details : Check symlink and name lengths before sending requests to MDS.
3910 Frequency : only if flock is enabled (not on by default)
3912 Description: posix locks not released on fd closure on 2.6.9+
3913 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3914 kernels, this caused such locks not to be released on fd close and
3915 then assertions on fs unmount about still used locks.
3918 Frequency : MDS failover only, very rarely
3920 Description: clients may get ASSERTION(granted_lock != NULL)
3921 Details : When request was taking a long time, and a client was resending
3922 a getattr by name lock request. The were multiple lock requests
3923 with the same client lock handle and
3924 mds_getattr_name->fixup_handle_for_resent_request found one of the
3925 lock handles but later failed with ASSERTION(granted_lock != NULL).
3930 Description: handle->h_buffer_credits > 0, assertion failure
3931 Details : h_buffer_credits is zero after truncate, causing assertion
3932 failure. This patch extends the transaction or creates a new
3936 Frequency : NFS re-export or patchless client
3937 Bugzilla : 11179, 10796
3938 Description: Crash on NFS re-export node (__d_move)
3939 Details : We do not want to hash the dentry if we don't have a lock.
3940 But if this dentry is later used in d_move, we'd hit uninitialised
3941 list head d_hash, so we just do this to init d_hash field but
3942 leave dentry unhashed.
3945 Frequency : NFS re-export or patchless client
3947 Description: NFS exports has problem with symbolic link
3948 Details : lustre client didn't properly install dentry when re-exported
3949 to NFS or running patchless client.
3952 Frequency : NFS re-export or patchless client
3954 Description: Various nfs/patchless fixes.
3955 Details : fixes reuse disconected alias for lookup process - this fixes
3956 warning "find_exported_dentry: npd != pd",
3957 fix permission error with open files at nfs.
3958 fix apply umask when do revalidate.
3961 Frequency : occasional
3963 Description: Crash on NFS re-export node
3964 Details : calling clear_page() on the wrong pointer triggered oops in
3965 generic_mapping_read().
3968 Frequency : rarely, using O_DIRECT IO
3970 Description: unaligned directio crashes client with LASSERT
3971 Details : check for unaligned buffers before trying any requests.
3974 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3976 Description: stale data returned from RAID cache
3977 Details : If only a small amount of IO is done to the RAID device before
3978 reading it again it is possible to get stale data from the RAID
3979 cache instead of reading it from disk.
3982 Frequency : always for sles10 kernel
3984 Description: sles10 support
3985 Details : ll_follow_link: compile fixes and using of nd_set_link
3986 under newer kernels.
3989 Frequency : depends on arch, kernel and compiler version, always on sles10
3992 Description: recursive or deep enough symlinks cause stack overflow
3993 Details : getting rid of large stack-allocated variable in
3997 Frequency : depends on hardware
3999 Description: lustre write performance loss in the SLES10 kernel
4000 Details : the performance loss is caused by using of write barriers in the
4001 ext3 code. The SLES10 kernel turns barrier support on by
4002 default. The fix is to undo that change for ldiskfs.
4004 ------------------------------------------------------------------------------
4006 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4008 * Support for kernels:
4009 2.6.9-42.0.3EL (RHEL 4)
4010 2.6.5-7.276 (SLES 9)
4011 2.4.21-47.0.1.EL (RHEL 3)
4012 2.6.12.6 vanilla (kernel.org)
4016 Frequency : quota enabled and large files being deleted
4018 Description: releasing more than 4GB of quota at once hangs OST
4019 Details : If a user deletes more than 4GB of files on a single OST it
4020 will cause the OST to spin in an infinite loop. Release
4021 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4026 Description: statfs data retrieved from /proc may be stale or zero
4027 Details : When reading per-device statfs data from /proc, in the
4028 {kbytes,files}_{total,free,avail} files, it may appear
4029 as zero or be out of date.
4032 Frequency : systems with MD RAID1 external journal devices
4034 Description: lconf's call to blkid is confused by RAID1 journal devices
4035 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4036 of returning all block devices that match the journal UUID.
4039 Frequency : always, for aggregate stripe size over 4GB
4041 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4042 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4043 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4046 Frequency : always if "lfs find" used on a local file/directory
4048 Description: "lfs find" segfaults if used on a local file/directory
4049 Details : The case where a directory component was not specified wasn't
4050 handled correctly. Handle this properly.
4053 Frequency : always on ppc64
4055 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4056 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4057 perform endianness conversion.
4060 Frequency : rarely (truncate to non-zero file size after write under load)
4061 Bugzilla : 10730, 10687
4062 Description: Files padded with zeros to next 4K multiple
4063 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4064 it is possible that files that are truncated to a non-zero size
4065 immediately after being written are filled with zero bytes beyond
4066 the truncated size. No file data is lost.
4068 Severity : enhancement
4070 Description: Allow recovery/failover for liblustre clients.
4071 Details : liblustre clients were unaware of failover configurations until
4074 Severity : enhancement
4076 Description: user file locks should fail when not mounting with flock option
4077 Details : Set up an error-returning stub in ll_file_operations.lock field
4078 to prevent incorrect behaviour when client is mounted without
4079 flock option. Also, set up properly f_op->flock field for
4083 Frequency : always on ia64
4085 Description: "lfs df" loops on printing out MDS statfs information
4086 Details : The obd_ioctl_data was not initialized and in some systems
4087 this caused a failure during the ioctl that did not return
4088 an error. Initialize the struct and return an error on failure.
4091 Frequency : SLES 9 only
4093 Description: Error of copying files with lustre special EAs as root
4094 Details : Client side always return success for setxattr call for lustre
4095 special xattr (currently only "trusted.lov").
4098 Frequency : rarely on clusters with both ia64+i386 clients
4100 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4101 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4102 clients are doing concurrent writes to the same file it is
4103 possible that the ia64 clients may LASSERT because the OST
4104 extent locks are not PAGE_SIZE aligned. Ensure that grown
4105 locks are always aligned on the request boundary.
4108 Frequency : specific use, occasional
4110 Description: Overwriting in use executable truncates on-disk binary image
4111 Details : If one node attempts to overwrite an executable in use by
4112 another node, we now correctly return ETXTBSY instead of
4113 truncating the file.
4115 Severity : enhancement
4117 Description: Async OSC create to avoid the blocking unnecessarily.
4118 Details : If an OST has no remaining object, system will block on the creating
4119 when need to create a new object on this OST. Now, ways use
4120 pre-created objects when available, instead of blocking on an
4121 empty osc while others are not empty. If we must block, we block
4122 for the shortest possible period of time.
4127 Description: chmod on Lustre root is propagated to other clients
4128 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4129 invalid by the follow_mount time.
4132 Frequency : liblustre clients only
4134 Description: Race in 'instant cancel' lock handling could lead to such locks
4135 never to be granted in case of SMP MDS
4136 Details : Do not destroy not yet granted but cbpending locks in
4140 Frequency : replay/resend of open
4142 Description: non null lock assetion failure in mds_intent_policy
4143 Details : Trying to replay/resend lockless open requests resulted in
4144 mds_open() returning 0 with no lock. Now it sets a flag if
4145 a lock is going to be returned.
4147 Severity : enhancement
4149 Description: Checksum enhancements
4150 Details : New checksum enhancements allow for resending RPCs that failed
4153 Severity : enhancement
4155 Description: Tunables on number of dirty pages in cacche
4156 Details : Allow to set limit on number of dirty pages cached.
4161 Description: client crash on unmount - lock still has references
4162 Details : In some error handling cases it was possible to leak a lock
4163 reference on a client while accessing a file. This was not
4164 harmful to the client during operation, but would cause the
4165 client to crash when the filesystem is unmounted.
4168 Frequency : specific case, rare
4170 Description: ETXTBSY on mds though file not in use by client
4171 Details : ETXTBSY is no longer incorrectly returned when attempting to
4172 chmod or chown a directory that the user previously tried to
4173 execute or a currently-executing binary.
4176 Frequency : extremely rare except on liblustre-based clients
4178 Description: Lustre space not freed when files are deleted
4179 Details : Clean up open-unlinked files after client eviction. Previously
4180 the unlink was skipped and the files remained as orphans.
4185 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4186 Details : In some cases it was possible to send a blocking callback to
4187 a client doing a glimpse, even though that client didn't get
4188 a lock granted. When the glimpse lock is cancelled on the OST
4189 the freed lock is left on the waiting list and corrupted the list.
4192 Frequency : all core dumps
4194 Description: Broke core dumps to lustre
4195 Details : Negative dentry may be unhashed if parent does not have UPDATE
4196 lock, but some callers, e.g. do_coredump, expect dentry to be
4197 hashed after successful create, hash it in ll_create_it.
4199 ------------------------------------------------------------------------------
4201 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4203 * Support for kernels:
4204 2.6.9-42.0.2.EL (RHEL 4)
4205 2.6.5-7.276 (SLES 9)
4206 2.4.21-40.EL (RHEL 3)
4207 2.6.12.6 vanilla (kernel.org)
4211 Frequency : always on RHEL 3
4213 Description: Number of open files grows over time
4214 Details : The number of open files grows over time, whether or not
4215 Lustre is started. This was due to a filp leak introduced
4216 by one of our kernel patches.
4218 ------------------------------------------------------------------------------
4220 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4222 * Support for kernels:
4223 2.6.9-42.EL (RHEL 4)
4224 2.6.5-7.276 (SLES 9)
4225 2.4.21-40.EL (RHEL 3)
4226 2.6.12.6 vanilla (kernel.org)
4231 Bugzilla : 5719, 9635, 9792, 9684
4232 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4233 Details : If a server is under heavy load and cannot reply to new
4234 connection requests before the client resends the (re)connect,
4235 the connection handling code can behave badly if two service
4236 threads are concurrently handing separate (re)connections from
4237 the same client. Add better locking to the connection handling
4238 code, and ensure that only a single connection will be processed
4239 for a given client UUID, even if the lock is dropped.
4241 Severity : enhancement
4243 Description: add TCP zero-copy support to kernel
4244 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4245 sends if the hardware supports scatter-gather and checksumming.
4246 This allows socklnd to do client-write and server-read more
4247 efficiently and reduce CPU utilization from skbuf copying.
4250 Frequency : only if NFS exporting from client
4252 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4253 Details : If setting attributes on a file created under NFS that had
4254 never been opened it would be possible to oops the client
4255 if the file had no objects.
4258 Frequency : always for liblustre
4260 Description: liblustre client does MDS+OSTs setattr RPC for each write
4261 Details : When doing a write from a liblustre client, the client
4262 incorrectly issued an RPC to the MDS and each OST the file was
4263 striped over in order to update the timestamps. When writing
4264 with small chunks and many clients this could overwhelm the MDS
4265 with RPCs. In all cases it would slow down the write because
4266 these RPCs are unnecessary.
4268 Severity : enhancement
4270 Description: allow number of MDS service threads to be changed at module load
4271 Details : It is now possible to change the number of MDS service threads
4272 running. Adding "options mds mds_num_threads={N}" to the MDS's
4273 /etc/modprobe.conf will set the number of threads for the next
4274 time Lustre is restarted (assuming the "mds" module is also
4275 reloaded at that time). The default number of threads will
4276 stay the same, 32 for most systems.
4281 Description: OST crash if filesystem is unformatted or corrupt
4282 Details : If an OST is started on a device that has never been formatted
4283 or if the filesystem is corrupt and cannot even mount then the
4284 error handling cleanup routines would dereference a NULL pointer.
4289 Description: NULL pointer deref in llap_from_page.
4290 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4291 incorrect) mapping if the page was truncated/reclaimed while it was
4292 searched for. Check for this condition and skip such pages when
4293 doing readahead. Introduce extra check to llap_from_page() to
4294 verify page->mapping->host is non-NULL (so page is not anonymous).
4297 Frequency : Sometimes when using sys_sendfile
4299 Description: "page not covered by a lock" warnings from ll_readpage
4300 Details : sendfile called ll_readpage without right page locks present.
4301 Now we introduced ll_file_sendfile that does necessary locking
4302 around call to generic_file_sendfile() much like we do in
4306 Frequency : with certain MDS communication failures at client mount time
4308 Description: NULL pointer deref after failed client mount
4309 Details : a client connection request may delayed by the network layer
4310 and not be sent until after the PTLRPC layer has timed out the
4311 request. If the client fails the mount immediately it will try
4312 to clean up before the network times out the request. Add a
4313 reference from the request import to the obd device and delay
4314 the cleanup until the network drops the request.
4317 Frequency : occasionally during client (re)connect
4319 Description: assertion failure during client (re)connect
4320 Details : processing a client connection request may be delayed by the
4321 client or server longer than the client connect timeout. This
4322 causes the client to resend the connection request. If the
4323 original connection request is replied in this interval, the
4324 client may trip an assertion failure in ptlrpc_connect_interpret()
4325 which thought it would be the only running connect process.
4328 Frequency : only with obd_echo servers and clients that are rebooted
4330 Description: kernel BUG accessing uninitialized data structure
4331 Details : When running an obd_echo server it did not start the ping_evictor
4332 thread, and when a client was evicted an uninitialized data
4333 structure was accessed. Start the ping_evictor in the RPC
4334 service startup instead of the OBD startup.
4336 Severity : enhancement
4337 Bugzilla : 10193 (patchless)
4338 Description: Remove dependency on various unexported kernel interfaces.
4339 Details : No longer need reparent_to_init, exit_mm, exit_files,
4340 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4343 Frequency : rare (only users of deprecated and unsupported LDAP config)
4345 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4346 Details : LDAP apparently contains 'lustreName' attributes instead of
4347 'name'. A simple remapping of the name is sufficient.
4350 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4352 Description: waiting_locks_callback trips kernel BUG if client is evicted
4353 Details : Running with the dump_on_timeout debug flag turned on makes
4354 it possible that the waiting_locks_callback() can try to dump
4355 the Lustre kernel debug logs from an interrupt handler. Defer
4356 this log dumping to the expired_lock_main() thread.
4358 Severity : enhancement
4360 Description: Support NFS exporting on 2.6 kernels.
4361 Details : Implement non-rawops metadata methods for NFS server to use without
4362 changing NFS server code.
4365 Frequency : very rare (synthetic metadata workload only)
4367 Description: two racing renames might cause an MDS thread to deadlock
4368 Details : Running the "racer" program may cause one MDS thread to rename
4369 a file from being the source of a rename to being the target of
4370 a rename at exactly the same time that another thread is doing
4371 so, and the second thread has already enqueued these locks after
4372 doing a lookup of the target and is trying to relock them in
4373 order. Ensure that we don't try to re-lock the same resource.
4376 Frequency : only very large systems with liblustre clients
4378 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4379 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4380 all OSTs in parallel. This allows the eviction of stale liblustre
4381 clients to proceed much faster than if they were done in series,
4382 and also offers similar improvements for other set_info RPCs.
4387 Description: excessive CPU usage during initial read phase on client
4388 Details : During the initial read phase on a client, it would agressively
4389 retry readahead on the file, consuming too much CPU and impacting
4390 performance (since 1.4.5.8). Improve the readahead algorithm
4391 to avoid this, and also improve some other common cases (read
4392 of small files in particular, where "small" is files smaller than
4393 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4398 Description: MDS crash when receiving packet with unknown intent.
4399 Details : Do not LBUG in unknown intent case, just return -EFAULT
4401 Severity : enhancement
4402 Bugzilla : 9293, 9385
4403 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4404 Details : Do not serialize getattr (non-intent version) and statfs.
4407 Frequency : occasional, when OST network is overloaded/intermittent
4409 Description: client evicted by OST after bulk IO timeout
4410 Details : If a client sends a bulk IO request (read or write) the OST
4411 may evict the client if it is unresposive to its data GET/PUT
4412 request. This is incorrect if the network is overloaded (takes
4413 too long to transfer the RPC data) or dropped the OST GET/PUT
4414 request. There is no need to evict the client at all, since
4415 the pinger and/or lock callbacks will handle this, and the
4416 client can restart the bulk request.
4419 Frequency : Always when mmapping file with no objects
4421 Description: client crashes when mmapping file with no objects
4422 Details : Check that we actually have objects in a file before doing any
4423 operations on objects in ll_vm_open, ll_vm_close and
4429 Description: Request leak when working with deleted CWD
4430 Details : Introduce advanced request refcount tracking for requests
4431 referenced from lustre intent.
4433 Severity : Enhancement
4435 Description: Cache open file handles on client.
4436 Details : MDS now will return special lock along with openhandle, if
4437 requested and client is allowed to hold openhandle, even if unused,
4438 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4439 closing files for every read/write openration.
4441 Severity : Enhancement
4443 Description: Cache open negative dentries on client when possible.
4444 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4445 negative dentries on lock revocation.
4450 Description: Remounting a client read-only wasn't possible with a zconf mount
4451 Details : It wasn't possible to remount a client read-only with llmount.
4453 Severity : enhancement
4454 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4455 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4456 MPICH is included to add Lustre-specific ADIO interfaces.
4457 This is based closely on the UFS ADIO layer and only differs
4458 in file creation, in order to allow the OST striping to be set.
4459 This is user-contributed code and not supported by CFS.
4464 Description: extended inode attributes (immutable, append-only) work improperly
4465 when 2.4 and 2.6 kernels are used on client/server or vice versa
4466 Details : Introduce kernel-independent values for these flags.
4468 Severity : enhancement
4471 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4472 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4473 granularity. Now, Fractional MB granularity is be supported,
4474 this is very useful for low memory system.
4476 Severity : enhancement
4478 Description: Getattr by fid
4479 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4480 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4481 have locks and avoids getattr rpc requests for it on every stat.
4484 Frequency : Always, for filesystems larger than 2TB
4486 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4487 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4488 filesystems larger than 2TB to be mounted with mballoc enabled.
4491 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4493 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4494 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4495 filesystems larger than 2TB is mounted then the kernel will
4496 silently corrupt the start of the filesystem. CONFIG_LBD is
4497 enabled for all CFS-supported kernels, but the possibility of
4498 this happening with a modified kernel config exists.
4500 Severity : enhancement
4502 Description: add client O_DIRECT support for 2.6 kernels
4503 Details : It is now possible to do O_DIRECT reads and writes to files
4504 in the Lustre client mountpoint on 2.6 kernel clients.
4506 Severity : enhancement
4508 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4509 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4510 parallel, not waiting for response from every OST before sending
4511 a rpc to the next OST.
4516 Description: setattr vs write race when updating file timestamps
4517 Details : Client processes that update a file timestamp into the past
4518 right after writing to the file (e.g. tar) it is possible that
4519 the updated file modification time can be reset to the current
4520 time due to a race between processing the setattr and write RPC.
4522 Severity : enhancement
4524 Description: Bring 'lfs find' closer in line with regular Linux find.
4525 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4526 -print0 options and obtains all the needed info through the lustre
4529 Severity : enhancement
4531 Description: support up to 1024 configured devices on one node
4532 Details : change obd_dev array from statically allocated to dynamically
4533 allocated structs as they are first used to reduce memory usage
4538 Description: Flush dirty partially truncated pages during truncate
4539 Details : Immediatelly flush partially truncated pages in filter_setattr,
4540 this way we completely avoid having any pages in page cache on OST
4541 and can retire ugly workarounds during writes to flush such pages.
4546 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4547 Details : protect inode from truncation within vfs_unlink() context
4548 just take a reference before calling vfs_unlink() and release it
4549 when parent's i_sem is free.
4554 Description: last_id value checked outside lock on OST caused LASSERT failure
4555 Details : If there were multiple MDS->OST object precreate requests in
4556 flight, it was possible that the OST's last object id was checked
4557 outside a lock and incorrectly tripped an assertion. Move checks
4558 inside locks, and discard old precreate requests.
4561 Frequency : always, if extents are used on OSTs
4563 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4564 Details : OSTs using the extents format would not zero the high 16 bits of
4565 the index physical block number. This is not a problem for any
4566 OST filesystems smaller than 16TB, and no kernels support ext3
4567 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4568 new/modified files) and can be fixed for existing filesystems
4569 with e2fsprogs-1.39-cfs1.
4574 Description: import connection selection may be incorrect if timer wraps
4575 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4576 import connections to be ignored if the 32-bit jiffies counter
4577 wraps. Use a 64-bit jiffies counter.
4580 Frequency : very large clusters immediately after boot
4582 Description: LNET request buffers exhausted under heavy short-term load
4583 Details : If a large number of client requests are generated on a service
4584 that has previously never seen so many requests it is possible
4585 that the request buffer growth cannot keep up with the spike in
4586 demand. Instead of dropping incoming requests, they are held in
4587 the LND until the RPC service can accept more requests.
4590 Frequency : Sometimes during replay
4592 Description: Assertion failure in ll_local_open after replay.
4593 Details : If replay happened on an open request reply before we were able
4594 to set replay handler, reply will become not swabbed tripping the
4595 assertion in ll_local_open. Now we set the handler right after
4596 recognising of open request
4599 Frequency : very rare
4601 Description: kernel reports "badness in vsnprintf"
4602 Details : Reading from the "recovery_status" /proc file in small chunks
4603 may cause a negative length in lprocfs_obd_rd_recovery_status()
4604 call to vsnprintf() (which is otherwise harmless). Exit early
4605 if there is no more space in the output buffer.
4607 Severity : enhancement
4609 Description: clear OBD RPC statistics by writing to them
4610 Details : It is now possible to clear the OBD RPC statistics by writing
4611 to the "stats" file.
4616 Description: Client mtime is not the same on different clients after utimes
4617 Details : In some cases, the client was using the utimes() syscall on
4618 a file cached on another node. The clients now validate the
4619 ctime from the MDS + OSTs to determine which one is right.
4624 Description: Inability to activate failout mode
4625 Details : lconf script incorrectly assumed that in pythong string's numeric
4626 value is used in comparisons.
4629 Frequency : always with multiple stripes per file
4631 Description: Inefficient object allocation for mutli-stripe files
4632 Details : When selecting which OSTs to stripe files over, for files with
4633 a stripe count that divides evenly into the number of OSTs,
4634 the MDS is always picking the same starting OST for each file.
4635 Return the OST selection heuristic to the original design.
4640 Description: mount failures may take full timeout to return an error
4641 Details : Under some heavy load conditions it is possible that a
4642 failed mount can wait for the full obd_timeout interval,
4643 possibly several minutes, before reporting an error.
4644 Instead return an error as soon as the status is known.
4646 Frequency : quota enabled and large files being deleted
4648 Description: releasing more than 4GB of quota at once hangs OST
4649 Details : If a user deletes more than 4GB of files on a single OST it
4650 will cause the OST to spin in an infinite loop. Release
4651 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4656 Description: statfs data retrieved from /proc may be stale or zero
4657 Details : When reading per-device statfs data from /proc, in the
4658 {kbytes,files}_{total,free,avail} files, it may appear
4659 as zero or be out of date.
4662 Frequency : systems with MD RAID1 external journal devices
4664 Description: lconf's call to blkid is confused by RAID1 journal devices
4665 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4666 of returning all block devices that match the journal UUID.
4669 Frequency : always, for aggregate stripe size over 4GB
4671 Description: assertion fails when trying to use 4GB stripe size
4672 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4673 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4676 Frequency : always on ppc64
4678 Description: the first write on an ext3 filesystem with mballoc got stuck
4679 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4680 perform endianness conversion.
4682 ------------------------------------------------------------------------------
4684 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4686 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4687 INTEROPERATE with older versions automatically. Please read the
4688 user documentation before upgrading any part of a live system.
4689 * WARNING: Lustre networking configuration changes are required with
4690 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4693 * Support for kernels:
4694 2.6.9-22.0.2.EL (RHEL 4)
4695 2.6.5-7.244 (SLES 9)
4696 2.6.12.6 vanilla (kernel.org)
4699 Severity : enhancement
4700 Bugzilla : 7981/8208
4701 Description: Introduced Lustre Networking (LNET)
4702 Details : LNET is new networking infrastructure for Lustre, it includes
4703 a reorganized network configuration mode (see the user
4704 documentation for full details) as well as support for routing
4705 between different network fabrics. Lustre Networking Devices
4706 (LNDs) for the supported network fabrics have also been
4707 created for this new infrastructure.
4709 Severity : enhancement
4710 Description: Introduced Access control lists
4711 Details : clients can set ACLs on files and directories in order to have
4712 more fine-grained permissions than the standard Unix UGO+RWX.
4713 The MDS must be started with the "-o acl" mount option.
4715 Severity : enhancement
4716 Description: Introduced filesystem quotas
4717 Details : Administrators may now establish per-user quotas on the
4720 Severity : enhancement
4722 Description: Configuration change for the XT3
4723 The PTLLND is now used to run Lustre over Portals on the XT3
4724 The configure option(s) --with-cray-portals are no longer used.
4725 Rather --with-portals=<path-to-portals-includes> is used to
4726 enable building on the XT3. In addition to enable XT3 specific
4727 features the option --enable-cray-xt3 must be used.
4732 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4733 Details : A race between allocating a new llcd and re-getting the llcd_lock
4734 allowed another thread to grab newly-allocated llcd.
4736 Severity : enhancement
4738 Description: 2.6 OST async journal commit and locking fix to improve performance
4739 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4740 2.6 kernels are now async as they already were in 2.4 kernels so
4741 that they can commit concurrently with the network bulk transfer.
4742 For block-allocated files the filter allocation semaphore is held
4743 to avoid filesystem fragmentation during allocation. BKL lock
4744 removed for 2.6 xattr operations where it is no longer needed.
4749 Description: lconf incorrectly determined whether two IP networks could talk
4750 Details : In some more complicated routing and multiple-network
4751 configurations, lconf will avoid trying to make a network
4752 connection to a disjoint part of the IP space. It was doing the
4753 math incorrectly for one set of cases.
4758 Description: Fix for potential infinite loop processing records in an llog.
4759 Details : If an llog record is corrupted/zeroed, it is possible to loop
4760 forever in llog_process(). Validate the llog record length
4761 and skip the remainder of the block on error.
4764 Frequency : occasional (liblustre only)
4766 Description: liblustre could not open files whose last component is a symlink
4767 Details : sysio_path_walk() would incorrectly pass the open intent to
4768 intermediate path components.
4771 Frequency : rare (liblustre only with non-standard tuning)
4772 Bugzilla : 7201 (7350)
4773 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4774 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4775 after it has been released, only do this for referenced locks
4777 Severity : enhancement
4779 Description: specifying an (invalid) directory default stripe_size of -1
4780 would reset the directory default striping
4781 Details : stripe_size -1 was used internally to signal directory stripe
4782 removal, now use "all default" to signal dir stripe removal
4783 as a directory striping of "all default" is not useful
4786 Frequency : common for large clusters running liblustre clients
4788 Description: doing an ls when liblustre clients are running is slow
4789 Details : sending a glimpse AST to a liblustre client waits for every AST
4790 to time out, as liblustre clients will not respond. Since they
4791 cannot cache data we refresh the OST lock LVB from disk instead.
4793 Severity : enhancement
4795 Description: doing an ls at the same time as file IO can be slow
4796 Details : enqueue and other "small" requests can be blocked behind many
4797 large IO requests. Create a new OST IO portal for non-IO
4798 requests so they can be processed faster.
4801 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4803 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4804 created file with mode 000
4805 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4806 allow an ftruncate() as a non-root user to a file with mode 000.
4807 Lustre now respects this flag to disable mode checks when
4808 truncating a file owned by the user
4811 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4814 Description: Revoking locks from clients that went dead or catatonic might take
4816 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4817 cancellation of such locks instant on servers without waiting for
4818 any reply from clients. Clients drops these locks when cancel
4819 notification from server is received without replying.
4822 Frequency : liblustre-only, when liblustre client dies or becomes busy
4824 Description: Doing ls on Linux clients can take a long time with active
4826 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4827 granting such locks to it in the first place if possible. Locks
4828 are taken by proxy on the OST during the read or write and
4829 dropped immediately afterward. Add connect flags handling, do
4830 not grant locks to liblustre clients for glimpse ASTs.
4832 Severity : enhancement
4834 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4835 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4836 sequential reads and behaves badly with applications doing random
4837 reads. Improve it by reading ahead at least read region, and
4838 avoiding excessive large RPC for small reads.
4840 Severity : enhancement
4842 Description: Creating more than 1000 files for a single job may cause a load
4843 imbalance on the OSTs if there are also a large number of OSTs.
4844 Details : qos_prep_create() uses an OST index reseed value that is an
4845 even multiple of the number of available OSTs so that if the
4846 reseed happens in the middle of the object allocation it will
4847 still utilize the OSTs as uniformly as possible.
4852 Description: OST or MDS may oops in ping_evictor_main()
4853 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4854 but doesn't restart at beginning of obd_exports_timed list
4857 Severity : enhancement
4859 Description: improve by-nid export eviction on the MDS and OST
4860 Details : allow multiple exports with the same NID to be evicted at one
4861 time without re-searching the exports list.
4864 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4866 Description: MDS may oops in groups_free()
4867 Details : in rare race conditions a newly allocated group_info struct is
4868 freed again, and this can be NULL. The 2.4 compatibility code
4869 for groups_free() checked for a NULL pointer, but 2.6 did not.
4872 Frequency : common for liblustre clients doing little filesystem IO
4873 Bugzilla : 9352, 7313
4874 Description: server may evict liblustre clients accessing contended locks
4875 Details : if a client is granted a lock or receives a completion AST
4876 with a blocking AST already set it would not reply to the AST
4877 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4880 Frequency : lfs setstripe, only systems with more than 160 OSTs
4882 Description: unable to set striping with a starting offset beyond OST 160
4883 Details : llapi_create_file() incorrectly limited the starting stripe
4884 index to the maximum single-file stripe count.
4887 Frequency : LDAP users only
4889 Description: lconf did not handle in-kernel recovery with LDAP properly
4890 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4892 Severity : enhancement
4894 Description: bind OST threads to NUMA nodes to improve performance
4895 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4896 node and do their allocations there to localize memory access
4898 Severity : enhancement
4900 Description: llmount can determine client NID directly from Myrinet (GM)
4901 Details : the client NID code from gmnalnid was moved directly into
4902 llmount, removing the need to use this or specifying the
4903 client NID explicitly when mounting GM clients with zeroconf
4906 Frequency : if client is started with down MDS
4908 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4909 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4910 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4911 ptlrpc_import_delayed_req() because we didn't check if the
4912 request was interrupted, and we also didn't break out of the
4913 event loop if there was no timeout
4918 Description: data loss during non-page-aligned writes to a single file from
4919 both multiple nodes and multiple threads on one node at same time
4920 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4921 inconsistency led to false short-reads, that were cached and later
4922 used by ->prepare_write() to fill in partially written page,
4923 leading to data loss.
4926 Frequency : always, if lconf --abort_recovery used
4928 Description: lconf --abort_recovery fails with 'Operation not supported'
4929 Details : lconf was attempting to abort recovery on the MDT device and not
4932 Severity : enhancement
4934 Description: remove cleanup logs
4935 Details : replace lconf-generated cleanup logs with lustre internal
4936 cleanup routines. Eliminates the need for client-cleanup and
4939 Severity : enhancement
4941 Description: add support for EAs (user and system) on lustre filesystems
4942 Details : it is now possible to store extended attributes in the Lustre
4943 client filesystem, and with the user_xattr mount option it
4944 is possible to allow users to store EAs on their files also
4946 Severity : enhancement
4948 Description: Add possibility (config option) to show minimal available OST free
4950 Details : When compiled with --enable-mindf configure option, statfs(2)
4951 (and so, df) will return least minimal free space available from
4952 all OSTs as amount of free space on FS, instead of summary of
4953 free spaces of all OSTs.
4955 Severity : enhancement
4957 Description: do not expand extent locks acquired on OST-side
4958 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4959 by server: they are not cached anyway.
4962 Frequency : when mmap is used/binaries executed from Lustre
4964 Description: Unmmap pages before throwing them away from read cache.
4965 Details : llap_shrink cache now attempts to unmap pages before discarding
4966 them (if unmapping failed - do not discard). SLES9 kernel has
4967 extra checks that trigger if this unmapping is not done first.
4972 Description: lconf didn't resolve symlinks before checking to see whether a
4973 given mountpoint was already in use
4976 Frequency : when migrating failover services
4977 Bugzilla : 6395, 9514
4978 Description: When migrating a subset of services from a node (e.g. failback
4979 from a failover service node) the remaining services would
4980 time out and evict clients.
4981 Details : lconf --force (implied by --failover) sets the global obd_timeout
4982 to 5 seconds in order to quickly disconnect, but this caused
4983 other RPCs to time out too quickly. Do not change the global
4984 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4986 Severity : enhancement
4987 Frequency : if MDS is started with down OST
4988 Bugzilla : 9439,5706
4989 Description: Allow startup/shutdown of an MDS without depending on the
4990 availability of the OSTs.
4991 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4992 Add appropriate locking and lov-osc refcounts for safe
4993 cleaning. Add osc abort_inflight calls in case the
4994 synchronize never started.
4997 Frequency : occasional (Cray XT3 only)
4999 Description: root not authorized to access files in CRAY_PORTALS environment
5000 Details : The client process capabilities were not honoured on the MDS in
5001 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5002 been accessed by an authorized user then root was able to access
5003 the file on the local client also. The root user capabilities
5004 are now allowed on the MDS, as this environment has secure UID.
5007 Frequency : occasional
5009 Description: ldiskfs "too long searching" message happens too often
5010 Details : A debugging message (otherwise harmless) prints too often on
5011 the OST console. This has been reduced to only happen when
5012 there are fragmentation problems on the filesystem.
5017 Description: Division by zero in statfs when all OSCs are inactive
5018 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5019 lov_statfs divides by value returned by lov_get_stripecnt().
5023 Bugzilla : 9489, 3273
5024 Description: First write from each client to each OST was only 4kB in size,
5025 to initialize client writeback cache, which caused sub-optimal
5026 RPCs and poor layout on disk for the first written file.
5027 Details : Clients now request an initial cache grant at (re)connect time
5028 and so that they can start streaming writes to the cache right
5029 away and always do full-sized RPCs if there is enough data.
5030 If the OST is rebooted the client also re-establishes its grant
5031 so that client cached writes will be honoured under the grant.
5036 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5037 Details : Now I/O RPCs go to different portal number and (presumably) fast
5038 lock requests (and glimses) and other RPCs get their own service
5039 threads pool that should be able to service those RPCs
5042 Severity : enhancement
5044 Description: Ability to exchange lustre version between client and servers and
5045 issue warnings at client side if client is too old. Also for
5046 liblustre clients there is ability to refuse connection of too old
5048 Details : New 'version' field is added to connect data structure that is
5049 filled with version info. That info is later checked by server and
5053 Frequency : rare, liblustre only.
5054 Bugzilla : 9296, 9581
5055 Description: Two simultaneous writes from liblustre at offset within same page
5056 might proceed at the same time overwriting eachother with stale
5058 Details : I/O lock withing llu_file_prwv was released too early, before data
5059 actually was hitting the wire. Extended lock-holding time until
5060 server acknowledges receiving data.
5063 Frequency : extremely rare. Never observed in practice.
5065 Description: avoid generating lustre_handle cookie of 0.
5066 Details : class_handle_hash() generates handle cookies by incrementing
5067 global counter, and can hit 0 occasionaly (this is unlikely, but
5068 not impossible, because initial value of cookie counter is
5069 selected randonly). Value of 0 is used as a sentinel meaning
5070 "unassigned handle" --- avoid it. Also coalesce two critical
5071 sections in this function into one.
5073 Severity : enhancement
5075 Description: allow liblustre clients to delegate truncate locking to OST
5076 Details : To avoid overhead of locking, liblustre client instructs OST to
5077 take extent lock in ost_punch() on client's behalf. New connection
5078 flag is added to handle backward compatibility.
5080 Severity : enhancement
5081 Bugzilla : 4928, 7341, 9758
5082 Description: allow number of OST service threads to be specified
5083 Details : a module parameter allows the number of OST service threads
5084 to be specified via "options ost ost_num_threads={N}" in the
5085 OSS's /etc/modules.conf or /etc/modprobe.conf.
5089 Bugzilla : 6146, 9635, 9895
5090 Description: servers crash with bad pointer in target_handle_connect()
5091 Details : In rare cases when a client is reconnecting it was possible that
5092 the connection request was the last reference for that export.
5093 We would temporarily drop the export reference and get a new
5094 one, but this may have been the last reference and the export
5095 was just destroyed. Get new reference before dropping old one.
5097 Severity : enhancement
5098 Frequency : if client is started with failover MDS
5100 Description: Allow multiple MDS hostnames in the mount command
5101 Details : Try to read the configuration from all specified MDS
5102 hostnames during a client mount in case the "primary"
5105 Severity : enhancement
5107 Description: Stop sending data to evicted clients as soon as possible.
5108 Details : Check if the client we are about to send or are sending data to
5109 was evicted already. (Check is done every second of waiting,
5110 for which l_wait_event interface was extended to allow checking
5111 of exit condition at specified intervals).
5114 Frequency : rare, normally only when NFS exporting is done from client
5116 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5118 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5121 Frequency : always, if srand() is called before liblustre initialization
5123 Description: Liblustre uses system PRNG disturbing its usage by user application
5124 Details : Introduce internal to lustre fast and high-quality PRNG for
5125 lustre usage and make liblustre and some other places in generic
5126 lustre code to use it.
5128 Severity : enhancement
5129 Bugzilla : 9477, 9557, 9870
5130 Description: Verify that the MDS configuration logs are updated when xml is
5131 Details : Check if the .xml configuration logs are newer than the config
5132 logs stored on the MDS and report an error if this is the case.
5133 Request --write-conf, or allow starting with --old_conf.
5135 Severity : enhancement
5137 Description: Handle symlinks in the path when checking if Lustre is mounted.
5138 Details : Resolve intermediate symlinks when checking if a client has
5139 mounted a filesystem to avoid duplicate client mounts.
5144 Description: lconf can hit an error exception but still return success.
5145 Details : The lconf command catches the Command error exception at the top
5146 level script context and will exit with the associated exit
5147 status, but doesn't ensure that this exit status is non-zero.
5152 Description: failure of ptlrpc thread startup can cause oops
5153 Details : Starting a ptlrpc service thread can fail if there are a large
5154 number of threads or the server memory is very fragmented.
5155 Handle this without oopsing.
5158 Frequency : always, only if liblustre and non-default acceptor port was used
5160 Description: liblustre cannot connect to servers with non-default acceptor port
5161 Details : tcpnal_set_default_params() was not called and was therefore
5162 ignoring the environment varaible TCPNAL_PORT, as well as other
5163 TCPNAL_ environment variables
5168 Description: two objects could be created on the same OST for a single file
5169 Details : If an OST is down, in some cases it was possible to create two
5170 objects on a single OST for a single file. No problems other
5171 than potential performance impact and spurious error messages.
5175 Bugzilla : 5681, 9562
5176 Description: Client may oops in ll_unhash_aliases
5177 Details : Client dcache may become inconsistent in race condition.
5178 In some cases "getcwd" can fail if the current directory is
5184 Description: Inode refcounting problems in NFS export code
5185 Details : link_raw functions used to call d_instantiate without obtaining
5186 extra inode reference first.
5190 Bugzilla : 9942, 9903
5191 Description: Referencing freed requests leading to crash, memleaks with NFS.
5192 Details : We used to require that call to ll_revalidate_it was always
5193 followed by ll_lookup_it. Also with revalidate_special() it is
5194 possible to call ll_revalidate_it() twice for the same dentry
5195 even if first occurence returned success. This fix changes semantic
5196 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5197 reference on a request referred from the intent.
5198 ll_intent_release() then releases such a request.
5201 Frequency : rare, normally benchmark loads only
5203 Description: unlinked inodes were kept in memory on the client
5204 Details : If a client is repeatedly creating and unlinking files it
5205 can accumulate a lot of stale inodes in the inode slab cache.
5206 If there is no other client load running this can cause the
5207 client node to run out of memory. Instead flush old inodes
5208 from client cache that have the same inode number as a new inode.
5211 Frequency : SLES9 2.6.5 kernel and long filenames only
5212 Bugzilla : 9969, 10379
5213 Description: utime reports stale NFS file handle
5214 Details : SLES9 uses out-of-dentry names in some cases, which confused
5215 the lustre dentry revalidation. Change it to always use the
5219 Frequency : rare, unless heavy write-truncate concurrency is continuous
5220 Bugzilla : 4180, 6984, 7171, 9963, 9331
5221 Description: OST becomes very slow and/or deadlocked during object unlink
5222 Details : filter_destroy() was holding onto the parent directory lock
5223 while truncating+unlinking objects. For very large objects this
5224 may block other threads for a long time and slow overall OST
5225 responsiveness. It may also be possible to get a lock ordering
5226 deadlock in this case, or run out of journal credits because of
5227 the combined truncate+unlink. Solution is to do object truncate
5228 first in one transaction without parent lock, and then do the
5229 final unlink in a new transaction with the parent lock. This
5230 reduces the lock hold time dramatically.
5233 Frequency : rare, 2.4 kernels only
5235 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5236 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5237 before unmounting their filesystem in order to avoid deadlock.
5238 The kernel_locked() function in 2.4 kernels only checks whether
5239 the kernel lock is held, not whether it is this process that is
5240 holding it as 2.6 kernels do.
5245 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5246 Details : The client ptlrpc code may be trying to reconnect to a down
5247 server before a previous connection attempt has timed out.
5248 Increase the reconnect interval to be longer than the connection
5249 timeout interval to avoid sending duplicate connections to
5253 Frequency : echo_client brw_test command
5255 Description: fix echo_client to work with OST preallocated code
5256 Details : OST preallocation code (5137) didn't take echo_client IO path
5257 into account: echo_client calls filter methods outside of any
5258 OST thread and, hence, there is no per-thread preallocated
5259 pages and buffers to use. Solution: hijack pga pages for IO. As
5260 a byproduct, this avoids unnecessary data copying.
5264 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5265 Description: Client can oops in mdc_commit_close() after open replay
5266 Details : It was possible for the MDS to return an open request with no
5267 transaction number in mds_finish_transno() if the client was
5268 evicted, but without actually returning an error. Clients
5269 would later try to replay that open and may trip an assertion
5270 Simplify the client close codepath, and always return an error
5271 from the MDS in case the open is not successful.
5274 Frequency : rare, 2.6 OSTs only
5276 Description: OST may deadlock under high load on fragmented files
5277 Details : If there was a heavy load and highly-fragmented OST filesystems
5278 it was possible to have all the OST threads deadlock waiting on
5279 allocation of biovecs, because the biovecs were not released
5280 until the entire RPC IO was completed. Instead, release biovecs
5281 as soon as they are complete to ensure forward IO progress.
5283 Severity : enhancement
5285 Description: Support for specifying external journal device at mount
5286 Details : If an OST or MDS device is formatted with an external journal
5287 device, this device major/minor is stored in the ext3 superblock
5288 and may not be valid for failover. Allow detecting and
5289 specifying the external journal at mount time.
5294 Description: Mounting an MDS with pending unlinked files may cause oops
5295 Details : target_finish_recovery() calls mds_postrecov() which returned
5296 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5297 considers this an error and immediately begins cleaning up the
5298 lov, just after starting the mds_lov process
5300 Severity : enhancement
5302 Description: Implement 'lfs df' to report actual free space on per-OST basis
5303 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5304 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5305 usage of objects; '-h' to report in human readable format.
5307 ------------------------------------------------------------------------------
5309 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5316 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5317 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5318 at startup during orphan unlinking. Instead, skip buddy update
5319 before setup, as it will be regenerated after recovery is complete.
5324 Description: If an OST is inactive, its locks might reference stale inodes.
5325 Details : lov_change_cbdata() must iterate over all namespaces, even if
5326 they are inactive to clear inode references from the lock.
5328 Severity : enhancement
5329 Frequency : occasional, if non-standard max_dirty_mb used
5331 Description: Client will block write RPCs if not enough grant
5332 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5333 then the client will block writes while waiting for another RPC
5334 to complete instead of consuming its dirty limit. With change
5335 we get improved performance when max_dirty_mb is small.
5337 Severity : enhancement
5338 Bugzilla : 3389, 6253
5339 Description: Add support for supplementary groups on the MDS.
5340 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5341 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5342 lookups for user supplementary groups into a cache.
5346 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5347 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5348 for liblustre O_EXCL is enforced by MDS.
5352 Description: readdir never returns NULL in liblustre.
5353 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5354 dentry in struct dirent.
5358 Frequency : liblustre only, depends on application IO pattern
5359 Description: liblustre clients evicted if not contacting servers
5360 Details : Don't put liblustre clients into the ping_evictor list, so
5361 they will not be evicted by the pinger ever.
5363 Severity : enhancement
5365 Description: Add ability to evict clients by NID from MDS.
5366 Details : By echoing "nid:$NID" string into
5367 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5368 $NID would be instantly evicted from this MDS and from all active
5369 OSTs connected to it.
5373 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5374 Details : lookup_it_finish() used to query file size from OSTs that was not
5379 Description: service threads change working directory to that of init
5380 Details : Starting lustre service threads may pin the working directory
5381 of the parent thread, making that filesystem busy. Threads
5382 now change to the working directory of init to avoid this.
5386 Frequency : during shutdown only
5387 Description: shutdown with a failed MDS or OST can cause unmount to hang
5388 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5393 Frequency : 2.6 only
5394 Description: chmod/chown may include an extra supplementary group
5395 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5396 supplementary group and if none is specified this is used.
5399 Bugzilla : 5479 (6816)
5401 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5402 Details : If lookup is in progress on a file that is unlinked we might try
5403 to revalidate the inode and fail in revalidate after lookup is
5404 complete and ll_file_open() enqueues the open again but
5405 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5408 Frequency : always, if lconf --abort_recovery used
5410 Description: lconf --abort_recovery fails with 'Operation not supported'
5411 Details : lconf was attempting to abort recovery on the MDT device and not
5414 ------------------------------------------------------------------------------
5416 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5421 Frequency : rare (only unsupported configurations with a node running as an
5423 Bugzilla : 6514, 5137
5424 Description: Mounting a Lustre file system on a node running as an OST could
5426 Details : OSTs now preallocates memory needed to write out data at
5427 startup, instead of when needed, to avoid having to
5428 allocate memory in possibly low memory situations.
5429 Specifically, if the file system is mounted on on OST,
5430 memory pressure could force it to try to write out data,
5431 which it needed to allocate memory to do. Due to the low
5432 memory, it would be unable to do so and the node would
5433 become unresponsive.
5435 Severity : enhancement
5437 Description: Addition of lconf --service command line option
5438 Details : lconf now accepts a '--service <arg>' option, which is
5439 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5441 Severity : enhancement
5443 Description: Failover mode is now the default for OSTs.
5444 Details : By default, OSTs will now run in failover mode. To return to
5445 the old behaviour, add '--failout' to the lmc line for OSTs.
5447 Severity : enhancement
5449 Description: Health checks are now provided for MDS and OSTs
5450 Details : Additional detailed health check information on MSD and OSTs
5451 is now provided through the procfs health_check value.
5454 Frequency : occasional, depends on IO load
5456 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5457 after numerous create/delete cycles
5458 Details : The ext3 inode allocation policy would not allocate new inodes
5459 very well on the OSTs because there are no new directories
5460 being created. Instead we look for groups with free space if
5461 the parent directories are nearly full.
5466 Description: Network or server problems during mount may cause partially
5467 mounted clients instead of returning an error.
5468 Details : The config llog parsing code may overwrite the error return
5469 code during mount error handling, returning success instead
5475 Description: MDS can fail to allocate large reply buffers
5476 Details : After long uptimes the MDS can fail to allocate large reply
5477 buffers (e.g. zconf client mount config records) due to memory
5478 fragmentation or consumption by the buffer cache. Preallocate
5479 some large reply buffers so that these replies can be sent even
5480 under memory pressure.
5484 Frequency : rare (liblustre)
5485 Description: fsx running with liblustre complained that using truncate() to
5486 extend the file doesn't work. This patch corrects that issue.
5487 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5488 ATTR_SIZE and lsm use in llu_setattr_raw.
5492 Frequency : rare, only 2.6 kernels
5493 Description: Unusual file access patterns on the MDS may result in inode
5494 data being lost in very rare circumstances.
5495 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5496 code in the 2.6 kernel caused the inode and/or EA data not to
5497 be read from disk, causing single-file corruption.
5501 Frequency : rare, only 2.6 filesystems using extents
5502 Description: Heavy concurrent write and delete load may cause data corruption.
5503 Details : It was possible under high-load situations to have an extent
5504 metadata block in the block device cache from a just-unlinked
5505 file overwrite a newly-allocated data block. We now unmap any
5506 metadata buffers that alias just-allocated data blocks.
5510 Frequency : filesystems with default stripe_count larger than 77
5511 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5512 Details : lconf specifies an inode size of 4096 bytes when the default
5513 stripe_count is larger than 77. This conflicts with the default
5514 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5515 this case to avoid pinning too much memory for large EAs.
5517 ------------------------------------------------------------------------------
5519 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5524 Frequency : rare (extremely heavy IO load with hundreds of clients)
5526 Description: Client is evicted, gets IO error writing to file
5527 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5528 caused clients to be evicted for AST timeouts. The fixes in
5529 bug 5192 mean we no longer need to have such short AST timeouts
5530 so ldlm_timeout has been increased.
5533 Frequency : occasional during --force or --failover shutdown under load
5534 Bugzilla : 5949, 4834
5535 Description: Server oops/LBUG if stopped with --force or --failover under load
5536 Details : a collection of import/export refcount and cleanup ordering
5537 issues fixed for safer force cleanup
5540 Frequency : only filesystems larger than 120 OSTs
5541 Bugzilla : 5990, 6223
5542 Description: lfs getstripe would oops on a very large filesystem
5543 Details : lov_getconfig used kfree on vmalloc'd memory
5546 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5547 Bugzilla : 6242, 6243
5548 Description: reading from files that had been truncated to a non-zero size
5549 but never opened returned no data
5550 Details : ll_file_read() reads zeros from no-object files to EOF
5555 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5556 Details : There's little harm in aborting MDS/OSS recovery and letting it
5557 try again, so I removed the LASSERT and return an error instead.
5559 Severity : enhancement
5561 Description: New debugging infrastructure for tracking down data corruption
5562 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5563 (b) cover more of the client-side code path, and (c) try to narrow
5564 down where problems occurred
5568 Bugzilla : 3819, 4364, 4397, 6313
5569 Description: Racing close and eviction MDS could cause assertion in mds_close
5570 Details : It was possible to get multiple mfd references during close and
5571 client eviction, leading to one thread referencing a freed mfd.
5573 Severity: : enhancement
5574 Bugzilla : 3262, 6359
5575 Description: Attempts to reconnect to servers are now more aggressive.
5576 Details : This builds on the enhanced upcall-less recovery that was added
5577 in 1.4.2. When trying to reconnect to servers, clients will
5578 now try each server in the failover group every 10 seconds. By
5579 default, clients would previously try one server every 25 seconds.
5584 Description: After recovery, certain operations trigger a failed
5585 assertion on a client.
5586 Details : Failing over an mds, using lconf -d --failover, while a
5587 client was doing a readdir() call would cause the client to
5588 LBUG after recovery completed and the readdir() was resent.
5590 Severity : enhancement
5592 Description: Default groups are now added by lconf
5593 Details : You can now run lconf --group <servicename> without having to
5594 manually add groups with lmc.
5597 Frequency : occasional
5599 Description: Nodes with an elan id of 0 trigger a failed assertion
5602 Frequency : always when accessing e.g. tty/console device nodes
5604 Description: tty and some other devices nodes cannot be used on lustre
5605 Details : file's private_data field is used by device data and lustre
5606 values in there got lost. New field was added to struct file to
5607 store fs-specific private data.
5610 Frequency : when exporting Lustre via NFS
5612 Description: NFSD failed occasionally when looking up a path component
5613 Details : NFSD is looking up ".." which was broken in ext3 directories
5614 that had grown large enough to become hashed.
5617 Frequency : Clusters with multiple interfaces not on the same subnet
5619 Description: Nodes will repeatedly try to reconnect to an interface which it
5620 cannot reach and report an error to the log.
5621 Details : Extra peer list entries will be created by lconf with some peers
5622 unreachable. lconf now validates the peer before adding it.
5625 Frequency : Only if a default stripe is set on the filesystem root.
5627 Description: Setting a default stripe on the filesystem root prevented the
5628 filesystem from being remounted.
5629 Details : The client was sending extra request flags in the root getattr
5630 request and did not allocate a reply buffer for the dir EA.
5633 Frequency : occasional, higher if lots of files are accessed by one client
5634 Bugzilla : 6159, 6097
5635 Description: Client trips assertion regarding lsm mismatch/magic
5636 Details : While revalidating inodes the VFS looks up inodes with ifind()
5637 and in rare cases can find an inode that is being freed.
5638 The ll_test_inode() code will free the lsm during ifind()
5639 when it finds an existing inode and then the VFS later attaches
5640 this free lsm to a new inode.
5644 Bugzilla : 6422, 7030
5645 Description: MDS deadlock between mkdir and client eviction
5646 Details : Creating a new file via mkdir or mknod (starting a transaction
5647 and getting the ns lock) can deadlock with client eviction
5648 (gets ns lock and trying to finish a synchronous transaction).
5651 Frequency : occasional
5652 Description: While starting a server, the fsfilt_ext3 module could not be
5654 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5655 kernels. Previously, lconf would still use the ext3 name
5656 when trying to load modules. Now, it will correctly use
5657 ext3 on 2.4 and ldiskfs on 2.6.
5659 Severity : enhancement
5660 Description: The default stripe count has been changed to 1
5661 Details : The interpretation of the default stripe count (0, to lfs
5662 or lmc) has been changed to mean striping across a single
5663 OST, rather than all available. For general usage we have
5664 found a stripe count of 1 or 2 works best.
5666 Severity : enhancement
5667 Description: Add support for compiling against Cray portals.
5668 Details : Conditional compiling for some areas that are different
5672 Frequency : occasional
5673 Bugzilla : 6409, 6834
5674 Description: Creating files with an explicit stripe count may lead to
5675 a failed assertion on the MDS
5676 Details : If some OSTs are full or unavailable, creating files may
5677 trigger a failed assertion on the MDS. Now, Lustre will
5678 try to use other servers or return an error to the
5682 Frequency : occasional
5684 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5685 caused lock timeout during readahead (since 1.4.2).
5686 Details : Processes doing readahead might match a lock that hasn't been
5687 granted yet if there are overlapping and conflicting lock
5688 requests. The readahead process waits on ungranted lock
5689 (original lock is CBPENDING), while OST waits for that process
5690 to cancel CBPENDING read lock and eventually evicts client.
5692 Severity : enhancement
5694 Description: Initial enabling of flock support for clients
5695 Details : Implements fcntl advisory locking and file status functions.
5696 This feature is provided as an optional mount flag (default
5697 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5698 locking are implemented yet, and those that are are not guaranteed
5699 to be completely correct in production environments.
5700 mount -t lustre -o [flock|noflock] ...
5703 Frequency : occasional
5705 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5706 assertion in the ext3 JBD (journaling) layer.
5707 Details : The b_committed_data struct is protected by the big kernel lock
5708 in 2.4 kernels, serializing journal_commit_transaction() and
5709 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5710 access to this struct. In 2.6 kernels there is finer grained
5711 locking to improve SMP performance of the JBD layer.
5715 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5720 Description: merge and cleanup kernel patches.
5723 -----------------------------------------------------------------------------
5725 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5727 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5728 versions of Lustre. Please update all servers and clients to
5729 version 1.4.2 or later at the same time. You must also run
5730 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5731 to update the configuration logs.
5733 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5734 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5735 - swab llog records of type '0' so we get proper header size/idx (5861)
5736 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5737 - fix rename of one directory over another leaking an inode (5953)
5738 - avoid SetPageDirty on 2.6 (5981)
5739 - don't re-add just-being-destroyed locks to the waiting list (5653)
5740 - when creating new directories, inherit the parent's custom
5741 striping settings if present parent (3048)
5742 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5743 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5744 - don't hold client locks on temporary worklist from l_lru (5666)
5745 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5746 - automatically evict dead clients (5921)
5747 - Update file size properly in create+truncate+fstat case (6196)
5748 - Do not unhash mountpoint dentries, do not allow removal of
5750 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5751 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5752 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5753 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5754 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5755 - Correctly handle reads of files with no objects (6243)
5756 - lctl recover will also mark a device active if deactivate used (5933)
5758 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5759 - allow --write-conf on an MDS with different nettype than client (5619)
5760 - don't write config llogs to MDS for mounts not from that MDS (5617)
5761 - lconf should create multiple TCP connections from a client (5201)
5762 - init scripts are now turned off by default; run chkconfig --on
5763 lustre and chkconfig --on lustrefs to use them
5764 - upcalls are no longer needed for clients to recover to failover
5766 - add --abort-recovery option to lconf to abort recovery on device
5768 - add support for an arbitrary number of OSTs (3026)
5769 - Quota support protocol changes.
5770 - forward compatibility changes to wire structs (6007)
5771 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5772 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5773 - improved "lustre status" script
5774 - initialize blocksize for non-regular files (6062)
5775 - added --disable-server and --disable-client configure options (5782)
5776 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5777 - Vanilla 2.4.29 support
5778 - increase maximum number of obd devices to 520 (6242)
5779 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5780 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5781 SLES 9 2.6.5 kernels
5782 - limit stripes per file to 160 (the maximum EA size) (6093)
5784 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5787 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5788 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5789 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5790 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5791 - small ext3 extents cleanups and fixes (5733)
5792 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5793 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5794 - avoid needless client->OST connect, fix handle mismatch (5317)
5795 - fix DLM error path that led to out-of-sync client, long delays (5779)
5796 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5797 - fix several locking issues related to i_size (5492,5624,5654,5672)
5798 - don't move pending lock onto export if it is already evicted (5683)
5799 - fix kernel oops when creating .foo in unlinked directory (5548)
5800 - fix deadlock in obdfilter statistics vs. object create (5811)
5801 - use time_{before,after} to avoid timer jiffies wrap (5882)
5802 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5803 - Do not leak request if server was not able to process it (5154)
5804 - If mds_open unable to find parent dir, make that negative lookup(5154)
5805 - don't create new directories with extent-mapping (5909, 5936)
5807 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5808 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5809 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5810 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5811 - support for RHEL4 kernel on i686 (5773)
5812 - provide error messages when incompatible logs are encountered (5898)
5814 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5815 * version 1.4.0.10 (1.4.1 release candidate 1)
5817 - don't keep a lock reference when lock is not granted (4238)
5818 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5819 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5820 - drop import inflight refcount on signal_completed_replay error (5255)
5821 - unlock page after async write error during send (3677)
5822 - handle missing objects in filter_preprw_read properly (5265)
5823 - no transno return for symlink open, don't save no-trasno open (3440)
5824 - don't try to complete elan receive that already failed (4012)
5825 - free RPC server reply state on error (5406)
5826 - clean up thread from ptlrpc_start_thread() on error (5160)
5827 - readahead could read extra page into cache that wasn't ejected (5388)
5828 - prevent races in class_attach/setup/cleanup/detach (5260)
5829 - don't dereference de->d_inode after l_dput of de (5458)
5830 - use "int" for stripe value returned from lock_to_stripe (5544)
5831 - mballoc allocation and error-checking fixes in 2.6 (5504)
5832 - block device patches to fix I/O request sizes in 2.6 (5482)
5833 - look up hostnames for IB nals (5602)
5834 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5835 - don't start multiple acceptors for the same port (5277)
5836 - fix incorrect LASSERT in mds_getattr_name (5635)
5837 - export a proc file for general "ping" checking (5628)
5838 - fix "lfs check" to not block when the MDS is down (5628)
5840 - service request history (4965)
5841 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5842 - create an "evict_client" /proc entry on OSTs, like the MDS has
5843 - fix mount usage message, return errors per mount(8) (5168)
5844 - change grep [] to grep "[]" in tests so they work in more UMLs
5845 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5846 - remove ancient LOV_MAGIC_V0 EA support (5047)
5847 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5848 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5849 - allow readahead allocations to fail when low on memory (5383)
5850 - mmap locking landed again, after considerable improvement (2828)
5851 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5853 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5856 - send OST transaction number in read/write reply to free req (4966)
5857 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5858 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5859 - the watchdog thread now runs as interruptible (5246)
5860 - flock/lockf fixes (but it's still disabled, pending 5135)
5861 - don't use EXT3 constants in llite code (5094)
5862 - memory shortage at startup could cause assertion (5176)
5864 - reorganization of lov code
5865 - single portals codebase
5867 - add extents/mballoc support (5025)
5868 - direct I/O reads in the obdfilter (4048)
5869 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5871 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5874 - send OST transaction number in read/write reply to free req (4966)
5875 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5876 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5877 - the watchdog thread now runs as interruptible (5246)
5878 - handle missing objects in filter_preprw_read properly (5265)
5879 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5880 - drop import inflight refcount on signal_completed_replay error (5255)
5881 - unlock page after async write error during send (3677)
5882 - return original error code on reconstructed replies (3761)
5883 - no transno return for symlink open, don't save no-trasno open (3440)
5885 - add pid to ldlm debugging output (4922)
5886 - bump the watchdog timeouts -- we can't handle 30sec yet
5887 - extra debugging for orphan dentry/inode bug (5259)
5889 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5892 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5893 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5894 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5895 - fix NULL dereference in /proc/sys/portals/routes (4827)
5896 - allow failed mdc_close() operations to be interrupted (4561)
5897 - stop precreate on OST before MDS would time out on it (4778)
5898 - don't send partial-page writes before EOF from client (4410)
5899 - discard client grant for sub-page writes on large-page clients (4520)
5900 - don't free dentries not owned by NFS code, check generation (4806)
5901 - fix lsm leak if mds_create_objects() fails (4801)
5902 - limit debug_daemon file size, always print CERROR messages (4789)
5903 - use transno after validating reply (3892)
5904 - process timed out requests if import state changes (3754)
5905 - update mtime on OST during writes, return in glimpse (4829)
5906 - add mkfsoptions to LDAP (4679)
5907 - use ->max_readahead method instead of zapping global ra (5039)
5908 - don't interrupt __l_wait_event() during strace
5910 - add software watchdogs to catch hung threads quickly (4941)
5911 - make lustrefs init script start after nfs is mounted
5912 - fix CWARN/ERROR duplication (4930)
5913 - return async write errors to application if possible (2248)
5914 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5915 - print NAL number in %x format (4645)
5916 - update barely-supported suse-2.4.21-171 series (4842)
5917 - support for sles 9 %post scripts
5918 - support for building 2.6 kernel-source packages
5919 - support for sles km_* packages
5921 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5924 - ignore -ENOENT errors in osc_destroy (3639)
5925 - notify osc create thread that OSC is being cleaned up (4600)
5926 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5927 - reconstruct ost_handle() like mds_handle() (4657)
5928 - create a new thread to do import eviction to avoid deadlock (3969)
5929 - let lconf resolve symlinked-to devices (4629)
5930 - don't unlink "objects" from directory with default EA (4554)
5931 - hold socknal file ref over connect in case target is down (4394)
5932 - allow more than 32000 subdirectories in a single directory (3244)
5933 - fix blocks count for O_DIRECT writes (3751)
5934 - OST returns ENOSPC from object create when no space left (4539)
5935 - don't send truncate RPC if file size isn't changing (4410)
5936 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5937 - bind to privileged port in socknal and tcpnal (3689)
5939 - rate limit CERROR/CWARN console message to avoid overload (4519)
5940 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5941 - basic mmap support (3918)
5942 - kernel patch series update from b1_4 (4711)
5944 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5947 - avoid crash during MDS cleanup with OST shut down (2775)
5948 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5949 - don't use bad inodes on the MDS (3744)
5950 - dynamic object preallocation to improve recovery speed (4236)
5951 - don't hold spinlock over lock dumping or change debug flags (4401)
5952 - don't zero obd_dev when it is force cleaned (3651)
5953 - print grants to console if they go negative (4431)
5954 - "lctl deactivate" will stop automatic recovery attempts (3406)
5955 - look for existing locks in ldlm_handle_enqueue() (3764)
5956 - don't resolve lock handle twice in recovery avoiding race (4401)
5957 - revalidate should check working dir is a directory (4134)
5959 - don't always mark "slow" obdfilter messages as errors (4418)
5961 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5964 - don't close LustreDB during write_conf until it is done (3860)
5965 - fix typo in lconf for_each_profile (3821)
5966 - allow dumping logs from multiple threads at one time (3820)
5967 - don't allow multiple threads in OSC recovery (3812)
5968 - fix debug_size parameters (3864)
5969 - fix mds_postrecov to initialize import for llog ctxt (3121)
5970 - replace config semaphore with spinlock (3306)
5971 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5972 - don't allow enqueue to complete on a destroyed export (3822)
5973 - down write_lock before checking llog header bitmap (3825)
5974 - recover from lock replay timeout (3764)
5975 - up llog sem before sending rpc (3652)
5976 - reduce ns lock hold times when setting kms (3267)
5977 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5978 - fix NULL deref and obd_dev leak on setup error (3312)
5979 - replace some LBUG about llog ops with error handling (3841)
5980 - don't match INVALID dentries from d_lookup and spin (3784)
5981 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5982 - fix invalid assertion in ptlrpc_set_wait (3880)
5984 - add libwrap support for the TCP acceptor (3996)
5985 - add /proc/sys/portals/routes for non-root route listing (3994)
5986 - allow setting MDS UUID in .xml (2580)
5987 - print the stack of a process that LBUGs (4228)
5989 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5992 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5993 - make sure to unset replay flag from failed open requests (3440)
5994 - if default stripe count is 0, use OST count for inode size (3636)
5995 - update parent mtime/ctime on client for create/unlink (2611)
5996 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5997 - free recovery state on server during a forced cleanup (3571)
5998 - unregister_reply for resent reqs (3063)
5999 - loop back devices mounting and status check on 2.6 (3563)
6000 - fix resource-creation race that can provoke i_size == 0 (3513)
6001 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6002 - more debugging for page-accounting assertion (3746)
6003 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6004 - avoid deadlock after precreation failure (3758)
6005 - fix race and lock order deadlock in orphan handling (3450, 3750)
6006 - add validity checks when grabbing inodes from l_ast_data (3599)
6008 - add /proc/.../recovery_status to obdfilter (3428)
6009 - lightweight CDEBUG infrastructure, debug daemon (3668)
6010 - change default OSC RPC parameters to be better on small clusters
6011 - turn off OST read cache for files smaller than 32MB
6012 - install man pages and include them in rpms (3100)
6013 - add new init script for (un)mounting lustre filesystems (2593)
6014 - run chkconfig in %post for init scripts (3701)
6015 - drop scimac NAL (unmaintained)
6017 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6020 - clean kiobufs before and after use (3485)
6021 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6022 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6023 - change init script to fail more clearly if not run as root (1528)
6024 - allow clients to reconnect during replay (1742)
6025 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6026 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6027 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6028 - fix cond_resched (3554)
6029 - only evict unfinished clients after recovery (3515)
6030 - allow bulk resend, prevent data loss (3570)
6031 - dynamic ptlrpc request buffer allocation (2102)
6032 - don't allow unlinking open directory if it isn't empty (2904)
6033 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6034 - remove extraneous obd dereference causing LASSERT failure (3334)
6035 - don't use get_cycles() when creating temp. files on the mds (3156)
6036 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6037 - handle EEXIST for set-stripe, set proper directory name (3336)
6039 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6040 - fix ksocknal_fmb_callback() error messages (2918)
6042 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6045 - don't copy lvb into (possibly NULL) reply on error (2983)
6046 - don't deref dentry after dput, don't free lvb on error (2922)
6047 - use the kms to determine writeback rpc length (2947)
6048 - increment oti_logcookies when osc is inactive (2948)
6049 - update client's i_blocks count via lvb messages (2543)
6050 - handle intent open/close of special files properly (1557)
6051 - mount MDS with errors=remount-ro, like obdfilter (2009)
6052 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6053 - don't use cancelling-locks' kms values (2947)
6054 - use highest lock extent for kms, not last one (2925)
6055 - don't dereference ERR_PTR() dentry in error handling path (3107)
6056 - fix thread race in portals_debug_dumplog() (3122)
6057 - create lprocfs device entries at setup instead of at attach (1519)
6058 - common AST error handler, don't evict client on completion race (3145)
6059 - zero nameidata in detach_mnt in 2.6 (3118)
6060 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6061 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6062 - initialize RPC timeout timer earlier for 2.6 (3219)
6063 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6064 - print nal/nid for unknown nid (3258)
6065 - additional checks for oscc recovery before doing precreate (3284)
6066 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6067 - don't crash in mdc_close for bad permissions on open (3285)
6068 - zero i_rdev for non-device files (3147)
6069 - clear page->private before handing to FS, better assertion (3119)
6070 - tune the read pipeline (3236)
6071 - fix incorrect decref of invalidated dentry (2350)
6072 - provide read-ahead stats and refine rpc in flight stats (3328)
6073 - don't hold journal transaction open across create RPC (3313)
6074 - update atime on MDS at close time (3265)
6075 - close LDAP connection when recovering to avoid server load (3315)
6076 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6077 - don't leak open file on MDS after open resend (3325)
6078 - serialize filter_precreate and filter_destroy_precreated (3329)
6079 - loop device shouldn't call sync_dev() for nul device (3092)
6080 - clear page cache after eviction (2766)
6081 - resynchronize MDS->OST in background (2824)
6082 - refuse to mount the same filesystem twice on same mountpoint (3394)
6083 - allow llmount to create routes for mounting behind routers (3320)
6084 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6085 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6086 - fix rare problem with rename on htree directories (3417)
6088 - allow default OST striping configuration per directory (1414)
6089 - fix compilation for qswnal for 2.6 kernels (3125)
6090 - increase maximum number of MDS request buffers for large systems
6091 - change liblustreapi to be useful for external progs like lfsck (3098)
6092 - increase local configuration timeout for slow disks (3353)
6093 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6095 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6098 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6099 - don't overwrite extent policy data in reply if lock was blocked (2901)
6100 - drop filter export grants atomically with removal from device (2663)
6101 - del obd_self_export from work_list in class_disconnect_exports (2908)
6102 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6103 - swab reply message in mdc_close, other PPC fixes (2464)
6104 - fix destroying of named logs (2325)
6105 - overwrite old logs when running lconf --write_conf (2264)
6106 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6107 - fix race in target_handle_connect (2898)
6108 - mds_reint_create() should take same inode create lock (2926)
6109 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6110 - don't close files for self_export to avoid uninitialized obd (2936)
6111 - allow MDS with the same name as client node (2939)
6112 - hold dentry reference for closed log files for unlink (2325)
6113 - reserve space for all logs during transactions (2059)
6114 - don't evict page beyond end of stripe extent (2925)
6115 - don't oops on a deleted current working directory (2399)
6116 - handle hard links to targets without a parent properly (2517)
6117 - don't dereference NULL lock when racing during eviction (2867)
6118 - don't grow lock extents when lots of conflicting locks (2919)
6120 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6123 - account for cache space usage on clients to avoid data loss (974)
6124 - lfsck support in lustre kernel code (2349)
6125 - reduce journal credits needed for BRW writes (2370)
6126 - orphan handling to avoid losing space on client/server crashes
6127 - ptlrpcd can be blocked, stopping ALL progress (2477)
6128 - use lock value blocks to assist in proper KMS, faster stat (1021)
6129 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6130 - recovery for initial connections (2355)
6131 - fixes for mds_cleanup_orphans (1934)
6132 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6133 - block all file creations until orphan recovery completes (1901)
6134 - client remove rq_connection from request struct (2423)
6135 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6136 - recovery timer race (2670)
6137 - mdc_close recovey bug (2532)
6138 - ptlrpc cleanup bug (2710)
6139 - mds timeout on local locks (2588)
6140 - namespace lock held during RPCs (2431)
6141 - handle interrupted sync write properly (2503)
6142 - don't try to handle a message that hasn't been replied to (2699)
6143 - client assert failure during cleanup after abort recovery (2701)
6144 - leak mdc device after failed mount (2712)
6145 - ptlrpc_check_set allows timedout requests to complete (2714)
6146 - wait for inflight reqs when ptlrpcd finishes (2710)
6147 - make sure unregistered services are removed from the srv_list
6148 - reset bulk XID's when resending them (caught by 1138 test)
6149 - unregister_bulk after timeout
6150 - fix lconf error (2694)
6151 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6152 - readahead locks pages, leaves pending causing memory pressure (2673)
6153 - increase OST request buffers to 4096 on large machines (2729)
6154 - fix up permission of existing directories in simple_mkdir (2661)
6155 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6156 - don't assign transno to errored transactions (2742)
6157 - don't delete objects on OST if given a bogus objid from MDS (2751)
6158 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6159 - if rq_no_resend, then timeout request after recovery (2432)
6160 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6161 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6162 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6163 - MDS recovery shouldn't delete objects that it has given out (2730)
6164 - if enqueue arrives after completion, don't clobber LVB (2819)
6165 - don't unlock pages twice when trigger_group_io returns error (2814)
6166 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6167 - don't write pages to disk if there was an error (1450)
6168 - don't ping imports that have recovery disabled (2676)
6169 - take buffered bytes into account when balancing socknal conn (2817)
6170 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6171 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6172 - remove little-endian swabbing of llog records (1987)
6173 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6174 - retry reposting request buffers if they fail (1191)
6175 - grow extent at grant time to avoid granting a revoked lock (2809)
6176 - lock revoke doesn't evict page if covered by a second lock (2765)
6177 - disable VM readahead to avoid reading outside lock extents (2805)
6179 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6180 - updated kernel patches for hp-2.4.20 kernel (2681)
6182 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6185 - fix truncated write corruption (2366)
6186 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6188 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6189 - ensure len(uuid) < 37 in lmc (1171)
6190 - fix ia64 OOPS in llog_test (2255)
6191 - zero end of page at obdfilter for partial page writes (2648)
6192 - don't leave stale dentries around after renames (bug 2428)
6193 - fix timeouts when evicting a client with a single lock held (2642)
6194 - set deadline for the initial HELLO message to drain (2634)
6195 - print out dotted-quad IP addresses in the socknal (2302)
6197 - additional debugging for MDS client eviction problem (2443)
6198 - fix mkfsoptions support for osts (2603, 2604)
6200 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6203 - add series for the vanilla 2.6.0 kernel
6204 - add series for the vanilla 2.4.24 kernel
6205 - add series for a cray x86/64 UL kernel drop
6206 - fix xattr patches for the vanilla 2.4.19 series
6208 - generate true UUIDs in lmc (1171)
6209 - have portals stack dumping break in UML (2466)
6210 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6211 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6212 - wake callback waiting threads on client eviction (2460)
6213 - Add --ptldebug and --subsystem to lmc (1719)
6214 - update assertion to allow safe interrupt allocation
6215 - set rq_no_resend for cancel requests (2432)
6216 - recalculate ptlrpcd timeout after resend (2494)
6217 - call vfs_rmdir when removing pending directories (2368)
6218 - fix renaming a file to itself (2429)
6219 - lmc creates a default one-stripe lov (2454)
6220 - expand procfs space to handle large clusters (2326)
6221 - increase UML stack to avoid overflow
6222 - update lconf's list of debug and subsystem masks
6223 - fix lfs find --obd (2510)
6224 - /proc tunable for disabling filter read caching (2591)
6225 - stop rpm packages from altering slapd.conf (2301)
6226 - disable nagle in the socknal under 0conf (2578)
6227 - choose mds inode size based on stripe count (2572)
6228 - fix kernel-source rpm problems (2516)
6230 - add --disable-doc to avoid pdf generation (2421)
6231 - update documentation, tests, type-os, comments
6232 - avoid format warnings on ia64
6233 - remove the TOE NAL
6234 - tiny code cleanups by removing unused fields
6236 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6239 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6240 - remove the most blatant lies from BUILDING file (2371)
6241 - change default debug level to reasonable production setting
6242 - reduce client side cache size to reduce cache flush time
6243 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6244 - make TCP zerocopy and pinger support enabled by default (2476)
6245 - sync writes completed after process exits caused crashes (2319)
6246 - maintain correct mount count on the MDS (2356)
6247 - backout 1557, because 2316 wasn't really fixed
6248 - better file I/O statistics gathering in /proc
6249 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6250 - another kernel patch to fix zero-copy TCP function export
6251 - don't take duplicate lock when processing re-sent getattr (2420)
6252 - lctl uses obd_self_export instead of creating new conn (2353)
6253 - MDS/OST recovery case which requires object creation asserted (2425)
6254 - move lfs from /usr/sbin to /usr/bin in packages
6255 - fix race between mds_client_add and mds_client_free (2417)
6256 - use kmalloc instead of slabs in portals (2430)
6257 - don't create duplicate records when a failover MDS is present (2442)
6258 - remove unnecessary mount age check (2332)
6259 - don't remove directory inodes from locks prematurely (2451)
6260 - don't break if MDS service name is the same as hostname (2103)
6261 - fix races in client write RPC generation when cache full (2482)
6263 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6266 - remove now-unused request->rq_obd (278)
6267 - if an allocation fails, print out how much memory we've used (1933)
6268 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6269 - add the "configurable stack size" patch to most series files (1256)
6270 - ability to write large log records, for 100+ OST configs (2306)
6271 - fix NULL deref when filter_prep fails (2314)
6272 - fix operator precedence error in filter_sync
6273 - dynamic allocation of socknal TX descriptors (2315)
6274 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6275 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6276 - ensure that CWARN messages in Portals always get to the syslog
6277 - __init/__exit are not for prototype decls (ldlm_init/exit)
6278 - x86-64 compile warning fixes
6279 - fix gateway LMC keyword conflict (2318)
6280 - fix MDS lock inversions in getattr/reint paths (1844)
6281 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6282 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6283 - DLM race condition prevented some lock evictions (2328)
6284 - ENOMEM detection and retry on socknal sends (2230)
6285 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6286 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6287 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6288 - better error messages when a client is rejected during recovery (1505)
6289 - avoid cancelling locks which were never granted, after failure (2330)
6290 - fix i_sem/journal inversion in mds_client_add (2333)
6291 - fix truncate/getattr lock cycle deadlock (2334)
6292 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6293 - fix two rare exit paths which could leak an l_lock() ref (2321)
6294 - fencepost error in MDS/OST orphan recovery (2226)
6295 - make log record alignment 8 bytes (1988)
6296 - lstripe now fails when requested offset > ost_count (2237)
6297 - ensure that all kernel series have a complete list.h (1607)
6298 - fix crashes in special-file operations (2316)
6299 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6301 - allow configurable automake binary, for testing new versions
6302 - small update to the lfs documentation
6304 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6306 * fix negative export reference count in fsfilt_sync (2312)
6308 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6309 * release candidate 0.9.1
6311 - orphans are moved into the PENDING directory for possible recovery
6312 - replayed opens now open by fid for orphan/rename safety (1042)
6313 - last close of an orphan inode generates a transno (683)
6314 - chdir() and mount() now pin the directory entry (1020)
6315 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6316 - discard very old requests without processing them (1502)
6317 - remove some common, well-understood CERRORs (1505)
6318 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6319 - clear "grant" flags in OST replies until OST grant code lands (1644)
6320 - fix read performance by not clobbering i_blksize on client (1598)
6321 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6322 - make LRU size a /proc tunable, clears locks when reduced (707)
6323 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6324 - server locks take references on exports now (1558)
6325 - build fixes for 2.4.20-rh trees (1663)
6326 - return an error from lov_create if all OSCs are inactive (1751)
6327 - fix import levels when a reconnect happens without a timeout (1597)
6328 - exit early from mds_open if we get a lookup error (1749)
6329 - partial page read at EOF wouldn't wait for disk before sending (1642)
6330 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6331 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6332 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6333 - write/truncate lock inversion (1639)
6334 - Don't auto-load obdclass, portals modules during cleanup (1495)
6335 - fix timestamps from jumping to "now" (1763)
6336 - extra journal assertions (1648)
6337 - add an extra multiunlink test (1771)
6338 - fix read_record/write_record API (1776)
6339 - fix leak of offset_extent, possible incorrect i_size later (1772)
6340 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6341 - Debugging for the kqswnal_get_idle_tx problems (1820)
6342 - Allow recovery to be attempted multiple times (1536)
6343 - Write out MDS last_rcvd file after it is first created (1600)
6344 - Fix tx_descriptor leak in failed transmit situations (1827)
6345 - ext3 journaling fixes for assertion failure after IO error (1871)
6346 - class_export_put() on freed export after completion AST error (1896)
6347 - Fix revalidate looping in VFS (1322)
6348 - Don't access a freed export during MDS_REINT timeout (1521)
6349 - Add open-unlink recovery support on the MDS (1673,1764)
6350 - Return an error if no MDS data was read from last_rcvd (1946)
6351 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6352 - Don't setup a disk device that doesn't match exported UUID (317)
6353 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6354 - avoid committing NULL handle in force close
6355 - local.sh is now a one-stripe LOV configuration
6356 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6357 - fix ext3 htree duplicate directory entry corruption (1516)
6358 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6359 - update to new LOV EA format (2097)
6360 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6362 2003-06-15 Phil Schwan <phil@clusterfs.com>
6365 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6366 - per-import/export recovery handling (958, 931, 959)
6367 - multiple last-rcvd slots, for serving multiple FSes (949)
6368 - connections are again shared between multiple imp/exports (963, 964)
6369 - "umount -f" would hang if any requests needed to be sent (393, 978)
6370 - avoid pinning large req buffer by copying for queued messages (989)
6371 - add "uuid" to "lctl device" command to help upcalls (991)
6372 - "open" RPCs with transnos would confuse recovery counters (1037)
6373 - do proper endian conversion of all wire messages (288, 340, 891)
6374 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6375 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6376 - fix ptlrpc_connection leak in target_handle_connect (1174)
6377 - fix import refcounting bug in OST and MDS cleanup (1134)
6378 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6379 - fix very unlikely obd_types race condition (501)
6380 - remove osc_open hack for echo_client (1187)
6381 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6382 - a failure in read_inode2 leads to deadlock (1139)
6383 - cancel ack-locks as soon as transaction is committed (1072)
6384 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6385 - make sure to commitrw after any preprw to avoid deadlock (1162)
6386 - failing to execute a file in a lustre FS would lock inode (1203)
6387 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6388 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6389 - obd_uuid2tgt would walk off the end of the list (1255)
6390 - on IA64 the osc would give portals incorrect bulk size (1258)
6391 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6392 - fix lock inversion caused by new llite matching code (1282)
6393 - limit the number of dirty pages on a client to 10MB (1286)
6394 - timed out locks were not being corrected cancelled (1289)
6395 - fix O_DIRECT above 4GB on IA-32 (1292)
6396 * major user-visible changes
6397 - fail out/fail over policy now controlled by the upcall (993)
6399 - add OBD_PING to check server availability and failure (954)
6400 - lustre messages are now sent in sending host order (288, 340, 891)
6401 - add eadatalen to MDS getattr reply (340)
6402 - OST read replies may contain second buffer, with per-page status (593)
6404 2003-03-11 Phil Schwan <phil@clusterfs.com>
6407 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6408 - failed open()s could cause deadlock; fixed (867, 869)
6409 - stop cancelling OST locks when files are closed (481)
6410 - overlapping XID spaces caused network corruption (851, 853)
6411 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6412 - setattr_raw added, to do single-RPC, server-side setattrs
6413 - lmc/lconf syntax change for OST UUIDs
6414 - fix crashy race condition between ptlrpc_free_req and osc_close
6415 - don't use request in mdc_enqueue if we hit a timeout (889)
6416 - don't set the inode i_size for regular files from the MDS (896)
6417 - handle out of order completion AST (842)
6418 - don't LBUG if a lock request times out after receiving AST (913)
6419 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6420 - if a bad lock AST arrives, send an error instead of dropping entirely
6421 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6422 - fix leak in bulk IO when only partially completed (899, 900, 926)
6423 - fix O_DIRECT for ia64 (55)
6424 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6425 - C-z after timeout could hang a process forever; fixed (977)
6427 - client-side I/O cache (678, 924, 929, 941, 970)
6429 - READPAGE and SETATTRs which don't take server-side locks get
6432 2003-02-11 Phil Schwan <phil@clusterfs.com>
6435 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6436 - Fix fsfilt_extN_readpage() to read a full page of directory
6437 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6438 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6439 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6440 - Most intent operations were removed, in favour of a new RPC mode
6441 that does a single RPC to the server and bypasses most of the VFS
6442 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6443 - Aggressively cancel local locks on DLM servers
6444 - mds_reint_unlink sends EA to the client if it's the last nlink.
6445 client uses that EA to unlink OST objects.
6446 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6447 - recursive symlinks were fixed (439)
6448 - fixed NULL deref in DEBUG_REQ
6449 - filter_update_lastobjid no longer calls sync, which annoyed extN
6450 - fixed multi-client small-writes to a single file problem (445)
6451 - fixed mtime updates during file writes (607)
6452 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6453 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6454 - fixed Linux OST size reporting problem (444, 656)
6455 - OST now updates object mtime with writes or setattr (607, 619)
6456 - client verifies file size before zeroing page past EOF (445)
6457 - OST now writes last allocated objid to disk with allocation (108)
6458 - LOV on echo now works (409)
6460 - mds_reint_unlink sends a new buffer, with the EA included. this
6461 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6462 set if a regular file was being unlinked, and it was the last link
6463 - use PtlGet from the target for bulk writes (315)
6464 - OST now updates object mtime with writes or setattr (607, 619)
6465 - LDLM now has a grant-time callback to revalidate locked items, if
6467 - Many MDS operations were reorganized to combat race conditions
6469 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6470 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6472 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6475 - Fully reactivate OST imports after reconnection (512, others)
6476 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6477 - More graceful error handling for truncating on dead OST (515)
6478 - Don't error out unless we're actually accessing dead stripes (474)
6479 - Fix garbage sizes when stripes are missing (410)
6480 - LRU counters were broken, causing constant lock purge (433, 432)
6481 - garbage on read from stripes with failed OSTs (441)
6482 - mark OSCs as active before reconnecting during recovery (438)
6483 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6484 - lfind did not preserve OST order in output (443)
6485 - symlinks cause hung clients, incorrect data (439)
6486 - stop dereferencing request after dropping refcount (457)
6487 - don't LASSERT(spin_is_locked) on non-SMP (455)
6488 - fixes for many rename() bugs
6489 - fstat didn't correctly synchronize attributes (399)
6490 - server must handle lock cancellation during blocking AST prep (487)
6491 - bulk descriptors were free()d too soon (511)
6492 - fix paths in lconf, which would load incorrect modules (451, 507)
6493 - fix confusing lconf 'host not found' error message (386)
6494 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6495 - fix race condition in mdc_blocking_ast() for inode access (526)
6496 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6497 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6498 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6499 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6500 - fix lstripe in several ways (kernel side) (527)
6501 - fix request leak in ldlm_cli_enqueue (262)
6502 - incorrect OSC was marked inactive after OST failure
6503 - call mds_fs_cleanup before unmounting filesystem (524)
6504 - fix races between taking ns_lock and ldlm_lock_change_resource
6505 - fix races updating LOV export open file list
6506 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6507 - fix recovery NULL deref in ldlm_cli_cancel_unused
6508 - fix some DLM races by using new hash table for lock handles (419)
6509 - permit the client to specify desired inodes, at replay
6510 - duplicate requests when we queue them for replay reintegration
6511 - fix last_rcvd offset calculation
6512 - sync after each recovered transaction, so we always make progress
6513 - never, not always, ERESTART requests without transnos
6514 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6516 - skip replay if the MDS says that the client is already connected
6517 - don't check for a recovery-enabled export to match lctl's UUID
6518 - don't INC_USE_COUNT for phantom exports
6519 - don't crash when cleaning up phantom exports (567)
6520 - don't double-finish or set replay data for errored mdc_open requests
6521 - abort requests when they time out, so we don't get old replies
6522 - send/receive replies for AST messages again
6523 - if the client says that it doesn't have the lock, cancel it on the
6525 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6526 mark it as destroyed, it will all work out when decref is called
6527 - fix module use counts (22, 581)
6529 - ASTs now expect a reply (server cancels lock on error reply)
6531 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6534 - fix many simultaneous client startup (392)
6535 - fix dentry->d_it clobbering
6536 - credentials weren't being shipped for readdir/getattr operations
6537 - remove invalid assertions triggered during some concurrent MD
6539 - proper Lustre versions added (336, 389)
6540 - fix memory leak for create error case (398)
6541 - fix LOV locking bug that would get cli/srv out of sync
6542 - fix echo client over LOV (409)
6543 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6544 - fix double-O_EXCL intent crash (424)
6545 - avoid sending multiple lock CANCELs (352)
6547 - MDS can do multi-client recovery (modulo bugs in new code)
6549 - many updates, edits, cleanups
6551 2002-11-18 Phil Schwan <phil@clusterfs.com>
6554 - fix null d_it dereference (346)
6555 - fix full OST/dbench hang (333)
6556 - fix permission problem with file removal (286)
6557 - fix removal of OSCs from LOV when they fail
6558 - fix NULL deref during bulk timeout (214)
6559 - fix problems related to multiple filesystems on one MDS (241)
6560 - fixed serious subtle metadata locking bugs
6561 - free locks on clients when inodes are removed due to memory
6563 - fix inode pointer in lock data (285)
6564 - partial support for multiple MDS on a single host (241)
6565 - data locks weren't cancelled at clear_inode time (290, 311)
6566 - intent locks could lead to unbounded lock growth (205)
6567 - added a maximum lock count, an LRU list, and a flusher
6568 - fix multiple rename (365)
6569 - properly abstracted the echo client
6570 - OSC locked 1 byte too many; fixed
6571 - rewrote brw callback code:
6572 - fixed recovery bugs related to LOVs (306)
6573 - fixed too-many-pages-in-one-write crash (191)
6574 - fixed (again) crash in sync_io_timeout (214)
6575 - probably fixed callback-related race (385)
6577 - Add capability to MDS protocol
6578 - LDLM cancellations and callbacks on different portals
6580 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6583 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6584 - defer open object destruction to close time (601981, 138)
6585 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6586 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6587 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6588 - don't call obd_setattr in ll_file_release for destroyed objects
6589 * protocol change to lustre_msg: move |version| and add |flags|
6590 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6591 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6592 * added replay of create, unlink, link and rename operations during
6593 MDS failover; recovery should be much more robust now
6594 * remove failed OSCs from LOVs (only lov_create uses this so far)
6595 * the lustre-HOWTO was brought (more) up to date (582544)
6597 2002-10-23 Phil Schwan <phil@clusterfs.com>
6600 - in-use dentries weren't being reused properly (617851)
6601 - prevent multiple LDLM setup (599178)
6602 - fix LOV size calculations for truncate (617853)
6603 - fix client handling of MDS intent errors (POSIX)
6604 - fix permission bug in lovstripe.c test (624321)
6605 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6606 - truncate past end of file could corrupt data
6607 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6608 - a race in recovery could return ETIMEDOUT to apps (623947)
6609 - building outside the source directory was fixed
6610 * the lustre-HOWTO was brought (more) up to date (582544)
6611 * major progress was made on recovery functionality
6613 2002-10-10 Phil Schwan <phil@clusterfs.com>
6616 - recovery deadlock fix
6617 - rm -rf causes LBUG fix (617817)
6618 - file open by multiple tasks fix (618962)
6619 - directory permissions bugs (602707 and 620007)
6620 - journal_stop fixed with locking (611313)
6621 - O_APPEND failures resolved (618273, perhaps 614459)
6622 - lconf PATH fix (619770)
6623 - IA64 build fix (621450)
6624 - RPC buffer sizes scale with amount of memory
6626 2002-10-01 Phil Schwan <phil@clusterfs.com>
6629 - locks would be cancelled without throwing away data pages,
6630 resulting in inconsistent data (605627)
6631 - inode attributes were not always being refreshed (605627, 612449)
6632 - lconf now continues to cleanup after lctl reports an error
6633 - MDS now enforces user permissions (602707)
6634 - lprocfs cleanup fixed, but not yet enabled (614157)
6635 - fixed infinite server hang, should a client not respond to an AST
6636 - avoid going into recovery if user calls readlink() with a buffer
6637 that's too small (613941)
6638 - AST RPCs no longer require replies (614867) -- this may be changed
6639 - don't crash server if client sends an IOV that's too big (611336)
6640 - fixed lock conversion deadlock (611892)
6641 - fixed the following of symlinks (614622)
6642 * recovery: the server can remove locks from a client that dies, other
6643 clients can make progress
6644 * more extN patch fixes
6645 * compile-time configurable ptlrpc buffer allocations
6647 - collaborative read cache document
6648 - Lustre Lite Performance CDR document-in-progress
6650 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6653 - fix typo in patch-2.4.18
6655 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6658 - clear ptlrpc request each time in handle_incoming_request()
6659 - unlink of files now destroys the object on the OST
6661 2002-09-19 Peter Braam <braam@clusterfs.com>
6663 * add hard link support
6664 * change obdfile creation method
6665 * kernel patch changed
6667 2002-09-19 Peter Braam <braam@clusterfs.com>
6670 - stack overflow bug in extN fixed
6672 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6674 * documentation updates
6675 - add man pages for config tools
6676 - update tests/README to describe testing with new config tools
6677 - finish metadata API descriptions
6678 * bug fixes and cleanups
6679 - statfs workaround for 16TB limit
6680 - LOV stripe allocation improved, can stripe on subset of OSTs
6681 - LOV file size/IO offset was wrong for files > 4GB in size
6682 - object EA data was being dropped, caused files to be unreadable
6683 - memory overflow with non-LOV OST caused memory corruption
6684 - fixed regression tests to work with new config tools, obdfilter
6685 - fixed bug when directory size became larger than 1 block
6686 - fixed bug (for single client case) when PWD was deleted
6687 - invalidate local directory pages when doing intent-based ops
6688 - avoid LDLM oops when lock callback contained bad data
6690 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6692 * documentation updates
6693 * bug fixes and cleanups
6694 - configuration tools
6697 - 64-bit compile warnings
6698 - 64-bit internal statfs data
6700 * test_brw on persistent OST devices
6702 * lprocfs (disabled)
6704 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6706 * documentation updates
6707 * bug fixes and cleanups
6708 * configuration tools
6710 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6713 * many small fixes to 0_5_4
6714 * io/network handling
6715 * thinkos in MDS operations
6717 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6720 * crucial basic fixes to 0.5.3
6721 * IOR, Iozone work over Elan
6724 2002-08-07 Phil Schwan <phil@clusterfs.com>
6725 * version 0_5_3, our first alpha
6726 * we use the new Portals iovs
6727 * documentation updates
6728 * bug fixes and cleanups
6729 * small changes in the DLM wire protocol
6731 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6732 * version 0_5_1 with some initial stability,
6733 * locking on MD and file I/O.
6734 * documentation updates
6735 * several bug fixes since 0.5.0
6736 * small changes in wire protocol
6738 2002-07-18 Phil Schwan <phil@clusterfs.com>
6740 * delivered as Lustre Light Alpha
6741 * fixed a crash after handling invalid MDS requests
6742 * fixed directory pages for architectures with non-4k pages sizes
6744 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6745 * release version v0_4_4
6746 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6748 * Fixes a number of interruption problems with OST operations.
6749 * Update documentation for portals header changes
6750 * Move all wire protocol structs/defines to lustre_idl.h
6751 * Fixes symlink length bug.
6752 * Add tcpdump to repository.
6754 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6755 * release version v0_4_3
6756 * Fixes statfs for inodes on extN.
6757 * Fixes bug in runtests which would delete /etc/hosts.
6758 * Use 64-bit object IDs wherever possible (not into VFS though)
6759 Remove ost_get_info, which is unused by lustre, and out of date.
6761 2002-07-03 Peter Braam <braam@clusterfs.com>
6762 * release version v0_4_2 Fixes a lookup error (type not passed)
6763 * move forward to head of Portals
6764 * move forward to latest Lustre kernel
6766 2002-06-25 Peter Braam <braam@clusterfs.com>
6767 * release version v0_4_1. Hopefully stable on single node use.