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 3.10.0-1160.25.1.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 4.18.0-240.22.1.el8 (RHEL8.3)
16 4.18.0-305.3.1.el8 (RHEL8.4)
17 4.4.120-92.70 (SLES12 SP2)
18 4.4.180-94.100 (SLES12 SP3)
19 4.4.0-131 (Ubuntu 16.04)
20 4.15.0-32 (Ubuntu 18.04)
21 5.4.0-48 (Ubuntu 20.04)
22 vanilla linux 5.4.0 (ZFS + ldiskfs)
23 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
24 * Client primary kernels built and tested during release cycle:
25 3.10.0-1062.18.1.el7 (RHEL7.7)
26 3.10.0-1127.19.1.el7 (RHEL7.8)
27 3.10.0-1160.25.1.el7 (RHEL7.9)
28 4.12.14-95.48 (SLES12 SP4)
29 4.12.14-122.66 (SLES12 SP5)
30 4.12.14-197.75 (SLES15 SP1)
31 5.3.18-24.61 (SLES15 SP2)
32 4.15.0-48 (Ubuntu 18.04)
33 * Other clients known to build on these kernels at some point (others may also work):
34 3.10.0-862.14.4.el7 (RHEL7.5)
35 3.10.0-957.27.2.el7 (RHEL7.6)
36 4.14.0-49.13.1.el7a (RHEL7.5)
37 4.14.0-115.2.2.el7a (RHEL7.6)
38 4.18.0-80.11.2.el8 (RHEL8.0)
39 4.18.0-147.8.1.el8 (RHEL8.1)
40 4.18.0-193.28.1.el8 (RHEL8.2)
41 4.18.0-240.22.1.el8 (RHEL8.3)
42 4.18.0-305.3.1.el8 (RHEL8.4)
43 4.4.120-92.70 (SLES12 SP2)
44 4.4.180-94.100 (SLES12 SP3)
45 4.4.0-131 (Ubuntu 16.04)
46 5.4.0-37 (Ubuntu 20.04)
47 5.4.0 (vanilla kernel.org)
48 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
49 * Recommended ZFS version: 2.0.0
50 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
51 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
52 stack overflow. For more information, please refer to bugzilla 17630.
53 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
54 the re-exporting nodes to be mounted with "32bitapi" mount option
56 --------------------------------------------------------------------------------
60 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
61 for currently supported client and server kernel versions.
62 * Server primary kernels built and tested during release cycle:
63 3.10.0-1062.18.1.el7 (RHEL7.7)
64 3.10.0-1127.19.1.el7 (RHEL7.8)
65 3.10.0-1160.6.1.el7 (RHEL7.9)
66 * Other server kernels known to build and work at some point (others may also work):
67 3.10.0-862.14.4.el7 (RHEL7.5)
68 3.10.0-957.27.2.el7 (RHEL7.6)
69 4.18.0-80.11.2.el8 (RHEL8.0)
70 4.18.0-147.8.1.el8 (RHEL8.1)
71 4.18.0-193.28.1.el8 (RHEL8.2)
72 4.18.0-240.1.1.el8 (RHEL8.3)
73 4.4.120-92.70 (SLES12 SP2)
74 4.4.180-94.100 (SLES12 SP3)
75 4.4.0-131 (Ubuntu 16.04)
76 4.15.0-32 (Ubuntu 18.04)
77 5.4.0-48 (Ubuntu 20.04)
78 vanilla linux 5.4.0 (ZFS + ldiskfs)
79 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
80 * Client primary kernels built and tested during release cycle:
81 3.10.0-1062.18.1.el7 (RHEL7.7)
82 3.10.0-1127.19.1.el7 (RHEL7.8)
83 3.10.0-1160.6.1.el7 (RHEL7.9)
84 4.12.14-95.48 (SLES12 SP4)
85 4.12.14-122.20 (SLES12 SP5)
86 4.12.14-197.75 (SLES15 SP1)
87 5.3.18-24.24 (SLES15 SP2)
88 4.15.0-48 (Ubuntu 18.04)
89 * Other clients known to build on these kernels at some point (others may also work):
90 3.10.0-862.14.4.el7 (RHEL7.5)
91 3.10.0-957.27.2.el7 (RHEL7.6)
92 4.14.0-49.13.1.el7a (RHEL7.5)
93 4.14.0-115.2.2.el7a (RHEL7.6)
94 4.18.0-80.11.2.el8 (RHEL8.0)
95 4.18.0-147.8.1.el8 (RHEL8.1)
96 4.18.0-193.28.1.el8 (RHEL8.2)
97 4.18.0-240.1.1.el8 (RHEL8.3)
98 4.4.120-92.70 (SLES12 SP2)
99 4.4.180-94.100 (SLES12 SP3)
100 4.4.0-131 (Ubuntu 16.04)
101 5.4.0-37 (Ubuntu 20.04)
102 5.4.0 (vanilla kernel.org)
103 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
104 * Recommended ZFS version: 2.0.0
105 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
106 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
107 stack overflow. For more information, please refer to bugzilla 17630.
108 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
109 the re-exporting nodes to be mounted with "32bitapi" mount option
111 --------------------------------------------------------------------------------
115 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
116 for currently supported client and server kernel versions.
117 * Server primary kernels built and tested during release cycle:
118 3.10.0-1062.1.1.el7 (RHEL7.7)
119 * Other server kernels known to build and work at some point (others may also work):
120 3.10.0-862.14.4.el7 (RHEL7.5)
121 3.10.0-957.27.2.el7 (RHEL7.6)
122 4.18.0-80.7.1.el8 (RHEL8.0)
123 4.4.120-92.70 (SLES12 SP2)
124 4.4.180-94.100 (SLES12 SP3)
125 4.4.0-131 (Ubuntu 16.04)
126 4.15.0-32 (Ubuntu 18.04)
127 vanilla linux 4.6.7 (ZFS only)
128 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
129 * Client primary unpatched kernels built and tested during release cycle:
130 3.10.0-1062.1.1.el7 (RHEL7.7)
131 4.12.14-95.29 (SLES12 SP4)
132 4.15.0-48 (Ubuntu 18.04)
133 * Other clients known to build on unpatched kernels at some point (others may also work):
134 3.10.0-862.14.4.el7 (RHEL7.5)
135 3.10.0-957.27.2.el7 (RHEL7.6)
136 4.14.0-49.13.1.el7a (RHEL7.5)
137 4.14.0-115.2.2.el7a (RHEL7.6)
138 4.18.0-80.7.1.el8 (RHEL8.0)
139 4.4.120-92.70 (SLES12 SP2)
140 4.4.180-94.100 (SLES12 SP3)
141 4.4.0-131 (Ubuntu 16.04)
142 vanilla linux 4.15.0 (ZFS only)
143 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
144 * Recommended ZFS / SPL version: 0.7.13
145 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
146 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
147 stack overflow. For more information, please refer to bugzilla 17630.
148 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
149 the re-exporting nodes to be mounted with "32bitapi" mount option
150 * Note for Ubuntu 14.04.5:
151 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
152 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
154 --------------------------------------------------------------------------------
158 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
159 for currently supported client and server kernel versions.
160 * Server known to build on the following kernels (others may also work):
161 2.6.32-696.18.7.el6 (RHEL6.9)
162 3.10.0-862.14.4.el7 (RHEL7.5)
163 3.10.0-957.el7 (RHEL7.6)
164 4.4.120-92.70 (SLES12 SP2)
165 4.4.162-94.69 (SLES12 SP3)
166 4.4.0-131 (Ubuntu 16.04)
167 4.15.0-32 (Ubuntu 18.04)
168 vanilla linux 4.6.7 (ZFS only)
169 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
170 * Client known to build on unpatched kernels (others may also work):
171 2.6.32-696.18.7.el6 (RHEL6.9)
172 2.6.32-754.6.3.el6 (RHEL6.10)
173 3.10.0-862.14.4.el7 (RHEL7.5)
174 3.10.0-957.el7 (RHEL7.6)
175 4.14.0-49.13.1.el7a (RHEL7.5)
176 4.14.0-115.2.2.el7a (RHEL7.6)
177 4.4.120-92.70 (SLES12 SP2)
178 4.4.162-94.69 (SLES12 SP3)
179 4.4.0-131 (Ubuntu 16.04)
180 4.15.0-32 (Ubuntu 18.04)
181 vanilla linux 4.15.0 (ZFS only)
182 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
183 * Recommended ZFS / SPL version: 0.7.9
184 * Tested with ZFS / SPL version: 0.7.9
185 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
186 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
187 stack overflow. For more information, please refer to bugzilla 17630.
188 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
189 the re-exporting nodes to be mounted with "32bitapi" mount option
190 * Note for Ubuntu 14.04.5:
191 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
192 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
194 --------------------------------------------------------------------------------
196 2018-04-03 Intel Corporation
198 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
199 for currently supported client and server kernel versions.
200 * Server known to build on patched kernels:
201 2.6.32-431.29.2.el6 (RHEL6.5)
202 2.6.32-504.30.3.el6 (RHEL6.6)
203 2.6.32-573.26.1.el6 (RHEL6.7)
204 2.6.32-642.15.1.el6 (RHEL6.8)
205 2.6.32-696.18.7.el6 (RHEL6.9)
206 3.10.0-693.21.1.el7 (RHEL7.4)
207 3.0.101-0.47.71 (SLES11 SP3)
208 3.0.101-107 (SLES11 SP4)
209 3.12.74-60.64.40 (SLES12 SP1)
210 4.4.114-92.64 (SLES12 SP2)
211 4.4.114-94.11 (SLES12 SP3)
212 3.13.0-101 (Ubuntu 14.04, ZFS only)
213 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
214 4.4.0-116 (Ubuntu 16.04)
215 vanilla linux 4.6.7 (ZFS only)
216 * Client known to build on unpatched kernels:
217 2.6.32-431.29.2.el6 (RHEL6.5)
218 2.6.32-504.30.3.el6 (RHEL6.6)
219 2.6.32-573.26.1.el6 (RHEL6.7)
220 2.6.32-642.15.1.el6 (RHEL6.8)
221 2.6.32-696.18.7.el6 (RHEL6.9)
222 3.10.0-693.21.1.el7 (RHEL7.4)
223 3.0.101-0.47.71 (SLES11 SP3)
224 3.0.101-107 (SLES11 SP4)
225 3.12.74-60.64.40 (SLES12 SP1)
226 4.4.114-92.64 (SLES12 SP2)
227 4.4.114-94.11 (SLES12 SP3)
228 3.13.0-101 (Ubuntu 14.04)
229 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
230 4.4.0-116 (Ubuntu 16.04)
232 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
233 * Recommended ZFS / SPL version: 0.7.6
234 * Tested with ZFS / SPL version: 0.7.6
235 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
236 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
237 stack overflow. For more information, please refer to bugzilla 17630.
238 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
239 the re-exporting nodes to be mounted with "32bitapi" mount option
240 * Note for Ubuntu 14.04.5:
241 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
242 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
244 --------------------------------------------------------------------------------
246 2017-07-05 Intel Corporation
248 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
249 for currently supported client and server kernel versions.
250 * Server known to build on patched kernels:
251 2.6.32-431.29.2.el6 (RHEL6.5)
252 2.6.32-504.30.3.el6 (RHEL6.6)
253 2.6.32-573.26.1.el6 (RHEL6.7)
254 2.6.32-642.15.1.el6 (RHEL6.8)
255 2.6.32-696.3.1.el6 (RHEL6.9)
256 3.10.0-514.21.1.el7 (RHEL7.3)
257 3.0.101-0.47.71 (SLES11 SP3)
258 3.0.101-100 (SLES11 SP4)
259 3.12.74-60.64.40 (SLES12 SP1)
260 4.4.59-92.17 (SLES12 SP2)
261 vanilla linux 4.6.7 (ZFS only)
262 * Client known to build on unpatched kernels:
263 2.6.32-431.29.2.el6 (RHEL6.5)
264 2.6.32-504.30.3.el6 (RHEL6.6)
265 2.6.32-573.26.1.el6 (RHEL6.7)
266 2.6.32-642.15.1.el6 (RHEL6.8)
267 2.6.32-696.3.1.el6 (RHEL6.9)
268 3.10.0-514.21.1.el7 (RHEL7.3)
269 3.0.101-0.47.71 (SLES11 SP3)
270 3.0.101-100 (SLES11 SP4)
271 3.12.74-60.64.40 (SLES12 SP1)
272 4.4.59-92.17 (SLES12 SP2)
274 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
275 * Recommended ZFS / SPL version: 0.6.5.9
276 * Tested with ZFS / SPL version: 0.6.5.9
277 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
278 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
279 stack overflow. For more information, please refer to bugzilla 17630.
280 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
281 the re-exporting nodes to be mounted with "32bitapi" mount option
283 --------------------------------------------------------------------------------
285 12-07-2016 Intel Corporation
287 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
288 for currently supported client and server kernel versions.
289 * Server known to build on patched kernels:
290 2.6.32-431.29.2.el6 (RHEL6.5)
291 2.6.32-504.30.3.el6 (RHEL6.6)
292 2.6.32-573.26.1.el6 (RHEL6.7)
293 2.6.32-642.11.1.el6 (RHEL6.8)
294 3.10.0-514.el7 (RHEL7.3)
295 3.0.101-0.47.71 (SLES11 SP3)
296 3.0.101-80 (SLES11 SP4)
297 3.12.62-60.64.8 (SLES12 SP1)
298 vanilla linux 4.5.7 (ZFS only)
299 * Client known to build on unpatched kernels:
300 2.6.32-431.29.2.el6 (RHEL6.5)
301 2.6.32-504.30.3.el6 (RHEL6.6)
302 2.6.32-573.26.1.el6 (RHEL6.7)
303 2.6.32-642.11.1.el6 (RHEL6.8)
304 3.10.0-514.el7 (RHEL7.3)
305 3.0.101-0.47.71 (SLES11 SP3)
306 3.0.101-80 (SLES11 SP4)
307 3.12.62-60.64.8 (SLES12 SP1)
309 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
310 * Recommended ZFS / SPL version: 0.6.5.8
311 * Tested with ZFS / SPL version: 0.6.5.8
312 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
313 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
314 stack overflow. For more information, please refer to bugzilla 17630.
315 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
316 the re-exporting nodes to be mounted with "32bitapi" mount option
318 --------------------------------------------------------------------------------
320 02-29-2016 Intel Corporation
322 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
323 for currently supported client and server kernel versions.
324 * Server known to build on patched kernels:
325 2.6.32-431.29.2.el6 (RHEL6.5)
326 2.6.32-504.30.3.el6 (RHEL6.6)
327 2.6.32-573.12.1.el6 (RHEL6.7)
328 3.10.0-327.3.1.el7 (RHEL7.2)
329 3.0.101-0.47.71 (SLES11 SP3)
330 3.0.101-68 (SLES11 SP4)
331 vanilla linux 4.2.1 (ZFS only)
332 * Client known to build on unpatched kernels:
333 2.6.32-431.29.2.el6 (RHEL6.5)
334 2.6.32-504.30.3.el6 (RHEL6.6)
335 2.6.32-573.12.1.el6 (RHEL6.7)
336 3.10.0-327.3.1.el7 (RHEL7.2)
337 3.0.101-0.47.71 (SLES11 SP3)
338 3.0.101-68 (SLES11 SP4)
341 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
342 * Recommended ZFS / SPL version: 0.6.4.2
343 * Tested with ZFS / SPL version: 0.6.4.2
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 03-10-2015 Intel Corporation
354 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
355 for currently supported client and server kernel versions.
356 * Server known to build on patched kernels:
357 2.6.32-431.29.2.el6 (RHEL6.5)
358 2.6.32-504.8.1.el6 (RHEL6.6)
359 3.0.101-0.46 (SLES11 SP3)
360 * Client known to build on unpatched kernels:
361 2.6.32-431.29.2.el6 (RHEL6.5)
362 2.6.32-504.8.1.el6 (RHEL6.6)
363 3.10.0-123.20.1.el7 (RHEL7)
364 3.0.101-0.46 (SLES11 SP3)
365 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
366 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
367 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
368 stack overflow. For more information, please refer to bugzilla 17630.
369 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
370 the re-exporting nodes to be mounted with "32bitapi" mount option
372 Severity : enhancement
374 Description: control OST-index in IDIF via ROCOMPAT flag.
375 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
376 last_rcvd file. For new formatted OST device, it will be auto set;
377 for the case of upgrading from old OST device, you can enable it
378 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
379 enabled, for new created OST-object, its IDIF-in-LMA will contain
380 the OST-index; for the existing OST-object, the OSD will convert
381 old format IDIF as new format IDIF with OST-index stored in the
382 LMA EA when accessing such OST-object or via OI scrub. Once such
383 flag is enabled, it cannot be reverted back, so the system cannot
384 be downgraded to the orignal incompatible version.
386 --------------------------------------------------------------------------------
388 07-30-2014 Intel Corporation
390 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
391 for currently supported client and server kernel versions.
392 * Server known to build on patched kernels:
393 2.6.32-431.20.3.el6 (RHEL6.5)
394 3.0.101-0.31 (SLES11 SP3)
395 * Client known to build on unpatched kernels:
396 2.6.32-431.20.3.el6 (RHEL6.5)
397 3.0.101-0.31 (SLES11 SP3)
400 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
401 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
402 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
403 stack overflow. For more information, please refer to bugzilla 17630.
404 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
405 the re-exporting nodes to be mounted with "32bitapi" mount option
407 --------------------------------------------------------------------------------
409 10-31-2013 Intel Corporation
411 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
412 * Server builds for kernels:
413 2.6.32-358.18.1.el6 (RHEL6.4)
414 * Client builds for unpatched kernels:
415 2.6.32-358.18.1.el6 (RHEL6.4)
416 3.0.92-0.8 (SLES11 SP3)
418 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
419 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
420 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
421 stack overflow. For more information, please refer to bugzilla 17630.
422 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
423 the re-exporting nodes to be mounted with "32bitapi" mount option
425 --------------------------------------------------------------------------------
427 05-30-2013 Intel Corporation
429 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
430 * Server builds for kernels:
431 2.6.32-358.6.2.el6 (RHEL6.4)
432 * Client builds for unpatched kernels:
433 2.6.32-358.6.2.el6 (RHEL6.4)
434 3.0.74-0.6.6 (SLES11 SP2)
436 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
437 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
438 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
439 stack overflow. For more information, please refer to bugzilla 17630.
440 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
441 the re-exporting nodes to be mounted with "32bitapi" mount option
443 --------------------------------------------------------------------------------
445 9-30-2012 Whamcloud, Inc.
447 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
448 * Server support for kernels:
449 2.6.32-279.5.1.el6 (RHEL6)
450 * Client support for unpatched kernels:
451 2.6.32-279.5.1.el6 (RHEL6)
453 2.6.18-238.19.1.el5 (RHEL5)
454 2.6.16.60-0.69.1 (SLES10)
455 * Recommended e2fsprogs version: 1.41.90.wc4
456 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
457 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
458 stack overflow. For more information, please refer to bugzilla 17630.
459 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
460 the re-exporting nodes to be mounted with "32bitapi" mount option
462 Severity : enhancement
464 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
466 Severity : enhancement
468 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
470 Severity : enhancement
472 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
474 Severity : enhancement
476 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
478 Severity : enhancement
480 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
482 --------------------------------------------------------------------------------
484 04-30-2012 Whamcloud, Inc.
486 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
487 * Server support for kernels:
488 2.6.32-220.4.2.el6 (RHEL6)
489 * Client support for unpatched kernels:
490 2.6.32-220.4.2.el6 (RHEL6)
492 2.6.18-238.19.1.el5 (RHEL5)
493 2.6.16.60-0.69.1 (SLES10)
494 * Recommended e2fsprogs version: 1.41.90.wc4
495 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
496 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
497 stack overflow. For more information, please refer to bugzilla 17630.
498 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
499 the re-exporting nodes to be mounted with "32bitapi" mount option
501 Severity : enhancement
503 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
505 Severity : enhancement
507 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
509 Severity : enhancement
511 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
513 Severity : enhancement
515 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
517 --------------------------------------------------------------------------------
519 09-30-2011 Whamcloud, Inc.
521 * Server support for kernels:
522 2.6.18-238.19.1.el5 (RHEL5)
523 2.6.32-131.6.1.el6 (RHEL6)
524 * Client support for unpatched kernels:
525 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
526 2.6.18-238.19.1.el5 (RHEL5)
527 2.6.32-131.6.1.el6 (RHEL6)
528 2.6.29.4-167.fc11 (FC11)
529 2.6.16.60-0.69.1 (SLES10)
531 * Recommended e2fsprogs version: 1.41.90.wc3
532 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
533 Lustre filesystem with 4K stack may cause a stack overflow. For more
534 information, please refer to bugzilla 17630.
535 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
536 mounted with "32bitapi" mount option
538 Severity : enhancement
539 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
541 Severity : enhancement
542 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
544 Severity : enhancement
545 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
548 Frequency : only with SLES10
550 Description: Use OFED "KMP" provided by Novell
551 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
552 Lustre is now built against this package. That means you need to
553 install the ofed-kmp package from Novell for the patchless client.
554 Note that the ofed-kmp that Novell ships may not exactly match the
555 kernel version but should still be compatible.
557 Severity : enhancement
559 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
560 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
562 --------------------------------------------------------------------------------
564 2010-07-15 Oracle, Inc.
566 * Server support for kernels:
567 2.6.18-164.11.1.el5 (RHEL 5)
568 2.6.18-164.11.1.0.1.el5 (OEL 5)
569 * Client support for unpatched kernels:
570 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
571 2.6.18-164.11.1.el5 (RHEL 5),
572 2.6.18-164.11.1.0.1.el5 (OEL 5)
573 2.6.16.60-0.42.8 (SLES 10),
575 2.6.29.4-167.fc11 (FC11)
576 * Recommended e2fsprogs version: 1.41.10-sun2
578 Severity : enhancement
580 Description: Distribute lfs_migrate script with lustre
581 Details : The lfs_migrate mentioned in the manual is now included in the
582 lustre tree in lustre/scripts and will distributed in lustre RPMs.
584 Severity : enhancement
586 Description: Failover nid list added to import proc file.
588 Severity : enhancement
590 Description: "som_preview" mount option required to use SOM
591 Details : The clients will need to use som_preview mount option
595 Description: Lustre does not do 1MB IOs to HW RAID
596 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
598 Severity : enhancement
600 Description: lfs quota output is not very convenient for awk/sed-parsing
601 Details : Some positions in lfs quota output table could be empty or
602 non-empty which made it hard to parse it with scripts, now
603 a dash is put instead of space where there is not supposed
608 Description: Update to OFED 1.5.1
609 Details : OFED 1.5.1 will be used in environments that don't use
610 vendor supplied OFED.
614 Description: Remove lnet. prefix from lctl params display
615 Details : Also renamed lnet.memused to lnet_memused to
620 Description: Use conf_param -d to remove permanant settings.
624 Description: Add list_param -R
625 Details : lctl list_param now supports recursive mode.
628 Bugzilla : 21829, 22312
629 Description: Fix llobdstat.
630 Details : 1. Now updating the write_bytes counter
631 2. Port script fixes from 1.8.
635 Description: decrease the usage of memory on clients.
636 Details : 1. On clients, recycle dentries and inodes unused.
637 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
642 Description: llobdstat fix and enhancement
643 Details : add a counter to set a limit to how many samples will be returned
644 fix a wildcard in the path to limit to obdfilter stats only
648 Description: lfs check only allowed for root
649 Details : Code cleanups and added a regresion test of non-root use of
654 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
655 Details : Fix hr->hr_index overload.
659 Description: Fix FIEMAP functionality.
663 Description: FC11 patchless client support added
667 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
668 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
672 Description: New acc-sm script and various test-framework enhancments.
676 Description: Previously landed patch was reverted.
680 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
684 Description: Add ldiskfs maxdirsize mount option.
688 Description: Fix mkfs.lustre support for 16TB LUNs.
692 Description: Print mmp_check_interval
693 Details : print mmp_check_interval and make it possible to abort
694 mount operation in case it takes too long.
698 Description: Add cascading_rw to lustre mpi tests.
702 Description: Store SOM data in EA.
703 Details : Also several SOM related cleanups.
707 Description: Add refcount to md_open_data
711 Description: Use refcount on lov_request_set to prevent memory corruption.
715 Description: Statahead: use parent dir's i_mutex for synchronization.
719 Description: loadgen improvements.
723 Description: Enable peer health detection in ksocklnd and ko2iblnd.
727 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
731 Description: Fix race on obd_devcie:md_stats
735 Description: Fix SLES 11 client builds.
736 Details : Resolved uninitalized pointer gcc warnings.
740 Description: Allow non-root access for "lfs check".
741 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
745 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
747 Severity : enhancement
749 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
750 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
754 Description: Sanity test_27p fails.
755 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
756 Sanity tests 27 were cleaned up and improved.
760 Description: Unify req format on client/servers
761 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
762 buffers and swabbers approach..
767 Description: drop join file stuff.
769 Severity : enhancement
771 Description: Add 2.6.27 and SLES11 (patchless client) support
775 Description: Fix in ptlrpc_expire_one_request() to print the signed time
778 Severity : enhancement
780 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
783 Severity : enhancement
785 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
787 Severity : enhancement
789 Description: correctly handle big reply message.
790 Details : send LNet event if reply is bigger than buffer and adjust this buffer
795 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
797 Severity : enhancement
799 Description: Add support for OEL5.
803 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
805 Severity : enhancement
807 Description: "weak-modules" support
808 Details : Implement "weak-modules" support which enables kernel modules
809 to be used with any kernel that implements the same kABI. In
810 order to achieve this modules are now installed in
811 /lib/modules/$(uname -r)/updates/kernel on all distributions.
815 Description: client could not reconnect to OST because of an active request.
816 Details : The client now retries to reconnect to the same server, if a
817 connect request failed with EBUSY or -EAGAIN. Test case 35b
818 conf-sanity.sh has been created.
822 Description: Don't allow make backward step on assignin osc next id.
823 Details : race between allocation next id and ll_sync thread can be cause
824 of set wrong osc next id and can be kill valid ost objects.
828 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
832 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
836 Description: Error handling in osc_statfs_interpret() has been improved.
837 Details : Check in osc_statfs_interpret() for EBADR.
841 Description: Do not send statfs() requests to OSTs disabled by administrator.
842 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
844 Severity : enhancement
846 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
848 Severity : enhancement
850 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
852 Severity : enhancement
854 Description: Enhancement for lfs(1) command to use numeric uid/gid.
858 Description: Improved cleanup procedure for conf-sanity test_46a.
862 Description: Do not start bulk transfer, if client tried to reconnect.
863 Details : Do not start bulk transfer, if client tried to reconnect,
864 but failed due to active rpcs. Check for exp_abort_active_req
865 has been implemented.
869 Description: Prevent inconsistences between linux and lustre mount structures.
870 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
871 Make the sleep interruptible.
873 Severity : enhancement
875 Description: provide server to client comms path
876 Details : server to client communications path via new LDLM_SET_INFO rpc
880 Description: truncate starts GFP_FS allocation under transaction causing deadlock
881 Details : ldiskfs_truncate calls grab_cache_page which may start page
882 allocation under an open transaction. This may lead to
883 calling prune_icache with consequent lustre reentrance.
887 Description: send correctly lsm on open replay
888 Details : MDS is trust to LSM size on replay open, but client can set wrong size
893 Description: enable client interface failover
894 Details : When a child reconnects from another NID, properly update export
895 nid hash position and ldlm reverse import.
897 Severity : enhancement
899 Description: Communicate OST degraded/readonly state via statfs to MDS
900 Details : Flags in the statfs returned from OSTs indicate whether the
901 OST is in a degraded RAID state, or if the filesystem has
902 turned read-only after a filesystem error is detected.
907 Description: don't panic if EPROTO was hit when reading symlink
908 Details : correctly handling request reference in error cases.
910 Severity : enhancement
912 Description: Add LustreNetLink, a kernel-userspace communcation path.
914 Severity : enhancement
916 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
921 Description: access to llog context before init.
922 Details : move handling CATALOGS file at osc layer and forbid access to llog
926 Frequency : rare, only when client is writing to OST on local node
928 Description: Avoid deadlock for local client writes
929 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
930 memory freeing context. This allows OST threads to set the
931 PF_MEMALLOC flag on task structures in order to allocate memory
932 from reserved pools and complete IO.
933 Use GFP_HIGHUSER for OST allocations for non-local client writes,
934 so that the OST threads generate memory pressure and allow
935 inactive pages to be reclaimed.
937 Severity : enhancement
939 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
944 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
945 Details : move ldlm namespace creation in setup phase to avoid grab
946 _lprocfs_lock with cli_sem held.
950 Description: Temporarily disable grant shrink.
951 Details : Disable the feature for debugging.
954 Frequency : only when formatting test filesystems
956 Description: Unable to run several mkfs.lustre on loop devices at the same time
957 Details : mkfs.lustre returns error 256 on the concurrent loop devices
958 formatting. The solution is to proper handle the error.
960 Severity : enhancement
962 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
964 Severity : enhancement
966 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
971 Description: resolve race between obd_disconnect and class_disconnect_exports
972 Details : if obd_disconnect will be called to already disconnected export he
973 forget release one reference and osc module can't unloaded.
975 Severity : enhancement
977 Description: Allow tuning service thread via /proc
978 Details : For each service a new
979 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
980 created that can be used to set min/max thread counts, and get the
981 current number of running threads.
986 Descriptoin: don't return error if have partially created objects for file.
987 Details : lov_update_create_set uses set->set_success as index for created
988 objects, so if some requests will be failed, they will have hole
989 at end of array and qos_shrink_lsm for allocate correct lsm.
991 Severity : enhancement
993 Description: Update OFED support to 1.4.1
998 Description: Reduce small size read RPC
999 Details : Set read-ahead limit for every file and only do read-ahead when
1000 available read-ahead pages are bigger than 1M to avoid small size
1003 Severity : enhancement
1005 Description: move AT tunable parameters for more consistent usage
1006 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1008 Severity : enhancement
1010 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1011 Details : allow skip disconnected ost for send statfs request and hide error
1017 Description: fix racy locking of mballoc block bitmaps causing BUG
1018 Details : The locking of the mballoc buddy bitmap and the in-memory
1019 block bitmap was using two different spin locks in some
1020 cases. This made it possible to incorrectly access the
1021 mballoc bitmap while another process was modifying it,
1022 causing a sanity assertion to fail. While no on-disk corruption
1023 was reported, there was some risk of this happening.
1026 Frequency : rare, on llog test 6
1028 Descriptoin: don't allow connect to already connected import
1029 Details : allowing connect to already connected import is hide connecting problem.
1034 Descriptoin: allow kill process which wait statahead result
1035 Details : for some reasons 'ls' can stick in waiting result from statahead,
1036 in this case need way for kill this process.
1038 Severity : enhancement
1040 Description: Add state history info file, enhance import info file
1041 Details : Track import connection state changes in a new osc/mdc proc file;
1042 add overview-type data to the osc/mdc import proc file.
1044 Severity : enhancement
1046 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1048 Severity : enhancement
1050 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1055 Descriptoin: OST grant too much space to client even there are not enough space.
1056 Details : Client will shrink its grant cache to OST if there are no write
1057 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1058 this grant cache if there are already not enough avaible space
1059 (left_space < total_clients * 32M).
1062 Frequency : start MDS on uncleanly shutdowned MDS device
1064 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1065 Details : stay in waiting mds<>ost recovery finished produce random bugs
1066 due race between two ll_sync thread for one lov target. send
1067 ACTIVATE event only if connect realy finished and import have
1071 Frequency : rare, connect and disconnect target at same time
1073 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1074 Details : don't call obd_disconnect under lov_lock. this long time
1075 operation and can block ptlrpcd which answer to connect request.
1080 Descriptoin: don't lose wakeup for imp_recovery_waitq
1081 Details : recover_import_no_retry or invalidate_import and import_close can
1082 both sleep on imp_recovery_waitq, but we was send only one wakeup
1086 Frequency : always with long access acl
1088 Descriptoin: mds can't pack reply with long acl.
1089 Details : mds don't control size of acl but they limited by reint/getattr
1092 Severity : enhancement
1094 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1096 Severity : enhancement
1098 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1101 Frequency : start MDS on uncleanly shutdowned MDS device
1103 Descriptoin: aborting recovery hang on MDS
1104 Details : don't throttle destroy RPCs for the MDT.
1107 Frequency : on remount
1109 Description: external journal device not working after the remount
1110 Details : clear dev_rdonly flag for external journal devices in
1116 Description: shutdown vs evict race
1117 Details : client_disconnect_export vs connect request race.
1118 if client will evicted at this time - we start invalidate
1119 thread without referece to import and import can be freed
1123 Frequency : rare, need acl's on inode.
1125 Description: client can't handle ost additional correctly
1126 Details : if ost was added after client connected to mds client can have
1127 hit lnet_try_match_md ... to big messages to wide striped files.
1128 in this case need teach client to handle config events about add
1129 lov target and update client max ea size at that event.
1131 Severity : enhancement
1133 Description: Changelogs
1134 Details : Changelogs are a lightweight mechanism to track filesystem
1135 metadata and namespace changes. The changelog is recorded
1136 permanently on the MDTs, and is periodically "consumed" / purged
1137 when records are no longer needed.
1139 Severity : enhancement
1141 Description: compact fld format with extents
1142 Details : Store range of seq rather than every seq in FLD. Seq
1143 controller update FLD rather than clients. In Case of CMD, mdt0
1144 has FLD, all other metadata server act as non persistent proxy
1145 for FLD queries and cache fld entries in fld cache.
1150 Description: don't skip ost target if they assigned to file
1151 Details : Drop slow OSCs if we can, but not for requested start idx.
1152 This means "if OSC is slow and it is not the requested
1153 start OST, then it can be skipped, otherwise skip it only
1154 if it is inactive/recovering/out-of-space.
1158 Description: more cleanup in mds_lov
1159 Details : not send LOV EA under replay, we can't know about they size at this
1160 time. Don't allow client connect to mds before any ost connected,
1161 for avoid problems with LOV EA size and returning EIO to client.
1163 Severity : enhancement
1165 Description: Interoperability at server side (Disk interoperability)
1167 Severity : enhancement
1169 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1171 Severity : enhancement
1173 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1175 Severity : enhancement
1177 Description: New client IO stack (CLIO).
1179 Severity : enhancement
1181 Description: Commit on sharing. Eliminate inter-client dependencies between
1182 uncommitted transactions by doing transaction commits.
1183 Thereby clients may recovery independently.
1186 Frequency : Create a symlink file with a very long name
1188 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1189 Details : If there is no extra space in the request for early cancels,
1190 ldlm_req_handles_avail() returns 0 instead of a negative value.
1192 Severity : enhancement
1194 Description: Add /proc entry for import status
1195 Details : The mdc, osc, and mgc import directories now have
1196 an import directory that contains useful import data for debugging
1197 connection problems.
1199 Severity : enhancement
1201 Description: Re-disable certain /proc logging
1202 Details : Enable and disable client's offset_stats, extents_stats and
1203 extents_stats_per_process stats logging on the fly.
1206 Frequency : Only on FC kernels 2.6.22+
1208 Description: oops in statahead
1209 Details : Do not drop reference count for the dentry from VFS when lookup,
1210 VFS will do that by itself.
1212 Severity : enhancement
1214 Description: Generic /proc file permissions
1215 Details : Set /Proc file permissions in a more generic way to enable non-
1216 root users operate on some /proc files.
1220 Description: Hitting mdc_commit_close() ASSERTION
1221 Details : Properly handle request reference release in
1222 ll_release_openhandle().
1226 Description: quota recovery deadlock during mds failover
1227 Details : This patch includes att18982, att18236, att18237 in bz14840.
1229 1. fix osts hang when mds does failover with quotaon
1230 2. prevent watchdog storm when osts threads wait for the
1235 Frequency : only patchless client
1236 Description: add workaround for race between add/remove dentry from hash
1238 Severity : enhancement
1240 Description: Allow OST glimpses to return PW locks
1244 Description: LBUG when llog conf file is full
1245 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1249 Description: Prevent import from entering FULL state when server in recovery
1253 Description: service mount cannot take device name with ":"
1254 Details : Only when device name contains ":/" will mount treat it as
1260 Description: replace ptlrpcd with the statahead thread to interpret the async
1261 statahead RPC callback
1265 Frequency : on recovery
1266 Description: I/O failures after umount during fail back
1267 Details : if client reconnected to restarted server we need join to recovery
1268 instead of find server handler is changed and process self
1269 eviction with cancel all locks.
1271 Severity : enhancement
1273 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1275 Severity : enhancement
1277 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1279 Severity : enhancement
1281 Description: Upcall on Lustre log has been dumped
1282 Details : Allow for a user mode script to be called once a Lustre log has
1283 been dumped. It passes the filename of the dumped log to the
1284 script, the location of the script can be specified via
1285 /proc/sys/lnet/debug_log_upcall.
1290 Description: avoid idr_remove called for id which is not allocated.
1291 Details : Move assigment s_dev for clustered nfs to end of initialization,
1292 to avoid problem with error handling.
1297 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1298 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1299 into CONN_USED_HASH and this prodice warning when put connection
1300 again in unused hash.
1303 Severity : enhancement
1305 Description: File striping can now be set to use an arbitrary pool of OSTs.
1307 Severity : enhancement
1309 Description: Export bytes_read/bytes_write count on OSC/OST.
1313 Description: Early reply size mismatch, MGC loses connection
1314 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1315 the connect flags are properly negotiated.
1320 Description: quotas are not honored with O_DIRECT
1321 Details : all writes with the flag O_DIRECT will use grants which leads to
1322 this problem. Now using OBD_BRW_SYNC to guard this.
1326 Description: add quota statistics
1327 Details : 1. sort out quota proc entries and proc code.
1328 2. add quota statistics
1330 Severity : enhancement
1332 Description: enable quota support for HEAD.
1336 Description: Properly propagate oinfo flags from lov to osc for statfs
1337 Details : restore missing copy oi_flags to lov requests.
1339 Severity : enhancement
1341 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1342 : l_getgroups(8), lst(8), routerstat(8)
1343 Details : included man pages for llobdstat(8), llstat(8),
1344 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1346 Severity : enhancement
1348 Description: configure's --enable-quota should check the
1349 : kernel .config for CONFIG_QUOTA
1350 Details : configure is terminated if --enable-quota is passed but
1351 : no quota support is in kernel
1355 Description: Remove portals compatibility
1356 Details : Remove portals compatibility, not interoperable with releases
1361 Description: Resolve device initialization race
1362 Details : Prevent proc handler from accessing devices added to the
1363 obd_devs array but yet be intialized.
1365 Severity : enhancement
1367 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1369 Severity : enhancement
1371 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1376 Description: Using wrong pointer in osc_brw_prep_request
1377 Details : Access to array[-1] can produce panic if kernel compiled with
1378 CONFIG_PAGE_ALLOC enabled
1382 Description: Client runs out of low memory
1383 Details : Consider only lowmem when counting initial number of llap pages
1387 Description: *optional* service tags registration
1388 Details : if the "service tags" package is installed on a Lustre node
1389 When the filesystem is mounted, a local-node service tag will
1390 be created. See http://inventory.sun.com/ for more information
1391 about the Service Tags asset management system.
1395 Description: Kernel BUG tries to release flock
1396 Details : Lustre does not destroy flock lock before last reference goes
1397 away. So always drop flock locks when client is evicted and
1398 perform unlock regardless of successfulness of speaking to MDS.
1402 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1406 Description: sanity test 65a fails if stripecount of -1 is set
1407 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1412 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1413 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1416 Severity : enhancement
1418 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1422 Description: do not process already freed flock
1423 Details : flock can possibly be freed by another thread before it reaches
1424 to ldlm_flock_completion_ast.
1428 Description: LBUG during stress test
1429 Details : Need properly lock accesses the flock deadlock detection list.
1433 Description: oops in page fault handler
1434 Details : kernel page fault handler can return two special 'pages' in error
1435 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1439 Description: timeout with invalidate import.
1440 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1441 should be handled by ptlrpcd. This produce long age waiting and
1442 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1444 Severity : enhancement
1446 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1450 Description: enable MGS and MDT services start separately
1451 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1452 a co-located MGS without starting the MGS, which is a complement
1453 to 'nosvc' mount option.
1457 Frequency : after recovery
1458 Description: precreate to many object's after del orphan.
1459 Details : del orphan st in oscc last_id == next_id and this triger growing
1460 count of precreated objects. Set flag LOW to skip increase count
1461 of precreated objects.
1465 Frequency : rare, on clear nid stats
1466 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1467 Details : when clean nid stats sometimes try destroy live entry,
1468 and this produce panic in free.
1472 Description: Stack overflow during MDS log replay
1473 ease stack pressure by using a thread dealing llog_process.
1477 Description: wait until IO finished before start new when do lock cancel.
1478 Details : VM protocol want old IO finished before start new, in this case
1479 need wait until PG_writeback is cleared until check dirty flag and
1480 call writepages in lock cancel callback.
1482 Severity : enhancement
1484 Description: using special macro for print time and cleanup in includes.
1488 Description: mds_mfd_close() ASSERTION(rc == 0)
1489 Details : In mds_mfd_close(), we need protect inode's writecount change
1490 within its orphan write semaphore to prevent possible races.
1494 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1498 Frequency : rare, on shutdown ost
1499 Description: don't hit live lock with umount ost.
1500 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1501 use shrink_dcache_sb instead.
1505 Description: don't panic with use echo client
1506 Details : echo client pass NULL as client nid pointer and this produce null
1507 pointer dereference.
1511 Description: fix build on ppc32
1512 Details : compile code with -m64 flag produce wrong object file for ppc32.
1516 Description: add message levels for liblustreapi
1520 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1521 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1522 will result in an -EAGAIN and in the exit call path we call
1523 alloc_rr() which will with increasing aggressiveness attempt to
1524 aquire precreated objects on the minimum number of required OSCs.
1528 Description: Use old size assignment to avoid deadlock
1529 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1530 the scheduling while holding a spinlock. We do not need locking
1531 for size in ll_update_inode() because size is only updated from
1532 the MDS for directories or files without objects, so there is no
1533 other place to do the update, and concurrent access to such inodes
1534 are protected by the inode lock.
1538 Description: resolve "_IOWR redefined" build error on SLES10
1542 Description: dump the memory debugging after all modules are unloaded to
1543 suppress false negative in conf_sanity test 39
1545 Severity : enhancement
1547 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1552 Description: fix SLES kernel versioning
1553 Details : the kernel version for our SLES 10 kernel did not include a "-"
1554 before the "smp" at the end. while this was not a problem in
1555 general, it did mean that software trying to use the kernel
1556 version to try to detect a vendor specific kernel would fail.
1557 this was most evident by the OFED build scripts.
1561 Description: Don't update lov_desc members until making sure they are valid
1562 Details : When updating lov_desc members via proc fs, need fix their
1563 validities before doing the real update.
1567 Description: don't put request into delay list while invalidate in flight.
1568 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1569 import in flight. this produce timeout for invalidate and sometimes
1570 can cause stale data.
1574 Frequency : on ppc only
1575 Description: not convert ost objects for directory because it's not exist.
1576 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1580 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1582 Severity : enhancement
1584 Description: Update to support 2.6.22.14 vanilla kernel.
1588 Frequency : rare, on recovery
1589 Description: read procfs can produce deadlock in some situation
1590 Details : Holding lprocfs lock which send rpc can produce block for destroy
1591 obd objects and this also block reconnect with -EALREADY. This
1592 isn't fix all lprocfs bugs - but make it rare.
1594 Severity : enhancement
1596 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1599 Frequency : frequent on X2 node
1601 Description: mdc_set_open_replay_data LBUG
1602 Details : Set replay data for requests that are eligible for replay.
1606 Description: lustre_mgs: operation 101 on unconnected MGS
1607 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1608 MGC, and late on MGC cannot successfully connect to MGS and a lot
1609 of the error messages complaining that MGS is not connected.
1613 Frequency : on network error
1614 Description: panic with double free request if network error
1615 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1616 but it's true only for synchronus enqueue, for async enqueue
1617 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1620 Severity : enhancement
1622 Description: client-side metadata stat-ahead during readdir(directory readahead)
1623 Details : perform client-side metadata stat-ahead when the client detects
1624 readdir and sequential stat of dir entries therein
1627 Frequency : on start mds
1629 Description: Implement get_info(last_id) in obdfilter.
1632 Frequency : occasional
1634 Description: Correctly check stale fid, not start epoch if ost not support SOM
1635 Details : open with flag O_CREATE need set old fid in op_fid3 because
1636 op_fid2 was overwritten with new generated fid, but mds can answer
1637 with one of these two fids and both is not stale. Setattr
1638 incorrectly started an epoch and assume will be called
1639 done_writing, but without SOM done_writing ever being called.
1642 Frequency : rare, depends on device drivers and load
1644 Description: MDS or OSS nodes crash due to stack overflow
1645 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1646 In some cases, in conjunction with device drivers that use a lot
1647 of stack the MDS (or possibly OSS) service threads could overflow
1648 the stack. One change which was identified to consume additional
1649 stack has been reworked to avoid the extra stack usage.
1652 Frequency : occasional
1654 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1655 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1656 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1657 oscc_has_objects and see OSC has no objests and call
1658 oscc_internal_create to resend create request.
1660 Severity : enhancement
1662 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1664 Severity : enhancement
1666 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1669 Frequency : very rare
1671 Description: Fix replay if there is an un-replied request and open
1672 Details : In some cases, older replay request will revert the
1673 mcd->mcd_last_xid on MDS which is used to record the client's
1674 latest sent request.
1676 Severity : enhancement
1678 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1680 Severity : enhancement
1682 Description: Add rhel5 support to HEAD.
1684 Serverity : enhancement
1686 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1691 Description: Don't allow skipping OSTs if index has been specified.
1692 Details : Don't allow skipping OSTs if index has been specified, make
1693 locking in internal create lots better.
1697 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1698 Details : access to bitfield in structure is always rounded to long
1699 and this produce problem with not atomic change any bit.
1703 Description: Lustre make rpms failed.
1704 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1705 builds Lustre rpms from tarball.
1708 Frequency : rare on shutdown ost
1710 Description: If llog cancel was not send before clean_exports phase, this can
1711 produce deadlock in llog code.
1712 Details : If llog thread has last reference to obd and call class_import_put
1713 this produce deadlock because llog_cleanup_commit_master wait when
1714 last llog_commit_thread exited, but this never success because was
1715 called from llog_commit_thread.
1719 Description: allow userland application know is lost one of stripes.
1720 Details : fill lvb_blocks with error code on ost and return it to
1721 application if error flag found.
1725 Description: NULL lov_tgts causing MDS oops
1726 Details : more safe checks for NULL lov_tgts for avoid oops.
1728 Severity : enhancement
1730 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1734 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1735 Details : If an OST is down the MDS will hang indefinitely in
1736 obd_statfs_rqset() waiting for the statfs data. While for
1737 MDS QOS usage of statfs, it should not stuck in waiting.
1739 Severity : enhancement
1741 Description: remote_acl support
1742 Details : Support ACL-based permission check for remote user.
1743 Support setfacl/getfacl for remote user with the utils
1744 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1745 the system "{s,g}etfacl" utils.
1747 Severity : enhancement
1749 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1751 Severity : enhancement
1753 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1757 Description: make mgs_setparam() handle fsname containing dash
1758 Details : fsname containing a dash does not work with lctl conf_param
1760 Severity : enhancement
1762 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1764 Severity : enhancement
1766 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1770 Frequency : rare, at shutdown
1771 Description: access already free / zero obd_namespace.
1772 Details : if client_disconnect_export was called without force flag set,
1773 and exist connect request in flight, this can produce access to
1774 NULL pointer (or already free pointer) when connect_interpret
1775 store ocd flags in obd_namespace.
1779 Frequency : only at startup
1780 Description: not alloc memory with spinlock held.
1781 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1782 if any spinlock held.
1784 Severity : enhancement
1786 Description: make lustre randomly fail allocating memory
1787 Details : Make lustre randomly failed allocating memory for testing purpose.
1789 Severity : enhancement
1791 Description: lost problems with lov objid file
1792 Details : Fixes some scability and access to not inited memory problems
1793 in work with lov objdid file.
1798 Description: lfs find does not continue on file error
1799 Details : Continue other files processing when a file/dir is absent.
1803 Description: Inconsistent usage of lustre_pack_reply()
1804 Details : Standardize the usage of lustre_pack_reply() such that it
1805 always generate a CERROR on failure.
1808 Frequency : occasional
1810 Description: MDS hang or stay in waiting lock
1811 Details : If client receive lock with CBPENDING flag ldlm need send lock
1812 cancel as separate rpc, to avoid situation when cancel request
1813 can't processed due all i/o threads stay in wait lock.
1817 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1821 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1825 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1829 Description: Update to RHEL4 latest kernel.
1831 Severity : enhancement
1833 Description: Build SLES10 patchless client fails
1834 Details : The configure was broken by run ./configure with
1835 --with-linux-obj=.... argument for patchless client. When the
1836 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1837 can't search header adequately. Use absolute path such as
1838 -I($LINUX)/include instead.
1842 Description: interrupt oig_wait produce painc on resend.
1843 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1844 private set, and this produce situation when rq_ptlrpcd_data not
1845 copyed to new allocated request and triggered LBUG on assert
1846 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1847 ptlrpcd set if request is changed and can be safety changed to use
1852 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1853 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1854 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1855 these macros and replace them with their difinitions instead.
1859 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1860 Details : Update which_patch & target file for SLES10 latest kernel.
1863 Description: Remove client patches from SLES 10 kernel.
1864 Details : This causes SLES 10 clients to behave as patchless clients
1865 even on a Lustre-patched (server) kernel.
1867 Severity : enhancement
1869 Description: self-adjustable client's lru lists
1870 Details : use adaptive algorithm for managing client cached locks lru
1871 lists according to current server load, other client's work
1872 pattern, memory activities, etc. Both, server and client
1873 side namespaces provide number of proc tunables for controlling
1876 Severity : enhancement
1878 Description: light-weight GSS support
1879 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1880 and reduce performance overhead.
1882 Severity : enhancement
1884 Description: Linux keyring support
1885 Details : Support using service of Linux keyring for Lustre GSS internal
1886 context refresh/cache mechanism.
1890 Description: Fix errors in lfs documentation
1891 Details : Fixes man pages
1895 Description: don't use GFP_* in generic Lustre code.
1896 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1900 Description: obdclass is limited by single OBD_ALLOC(idarray)
1901 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1905 Description: warnings with build patchless client with vanila 2.6.19 and up
1906 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1907 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1911 Frequency : Only for SLES
1912 Description: sanity_quota fail test_1
1913 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1914 file, which makes TSTID[2] inunique.
1918 Frequency : Only for SLES9
1919 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1920 Details : Update target/ChangeLog/which_patch .
1924 Frequency : Only for test_10 in sanity-quota.sh
1925 Description: a bug in quota test code
1926 Details : checking a test flag in a wrong place
1930 Frequency : Only for test_13 in sanity-quota.sh
1931 Description: a bug in quota test script
1932 Details : shouldn't check overfull allocated quota
1936 Description: jbd statistics
1937 Details : Port older jbd statistics patch for sles10
1939 Severity : enhancement
1941 Description: Add printing inode info into message about error in writepage.
1946 Description: reply_lock_interpret crash due to race with it and lock cancel.
1947 Details : Do not replay locks that are being cancelled. Do not reference
1948 locks by their address during replay, just by their handle.
1952 Frequency : When flocks are used.
1953 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1954 Details : Flock locks might destroy just granted lock if it could be merged
1955 with another existing flock, this is done in completion handler,
1956 so teach ldlm_cli_enquque_fini that this is a valid case for
1962 Description: Oops in read and write path when failing to allocate lock.
1963 Details : Check if lock allocation failed and return error back.
1967 Description: lstripe command fails for valid OST index
1968 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1969 instead of lov->desc.ld_active_tgt_count.
1973 Description: sanity.sh failed test 103
1974 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1979 Description: df doesn't work properly if diskfs blocksize != 4K
1980 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1984 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1985 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1987 Severity : enhancement
1989 Description: Adaptive timeouts
1990 Details : RPC timeouts adapt to changing server load and network
1991 conditions to reduce resend attempts and improve recovery time.
1995 Description: llapi_file_create() does not allow some changes
1996 Details : add llapi_file_open() that allows specifying the mode and
1997 open flags, and also returns an open file handle.
2001 Description: merge and cleanup kernel patches.
2002 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2006 Description: Add journal checksum support.(Kernel part)
2007 Details : The journal checksum feature adds two new flags i.e
2008 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2009 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2010 indicates that the commit block contains the checksum for
2011 the blocks described by the descriptor blocks. Now commit
2012 record can be sent to disk without waiting for descriptor
2013 blocks to be written to disk. This behavior is controlled
2014 using JBD2_FEATURE_ASYNC_COMMIT flag.
2018 Description: OSS needs mutliple precreate threads
2019 Details : Add ability to start more than one create thread per OSS.
2023 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2024 Details : execute lfs setstripe on client
2028 Description: mds_obd_create error creating tmp object
2029 Details : When the user sets quota on root, llog will be affected and can't
2030 create files and write files.
2033 Frequency : Always on ia64 patchless client, and possibly others.
2035 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2036 Details : This allows the patchless client to be loaded on architectures
2037 without this export.
2041 Description: RedHat Update kernel for RHEL5
2042 Details : Add the kernel config file for RHEL5.
2046 Description: RedHat Update kernel for RHEL5
2047 Details : Modify the kernel config file more closer RHEL5.
2051 Description: Build failure against Centos5 (RHEL5)
2052 Details : Define PAGE_SIZE when it isn't present.
2056 Description: test_7 on sanity-quota.sh can't be used on separate machines
2057 Details : add facet to handle it
2061 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2062 Details : with NFS, the anon dentry's parent was set to itself in
2063 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2064 corresponding dentry other than use rec->ur_name.
2066 Severity : enhancement
2068 Description: lfs setstripe enhancement
2069 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2073 Description: enable data checksumming by default
2074 Details : enable checksum by default, allow --disable-checksum
2075 configure option and "-o nochecksum" mount option. Checksums
2076 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2080 Description: lustre support for RHEL5
2081 Details : Add support for RHEL5.
2085 Description: Client eviction due to failover config
2086 Details : after a connection loss, the lustre client should attempt to
2087 reconnect to the last active server first before trying the
2088 other potential connections.
2092 Description: when mds and osts use different quota unit(32bit and 64bit),
2093 quota will be released repeatly.
2094 Details : void sending multiple quota reqs to mds, which will keep the
2095 status between the reqs.
2099 Description: osts not allocated evenly to files
2100 Details : change the condition to increase offset_idx
2103 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2105 Description: Data corruption for OSTs that are formatted larger than 2TB
2107 Details : When generating the bio request for lustre file writes the
2108 sector number would overflow a temporary variable before being
2109 used for the IO. The data reads correctly from Lustre (which
2110 will overflow in a similar manner) but other file data or
2111 filesystem metadata may be corrupted in some cases.
2115 Description: Tune the kernel for good SCSI performance.
2116 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2117 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2122 Description: rewrite ext2-derived code in obdclass/uuid.c
2123 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2124 from scratch preserving functionality.
2128 Description: "lfs find -obd UUID" prints directories
2129 Details : "lfs find -obd UUID" will return all directory names instead
2130 of just file names. It is incorrect because the directories
2131 do not reside on the OSTs.
2135 Description: use i_size_read and i_size_write in 2.6 port
2136 Details : replace inode->i_size access with i_size_read/write()
2139 Frequency : when removing large files
2141 Description: scheduling issue during removal of large Lustre files
2142 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2143 It causes scheduling issues when removing large files (17TB in the
2147 Frequency : only with liblustre clients on XT3
2149 Description: evictions taking too long
2150 Details : allow llrd to evict clients directly on OSTs
2153 Frequency : only on ppc
2155 Description: /proc/fs/lustre/devices broken on ppc
2156 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2157 We should make sure the type of 'index' is loff_t and then cast
2158 explicitly as needed below. Do not assign an explicitly cast
2162 Frequency : only for rhel5
2164 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2165 Details : Modify the target file & which_kernel.
2167 Severity : enhancement
2169 Description: omit set fsid for export NFS
2170 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2171 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2175 Frequency : Always, for kernels after 2.6.16
2176 Description: Fix warning idr_remove called for id=.. which is not allocated.
2177 Details : Last kernels save old s_dev before kill super and not allow
2178 to restore from callback - restore it before call kill_anon_super.
2182 Description: buffer overruns could theoretically occur
2183 Details : llapi_semantic_traverse() modifies the "path" argument by
2184 appending values to the end of the origin string, and a
2185 overrun may occur. Adding buffer overrun check in liblustreapi.
2189 Description: Fix error on 'ls .' at the top of the Lustre mount.
2190 Details : Don't revalidate dentry if it is a root dentry.
2194 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2195 Details : Modify vm-tunables-rhel4.patch.
2199 Description: Kernel config for 2.6.18-vanilla.
2200 Details : Modify targets/2.6-vanilla.target.in.
2201 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2202 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2203 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2204 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2208 Description: improve handling recoverable errors
2209 Details : if request processig with error which can be recoverable on server
2210 request should be resend, otherwise page released from cache and
2215 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2216 Details : Modify target file & which_patch.
2217 A flaw was found in the IA32 system call emulation provided
2218 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2219 value could be stored in the %RAX register, which could trigger an
2220 out-of-bounds system call table access. An untrusted local user
2221 could exploit this flaw to run code in the kernel
2222 (ie a root privilege escalation). (CVE-2007-4573).
2226 Description: O_DIRECT bypasses client statistics.
2227 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2228 read_ahead_stats were updated. copied the stats section from
2229 osc_send_oap_rpc() into async_internal().
2233 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2237 Description: change order of libsysio includes
2238 Details : '#include sysio.h' should always come before '#include xtio.h'
2242 Description: handle "serious error: objid * already exists" more gracefully
2243 Details : If LAST_ID value on disk is smaller than the objects existing in
2244 the O/0/d* directories, it indicates disk corruption and causes an
2245 LBUG(). If the object is 0-length, then we should use the existing
2246 object. This will help to avoid a full fsck in most cases.
2248 Severity : enhancement
2250 Description: adapt the lustre_config script to support the upgrade case
2251 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2252 server targets to 1.6 in parallel.
2256 Description: To avoid grant space > avaible space when the disk is almost
2257 full. Without this patch you might see the error "grant XXXX >
2258 available" or some LBUG about grant, when the disk is almost
2260 Details : In filter_check_grant, for non_grant cache write, we should
2261 check the left space by if (*left > ungranted + bytes), instead
2262 of (*left > ungranted), because only we are sure the left space
2263 is enough for another "bytes", then the ungrant space should be
2264 increase. In client, we should update cl_avail_grant only there
2265 is OBD_MD_FLGRANT in the reply.
2269 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2270 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2271 described in CVE-2007-4573. This problem could allow untrusted
2272 local users to gain root access.
2275 Frequency : when using O_DIRECT and quotas
2277 Description: Incorrect file ownership on O_DIRECT output files
2278 Details : block usage reported by 'lfs quota' does not take into account
2279 files that have been written with O_DIRECT.
2284 Description: touch file failed when fs is not full
2285 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2286 otherwise we can get ENOSP while fs is not full.
2290 Description: parallel lock callbacks
2291 Details : Instead of sending blocking and completion callbacks as separated
2292 requests, adding them to a set and sending in parallel.
2295 Frequency : only for Cray XT3
2296 Bugzilla : 12829/13455
2297 Description: Changing primary group doesn't change the group lustre assigns to
2299 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2300 overridden with the primary group provided by the group upcall,
2301 whereas the supplied fsgid can be trusted if it is in the list of
2302 supplementary groups returned by the group upcall.
2304 Severity : enhancement
2306 Description: Allow masking D_WARNING, D_ERROR messages from console
2307 Details : Console messages can now be disabled via lnet.printk.
2311 Description: User code with malformed file open parameter crashes client node
2312 Details : Before packing join_file req, all the related reference should be
2313 checked carefully in case some malformed flags cause fake
2314 join_file req on client.
2318 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2319 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2320 mode is equal to the granted mode. However, it should also check
2321 that LDLM_FL_CP_REQD is not set.
2325 Description: Heavy nfs access might result in deadlocks
2326 Details : After ELC code landed, it is now improper to enqueue any mds
2327 locks under och_sem, because enqueue might want to decide to
2328 cancel open locks for same inode we are holding och_sem for.
2332 Description: Client eviction while running blogbench
2333 Details : A lot of unlink operations with concurrent I/O can lead to a
2334 deadlock causing evictions. To address the problem, the number of
2335 oustanding OST_DESTROY requests is now throttled to
2336 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2337 callbacks are processed in priority.
2341 Description: enable ACLs on MDS by default
2342 Details : ACLs must be enabled on MDS by default.
2345 Frequency : PPC/PPC64 only
2347 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2348 Details : fix duplicated definitions between asm-ppc64/types.h and
2349 lustre_types.h on PPC.
2352 Frequency : PPC/PPC64 only
2354 Description: asm-ppc/segment.h does not exist
2355 Details : fix compile issue on PPC.
2359 Description: better handle error messages in extents code
2362 Frequency : RHEL4 only
2364 Description: mkfs is very slow on IA64/RHEL4
2365 Details : A performance regression has been discovered in the MPT Fusion
2366 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2367 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2368 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2370 Severity : enhancement
2372 Description: SNMP support enhancement
2373 Details : Adding total number of sampled request for an MDS node in snmp
2376 Severity : enhancement
2378 Description: Optimize ldlm waiting list processing for PR extent locks
2379 Details : When processing waiting list for read extent lock and meeting read
2380 lock that is same or wider to it that is not contended, skip
2381 processing rest of the list and immediatelly return current
2382 status of conflictness, since we are guaranteed there are no
2383 conflicting locks in the rest of the list.
2387 Description: Time out and refuse to reconnect
2388 Details : When the failover node is the primary node, it is possible
2389 to have two identical connections in imp_conn_list. We must
2390 compare not conn's pointers but NIDs, otherwise we can defeat
2391 connection throttling.
2395 Description: port llog fixes from b1_6 into HEAD
2396 Details : Port llog reference couting and some llog cleanups from b1_6
2397 (bug 10800) into HEAD, for protect from panic and access to already
2398 free llog structures.
2402 Description: Detect stride IO mode in read-ahead
2403 Details : When a client does stride read, read-ahead should detect that and
2404 read-ahead pages according to the detected stride pattern.
2408 Description: data checksumming impacts single node performance
2409 Details : add support for several checksum algorithm. Currently, only CRC32
2410 and Adler-32 are supported. The checksum type can be changed on
2411 the fly via /proc/fs/lustre/osc/*/checksum_type.
2415 Description: use adler32 for page checksums
2416 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2421 Description: build for x2 fails
2422 Details : fix compile issue on Cray systems.
2426 Description: Properly match for duplicate locks
2427 Details : Due to different lock order from skiplists code, we need to
2428 traverse entire list for now
2431 Frequency : only on PPC/SLES10
2433 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2434 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2435 be defined. Add a hack in mkfs_lustre.c to work around this compile
2440 Description: LASSERT on MDS when client holding flock lock dies
2441 Details : ldlm pool logic depends on number of granted locks equal to
2442 number of released locks which is not true for flock locks, so
2443 just exclude such locks from consideration.
2447 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2448 Details : Use fsfilt_sync() for both the whole filesystem sync and
2449 individual file sync to eliminate dangerous inode locking
2450 with I_LOCK that can lead to a deadlock.
2454 Description: performance in 1.6.3
2455 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2456 sure that 1MB requests are not fragmented by the block layer.
2458 Severity : enhancement
2460 Description: organize the server-side client stats on per-nid basis
2461 Details : Change the structure of stats under obdfilter and mds to
2469 The "uuid"s file would list the uuids of _active_ exports.
2470 And the clear entry is to clear all stats and stale nids.
2472 Severity : enhancement
2474 Description: eliminate client locks in face of contention
2475 Details : file contention detection and lockless i/o implementation
2476 for contended files.
2480 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2483 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2485 Description: MDS may lose file striping (and hence file data) in some cases
2486 Details : If there are additional extended attributes stored on the MDS,
2487 in particular ACLs, SELinux, or user attributes (if user_xattr
2488 is specified for the client mount options) then there is a risk
2489 of attribute loss. Additionally, the Lustre file striping
2490 needs to be larger than default (e.g. striped over all OSTs),
2491 and an additional attribute must be stored initially in the
2492 inode and then increase in size enough to be moved to the
2493 external attribute block (e.g. ACL growing in size) for file
2498 Description: skiplist implementation simplification
2499 Details : skiplists are used to group compatible locks on granted list
2500 that was implemented as tracking first and last lock of each
2501 lock group the patch changes that to using doubly linked lists
2505 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2506 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2507 because LASSERT on both the data supplied by a client, and the
2508 data on disk is dangerous and incorrect.
2510 Severity : enhancement
2512 Description: Slow truncate/writes to huge files at high offsets.
2513 Details : Directly associate cached pages to lock that protect those pages,
2514 this allows us to quickly find what pages to write and remove
2515 once lock callback is received.
2519 Description: more ldlm soft lockups
2520 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2521 starve other threads from the resource lock for a long time in
2522 case of long waiting queue, so change the debug level from
2523 D_OTHER to the less frequently used D_INFO.
2525 Severity : enhancement
2527 Description: add -gid, -group, -uid, -user options to lfs find
2531 Description: Hung threads in invalidate_inode_pages2_range
2532 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2533 once one is completed. As a result, some RPCs are stuck in the
2534 queue and are never sent.
2538 Description: filter threads hungs on waiting journal commit
2539 Details : Cleanup filter group llog code, then only filter group llog will
2540 be only created in the MDS/OST syncing process.
2544 Description: Procfs and llog threads access destoryed import sometimes.
2545 Details : Sync the import destoryed process with procfs and llog threads by
2546 the import refcount and semaphore.
2548 Severity : enhancement
2550 Description: openlock cache of b1_6 port to HEAD
2555 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2556 Details : If insertion of an extent fails, then discard the inode
2557 preallocation and free data blocks else it can lead to duplicate
2562 Description: don't always update ctime in ext3_xattr_set_handle()
2563 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2564 In some cases the ctime should not be updated, for example for
2565 2.0->1.8 compatibility it is necessary to delete an xattr and it
2566 should not update the ctime.
2570 Bugzilla : 15713/16362
2571 Description: Assertion in iopen_connect_dentry in 1.6.3
2572 Details : looking up an inode via iopen with the wrong generation number can
2573 populate the dcache with a disconneced dentry while the inode
2574 number is in the process of being reallocated. This causes an
2575 assertion failure in iopen since the inode's dentry list contains
2576 both a connected and disconnected dentry.
2580 Description: assertion failure in ldlm_handle2lock()
2581 Details : fix a race between class_handle_unhash() and class_handle2object()
2582 introduced in lustre 1.6.5 by bug 13622.
2587 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2588 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2593 Description: oops in sd_iostats_seq_show()
2594 Details : unloading/reloading the scsi low level driver triggers a kernel
2595 bug when trying to access the sd iostat file.
2600 Description: Kernel panics during QLogic driver reload
2601 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2602 patch, causing memory corruption.
2607 Description: journal_dev option does not work in b1_6
2608 Details : pass mount option during pre-mount.
2610 Severity : enhancement
2612 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2613 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2614 extent information of a file. It can be used to map logical blocks
2615 in a file to physical blocks in the block device.
2619 Description: LDLM soft lockups - improvement
2620 Details : It is be possible to send the lock handle along with each read
2621 or write request because the client is already doing a lock match
2622 itself so there isn't any reason the OST should have to re-do that
2626 Frequency : only on Cray X2
2628 Description: X2 build failures
2629 Details : fix build failures on Cray X2.
2633 Description: xid & resent requests
2634 Details : Initialize RPC XID from clock at startup (randomly if clock is
2637 Severity : enhancement
2639 Description: Add lustre_start utility to start or stop multiple Lustre servers
2644 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2645 Details : In case of memory pressure, list_del() can be called twice on
2646 req->rq_history_list, causing a kernel oops.
2650 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2651 Details : fix stack overflow in the distributed lock manager by defering
2652 export eviction after a failed AST to the elt thread instead of
2653 handling it in the dlm interpret routine.
2657 Description: Convert some comments to new format.
2658 Details : Update documenting comments to match doxygen conventions.
2662 Description: Grammar fixes.
2663 Details : A couple of trivial spelling fixes.
2667 Description: OSD_COUNTERS-mandatory
2668 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2669 to simplify the code.
2673 Description: simplify cmm_device freeing logic.
2674 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2678 Description: Add lockdep support to dt_object_operations locking interface.
2679 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2680 parameter indicating lock ordering. Update mdd code to use new
2685 Description: Introduce failloc constants for lockless IO tests.
2686 Details : Add two new failloc constants to test lockless IO. Only one of
2687 them in implemented---another is checked in yet to be landed
2692 Description: Add lockdep support for inode mutex.
2693 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2694 in the situations where Lustre has to lock more than one inode
2699 Description: Add optional invariants checking support.
2700 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2701 configure switch. This macro is to be used for consistency and
2702 sanity checks that are too expensive to be left in `production'
2707 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2708 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2713 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2714 Details : Use spin_lock_nested() in (the only) situation where more than
2715 one ldlm_lock is locked simultaneously. Also, fix possible
2716 dead-lock in ldlm_lock_change_resource() by enforcing particular
2721 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2722 Details : Instead of specifying each ldlm_lock call-back through separate
2723 parameter, wrap them into struct ldlm_callback_suite.
2727 Description: Kill join_lru obd method and its callers.
2728 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2729 in memory---a requirement implemented through ->o_join_lru() obd
2730 method in HEAD. Remove this method and its users.
2734 Description: Add asynchronous ldlm ENQUEUE completion handler.
2735 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2736 case that stock ldlm_completion_ast() cannot handle as it waits
2737 until lock is granted. Introduce new ldlm_completion_ast_async()
2738 for this. Also comment ldlm_completion_ast().
2742 Description: ldlm_error <-> errno conversion.
2743 Details : Add functions to map (rather arbitrary) between LDLM error codes
2744 and standard errno values. CLIO needs this to prevent LDLM specific
2745 constants from escaping ldlm and osc.
2749 Description: Kill unused ldlm_handle2lock_ns() function.
2750 Details : Kill unused ldlm_handle2lock_ns() function.
2754 Description: Add lu_ref support to ldlm_lock
2755 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2756 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2757 and ldlm_resource. LDLM interface has to be changed, because code that
2758 releases a reference on a lock, has to "know" what reference this is.
2759 In the most frequent case
2761 lock = ldlm_handle2lock(handle);
2763 LDLM_LOCK_PUT(lock);
2765 no changes are required. When any other reference (received _not_ from
2766 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2767 instead of LDLM_LOCK_PUT().
2769 Arguably, changes are pervasive, and interface requires some discipline
2770 for proper use. On the other hand, it was very instrumental in finding
2771 a few leaked lock references.
2775 Description: Add ldlm_lock_addref_try().
2776 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2777 attempts to addref a lock that might be being canceled
2782 Description: Add ldlm_weigh_callback().
2783 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2784 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2785 hard-coded `number of pages' logic.
2789 Description: Add lockdep annotations to llog code.
2790 Details : Use appropriately tagged _nested() locking calls in the places
2791 where llog takes more than one ->lgh_lock lock.
2795 Description: Add loi_kms_set().
2796 Details : Wrap kms updates into a helper function.
2800 Description: Constify instances of struct lsm_operations.
2801 Details : Constify instances of struct lsm_operations.
2805 Description: lu_conf support.
2806 Details : On a server, a file system object is uniquely identified
2807 by a fid, which is sufficient to locate and load all object
2808 state (inode). On a client, on the other hand, more data are
2809 necessary instantiate an object. Change lu_object_find() and
2810 friends to take additional `lu_conf' argument describing object.
2811 Typically this includes layout information.
2815 Description: lu_context fixes.
2816 Details : Introduce new lu_context functions that are needed on the client
2817 side, where some system threads (ptlrpcd) are shared by multiple
2818 modules, and so cannot be stopped during module shutdown.
2822 Description: Add start and stop methods to lu_device_type_operations.
2823 Details : Introduce two new methods in lu_device_type_operations, that are
2824 invoked when first instance of a given type is created and last one
2825 is destroyed respectively. This is need by CLIO.
2829 Description: Add lu_ref support to struct lu_device.
2830 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2831 track leaked references.
2835 Description: Introduce lu_kmem_descr.
2836 Details : lu_kmem_descr and its companion interface allow to create
2837 and destroy a number of kmem caches at once.
2841 Description: Fix lu_object finalization race.
2842 Details : Fix a race between lu_object_find() finding an object and its
2843 concurrent finalization. This race is (most likely) not possible
2844 on the server, but might happen on the client.
2848 Description: Introduce lu_ref interface.
2849 Details : lu_ref is a debugging module allowing to track references to
2850 a given object. It is quite cpu expensive, and has to be
2851 explicitly enabled with --enable-lu_ref. See usage description
2856 Description: Factor lu_site procfs stats into a separate function.
2857 Details : Separate lu_site stats printing code into a separate function
2858 that can be reused on a client.
2862 Description: Constify instances of struct {lu,dt,md}_device_operations.
2863 Details : Constify instances of struct {lu,dt,md}_device_operations.
2867 Description: Introduce struct md_site and move meta-data specific parts of
2868 struct lu_site here.
2869 Details : Move md-specific fields out of struct lu_site into special struct
2870 md_site, so that lu_site can be used on a client.
2874 Description: Kill mdd_lov_destroy().
2875 Details : Remove unused mdd code.
2879 Description: Add st_block checking to multistat.c.
2880 Details : Add st_block checking to multistat.c.
2884 Description: Add lu_ref support to struct obd_device.
2885 Details : Add lu_ref tracking to obd_device.
2889 Description: Kill obd_set_fail_loc().
2890 Details : Remove unused code.
2894 Description: Add special type for ptlrpc_request interpret functions.
2895 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2896 there. Actual usage will be in CLIO.
2900 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2901 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2902 doesn't work with lockdep.
2906 Description: Add rwv.c test program.
2907 Details : New testing program exercising readv(2) and writev(2) (Qian).
2911 Description: Add sendfile.c test program.
2912 Details : New testing program exercising sendfile(2) (Jay).
2916 Description: Ratelimit a message that can be very frequent.
2917 Details : Ratelimit a memory allocation failure message that can
2922 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2923 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2927 Description: Liblustre build fixes.
2928 Details : Liblustre build fixes.
2932 Description: libcfs: add cfs_{need,cond}_resched() interface.
2933 Details : libcfs: add cfs_{need,cond}_resched() definition and
2934 implementations for Linux, NT, and liblustre.
2936 Severity : enhancement
2938 Description: More exported tunables for mballoc
2939 Details : Add support for tunable preallocation window and new tunables for
2940 large/small requests
2944 Description: Detect corruption of block bitmap and checking for preallocations
2945 Details : Checks validity of on-disk block bitmap. Also it does better
2946 checking of number of applied preallocations. When corruption is
2947 found, it turns filesystem readonly to prevent further corruptions.
2951 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2952 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2953 window to 0. If it does hit cache miss, and read-pattern is still
2954 stride-io mode, does not reset the stride window, but also does not
2955 increase the stride window length in this case.
2959 Frequency : only for big-endian servers
2960 Description: Check if system is big-endian while mounting fs with extents feature
2961 Details : Mounting a filesystem with extents feature will fail on big-endian
2962 systems since ext3-based ldiskfs is not supported on big-endian
2963 systems. This can be over-riden with "bigendian_extents" mount option.
2965 Severity : enhancement
2967 Description: The root squash functionality
2968 Details : A security feature, which is to prevent users from being able
2969 to mount lustre on their desktop, run as root, and delete
2970 all of the files in the filesystem. The goal is accomplished by
2971 remapping user id (UID) and group id (GID) of the root user to
2972 a UID and GID specified by the system administartor via Lustre
2973 configuration management server (MGS). The functionality also
2974 allows to specify sets of clients for which the remapping does
2979 Description: Excessive recovery window
2980 Details : With AT enabled, the recovery window can be excessively long (6000+
2981 seconds). To address this problem, we no longer use
2982 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2983 timeout no longer depends on the service time, it is set to
2984 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2985 time via pb_service_time.
2989 Description: Watchdog triggered on MDS failover
2990 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2991 the OSTs know that the MDS "UUID" can be reused for the same export
2992 from a different NID, so we do not need to wait for the export to be
2996 Frequency : rare, only if using MMP with Linux RAID
2998 Description: MMP doesn't work with Linux RAID
2999 Details : While using HA for Lustre servers with Linux RAID, it is possible
3000 that MMP will not detect multiple mounts. To make this work we
3001 need to unplug the device queue in RAID when the MMP block is being
3002 written. Also while reading the MMP block, we should read it from
3003 disk and not the cached one.
3005 Severity : enhancement
3007 Description: open file using fid
3008 Details : A file can be opened using just its fid, like
3009 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3012 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3015 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3016 Details : add proper locking for creation of jbd_4k slab cache
3020 Description: MMP check in ext3_remount() fails without displaying any error
3021 Details : When multiple mount protection fails during remount, proper error
3024 Severity : enhancement
3026 Description: Allow stripe size to be up to 4G-64k
3027 Details : Fix math logic to allow large stripe sizes.
3031 Description: add check for >8TB ldiskfs filesystems
3032 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3033 Don't allow >8TB ldiskfs filesystems to be mounted without
3034 force_over_8tb mount option
3039 Descriptoin: OSS complains "too many blocks" and crashes
3040 Details : In filter_commitrw_write():
3041 1. lquota_chkquota();
3042 2. obdo_from_inode();
3043 3. lquota_pending_commit();
3044 "2." has changed uid and gid in this case, then caused
3045 lquota_chkquota() and lquota_pending_commit() to use
3046 different uid and gid. That is the root of the bug.
3050 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3051 Details : Detect and print out the 'stripe count = -1' case properly.
3055 Description: Out or order replies might be lost on replay
3056 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3057 request with tid smaller than one currently being added, add it
3058 to the start, not end of the list.
3060 --------------------------------------------------------------------------------
3062 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3064 * Support for kernels:
3065 2.6.5-7.283 (SLES 9),
3066 2.6.9-55.EL (RHEL 4),
3067 2.6.16.46-0.14 (SLES 10),
3068 2.6.18.8 vanilla (kernel.org)
3069 * Client support for unpatched kernels:
3070 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3071 2.6.16 - 2.6.22 vanilla (kernel.org)
3072 * Due to recently discovered recovery problems, we do not recommend
3073 using patchless RHEL 4 clients with this or any earlier release.
3074 * Recommended e2fsprogs version: 1.39.cfs8
3075 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3076 * Starting with this release, the ldiskfs backing filesystem required
3077 by Lustre is now in its own package, lustre-ldiskfs. This package
3078 should be installed. It is versioned separately from Lustre and
3079 may be released separately in future.
3083 Description: block reactivating mgc import until all deactivates complete
3084 Details : Fix race when failing back MDT/MGS to itself (testing)
3087 Frequency : at statup only
3089 Description: mds_lov_synchronize race leads to various problems
3090 Details : simultaneous MDT->OST connections at startup can cause the
3091 sync to abort, leaving the OSC in a bad state.
3093 Severity : enhancement
3095 Description: add optional extra BUILD_VERSION info
3096 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3097 to override the lustre version.
3100 Frequency : 2.6.18 servers only
3102 Description: ll_kern_mount() doesn't release the module reference
3103 Details : The ldiskfs module reference count never drops down to 0
3104 because ll_kern_mount() doesn't release the module reference.
3109 Description: server LBUG when using old ost_num_threads parameter
3110 Details : Accept the old ost_num_threads parameter but warn that it
3111 is deprecated, and fix an off-by-one error that caused an LBUG.
3116 Description: Transient SCSI error results in persistent IO issue
3117 Details : iobuf->dr_error is not reinitialized to 0 between two
3121 Frequency : sometimes when underlying device returns I/O errors
3123 Description: OSTs not going read-only during write failures
3124 Details : OSTs are not remounted read-only when the journal commit threads
3125 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3126 instead of the ext3 wrappers.
3130 Description: poor connect scaling with increasing client count
3131 Details : Don't run filter_grant_sanity_check for more than 100 exports
3132 to improve scaling for large numbers of clients.
3135 Frequency : SLES10 only
3137 Description: sanity-quota.sh quotacheck failed: rc = -22
3138 Details : Quotas cannot be enabled on SLES10.
3141 Frequency : liblustre clients only
3143 Description: getdirentries does not give error when run on compute nodes
3144 Details : getdirentries does not fail when the size specified as an argument
3145 is too small to contain at least one entry
3147 Severity : enhancement
3149 Description: Add LNET router traceability for debug purposes
3150 Details : If a checksum failure occurs with a router as part of the
3151 IO path, the NID of the last router that forwarded the bulk data
3152 is printed so it can be identified.
3157 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3158 Details : Due to a race condition, liblustre clients were occasionally
3159 evicted incorrectly.
3161 Severity : enhancement
3163 Description: lfs setstripe use optional parameters instead of postional
3166 Severity : enhancement
3168 Description: Nanosecond timestamp support for ldiskfs
3169 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3170 resolution timestamps. There is not yet support for this at
3171 the Lustre filesystem level.
3174 Frequency : during server recovery
3176 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3177 Details : request with rq_no_resend flag not awake l_wait_event if they get
3181 Frequency : nfs export on patchless client
3183 Description: connectathon hang when test nfs export over patchless client
3184 Details : Disconnected dentry cannot be found with lookup, so we do not need
3185 to unhash it or make it invalid
3188 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3189 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3190 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3195 Description: open req refcounting wrong on reconnect
3196 Details : If reconnect happened between getting open reply from server and
3197 call to mdc_set_replay_data in ll_file_open, we will schedule
3198 replay for unreferenced request that we are about to free.
3199 Subsequent close will crash in variety of ways.
3200 Check that request is still eligible for replay in
3201 mdc_set_replay_data().
3206 Description: disable writes to filesystem when reading health_check file
3207 Details : the default for reading the health_check proc file has changed
3208 to NOT do a journal transaction and write to disk, because this
3209 can cause reads of the /proc file to hang and block HA state
3210 checking on a healthy but otherwise heavily loaded system. It
3211 is possible to return to the previous behaviour during configure
3212 with --enable-health-write.
3214 Severity : enhancement
3216 Description: 64-bit inode version
3217 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3218 to the inode. This will be required for version-based recovery.
3223 Description: MDS fails to start if a duplicate client export is detected
3224 Details : in some rare cases it was possible for a client to connect to
3225 an MDS multiple times. Upon recovery the MDS would detect this
3226 and fail during startup. Handle this more gracefully.
3228 Severity : enhancement
3230 Description: Add -o localflock option to simulate old noflock
3232 Details : This will achieve local-only flock/fcntl locks
3238 Description: log_commit_thread vs filter_destroy race leads to crash
3239 Details : Take import reference before releasing llog record semaphore
3244 Description: Wrong request locking in request set processing
3245 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3246 imp_delayed_list, in this place should be used imp_lock.
3249 Frequency : when reconnection
3251 Description: Grant Leak when osc reconnect to OST
3252 Details : When osc reconnect ost, OST(filter) should check whether it
3253 should grant more space to client by comparing fed_grant and
3254 cl_avail_grant, and return the granted space to client instead
3255 of "new granted" space, because client will call osc_init_grant
3256 to update the client grant space info.
3259 Frequency : when client reconnect to OST
3261 Description: Grant Leak when osc do resend and replay bulk write
3262 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3263 bulk write request, because the grant info while be sync between
3264 OSC and OST when reconnect, and we should ignore the grant info
3265 these of resend/replay write req.
3270 Description: Grant space more than available left space sometimes.
3271 Details : When then OST is about to be full, if two bulk writing from
3272 different clients came to OST. Accord the avaliable space of the
3273 OST, the first req should be permitted, and the second one
3274 should be denied by ENOSPC. But if the seconde arrived before
3275 the first one is committed. The OST might wrongly permit second
3276 writing, which will cause grant space > available space.
3279 Frequency : when client is evicted
3281 Description: Grant might be wrongly erased when osc is evicted by OST
3282 Details : when the import is evicted by server, it will fork another
3283 thread ptlrpc_invalidate_import_thread to invalidate the
3284 import, where the grant will be set to 0. While the original
3285 thread will update the grant it got when connecting. So if
3286 the former happened latter, the grant will be wrongly errased
3287 because of this race.
3292 Description: Checking Stale with correct fid
3293 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3294 to check whether it is stale, because sometimes, we want the
3295 enqueue happened anyway, and op_data.fid2 will not be initialized.
3297 Severity : enhancement
3299 Description: update patchless client
3300 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3303 Frequency : only with 2.4 kernel
3305 Description: random memory corruption
3306 Details : size of struct ll_inode_info is to big for union inode.u and this
3307 can be cause of random memory corruption.
3312 Description: Memory leak in recovery
3313 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3314 It should also check obd_fail before fsfilt_start, otherwise if
3315 fsfilt_start return -EROFS,(failover mds during mds recovery).
3316 then the req will return with repmsg->transno = 0 and rc = EROFS.
3317 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3318 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3319 be freed no matter whether fsfilt_commit success or not.
3322 Frequency : only with huge count clients
3324 Description: Prevents from taking the superblock lock in llap_from_page for
3326 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3327 for llap_from_page call in ll_removepage prevents from taking the
3328 superblock lock for a soon died page.
3333 Description: Not check open intent error before release open handle
3334 Details : in some rare cases, the open intent error is not checked before
3335 release open handle, which may cause
3336 ASSERTION(open_req->rq_transno != 0), because it tries to release
3337 the failed open handle.
3342 Description: Set cat log bitmap only after create log success.
3343 Details : in some rare cases, the cat log bitmap is set too early. and it
3344 should be set only after create log success.
3348 Description: Accessing a block bevice can re-enable I/O when Lustre is
3349 tearing down a device.
3350 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3355 Description: service threads may hog cpus when there are a lot of requests
3356 Details : Insert cond_resched to give other threads a chance to use some CPU
3361 Description: the cat log was not initialized in recovery
3362 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3363 unlink log on mds will not be initialized until mds post
3364 recovery. And also in mds post recovery, the unlink log will
3365 initialization will be done asynchronausly, so there will be race
3366 between add unlink log and unlink log initialization.
3370 Description: brw_stats were being printed incorrectly
3371 Details : brw_stats were being printed as log2 but all of them were not
3372 recorded as log2. Also remove some code duplication arising from
3373 filter_tally_{read,write}.
3377 Frequency : rare, only in recovery.
3378 Description: ASSERTION(req->rq_type != LI_POISON) failed
3379 Details : imp_lock should be held while iterating over imp_sending_list for
3380 prevent destroy request after get timeout in ptlrpc_queue_wait.
3384 Description: replay-single.sh test 52 fails
3385 Details : A lock's skiplist need to be cleanup when it being unlinked
3386 from its resource list.
3390 Description: Short directio read returns full requested size rather than
3392 Details : Direct I/O operations should return actual amount of bytes
3393 transferred rather than requested size.
3395 Severity : enhancement
3397 Description: metadata RPC reduction (e.g. for rm performance)
3398 Details : decrease the amount of synchronous RPC between clients and servers
3399 by canceling conflicing lock before the operation on the client
3400 and packing thier handles into the main operation RPC to server.
3402 Severity : enhancement
3404 Description: Async OSC create to avoid the blocking unnecessarily.
3405 Details : If an OST has no remaining object, system will block on the
3406 creation when it needs to create a new object on this OST. Now,
3407 ways use pre-created objects when available, instead of blocking on
3408 an empty osc while others are not empty. If we must block, we
3409 block for the shortest possible period of time.
3413 Description: improve handling recoverable errors
3414 Details : if request processig with error which can be recoverable on server
3415 request should be resend, otherwise page released from cache and
3418 Severity : enhancement
3420 Description: refine locking for avoid write wrong info into lov_objid file
3421 Details : fix possible races with add new target and write/update data in
3424 --------------------------------------------------------------------------------
3426 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3431 Frequency : on some architectures
3433 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3434 Details : Uninitialized flags sometimes cause configuration commands to
3438 Frequency : patchless clients only
3440 Description: missing __iget() symbol export
3441 Details : The __iget() symbol export is missing. To avoid the need for
3442 this on patchless clients the deathrow inode reaper is turned
3443 off, and we depend on the VM to clean up old inodes. This
3444 dependency was during via the fix for bug 12181.
3446 --------------------------------------------------------------------------------
3448 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3450 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3451 INTEROPERATE with older versions automatically. In many cases a
3452 special upgrade step is needed. Please read the
3453 user documentation before upgrading any part of a live system.
3454 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3455 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3456 * WARNING: Lustre configuration and startup changes are required with
3457 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3461 Severity : enhancement
3463 Description: MountConf
3464 Details : Lustre configuration is now managed via mkfs and mount
3465 commands instead of lmc and lconf. New obd types (MGS, MGC)
3466 are added for dynamic configuration management. See
3467 https://mail.clusterfs.com/wikis/lustre/MountConf for
3470 Severity : enhancement
3472 Description: dynamic OST addition
3473 Details : OSTs can now be added to a live filesystem
3475 Severity : enhancement
3477 Description: startup order invariance
3478 Details : MDTs and OSTs can be started in any order. Clients only
3479 require the MDT to complete startup.
3481 Severity : enhancement
3483 Description: parallel, asynchronous orphan cleanup
3484 Details : orphan cleanup is now performed in separate threads for each
3485 OST, allowing parallel non-blocking operation.
3487 Severity : enhancement
3489 Description: optimized stripe assignment
3490 Details : stripe assignments are now made based on ost space available,
3491 ost previous usage, and OSS previous usage, in order to try
3492 to optimize storage space and networking resources.
3494 Severity : enhancement
3496 Description: Permanently set tunables
3497 Details : All writable /proc/fs/lustre tunables can now be permanently
3498 set on a per-server basis, at mkfs time or on a live system.
3500 Severity : enhancement
3502 Description: Lustre message v2
3503 Details : Add lustre message format v2.
3505 Severity : enhancement
3507 Description: client OST exclusion list
3508 Details : Clients can be started with a list of OSTs that should be
3509 declared "inactive" for known non-responsive OSTs.
3513 Description: SPEC SFS validation failure on NFS v2 over lustre.
3514 Details : Changes the blocksize for regular files to be 2x RPC size,
3515 and not depend on stripe size.
3517 Severity : enhancement
3519 Description: Multiple MD RPCs in flight.
3520 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3521 To avoid overly-overloading MDS, introduce a limit on number of
3522 MDS RPCs in flight for a single client and add /proc controls
3523 to adjust this limit.
3525 Severity : enhancement
3527 Description: client read/write statistics
3528 Details : Add client read/write call usage stats for performance
3529 analysis of user processes.
3530 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3531 file access. extents_stats shows chunk size distribution.
3532 extents_stats_per_process show chunk size distribution per
3535 Severity : enhancement
3537 Description: mds statistics
3538 Details : Add detailed mds operations statistics in
3539 /proc/fs/lustre/mds/*/stats.
3543 Description: Failure of copying files with lustre special EAs.
3544 Details : Client side always return success for setxattr call for lustre
3545 special xattr (currently only "trusted.lov").
3550 Description: Refcount LNET uuids
3551 Details : The global LNET uuid list grew linearly with every startup;
3552 refcount repeated list entries instead of always adding to
3555 Severity : enhancement
3557 Description: Dynamic service threads
3558 Details : Within a small range, start extra service threads
3559 automatically when the request queue builds up.
3562 Frequency : mixed-endian client/server environments
3564 Description: mixed-endian crashes
3565 Details : The new msg_v2 system had some failures in mixed-endian
3568 Severity : enhancement
3570 Description: Easy OST removal
3571 Details : OSTs can be permanently deactivated with e.g. 'lctl
3572 conf_param lustre-OST0001.osc.active=0'
3574 Severity : enhancement
3576 Description: MGS proc entries
3577 Details : Added basic proc entries for the MGS showing what filesystems
3580 Severity : enhancement
3582 Description: provide MGS failover
3583 Details : Added config lock reacquisition after MGS server failover.
3585 Severity : enhancement
3587 Description: add Linux 2.4 support
3588 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3592 Description: a large application tries to do I/O to the same resource and dies
3593 in the middle of it.
3594 Details : Check the req->rq_arrival time after the call to
3595 ost_brw_lock_get(), but before we do anything about
3596 processing it & sending the BULK transfer request. This
3597 should help move old stale pending locks off the queue as
3598 quickly as obd_timeout.
3601 Frequency : when an incorrect nid is specified during startup
3603 Description: ptlrpc connect to non-existant node causes kernel crash
3604 Details : LNET can't be re-entered from an event callback, which
3605 happened when we expire a message after the export has been
3606 cleaned up. Instead, hand the zombie cleanup off to another
3609 Severity : enhancement
3611 Description: plain/inodebits lock performance improvement
3612 Details : Grouping plain/inodebits in granted list by their request modes
3613 and bits policy, thus improving the performance of search through
3617 Frequency : only if OST filesystem is corrupted
3619 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3620 Details : for a short time RPCs with bulk IO are in the replay list,
3621 but replay of bulk IOs is unimplemented. If the OST filesystem
3622 is corrupted due to disk cache incoherency and then replay is
3623 started it is possible to trip an assertion. Avoid putting
3624 committed RPCs into the replay list at all to avoid this issue.
3627 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3629 Description: System hang on startup
3630 Details : This bug allowed the liblustre (e.g. catamount) client to
3631 return to the app before handling all startup RPCs. This
3632 could leave the node unresponsive to lustre network traffic
3633 and manifested as a server ptllnd timeout.
3635 Severity : enhancement
3637 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3638 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3639 boolean to control whether to print peer debug info when a
3640 client's RPC times out.
3643 Frequency : only for kernels with patches from Lustre below 1.4.3
3645 Description: Remove old rdonly API
3646 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3649 Frequency : only for devices with external journals
3651 Description: Set external device read-only also
3652 Details : During a commanded failover stop, we set the disk device
3653 read-only while the server shuts down. We now also set any
3654 external journal device read-only at the same time.
3657 Frequency : when upgrading from 1.4 while trying to change parameters
3659 Description: The wrong (new) MDC name was used when setting parameters for
3660 upgraded MDT's. Also allows changing of OSC (and MDC)
3661 parameters if --writeconf is specified at tunefs upgrade time.
3664 Frequency : when setting specific ost indicies
3666 Description: QOS code breaks on skipped indicies
3667 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3668 created with --index need not be sequential.
3670 Severity : enhancement
3672 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3673 Details : The uninit_groups feature works in conjunction with the kernel
3674 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3675 up the pass1 processing of e2fsck. This is a read-only feature
3676 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3677 mount filesystems that have had this feature enabled.
3679 Severity : enhancement
3681 Description: Improve multi-block allocation algorithm to avoid fragmentation
3682 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3683 allocation locality and avoid filesystem fragmentation.
3685 ------------------------------------------------------------------------------
3687 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3689 * Support for kernels:
3690 2.4.21-47.0.1.EL (RHEL 3)
3691 2.6.5-7.283 (SLES 9)
3692 2.6.9-42.0.10.EL (RHEL 4)
3693 2.6.12.6 vanilla (kernel.org)
3694 2.6.16.27-0.9 (SLES 10)
3695 * Recommended e2fsprogs version: 1.39.cfs5
3697 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3701 Frequency : occasional, depends on client load and configuration
3702 Bugzilla : 12181, 12203
3703 Description: data loss for recently-modified files
3705 Details : In some cases it is possible that recently written or created
3706 files may not be written to disk in a timely manner (this should
3707 normally be within 30s unless client IO load is very high).
3708 The problem appears as zero-length files or files that are a
3709 multiple of 1MB in size after a client crash or client eviction
3710 that are missing data at the end of the file.
3712 This problem is more likely to be hit on clients where files are
3713 repeatedly created and unlinked in the same directory, clients
3714 have a large amount of RAM, have many CPUs, the filesystem has
3715 many OSTs, the clients are rebooted frequently, and/or the files
3716 are not accessed by other nodes after being written.
3718 The presence of the problem can be detected by looking at
3719 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3720 smaller than the second (nr_unused) then dirty files will not
3721 be flushed automatically to disk. "sync; sleep 10" should be
3722 run several times on the node before unmounting it to update
3723 Lustre (this is also safe to run on nodes without this problem).
3725 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3726 that can cause this same problem, so customers using that kernel
3727 also need to update the kernel in addition to Lustre. In order
3728 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3730 It is normal that files written just before a client crash (less
3731 than 30s) may not yet have been flushed to disk, even for local
3735 Frequency : frequent on thin XT3 nodes
3737 Description: UUID collision on thin XT3 Linux nodes
3738 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3739 randomly, since we relied on an insufficiently-seeded PRNG.
3744 Description: OSS hangs after "All ost request buffers busy"
3745 Details : A deadlock between quota and journal operations caused OSS
3746 hangs after printing "All ost request buffers busy."
3749 Frequency : always on liblustre builds
3751 Description: Cleanup compiler warnings on liblustre
3754 Frequency : always on liblustre builds on XT3
3756 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3760 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3761 > 32000 subdirectories
3762 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3763 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3764 subdirectory count crosses 32000. This will aid e2fsck to
3765 correctly handle more than 32000 subdirectories.
3768 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3770 Description: System hang on startup
3771 Details : This bug allowed the liblustre (e.g. catamount) client to
3772 return to the app before handling all startup RPCs. This
3773 could leave the node unresponsive to lustre network traffic
3774 and manifested as a server ptllnd timeout.
3776 Severity : enhancement
3778 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3779 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3780 boolean to control whether to print peer debug info when a
3781 client's RPC times out.
3786 Description: make O_SYNC working on 2.6 kernels
3787 Details : 2.6 kernels use different method for mark pages for write,
3788 so need add a code to lustre for O_SYNC work.
3793 Description: Failure to close file and release space on NFS
3794 Details : Put inode details into lock acquired in ll_intent_file_open.
3795 Use mdc_intent_lock in ll_intent_open to properly
3796 detect all kind of errors unhandled by mdc_enqueue.
3801 Description: proc file read during shutdown sometimes raced obd removal,
3803 Details : Add lock to prevent obd access after proc file removal.
3806 Frequency : Only for files larger than 4GB on 32-bit clients.
3808 Description: improperly doing page alignment of locks
3809 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3810 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3815 Description: RPCs being resent when they shouldn't be
3816 Details : Some RPCs that should not be resent are being resent. This
3817 can cause inconsistencies in the RPC state machine. Do not
3818 resend such requests.
3821 Frequency : rare, only with NFS export
3823 Description: Crash on NFS re-export node
3824 Details : under very unusual load conditions an assertion is hit in
3825 ll_intent_file_open()
3828 Frequency : only if OST filesystem is corrupted
3830 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3831 Details : for a short time RPCs with bulk IO are in the replay list,
3832 but replay of bulk IOs is unimplemented. If the OST filesystem
3833 is corrupted due to disk cache incoherency and then replay is
3834 started it is possible to trip an assertion. Avoid putting
3835 committed RPCs into the replay list at all to avoid this issue.
3840 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3841 Details : Large single O_DIRECT read and write calls can fail to allocate
3842 a sufficiently large buffer to process the request. In case of
3843 allocation failure the allocation is retried with a smaller
3844 buffer and broken into smaller requests.
3846 Severity : enhancement
3848 Description: Add -o localflock option to simulate old noflock behaviour.
3849 Details : This will achieve local-only flock/fcntl locks coherentness.
3854 Description: versioning check is incomplete
3855 Details : Checking the version difference of client vs. server, report
3856 error if the gap is too big.
3861 Description: add support PG_writeback bit
3862 Details : add support for PG_writeback bit for Lustre, for more carefull
3863 work with page cache in 2.6 kernel. This also fix some deadlocks
3864 and remove hack for work O_SYNC with 2.6 kernel.
3866 Severity : enhancement
3868 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3869 Details : The uninit_groups feature works in conjunction with the kernel
3870 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3871 up the pass1 processing of e2fsck. This is a read-only feature
3872 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3873 mount filesystems that have had this feature enabled.
3875 Severity : enhancement
3877 Description: Improve multi-block allocation algorithm to avoid fragmentation
3878 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3879 allocation locality and avoid filesystem fragmentation.
3881 ------------------------------------------------------------------------------
3883 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3885 * Support for kernels:
3886 2.6.9-42.0.3.EL (RHEL 4)
3887 2.6.5-7.276 (SLES 9)
3888 2.4.21-47.0.1.EL (RHEL 3)
3889 2.6.12.6 vanilla (kernel.org)
3890 2.6.16.21-0.8 (SLES10)
3891 * Recommended e2fsprogs version: 1.39.cfs2-0
3893 * The backwards-compatible /proc/sys/portals symlink has been removed
3894 in this release. Before upgrading, please ensure that you change
3895 any configuration scripts or /etc/sysctl.conf files that access
3896 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3897 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3898 in advance of the upgrade on any system running Lustre 1.4.6 or
3899 newer, since /proc/sys/lnet was added in that version.
3900 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3904 Frequency : only when quota is used
3906 Description: avoid scanning export list for quota master
3907 Details : Change the algorithms to avoid scanning export list in order
3908 to improve the efficiency.
3911 Frequency : MDS failover only, very rarely
3913 Description: "went back in time" messages on mds failover
3914 Details : The greatest transno may be lost when the current operation
3915 finishes with an error (transno==0) and the client's last_rcvd
3916 record is over-written. Save the greatest transno in the
3917 mds_last_transno for this case.
3920 Frequency : always for specific kernels and striping counts
3922 Description: client may get "Matching packet too big" without ACL support
3923 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3924 when trying to access files in certain configurations. The
3925 clients should in fact be denied when mounting because they do
3926 not understand ACLs.
3929 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3931 Description: many clients connecting with IO in progress causes connect timeouts
3932 Details : Avoid synchronous journal commits to avoid delays caused by many
3933 clients connecting/disconnecting when bulk IO is in progress.
3934 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3935 OST_IO_PORTAL to avoid delays behind potentially many pending
3939 Frequency : occasionally with multiple writers to a single file
3941 Description: shared writes to file may result in wrong size reported by stat()
3942 Details : Allow growing of kms when extent lock is cancelled
3945 Frequency : always with random mmap IO to multi-striped file
3947 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3948 Details : Only if the hole is at the end of OST object so that kms is too
3949 small. Fix is to increase kms accordingly in ll_nopage.
3952 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3954 Description: writes to a missing object would leak memory on the OST
3955 Details : If there is an inconsistency between the MDS and OST filesystems,
3956 such that the MDS references an object that doesn't exist, writes
3957 to that object will leak memory due to incorrect cleanup in the
3958 error handling path, eventually running out of memory on the OST.
3963 Description: Creating too long symlink causes lustre errors
3964 Details : Check symlink and name lengths before sending requests to MDS.
3967 Frequency : only if flock is enabled (not on by default)
3969 Description: posix locks not released on fd closure on 2.6.9+
3970 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3971 kernels, this caused such locks not to be released on fd close and
3972 then assertions on fs unmount about still used locks.
3975 Frequency : MDS failover only, very rarely
3977 Description: clients may get ASSERTION(granted_lock != NULL)
3978 Details : When request was taking a long time, and a client was resending
3979 a getattr by name lock request. The were multiple lock requests
3980 with the same client lock handle and
3981 mds_getattr_name->fixup_handle_for_resent_request found one of the
3982 lock handles but later failed with ASSERTION(granted_lock != NULL).
3987 Description: handle->h_buffer_credits > 0, assertion failure
3988 Details : h_buffer_credits is zero after truncate, causing assertion
3989 failure. This patch extends the transaction or creates a new
3993 Frequency : NFS re-export or patchless client
3994 Bugzilla : 11179, 10796
3995 Description: Crash on NFS re-export node (__d_move)
3996 Details : We do not want to hash the dentry if we don't have a lock.
3997 But if this dentry is later used in d_move, we'd hit uninitialised
3998 list head d_hash, so we just do this to init d_hash field but
3999 leave dentry unhashed.
4002 Frequency : NFS re-export or patchless client
4004 Description: NFS exports has problem with symbolic link
4005 Details : lustre client didn't properly install dentry when re-exported
4006 to NFS or running patchless client.
4009 Frequency : NFS re-export or patchless client
4011 Description: Various nfs/patchless fixes.
4012 Details : fixes reuse disconected alias for lookup process - this fixes
4013 warning "find_exported_dentry: npd != pd",
4014 fix permission error with open files at nfs.
4015 fix apply umask when do revalidate.
4018 Frequency : occasional
4020 Description: Crash on NFS re-export node
4021 Details : calling clear_page() on the wrong pointer triggered oops in
4022 generic_mapping_read().
4025 Frequency : rarely, using O_DIRECT IO
4027 Description: unaligned directio crashes client with LASSERT
4028 Details : check for unaligned buffers before trying any requests.
4031 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4033 Description: stale data returned from RAID cache
4034 Details : If only a small amount of IO is done to the RAID device before
4035 reading it again it is possible to get stale data from the RAID
4036 cache instead of reading it from disk.
4039 Frequency : always for sles10 kernel
4041 Description: sles10 support
4042 Details : ll_follow_link: compile fixes and using of nd_set_link
4043 under newer kernels.
4046 Frequency : depends on arch, kernel and compiler version, always on sles10
4049 Description: recursive or deep enough symlinks cause stack overflow
4050 Details : getting rid of large stack-allocated variable in
4054 Frequency : depends on hardware
4056 Description: lustre write performance loss in the SLES10 kernel
4057 Details : the performance loss is caused by using of write barriers in the
4058 ext3 code. The SLES10 kernel turns barrier support on by
4059 default. The fix is to undo that change for ldiskfs.
4061 ------------------------------------------------------------------------------
4063 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4065 * Support for kernels:
4066 2.6.9-42.0.3EL (RHEL 4)
4067 2.6.5-7.276 (SLES 9)
4068 2.4.21-47.0.1.EL (RHEL 3)
4069 2.6.12.6 vanilla (kernel.org)
4073 Frequency : quota enabled and large files being deleted
4075 Description: releasing more than 4GB of quota at once hangs OST
4076 Details : If a user deletes more than 4GB of files on a single OST it
4077 will cause the OST to spin in an infinite loop. Release
4078 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4083 Description: statfs data retrieved from /proc may be stale or zero
4084 Details : When reading per-device statfs data from /proc, in the
4085 {kbytes,files}_{total,free,avail} files, it may appear
4086 as zero or be out of date.
4089 Frequency : systems with MD RAID1 external journal devices
4091 Description: lconf's call to blkid is confused by RAID1 journal devices
4092 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4093 of returning all block devices that match the journal UUID.
4096 Frequency : always, for aggregate stripe size over 4GB
4098 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4099 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4100 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4103 Frequency : always if "lfs find" used on a local file/directory
4105 Description: "lfs find" segfaults if used on a local file/directory
4106 Details : The case where a directory component was not specified wasn't
4107 handled correctly. Handle this properly.
4110 Frequency : always on ppc64
4112 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4113 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4114 perform endianness conversion.
4117 Frequency : rarely (truncate to non-zero file size after write under load)
4118 Bugzilla : 10730, 10687
4119 Description: Files padded with zeros to next 4K multiple
4120 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4121 it is possible that files that are truncated to a non-zero size
4122 immediately after being written are filled with zero bytes beyond
4123 the truncated size. No file data is lost.
4125 Severity : enhancement
4127 Description: Allow recovery/failover for liblustre clients.
4128 Details : liblustre clients were unaware of failover configurations until
4131 Severity : enhancement
4133 Description: user file locks should fail when not mounting with flock option
4134 Details : Set up an error-returning stub in ll_file_operations.lock field
4135 to prevent incorrect behaviour when client is mounted without
4136 flock option. Also, set up properly f_op->flock field for
4140 Frequency : always on ia64
4142 Description: "lfs df" loops on printing out MDS statfs information
4143 Details : The obd_ioctl_data was not initialized and in some systems
4144 this caused a failure during the ioctl that did not return
4145 an error. Initialize the struct and return an error on failure.
4148 Frequency : SLES 9 only
4150 Description: Error of copying files with lustre special EAs as root
4151 Details : Client side always return success for setxattr call for lustre
4152 special xattr (currently only "trusted.lov").
4155 Frequency : rarely on clusters with both ia64+i386 clients
4157 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4158 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4159 clients are doing concurrent writes to the same file it is
4160 possible that the ia64 clients may LASSERT because the OST
4161 extent locks are not PAGE_SIZE aligned. Ensure that grown
4162 locks are always aligned on the request boundary.
4165 Frequency : specific use, occasional
4167 Description: Overwriting in use executable truncates on-disk binary image
4168 Details : If one node attempts to overwrite an executable in use by
4169 another node, we now correctly return ETXTBSY instead of
4170 truncating the file.
4172 Severity : enhancement
4174 Description: Async OSC create to avoid the blocking unnecessarily.
4175 Details : If an OST has no remaining object, system will block on the creating
4176 when need to create a new object on this OST. Now, ways use
4177 pre-created objects when available, instead of blocking on an
4178 empty osc while others are not empty. If we must block, we block
4179 for the shortest possible period of time.
4184 Description: chmod on Lustre root is propagated to other clients
4185 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4186 invalid by the follow_mount time.
4189 Frequency : liblustre clients only
4191 Description: Race in 'instant cancel' lock handling could lead to such locks
4192 never to be granted in case of SMP MDS
4193 Details : Do not destroy not yet granted but cbpending locks in
4197 Frequency : replay/resend of open
4199 Description: non null lock assetion failure in mds_intent_policy
4200 Details : Trying to replay/resend lockless open requests resulted in
4201 mds_open() returning 0 with no lock. Now it sets a flag if
4202 a lock is going to be returned.
4204 Severity : enhancement
4206 Description: Checksum enhancements
4207 Details : New checksum enhancements allow for resending RPCs that failed
4210 Severity : enhancement
4212 Description: Tunables on number of dirty pages in cacche
4213 Details : Allow to set limit on number of dirty pages cached.
4218 Description: client crash on unmount - lock still has references
4219 Details : In some error handling cases it was possible to leak a lock
4220 reference on a client while accessing a file. This was not
4221 harmful to the client during operation, but would cause the
4222 client to crash when the filesystem is unmounted.
4225 Frequency : specific case, rare
4227 Description: ETXTBSY on mds though file not in use by client
4228 Details : ETXTBSY is no longer incorrectly returned when attempting to
4229 chmod or chown a directory that the user previously tried to
4230 execute or a currently-executing binary.
4233 Frequency : extremely rare except on liblustre-based clients
4235 Description: Lustre space not freed when files are deleted
4236 Details : Clean up open-unlinked files after client eviction. Previously
4237 the unlink was skipped and the files remained as orphans.
4242 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4243 Details : In some cases it was possible to send a blocking callback to
4244 a client doing a glimpse, even though that client didn't get
4245 a lock granted. When the glimpse lock is cancelled on the OST
4246 the freed lock is left on the waiting list and corrupted the list.
4249 Frequency : all core dumps
4251 Description: Broke core dumps to lustre
4252 Details : Negative dentry may be unhashed if parent does not have UPDATE
4253 lock, but some callers, e.g. do_coredump, expect dentry to be
4254 hashed after successful create, hash it in ll_create_it.
4256 ------------------------------------------------------------------------------
4258 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4260 * Support for kernels:
4261 2.6.9-42.0.2.EL (RHEL 4)
4262 2.6.5-7.276 (SLES 9)
4263 2.4.21-40.EL (RHEL 3)
4264 2.6.12.6 vanilla (kernel.org)
4268 Frequency : always on RHEL 3
4270 Description: Number of open files grows over time
4271 Details : The number of open files grows over time, whether or not
4272 Lustre is started. This was due to a filp leak introduced
4273 by one of our kernel patches.
4275 ------------------------------------------------------------------------------
4277 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4279 * Support for kernels:
4280 2.6.9-42.EL (RHEL 4)
4281 2.6.5-7.276 (SLES 9)
4282 2.4.21-40.EL (RHEL 3)
4283 2.6.12.6 vanilla (kernel.org)
4288 Bugzilla : 5719, 9635, 9792, 9684
4289 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4290 Details : If a server is under heavy load and cannot reply to new
4291 connection requests before the client resends the (re)connect,
4292 the connection handling code can behave badly if two service
4293 threads are concurrently handing separate (re)connections from
4294 the same client. Add better locking to the connection handling
4295 code, and ensure that only a single connection will be processed
4296 for a given client UUID, even if the lock is dropped.
4298 Severity : enhancement
4300 Description: add TCP zero-copy support to kernel
4301 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4302 sends if the hardware supports scatter-gather and checksumming.
4303 This allows socklnd to do client-write and server-read more
4304 efficiently and reduce CPU utilization from skbuf copying.
4307 Frequency : only if NFS exporting from client
4309 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4310 Details : If setting attributes on a file created under NFS that had
4311 never been opened it would be possible to oops the client
4312 if the file had no objects.
4315 Frequency : always for liblustre
4317 Description: liblustre client does MDS+OSTs setattr RPC for each write
4318 Details : When doing a write from a liblustre client, the client
4319 incorrectly issued an RPC to the MDS and each OST the file was
4320 striped over in order to update the timestamps. When writing
4321 with small chunks and many clients this could overwhelm the MDS
4322 with RPCs. In all cases it would slow down the write because
4323 these RPCs are unnecessary.
4325 Severity : enhancement
4327 Description: allow number of MDS service threads to be changed at module load
4328 Details : It is now possible to change the number of MDS service threads
4329 running. Adding "options mds mds_num_threads={N}" to the MDS's
4330 /etc/modprobe.conf will set the number of threads for the next
4331 time Lustre is restarted (assuming the "mds" module is also
4332 reloaded at that time). The default number of threads will
4333 stay the same, 32 for most systems.
4338 Description: OST crash if filesystem is unformatted or corrupt
4339 Details : If an OST is started on a device that has never been formatted
4340 or if the filesystem is corrupt and cannot even mount then the
4341 error handling cleanup routines would dereference a NULL pointer.
4346 Description: NULL pointer deref in llap_from_page.
4347 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4348 incorrect) mapping if the page was truncated/reclaimed while it was
4349 searched for. Check for this condition and skip such pages when
4350 doing readahead. Introduce extra check to llap_from_page() to
4351 verify page->mapping->host is non-NULL (so page is not anonymous).
4354 Frequency : Sometimes when using sys_sendfile
4356 Description: "page not covered by a lock" warnings from ll_readpage
4357 Details : sendfile called ll_readpage without right page locks present.
4358 Now we introduced ll_file_sendfile that does necessary locking
4359 around call to generic_file_sendfile() much like we do in
4363 Frequency : with certain MDS communication failures at client mount time
4365 Description: NULL pointer deref after failed client mount
4366 Details : a client connection request may delayed by the network layer
4367 and not be sent until after the PTLRPC layer has timed out the
4368 request. If the client fails the mount immediately it will try
4369 to clean up before the network times out the request. Add a
4370 reference from the request import to the obd device and delay
4371 the cleanup until the network drops the request.
4374 Frequency : occasionally during client (re)connect
4376 Description: assertion failure during client (re)connect
4377 Details : processing a client connection request may be delayed by the
4378 client or server longer than the client connect timeout. This
4379 causes the client to resend the connection request. If the
4380 original connection request is replied in this interval, the
4381 client may trip an assertion failure in ptlrpc_connect_interpret()
4382 which thought it would be the only running connect process.
4385 Frequency : only with obd_echo servers and clients that are rebooted
4387 Description: kernel BUG accessing uninitialized data structure
4388 Details : When running an obd_echo server it did not start the ping_evictor
4389 thread, and when a client was evicted an uninitialized data
4390 structure was accessed. Start the ping_evictor in the RPC
4391 service startup instead of the OBD startup.
4393 Severity : enhancement
4394 Bugzilla : 10193 (patchless)
4395 Description: Remove dependency on various unexported kernel interfaces.
4396 Details : No longer need reparent_to_init, exit_mm, exit_files,
4397 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4400 Frequency : rare (only users of deprecated and unsupported LDAP config)
4402 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4403 Details : LDAP apparently contains 'lustreName' attributes instead of
4404 'name'. A simple remapping of the name is sufficient.
4407 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4409 Description: waiting_locks_callback trips kernel BUG if client is evicted
4410 Details : Running with the dump_on_timeout debug flag turned on makes
4411 it possible that the waiting_locks_callback() can try to dump
4412 the Lustre kernel debug logs from an interrupt handler. Defer
4413 this log dumping to the expired_lock_main() thread.
4415 Severity : enhancement
4417 Description: Support NFS exporting on 2.6 kernels.
4418 Details : Implement non-rawops metadata methods for NFS server to use without
4419 changing NFS server code.
4422 Frequency : very rare (synthetic metadata workload only)
4424 Description: two racing renames might cause an MDS thread to deadlock
4425 Details : Running the "racer" program may cause one MDS thread to rename
4426 a file from being the source of a rename to being the target of
4427 a rename at exactly the same time that another thread is doing
4428 so, and the second thread has already enqueued these locks after
4429 doing a lookup of the target and is trying to relock them in
4430 order. Ensure that we don't try to re-lock the same resource.
4433 Frequency : only very large systems with liblustre clients
4435 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4436 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4437 all OSTs in parallel. This allows the eviction of stale liblustre
4438 clients to proceed much faster than if they were done in series,
4439 and also offers similar improvements for other set_info RPCs.
4444 Description: excessive CPU usage during initial read phase on client
4445 Details : During the initial read phase on a client, it would agressively
4446 retry readahead on the file, consuming too much CPU and impacting
4447 performance (since 1.4.5.8). Improve the readahead algorithm
4448 to avoid this, and also improve some other common cases (read
4449 of small files in particular, where "small" is files smaller than
4450 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4455 Description: MDS crash when receiving packet with unknown intent.
4456 Details : Do not LBUG in unknown intent case, just return -EFAULT
4458 Severity : enhancement
4459 Bugzilla : 9293, 9385
4460 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4461 Details : Do not serialize getattr (non-intent version) and statfs.
4464 Frequency : occasional, when OST network is overloaded/intermittent
4466 Description: client evicted by OST after bulk IO timeout
4467 Details : If a client sends a bulk IO request (read or write) the OST
4468 may evict the client if it is unresposive to its data GET/PUT
4469 request. This is incorrect if the network is overloaded (takes
4470 too long to transfer the RPC data) or dropped the OST GET/PUT
4471 request. There is no need to evict the client at all, since
4472 the pinger and/or lock callbacks will handle this, and the
4473 client can restart the bulk request.
4476 Frequency : Always when mmapping file with no objects
4478 Description: client crashes when mmapping file with no objects
4479 Details : Check that we actually have objects in a file before doing any
4480 operations on objects in ll_vm_open, ll_vm_close and
4486 Description: Request leak when working with deleted CWD
4487 Details : Introduce advanced request refcount tracking for requests
4488 referenced from lustre intent.
4490 Severity : Enhancement
4492 Description: Cache open file handles on client.
4493 Details : MDS now will return special lock along with openhandle, if
4494 requested and client is allowed to hold openhandle, even if unused,
4495 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4496 closing files for every read/write openration.
4498 Severity : Enhancement
4500 Description: Cache open negative dentries on client when possible.
4501 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4502 negative dentries on lock revocation.
4507 Description: Remounting a client read-only wasn't possible with a zconf mount
4508 Details : It wasn't possible to remount a client read-only with llmount.
4510 Severity : enhancement
4511 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4512 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4513 MPICH is included to add Lustre-specific ADIO interfaces.
4514 This is based closely on the UFS ADIO layer and only differs
4515 in file creation, in order to allow the OST striping to be set.
4516 This is user-contributed code and not supported by CFS.
4521 Description: extended inode attributes (immutable, append-only) work improperly
4522 when 2.4 and 2.6 kernels are used on client/server or vice versa
4523 Details : Introduce kernel-independent values for these flags.
4525 Severity : enhancement
4528 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4529 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4530 granularity. Now, Fractional MB granularity is be supported,
4531 this is very useful for low memory system.
4533 Severity : enhancement
4535 Description: Getattr by fid
4536 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4537 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4538 have locks and avoids getattr rpc requests for it on every stat.
4541 Frequency : Always, for filesystems larger than 2TB
4543 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4544 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4545 filesystems larger than 2TB to be mounted with mballoc enabled.
4548 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4550 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4551 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4552 filesystems larger than 2TB is mounted then the kernel will
4553 silently corrupt the start of the filesystem. CONFIG_LBD is
4554 enabled for all CFS-supported kernels, but the possibility of
4555 this happening with a modified kernel config exists.
4557 Severity : enhancement
4559 Description: add client O_DIRECT support for 2.6 kernels
4560 Details : It is now possible to do O_DIRECT reads and writes to files
4561 in the Lustre client mountpoint on 2.6 kernel clients.
4563 Severity : enhancement
4565 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4566 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4567 parallel, not waiting for response from every OST before sending
4568 a rpc to the next OST.
4573 Description: setattr vs write race when updating file timestamps
4574 Details : Client processes that update a file timestamp into the past
4575 right after writing to the file (e.g. tar) it is possible that
4576 the updated file modification time can be reset to the current
4577 time due to a race between processing the setattr and write RPC.
4579 Severity : enhancement
4581 Description: Bring 'lfs find' closer in line with regular Linux find.
4582 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4583 -print0 options and obtains all the needed info through the lustre
4586 Severity : enhancement
4588 Description: support up to 1024 configured devices on one node
4589 Details : change obd_dev array from statically allocated to dynamically
4590 allocated structs as they are first used to reduce memory usage
4595 Description: Flush dirty partially truncated pages during truncate
4596 Details : Immediatelly flush partially truncated pages in filter_setattr,
4597 this way we completely avoid having any pages in page cache on OST
4598 and can retire ugly workarounds during writes to flush such pages.
4603 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4604 Details : protect inode from truncation within vfs_unlink() context
4605 just take a reference before calling vfs_unlink() and release it
4606 when parent's i_sem is free.
4611 Description: last_id value checked outside lock on OST caused LASSERT failure
4612 Details : If there were multiple MDS->OST object precreate requests in
4613 flight, it was possible that the OST's last object id was checked
4614 outside a lock and incorrectly tripped an assertion. Move checks
4615 inside locks, and discard old precreate requests.
4618 Frequency : always, if extents are used on OSTs
4620 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4621 Details : OSTs using the extents format would not zero the high 16 bits of
4622 the index physical block number. This is not a problem for any
4623 OST filesystems smaller than 16TB, and no kernels support ext3
4624 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4625 new/modified files) and can be fixed for existing filesystems
4626 with e2fsprogs-1.39-cfs1.
4631 Description: import connection selection may be incorrect if timer wraps
4632 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4633 import connections to be ignored if the 32-bit jiffies counter
4634 wraps. Use a 64-bit jiffies counter.
4637 Frequency : very large clusters immediately after boot
4639 Description: LNET request buffers exhausted under heavy short-term load
4640 Details : If a large number of client requests are generated on a service
4641 that has previously never seen so many requests it is possible
4642 that the request buffer growth cannot keep up with the spike in
4643 demand. Instead of dropping incoming requests, they are held in
4644 the LND until the RPC service can accept more requests.
4647 Frequency : Sometimes during replay
4649 Description: Assertion failure in ll_local_open after replay.
4650 Details : If replay happened on an open request reply before we were able
4651 to set replay handler, reply will become not swabbed tripping the
4652 assertion in ll_local_open. Now we set the handler right after
4653 recognising of open request
4656 Frequency : very rare
4658 Description: kernel reports "badness in vsnprintf"
4659 Details : Reading from the "recovery_status" /proc file in small chunks
4660 may cause a negative length in lprocfs_obd_rd_recovery_status()
4661 call to vsnprintf() (which is otherwise harmless). Exit early
4662 if there is no more space in the output buffer.
4664 Severity : enhancement
4666 Description: clear OBD RPC statistics by writing to them
4667 Details : It is now possible to clear the OBD RPC statistics by writing
4668 to the "stats" file.
4673 Description: Client mtime is not the same on different clients after utimes
4674 Details : In some cases, the client was using the utimes() syscall on
4675 a file cached on another node. The clients now validate the
4676 ctime from the MDS + OSTs to determine which one is right.
4681 Description: Inability to activate failout mode
4682 Details : lconf script incorrectly assumed that in pythong string's numeric
4683 value is used in comparisons.
4686 Frequency : always with multiple stripes per file
4688 Description: Inefficient object allocation for mutli-stripe files
4689 Details : When selecting which OSTs to stripe files over, for files with
4690 a stripe count that divides evenly into the number of OSTs,
4691 the MDS is always picking the same starting OST for each file.
4692 Return the OST selection heuristic to the original design.
4697 Description: mount failures may take full timeout to return an error
4698 Details : Under some heavy load conditions it is possible that a
4699 failed mount can wait for the full obd_timeout interval,
4700 possibly several minutes, before reporting an error.
4701 Instead return an error as soon as the status is known.
4703 Frequency : quota enabled and large files being deleted
4705 Description: releasing more than 4GB of quota at once hangs OST
4706 Details : If a user deletes more than 4GB of files on a single OST it
4707 will cause the OST to spin in an infinite loop. Release
4708 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4713 Description: statfs data retrieved from /proc may be stale or zero
4714 Details : When reading per-device statfs data from /proc, in the
4715 {kbytes,files}_{total,free,avail} files, it may appear
4716 as zero or be out of date.
4719 Frequency : systems with MD RAID1 external journal devices
4721 Description: lconf's call to blkid is confused by RAID1 journal devices
4722 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4723 of returning all block devices that match the journal UUID.
4726 Frequency : always, for aggregate stripe size over 4GB
4728 Description: assertion fails when trying to use 4GB stripe size
4729 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4730 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4733 Frequency : always on ppc64
4735 Description: the first write on an ext3 filesystem with mballoc got stuck
4736 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4737 perform endianness conversion.
4739 ------------------------------------------------------------------------------
4741 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4743 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4744 INTEROPERATE with older versions automatically. Please read the
4745 user documentation before upgrading any part of a live system.
4746 * WARNING: Lustre networking configuration changes are required with
4747 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4750 * Support for kernels:
4751 2.6.9-22.0.2.EL (RHEL 4)
4752 2.6.5-7.244 (SLES 9)
4753 2.6.12.6 vanilla (kernel.org)
4756 Severity : enhancement
4757 Bugzilla : 7981/8208
4758 Description: Introduced Lustre Networking (LNET)
4759 Details : LNET is new networking infrastructure for Lustre, it includes
4760 a reorganized network configuration mode (see the user
4761 documentation for full details) as well as support for routing
4762 between different network fabrics. Lustre Networking Devices
4763 (LNDs) for the supported network fabrics have also been
4764 created for this new infrastructure.
4766 Severity : enhancement
4767 Description: Introduced Access control lists
4768 Details : clients can set ACLs on files and directories in order to have
4769 more fine-grained permissions than the standard Unix UGO+RWX.
4770 The MDS must be started with the "-o acl" mount option.
4772 Severity : enhancement
4773 Description: Introduced filesystem quotas
4774 Details : Administrators may now establish per-user quotas on the
4777 Severity : enhancement
4779 Description: Configuration change for the XT3
4780 The PTLLND is now used to run Lustre over Portals on the XT3
4781 The configure option(s) --with-cray-portals are no longer used.
4782 Rather --with-portals=<path-to-portals-includes> is used to
4783 enable building on the XT3. In addition to enable XT3 specific
4784 features the option --enable-cray-xt3 must be used.
4789 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4790 Details : A race between allocating a new llcd and re-getting the llcd_lock
4791 allowed another thread to grab newly-allocated llcd.
4793 Severity : enhancement
4795 Description: 2.6 OST async journal commit and locking fix to improve performance
4796 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4797 2.6 kernels are now async as they already were in 2.4 kernels so
4798 that they can commit concurrently with the network bulk transfer.
4799 For block-allocated files the filter allocation semaphore is held
4800 to avoid filesystem fragmentation during allocation. BKL lock
4801 removed for 2.6 xattr operations where it is no longer needed.
4806 Description: lconf incorrectly determined whether two IP networks could talk
4807 Details : In some more complicated routing and multiple-network
4808 configurations, lconf will avoid trying to make a network
4809 connection to a disjoint part of the IP space. It was doing the
4810 math incorrectly for one set of cases.
4815 Description: Fix for potential infinite loop processing records in an llog.
4816 Details : If an llog record is corrupted/zeroed, it is possible to loop
4817 forever in llog_process(). Validate the llog record length
4818 and skip the remainder of the block on error.
4821 Frequency : occasional (liblustre only)
4823 Description: liblustre could not open files whose last component is a symlink
4824 Details : sysio_path_walk() would incorrectly pass the open intent to
4825 intermediate path components.
4828 Frequency : rare (liblustre only with non-standard tuning)
4829 Bugzilla : 7201 (7350)
4830 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4831 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4832 after it has been released, only do this for referenced locks
4834 Severity : enhancement
4836 Description: specifying an (invalid) directory default stripe_size of -1
4837 would reset the directory default striping
4838 Details : stripe_size -1 was used internally to signal directory stripe
4839 removal, now use "all default" to signal dir stripe removal
4840 as a directory striping of "all default" is not useful
4843 Frequency : common for large clusters running liblustre clients
4845 Description: doing an ls when liblustre clients are running is slow
4846 Details : sending a glimpse AST to a liblustre client waits for every AST
4847 to time out, as liblustre clients will not respond. Since they
4848 cannot cache data we refresh the OST lock LVB from disk instead.
4850 Severity : enhancement
4852 Description: doing an ls at the same time as file IO can be slow
4853 Details : enqueue and other "small" requests can be blocked behind many
4854 large IO requests. Create a new OST IO portal for non-IO
4855 requests so they can be processed faster.
4858 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4860 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4861 created file with mode 000
4862 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4863 allow an ftruncate() as a non-root user to a file with mode 000.
4864 Lustre now respects this flag to disable mode checks when
4865 truncating a file owned by the user
4868 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4871 Description: Revoking locks from clients that went dead or catatonic might take
4873 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4874 cancellation of such locks instant on servers without waiting for
4875 any reply from clients. Clients drops these locks when cancel
4876 notification from server is received without replying.
4879 Frequency : liblustre-only, when liblustre client dies or becomes busy
4881 Description: Doing ls on Linux clients can take a long time with active
4883 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4884 granting such locks to it in the first place if possible. Locks
4885 are taken by proxy on the OST during the read or write and
4886 dropped immediately afterward. Add connect flags handling, do
4887 not grant locks to liblustre clients for glimpse ASTs.
4889 Severity : enhancement
4891 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4892 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4893 sequential reads and behaves badly with applications doing random
4894 reads. Improve it by reading ahead at least read region, and
4895 avoiding excessive large RPC for small reads.
4897 Severity : enhancement
4899 Description: Creating more than 1000 files for a single job may cause a load
4900 imbalance on the OSTs if there are also a large number of OSTs.
4901 Details : qos_prep_create() uses an OST index reseed value that is an
4902 even multiple of the number of available OSTs so that if the
4903 reseed happens in the middle of the object allocation it will
4904 still utilize the OSTs as uniformly as possible.
4909 Description: OST or MDS may oops in ping_evictor_main()
4910 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4911 but doesn't restart at beginning of obd_exports_timed list
4914 Severity : enhancement
4916 Description: improve by-nid export eviction on the MDS and OST
4917 Details : allow multiple exports with the same NID to be evicted at one
4918 time without re-searching the exports list.
4921 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4923 Description: MDS may oops in groups_free()
4924 Details : in rare race conditions a newly allocated group_info struct is
4925 freed again, and this can be NULL. The 2.4 compatibility code
4926 for groups_free() checked for a NULL pointer, but 2.6 did not.
4929 Frequency : common for liblustre clients doing little filesystem IO
4930 Bugzilla : 9352, 7313
4931 Description: server may evict liblustre clients accessing contended locks
4932 Details : if a client is granted a lock or receives a completion AST
4933 with a blocking AST already set it would not reply to the AST
4934 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4937 Frequency : lfs setstripe, only systems with more than 160 OSTs
4939 Description: unable to set striping with a starting offset beyond OST 160
4940 Details : llapi_create_file() incorrectly limited the starting stripe
4941 index to the maximum single-file stripe count.
4944 Frequency : LDAP users only
4946 Description: lconf did not handle in-kernel recovery with LDAP properly
4947 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4949 Severity : enhancement
4951 Description: bind OST threads to NUMA nodes to improve performance
4952 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4953 node and do their allocations there to localize memory access
4955 Severity : enhancement
4957 Description: llmount can determine client NID directly from Myrinet (GM)
4958 Details : the client NID code from gmnalnid was moved directly into
4959 llmount, removing the need to use this or specifying the
4960 client NID explicitly when mounting GM clients with zeroconf
4963 Frequency : if client is started with down MDS
4965 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4966 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4967 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4968 ptlrpc_import_delayed_req() because we didn't check if the
4969 request was interrupted, and we also didn't break out of the
4970 event loop if there was no timeout
4975 Description: data loss during non-page-aligned writes to a single file from
4976 both multiple nodes and multiple threads on one node at same time
4977 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4978 inconsistency led to false short-reads, that were cached and later
4979 used by ->prepare_write() to fill in partially written page,
4980 leading to data loss.
4983 Frequency : always, if lconf --abort_recovery used
4985 Description: lconf --abort_recovery fails with 'Operation not supported'
4986 Details : lconf was attempting to abort recovery on the MDT device and not
4989 Severity : enhancement
4991 Description: remove cleanup logs
4992 Details : replace lconf-generated cleanup logs with lustre internal
4993 cleanup routines. Eliminates the need for client-cleanup and
4996 Severity : enhancement
4998 Description: add support for EAs (user and system) on lustre filesystems
4999 Details : it is now possible to store extended attributes in the Lustre
5000 client filesystem, and with the user_xattr mount option it
5001 is possible to allow users to store EAs on their files also
5003 Severity : enhancement
5005 Description: Add possibility (config option) to show minimal available OST free
5007 Details : When compiled with --enable-mindf configure option, statfs(2)
5008 (and so, df) will return least minimal free space available from
5009 all OSTs as amount of free space on FS, instead of summary of
5010 free spaces of all OSTs.
5012 Severity : enhancement
5014 Description: do not expand extent locks acquired on OST-side
5015 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5016 by server: they are not cached anyway.
5019 Frequency : when mmap is used/binaries executed from Lustre
5021 Description: Unmmap pages before throwing them away from read cache.
5022 Details : llap_shrink cache now attempts to unmap pages before discarding
5023 them (if unmapping failed - do not discard). SLES9 kernel has
5024 extra checks that trigger if this unmapping is not done first.
5029 Description: lconf didn't resolve symlinks before checking to see whether a
5030 given mountpoint was already in use
5033 Frequency : when migrating failover services
5034 Bugzilla : 6395, 9514
5035 Description: When migrating a subset of services from a node (e.g. failback
5036 from a failover service node) the remaining services would
5037 time out and evict clients.
5038 Details : lconf --force (implied by --failover) sets the global obd_timeout
5039 to 5 seconds in order to quickly disconnect, but this caused
5040 other RPCs to time out too quickly. Do not change the global
5041 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5043 Severity : enhancement
5044 Frequency : if MDS is started with down OST
5045 Bugzilla : 9439,5706
5046 Description: Allow startup/shutdown of an MDS without depending on the
5047 availability of the OSTs.
5048 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5049 Add appropriate locking and lov-osc refcounts for safe
5050 cleaning. Add osc abort_inflight calls in case the
5051 synchronize never started.
5054 Frequency : occasional (Cray XT3 only)
5056 Description: root not authorized to access files in CRAY_PORTALS environment
5057 Details : The client process capabilities were not honoured on the MDS in
5058 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5059 been accessed by an authorized user then root was able to access
5060 the file on the local client also. The root user capabilities
5061 are now allowed on the MDS, as this environment has secure UID.
5064 Frequency : occasional
5066 Description: ldiskfs "too long searching" message happens too often
5067 Details : A debugging message (otherwise harmless) prints too often on
5068 the OST console. This has been reduced to only happen when
5069 there are fragmentation problems on the filesystem.
5074 Description: Division by zero in statfs when all OSCs are inactive
5075 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5076 lov_statfs divides by value returned by lov_get_stripecnt().
5080 Bugzilla : 9489, 3273
5081 Description: First write from each client to each OST was only 4kB in size,
5082 to initialize client writeback cache, which caused sub-optimal
5083 RPCs and poor layout on disk for the first written file.
5084 Details : Clients now request an initial cache grant at (re)connect time
5085 and so that they can start streaming writes to the cache right
5086 away and always do full-sized RPCs if there is enough data.
5087 If the OST is rebooted the client also re-establishes its grant
5088 so that client cached writes will be honoured under the grant.
5093 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5094 Details : Now I/O RPCs go to different portal number and (presumably) fast
5095 lock requests (and glimses) and other RPCs get their own service
5096 threads pool that should be able to service those RPCs
5099 Severity : enhancement
5101 Description: Ability to exchange lustre version between client and servers and
5102 issue warnings at client side if client is too old. Also for
5103 liblustre clients there is ability to refuse connection of too old
5105 Details : New 'version' field is added to connect data structure that is
5106 filled with version info. That info is later checked by server and
5110 Frequency : rare, liblustre only.
5111 Bugzilla : 9296, 9581
5112 Description: Two simultaneous writes from liblustre at offset within same page
5113 might proceed at the same time overwriting eachother with stale
5115 Details : I/O lock withing llu_file_prwv was released too early, before data
5116 actually was hitting the wire. Extended lock-holding time until
5117 server acknowledges receiving data.
5120 Frequency : extremely rare. Never observed in practice.
5122 Description: avoid generating lustre_handle cookie of 0.
5123 Details : class_handle_hash() generates handle cookies by incrementing
5124 global counter, and can hit 0 occasionaly (this is unlikely, but
5125 not impossible, because initial value of cookie counter is
5126 selected randonly). Value of 0 is used as a sentinel meaning
5127 "unassigned handle" --- avoid it. Also coalesce two critical
5128 sections in this function into one.
5130 Severity : enhancement
5132 Description: allow liblustre clients to delegate truncate locking to OST
5133 Details : To avoid overhead of locking, liblustre client instructs OST to
5134 take extent lock in ost_punch() on client's behalf. New connection
5135 flag is added to handle backward compatibility.
5137 Severity : enhancement
5138 Bugzilla : 4928, 7341, 9758
5139 Description: allow number of OST service threads to be specified
5140 Details : a module parameter allows the number of OST service threads
5141 to be specified via "options ost ost_num_threads={N}" in the
5142 OSS's /etc/modules.conf or /etc/modprobe.conf.
5146 Bugzilla : 6146, 9635, 9895
5147 Description: servers crash with bad pointer in target_handle_connect()
5148 Details : In rare cases when a client is reconnecting it was possible that
5149 the connection request was the last reference for that export.
5150 We would temporarily drop the export reference and get a new
5151 one, but this may have been the last reference and the export
5152 was just destroyed. Get new reference before dropping old one.
5154 Severity : enhancement
5155 Frequency : if client is started with failover MDS
5157 Description: Allow multiple MDS hostnames in the mount command
5158 Details : Try to read the configuration from all specified MDS
5159 hostnames during a client mount in case the "primary"
5162 Severity : enhancement
5164 Description: Stop sending data to evicted clients as soon as possible.
5165 Details : Check if the client we are about to send or are sending data to
5166 was evicted already. (Check is done every second of waiting,
5167 for which l_wait_event interface was extended to allow checking
5168 of exit condition at specified intervals).
5171 Frequency : rare, normally only when NFS exporting is done from client
5173 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5175 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5178 Frequency : always, if srand() is called before liblustre initialization
5180 Description: Liblustre uses system PRNG disturbing its usage by user application
5181 Details : Introduce internal to lustre fast and high-quality PRNG for
5182 lustre usage and make liblustre and some other places in generic
5183 lustre code to use it.
5185 Severity : enhancement
5186 Bugzilla : 9477, 9557, 9870
5187 Description: Verify that the MDS configuration logs are updated when xml is
5188 Details : Check if the .xml configuration logs are newer than the config
5189 logs stored on the MDS and report an error if this is the case.
5190 Request --write-conf, or allow starting with --old_conf.
5192 Severity : enhancement
5194 Description: Handle symlinks in the path when checking if Lustre is mounted.
5195 Details : Resolve intermediate symlinks when checking if a client has
5196 mounted a filesystem to avoid duplicate client mounts.
5201 Description: lconf can hit an error exception but still return success.
5202 Details : The lconf command catches the Command error exception at the top
5203 level script context and will exit with the associated exit
5204 status, but doesn't ensure that this exit status is non-zero.
5209 Description: failure of ptlrpc thread startup can cause oops
5210 Details : Starting a ptlrpc service thread can fail if there are a large
5211 number of threads or the server memory is very fragmented.
5212 Handle this without oopsing.
5215 Frequency : always, only if liblustre and non-default acceptor port was used
5217 Description: liblustre cannot connect to servers with non-default acceptor port
5218 Details : tcpnal_set_default_params() was not called and was therefore
5219 ignoring the environment varaible TCPNAL_PORT, as well as other
5220 TCPNAL_ environment variables
5225 Description: two objects could be created on the same OST for a single file
5226 Details : If an OST is down, in some cases it was possible to create two
5227 objects on a single OST for a single file. No problems other
5228 than potential performance impact and spurious error messages.
5232 Bugzilla : 5681, 9562
5233 Description: Client may oops in ll_unhash_aliases
5234 Details : Client dcache may become inconsistent in race condition.
5235 In some cases "getcwd" can fail if the current directory is
5241 Description: Inode refcounting problems in NFS export code
5242 Details : link_raw functions used to call d_instantiate without obtaining
5243 extra inode reference first.
5247 Bugzilla : 9942, 9903
5248 Description: Referencing freed requests leading to crash, memleaks with NFS.
5249 Details : We used to require that call to ll_revalidate_it was always
5250 followed by ll_lookup_it. Also with revalidate_special() it is
5251 possible to call ll_revalidate_it() twice for the same dentry
5252 even if first occurence returned success. This fix changes semantic
5253 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5254 reference on a request referred from the intent.
5255 ll_intent_release() then releases such a request.
5258 Frequency : rare, normally benchmark loads only
5260 Description: unlinked inodes were kept in memory on the client
5261 Details : If a client is repeatedly creating and unlinking files it
5262 can accumulate a lot of stale inodes in the inode slab cache.
5263 If there is no other client load running this can cause the
5264 client node to run out of memory. Instead flush old inodes
5265 from client cache that have the same inode number as a new inode.
5268 Frequency : SLES9 2.6.5 kernel and long filenames only
5269 Bugzilla : 9969, 10379
5270 Description: utime reports stale NFS file handle
5271 Details : SLES9 uses out-of-dentry names in some cases, which confused
5272 the lustre dentry revalidation. Change it to always use the
5276 Frequency : rare, unless heavy write-truncate concurrency is continuous
5277 Bugzilla : 4180, 6984, 7171, 9963, 9331
5278 Description: OST becomes very slow and/or deadlocked during object unlink
5279 Details : filter_destroy() was holding onto the parent directory lock
5280 while truncating+unlinking objects. For very large objects this
5281 may block other threads for a long time and slow overall OST
5282 responsiveness. It may also be possible to get a lock ordering
5283 deadlock in this case, or run out of journal credits because of
5284 the combined truncate+unlink. Solution is to do object truncate
5285 first in one transaction without parent lock, and then do the
5286 final unlink in a new transaction with the parent lock. This
5287 reduces the lock hold time dramatically.
5290 Frequency : rare, 2.4 kernels only
5292 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5293 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5294 before unmounting their filesystem in order to avoid deadlock.
5295 The kernel_locked() function in 2.4 kernels only checks whether
5296 the kernel lock is held, not whether it is this process that is
5297 holding it as 2.6 kernels do.
5302 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5303 Details : The client ptlrpc code may be trying to reconnect to a down
5304 server before a previous connection attempt has timed out.
5305 Increase the reconnect interval to be longer than the connection
5306 timeout interval to avoid sending duplicate connections to
5310 Frequency : echo_client brw_test command
5312 Description: fix echo_client to work with OST preallocated code
5313 Details : OST preallocation code (5137) didn't take echo_client IO path
5314 into account: echo_client calls filter methods outside of any
5315 OST thread and, hence, there is no per-thread preallocated
5316 pages and buffers to use. Solution: hijack pga pages for IO. As
5317 a byproduct, this avoids unnecessary data copying.
5321 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5322 Description: Client can oops in mdc_commit_close() after open replay
5323 Details : It was possible for the MDS to return an open request with no
5324 transaction number in mds_finish_transno() if the client was
5325 evicted, but without actually returning an error. Clients
5326 would later try to replay that open and may trip an assertion
5327 Simplify the client close codepath, and always return an error
5328 from the MDS in case the open is not successful.
5331 Frequency : rare, 2.6 OSTs only
5333 Description: OST may deadlock under high load on fragmented files
5334 Details : If there was a heavy load and highly-fragmented OST filesystems
5335 it was possible to have all the OST threads deadlock waiting on
5336 allocation of biovecs, because the biovecs were not released
5337 until the entire RPC IO was completed. Instead, release biovecs
5338 as soon as they are complete to ensure forward IO progress.
5340 Severity : enhancement
5342 Description: Support for specifying external journal device at mount
5343 Details : If an OST or MDS device is formatted with an external journal
5344 device, this device major/minor is stored in the ext3 superblock
5345 and may not be valid for failover. Allow detecting and
5346 specifying the external journal at mount time.
5351 Description: Mounting an MDS with pending unlinked files may cause oops
5352 Details : target_finish_recovery() calls mds_postrecov() which returned
5353 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5354 considers this an error and immediately begins cleaning up the
5355 lov, just after starting the mds_lov process
5357 Severity : enhancement
5359 Description: Implement 'lfs df' to report actual free space on per-OST basis
5360 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5361 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5362 usage of objects; '-h' to report in human readable format.
5364 ------------------------------------------------------------------------------
5366 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5373 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5374 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5375 at startup during orphan unlinking. Instead, skip buddy update
5376 before setup, as it will be regenerated after recovery is complete.
5381 Description: If an OST is inactive, its locks might reference stale inodes.
5382 Details : lov_change_cbdata() must iterate over all namespaces, even if
5383 they are inactive to clear inode references from the lock.
5385 Severity : enhancement
5386 Frequency : occasional, if non-standard max_dirty_mb used
5388 Description: Client will block write RPCs if not enough grant
5389 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5390 then the client will block writes while waiting for another RPC
5391 to complete instead of consuming its dirty limit. With change
5392 we get improved performance when max_dirty_mb is small.
5394 Severity : enhancement
5395 Bugzilla : 3389, 6253
5396 Description: Add support for supplementary groups on the MDS.
5397 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5398 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5399 lookups for user supplementary groups into a cache.
5403 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5404 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5405 for liblustre O_EXCL is enforced by MDS.
5409 Description: readdir never returns NULL in liblustre.
5410 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5411 dentry in struct dirent.
5415 Frequency : liblustre only, depends on application IO pattern
5416 Description: liblustre clients evicted if not contacting servers
5417 Details : Don't put liblustre clients into the ping_evictor list, so
5418 they will not be evicted by the pinger ever.
5420 Severity : enhancement
5422 Description: Add ability to evict clients by NID from MDS.
5423 Details : By echoing "nid:$NID" string into
5424 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5425 $NID would be instantly evicted from this MDS and from all active
5426 OSTs connected to it.
5430 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5431 Details : lookup_it_finish() used to query file size from OSTs that was not
5436 Description: service threads change working directory to that of init
5437 Details : Starting lustre service threads may pin the working directory
5438 of the parent thread, making that filesystem busy. Threads
5439 now change to the working directory of init to avoid this.
5443 Frequency : during shutdown only
5444 Description: shutdown with a failed MDS or OST can cause unmount to hang
5445 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5450 Frequency : 2.6 only
5451 Description: chmod/chown may include an extra supplementary group
5452 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5453 supplementary group and if none is specified this is used.
5456 Bugzilla : 5479 (6816)
5458 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5459 Details : If lookup is in progress on a file that is unlinked we might try
5460 to revalidate the inode and fail in revalidate after lookup is
5461 complete and ll_file_open() enqueues the open again but
5462 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5465 Frequency : always, if lconf --abort_recovery used
5467 Description: lconf --abort_recovery fails with 'Operation not supported'
5468 Details : lconf was attempting to abort recovery on the MDT device and not
5471 ------------------------------------------------------------------------------
5473 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5478 Frequency : rare (only unsupported configurations with a node running as an
5480 Bugzilla : 6514, 5137
5481 Description: Mounting a Lustre file system on a node running as an OST could
5483 Details : OSTs now preallocates memory needed to write out data at
5484 startup, instead of when needed, to avoid having to
5485 allocate memory in possibly low memory situations.
5486 Specifically, if the file system is mounted on on OST,
5487 memory pressure could force it to try to write out data,
5488 which it needed to allocate memory to do. Due to the low
5489 memory, it would be unable to do so and the node would
5490 become unresponsive.
5492 Severity : enhancement
5494 Description: Addition of lconf --service command line option
5495 Details : lconf now accepts a '--service <arg>' option, which is
5496 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5498 Severity : enhancement
5500 Description: Failover mode is now the default for OSTs.
5501 Details : By default, OSTs will now run in failover mode. To return to
5502 the old behaviour, add '--failout' to the lmc line for OSTs.
5504 Severity : enhancement
5506 Description: Health checks are now provided for MDS and OSTs
5507 Details : Additional detailed health check information on MSD and OSTs
5508 is now provided through the procfs health_check value.
5511 Frequency : occasional, depends on IO load
5513 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5514 after numerous create/delete cycles
5515 Details : The ext3 inode allocation policy would not allocate new inodes
5516 very well on the OSTs because there are no new directories
5517 being created. Instead we look for groups with free space if
5518 the parent directories are nearly full.
5523 Description: Network or server problems during mount may cause partially
5524 mounted clients instead of returning an error.
5525 Details : The config llog parsing code may overwrite the error return
5526 code during mount error handling, returning success instead
5532 Description: MDS can fail to allocate large reply buffers
5533 Details : After long uptimes the MDS can fail to allocate large reply
5534 buffers (e.g. zconf client mount config records) due to memory
5535 fragmentation or consumption by the buffer cache. Preallocate
5536 some large reply buffers so that these replies can be sent even
5537 under memory pressure.
5541 Frequency : rare (liblustre)
5542 Description: fsx running with liblustre complained that using truncate() to
5543 extend the file doesn't work. This patch corrects that issue.
5544 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5545 ATTR_SIZE and lsm use in llu_setattr_raw.
5549 Frequency : rare, only 2.6 kernels
5550 Description: Unusual file access patterns on the MDS may result in inode
5551 data being lost in very rare circumstances.
5552 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5553 code in the 2.6 kernel caused the inode and/or EA data not to
5554 be read from disk, causing single-file corruption.
5558 Frequency : rare, only 2.6 filesystems using extents
5559 Description: Heavy concurrent write and delete load may cause data corruption.
5560 Details : It was possible under high-load situations to have an extent
5561 metadata block in the block device cache from a just-unlinked
5562 file overwrite a newly-allocated data block. We now unmap any
5563 metadata buffers that alias just-allocated data blocks.
5567 Frequency : filesystems with default stripe_count larger than 77
5568 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5569 Details : lconf specifies an inode size of 4096 bytes when the default
5570 stripe_count is larger than 77. This conflicts with the default
5571 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5572 this case to avoid pinning too much memory for large EAs.
5574 ------------------------------------------------------------------------------
5576 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5581 Frequency : rare (extremely heavy IO load with hundreds of clients)
5583 Description: Client is evicted, gets IO error writing to file
5584 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5585 caused clients to be evicted for AST timeouts. The fixes in
5586 bug 5192 mean we no longer need to have such short AST timeouts
5587 so ldlm_timeout has been increased.
5590 Frequency : occasional during --force or --failover shutdown under load
5591 Bugzilla : 5949, 4834
5592 Description: Server oops/LBUG if stopped with --force or --failover under load
5593 Details : a collection of import/export refcount and cleanup ordering
5594 issues fixed for safer force cleanup
5597 Frequency : only filesystems larger than 120 OSTs
5598 Bugzilla : 5990, 6223
5599 Description: lfs getstripe would oops on a very large filesystem
5600 Details : lov_getconfig used kfree on vmalloc'd memory
5603 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5604 Bugzilla : 6242, 6243
5605 Description: reading from files that had been truncated to a non-zero size
5606 but never opened returned no data
5607 Details : ll_file_read() reads zeros from no-object files to EOF
5612 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5613 Details : There's little harm in aborting MDS/OSS recovery and letting it
5614 try again, so I removed the LASSERT and return an error instead.
5616 Severity : enhancement
5618 Description: New debugging infrastructure for tracking down data corruption
5619 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5620 (b) cover more of the client-side code path, and (c) try to narrow
5621 down where problems occurred
5625 Bugzilla : 3819, 4364, 4397, 6313
5626 Description: Racing close and eviction MDS could cause assertion in mds_close
5627 Details : It was possible to get multiple mfd references during close and
5628 client eviction, leading to one thread referencing a freed mfd.
5630 Severity: : enhancement
5631 Bugzilla : 3262, 6359
5632 Description: Attempts to reconnect to servers are now more aggressive.
5633 Details : This builds on the enhanced upcall-less recovery that was added
5634 in 1.4.2. When trying to reconnect to servers, clients will
5635 now try each server in the failover group every 10 seconds. By
5636 default, clients would previously try one server every 25 seconds.
5641 Description: After recovery, certain operations trigger a failed
5642 assertion on a client.
5643 Details : Failing over an mds, using lconf -d --failover, while a
5644 client was doing a readdir() call would cause the client to
5645 LBUG after recovery completed and the readdir() was resent.
5647 Severity : enhancement
5649 Description: Default groups are now added by lconf
5650 Details : You can now run lconf --group <servicename> without having to
5651 manually add groups with lmc.
5654 Frequency : occasional
5656 Description: Nodes with an elan id of 0 trigger a failed assertion
5659 Frequency : always when accessing e.g. tty/console device nodes
5661 Description: tty and some other devices nodes cannot be used on lustre
5662 Details : file's private_data field is used by device data and lustre
5663 values in there got lost. New field was added to struct file to
5664 store fs-specific private data.
5667 Frequency : when exporting Lustre via NFS
5669 Description: NFSD failed occasionally when looking up a path component
5670 Details : NFSD is looking up ".." which was broken in ext3 directories
5671 that had grown large enough to become hashed.
5674 Frequency : Clusters with multiple interfaces not on the same subnet
5676 Description: Nodes will repeatedly try to reconnect to an interface which it
5677 cannot reach and report an error to the log.
5678 Details : Extra peer list entries will be created by lconf with some peers
5679 unreachable. lconf now validates the peer before adding it.
5682 Frequency : Only if a default stripe is set on the filesystem root.
5684 Description: Setting a default stripe on the filesystem root prevented the
5685 filesystem from being remounted.
5686 Details : The client was sending extra request flags in the root getattr
5687 request and did not allocate a reply buffer for the dir EA.
5690 Frequency : occasional, higher if lots of files are accessed by one client
5691 Bugzilla : 6159, 6097
5692 Description: Client trips assertion regarding lsm mismatch/magic
5693 Details : While revalidating inodes the VFS looks up inodes with ifind()
5694 and in rare cases can find an inode that is being freed.
5695 The ll_test_inode() code will free the lsm during ifind()
5696 when it finds an existing inode and then the VFS later attaches
5697 this free lsm to a new inode.
5701 Bugzilla : 6422, 7030
5702 Description: MDS deadlock between mkdir and client eviction
5703 Details : Creating a new file via mkdir or mknod (starting a transaction
5704 and getting the ns lock) can deadlock with client eviction
5705 (gets ns lock and trying to finish a synchronous transaction).
5708 Frequency : occasional
5709 Description: While starting a server, the fsfilt_ext3 module could not be
5711 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5712 kernels. Previously, lconf would still use the ext3 name
5713 when trying to load modules. Now, it will correctly use
5714 ext3 on 2.4 and ldiskfs on 2.6.
5716 Severity : enhancement
5717 Description: The default stripe count has been changed to 1
5718 Details : The interpretation of the default stripe count (0, to lfs
5719 or lmc) has been changed to mean striping across a single
5720 OST, rather than all available. For general usage we have
5721 found a stripe count of 1 or 2 works best.
5723 Severity : enhancement
5724 Description: Add support for compiling against Cray portals.
5725 Details : Conditional compiling for some areas that are different
5729 Frequency : occasional
5730 Bugzilla : 6409, 6834
5731 Description: Creating files with an explicit stripe count may lead to
5732 a failed assertion on the MDS
5733 Details : If some OSTs are full or unavailable, creating files may
5734 trigger a failed assertion on the MDS. Now, Lustre will
5735 try to use other servers or return an error to the
5739 Frequency : occasional
5741 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5742 caused lock timeout during readahead (since 1.4.2).
5743 Details : Processes doing readahead might match a lock that hasn't been
5744 granted yet if there are overlapping and conflicting lock
5745 requests. The readahead process waits on ungranted lock
5746 (original lock is CBPENDING), while OST waits for that process
5747 to cancel CBPENDING read lock and eventually evicts client.
5749 Severity : enhancement
5751 Description: Initial enabling of flock support for clients
5752 Details : Implements fcntl advisory locking and file status functions.
5753 This feature is provided as an optional mount flag (default
5754 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5755 locking are implemented yet, and those that are are not guaranteed
5756 to be completely correct in production environments.
5757 mount -t lustre -o [flock|noflock] ...
5760 Frequency : occasional
5762 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5763 assertion in the ext3 JBD (journaling) layer.
5764 Details : The b_committed_data struct is protected by the big kernel lock
5765 in 2.4 kernels, serializing journal_commit_transaction() and
5766 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5767 access to this struct. In 2.6 kernels there is finer grained
5768 locking to improve SMP performance of the JBD layer.
5772 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5777 Description: merge and cleanup kernel patches.
5780 -----------------------------------------------------------------------------
5782 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5784 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5785 versions of Lustre. Please update all servers and clients to
5786 version 1.4.2 or later at the same time. You must also run
5787 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5788 to update the configuration logs.
5790 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5791 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5792 - swab llog records of type '0' so we get proper header size/idx (5861)
5793 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5794 - fix rename of one directory over another leaking an inode (5953)
5795 - avoid SetPageDirty on 2.6 (5981)
5796 - don't re-add just-being-destroyed locks to the waiting list (5653)
5797 - when creating new directories, inherit the parent's custom
5798 striping settings if present parent (3048)
5799 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5800 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5801 - don't hold client locks on temporary worklist from l_lru (5666)
5802 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5803 - automatically evict dead clients (5921)
5804 - Update file size properly in create+truncate+fstat case (6196)
5805 - Do not unhash mountpoint dentries, do not allow removal of
5807 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5808 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5809 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5810 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5811 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5812 - Correctly handle reads of files with no objects (6243)
5813 - lctl recover will also mark a device active if deactivate used (5933)
5815 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5816 - allow --write-conf on an MDS with different nettype than client (5619)
5817 - don't write config llogs to MDS for mounts not from that MDS (5617)
5818 - lconf should create multiple TCP connections from a client (5201)
5819 - init scripts are now turned off by default; run chkconfig --on
5820 lustre and chkconfig --on lustrefs to use them
5821 - upcalls are no longer needed for clients to recover to failover
5823 - add --abort-recovery option to lconf to abort recovery on device
5825 - add support for an arbitrary number of OSTs (3026)
5826 - Quota support protocol changes.
5827 - forward compatibility changes to wire structs (6007)
5828 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5829 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5830 - improved "lustre status" script
5831 - initialize blocksize for non-regular files (6062)
5832 - added --disable-server and --disable-client configure options (5782)
5833 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5834 - Vanilla 2.4.29 support
5835 - increase maximum number of obd devices to 520 (6242)
5836 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5837 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5838 SLES 9 2.6.5 kernels
5839 - limit stripes per file to 160 (the maximum EA size) (6093)
5841 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5844 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5845 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5846 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5847 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5848 - small ext3 extents cleanups and fixes (5733)
5849 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5850 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5851 - avoid needless client->OST connect, fix handle mismatch (5317)
5852 - fix DLM error path that led to out-of-sync client, long delays (5779)
5853 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5854 - fix several locking issues related to i_size (5492,5624,5654,5672)
5855 - don't move pending lock onto export if it is already evicted (5683)
5856 - fix kernel oops when creating .foo in unlinked directory (5548)
5857 - fix deadlock in obdfilter statistics vs. object create (5811)
5858 - use time_{before,after} to avoid timer jiffies wrap (5882)
5859 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5860 - Do not leak request if server was not able to process it (5154)
5861 - If mds_open unable to find parent dir, make that negative lookup(5154)
5862 - don't create new directories with extent-mapping (5909, 5936)
5864 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5865 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5866 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5867 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5868 - support for RHEL4 kernel on i686 (5773)
5869 - provide error messages when incompatible logs are encountered (5898)
5871 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5872 * version 1.4.0.10 (1.4.1 release candidate 1)
5874 - don't keep a lock reference when lock is not granted (4238)
5875 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5876 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5877 - drop import inflight refcount on signal_completed_replay error (5255)
5878 - unlock page after async write error during send (3677)
5879 - handle missing objects in filter_preprw_read properly (5265)
5880 - no transno return for symlink open, don't save no-trasno open (3440)
5881 - don't try to complete elan receive that already failed (4012)
5882 - free RPC server reply state on error (5406)
5883 - clean up thread from ptlrpc_start_thread() on error (5160)
5884 - readahead could read extra page into cache that wasn't ejected (5388)
5885 - prevent races in class_attach/setup/cleanup/detach (5260)
5886 - don't dereference de->d_inode after l_dput of de (5458)
5887 - use "int" for stripe value returned from lock_to_stripe (5544)
5888 - mballoc allocation and error-checking fixes in 2.6 (5504)
5889 - block device patches to fix I/O request sizes in 2.6 (5482)
5890 - look up hostnames for IB nals (5602)
5891 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5892 - don't start multiple acceptors for the same port (5277)
5893 - fix incorrect LASSERT in mds_getattr_name (5635)
5894 - export a proc file for general "ping" checking (5628)
5895 - fix "lfs check" to not block when the MDS is down (5628)
5897 - service request history (4965)
5898 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5899 - create an "evict_client" /proc entry on OSTs, like the MDS has
5900 - fix mount usage message, return errors per mount(8) (5168)
5901 - change grep [] to grep "[]" in tests so they work in more UMLs
5902 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5903 - remove ancient LOV_MAGIC_V0 EA support (5047)
5904 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5905 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5906 - allow readahead allocations to fail when low on memory (5383)
5907 - mmap locking landed again, after considerable improvement (2828)
5908 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5910 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5913 - send OST transaction number in read/write reply to free req (4966)
5914 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5915 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5916 - the watchdog thread now runs as interruptible (5246)
5917 - flock/lockf fixes (but it's still disabled, pending 5135)
5918 - don't use EXT3 constants in llite code (5094)
5919 - memory shortage at startup could cause assertion (5176)
5921 - reorganization of lov code
5922 - single portals codebase
5924 - add extents/mballoc support (5025)
5925 - direct I/O reads in the obdfilter (4048)
5926 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5928 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5931 - send OST transaction number in read/write reply to free req (4966)
5932 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5933 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5934 - the watchdog thread now runs as interruptible (5246)
5935 - handle missing objects in filter_preprw_read properly (5265)
5936 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5937 - drop import inflight refcount on signal_completed_replay error (5255)
5938 - unlock page after async write error during send (3677)
5939 - return original error code on reconstructed replies (3761)
5940 - no transno return for symlink open, don't save no-trasno open (3440)
5942 - add pid to ldlm debugging output (4922)
5943 - bump the watchdog timeouts -- we can't handle 30sec yet
5944 - extra debugging for orphan dentry/inode bug (5259)
5946 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5949 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5950 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5951 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5952 - fix NULL dereference in /proc/sys/portals/routes (4827)
5953 - allow failed mdc_close() operations to be interrupted (4561)
5954 - stop precreate on OST before MDS would time out on it (4778)
5955 - don't send partial-page writes before EOF from client (4410)
5956 - discard client grant for sub-page writes on large-page clients (4520)
5957 - don't free dentries not owned by NFS code, check generation (4806)
5958 - fix lsm leak if mds_create_objects() fails (4801)
5959 - limit debug_daemon file size, always print CERROR messages (4789)
5960 - use transno after validating reply (3892)
5961 - process timed out requests if import state changes (3754)
5962 - update mtime on OST during writes, return in glimpse (4829)
5963 - add mkfsoptions to LDAP (4679)
5964 - use ->max_readahead method instead of zapping global ra (5039)
5965 - don't interrupt __l_wait_event() during strace
5967 - add software watchdogs to catch hung threads quickly (4941)
5968 - make lustrefs init script start after nfs is mounted
5969 - fix CWARN/ERROR duplication (4930)
5970 - return async write errors to application if possible (2248)
5971 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5972 - print NAL number in %x format (4645)
5973 - update barely-supported suse-2.4.21-171 series (4842)
5974 - support for sles 9 %post scripts
5975 - support for building 2.6 kernel-source packages
5976 - support for sles km_* packages
5978 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5981 - ignore -ENOENT errors in osc_destroy (3639)
5982 - notify osc create thread that OSC is being cleaned up (4600)
5983 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5984 - reconstruct ost_handle() like mds_handle() (4657)
5985 - create a new thread to do import eviction to avoid deadlock (3969)
5986 - let lconf resolve symlinked-to devices (4629)
5987 - don't unlink "objects" from directory with default EA (4554)
5988 - hold socknal file ref over connect in case target is down (4394)
5989 - allow more than 32000 subdirectories in a single directory (3244)
5990 - fix blocks count for O_DIRECT writes (3751)
5991 - OST returns ENOSPC from object create when no space left (4539)
5992 - don't send truncate RPC if file size isn't changing (4410)
5993 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5994 - bind to privileged port in socknal and tcpnal (3689)
5996 - rate limit CERROR/CWARN console message to avoid overload (4519)
5997 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5998 - basic mmap support (3918)
5999 - kernel patch series update from b1_4 (4711)
6001 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6004 - avoid crash during MDS cleanup with OST shut down (2775)
6005 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6006 - don't use bad inodes on the MDS (3744)
6007 - dynamic object preallocation to improve recovery speed (4236)
6008 - don't hold spinlock over lock dumping or change debug flags (4401)
6009 - don't zero obd_dev when it is force cleaned (3651)
6010 - print grants to console if they go negative (4431)
6011 - "lctl deactivate" will stop automatic recovery attempts (3406)
6012 - look for existing locks in ldlm_handle_enqueue() (3764)
6013 - don't resolve lock handle twice in recovery avoiding race (4401)
6014 - revalidate should check working dir is a directory (4134)
6016 - don't always mark "slow" obdfilter messages as errors (4418)
6018 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6021 - don't close LustreDB during write_conf until it is done (3860)
6022 - fix typo in lconf for_each_profile (3821)
6023 - allow dumping logs from multiple threads at one time (3820)
6024 - don't allow multiple threads in OSC recovery (3812)
6025 - fix debug_size parameters (3864)
6026 - fix mds_postrecov to initialize import for llog ctxt (3121)
6027 - replace config semaphore with spinlock (3306)
6028 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6029 - don't allow enqueue to complete on a destroyed export (3822)
6030 - down write_lock before checking llog header bitmap (3825)
6031 - recover from lock replay timeout (3764)
6032 - up llog sem before sending rpc (3652)
6033 - reduce ns lock hold times when setting kms (3267)
6034 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6035 - fix NULL deref and obd_dev leak on setup error (3312)
6036 - replace some LBUG about llog ops with error handling (3841)
6037 - don't match INVALID dentries from d_lookup and spin (3784)
6038 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6039 - fix invalid assertion in ptlrpc_set_wait (3880)
6041 - add libwrap support for the TCP acceptor (3996)
6042 - add /proc/sys/portals/routes for non-root route listing (3994)
6043 - allow setting MDS UUID in .xml (2580)
6044 - print the stack of a process that LBUGs (4228)
6046 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6049 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6050 - make sure to unset replay flag from failed open requests (3440)
6051 - if default stripe count is 0, use OST count for inode size (3636)
6052 - update parent mtime/ctime on client for create/unlink (2611)
6053 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6054 - free recovery state on server during a forced cleanup (3571)
6055 - unregister_reply for resent reqs (3063)
6056 - loop back devices mounting and status check on 2.6 (3563)
6057 - fix resource-creation race that can provoke i_size == 0 (3513)
6058 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6059 - more debugging for page-accounting assertion (3746)
6060 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6061 - avoid deadlock after precreation failure (3758)
6062 - fix race and lock order deadlock in orphan handling (3450, 3750)
6063 - add validity checks when grabbing inodes from l_ast_data (3599)
6065 - add /proc/.../recovery_status to obdfilter (3428)
6066 - lightweight CDEBUG infrastructure, debug daemon (3668)
6067 - change default OSC RPC parameters to be better on small clusters
6068 - turn off OST read cache for files smaller than 32MB
6069 - install man pages and include them in rpms (3100)
6070 - add new init script for (un)mounting lustre filesystems (2593)
6071 - run chkconfig in %post for init scripts (3701)
6072 - drop scimac NAL (unmaintained)
6074 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6077 - clean kiobufs before and after use (3485)
6078 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6079 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6080 - change init script to fail more clearly if not run as root (1528)
6081 - allow clients to reconnect during replay (1742)
6082 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6083 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6084 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6085 - fix cond_resched (3554)
6086 - only evict unfinished clients after recovery (3515)
6087 - allow bulk resend, prevent data loss (3570)
6088 - dynamic ptlrpc request buffer allocation (2102)
6089 - don't allow unlinking open directory if it isn't empty (2904)
6090 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6091 - remove extraneous obd dereference causing LASSERT failure (3334)
6092 - don't use get_cycles() when creating temp. files on the mds (3156)
6093 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6094 - handle EEXIST for set-stripe, set proper directory name (3336)
6096 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6097 - fix ksocknal_fmb_callback() error messages (2918)
6099 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6102 - don't copy lvb into (possibly NULL) reply on error (2983)
6103 - don't deref dentry after dput, don't free lvb on error (2922)
6104 - use the kms to determine writeback rpc length (2947)
6105 - increment oti_logcookies when osc is inactive (2948)
6106 - update client's i_blocks count via lvb messages (2543)
6107 - handle intent open/close of special files properly (1557)
6108 - mount MDS with errors=remount-ro, like obdfilter (2009)
6109 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6110 - don't use cancelling-locks' kms values (2947)
6111 - use highest lock extent for kms, not last one (2925)
6112 - don't dereference ERR_PTR() dentry in error handling path (3107)
6113 - fix thread race in portals_debug_dumplog() (3122)
6114 - create lprocfs device entries at setup instead of at attach (1519)
6115 - common AST error handler, don't evict client on completion race (3145)
6116 - zero nameidata in detach_mnt in 2.6 (3118)
6117 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6118 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6119 - initialize RPC timeout timer earlier for 2.6 (3219)
6120 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6121 - print nal/nid for unknown nid (3258)
6122 - additional checks for oscc recovery before doing precreate (3284)
6123 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6124 - don't crash in mdc_close for bad permissions on open (3285)
6125 - zero i_rdev for non-device files (3147)
6126 - clear page->private before handing to FS, better assertion (3119)
6127 - tune the read pipeline (3236)
6128 - fix incorrect decref of invalidated dentry (2350)
6129 - provide read-ahead stats and refine rpc in flight stats (3328)
6130 - don't hold journal transaction open across create RPC (3313)
6131 - update atime on MDS at close time (3265)
6132 - close LDAP connection when recovering to avoid server load (3315)
6133 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6134 - don't leak open file on MDS after open resend (3325)
6135 - serialize filter_precreate and filter_destroy_precreated (3329)
6136 - loop device shouldn't call sync_dev() for nul device (3092)
6137 - clear page cache after eviction (2766)
6138 - resynchronize MDS->OST in background (2824)
6139 - refuse to mount the same filesystem twice on same mountpoint (3394)
6140 - allow llmount to create routes for mounting behind routers (3320)
6141 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6142 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6143 - fix rare problem with rename on htree directories (3417)
6145 - allow default OST striping configuration per directory (1414)
6146 - fix compilation for qswnal for 2.6 kernels (3125)
6147 - increase maximum number of MDS request buffers for large systems
6148 - change liblustreapi to be useful for external progs like lfsck (3098)
6149 - increase local configuration timeout for slow disks (3353)
6150 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6152 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6155 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6156 - don't overwrite extent policy data in reply if lock was blocked (2901)
6157 - drop filter export grants atomically with removal from device (2663)
6158 - del obd_self_export from work_list in class_disconnect_exports (2908)
6159 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6160 - swab reply message in mdc_close, other PPC fixes (2464)
6161 - fix destroying of named logs (2325)
6162 - overwrite old logs when running lconf --write_conf (2264)
6163 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6164 - fix race in target_handle_connect (2898)
6165 - mds_reint_create() should take same inode create lock (2926)
6166 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6167 - don't close files for self_export to avoid uninitialized obd (2936)
6168 - allow MDS with the same name as client node (2939)
6169 - hold dentry reference for closed log files for unlink (2325)
6170 - reserve space for all logs during transactions (2059)
6171 - don't evict page beyond end of stripe extent (2925)
6172 - don't oops on a deleted current working directory (2399)
6173 - handle hard links to targets without a parent properly (2517)
6174 - don't dereference NULL lock when racing during eviction (2867)
6175 - don't grow lock extents when lots of conflicting locks (2919)
6177 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6180 - account for cache space usage on clients to avoid data loss (974)
6181 - lfsck support in lustre kernel code (2349)
6182 - reduce journal credits needed for BRW writes (2370)
6183 - orphan handling to avoid losing space on client/server crashes
6184 - ptlrpcd can be blocked, stopping ALL progress (2477)
6185 - use lock value blocks to assist in proper KMS, faster stat (1021)
6186 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6187 - recovery for initial connections (2355)
6188 - fixes for mds_cleanup_orphans (1934)
6189 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6190 - block all file creations until orphan recovery completes (1901)
6191 - client remove rq_connection from request struct (2423)
6192 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6193 - recovery timer race (2670)
6194 - mdc_close recovey bug (2532)
6195 - ptlrpc cleanup bug (2710)
6196 - mds timeout on local locks (2588)
6197 - namespace lock held during RPCs (2431)
6198 - handle interrupted sync write properly (2503)
6199 - don't try to handle a message that hasn't been replied to (2699)
6200 - client assert failure during cleanup after abort recovery (2701)
6201 - leak mdc device after failed mount (2712)
6202 - ptlrpc_check_set allows timedout requests to complete (2714)
6203 - wait for inflight reqs when ptlrpcd finishes (2710)
6204 - make sure unregistered services are removed from the srv_list
6205 - reset bulk XID's when resending them (caught by 1138 test)
6206 - unregister_bulk after timeout
6207 - fix lconf error (2694)
6208 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6209 - readahead locks pages, leaves pending causing memory pressure (2673)
6210 - increase OST request buffers to 4096 on large machines (2729)
6211 - fix up permission of existing directories in simple_mkdir (2661)
6212 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6213 - don't assign transno to errored transactions (2742)
6214 - don't delete objects on OST if given a bogus objid from MDS (2751)
6215 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6216 - if rq_no_resend, then timeout request after recovery (2432)
6217 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6218 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6219 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6220 - MDS recovery shouldn't delete objects that it has given out (2730)
6221 - if enqueue arrives after completion, don't clobber LVB (2819)
6222 - don't unlock pages twice when trigger_group_io returns error (2814)
6223 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6224 - don't write pages to disk if there was an error (1450)
6225 - don't ping imports that have recovery disabled (2676)
6226 - take buffered bytes into account when balancing socknal conn (2817)
6227 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6228 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6229 - remove little-endian swabbing of llog records (1987)
6230 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6231 - retry reposting request buffers if they fail (1191)
6232 - grow extent at grant time to avoid granting a revoked lock (2809)
6233 - lock revoke doesn't evict page if covered by a second lock (2765)
6234 - disable VM readahead to avoid reading outside lock extents (2805)
6236 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6237 - updated kernel patches for hp-2.4.20 kernel (2681)
6239 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6242 - fix truncated write corruption (2366)
6243 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6245 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6246 - ensure len(uuid) < 37 in lmc (1171)
6247 - fix ia64 OOPS in llog_test (2255)
6248 - zero end of page at obdfilter for partial page writes (2648)
6249 - don't leave stale dentries around after renames (bug 2428)
6250 - fix timeouts when evicting a client with a single lock held (2642)
6251 - set deadline for the initial HELLO message to drain (2634)
6252 - print out dotted-quad IP addresses in the socknal (2302)
6254 - additional debugging for MDS client eviction problem (2443)
6255 - fix mkfsoptions support for osts (2603, 2604)
6257 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6260 - add series for the vanilla 2.6.0 kernel
6261 - add series for the vanilla 2.4.24 kernel
6262 - add series for a cray x86/64 UL kernel drop
6263 - fix xattr patches for the vanilla 2.4.19 series
6265 - generate true UUIDs in lmc (1171)
6266 - have portals stack dumping break in UML (2466)
6267 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6268 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6269 - wake callback waiting threads on client eviction (2460)
6270 - Add --ptldebug and --subsystem to lmc (1719)
6271 - update assertion to allow safe interrupt allocation
6272 - set rq_no_resend for cancel requests (2432)
6273 - recalculate ptlrpcd timeout after resend (2494)
6274 - call vfs_rmdir when removing pending directories (2368)
6275 - fix renaming a file to itself (2429)
6276 - lmc creates a default one-stripe lov (2454)
6277 - expand procfs space to handle large clusters (2326)
6278 - increase UML stack to avoid overflow
6279 - update lconf's list of debug and subsystem masks
6280 - fix lfs find --obd (2510)
6281 - /proc tunable for disabling filter read caching (2591)
6282 - stop rpm packages from altering slapd.conf (2301)
6283 - disable nagle in the socknal under 0conf (2578)
6284 - choose mds inode size based on stripe count (2572)
6285 - fix kernel-source rpm problems (2516)
6287 - add --disable-doc to avoid pdf generation (2421)
6288 - update documentation, tests, type-os, comments
6289 - avoid format warnings on ia64
6290 - remove the TOE NAL
6291 - tiny code cleanups by removing unused fields
6293 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6296 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6297 - remove the most blatant lies from BUILDING file (2371)
6298 - change default debug level to reasonable production setting
6299 - reduce client side cache size to reduce cache flush time
6300 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6301 - make TCP zerocopy and pinger support enabled by default (2476)
6302 - sync writes completed after process exits caused crashes (2319)
6303 - maintain correct mount count on the MDS (2356)
6304 - backout 1557, because 2316 wasn't really fixed
6305 - better file I/O statistics gathering in /proc
6306 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6307 - another kernel patch to fix zero-copy TCP function export
6308 - don't take duplicate lock when processing re-sent getattr (2420)
6309 - lctl uses obd_self_export instead of creating new conn (2353)
6310 - MDS/OST recovery case which requires object creation asserted (2425)
6311 - move lfs from /usr/sbin to /usr/bin in packages
6312 - fix race between mds_client_add and mds_client_free (2417)
6313 - use kmalloc instead of slabs in portals (2430)
6314 - don't create duplicate records when a failover MDS is present (2442)
6315 - remove unnecessary mount age check (2332)
6316 - don't remove directory inodes from locks prematurely (2451)
6317 - don't break if MDS service name is the same as hostname (2103)
6318 - fix races in client write RPC generation when cache full (2482)
6320 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6323 - remove now-unused request->rq_obd (278)
6324 - if an allocation fails, print out how much memory we've used (1933)
6325 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6326 - add the "configurable stack size" patch to most series files (1256)
6327 - ability to write large log records, for 100+ OST configs (2306)
6328 - fix NULL deref when filter_prep fails (2314)
6329 - fix operator precedence error in filter_sync
6330 - dynamic allocation of socknal TX descriptors (2315)
6331 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6332 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6333 - ensure that CWARN messages in Portals always get to the syslog
6334 - __init/__exit are not for prototype decls (ldlm_init/exit)
6335 - x86-64 compile warning fixes
6336 - fix gateway LMC keyword conflict (2318)
6337 - fix MDS lock inversions in getattr/reint paths (1844)
6338 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6339 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6340 - DLM race condition prevented some lock evictions (2328)
6341 - ENOMEM detection and retry on socknal sends (2230)
6342 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6343 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6344 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6345 - better error messages when a client is rejected during recovery (1505)
6346 - avoid cancelling locks which were never granted, after failure (2330)
6347 - fix i_sem/journal inversion in mds_client_add (2333)
6348 - fix truncate/getattr lock cycle deadlock (2334)
6349 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6350 - fix two rare exit paths which could leak an l_lock() ref (2321)
6351 - fencepost error in MDS/OST orphan recovery (2226)
6352 - make log record alignment 8 bytes (1988)
6353 - lstripe now fails when requested offset > ost_count (2237)
6354 - ensure that all kernel series have a complete list.h (1607)
6355 - fix crashes in special-file operations (2316)
6356 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6358 - allow configurable automake binary, for testing new versions
6359 - small update to the lfs documentation
6361 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6363 * fix negative export reference count in fsfilt_sync (2312)
6365 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6366 * release candidate 0.9.1
6368 - orphans are moved into the PENDING directory for possible recovery
6369 - replayed opens now open by fid for orphan/rename safety (1042)
6370 - last close of an orphan inode generates a transno (683)
6371 - chdir() and mount() now pin the directory entry (1020)
6372 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6373 - discard very old requests without processing them (1502)
6374 - remove some common, well-understood CERRORs (1505)
6375 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6376 - clear "grant" flags in OST replies until OST grant code lands (1644)
6377 - fix read performance by not clobbering i_blksize on client (1598)
6378 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6379 - make LRU size a /proc tunable, clears locks when reduced (707)
6380 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6381 - server locks take references on exports now (1558)
6382 - build fixes for 2.4.20-rh trees (1663)
6383 - return an error from lov_create if all OSCs are inactive (1751)
6384 - fix import levels when a reconnect happens without a timeout (1597)
6385 - exit early from mds_open if we get a lookup error (1749)
6386 - partial page read at EOF wouldn't wait for disk before sending (1642)
6387 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6388 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6389 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6390 - write/truncate lock inversion (1639)
6391 - Don't auto-load obdclass, portals modules during cleanup (1495)
6392 - fix timestamps from jumping to "now" (1763)
6393 - extra journal assertions (1648)
6394 - add an extra multiunlink test (1771)
6395 - fix read_record/write_record API (1776)
6396 - fix leak of offset_extent, possible incorrect i_size later (1772)
6397 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6398 - Debugging for the kqswnal_get_idle_tx problems (1820)
6399 - Allow recovery to be attempted multiple times (1536)
6400 - Write out MDS last_rcvd file after it is first created (1600)
6401 - Fix tx_descriptor leak in failed transmit situations (1827)
6402 - ext3 journaling fixes for assertion failure after IO error (1871)
6403 - class_export_put() on freed export after completion AST error (1896)
6404 - Fix revalidate looping in VFS (1322)
6405 - Don't access a freed export during MDS_REINT timeout (1521)
6406 - Add open-unlink recovery support on the MDS (1673,1764)
6407 - Return an error if no MDS data was read from last_rcvd (1946)
6408 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6409 - Don't setup a disk device that doesn't match exported UUID (317)
6410 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6411 - avoid committing NULL handle in force close
6412 - local.sh is now a one-stripe LOV configuration
6413 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6414 - fix ext3 htree duplicate directory entry corruption (1516)
6415 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6416 - update to new LOV EA format (2097)
6417 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6419 2003-06-15 Phil Schwan <phil@clusterfs.com>
6422 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6423 - per-import/export recovery handling (958, 931, 959)
6424 - multiple last-rcvd slots, for serving multiple FSes (949)
6425 - connections are again shared between multiple imp/exports (963, 964)
6426 - "umount -f" would hang if any requests needed to be sent (393, 978)
6427 - avoid pinning large req buffer by copying for queued messages (989)
6428 - add "uuid" to "lctl device" command to help upcalls (991)
6429 - "open" RPCs with transnos would confuse recovery counters (1037)
6430 - do proper endian conversion of all wire messages (288, 340, 891)
6431 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6432 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6433 - fix ptlrpc_connection leak in target_handle_connect (1174)
6434 - fix import refcounting bug in OST and MDS cleanup (1134)
6435 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6436 - fix very unlikely obd_types race condition (501)
6437 - remove osc_open hack for echo_client (1187)
6438 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6439 - a failure in read_inode2 leads to deadlock (1139)
6440 - cancel ack-locks as soon as transaction is committed (1072)
6441 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6442 - make sure to commitrw after any preprw to avoid deadlock (1162)
6443 - failing to execute a file in a lustre FS would lock inode (1203)
6444 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6445 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6446 - obd_uuid2tgt would walk off the end of the list (1255)
6447 - on IA64 the osc would give portals incorrect bulk size (1258)
6448 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6449 - fix lock inversion caused by new llite matching code (1282)
6450 - limit the number of dirty pages on a client to 10MB (1286)
6451 - timed out locks were not being corrected cancelled (1289)
6452 - fix O_DIRECT above 4GB on IA-32 (1292)
6453 * major user-visible changes
6454 - fail out/fail over policy now controlled by the upcall (993)
6456 - add OBD_PING to check server availability and failure (954)
6457 - lustre messages are now sent in sending host order (288, 340, 891)
6458 - add eadatalen to MDS getattr reply (340)
6459 - OST read replies may contain second buffer, with per-page status (593)
6461 2003-03-11 Phil Schwan <phil@clusterfs.com>
6464 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6465 - failed open()s could cause deadlock; fixed (867, 869)
6466 - stop cancelling OST locks when files are closed (481)
6467 - overlapping XID spaces caused network corruption (851, 853)
6468 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6469 - setattr_raw added, to do single-RPC, server-side setattrs
6470 - lmc/lconf syntax change for OST UUIDs
6471 - fix crashy race condition between ptlrpc_free_req and osc_close
6472 - don't use request in mdc_enqueue if we hit a timeout (889)
6473 - don't set the inode i_size for regular files from the MDS (896)
6474 - handle out of order completion AST (842)
6475 - don't LBUG if a lock request times out after receiving AST (913)
6476 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6477 - if a bad lock AST arrives, send an error instead of dropping entirely
6478 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6479 - fix leak in bulk IO when only partially completed (899, 900, 926)
6480 - fix O_DIRECT for ia64 (55)
6481 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6482 - C-z after timeout could hang a process forever; fixed (977)
6484 - client-side I/O cache (678, 924, 929, 941, 970)
6486 - READPAGE and SETATTRs which don't take server-side locks get
6489 2003-02-11 Phil Schwan <phil@clusterfs.com>
6492 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6493 - Fix fsfilt_extN_readpage() to read a full page of directory
6494 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6495 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6496 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6497 - Most intent operations were removed, in favour of a new RPC mode
6498 that does a single RPC to the server and bypasses most of the VFS
6499 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6500 - Aggressively cancel local locks on DLM servers
6501 - mds_reint_unlink sends EA to the client if it's the last nlink.
6502 client uses that EA to unlink OST objects.
6503 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6504 - recursive symlinks were fixed (439)
6505 - fixed NULL deref in DEBUG_REQ
6506 - filter_update_lastobjid no longer calls sync, which annoyed extN
6507 - fixed multi-client small-writes to a single file problem (445)
6508 - fixed mtime updates during file writes (607)
6509 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6510 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6511 - fixed Linux OST size reporting problem (444, 656)
6512 - OST now updates object mtime with writes or setattr (607, 619)
6513 - client verifies file size before zeroing page past EOF (445)
6514 - OST now writes last allocated objid to disk with allocation (108)
6515 - LOV on echo now works (409)
6517 - mds_reint_unlink sends a new buffer, with the EA included. this
6518 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6519 set if a regular file was being unlinked, and it was the last link
6520 - use PtlGet from the target for bulk writes (315)
6521 - OST now updates object mtime with writes or setattr (607, 619)
6522 - LDLM now has a grant-time callback to revalidate locked items, if
6524 - Many MDS operations were reorganized to combat race conditions
6526 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6527 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6529 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6532 - Fully reactivate OST imports after reconnection (512, others)
6533 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6534 - More graceful error handling for truncating on dead OST (515)
6535 - Don't error out unless we're actually accessing dead stripes (474)
6536 - Fix garbage sizes when stripes are missing (410)
6537 - LRU counters were broken, causing constant lock purge (433, 432)
6538 - garbage on read from stripes with failed OSTs (441)
6539 - mark OSCs as active before reconnecting during recovery (438)
6540 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6541 - lfind did not preserve OST order in output (443)
6542 - symlinks cause hung clients, incorrect data (439)
6543 - stop dereferencing request after dropping refcount (457)
6544 - don't LASSERT(spin_is_locked) on non-SMP (455)
6545 - fixes for many rename() bugs
6546 - fstat didn't correctly synchronize attributes (399)
6547 - server must handle lock cancellation during blocking AST prep (487)
6548 - bulk descriptors were free()d too soon (511)
6549 - fix paths in lconf, which would load incorrect modules (451, 507)
6550 - fix confusing lconf 'host not found' error message (386)
6551 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6552 - fix race condition in mdc_blocking_ast() for inode access (526)
6553 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6554 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6555 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6556 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6557 - fix lstripe in several ways (kernel side) (527)
6558 - fix request leak in ldlm_cli_enqueue (262)
6559 - incorrect OSC was marked inactive after OST failure
6560 - call mds_fs_cleanup before unmounting filesystem (524)
6561 - fix races between taking ns_lock and ldlm_lock_change_resource
6562 - fix races updating LOV export open file list
6563 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6564 - fix recovery NULL deref in ldlm_cli_cancel_unused
6565 - fix some DLM races by using new hash table for lock handles (419)
6566 - permit the client to specify desired inodes, at replay
6567 - duplicate requests when we queue them for replay reintegration
6568 - fix last_rcvd offset calculation
6569 - sync after each recovered transaction, so we always make progress
6570 - never, not always, ERESTART requests without transnos
6571 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6573 - skip replay if the MDS says that the client is already connected
6574 - don't check for a recovery-enabled export to match lctl's UUID
6575 - don't INC_USE_COUNT for phantom exports
6576 - don't crash when cleaning up phantom exports (567)
6577 - don't double-finish or set replay data for errored mdc_open requests
6578 - abort requests when they time out, so we don't get old replies
6579 - send/receive replies for AST messages again
6580 - if the client says that it doesn't have the lock, cancel it on the
6582 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6583 mark it as destroyed, it will all work out when decref is called
6584 - fix module use counts (22, 581)
6586 - ASTs now expect a reply (server cancels lock on error reply)
6588 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6591 - fix many simultaneous client startup (392)
6592 - fix dentry->d_it clobbering
6593 - credentials weren't being shipped for readdir/getattr operations
6594 - remove invalid assertions triggered during some concurrent MD
6596 - proper Lustre versions added (336, 389)
6597 - fix memory leak for create error case (398)
6598 - fix LOV locking bug that would get cli/srv out of sync
6599 - fix echo client over LOV (409)
6600 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6601 - fix double-O_EXCL intent crash (424)
6602 - avoid sending multiple lock CANCELs (352)
6604 - MDS can do multi-client recovery (modulo bugs in new code)
6606 - many updates, edits, cleanups
6608 2002-11-18 Phil Schwan <phil@clusterfs.com>
6611 - fix null d_it dereference (346)
6612 - fix full OST/dbench hang (333)
6613 - fix permission problem with file removal (286)
6614 - fix removal of OSCs from LOV when they fail
6615 - fix NULL deref during bulk timeout (214)
6616 - fix problems related to multiple filesystems on one MDS (241)
6617 - fixed serious subtle metadata locking bugs
6618 - free locks on clients when inodes are removed due to memory
6620 - fix inode pointer in lock data (285)
6621 - partial support for multiple MDS on a single host (241)
6622 - data locks weren't cancelled at clear_inode time (290, 311)
6623 - intent locks could lead to unbounded lock growth (205)
6624 - added a maximum lock count, an LRU list, and a flusher
6625 - fix multiple rename (365)
6626 - properly abstracted the echo client
6627 - OSC locked 1 byte too many; fixed
6628 - rewrote brw callback code:
6629 - fixed recovery bugs related to LOVs (306)
6630 - fixed too-many-pages-in-one-write crash (191)
6631 - fixed (again) crash in sync_io_timeout (214)
6632 - probably fixed callback-related race (385)
6634 - Add capability to MDS protocol
6635 - LDLM cancellations and callbacks on different portals
6637 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6640 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6641 - defer open object destruction to close time (601981, 138)
6642 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6643 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6644 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6645 - don't call obd_setattr in ll_file_release for destroyed objects
6646 * protocol change to lustre_msg: move |version| and add |flags|
6647 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6648 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6649 * added replay of create, unlink, link and rename operations during
6650 MDS failover; recovery should be much more robust now
6651 * remove failed OSCs from LOVs (only lov_create uses this so far)
6652 * the lustre-HOWTO was brought (more) up to date (582544)
6654 2002-10-23 Phil Schwan <phil@clusterfs.com>
6657 - in-use dentries weren't being reused properly (617851)
6658 - prevent multiple LDLM setup (599178)
6659 - fix LOV size calculations for truncate (617853)
6660 - fix client handling of MDS intent errors (POSIX)
6661 - fix permission bug in lovstripe.c test (624321)
6662 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6663 - truncate past end of file could corrupt data
6664 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6665 - a race in recovery could return ETIMEDOUT to apps (623947)
6666 - building outside the source directory was fixed
6667 * the lustre-HOWTO was brought (more) up to date (582544)
6668 * major progress was made on recovery functionality
6670 2002-10-10 Phil Schwan <phil@clusterfs.com>
6673 - recovery deadlock fix
6674 - rm -rf causes LBUG fix (617817)
6675 - file open by multiple tasks fix (618962)
6676 - directory permissions bugs (602707 and 620007)
6677 - journal_stop fixed with locking (611313)
6678 - O_APPEND failures resolved (618273, perhaps 614459)
6679 - lconf PATH fix (619770)
6680 - IA64 build fix (621450)
6681 - RPC buffer sizes scale with amount of memory
6683 2002-10-01 Phil Schwan <phil@clusterfs.com>
6686 - locks would be cancelled without throwing away data pages,
6687 resulting in inconsistent data (605627)
6688 - inode attributes were not always being refreshed (605627, 612449)
6689 - lconf now continues to cleanup after lctl reports an error
6690 - MDS now enforces user permissions (602707)
6691 - lprocfs cleanup fixed, but not yet enabled (614157)
6692 - fixed infinite server hang, should a client not respond to an AST
6693 - avoid going into recovery if user calls readlink() with a buffer
6694 that's too small (613941)
6695 - AST RPCs no longer require replies (614867) -- this may be changed
6696 - don't crash server if client sends an IOV that's too big (611336)
6697 - fixed lock conversion deadlock (611892)
6698 - fixed the following of symlinks (614622)
6699 * recovery: the server can remove locks from a client that dies, other
6700 clients can make progress
6701 * more extN patch fixes
6702 * compile-time configurable ptlrpc buffer allocations
6704 - collaborative read cache document
6705 - Lustre Lite Performance CDR document-in-progress
6707 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6710 - fix typo in patch-2.4.18
6712 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6715 - clear ptlrpc request each time in handle_incoming_request()
6716 - unlink of files now destroys the object on the OST
6718 2002-09-19 Peter Braam <braam@clusterfs.com>
6720 * add hard link support
6721 * change obdfile creation method
6722 * kernel patch changed
6724 2002-09-19 Peter Braam <braam@clusterfs.com>
6727 - stack overflow bug in extN fixed
6729 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6731 * documentation updates
6732 - add man pages for config tools
6733 - update tests/README to describe testing with new config tools
6734 - finish metadata API descriptions
6735 * bug fixes and cleanups
6736 - statfs workaround for 16TB limit
6737 - LOV stripe allocation improved, can stripe on subset of OSTs
6738 - LOV file size/IO offset was wrong for files > 4GB in size
6739 - object EA data was being dropped, caused files to be unreadable
6740 - memory overflow with non-LOV OST caused memory corruption
6741 - fixed regression tests to work with new config tools, obdfilter
6742 - fixed bug when directory size became larger than 1 block
6743 - fixed bug (for single client case) when PWD was deleted
6744 - invalidate local directory pages when doing intent-based ops
6745 - avoid LDLM oops when lock callback contained bad data
6747 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6749 * documentation updates
6750 * bug fixes and cleanups
6751 - configuration tools
6754 - 64-bit compile warnings
6755 - 64-bit internal statfs data
6757 * test_brw on persistent OST devices
6759 * lprocfs (disabled)
6761 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6763 * documentation updates
6764 * bug fixes and cleanups
6765 * configuration tools
6767 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6770 * many small fixes to 0_5_4
6771 * io/network handling
6772 * thinkos in MDS operations
6774 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6777 * crucial basic fixes to 0.5.3
6778 * IOR, Iozone work over Elan
6781 2002-08-07 Phil Schwan <phil@clusterfs.com>
6782 * version 0_5_3, our first alpha
6783 * we use the new Portals iovs
6784 * documentation updates
6785 * bug fixes and cleanups
6786 * small changes in the DLM wire protocol
6788 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6789 * version 0_5_1 with some initial stability,
6790 * locking on MD and file I/O.
6791 * documentation updates
6792 * several bug fixes since 0.5.0
6793 * small changes in wire protocol
6795 2002-07-18 Phil Schwan <phil@clusterfs.com>
6797 * delivered as Lustre Light Alpha
6798 * fixed a crash after handling invalid MDS requests
6799 * fixed directory pages for architectures with non-4k pages sizes
6801 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6802 * release version v0_4_4
6803 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6805 * Fixes a number of interruption problems with OST operations.
6806 * Update documentation for portals header changes
6807 * Move all wire protocol structs/defines to lustre_idl.h
6808 * Fixes symlink length bug.
6809 * Add tcpdump to repository.
6811 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6812 * release version v0_4_3
6813 * Fixes statfs for inodes on extN.
6814 * Fixes bug in runtests which would delete /etc/hosts.
6815 * Use 64-bit object IDs wherever possible (not into VFS though)
6816 Remove ost_get_info, which is unused by lustre, and out of date.
6818 2002-07-03 Peter Braam <braam@clusterfs.com>
6819 * release version v0_4_2 Fixes a lookup error (type not passed)
6820 * move forward to head of Portals
6821 * move forward to latest Lustre kernel
6823 2002-06-25 Peter Braam <braam@clusterfs.com>
6824 * release version v0_4_1. Hopefully stable on single node use.