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 3.10.0-1160.36.2.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 vanilla linux 5.4.21 (ZFS + ldiskfs)
24 vanilla linux 5.4.136 (ZFS + ldiskfs)
25 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
26 * Client primary kernels built and tested during release cycle:
27 3.10.0-1062.18.1.el7 (RHEL7.7)
28 3.10.0-1127.19.1.el7 (RHEL7.8)
29 3.10.0-1160.36.2.el7 (RHEL7.9)
30 4.12.14-95.48 (SLES12 SP4)
31 4.12.14-122.74 (SLES12 SP5)
32 4.12.14-197.75 (SLES15 SP1)
33 5.3.18-24.61 (SLES15 SP2)
34 4.15.0-48 (Ubuntu 18.04)
35 * Other clients known to build on these kernels at some point (others may also work):
36 3.10.0-862.14.4.el7 (RHEL7.5)
37 3.10.0-957.27.2.el7 (RHEL7.6)
38 4.14.0-49.13.1.el7a (RHEL7.5)
39 4.14.0-115.2.2.el7a (RHEL7.6)
40 4.18.0-80.11.2.el8 (RHEL8.0)
41 4.18.0-147.8.1.el8 (RHEL8.1)
42 4.18.0-193.28.1.el8 (RHEL8.2)
43 4.18.0-240.22.1.el8 (RHEL8.3)
44 4.18.0-305.3.1.el8 (RHEL8.4)
45 4.4.120-92.70 (SLES12 SP2)
46 4.4.180-94.100 (SLES12 SP3)
47 4.4.0-131 (Ubuntu 16.04)
48 5.4.0-37 (Ubuntu 20.04)
49 5.8.0-53 (Ubuntu 20.04.2 HWE)
50 5.11.0-31 (Ubuntu 20.04.3 HWE)
51 5.11.0 (vanilla kernel.org)
52 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
53 * Recommended ZFS version: 2.0.0
54 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
55 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
56 stack overflow. For more information, please refer to bugzilla 17630.
57 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
58 the re-exporting nodes to be mounted with "32bitapi" mount option
60 --------------------------------------------------------------------------------
64 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
65 for currently supported client and server kernel versions.
66 * Server primary kernels built and tested during release cycle:
67 3.10.0-1062.18.1.el7 (RHEL7.7)
68 3.10.0-1127.19.1.el7 (RHEL7.8)
69 3.10.0-1160.6.1.el7 (RHEL7.9)
70 * Other server kernels known to build and work at some point (others may also work):
71 3.10.0-862.14.4.el7 (RHEL7.5)
72 3.10.0-957.27.2.el7 (RHEL7.6)
73 4.18.0-80.11.2.el8 (RHEL8.0)
74 4.18.0-147.8.1.el8 (RHEL8.1)
75 4.18.0-193.28.1.el8 (RHEL8.2)
76 4.18.0-240.1.1.el8 (RHEL8.3)
77 4.4.120-92.70 (SLES12 SP2)
78 4.4.180-94.100 (SLES12 SP3)
79 4.4.0-131 (Ubuntu 16.04)
80 4.15.0-32 (Ubuntu 18.04)
81 5.4.0-48 (Ubuntu 20.04)
82 vanilla linux 5.4.0 (ZFS + ldiskfs)
83 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
84 * Client primary kernels built and tested during release cycle:
85 3.10.0-1062.18.1.el7 (RHEL7.7)
86 3.10.0-1127.19.1.el7 (RHEL7.8)
87 3.10.0-1160.6.1.el7 (RHEL7.9)
88 4.12.14-95.48 (SLES12 SP4)
89 4.12.14-122.20 (SLES12 SP5)
90 4.12.14-197.75 (SLES15 SP1)
91 5.3.18-24.24 (SLES15 SP2)
92 4.15.0-48 (Ubuntu 18.04)
93 * Other clients known to build on these kernels at some point (others may also work):
94 3.10.0-862.14.4.el7 (RHEL7.5)
95 3.10.0-957.27.2.el7 (RHEL7.6)
96 4.14.0-49.13.1.el7a (RHEL7.5)
97 4.14.0-115.2.2.el7a (RHEL7.6)
98 4.18.0-80.11.2.el8 (RHEL8.0)
99 4.18.0-147.8.1.el8 (RHEL8.1)
100 4.18.0-193.28.1.el8 (RHEL8.2)
101 4.18.0-240.1.1.el8 (RHEL8.3)
102 4.4.120-92.70 (SLES12 SP2)
103 4.4.180-94.100 (SLES12 SP3)
104 4.4.0-131 (Ubuntu 16.04)
105 5.4.0-37 (Ubuntu 20.04)
106 5.4.0 (vanilla kernel.org)
107 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
108 * Recommended ZFS version: 2.0.0
109 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
110 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
111 stack overflow. For more information, please refer to bugzilla 17630.
112 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
113 the re-exporting nodes to be mounted with "32bitapi" mount option
115 --------------------------------------------------------------------------------
119 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
120 for currently supported client and server kernel versions.
121 * Server primary kernels built and tested during release cycle:
122 3.10.0-1062.1.1.el7 (RHEL7.7)
123 * Other server kernels known to build and work at some point (others may also work):
124 3.10.0-862.14.4.el7 (RHEL7.5)
125 3.10.0-957.27.2.el7 (RHEL7.6)
126 4.18.0-80.7.1.el8 (RHEL8.0)
127 4.4.120-92.70 (SLES12 SP2)
128 4.4.180-94.100 (SLES12 SP3)
129 4.4.0-131 (Ubuntu 16.04)
130 4.15.0-32 (Ubuntu 18.04)
131 vanilla linux 4.6.7 (ZFS only)
132 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
133 * Client primary unpatched kernels built and tested during release cycle:
134 3.10.0-1062.1.1.el7 (RHEL7.7)
135 4.12.14-95.29 (SLES12 SP4)
136 4.15.0-48 (Ubuntu 18.04)
137 * Other clients known to build on unpatched kernels at some point (others may also work):
138 3.10.0-862.14.4.el7 (RHEL7.5)
139 3.10.0-957.27.2.el7 (RHEL7.6)
140 4.14.0-49.13.1.el7a (RHEL7.5)
141 4.14.0-115.2.2.el7a (RHEL7.6)
142 4.18.0-80.7.1.el8 (RHEL8.0)
143 4.4.120-92.70 (SLES12 SP2)
144 4.4.180-94.100 (SLES12 SP3)
145 4.4.0-131 (Ubuntu 16.04)
146 vanilla linux 4.15.0 (ZFS only)
147 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
148 * Recommended ZFS / SPL version: 0.7.13
149 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
150 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
151 stack overflow. For more information, please refer to bugzilla 17630.
152 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
153 the re-exporting nodes to be mounted with "32bitapi" mount option
154 * Note for Ubuntu 14.04.5:
155 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
156 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
158 --------------------------------------------------------------------------------
162 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
163 for currently supported client and server kernel versions.
164 * Server known to build on the following kernels (others may also work):
165 2.6.32-696.18.7.el6 (RHEL6.9)
166 3.10.0-862.14.4.el7 (RHEL7.5)
167 3.10.0-957.el7 (RHEL7.6)
168 4.4.120-92.70 (SLES12 SP2)
169 4.4.162-94.69 (SLES12 SP3)
170 4.4.0-131 (Ubuntu 16.04)
171 4.15.0-32 (Ubuntu 18.04)
172 vanilla linux 4.6.7 (ZFS only)
173 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
174 * Client known to build on unpatched kernels (others may also work):
175 2.6.32-696.18.7.el6 (RHEL6.9)
176 2.6.32-754.6.3.el6 (RHEL6.10)
177 3.10.0-862.14.4.el7 (RHEL7.5)
178 3.10.0-957.el7 (RHEL7.6)
179 4.14.0-49.13.1.el7a (RHEL7.5)
180 4.14.0-115.2.2.el7a (RHEL7.6)
181 4.4.120-92.70 (SLES12 SP2)
182 4.4.162-94.69 (SLES12 SP3)
183 4.4.0-131 (Ubuntu 16.04)
184 4.15.0-32 (Ubuntu 18.04)
185 vanilla linux 4.15.0 (ZFS only)
186 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
187 * Recommended ZFS / SPL version: 0.7.9
188 * Tested with ZFS / SPL version: 0.7.9
189 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
190 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
191 stack overflow. For more information, please refer to bugzilla 17630.
192 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
193 the re-exporting nodes to be mounted with "32bitapi" mount option
194 * Note for Ubuntu 14.04.5:
195 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
196 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
198 --------------------------------------------------------------------------------
200 2018-04-03 Intel Corporation
202 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
203 for currently supported client and server kernel versions.
204 * Server known to build on patched kernels:
205 2.6.32-431.29.2.el6 (RHEL6.5)
206 2.6.32-504.30.3.el6 (RHEL6.6)
207 2.6.32-573.26.1.el6 (RHEL6.7)
208 2.6.32-642.15.1.el6 (RHEL6.8)
209 2.6.32-696.18.7.el6 (RHEL6.9)
210 3.10.0-693.21.1.el7 (RHEL7.4)
211 3.0.101-0.47.71 (SLES11 SP3)
212 3.0.101-107 (SLES11 SP4)
213 3.12.74-60.64.40 (SLES12 SP1)
214 4.4.114-92.64 (SLES12 SP2)
215 4.4.114-94.11 (SLES12 SP3)
216 3.13.0-101 (Ubuntu 14.04, ZFS only)
217 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
218 4.4.0-116 (Ubuntu 16.04)
219 vanilla linux 4.6.7 (ZFS only)
220 * Client known to build on unpatched kernels:
221 2.6.32-431.29.2.el6 (RHEL6.5)
222 2.6.32-504.30.3.el6 (RHEL6.6)
223 2.6.32-573.26.1.el6 (RHEL6.7)
224 2.6.32-642.15.1.el6 (RHEL6.8)
225 2.6.32-696.18.7.el6 (RHEL6.9)
226 3.10.0-693.21.1.el7 (RHEL7.4)
227 3.0.101-0.47.71 (SLES11 SP3)
228 3.0.101-107 (SLES11 SP4)
229 3.12.74-60.64.40 (SLES12 SP1)
230 4.4.114-92.64 (SLES12 SP2)
231 4.4.114-94.11 (SLES12 SP3)
232 3.13.0-101 (Ubuntu 14.04)
233 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
234 4.4.0-116 (Ubuntu 16.04)
236 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
237 * Recommended ZFS / SPL version: 0.7.6
238 * Tested with ZFS / SPL version: 0.7.6
239 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
240 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
241 stack overflow. For more information, please refer to bugzilla 17630.
242 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
243 the re-exporting nodes to be mounted with "32bitapi" mount option
244 * Note for Ubuntu 14.04.5:
245 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
246 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
248 --------------------------------------------------------------------------------
250 2017-07-05 Intel Corporation
252 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
253 for currently supported client and server kernel versions.
254 * Server known to build on patched kernels:
255 2.6.32-431.29.2.el6 (RHEL6.5)
256 2.6.32-504.30.3.el6 (RHEL6.6)
257 2.6.32-573.26.1.el6 (RHEL6.7)
258 2.6.32-642.15.1.el6 (RHEL6.8)
259 2.6.32-696.3.1.el6 (RHEL6.9)
260 3.10.0-514.21.1.el7 (RHEL7.3)
261 3.0.101-0.47.71 (SLES11 SP3)
262 3.0.101-100 (SLES11 SP4)
263 3.12.74-60.64.40 (SLES12 SP1)
264 4.4.59-92.17 (SLES12 SP2)
265 vanilla linux 4.6.7 (ZFS only)
266 * Client known to build on unpatched kernels:
267 2.6.32-431.29.2.el6 (RHEL6.5)
268 2.6.32-504.30.3.el6 (RHEL6.6)
269 2.6.32-573.26.1.el6 (RHEL6.7)
270 2.6.32-642.15.1.el6 (RHEL6.8)
271 2.6.32-696.3.1.el6 (RHEL6.9)
272 3.10.0-514.21.1.el7 (RHEL7.3)
273 3.0.101-0.47.71 (SLES11 SP3)
274 3.0.101-100 (SLES11 SP4)
275 3.12.74-60.64.40 (SLES12 SP1)
276 4.4.59-92.17 (SLES12 SP2)
278 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
279 * Recommended ZFS / SPL version: 0.6.5.9
280 * Tested with ZFS / SPL version: 0.6.5.9
281 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
282 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
283 stack overflow. For more information, please refer to bugzilla 17630.
284 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
285 the re-exporting nodes to be mounted with "32bitapi" mount option
287 --------------------------------------------------------------------------------
289 12-07-2016 Intel Corporation
291 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
292 for currently supported client and server kernel versions.
293 * Server known to build on patched kernels:
294 2.6.32-431.29.2.el6 (RHEL6.5)
295 2.6.32-504.30.3.el6 (RHEL6.6)
296 2.6.32-573.26.1.el6 (RHEL6.7)
297 2.6.32-642.11.1.el6 (RHEL6.8)
298 3.10.0-514.el7 (RHEL7.3)
299 3.0.101-0.47.71 (SLES11 SP3)
300 3.0.101-80 (SLES11 SP4)
301 3.12.62-60.64.8 (SLES12 SP1)
302 vanilla linux 4.5.7 (ZFS only)
303 * Client known to build on unpatched kernels:
304 2.6.32-431.29.2.el6 (RHEL6.5)
305 2.6.32-504.30.3.el6 (RHEL6.6)
306 2.6.32-573.26.1.el6 (RHEL6.7)
307 2.6.32-642.11.1.el6 (RHEL6.8)
308 3.10.0-514.el7 (RHEL7.3)
309 3.0.101-0.47.71 (SLES11 SP3)
310 3.0.101-80 (SLES11 SP4)
311 3.12.62-60.64.8 (SLES12 SP1)
313 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
314 * Recommended ZFS / SPL version: 0.6.5.8
315 * Tested with ZFS / SPL version: 0.6.5.8
316 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
317 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
318 stack overflow. For more information, please refer to bugzilla 17630.
319 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
320 the re-exporting nodes to be mounted with "32bitapi" mount option
322 --------------------------------------------------------------------------------
324 02-29-2016 Intel Corporation
326 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
327 for currently supported client and server kernel versions.
328 * Server known to build on patched kernels:
329 2.6.32-431.29.2.el6 (RHEL6.5)
330 2.6.32-504.30.3.el6 (RHEL6.6)
331 2.6.32-573.12.1.el6 (RHEL6.7)
332 3.10.0-327.3.1.el7 (RHEL7.2)
333 3.0.101-0.47.71 (SLES11 SP3)
334 3.0.101-68 (SLES11 SP4)
335 vanilla linux 4.2.1 (ZFS only)
336 * Client known to build on unpatched kernels:
337 2.6.32-431.29.2.el6 (RHEL6.5)
338 2.6.32-504.30.3.el6 (RHEL6.6)
339 2.6.32-573.12.1.el6 (RHEL6.7)
340 3.10.0-327.3.1.el7 (RHEL7.2)
341 3.0.101-0.47.71 (SLES11 SP3)
342 3.0.101-68 (SLES11 SP4)
345 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
346 * Recommended ZFS / SPL version: 0.6.4.2
347 * Tested with ZFS / SPL version: 0.6.4.2
348 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
349 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
350 stack overflow. For more information, please refer to bugzilla 17630.
351 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
352 the re-exporting nodes to be mounted with "32bitapi" mount option
354 --------------------------------------------------------------------------------
356 03-10-2015 Intel Corporation
358 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
359 for currently supported client and server kernel versions.
360 * Server known to build on patched kernels:
361 2.6.32-431.29.2.el6 (RHEL6.5)
362 2.6.32-504.8.1.el6 (RHEL6.6)
363 3.0.101-0.46 (SLES11 SP3)
364 * Client known to build on unpatched kernels:
365 2.6.32-431.29.2.el6 (RHEL6.5)
366 2.6.32-504.8.1.el6 (RHEL6.6)
367 3.10.0-123.20.1.el7 (RHEL7)
368 3.0.101-0.46 (SLES11 SP3)
369 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
370 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
371 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
372 stack overflow. For more information, please refer to bugzilla 17630.
373 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
374 the re-exporting nodes to be mounted with "32bitapi" mount option
376 Severity : enhancement
378 Description: control OST-index in IDIF via ROCOMPAT flag.
379 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
380 last_rcvd file. For new formatted OST device, it will be auto set;
381 for the case of upgrading from old OST device, you can enable it
382 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
383 enabled, for new created OST-object, its IDIF-in-LMA will contain
384 the OST-index; for the existing OST-object, the OSD will convert
385 old format IDIF as new format IDIF with OST-index stored in the
386 LMA EA when accessing such OST-object or via OI scrub. Once such
387 flag is enabled, it cannot be reverted back, so the system cannot
388 be downgraded to the orignal incompatible version.
390 --------------------------------------------------------------------------------
392 07-30-2014 Intel Corporation
394 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
395 for currently supported client and server kernel versions.
396 * Server known to build on patched kernels:
397 2.6.32-431.20.3.el6 (RHEL6.5)
398 3.0.101-0.31 (SLES11 SP3)
399 * Client known to build on unpatched kernels:
400 2.6.32-431.20.3.el6 (RHEL6.5)
401 3.0.101-0.31 (SLES11 SP3)
404 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
405 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
406 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
407 stack overflow. For more information, please refer to bugzilla 17630.
408 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
409 the re-exporting nodes to be mounted with "32bitapi" mount option
411 --------------------------------------------------------------------------------
413 10-31-2013 Intel Corporation
415 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
416 * Server builds for kernels:
417 2.6.32-358.18.1.el6 (RHEL6.4)
418 * Client builds for unpatched kernels:
419 2.6.32-358.18.1.el6 (RHEL6.4)
420 3.0.92-0.8 (SLES11 SP3)
422 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
423 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
424 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
425 stack overflow. For more information, please refer to bugzilla 17630.
426 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
427 the re-exporting nodes to be mounted with "32bitapi" mount option
429 --------------------------------------------------------------------------------
431 05-30-2013 Intel Corporation
433 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
434 * Server builds for kernels:
435 2.6.32-358.6.2.el6 (RHEL6.4)
436 * Client builds for unpatched kernels:
437 2.6.32-358.6.2.el6 (RHEL6.4)
438 3.0.74-0.6.6 (SLES11 SP2)
440 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
441 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
442 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
443 stack overflow. For more information, please refer to bugzilla 17630.
444 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
445 the re-exporting nodes to be mounted with "32bitapi" mount option
447 --------------------------------------------------------------------------------
449 9-30-2012 Whamcloud, Inc.
451 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
452 * Server support for kernels:
453 2.6.32-279.5.1.el6 (RHEL6)
454 * Client support for unpatched kernels:
455 2.6.32-279.5.1.el6 (RHEL6)
457 2.6.18-238.19.1.el5 (RHEL5)
458 2.6.16.60-0.69.1 (SLES10)
459 * Recommended e2fsprogs version: 1.41.90.wc4
460 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
461 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
462 stack overflow. For more information, please refer to bugzilla 17630.
463 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
464 the re-exporting nodes to be mounted with "32bitapi" mount option
466 Severity : enhancement
468 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
470 Severity : enhancement
472 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
474 Severity : enhancement
476 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
478 Severity : enhancement
480 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
482 Severity : enhancement
484 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
486 --------------------------------------------------------------------------------
488 04-30-2012 Whamcloud, Inc.
490 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
491 * Server support for kernels:
492 2.6.32-220.4.2.el6 (RHEL6)
493 * Client support for unpatched kernels:
494 2.6.32-220.4.2.el6 (RHEL6)
496 2.6.18-238.19.1.el5 (RHEL5)
497 2.6.16.60-0.69.1 (SLES10)
498 * Recommended e2fsprogs version: 1.41.90.wc4
499 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
500 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
501 stack overflow. For more information, please refer to bugzilla 17630.
502 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
503 the re-exporting nodes to be mounted with "32bitapi" mount option
505 Severity : enhancement
507 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
509 Severity : enhancement
511 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
513 Severity : enhancement
515 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
517 Severity : enhancement
519 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
521 --------------------------------------------------------------------------------
523 09-30-2011 Whamcloud, Inc.
525 * Server support for kernels:
526 2.6.18-238.19.1.el5 (RHEL5)
527 2.6.32-131.6.1.el6 (RHEL6)
528 * Client support for unpatched kernels:
529 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
530 2.6.18-238.19.1.el5 (RHEL5)
531 2.6.32-131.6.1.el6 (RHEL6)
532 2.6.29.4-167.fc11 (FC11)
533 2.6.16.60-0.69.1 (SLES10)
535 * Recommended e2fsprogs version: 1.41.90.wc3
536 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
537 Lustre filesystem with 4K stack may cause a stack overflow. For more
538 information, please refer to bugzilla 17630.
539 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
540 mounted with "32bitapi" mount option
542 Severity : enhancement
543 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
545 Severity : enhancement
546 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
548 Severity : enhancement
549 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
552 Frequency : only with SLES10
554 Description: Use OFED "KMP" provided by Novell
555 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
556 Lustre is now built against this package. That means you need to
557 install the ofed-kmp package from Novell for the patchless client.
558 Note that the ofed-kmp that Novell ships may not exactly match the
559 kernel version but should still be compatible.
561 Severity : enhancement
563 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
564 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
566 --------------------------------------------------------------------------------
568 2010-07-15 Oracle, Inc.
570 * Server support for kernels:
571 2.6.18-164.11.1.el5 (RHEL 5)
572 2.6.18-164.11.1.0.1.el5 (OEL 5)
573 * Client support for unpatched kernels:
574 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
575 2.6.18-164.11.1.el5 (RHEL 5),
576 2.6.18-164.11.1.0.1.el5 (OEL 5)
577 2.6.16.60-0.42.8 (SLES 10),
579 2.6.29.4-167.fc11 (FC11)
580 * Recommended e2fsprogs version: 1.41.10-sun2
582 Severity : enhancement
584 Description: Distribute lfs_migrate script with lustre
585 Details : The lfs_migrate mentioned in the manual is now included in the
586 lustre tree in lustre/scripts and will distributed in lustre RPMs.
588 Severity : enhancement
590 Description: Failover nid list added to import proc file.
592 Severity : enhancement
594 Description: "som_preview" mount option required to use SOM
595 Details : The clients will need to use som_preview mount option
599 Description: Lustre does not do 1MB IOs to HW RAID
600 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
602 Severity : enhancement
604 Description: lfs quota output is not very convenient for awk/sed-parsing
605 Details : Some positions in lfs quota output table could be empty or
606 non-empty which made it hard to parse it with scripts, now
607 a dash is put instead of space where there is not supposed
612 Description: Update to OFED 1.5.1
613 Details : OFED 1.5.1 will be used in environments that don't use
614 vendor supplied OFED.
618 Description: Remove lnet. prefix from lctl params display
619 Details : Also renamed lnet.memused to lnet_memused to
624 Description: Use conf_param -d to remove permanant settings.
628 Description: Add list_param -R
629 Details : lctl list_param now supports recursive mode.
632 Bugzilla : 21829, 22312
633 Description: Fix llobdstat.
634 Details : 1. Now updating the write_bytes counter
635 2. Port script fixes from 1.8.
639 Description: decrease the usage of memory on clients.
640 Details : 1. On clients, recycle dentries and inodes unused.
641 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
646 Description: llobdstat fix and enhancement
647 Details : add a counter to set a limit to how many samples will be returned
648 fix a wildcard in the path to limit to obdfilter stats only
652 Description: lfs check only allowed for root
653 Details : Code cleanups and added a regresion test of non-root use of
658 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
659 Details : Fix hr->hr_index overload.
663 Description: Fix FIEMAP functionality.
667 Description: FC11 patchless client support added
671 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
672 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
676 Description: New acc-sm script and various test-framework enhancments.
680 Description: Previously landed patch was reverted.
684 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
688 Description: Add ldiskfs maxdirsize mount option.
692 Description: Fix mkfs.lustre support for 16TB LUNs.
696 Description: Print mmp_check_interval
697 Details : print mmp_check_interval and make it possible to abort
698 mount operation in case it takes too long.
702 Description: Add cascading_rw to lustre mpi tests.
706 Description: Store SOM data in EA.
707 Details : Also several SOM related cleanups.
711 Description: Add refcount to md_open_data
715 Description: Use refcount on lov_request_set to prevent memory corruption.
719 Description: Statahead: use parent dir's i_mutex for synchronization.
723 Description: loadgen improvements.
727 Description: Enable peer health detection in ksocklnd and ko2iblnd.
731 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
735 Description: Fix race on obd_devcie:md_stats
739 Description: Fix SLES 11 client builds.
740 Details : Resolved uninitalized pointer gcc warnings.
744 Description: Allow non-root access for "lfs check".
745 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
749 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
751 Severity : enhancement
753 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
754 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
758 Description: Sanity test_27p fails.
759 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
760 Sanity tests 27 were cleaned up and improved.
764 Description: Unify req format on client/servers
765 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
766 buffers and swabbers approach..
771 Description: drop join file stuff.
773 Severity : enhancement
775 Description: Add 2.6.27 and SLES11 (patchless client) support
779 Description: Fix in ptlrpc_expire_one_request() to print the signed time
782 Severity : enhancement
784 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
787 Severity : enhancement
789 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
791 Severity : enhancement
793 Description: correctly handle big reply message.
794 Details : send LNet event if reply is bigger than buffer and adjust this buffer
799 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
801 Severity : enhancement
803 Description: Add support for OEL5.
807 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
809 Severity : enhancement
811 Description: "weak-modules" support
812 Details : Implement "weak-modules" support which enables kernel modules
813 to be used with any kernel that implements the same kABI. In
814 order to achieve this modules are now installed in
815 /lib/modules/$(uname -r)/updates/kernel on all distributions.
819 Description: client could not reconnect to OST because of an active request.
820 Details : The client now retries to reconnect to the same server, if a
821 connect request failed with EBUSY or -EAGAIN. Test case 35b
822 conf-sanity.sh has been created.
826 Description: Don't allow make backward step on assignin osc next id.
827 Details : race between allocation next id and ll_sync thread can be cause
828 of set wrong osc next id and can be kill valid ost objects.
832 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
836 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
840 Description: Error handling in osc_statfs_interpret() has been improved.
841 Details : Check in osc_statfs_interpret() for EBADR.
845 Description: Do not send statfs() requests to OSTs disabled by administrator.
846 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
848 Severity : enhancement
850 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
852 Severity : enhancement
854 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
856 Severity : enhancement
858 Description: Enhancement for lfs(1) command to use numeric uid/gid.
862 Description: Improved cleanup procedure for conf-sanity test_46a.
866 Description: Do not start bulk transfer, if client tried to reconnect.
867 Details : Do not start bulk transfer, if client tried to reconnect,
868 but failed due to active rpcs. Check for exp_abort_active_req
869 has been implemented.
873 Description: Prevent inconsistences between linux and lustre mount structures.
874 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
875 Make the sleep interruptible.
877 Severity : enhancement
879 Description: provide server to client comms path
880 Details : server to client communications path via new LDLM_SET_INFO rpc
884 Description: truncate starts GFP_FS allocation under transaction causing deadlock
885 Details : ldiskfs_truncate calls grab_cache_page which may start page
886 allocation under an open transaction. This may lead to
887 calling prune_icache with consequent lustre reentrance.
891 Description: send correctly lsm on open replay
892 Details : MDS is trust to LSM size on replay open, but client can set wrong size
897 Description: enable client interface failover
898 Details : When a child reconnects from another NID, properly update export
899 nid hash position and ldlm reverse import.
901 Severity : enhancement
903 Description: Communicate OST degraded/readonly state via statfs to MDS
904 Details : Flags in the statfs returned from OSTs indicate whether the
905 OST is in a degraded RAID state, or if the filesystem has
906 turned read-only after a filesystem error is detected.
911 Description: don't panic if EPROTO was hit when reading symlink
912 Details : correctly handling request reference in error cases.
914 Severity : enhancement
916 Description: Add LustreNetLink, a kernel-userspace communcation path.
918 Severity : enhancement
920 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
925 Description: access to llog context before init.
926 Details : move handling CATALOGS file at osc layer and forbid access to llog
930 Frequency : rare, only when client is writing to OST on local node
932 Description: Avoid deadlock for local client writes
933 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
934 memory freeing context. This allows OST threads to set the
935 PF_MEMALLOC flag on task structures in order to allocate memory
936 from reserved pools and complete IO.
937 Use GFP_HIGHUSER for OST allocations for non-local client writes,
938 so that the OST threads generate memory pressure and allow
939 inactive pages to be reclaimed.
941 Severity : enhancement
943 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
948 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
949 Details : move ldlm namespace creation in setup phase to avoid grab
950 _lprocfs_lock with cli_sem held.
954 Description: Temporarily disable grant shrink.
955 Details : Disable the feature for debugging.
958 Frequency : only when formatting test filesystems
960 Description: Unable to run several mkfs.lustre on loop devices at the same time
961 Details : mkfs.lustre returns error 256 on the concurrent loop devices
962 formatting. The solution is to proper handle the error.
964 Severity : enhancement
966 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
968 Severity : enhancement
970 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
975 Description: resolve race between obd_disconnect and class_disconnect_exports
976 Details : if obd_disconnect will be called to already disconnected export he
977 forget release one reference and osc module can't unloaded.
979 Severity : enhancement
981 Description: Allow tuning service thread via /proc
982 Details : For each service a new
983 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
984 created that can be used to set min/max thread counts, and get the
985 current number of running threads.
990 Descriptoin: don't return error if have partially created objects for file.
991 Details : lov_update_create_set uses set->set_success as index for created
992 objects, so if some requests will be failed, they will have hole
993 at end of array and qos_shrink_lsm for allocate correct lsm.
995 Severity : enhancement
997 Description: Update OFED support to 1.4.1
1002 Description: Reduce small size read RPC
1003 Details : Set read-ahead limit for every file and only do read-ahead when
1004 available read-ahead pages are bigger than 1M to avoid small size
1007 Severity : enhancement
1009 Description: move AT tunable parameters for more consistent usage
1010 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1012 Severity : enhancement
1014 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1015 Details : allow skip disconnected ost for send statfs request and hide error
1021 Description: fix racy locking of mballoc block bitmaps causing BUG
1022 Details : The locking of the mballoc buddy bitmap and the in-memory
1023 block bitmap was using two different spin locks in some
1024 cases. This made it possible to incorrectly access the
1025 mballoc bitmap while another process was modifying it,
1026 causing a sanity assertion to fail. While no on-disk corruption
1027 was reported, there was some risk of this happening.
1030 Frequency : rare, on llog test 6
1032 Descriptoin: don't allow connect to already connected import
1033 Details : allowing connect to already connected import is hide connecting problem.
1038 Descriptoin: allow kill process which wait statahead result
1039 Details : for some reasons 'ls' can stick in waiting result from statahead,
1040 in this case need way for kill this process.
1042 Severity : enhancement
1044 Description: Add state history info file, enhance import info file
1045 Details : Track import connection state changes in a new osc/mdc proc file;
1046 add overview-type data to the osc/mdc import proc file.
1048 Severity : enhancement
1050 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1052 Severity : enhancement
1054 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1059 Descriptoin: OST grant too much space to client even there are not enough space.
1060 Details : Client will shrink its grant cache to OST if there are no write
1061 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1062 this grant cache if there are already not enough avaible space
1063 (left_space < total_clients * 32M).
1066 Frequency : start MDS on uncleanly shutdowned MDS device
1068 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1069 Details : stay in waiting mds<>ost recovery finished produce random bugs
1070 due race between two ll_sync thread for one lov target. send
1071 ACTIVATE event only if connect realy finished and import have
1075 Frequency : rare, connect and disconnect target at same time
1077 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1078 Details : don't call obd_disconnect under lov_lock. this long time
1079 operation and can block ptlrpcd which answer to connect request.
1084 Descriptoin: don't lose wakeup for imp_recovery_waitq
1085 Details : recover_import_no_retry or invalidate_import and import_close can
1086 both sleep on imp_recovery_waitq, but we was send only one wakeup
1090 Frequency : always with long access acl
1092 Descriptoin: mds can't pack reply with long acl.
1093 Details : mds don't control size of acl but they limited by reint/getattr
1096 Severity : enhancement
1098 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1100 Severity : enhancement
1102 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1105 Frequency : start MDS on uncleanly shutdowned MDS device
1107 Descriptoin: aborting recovery hang on MDS
1108 Details : don't throttle destroy RPCs for the MDT.
1111 Frequency : on remount
1113 Description: external journal device not working after the remount
1114 Details : clear dev_rdonly flag for external journal devices in
1120 Description: shutdown vs evict race
1121 Details : client_disconnect_export vs connect request race.
1122 if client will evicted at this time - we start invalidate
1123 thread without referece to import and import can be freed
1127 Frequency : rare, need acl's on inode.
1129 Description: client can't handle ost additional correctly
1130 Details : if ost was added after client connected to mds client can have
1131 hit lnet_try_match_md ... to big messages to wide striped files.
1132 in this case need teach client to handle config events about add
1133 lov target and update client max ea size at that event.
1135 Severity : enhancement
1137 Description: Changelogs
1138 Details : Changelogs are a lightweight mechanism to track filesystem
1139 metadata and namespace changes. The changelog is recorded
1140 permanently on the MDTs, and is periodically "consumed" / purged
1141 when records are no longer needed.
1143 Severity : enhancement
1145 Description: compact fld format with extents
1146 Details : Store range of seq rather than every seq in FLD. Seq
1147 controller update FLD rather than clients. In Case of CMD, mdt0
1148 has FLD, all other metadata server act as non persistent proxy
1149 for FLD queries and cache fld entries in fld cache.
1154 Description: don't skip ost target if they assigned to file
1155 Details : Drop slow OSCs if we can, but not for requested start idx.
1156 This means "if OSC is slow and it is not the requested
1157 start OST, then it can be skipped, otherwise skip it only
1158 if it is inactive/recovering/out-of-space.
1162 Description: more cleanup in mds_lov
1163 Details : not send LOV EA under replay, we can't know about they size at this
1164 time. Don't allow client connect to mds before any ost connected,
1165 for avoid problems with LOV EA size and returning EIO to client.
1167 Severity : enhancement
1169 Description: Interoperability at server side (Disk interoperability)
1171 Severity : enhancement
1173 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1175 Severity : enhancement
1177 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1179 Severity : enhancement
1181 Description: New client IO stack (CLIO).
1183 Severity : enhancement
1185 Description: Commit on sharing. Eliminate inter-client dependencies between
1186 uncommitted transactions by doing transaction commits.
1187 Thereby clients may recovery independently.
1190 Frequency : Create a symlink file with a very long name
1192 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1193 Details : If there is no extra space in the request for early cancels,
1194 ldlm_req_handles_avail() returns 0 instead of a negative value.
1196 Severity : enhancement
1198 Description: Add /proc entry for import status
1199 Details : The mdc, osc, and mgc import directories now have
1200 an import directory that contains useful import data for debugging
1201 connection problems.
1203 Severity : enhancement
1205 Description: Re-disable certain /proc logging
1206 Details : Enable and disable client's offset_stats, extents_stats and
1207 extents_stats_per_process stats logging on the fly.
1210 Frequency : Only on FC kernels 2.6.22+
1212 Description: oops in statahead
1213 Details : Do not drop reference count for the dentry from VFS when lookup,
1214 VFS will do that by itself.
1216 Severity : enhancement
1218 Description: Generic /proc file permissions
1219 Details : Set /Proc file permissions in a more generic way to enable non-
1220 root users operate on some /proc files.
1224 Description: Hitting mdc_commit_close() ASSERTION
1225 Details : Properly handle request reference release in
1226 ll_release_openhandle().
1230 Description: quota recovery deadlock during mds failover
1231 Details : This patch includes att18982, att18236, att18237 in bz14840.
1233 1. fix osts hang when mds does failover with quotaon
1234 2. prevent watchdog storm when osts threads wait for the
1239 Frequency : only patchless client
1240 Description: add workaround for race between add/remove dentry from hash
1242 Severity : enhancement
1244 Description: Allow OST glimpses to return PW locks
1248 Description: LBUG when llog conf file is full
1249 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1253 Description: Prevent import from entering FULL state when server in recovery
1257 Description: service mount cannot take device name with ":"
1258 Details : Only when device name contains ":/" will mount treat it as
1264 Description: replace ptlrpcd with the statahead thread to interpret the async
1265 statahead RPC callback
1269 Frequency : on recovery
1270 Description: I/O failures after umount during fail back
1271 Details : if client reconnected to restarted server we need join to recovery
1272 instead of find server handler is changed and process self
1273 eviction with cancel all locks.
1275 Severity : enhancement
1277 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1279 Severity : enhancement
1281 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1283 Severity : enhancement
1285 Description: Upcall on Lustre log has been dumped
1286 Details : Allow for a user mode script to be called once a Lustre log has
1287 been dumped. It passes the filename of the dumped log to the
1288 script, the location of the script can be specified via
1289 /proc/sys/lnet/debug_log_upcall.
1294 Description: avoid idr_remove called for id which is not allocated.
1295 Details : Move assigment s_dev for clustered nfs to end of initialization,
1296 to avoid problem with error handling.
1301 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1302 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1303 into CONN_USED_HASH and this prodice warning when put connection
1304 again in unused hash.
1307 Severity : enhancement
1309 Description: File striping can now be set to use an arbitrary pool of OSTs.
1311 Severity : enhancement
1313 Description: Export bytes_read/bytes_write count on OSC/OST.
1317 Description: Early reply size mismatch, MGC loses connection
1318 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1319 the connect flags are properly negotiated.
1324 Description: quotas are not honored with O_DIRECT
1325 Details : all writes with the flag O_DIRECT will use grants which leads to
1326 this problem. Now using OBD_BRW_SYNC to guard this.
1330 Description: add quota statistics
1331 Details : 1. sort out quota proc entries and proc code.
1332 2. add quota statistics
1334 Severity : enhancement
1336 Description: enable quota support for HEAD.
1340 Description: Properly propagate oinfo flags from lov to osc for statfs
1341 Details : restore missing copy oi_flags to lov requests.
1343 Severity : enhancement
1345 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1346 : l_getgroups(8), lst(8), routerstat(8)
1347 Details : included man pages for llobdstat(8), llstat(8),
1348 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1350 Severity : enhancement
1352 Description: configure's --enable-quota should check the
1353 : kernel .config for CONFIG_QUOTA
1354 Details : configure is terminated if --enable-quota is passed but
1355 : no quota support is in kernel
1359 Description: Remove portals compatibility
1360 Details : Remove portals compatibility, not interoperable with releases
1365 Description: Resolve device initialization race
1366 Details : Prevent proc handler from accessing devices added to the
1367 obd_devs array but yet be intialized.
1369 Severity : enhancement
1371 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1373 Severity : enhancement
1375 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1380 Description: Using wrong pointer in osc_brw_prep_request
1381 Details : Access to array[-1] can produce panic if kernel compiled with
1382 CONFIG_PAGE_ALLOC enabled
1386 Description: Client runs out of low memory
1387 Details : Consider only lowmem when counting initial number of llap pages
1391 Description: *optional* service tags registration
1392 Details : if the "service tags" package is installed on a Lustre node
1393 When the filesystem is mounted, a local-node service tag will
1394 be created. See http://inventory.sun.com/ for more information
1395 about the Service Tags asset management system.
1399 Description: Kernel BUG tries to release flock
1400 Details : Lustre does not destroy flock lock before last reference goes
1401 away. So always drop flock locks when client is evicted and
1402 perform unlock regardless of successfulness of speaking to MDS.
1406 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1410 Description: sanity test 65a fails if stripecount of -1 is set
1411 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1416 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1417 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1420 Severity : enhancement
1422 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1426 Description: do not process already freed flock
1427 Details : flock can possibly be freed by another thread before it reaches
1428 to ldlm_flock_completion_ast.
1432 Description: LBUG during stress test
1433 Details : Need properly lock accesses the flock deadlock detection list.
1437 Description: oops in page fault handler
1438 Details : kernel page fault handler can return two special 'pages' in error
1439 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1443 Description: timeout with invalidate import.
1444 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1445 should be handled by ptlrpcd. This produce long age waiting and
1446 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1448 Severity : enhancement
1450 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1454 Description: enable MGS and MDT services start separately
1455 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1456 a co-located MGS without starting the MGS, which is a complement
1457 to 'nosvc' mount option.
1461 Frequency : after recovery
1462 Description: precreate to many object's after del orphan.
1463 Details : del orphan st in oscc last_id == next_id and this triger growing
1464 count of precreated objects. Set flag LOW to skip increase count
1465 of precreated objects.
1469 Frequency : rare, on clear nid stats
1470 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1471 Details : when clean nid stats sometimes try destroy live entry,
1472 and this produce panic in free.
1476 Description: Stack overflow during MDS log replay
1477 ease stack pressure by using a thread dealing llog_process.
1481 Description: wait until IO finished before start new when do lock cancel.
1482 Details : VM protocol want old IO finished before start new, in this case
1483 need wait until PG_writeback is cleared until check dirty flag and
1484 call writepages in lock cancel callback.
1486 Severity : enhancement
1488 Description: using special macro for print time and cleanup in includes.
1492 Description: mds_mfd_close() ASSERTION(rc == 0)
1493 Details : In mds_mfd_close(), we need protect inode's writecount change
1494 within its orphan write semaphore to prevent possible races.
1498 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1502 Frequency : rare, on shutdown ost
1503 Description: don't hit live lock with umount ost.
1504 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1505 use shrink_dcache_sb instead.
1509 Description: don't panic with use echo client
1510 Details : echo client pass NULL as client nid pointer and this produce null
1511 pointer dereference.
1515 Description: fix build on ppc32
1516 Details : compile code with -m64 flag produce wrong object file for ppc32.
1520 Description: add message levels for liblustreapi
1524 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1525 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1526 will result in an -EAGAIN and in the exit call path we call
1527 alloc_rr() which will with increasing aggressiveness attempt to
1528 aquire precreated objects on the minimum number of required OSCs.
1532 Description: Use old size assignment to avoid deadlock
1533 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1534 the scheduling while holding a spinlock. We do not need locking
1535 for size in ll_update_inode() because size is only updated from
1536 the MDS for directories or files without objects, so there is no
1537 other place to do the update, and concurrent access to such inodes
1538 are protected by the inode lock.
1542 Description: resolve "_IOWR redefined" build error on SLES10
1546 Description: dump the memory debugging after all modules are unloaded to
1547 suppress false negative in conf_sanity test 39
1549 Severity : enhancement
1551 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1556 Description: fix SLES kernel versioning
1557 Details : the kernel version for our SLES 10 kernel did not include a "-"
1558 before the "smp" at the end. while this was not a problem in
1559 general, it did mean that software trying to use the kernel
1560 version to try to detect a vendor specific kernel would fail.
1561 this was most evident by the OFED build scripts.
1565 Description: Don't update lov_desc members until making sure they are valid
1566 Details : When updating lov_desc members via proc fs, need fix their
1567 validities before doing the real update.
1571 Description: don't put request into delay list while invalidate in flight.
1572 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1573 import in flight. this produce timeout for invalidate and sometimes
1574 can cause stale data.
1578 Frequency : on ppc only
1579 Description: not convert ost objects for directory because it's not exist.
1580 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1584 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1586 Severity : enhancement
1588 Description: Update to support 2.6.22.14 vanilla kernel.
1592 Frequency : rare, on recovery
1593 Description: read procfs can produce deadlock in some situation
1594 Details : Holding lprocfs lock which send rpc can produce block for destroy
1595 obd objects and this also block reconnect with -EALREADY. This
1596 isn't fix all lprocfs bugs - but make it rare.
1598 Severity : enhancement
1600 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1603 Frequency : frequent on X2 node
1605 Description: mdc_set_open_replay_data LBUG
1606 Details : Set replay data for requests that are eligible for replay.
1610 Description: lustre_mgs: operation 101 on unconnected MGS
1611 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1612 MGC, and late on MGC cannot successfully connect to MGS and a lot
1613 of the error messages complaining that MGS is not connected.
1617 Frequency : on network error
1618 Description: panic with double free request if network error
1619 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1620 but it's true only for synchronus enqueue, for async enqueue
1621 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1624 Severity : enhancement
1626 Description: client-side metadata stat-ahead during readdir(directory readahead)
1627 Details : perform client-side metadata stat-ahead when the client detects
1628 readdir and sequential stat of dir entries therein
1631 Frequency : on start mds
1633 Description: Implement get_info(last_id) in obdfilter.
1636 Frequency : occasional
1638 Description: Correctly check stale fid, not start epoch if ost not support SOM
1639 Details : open with flag O_CREATE need set old fid in op_fid3 because
1640 op_fid2 was overwritten with new generated fid, but mds can answer
1641 with one of these two fids and both is not stale. Setattr
1642 incorrectly started an epoch and assume will be called
1643 done_writing, but without SOM done_writing ever being called.
1646 Frequency : rare, depends on device drivers and load
1648 Description: MDS or OSS nodes crash due to stack overflow
1649 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1650 In some cases, in conjunction with device drivers that use a lot
1651 of stack the MDS (or possibly OSS) service threads could overflow
1652 the stack. One change which was identified to consume additional
1653 stack has been reworked to avoid the extra stack usage.
1656 Frequency : occasional
1658 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1659 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1660 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1661 oscc_has_objects and see OSC has no objests and call
1662 oscc_internal_create to resend create request.
1664 Severity : enhancement
1666 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1668 Severity : enhancement
1670 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1673 Frequency : very rare
1675 Description: Fix replay if there is an un-replied request and open
1676 Details : In some cases, older replay request will revert the
1677 mcd->mcd_last_xid on MDS which is used to record the client's
1678 latest sent request.
1680 Severity : enhancement
1682 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1684 Severity : enhancement
1686 Description: Add rhel5 support to HEAD.
1688 Serverity : enhancement
1690 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1695 Description: Don't allow skipping OSTs if index has been specified.
1696 Details : Don't allow skipping OSTs if index has been specified, make
1697 locking in internal create lots better.
1701 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1702 Details : access to bitfield in structure is always rounded to long
1703 and this produce problem with not atomic change any bit.
1707 Description: Lustre make rpms failed.
1708 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1709 builds Lustre rpms from tarball.
1712 Frequency : rare on shutdown ost
1714 Description: If llog cancel was not send before clean_exports phase, this can
1715 produce deadlock in llog code.
1716 Details : If llog thread has last reference to obd and call class_import_put
1717 this produce deadlock because llog_cleanup_commit_master wait when
1718 last llog_commit_thread exited, but this never success because was
1719 called from llog_commit_thread.
1723 Description: allow userland application know is lost one of stripes.
1724 Details : fill lvb_blocks with error code on ost and return it to
1725 application if error flag found.
1729 Description: NULL lov_tgts causing MDS oops
1730 Details : more safe checks for NULL lov_tgts for avoid oops.
1732 Severity : enhancement
1734 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1738 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1739 Details : If an OST is down the MDS will hang indefinitely in
1740 obd_statfs_rqset() waiting for the statfs data. While for
1741 MDS QOS usage of statfs, it should not stuck in waiting.
1743 Severity : enhancement
1745 Description: remote_acl support
1746 Details : Support ACL-based permission check for remote user.
1747 Support setfacl/getfacl for remote user with the utils
1748 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1749 the system "{s,g}etfacl" utils.
1751 Severity : enhancement
1753 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1755 Severity : enhancement
1757 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1761 Description: make mgs_setparam() handle fsname containing dash
1762 Details : fsname containing a dash does not work with lctl conf_param
1764 Severity : enhancement
1766 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1768 Severity : enhancement
1770 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1774 Frequency : rare, at shutdown
1775 Description: access already free / zero obd_namespace.
1776 Details : if client_disconnect_export was called without force flag set,
1777 and exist connect request in flight, this can produce access to
1778 NULL pointer (or already free pointer) when connect_interpret
1779 store ocd flags in obd_namespace.
1783 Frequency : only at startup
1784 Description: not alloc memory with spinlock held.
1785 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1786 if any spinlock held.
1788 Severity : enhancement
1790 Description: make lustre randomly fail allocating memory
1791 Details : Make lustre randomly failed allocating memory for testing purpose.
1793 Severity : enhancement
1795 Description: lost problems with lov objid file
1796 Details : Fixes some scability and access to not inited memory problems
1797 in work with lov objdid file.
1802 Description: lfs find does not continue on file error
1803 Details : Continue other files processing when a file/dir is absent.
1807 Description: Inconsistent usage of lustre_pack_reply()
1808 Details : Standardize the usage of lustre_pack_reply() such that it
1809 always generate a CERROR on failure.
1812 Frequency : occasional
1814 Description: MDS hang or stay in waiting lock
1815 Details : If client receive lock with CBPENDING flag ldlm need send lock
1816 cancel as separate rpc, to avoid situation when cancel request
1817 can't processed due all i/o threads stay in wait lock.
1821 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1825 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1829 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1833 Description: Update to RHEL4 latest kernel.
1835 Severity : enhancement
1837 Description: Build SLES10 patchless client fails
1838 Details : The configure was broken by run ./configure with
1839 --with-linux-obj=.... argument for patchless client. When the
1840 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1841 can't search header adequately. Use absolute path such as
1842 -I($LINUX)/include instead.
1846 Description: interrupt oig_wait produce painc on resend.
1847 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1848 private set, and this produce situation when rq_ptlrpcd_data not
1849 copyed to new allocated request and triggered LBUG on assert
1850 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1851 ptlrpcd set if request is changed and can be safety changed to use
1856 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1857 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1858 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1859 these macros and replace them with their difinitions instead.
1863 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1864 Details : Update which_patch & target file for SLES10 latest kernel.
1867 Description: Remove client patches from SLES 10 kernel.
1868 Details : This causes SLES 10 clients to behave as patchless clients
1869 even on a Lustre-patched (server) kernel.
1871 Severity : enhancement
1873 Description: self-adjustable client's lru lists
1874 Details : use adaptive algorithm for managing client cached locks lru
1875 lists according to current server load, other client's work
1876 pattern, memory activities, etc. Both, server and client
1877 side namespaces provide number of proc tunables for controlling
1880 Severity : enhancement
1882 Description: light-weight GSS support
1883 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1884 and reduce performance overhead.
1886 Severity : enhancement
1888 Description: Linux keyring support
1889 Details : Support using service of Linux keyring for Lustre GSS internal
1890 context refresh/cache mechanism.
1894 Description: Fix errors in lfs documentation
1895 Details : Fixes man pages
1899 Description: don't use GFP_* in generic Lustre code.
1900 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1904 Description: obdclass is limited by single OBD_ALLOC(idarray)
1905 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1909 Description: warnings with build patchless client with vanila 2.6.19 and up
1910 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1911 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1915 Frequency : Only for SLES
1916 Description: sanity_quota fail test_1
1917 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1918 file, which makes TSTID[2] inunique.
1922 Frequency : Only for SLES9
1923 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1924 Details : Update target/ChangeLog/which_patch .
1928 Frequency : Only for test_10 in sanity-quota.sh
1929 Description: a bug in quota test code
1930 Details : checking a test flag in a wrong place
1934 Frequency : Only for test_13 in sanity-quota.sh
1935 Description: a bug in quota test script
1936 Details : shouldn't check overfull allocated quota
1940 Description: jbd statistics
1941 Details : Port older jbd statistics patch for sles10
1943 Severity : enhancement
1945 Description: Add printing inode info into message about error in writepage.
1950 Description: reply_lock_interpret crash due to race with it and lock cancel.
1951 Details : Do not replay locks that are being cancelled. Do not reference
1952 locks by their address during replay, just by their handle.
1956 Frequency : When flocks are used.
1957 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1958 Details : Flock locks might destroy just granted lock if it could be merged
1959 with another existing flock, this is done in completion handler,
1960 so teach ldlm_cli_enquque_fini that this is a valid case for
1966 Description: Oops in read and write path when failing to allocate lock.
1967 Details : Check if lock allocation failed and return error back.
1971 Description: lstripe command fails for valid OST index
1972 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1973 instead of lov->desc.ld_active_tgt_count.
1977 Description: sanity.sh failed test 103
1978 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1983 Description: df doesn't work properly if diskfs blocksize != 4K
1984 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1988 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1989 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1991 Severity : enhancement
1993 Description: Adaptive timeouts
1994 Details : RPC timeouts adapt to changing server load and network
1995 conditions to reduce resend attempts and improve recovery time.
1999 Description: llapi_file_create() does not allow some changes
2000 Details : add llapi_file_open() that allows specifying the mode and
2001 open flags, and also returns an open file handle.
2005 Description: merge and cleanup kernel patches.
2006 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2010 Description: Add journal checksum support.(Kernel part)
2011 Details : The journal checksum feature adds two new flags i.e
2012 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2013 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2014 indicates that the commit block contains the checksum for
2015 the blocks described by the descriptor blocks. Now commit
2016 record can be sent to disk without waiting for descriptor
2017 blocks to be written to disk. This behavior is controlled
2018 using JBD2_FEATURE_ASYNC_COMMIT flag.
2022 Description: OSS needs mutliple precreate threads
2023 Details : Add ability to start more than one create thread per OSS.
2027 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2028 Details : execute lfs setstripe on client
2032 Description: mds_obd_create error creating tmp object
2033 Details : When the user sets quota on root, llog will be affected and can't
2034 create files and write files.
2037 Frequency : Always on ia64 patchless client, and possibly others.
2039 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2040 Details : This allows the patchless client to be loaded on architectures
2041 without this export.
2045 Description: RedHat Update kernel for RHEL5
2046 Details : Add the kernel config file for RHEL5.
2050 Description: RedHat Update kernel for RHEL5
2051 Details : Modify the kernel config file more closer RHEL5.
2055 Description: Build failure against Centos5 (RHEL5)
2056 Details : Define PAGE_SIZE when it isn't present.
2060 Description: test_7 on sanity-quota.sh can't be used on separate machines
2061 Details : add facet to handle it
2065 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2066 Details : with NFS, the anon dentry's parent was set to itself in
2067 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2068 corresponding dentry other than use rec->ur_name.
2070 Severity : enhancement
2072 Description: lfs setstripe enhancement
2073 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2077 Description: enable data checksumming by default
2078 Details : enable checksum by default, allow --disable-checksum
2079 configure option and "-o nochecksum" mount option. Checksums
2080 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2084 Description: lustre support for RHEL5
2085 Details : Add support for RHEL5.
2089 Description: Client eviction due to failover config
2090 Details : after a connection loss, the lustre client should attempt to
2091 reconnect to the last active server first before trying the
2092 other potential connections.
2096 Description: when mds and osts use different quota unit(32bit and 64bit),
2097 quota will be released repeatly.
2098 Details : void sending multiple quota reqs to mds, which will keep the
2099 status between the reqs.
2103 Description: osts not allocated evenly to files
2104 Details : change the condition to increase offset_idx
2107 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2109 Description: Data corruption for OSTs that are formatted larger than 2TB
2111 Details : When generating the bio request for lustre file writes the
2112 sector number would overflow a temporary variable before being
2113 used for the IO. The data reads correctly from Lustre (which
2114 will overflow in a similar manner) but other file data or
2115 filesystem metadata may be corrupted in some cases.
2119 Description: Tune the kernel for good SCSI performance.
2120 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2121 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2126 Description: rewrite ext2-derived code in obdclass/uuid.c
2127 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2128 from scratch preserving functionality.
2132 Description: "lfs find -obd UUID" prints directories
2133 Details : "lfs find -obd UUID" will return all directory names instead
2134 of just file names. It is incorrect because the directories
2135 do not reside on the OSTs.
2139 Description: use i_size_read and i_size_write in 2.6 port
2140 Details : replace inode->i_size access with i_size_read/write()
2143 Frequency : when removing large files
2145 Description: scheduling issue during removal of large Lustre files
2146 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2147 It causes scheduling issues when removing large files (17TB in the
2151 Frequency : only with liblustre clients on XT3
2153 Description: evictions taking too long
2154 Details : allow llrd to evict clients directly on OSTs
2157 Frequency : only on ppc
2159 Description: /proc/fs/lustre/devices broken on ppc
2160 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2161 We should make sure the type of 'index' is loff_t and then cast
2162 explicitly as needed below. Do not assign an explicitly cast
2166 Frequency : only for rhel5
2168 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2169 Details : Modify the target file & which_kernel.
2171 Severity : enhancement
2173 Description: omit set fsid for export NFS
2174 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2175 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2179 Frequency : Always, for kernels after 2.6.16
2180 Description: Fix warning idr_remove called for id=.. which is not allocated.
2181 Details : Last kernels save old s_dev before kill super and not allow
2182 to restore from callback - restore it before call kill_anon_super.
2186 Description: buffer overruns could theoretically occur
2187 Details : llapi_semantic_traverse() modifies the "path" argument by
2188 appending values to the end of the origin string, and a
2189 overrun may occur. Adding buffer overrun check in liblustreapi.
2193 Description: Fix error on 'ls .' at the top of the Lustre mount.
2194 Details : Don't revalidate dentry if it is a root dentry.
2198 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2199 Details : Modify vm-tunables-rhel4.patch.
2203 Description: Kernel config for 2.6.18-vanilla.
2204 Details : Modify targets/2.6-vanilla.target.in.
2205 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2206 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2207 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2208 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2212 Description: improve handling recoverable errors
2213 Details : if request processig with error which can be recoverable on server
2214 request should be resend, otherwise page released from cache and
2219 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2220 Details : Modify target file & which_patch.
2221 A flaw was found in the IA32 system call emulation provided
2222 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2223 value could be stored in the %RAX register, which could trigger an
2224 out-of-bounds system call table access. An untrusted local user
2225 could exploit this flaw to run code in the kernel
2226 (ie a root privilege escalation). (CVE-2007-4573).
2230 Description: O_DIRECT bypasses client statistics.
2231 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2232 read_ahead_stats were updated. copied the stats section from
2233 osc_send_oap_rpc() into async_internal().
2237 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2241 Description: change order of libsysio includes
2242 Details : '#include sysio.h' should always come before '#include xtio.h'
2246 Description: handle "serious error: objid * already exists" more gracefully
2247 Details : If LAST_ID value on disk is smaller than the objects existing in
2248 the O/0/d* directories, it indicates disk corruption and causes an
2249 LBUG(). If the object is 0-length, then we should use the existing
2250 object. This will help to avoid a full fsck in most cases.
2252 Severity : enhancement
2254 Description: adapt the lustre_config script to support the upgrade case
2255 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2256 server targets to 1.6 in parallel.
2260 Description: To avoid grant space > avaible space when the disk is almost
2261 full. Without this patch you might see the error "grant XXXX >
2262 available" or some LBUG about grant, when the disk is almost
2264 Details : In filter_check_grant, for non_grant cache write, we should
2265 check the left space by if (*left > ungranted + bytes), instead
2266 of (*left > ungranted), because only we are sure the left space
2267 is enough for another "bytes", then the ungrant space should be
2268 increase. In client, we should update cl_avail_grant only there
2269 is OBD_MD_FLGRANT in the reply.
2273 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2274 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2275 described in CVE-2007-4573. This problem could allow untrusted
2276 local users to gain root access.
2279 Frequency : when using O_DIRECT and quotas
2281 Description: Incorrect file ownership on O_DIRECT output files
2282 Details : block usage reported by 'lfs quota' does not take into account
2283 files that have been written with O_DIRECT.
2288 Description: touch file failed when fs is not full
2289 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2290 otherwise we can get ENOSP while fs is not full.
2294 Description: parallel lock callbacks
2295 Details : Instead of sending blocking and completion callbacks as separated
2296 requests, adding them to a set and sending in parallel.
2299 Frequency : only for Cray XT3
2300 Bugzilla : 12829/13455
2301 Description: Changing primary group doesn't change the group lustre assigns to
2303 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2304 overridden with the primary group provided by the group upcall,
2305 whereas the supplied fsgid can be trusted if it is in the list of
2306 supplementary groups returned by the group upcall.
2308 Severity : enhancement
2310 Description: Allow masking D_WARNING, D_ERROR messages from console
2311 Details : Console messages can now be disabled via lnet.printk.
2315 Description: User code with malformed file open parameter crashes client node
2316 Details : Before packing join_file req, all the related reference should be
2317 checked carefully in case some malformed flags cause fake
2318 join_file req on client.
2322 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2323 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2324 mode is equal to the granted mode. However, it should also check
2325 that LDLM_FL_CP_REQD is not set.
2329 Description: Heavy nfs access might result in deadlocks
2330 Details : After ELC code landed, it is now improper to enqueue any mds
2331 locks under och_sem, because enqueue might want to decide to
2332 cancel open locks for same inode we are holding och_sem for.
2336 Description: Client eviction while running blogbench
2337 Details : A lot of unlink operations with concurrent I/O can lead to a
2338 deadlock causing evictions. To address the problem, the number of
2339 oustanding OST_DESTROY requests is now throttled to
2340 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2341 callbacks are processed in priority.
2345 Description: enable ACLs on MDS by default
2346 Details : ACLs must be enabled on MDS by default.
2349 Frequency : PPC/PPC64 only
2351 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2352 Details : fix duplicated definitions between asm-ppc64/types.h and
2353 lustre_types.h on PPC.
2356 Frequency : PPC/PPC64 only
2358 Description: asm-ppc/segment.h does not exist
2359 Details : fix compile issue on PPC.
2363 Description: better handle error messages in extents code
2366 Frequency : RHEL4 only
2368 Description: mkfs is very slow on IA64/RHEL4
2369 Details : A performance regression has been discovered in the MPT Fusion
2370 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2371 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2372 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2374 Severity : enhancement
2376 Description: SNMP support enhancement
2377 Details : Adding total number of sampled request for an MDS node in snmp
2380 Severity : enhancement
2382 Description: Optimize ldlm waiting list processing for PR extent locks
2383 Details : When processing waiting list for read extent lock and meeting read
2384 lock that is same or wider to it that is not contended, skip
2385 processing rest of the list and immediatelly return current
2386 status of conflictness, since we are guaranteed there are no
2387 conflicting locks in the rest of the list.
2391 Description: Time out and refuse to reconnect
2392 Details : When the failover node is the primary node, it is possible
2393 to have two identical connections in imp_conn_list. We must
2394 compare not conn's pointers but NIDs, otherwise we can defeat
2395 connection throttling.
2399 Description: port llog fixes from b1_6 into HEAD
2400 Details : Port llog reference couting and some llog cleanups from b1_6
2401 (bug 10800) into HEAD, for protect from panic and access to already
2402 free llog structures.
2406 Description: Detect stride IO mode in read-ahead
2407 Details : When a client does stride read, read-ahead should detect that and
2408 read-ahead pages according to the detected stride pattern.
2412 Description: data checksumming impacts single node performance
2413 Details : add support for several checksum algorithm. Currently, only CRC32
2414 and Adler-32 are supported. The checksum type can be changed on
2415 the fly via /proc/fs/lustre/osc/*/checksum_type.
2419 Description: use adler32 for page checksums
2420 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2425 Description: build for x2 fails
2426 Details : fix compile issue on Cray systems.
2430 Description: Properly match for duplicate locks
2431 Details : Due to different lock order from skiplists code, we need to
2432 traverse entire list for now
2435 Frequency : only on PPC/SLES10
2437 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2438 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2439 be defined. Add a hack in mkfs_lustre.c to work around this compile
2444 Description: LASSERT on MDS when client holding flock lock dies
2445 Details : ldlm pool logic depends on number of granted locks equal to
2446 number of released locks which is not true for flock locks, so
2447 just exclude such locks from consideration.
2451 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2452 Details : Use fsfilt_sync() for both the whole filesystem sync and
2453 individual file sync to eliminate dangerous inode locking
2454 with I_LOCK that can lead to a deadlock.
2458 Description: performance in 1.6.3
2459 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2460 sure that 1MB requests are not fragmented by the block layer.
2462 Severity : enhancement
2464 Description: organize the server-side client stats on per-nid basis
2465 Details : Change the structure of stats under obdfilter and mds to
2473 The "uuid"s file would list the uuids of _active_ exports.
2474 And the clear entry is to clear all stats and stale nids.
2476 Severity : enhancement
2478 Description: eliminate client locks in face of contention
2479 Details : file contention detection and lockless i/o implementation
2480 for contended files.
2484 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2487 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2489 Description: MDS may lose file striping (and hence file data) in some cases
2490 Details : If there are additional extended attributes stored on the MDS,
2491 in particular ACLs, SELinux, or user attributes (if user_xattr
2492 is specified for the client mount options) then there is a risk
2493 of attribute loss. Additionally, the Lustre file striping
2494 needs to be larger than default (e.g. striped over all OSTs),
2495 and an additional attribute must be stored initially in the
2496 inode and then increase in size enough to be moved to the
2497 external attribute block (e.g. ACL growing in size) for file
2502 Description: skiplist implementation simplification
2503 Details : skiplists are used to group compatible locks on granted list
2504 that was implemented as tracking first and last lock of each
2505 lock group the patch changes that to using doubly linked lists
2509 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2510 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2511 because LASSERT on both the data supplied by a client, and the
2512 data on disk is dangerous and incorrect.
2514 Severity : enhancement
2516 Description: Slow truncate/writes to huge files at high offsets.
2517 Details : Directly associate cached pages to lock that protect those pages,
2518 this allows us to quickly find what pages to write and remove
2519 once lock callback is received.
2523 Description: more ldlm soft lockups
2524 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2525 starve other threads from the resource lock for a long time in
2526 case of long waiting queue, so change the debug level from
2527 D_OTHER to the less frequently used D_INFO.
2529 Severity : enhancement
2531 Description: add -gid, -group, -uid, -user options to lfs find
2535 Description: Hung threads in invalidate_inode_pages2_range
2536 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2537 once one is completed. As a result, some RPCs are stuck in the
2538 queue and are never sent.
2542 Description: filter threads hungs on waiting journal commit
2543 Details : Cleanup filter group llog code, then only filter group llog will
2544 be only created in the MDS/OST syncing process.
2548 Description: Procfs and llog threads access destoryed import sometimes.
2549 Details : Sync the import destoryed process with procfs and llog threads by
2550 the import refcount and semaphore.
2552 Severity : enhancement
2554 Description: openlock cache of b1_6 port to HEAD
2559 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2560 Details : If insertion of an extent fails, then discard the inode
2561 preallocation and free data blocks else it can lead to duplicate
2566 Description: don't always update ctime in ext3_xattr_set_handle()
2567 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2568 In some cases the ctime should not be updated, for example for
2569 2.0->1.8 compatibility it is necessary to delete an xattr and it
2570 should not update the ctime.
2574 Bugzilla : 15713/16362
2575 Description: Assertion in iopen_connect_dentry in 1.6.3
2576 Details : looking up an inode via iopen with the wrong generation number can
2577 populate the dcache with a disconneced dentry while the inode
2578 number is in the process of being reallocated. This causes an
2579 assertion failure in iopen since the inode's dentry list contains
2580 both a connected and disconnected dentry.
2584 Description: assertion failure in ldlm_handle2lock()
2585 Details : fix a race between class_handle_unhash() and class_handle2object()
2586 introduced in lustre 1.6.5 by bug 13622.
2591 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2592 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2597 Description: oops in sd_iostats_seq_show()
2598 Details : unloading/reloading the scsi low level driver triggers a kernel
2599 bug when trying to access the sd iostat file.
2604 Description: Kernel panics during QLogic driver reload
2605 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2606 patch, causing memory corruption.
2611 Description: journal_dev option does not work in b1_6
2612 Details : pass mount option during pre-mount.
2614 Severity : enhancement
2616 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2617 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2618 extent information of a file. It can be used to map logical blocks
2619 in a file to physical blocks in the block device.
2623 Description: LDLM soft lockups - improvement
2624 Details : It is be possible to send the lock handle along with each read
2625 or write request because the client is already doing a lock match
2626 itself so there isn't any reason the OST should have to re-do that
2630 Frequency : only on Cray X2
2632 Description: X2 build failures
2633 Details : fix build failures on Cray X2.
2637 Description: xid & resent requests
2638 Details : Initialize RPC XID from clock at startup (randomly if clock is
2641 Severity : enhancement
2643 Description: Add lustre_start utility to start or stop multiple Lustre servers
2648 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2649 Details : In case of memory pressure, list_del() can be called twice on
2650 req->rq_history_list, causing a kernel oops.
2654 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2655 Details : fix stack overflow in the distributed lock manager by defering
2656 export eviction after a failed AST to the elt thread instead of
2657 handling it in the dlm interpret routine.
2661 Description: Convert some comments to new format.
2662 Details : Update documenting comments to match doxygen conventions.
2666 Description: Grammar fixes.
2667 Details : A couple of trivial spelling fixes.
2671 Description: OSD_COUNTERS-mandatory
2672 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2673 to simplify the code.
2677 Description: simplify cmm_device freeing logic.
2678 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2682 Description: Add lockdep support to dt_object_operations locking interface.
2683 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2684 parameter indicating lock ordering. Update mdd code to use new
2689 Description: Introduce failloc constants for lockless IO tests.
2690 Details : Add two new failloc constants to test lockless IO. Only one of
2691 them in implemented---another is checked in yet to be landed
2696 Description: Add lockdep support for inode mutex.
2697 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2698 in the situations where Lustre has to lock more than one inode
2703 Description: Add optional invariants checking support.
2704 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2705 configure switch. This macro is to be used for consistency and
2706 sanity checks that are too expensive to be left in `production'
2711 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2712 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2717 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2718 Details : Use spin_lock_nested() in (the only) situation where more than
2719 one ldlm_lock is locked simultaneously. Also, fix possible
2720 dead-lock in ldlm_lock_change_resource() by enforcing particular
2725 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2726 Details : Instead of specifying each ldlm_lock call-back through separate
2727 parameter, wrap them into struct ldlm_callback_suite.
2731 Description: Kill join_lru obd method and its callers.
2732 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2733 in memory---a requirement implemented through ->o_join_lru() obd
2734 method in HEAD. Remove this method and its users.
2738 Description: Add asynchronous ldlm ENQUEUE completion handler.
2739 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2740 case that stock ldlm_completion_ast() cannot handle as it waits
2741 until lock is granted. Introduce new ldlm_completion_ast_async()
2742 for this. Also comment ldlm_completion_ast().
2746 Description: ldlm_error <-> errno conversion.
2747 Details : Add functions to map (rather arbitrary) between LDLM error codes
2748 and standard errno values. CLIO needs this to prevent LDLM specific
2749 constants from escaping ldlm and osc.
2753 Description: Kill unused ldlm_handle2lock_ns() function.
2754 Details : Kill unused ldlm_handle2lock_ns() function.
2758 Description: Add lu_ref support to ldlm_lock
2759 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2760 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2761 and ldlm_resource. LDLM interface has to be changed, because code that
2762 releases a reference on a lock, has to "know" what reference this is.
2763 In the most frequent case
2765 lock = ldlm_handle2lock(handle);
2767 LDLM_LOCK_PUT(lock);
2769 no changes are required. When any other reference (received _not_ from
2770 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2771 instead of LDLM_LOCK_PUT().
2773 Arguably, changes are pervasive, and interface requires some discipline
2774 for proper use. On the other hand, it was very instrumental in finding
2775 a few leaked lock references.
2779 Description: Add ldlm_lock_addref_try().
2780 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2781 attempts to addref a lock that might be being canceled
2786 Description: Add ldlm_weigh_callback().
2787 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2788 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2789 hard-coded `number of pages' logic.
2793 Description: Add lockdep annotations to llog code.
2794 Details : Use appropriately tagged _nested() locking calls in the places
2795 where llog takes more than one ->lgh_lock lock.
2799 Description: Add loi_kms_set().
2800 Details : Wrap kms updates into a helper function.
2804 Description: Constify instances of struct lsm_operations.
2805 Details : Constify instances of struct lsm_operations.
2809 Description: lu_conf support.
2810 Details : On a server, a file system object is uniquely identified
2811 by a fid, which is sufficient to locate and load all object
2812 state (inode). On a client, on the other hand, more data are
2813 necessary instantiate an object. Change lu_object_find() and
2814 friends to take additional `lu_conf' argument describing object.
2815 Typically this includes layout information.
2819 Description: lu_context fixes.
2820 Details : Introduce new lu_context functions that are needed on the client
2821 side, where some system threads (ptlrpcd) are shared by multiple
2822 modules, and so cannot be stopped during module shutdown.
2826 Description: Add start and stop methods to lu_device_type_operations.
2827 Details : Introduce two new methods in lu_device_type_operations, that are
2828 invoked when first instance of a given type is created and last one
2829 is destroyed respectively. This is need by CLIO.
2833 Description: Add lu_ref support to struct lu_device.
2834 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2835 track leaked references.
2839 Description: Introduce lu_kmem_descr.
2840 Details : lu_kmem_descr and its companion interface allow to create
2841 and destroy a number of kmem caches at once.
2845 Description: Fix lu_object finalization race.
2846 Details : Fix a race between lu_object_find() finding an object and its
2847 concurrent finalization. This race is (most likely) not possible
2848 on the server, but might happen on the client.
2852 Description: Introduce lu_ref interface.
2853 Details : lu_ref is a debugging module allowing to track references to
2854 a given object. It is quite cpu expensive, and has to be
2855 explicitly enabled with --enable-lu_ref. See usage description
2860 Description: Factor lu_site procfs stats into a separate function.
2861 Details : Separate lu_site stats printing code into a separate function
2862 that can be reused on a client.
2866 Description: Constify instances of struct {lu,dt,md}_device_operations.
2867 Details : Constify instances of struct {lu,dt,md}_device_operations.
2871 Description: Introduce struct md_site and move meta-data specific parts of
2872 struct lu_site here.
2873 Details : Move md-specific fields out of struct lu_site into special struct
2874 md_site, so that lu_site can be used on a client.
2878 Description: Kill mdd_lov_destroy().
2879 Details : Remove unused mdd code.
2883 Description: Add st_block checking to multistat.c.
2884 Details : Add st_block checking to multistat.c.
2888 Description: Add lu_ref support to struct obd_device.
2889 Details : Add lu_ref tracking to obd_device.
2893 Description: Kill obd_set_fail_loc().
2894 Details : Remove unused code.
2898 Description: Add special type for ptlrpc_request interpret functions.
2899 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2900 there. Actual usage will be in CLIO.
2904 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2905 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2906 doesn't work with lockdep.
2910 Description: Add rwv.c test program.
2911 Details : New testing program exercising readv(2) and writev(2) (Qian).
2915 Description: Add sendfile.c test program.
2916 Details : New testing program exercising sendfile(2) (Jay).
2920 Description: Ratelimit a message that can be very frequent.
2921 Details : Ratelimit a memory allocation failure message that can
2926 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2927 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2931 Description: Liblustre build fixes.
2932 Details : Liblustre build fixes.
2936 Description: libcfs: add cfs_{need,cond}_resched() interface.
2937 Details : libcfs: add cfs_{need,cond}_resched() definition and
2938 implementations for Linux, NT, and liblustre.
2940 Severity : enhancement
2942 Description: More exported tunables for mballoc
2943 Details : Add support for tunable preallocation window and new tunables for
2944 large/small requests
2948 Description: Detect corruption of block bitmap and checking for preallocations
2949 Details : Checks validity of on-disk block bitmap. Also it does better
2950 checking of number of applied preallocations. When corruption is
2951 found, it turns filesystem readonly to prevent further corruptions.
2955 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2956 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2957 window to 0. If it does hit cache miss, and read-pattern is still
2958 stride-io mode, does not reset the stride window, but also does not
2959 increase the stride window length in this case.
2963 Frequency : only for big-endian servers
2964 Description: Check if system is big-endian while mounting fs with extents feature
2965 Details : Mounting a filesystem with extents feature will fail on big-endian
2966 systems since ext3-based ldiskfs is not supported on big-endian
2967 systems. This can be over-riden with "bigendian_extents" mount option.
2969 Severity : enhancement
2971 Description: The root squash functionality
2972 Details : A security feature, which is to prevent users from being able
2973 to mount lustre on their desktop, run as root, and delete
2974 all of the files in the filesystem. The goal is accomplished by
2975 remapping user id (UID) and group id (GID) of the root user to
2976 a UID and GID specified by the system administartor via Lustre
2977 configuration management server (MGS). The functionality also
2978 allows to specify sets of clients for which the remapping does
2983 Description: Excessive recovery window
2984 Details : With AT enabled, the recovery window can be excessively long (6000+
2985 seconds). To address this problem, we no longer use
2986 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2987 timeout no longer depends on the service time, it is set to
2988 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2989 time via pb_service_time.
2993 Description: Watchdog triggered on MDS failover
2994 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2995 the OSTs know that the MDS "UUID" can be reused for the same export
2996 from a different NID, so we do not need to wait for the export to be
3000 Frequency : rare, only if using MMP with Linux RAID
3002 Description: MMP doesn't work with Linux RAID
3003 Details : While using HA for Lustre servers with Linux RAID, it is possible
3004 that MMP will not detect multiple mounts. To make this work we
3005 need to unplug the device queue in RAID when the MMP block is being
3006 written. Also while reading the MMP block, we should read it from
3007 disk and not the cached one.
3009 Severity : enhancement
3011 Description: open file using fid
3012 Details : A file can be opened using just its fid, like
3013 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3016 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3019 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3020 Details : add proper locking for creation of jbd_4k slab cache
3024 Description: MMP check in ext3_remount() fails without displaying any error
3025 Details : When multiple mount protection fails during remount, proper error
3028 Severity : enhancement
3030 Description: Allow stripe size to be up to 4G-64k
3031 Details : Fix math logic to allow large stripe sizes.
3035 Description: add check for >8TB ldiskfs filesystems
3036 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3037 Don't allow >8TB ldiskfs filesystems to be mounted without
3038 force_over_8tb mount option
3043 Descriptoin: OSS complains "too many blocks" and crashes
3044 Details : In filter_commitrw_write():
3045 1. lquota_chkquota();
3046 2. obdo_from_inode();
3047 3. lquota_pending_commit();
3048 "2." has changed uid and gid in this case, then caused
3049 lquota_chkquota() and lquota_pending_commit() to use
3050 different uid and gid. That is the root of the bug.
3054 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3055 Details : Detect and print out the 'stripe count = -1' case properly.
3059 Description: Out or order replies might be lost on replay
3060 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3061 request with tid smaller than one currently being added, add it
3062 to the start, not end of the list.
3064 --------------------------------------------------------------------------------
3066 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3068 * Support for kernels:
3069 2.6.5-7.283 (SLES 9),
3070 2.6.9-55.EL (RHEL 4),
3071 2.6.16.46-0.14 (SLES 10),
3072 2.6.18.8 vanilla (kernel.org)
3073 * Client support for unpatched kernels:
3074 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3075 2.6.16 - 2.6.22 vanilla (kernel.org)
3076 * Due to recently discovered recovery problems, we do not recommend
3077 using patchless RHEL 4 clients with this or any earlier release.
3078 * Recommended e2fsprogs version: 1.39.cfs8
3079 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3080 * Starting with this release, the ldiskfs backing filesystem required
3081 by Lustre is now in its own package, lustre-ldiskfs. This package
3082 should be installed. It is versioned separately from Lustre and
3083 may be released separately in future.
3087 Description: block reactivating mgc import until all deactivates complete
3088 Details : Fix race when failing back MDT/MGS to itself (testing)
3091 Frequency : at statup only
3093 Description: mds_lov_synchronize race leads to various problems
3094 Details : simultaneous MDT->OST connections at startup can cause the
3095 sync to abort, leaving the OSC in a bad state.
3097 Severity : enhancement
3099 Description: add optional extra BUILD_VERSION info
3100 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3101 to override the lustre version.
3104 Frequency : 2.6.18 servers only
3106 Description: ll_kern_mount() doesn't release the module reference
3107 Details : The ldiskfs module reference count never drops down to 0
3108 because ll_kern_mount() doesn't release the module reference.
3113 Description: server LBUG when using old ost_num_threads parameter
3114 Details : Accept the old ost_num_threads parameter but warn that it
3115 is deprecated, and fix an off-by-one error that caused an LBUG.
3120 Description: Transient SCSI error results in persistent IO issue
3121 Details : iobuf->dr_error is not reinitialized to 0 between two
3125 Frequency : sometimes when underlying device returns I/O errors
3127 Description: OSTs not going read-only during write failures
3128 Details : OSTs are not remounted read-only when the journal commit threads
3129 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3130 instead of the ext3 wrappers.
3134 Description: poor connect scaling with increasing client count
3135 Details : Don't run filter_grant_sanity_check for more than 100 exports
3136 to improve scaling for large numbers of clients.
3139 Frequency : SLES10 only
3141 Description: sanity-quota.sh quotacheck failed: rc = -22
3142 Details : Quotas cannot be enabled on SLES10.
3145 Frequency : liblustre clients only
3147 Description: getdirentries does not give error when run on compute nodes
3148 Details : getdirentries does not fail when the size specified as an argument
3149 is too small to contain at least one entry
3151 Severity : enhancement
3153 Description: Add LNET router traceability for debug purposes
3154 Details : If a checksum failure occurs with a router as part of the
3155 IO path, the NID of the last router that forwarded the bulk data
3156 is printed so it can be identified.
3161 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3162 Details : Due to a race condition, liblustre clients were occasionally
3163 evicted incorrectly.
3165 Severity : enhancement
3167 Description: lfs setstripe use optional parameters instead of postional
3170 Severity : enhancement
3172 Description: Nanosecond timestamp support for ldiskfs
3173 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3174 resolution timestamps. There is not yet support for this at
3175 the Lustre filesystem level.
3178 Frequency : during server recovery
3180 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3181 Details : request with rq_no_resend flag not awake l_wait_event if they get
3185 Frequency : nfs export on patchless client
3187 Description: connectathon hang when test nfs export over patchless client
3188 Details : Disconnected dentry cannot be found with lookup, so we do not need
3189 to unhash it or make it invalid
3192 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3193 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3194 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3199 Description: open req refcounting wrong on reconnect
3200 Details : If reconnect happened between getting open reply from server and
3201 call to mdc_set_replay_data in ll_file_open, we will schedule
3202 replay for unreferenced request that we are about to free.
3203 Subsequent close will crash in variety of ways.
3204 Check that request is still eligible for replay in
3205 mdc_set_replay_data().
3210 Description: disable writes to filesystem when reading health_check file
3211 Details : the default for reading the health_check proc file has changed
3212 to NOT do a journal transaction and write to disk, because this
3213 can cause reads of the /proc file to hang and block HA state
3214 checking on a healthy but otherwise heavily loaded system. It
3215 is possible to return to the previous behaviour during configure
3216 with --enable-health-write.
3218 Severity : enhancement
3220 Description: 64-bit inode version
3221 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3222 to the inode. This will be required for version-based recovery.
3227 Description: MDS fails to start if a duplicate client export is detected
3228 Details : in some rare cases it was possible for a client to connect to
3229 an MDS multiple times. Upon recovery the MDS would detect this
3230 and fail during startup. Handle this more gracefully.
3232 Severity : enhancement
3234 Description: Add -o localflock option to simulate old noflock
3236 Details : This will achieve local-only flock/fcntl locks
3242 Description: log_commit_thread vs filter_destroy race leads to crash
3243 Details : Take import reference before releasing llog record semaphore
3248 Description: Wrong request locking in request set processing
3249 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3250 imp_delayed_list, in this place should be used imp_lock.
3253 Frequency : when reconnection
3255 Description: Grant Leak when osc reconnect to OST
3256 Details : When osc reconnect ost, OST(filter) should check whether it
3257 should grant more space to client by comparing fed_grant and
3258 cl_avail_grant, and return the granted space to client instead
3259 of "new granted" space, because client will call osc_init_grant
3260 to update the client grant space info.
3263 Frequency : when client reconnect to OST
3265 Description: Grant Leak when osc do resend and replay bulk write
3266 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3267 bulk write request, because the grant info while be sync between
3268 OSC and OST when reconnect, and we should ignore the grant info
3269 these of resend/replay write req.
3274 Description: Grant space more than available left space sometimes.
3275 Details : When then OST is about to be full, if two bulk writing from
3276 different clients came to OST. Accord the avaliable space of the
3277 OST, the first req should be permitted, and the second one
3278 should be denied by ENOSPC. But if the seconde arrived before
3279 the first one is committed. The OST might wrongly permit second
3280 writing, which will cause grant space > available space.
3283 Frequency : when client is evicted
3285 Description: Grant might be wrongly erased when osc is evicted by OST
3286 Details : when the import is evicted by server, it will fork another
3287 thread ptlrpc_invalidate_import_thread to invalidate the
3288 import, where the grant will be set to 0. While the original
3289 thread will update the grant it got when connecting. So if
3290 the former happened latter, the grant will be wrongly errased
3291 because of this race.
3296 Description: Checking Stale with correct fid
3297 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3298 to check whether it is stale, because sometimes, we want the
3299 enqueue happened anyway, and op_data.fid2 will not be initialized.
3301 Severity : enhancement
3303 Description: update patchless client
3304 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3307 Frequency : only with 2.4 kernel
3309 Description: random memory corruption
3310 Details : size of struct ll_inode_info is to big for union inode.u and this
3311 can be cause of random memory corruption.
3316 Description: Memory leak in recovery
3317 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3318 It should also check obd_fail before fsfilt_start, otherwise if
3319 fsfilt_start return -EROFS,(failover mds during mds recovery).
3320 then the req will return with repmsg->transno = 0 and rc = EROFS.
3321 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3322 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3323 be freed no matter whether fsfilt_commit success or not.
3326 Frequency : only with huge count clients
3328 Description: Prevents from taking the superblock lock in llap_from_page for
3330 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3331 for llap_from_page call in ll_removepage prevents from taking the
3332 superblock lock for a soon died page.
3337 Description: Not check open intent error before release open handle
3338 Details : in some rare cases, the open intent error is not checked before
3339 release open handle, which may cause
3340 ASSERTION(open_req->rq_transno != 0), because it tries to release
3341 the failed open handle.
3346 Description: Set cat log bitmap only after create log success.
3347 Details : in some rare cases, the cat log bitmap is set too early. and it
3348 should be set only after create log success.
3352 Description: Accessing a block bevice can re-enable I/O when Lustre is
3353 tearing down a device.
3354 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3359 Description: service threads may hog cpus when there are a lot of requests
3360 Details : Insert cond_resched to give other threads a chance to use some CPU
3365 Description: the cat log was not initialized in recovery
3366 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3367 unlink log on mds will not be initialized until mds post
3368 recovery. And also in mds post recovery, the unlink log will
3369 initialization will be done asynchronausly, so there will be race
3370 between add unlink log and unlink log initialization.
3374 Description: brw_stats were being printed incorrectly
3375 Details : brw_stats were being printed as log2 but all of them were not
3376 recorded as log2. Also remove some code duplication arising from
3377 filter_tally_{read,write}.
3381 Frequency : rare, only in recovery.
3382 Description: ASSERTION(req->rq_type != LI_POISON) failed
3383 Details : imp_lock should be held while iterating over imp_sending_list for
3384 prevent destroy request after get timeout in ptlrpc_queue_wait.
3388 Description: replay-single.sh test 52 fails
3389 Details : A lock's skiplist need to be cleanup when it being unlinked
3390 from its resource list.
3394 Description: Short directio read returns full requested size rather than
3396 Details : Direct I/O operations should return actual amount of bytes
3397 transferred rather than requested size.
3399 Severity : enhancement
3401 Description: metadata RPC reduction (e.g. for rm performance)
3402 Details : decrease the amount of synchronous RPC between clients and servers
3403 by canceling conflicing lock before the operation on the client
3404 and packing thier handles into the main operation RPC to server.
3406 Severity : enhancement
3408 Description: Async OSC create to avoid the blocking unnecessarily.
3409 Details : If an OST has no remaining object, system will block on the
3410 creation when it needs to create a new object on this OST. Now,
3411 ways use pre-created objects when available, instead of blocking on
3412 an empty osc while others are not empty. If we must block, we
3413 block for the shortest possible period of time.
3417 Description: improve handling recoverable errors
3418 Details : if request processig with error which can be recoverable on server
3419 request should be resend, otherwise page released from cache and
3422 Severity : enhancement
3424 Description: refine locking for avoid write wrong info into lov_objid file
3425 Details : fix possible races with add new target and write/update data in
3428 --------------------------------------------------------------------------------
3430 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3435 Frequency : on some architectures
3437 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3438 Details : Uninitialized flags sometimes cause configuration commands to
3442 Frequency : patchless clients only
3444 Description: missing __iget() symbol export
3445 Details : The __iget() symbol export is missing. To avoid the need for
3446 this on patchless clients the deathrow inode reaper is turned
3447 off, and we depend on the VM to clean up old inodes. This
3448 dependency was during via the fix for bug 12181.
3450 --------------------------------------------------------------------------------
3452 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3454 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3455 INTEROPERATE with older versions automatically. In many cases a
3456 special upgrade step is needed. Please read the
3457 user documentation before upgrading any part of a live system.
3458 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3459 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3460 * WARNING: Lustre configuration and startup changes are required with
3461 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3465 Severity : enhancement
3467 Description: MountConf
3468 Details : Lustre configuration is now managed via mkfs and mount
3469 commands instead of lmc and lconf. New obd types (MGS, MGC)
3470 are added for dynamic configuration management. See
3471 https://mail.clusterfs.com/wikis/lustre/MountConf for
3474 Severity : enhancement
3476 Description: dynamic OST addition
3477 Details : OSTs can now be added to a live filesystem
3479 Severity : enhancement
3481 Description: startup order invariance
3482 Details : MDTs and OSTs can be started in any order. Clients only
3483 require the MDT to complete startup.
3485 Severity : enhancement
3487 Description: parallel, asynchronous orphan cleanup
3488 Details : orphan cleanup is now performed in separate threads for each
3489 OST, allowing parallel non-blocking operation.
3491 Severity : enhancement
3493 Description: optimized stripe assignment
3494 Details : stripe assignments are now made based on ost space available,
3495 ost previous usage, and OSS previous usage, in order to try
3496 to optimize storage space and networking resources.
3498 Severity : enhancement
3500 Description: Permanently set tunables
3501 Details : All writable /proc/fs/lustre tunables can now be permanently
3502 set on a per-server basis, at mkfs time or on a live system.
3504 Severity : enhancement
3506 Description: Lustre message v2
3507 Details : Add lustre message format v2.
3509 Severity : enhancement
3511 Description: client OST exclusion list
3512 Details : Clients can be started with a list of OSTs that should be
3513 declared "inactive" for known non-responsive OSTs.
3517 Description: SPEC SFS validation failure on NFS v2 over lustre.
3518 Details : Changes the blocksize for regular files to be 2x RPC size,
3519 and not depend on stripe size.
3521 Severity : enhancement
3523 Description: Multiple MD RPCs in flight.
3524 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3525 To avoid overly-overloading MDS, introduce a limit on number of
3526 MDS RPCs in flight for a single client and add /proc controls
3527 to adjust this limit.
3529 Severity : enhancement
3531 Description: client read/write statistics
3532 Details : Add client read/write call usage stats for performance
3533 analysis of user processes.
3534 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3535 file access. extents_stats shows chunk size distribution.
3536 extents_stats_per_process show chunk size distribution per
3539 Severity : enhancement
3541 Description: mds statistics
3542 Details : Add detailed mds operations statistics in
3543 /proc/fs/lustre/mds/*/stats.
3547 Description: Failure of copying files with lustre special EAs.
3548 Details : Client side always return success for setxattr call for lustre
3549 special xattr (currently only "trusted.lov").
3554 Description: Refcount LNET uuids
3555 Details : The global LNET uuid list grew linearly with every startup;
3556 refcount repeated list entries instead of always adding to
3559 Severity : enhancement
3561 Description: Dynamic service threads
3562 Details : Within a small range, start extra service threads
3563 automatically when the request queue builds up.
3566 Frequency : mixed-endian client/server environments
3568 Description: mixed-endian crashes
3569 Details : The new msg_v2 system had some failures in mixed-endian
3572 Severity : enhancement
3574 Description: Easy OST removal
3575 Details : OSTs can be permanently deactivated with e.g. 'lctl
3576 conf_param lustre-OST0001.osc.active=0'
3578 Severity : enhancement
3580 Description: MGS proc entries
3581 Details : Added basic proc entries for the MGS showing what filesystems
3584 Severity : enhancement
3586 Description: provide MGS failover
3587 Details : Added config lock reacquisition after MGS server failover.
3589 Severity : enhancement
3591 Description: add Linux 2.4 support
3592 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3596 Description: a large application tries to do I/O to the same resource and dies
3597 in the middle of it.
3598 Details : Check the req->rq_arrival time after the call to
3599 ost_brw_lock_get(), but before we do anything about
3600 processing it & sending the BULK transfer request. This
3601 should help move old stale pending locks off the queue as
3602 quickly as obd_timeout.
3605 Frequency : when an incorrect nid is specified during startup
3607 Description: ptlrpc connect to non-existant node causes kernel crash
3608 Details : LNET can't be re-entered from an event callback, which
3609 happened when we expire a message after the export has been
3610 cleaned up. Instead, hand the zombie cleanup off to another
3613 Severity : enhancement
3615 Description: plain/inodebits lock performance improvement
3616 Details : Grouping plain/inodebits in granted list by their request modes
3617 and bits policy, thus improving the performance of search through
3621 Frequency : only if OST filesystem is corrupted
3623 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3624 Details : for a short time RPCs with bulk IO are in the replay list,
3625 but replay of bulk IOs is unimplemented. If the OST filesystem
3626 is corrupted due to disk cache incoherency and then replay is
3627 started it is possible to trip an assertion. Avoid putting
3628 committed RPCs into the replay list at all to avoid this issue.
3631 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3633 Description: System hang on startup
3634 Details : This bug allowed the liblustre (e.g. catamount) client to
3635 return to the app before handling all startup RPCs. This
3636 could leave the node unresponsive to lustre network traffic
3637 and manifested as a server ptllnd timeout.
3639 Severity : enhancement
3641 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3642 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3643 boolean to control whether to print peer debug info when a
3644 client's RPC times out.
3647 Frequency : only for kernels with patches from Lustre below 1.4.3
3649 Description: Remove old rdonly API
3650 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3653 Frequency : only for devices with external journals
3655 Description: Set external device read-only also
3656 Details : During a commanded failover stop, we set the disk device
3657 read-only while the server shuts down. We now also set any
3658 external journal device read-only at the same time.
3661 Frequency : when upgrading from 1.4 while trying to change parameters
3663 Description: The wrong (new) MDC name was used when setting parameters for
3664 upgraded MDT's. Also allows changing of OSC (and MDC)
3665 parameters if --writeconf is specified at tunefs upgrade time.
3668 Frequency : when setting specific ost indicies
3670 Description: QOS code breaks on skipped indicies
3671 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3672 created with --index need not be sequential.
3674 Severity : enhancement
3676 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3677 Details : The uninit_groups feature works in conjunction with the kernel
3678 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3679 up the pass1 processing of e2fsck. This is a read-only feature
3680 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3681 mount filesystems that have had this feature enabled.
3683 Severity : enhancement
3685 Description: Improve multi-block allocation algorithm to avoid fragmentation
3686 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3687 allocation locality and avoid filesystem fragmentation.
3689 ------------------------------------------------------------------------------
3691 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3693 * Support for kernels:
3694 2.4.21-47.0.1.EL (RHEL 3)
3695 2.6.5-7.283 (SLES 9)
3696 2.6.9-42.0.10.EL (RHEL 4)
3697 2.6.12.6 vanilla (kernel.org)
3698 2.6.16.27-0.9 (SLES 10)
3699 * Recommended e2fsprogs version: 1.39.cfs5
3701 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3705 Frequency : occasional, depends on client load and configuration
3706 Bugzilla : 12181, 12203
3707 Description: data loss for recently-modified files
3709 Details : In some cases it is possible that recently written or created
3710 files may not be written to disk in a timely manner (this should
3711 normally be within 30s unless client IO load is very high).
3712 The problem appears as zero-length files or files that are a
3713 multiple of 1MB in size after a client crash or client eviction
3714 that are missing data at the end of the file.
3716 This problem is more likely to be hit on clients where files are
3717 repeatedly created and unlinked in the same directory, clients
3718 have a large amount of RAM, have many CPUs, the filesystem has
3719 many OSTs, the clients are rebooted frequently, and/or the files
3720 are not accessed by other nodes after being written.
3722 The presence of the problem can be detected by looking at
3723 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3724 smaller than the second (nr_unused) then dirty files will not
3725 be flushed automatically to disk. "sync; sleep 10" should be
3726 run several times on the node before unmounting it to update
3727 Lustre (this is also safe to run on nodes without this problem).
3729 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3730 that can cause this same problem, so customers using that kernel
3731 also need to update the kernel in addition to Lustre. In order
3732 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3734 It is normal that files written just before a client crash (less
3735 than 30s) may not yet have been flushed to disk, even for local
3739 Frequency : frequent on thin XT3 nodes
3741 Description: UUID collision on thin XT3 Linux nodes
3742 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3743 randomly, since we relied on an insufficiently-seeded PRNG.
3748 Description: OSS hangs after "All ost request buffers busy"
3749 Details : A deadlock between quota and journal operations caused OSS
3750 hangs after printing "All ost request buffers busy."
3753 Frequency : always on liblustre builds
3755 Description: Cleanup compiler warnings on liblustre
3758 Frequency : always on liblustre builds on XT3
3760 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3764 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3765 > 32000 subdirectories
3766 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3767 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3768 subdirectory count crosses 32000. This will aid e2fsck to
3769 correctly handle more than 32000 subdirectories.
3772 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3774 Description: System hang on startup
3775 Details : This bug allowed the liblustre (e.g. catamount) client to
3776 return to the app before handling all startup RPCs. This
3777 could leave the node unresponsive to lustre network traffic
3778 and manifested as a server ptllnd timeout.
3780 Severity : enhancement
3782 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3783 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3784 boolean to control whether to print peer debug info when a
3785 client's RPC times out.
3790 Description: make O_SYNC working on 2.6 kernels
3791 Details : 2.6 kernels use different method for mark pages for write,
3792 so need add a code to lustre for O_SYNC work.
3797 Description: Failure to close file and release space on NFS
3798 Details : Put inode details into lock acquired in ll_intent_file_open.
3799 Use mdc_intent_lock in ll_intent_open to properly
3800 detect all kind of errors unhandled by mdc_enqueue.
3805 Description: proc file read during shutdown sometimes raced obd removal,
3807 Details : Add lock to prevent obd access after proc file removal.
3810 Frequency : Only for files larger than 4GB on 32-bit clients.
3812 Description: improperly doing page alignment of locks
3813 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3814 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3819 Description: RPCs being resent when they shouldn't be
3820 Details : Some RPCs that should not be resent are being resent. This
3821 can cause inconsistencies in the RPC state machine. Do not
3822 resend such requests.
3825 Frequency : rare, only with NFS export
3827 Description: Crash on NFS re-export node
3828 Details : under very unusual load conditions an assertion is hit in
3829 ll_intent_file_open()
3832 Frequency : only if OST filesystem is corrupted
3834 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3835 Details : for a short time RPCs with bulk IO are in the replay list,
3836 but replay of bulk IOs is unimplemented. If the OST filesystem
3837 is corrupted due to disk cache incoherency and then replay is
3838 started it is possible to trip an assertion. Avoid putting
3839 committed RPCs into the replay list at all to avoid this issue.
3844 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3845 Details : Large single O_DIRECT read and write calls can fail to allocate
3846 a sufficiently large buffer to process the request. In case of
3847 allocation failure the allocation is retried with a smaller
3848 buffer and broken into smaller requests.
3850 Severity : enhancement
3852 Description: Add -o localflock option to simulate old noflock behaviour.
3853 Details : This will achieve local-only flock/fcntl locks coherentness.
3858 Description: versioning check is incomplete
3859 Details : Checking the version difference of client vs. server, report
3860 error if the gap is too big.
3865 Description: add support PG_writeback bit
3866 Details : add support for PG_writeback bit for Lustre, for more carefull
3867 work with page cache in 2.6 kernel. This also fix some deadlocks
3868 and remove hack for work O_SYNC with 2.6 kernel.
3870 Severity : enhancement
3872 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3873 Details : The uninit_groups feature works in conjunction with the kernel
3874 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3875 up the pass1 processing of e2fsck. This is a read-only feature
3876 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3877 mount filesystems that have had this feature enabled.
3879 Severity : enhancement
3881 Description: Improve multi-block allocation algorithm to avoid fragmentation
3882 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3883 allocation locality and avoid filesystem fragmentation.
3885 ------------------------------------------------------------------------------
3887 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3889 * Support for kernels:
3890 2.6.9-42.0.3.EL (RHEL 4)
3891 2.6.5-7.276 (SLES 9)
3892 2.4.21-47.0.1.EL (RHEL 3)
3893 2.6.12.6 vanilla (kernel.org)
3894 2.6.16.21-0.8 (SLES10)
3895 * Recommended e2fsprogs version: 1.39.cfs2-0
3897 * The backwards-compatible /proc/sys/portals symlink has been removed
3898 in this release. Before upgrading, please ensure that you change
3899 any configuration scripts or /etc/sysctl.conf files that access
3900 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3901 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3902 in advance of the upgrade on any system running Lustre 1.4.6 or
3903 newer, since /proc/sys/lnet was added in that version.
3904 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3908 Frequency : only when quota is used
3910 Description: avoid scanning export list for quota master
3911 Details : Change the algorithms to avoid scanning export list in order
3912 to improve the efficiency.
3915 Frequency : MDS failover only, very rarely
3917 Description: "went back in time" messages on mds failover
3918 Details : The greatest transno may be lost when the current operation
3919 finishes with an error (transno==0) and the client's last_rcvd
3920 record is over-written. Save the greatest transno in the
3921 mds_last_transno for this case.
3924 Frequency : always for specific kernels and striping counts
3926 Description: client may get "Matching packet too big" without ACL support
3927 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3928 when trying to access files in certain configurations. The
3929 clients should in fact be denied when mounting because they do
3930 not understand ACLs.
3933 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3935 Description: many clients connecting with IO in progress causes connect timeouts
3936 Details : Avoid synchronous journal commits to avoid delays caused by many
3937 clients connecting/disconnecting when bulk IO is in progress.
3938 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3939 OST_IO_PORTAL to avoid delays behind potentially many pending
3943 Frequency : occasionally with multiple writers to a single file
3945 Description: shared writes to file may result in wrong size reported by stat()
3946 Details : Allow growing of kms when extent lock is cancelled
3949 Frequency : always with random mmap IO to multi-striped file
3951 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3952 Details : Only if the hole is at the end of OST object so that kms is too
3953 small. Fix is to increase kms accordingly in ll_nopage.
3956 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3958 Description: writes to a missing object would leak memory on the OST
3959 Details : If there is an inconsistency between the MDS and OST filesystems,
3960 such that the MDS references an object that doesn't exist, writes
3961 to that object will leak memory due to incorrect cleanup in the
3962 error handling path, eventually running out of memory on the OST.
3967 Description: Creating too long symlink causes lustre errors
3968 Details : Check symlink and name lengths before sending requests to MDS.
3971 Frequency : only if flock is enabled (not on by default)
3973 Description: posix locks not released on fd closure on 2.6.9+
3974 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3975 kernels, this caused such locks not to be released on fd close and
3976 then assertions on fs unmount about still used locks.
3979 Frequency : MDS failover only, very rarely
3981 Description: clients may get ASSERTION(granted_lock != NULL)
3982 Details : When request was taking a long time, and a client was resending
3983 a getattr by name lock request. The were multiple lock requests
3984 with the same client lock handle and
3985 mds_getattr_name->fixup_handle_for_resent_request found one of the
3986 lock handles but later failed with ASSERTION(granted_lock != NULL).
3991 Description: handle->h_buffer_credits > 0, assertion failure
3992 Details : h_buffer_credits is zero after truncate, causing assertion
3993 failure. This patch extends the transaction or creates a new
3997 Frequency : NFS re-export or patchless client
3998 Bugzilla : 11179, 10796
3999 Description: Crash on NFS re-export node (__d_move)
4000 Details : We do not want to hash the dentry if we don't have a lock.
4001 But if this dentry is later used in d_move, we'd hit uninitialised
4002 list head d_hash, so we just do this to init d_hash field but
4003 leave dentry unhashed.
4006 Frequency : NFS re-export or patchless client
4008 Description: NFS exports has problem with symbolic link
4009 Details : lustre client didn't properly install dentry when re-exported
4010 to NFS or running patchless client.
4013 Frequency : NFS re-export or patchless client
4015 Description: Various nfs/patchless fixes.
4016 Details : fixes reuse disconected alias for lookup process - this fixes
4017 warning "find_exported_dentry: npd != pd",
4018 fix permission error with open files at nfs.
4019 fix apply umask when do revalidate.
4022 Frequency : occasional
4024 Description: Crash on NFS re-export node
4025 Details : calling clear_page() on the wrong pointer triggered oops in
4026 generic_mapping_read().
4029 Frequency : rarely, using O_DIRECT IO
4031 Description: unaligned directio crashes client with LASSERT
4032 Details : check for unaligned buffers before trying any requests.
4035 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4037 Description: stale data returned from RAID cache
4038 Details : If only a small amount of IO is done to the RAID device before
4039 reading it again it is possible to get stale data from the RAID
4040 cache instead of reading it from disk.
4043 Frequency : always for sles10 kernel
4045 Description: sles10 support
4046 Details : ll_follow_link: compile fixes and using of nd_set_link
4047 under newer kernels.
4050 Frequency : depends on arch, kernel and compiler version, always on sles10
4053 Description: recursive or deep enough symlinks cause stack overflow
4054 Details : getting rid of large stack-allocated variable in
4058 Frequency : depends on hardware
4060 Description: lustre write performance loss in the SLES10 kernel
4061 Details : the performance loss is caused by using of write barriers in the
4062 ext3 code. The SLES10 kernel turns barrier support on by
4063 default. The fix is to undo that change for ldiskfs.
4065 ------------------------------------------------------------------------------
4067 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4069 * Support for kernels:
4070 2.6.9-42.0.3EL (RHEL 4)
4071 2.6.5-7.276 (SLES 9)
4072 2.4.21-47.0.1.EL (RHEL 3)
4073 2.6.12.6 vanilla (kernel.org)
4077 Frequency : quota enabled and large files being deleted
4079 Description: releasing more than 4GB of quota at once hangs OST
4080 Details : If a user deletes more than 4GB of files on a single OST it
4081 will cause the OST to spin in an infinite loop. Release
4082 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4087 Description: statfs data retrieved from /proc may be stale or zero
4088 Details : When reading per-device statfs data from /proc, in the
4089 {kbytes,files}_{total,free,avail} files, it may appear
4090 as zero or be out of date.
4093 Frequency : systems with MD RAID1 external journal devices
4095 Description: lconf's call to blkid is confused by RAID1 journal devices
4096 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4097 of returning all block devices that match the journal UUID.
4100 Frequency : always, for aggregate stripe size over 4GB
4102 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4103 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4104 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4107 Frequency : always if "lfs find" used on a local file/directory
4109 Description: "lfs find" segfaults if used on a local file/directory
4110 Details : The case where a directory component was not specified wasn't
4111 handled correctly. Handle this properly.
4114 Frequency : always on ppc64
4116 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4117 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4118 perform endianness conversion.
4121 Frequency : rarely (truncate to non-zero file size after write under load)
4122 Bugzilla : 10730, 10687
4123 Description: Files padded with zeros to next 4K multiple
4124 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4125 it is possible that files that are truncated to a non-zero size
4126 immediately after being written are filled with zero bytes beyond
4127 the truncated size. No file data is lost.
4129 Severity : enhancement
4131 Description: Allow recovery/failover for liblustre clients.
4132 Details : liblustre clients were unaware of failover configurations until
4135 Severity : enhancement
4137 Description: user file locks should fail when not mounting with flock option
4138 Details : Set up an error-returning stub in ll_file_operations.lock field
4139 to prevent incorrect behaviour when client is mounted without
4140 flock option. Also, set up properly f_op->flock field for
4144 Frequency : always on ia64
4146 Description: "lfs df" loops on printing out MDS statfs information
4147 Details : The obd_ioctl_data was not initialized and in some systems
4148 this caused a failure during the ioctl that did not return
4149 an error. Initialize the struct and return an error on failure.
4152 Frequency : SLES 9 only
4154 Description: Error of copying files with lustre special EAs as root
4155 Details : Client side always return success for setxattr call for lustre
4156 special xattr (currently only "trusted.lov").
4159 Frequency : rarely on clusters with both ia64+i386 clients
4161 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4162 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4163 clients are doing concurrent writes to the same file it is
4164 possible that the ia64 clients may LASSERT because the OST
4165 extent locks are not PAGE_SIZE aligned. Ensure that grown
4166 locks are always aligned on the request boundary.
4169 Frequency : specific use, occasional
4171 Description: Overwriting in use executable truncates on-disk binary image
4172 Details : If one node attempts to overwrite an executable in use by
4173 another node, we now correctly return ETXTBSY instead of
4174 truncating the file.
4176 Severity : enhancement
4178 Description: Async OSC create to avoid the blocking unnecessarily.
4179 Details : If an OST has no remaining object, system will block on the creating
4180 when need to create a new object on this OST. Now, ways use
4181 pre-created objects when available, instead of blocking on an
4182 empty osc while others are not empty. If we must block, we block
4183 for the shortest possible period of time.
4188 Description: chmod on Lustre root is propagated to other clients
4189 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4190 invalid by the follow_mount time.
4193 Frequency : liblustre clients only
4195 Description: Race in 'instant cancel' lock handling could lead to such locks
4196 never to be granted in case of SMP MDS
4197 Details : Do not destroy not yet granted but cbpending locks in
4201 Frequency : replay/resend of open
4203 Description: non null lock assetion failure in mds_intent_policy
4204 Details : Trying to replay/resend lockless open requests resulted in
4205 mds_open() returning 0 with no lock. Now it sets a flag if
4206 a lock is going to be returned.
4208 Severity : enhancement
4210 Description: Checksum enhancements
4211 Details : New checksum enhancements allow for resending RPCs that failed
4214 Severity : enhancement
4216 Description: Tunables on number of dirty pages in cacche
4217 Details : Allow to set limit on number of dirty pages cached.
4222 Description: client crash on unmount - lock still has references
4223 Details : In some error handling cases it was possible to leak a lock
4224 reference on a client while accessing a file. This was not
4225 harmful to the client during operation, but would cause the
4226 client to crash when the filesystem is unmounted.
4229 Frequency : specific case, rare
4231 Description: ETXTBSY on mds though file not in use by client
4232 Details : ETXTBSY is no longer incorrectly returned when attempting to
4233 chmod or chown a directory that the user previously tried to
4234 execute or a currently-executing binary.
4237 Frequency : extremely rare except on liblustre-based clients
4239 Description: Lustre space not freed when files are deleted
4240 Details : Clean up open-unlinked files after client eviction. Previously
4241 the unlink was skipped and the files remained as orphans.
4246 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4247 Details : In some cases it was possible to send a blocking callback to
4248 a client doing a glimpse, even though that client didn't get
4249 a lock granted. When the glimpse lock is cancelled on the OST
4250 the freed lock is left on the waiting list and corrupted the list.
4253 Frequency : all core dumps
4255 Description: Broke core dumps to lustre
4256 Details : Negative dentry may be unhashed if parent does not have UPDATE
4257 lock, but some callers, e.g. do_coredump, expect dentry to be
4258 hashed after successful create, hash it in ll_create_it.
4260 ------------------------------------------------------------------------------
4262 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4264 * Support for kernels:
4265 2.6.9-42.0.2.EL (RHEL 4)
4266 2.6.5-7.276 (SLES 9)
4267 2.4.21-40.EL (RHEL 3)
4268 2.6.12.6 vanilla (kernel.org)
4272 Frequency : always on RHEL 3
4274 Description: Number of open files grows over time
4275 Details : The number of open files grows over time, whether or not
4276 Lustre is started. This was due to a filp leak introduced
4277 by one of our kernel patches.
4279 ------------------------------------------------------------------------------
4281 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4283 * Support for kernels:
4284 2.6.9-42.EL (RHEL 4)
4285 2.6.5-7.276 (SLES 9)
4286 2.4.21-40.EL (RHEL 3)
4287 2.6.12.6 vanilla (kernel.org)
4292 Bugzilla : 5719, 9635, 9792, 9684
4293 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4294 Details : If a server is under heavy load and cannot reply to new
4295 connection requests before the client resends the (re)connect,
4296 the connection handling code can behave badly if two service
4297 threads are concurrently handing separate (re)connections from
4298 the same client. Add better locking to the connection handling
4299 code, and ensure that only a single connection will be processed
4300 for a given client UUID, even if the lock is dropped.
4302 Severity : enhancement
4304 Description: add TCP zero-copy support to kernel
4305 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4306 sends if the hardware supports scatter-gather and checksumming.
4307 This allows socklnd to do client-write and server-read more
4308 efficiently and reduce CPU utilization from skbuf copying.
4311 Frequency : only if NFS exporting from client
4313 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4314 Details : If setting attributes on a file created under NFS that had
4315 never been opened it would be possible to oops the client
4316 if the file had no objects.
4319 Frequency : always for liblustre
4321 Description: liblustre client does MDS+OSTs setattr RPC for each write
4322 Details : When doing a write from a liblustre client, the client
4323 incorrectly issued an RPC to the MDS and each OST the file was
4324 striped over in order to update the timestamps. When writing
4325 with small chunks and many clients this could overwhelm the MDS
4326 with RPCs. In all cases it would slow down the write because
4327 these RPCs are unnecessary.
4329 Severity : enhancement
4331 Description: allow number of MDS service threads to be changed at module load
4332 Details : It is now possible to change the number of MDS service threads
4333 running. Adding "options mds mds_num_threads={N}" to the MDS's
4334 /etc/modprobe.conf will set the number of threads for the next
4335 time Lustre is restarted (assuming the "mds" module is also
4336 reloaded at that time). The default number of threads will
4337 stay the same, 32 for most systems.
4342 Description: OST crash if filesystem is unformatted or corrupt
4343 Details : If an OST is started on a device that has never been formatted
4344 or if the filesystem is corrupt and cannot even mount then the
4345 error handling cleanup routines would dereference a NULL pointer.
4350 Description: NULL pointer deref in llap_from_page.
4351 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4352 incorrect) mapping if the page was truncated/reclaimed while it was
4353 searched for. Check for this condition and skip such pages when
4354 doing readahead. Introduce extra check to llap_from_page() to
4355 verify page->mapping->host is non-NULL (so page is not anonymous).
4358 Frequency : Sometimes when using sys_sendfile
4360 Description: "page not covered by a lock" warnings from ll_readpage
4361 Details : sendfile called ll_readpage without right page locks present.
4362 Now we introduced ll_file_sendfile that does necessary locking
4363 around call to generic_file_sendfile() much like we do in
4367 Frequency : with certain MDS communication failures at client mount time
4369 Description: NULL pointer deref after failed client mount
4370 Details : a client connection request may delayed by the network layer
4371 and not be sent until after the PTLRPC layer has timed out the
4372 request. If the client fails the mount immediately it will try
4373 to clean up before the network times out the request. Add a
4374 reference from the request import to the obd device and delay
4375 the cleanup until the network drops the request.
4378 Frequency : occasionally during client (re)connect
4380 Description: assertion failure during client (re)connect
4381 Details : processing a client connection request may be delayed by the
4382 client or server longer than the client connect timeout. This
4383 causes the client to resend the connection request. If the
4384 original connection request is replied in this interval, the
4385 client may trip an assertion failure in ptlrpc_connect_interpret()
4386 which thought it would be the only running connect process.
4389 Frequency : only with obd_echo servers and clients that are rebooted
4391 Description: kernel BUG accessing uninitialized data structure
4392 Details : When running an obd_echo server it did not start the ping_evictor
4393 thread, and when a client was evicted an uninitialized data
4394 structure was accessed. Start the ping_evictor in the RPC
4395 service startup instead of the OBD startup.
4397 Severity : enhancement
4398 Bugzilla : 10193 (patchless)
4399 Description: Remove dependency on various unexported kernel interfaces.
4400 Details : No longer need reparent_to_init, exit_mm, exit_files,
4401 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4404 Frequency : rare (only users of deprecated and unsupported LDAP config)
4406 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4407 Details : LDAP apparently contains 'lustreName' attributes instead of
4408 'name'. A simple remapping of the name is sufficient.
4411 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4413 Description: waiting_locks_callback trips kernel BUG if client is evicted
4414 Details : Running with the dump_on_timeout debug flag turned on makes
4415 it possible that the waiting_locks_callback() can try to dump
4416 the Lustre kernel debug logs from an interrupt handler. Defer
4417 this log dumping to the expired_lock_main() thread.
4419 Severity : enhancement
4421 Description: Support NFS exporting on 2.6 kernels.
4422 Details : Implement non-rawops metadata methods for NFS server to use without
4423 changing NFS server code.
4426 Frequency : very rare (synthetic metadata workload only)
4428 Description: two racing renames might cause an MDS thread to deadlock
4429 Details : Running the "racer" program may cause one MDS thread to rename
4430 a file from being the source of a rename to being the target of
4431 a rename at exactly the same time that another thread is doing
4432 so, and the second thread has already enqueued these locks after
4433 doing a lookup of the target and is trying to relock them in
4434 order. Ensure that we don't try to re-lock the same resource.
4437 Frequency : only very large systems with liblustre clients
4439 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4440 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4441 all OSTs in parallel. This allows the eviction of stale liblustre
4442 clients to proceed much faster than if they were done in series,
4443 and also offers similar improvements for other set_info RPCs.
4448 Description: excessive CPU usage during initial read phase on client
4449 Details : During the initial read phase on a client, it would agressively
4450 retry readahead on the file, consuming too much CPU and impacting
4451 performance (since 1.4.5.8). Improve the readahead algorithm
4452 to avoid this, and also improve some other common cases (read
4453 of small files in particular, where "small" is files smaller than
4454 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4459 Description: MDS crash when receiving packet with unknown intent.
4460 Details : Do not LBUG in unknown intent case, just return -EFAULT
4462 Severity : enhancement
4463 Bugzilla : 9293, 9385
4464 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4465 Details : Do not serialize getattr (non-intent version) and statfs.
4468 Frequency : occasional, when OST network is overloaded/intermittent
4470 Description: client evicted by OST after bulk IO timeout
4471 Details : If a client sends a bulk IO request (read or write) the OST
4472 may evict the client if it is unresposive to its data GET/PUT
4473 request. This is incorrect if the network is overloaded (takes
4474 too long to transfer the RPC data) or dropped the OST GET/PUT
4475 request. There is no need to evict the client at all, since
4476 the pinger and/or lock callbacks will handle this, and the
4477 client can restart the bulk request.
4480 Frequency : Always when mmapping file with no objects
4482 Description: client crashes when mmapping file with no objects
4483 Details : Check that we actually have objects in a file before doing any
4484 operations on objects in ll_vm_open, ll_vm_close and
4490 Description: Request leak when working with deleted CWD
4491 Details : Introduce advanced request refcount tracking for requests
4492 referenced from lustre intent.
4494 Severity : Enhancement
4496 Description: Cache open file handles on client.
4497 Details : MDS now will return special lock along with openhandle, if
4498 requested and client is allowed to hold openhandle, even if unused,
4499 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4500 closing files for every read/write openration.
4502 Severity : Enhancement
4504 Description: Cache open negative dentries on client when possible.
4505 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4506 negative dentries on lock revocation.
4511 Description: Remounting a client read-only wasn't possible with a zconf mount
4512 Details : It wasn't possible to remount a client read-only with llmount.
4514 Severity : enhancement
4515 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4516 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4517 MPICH is included to add Lustre-specific ADIO interfaces.
4518 This is based closely on the UFS ADIO layer and only differs
4519 in file creation, in order to allow the OST striping to be set.
4520 This is user-contributed code and not supported by CFS.
4525 Description: extended inode attributes (immutable, append-only) work improperly
4526 when 2.4 and 2.6 kernels are used on client/server or vice versa
4527 Details : Introduce kernel-independent values for these flags.
4529 Severity : enhancement
4532 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4533 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4534 granularity. Now, Fractional MB granularity is be supported,
4535 this is very useful for low memory system.
4537 Severity : enhancement
4539 Description: Getattr by fid
4540 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4541 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4542 have locks and avoids getattr rpc requests for it on every stat.
4545 Frequency : Always, for filesystems larger than 2TB
4547 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4548 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4549 filesystems larger than 2TB to be mounted with mballoc enabled.
4552 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4554 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4555 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4556 filesystems larger than 2TB is mounted then the kernel will
4557 silently corrupt the start of the filesystem. CONFIG_LBD is
4558 enabled for all CFS-supported kernels, but the possibility of
4559 this happening with a modified kernel config exists.
4561 Severity : enhancement
4563 Description: add client O_DIRECT support for 2.6 kernels
4564 Details : It is now possible to do O_DIRECT reads and writes to files
4565 in the Lustre client mountpoint on 2.6 kernel clients.
4567 Severity : enhancement
4569 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4570 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4571 parallel, not waiting for response from every OST before sending
4572 a rpc to the next OST.
4577 Description: setattr vs write race when updating file timestamps
4578 Details : Client processes that update a file timestamp into the past
4579 right after writing to the file (e.g. tar) it is possible that
4580 the updated file modification time can be reset to the current
4581 time due to a race between processing the setattr and write RPC.
4583 Severity : enhancement
4585 Description: Bring 'lfs find' closer in line with regular Linux find.
4586 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4587 -print0 options and obtains all the needed info through the lustre
4590 Severity : enhancement
4592 Description: support up to 1024 configured devices on one node
4593 Details : change obd_dev array from statically allocated to dynamically
4594 allocated structs as they are first used to reduce memory usage
4599 Description: Flush dirty partially truncated pages during truncate
4600 Details : Immediatelly flush partially truncated pages in filter_setattr,
4601 this way we completely avoid having any pages in page cache on OST
4602 and can retire ugly workarounds during writes to flush such pages.
4607 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4608 Details : protect inode from truncation within vfs_unlink() context
4609 just take a reference before calling vfs_unlink() and release it
4610 when parent's i_sem is free.
4615 Description: last_id value checked outside lock on OST caused LASSERT failure
4616 Details : If there were multiple MDS->OST object precreate requests in
4617 flight, it was possible that the OST's last object id was checked
4618 outside a lock and incorrectly tripped an assertion. Move checks
4619 inside locks, and discard old precreate requests.
4622 Frequency : always, if extents are used on OSTs
4624 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4625 Details : OSTs using the extents format would not zero the high 16 bits of
4626 the index physical block number. This is not a problem for any
4627 OST filesystems smaller than 16TB, and no kernels support ext3
4628 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4629 new/modified files) and can be fixed for existing filesystems
4630 with e2fsprogs-1.39-cfs1.
4635 Description: import connection selection may be incorrect if timer wraps
4636 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4637 import connections to be ignored if the 32-bit jiffies counter
4638 wraps. Use a 64-bit jiffies counter.
4641 Frequency : very large clusters immediately after boot
4643 Description: LNET request buffers exhausted under heavy short-term load
4644 Details : If a large number of client requests are generated on a service
4645 that has previously never seen so many requests it is possible
4646 that the request buffer growth cannot keep up with the spike in
4647 demand. Instead of dropping incoming requests, they are held in
4648 the LND until the RPC service can accept more requests.
4651 Frequency : Sometimes during replay
4653 Description: Assertion failure in ll_local_open after replay.
4654 Details : If replay happened on an open request reply before we were able
4655 to set replay handler, reply will become not swabbed tripping the
4656 assertion in ll_local_open. Now we set the handler right after
4657 recognising of open request
4660 Frequency : very rare
4662 Description: kernel reports "badness in vsnprintf"
4663 Details : Reading from the "recovery_status" /proc file in small chunks
4664 may cause a negative length in lprocfs_obd_rd_recovery_status()
4665 call to vsnprintf() (which is otherwise harmless). Exit early
4666 if there is no more space in the output buffer.
4668 Severity : enhancement
4670 Description: clear OBD RPC statistics by writing to them
4671 Details : It is now possible to clear the OBD RPC statistics by writing
4672 to the "stats" file.
4677 Description: Client mtime is not the same on different clients after utimes
4678 Details : In some cases, the client was using the utimes() syscall on
4679 a file cached on another node. The clients now validate the
4680 ctime from the MDS + OSTs to determine which one is right.
4685 Description: Inability to activate failout mode
4686 Details : lconf script incorrectly assumed that in pythong string's numeric
4687 value is used in comparisons.
4690 Frequency : always with multiple stripes per file
4692 Description: Inefficient object allocation for mutli-stripe files
4693 Details : When selecting which OSTs to stripe files over, for files with
4694 a stripe count that divides evenly into the number of OSTs,
4695 the MDS is always picking the same starting OST for each file.
4696 Return the OST selection heuristic to the original design.
4701 Description: mount failures may take full timeout to return an error
4702 Details : Under some heavy load conditions it is possible that a
4703 failed mount can wait for the full obd_timeout interval,
4704 possibly several minutes, before reporting an error.
4705 Instead return an error as soon as the status is known.
4707 Frequency : quota enabled and large files being deleted
4709 Description: releasing more than 4GB of quota at once hangs OST
4710 Details : If a user deletes more than 4GB of files on a single OST it
4711 will cause the OST to spin in an infinite loop. Release
4712 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4717 Description: statfs data retrieved from /proc may be stale or zero
4718 Details : When reading per-device statfs data from /proc, in the
4719 {kbytes,files}_{total,free,avail} files, it may appear
4720 as zero or be out of date.
4723 Frequency : systems with MD RAID1 external journal devices
4725 Description: lconf's call to blkid is confused by RAID1 journal devices
4726 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4727 of returning all block devices that match the journal UUID.
4730 Frequency : always, for aggregate stripe size over 4GB
4732 Description: assertion fails when trying to use 4GB stripe size
4733 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4734 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4737 Frequency : always on ppc64
4739 Description: the first write on an ext3 filesystem with mballoc got stuck
4740 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4741 perform endianness conversion.
4743 ------------------------------------------------------------------------------
4745 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4747 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4748 INTEROPERATE with older versions automatically. Please read the
4749 user documentation before upgrading any part of a live system.
4750 * WARNING: Lustre networking configuration changes are required with
4751 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4754 * Support for kernels:
4755 2.6.9-22.0.2.EL (RHEL 4)
4756 2.6.5-7.244 (SLES 9)
4757 2.6.12.6 vanilla (kernel.org)
4760 Severity : enhancement
4761 Bugzilla : 7981/8208
4762 Description: Introduced Lustre Networking (LNET)
4763 Details : LNET is new networking infrastructure for Lustre, it includes
4764 a reorganized network configuration mode (see the user
4765 documentation for full details) as well as support for routing
4766 between different network fabrics. Lustre Networking Devices
4767 (LNDs) for the supported network fabrics have also been
4768 created for this new infrastructure.
4770 Severity : enhancement
4771 Description: Introduced Access control lists
4772 Details : clients can set ACLs on files and directories in order to have
4773 more fine-grained permissions than the standard Unix UGO+RWX.
4774 The MDS must be started with the "-o acl" mount option.
4776 Severity : enhancement
4777 Description: Introduced filesystem quotas
4778 Details : Administrators may now establish per-user quotas on the
4781 Severity : enhancement
4783 Description: Configuration change for the XT3
4784 The PTLLND is now used to run Lustre over Portals on the XT3
4785 The configure option(s) --with-cray-portals are no longer used.
4786 Rather --with-portals=<path-to-portals-includes> is used to
4787 enable building on the XT3. In addition to enable XT3 specific
4788 features the option --enable-cray-xt3 must be used.
4793 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4794 Details : A race between allocating a new llcd and re-getting the llcd_lock
4795 allowed another thread to grab newly-allocated llcd.
4797 Severity : enhancement
4799 Description: 2.6 OST async journal commit and locking fix to improve performance
4800 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4801 2.6 kernels are now async as they already were in 2.4 kernels so
4802 that they can commit concurrently with the network bulk transfer.
4803 For block-allocated files the filter allocation semaphore is held
4804 to avoid filesystem fragmentation during allocation. BKL lock
4805 removed for 2.6 xattr operations where it is no longer needed.
4810 Description: lconf incorrectly determined whether two IP networks could talk
4811 Details : In some more complicated routing and multiple-network
4812 configurations, lconf will avoid trying to make a network
4813 connection to a disjoint part of the IP space. It was doing the
4814 math incorrectly for one set of cases.
4819 Description: Fix for potential infinite loop processing records in an llog.
4820 Details : If an llog record is corrupted/zeroed, it is possible to loop
4821 forever in llog_process(). Validate the llog record length
4822 and skip the remainder of the block on error.
4825 Frequency : occasional (liblustre only)
4827 Description: liblustre could not open files whose last component is a symlink
4828 Details : sysio_path_walk() would incorrectly pass the open intent to
4829 intermediate path components.
4832 Frequency : rare (liblustre only with non-standard tuning)
4833 Bugzilla : 7201 (7350)
4834 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4835 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4836 after it has been released, only do this for referenced locks
4838 Severity : enhancement
4840 Description: specifying an (invalid) directory default stripe_size of -1
4841 would reset the directory default striping
4842 Details : stripe_size -1 was used internally to signal directory stripe
4843 removal, now use "all default" to signal dir stripe removal
4844 as a directory striping of "all default" is not useful
4847 Frequency : common for large clusters running liblustre clients
4849 Description: doing an ls when liblustre clients are running is slow
4850 Details : sending a glimpse AST to a liblustre client waits for every AST
4851 to time out, as liblustre clients will not respond. Since they
4852 cannot cache data we refresh the OST lock LVB from disk instead.
4854 Severity : enhancement
4856 Description: doing an ls at the same time as file IO can be slow
4857 Details : enqueue and other "small" requests can be blocked behind many
4858 large IO requests. Create a new OST IO portal for non-IO
4859 requests so they can be processed faster.
4862 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4864 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4865 created file with mode 000
4866 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4867 allow an ftruncate() as a non-root user to a file with mode 000.
4868 Lustre now respects this flag to disable mode checks when
4869 truncating a file owned by the user
4872 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4875 Description: Revoking locks from clients that went dead or catatonic might take
4877 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4878 cancellation of such locks instant on servers without waiting for
4879 any reply from clients. Clients drops these locks when cancel
4880 notification from server is received without replying.
4883 Frequency : liblustre-only, when liblustre client dies or becomes busy
4885 Description: Doing ls on Linux clients can take a long time with active
4887 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4888 granting such locks to it in the first place if possible. Locks
4889 are taken by proxy on the OST during the read or write and
4890 dropped immediately afterward. Add connect flags handling, do
4891 not grant locks to liblustre clients for glimpse ASTs.
4893 Severity : enhancement
4895 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4896 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4897 sequential reads and behaves badly with applications doing random
4898 reads. Improve it by reading ahead at least read region, and
4899 avoiding excessive large RPC for small reads.
4901 Severity : enhancement
4903 Description: Creating more than 1000 files for a single job may cause a load
4904 imbalance on the OSTs if there are also a large number of OSTs.
4905 Details : qos_prep_create() uses an OST index reseed value that is an
4906 even multiple of the number of available OSTs so that if the
4907 reseed happens in the middle of the object allocation it will
4908 still utilize the OSTs as uniformly as possible.
4913 Description: OST or MDS may oops in ping_evictor_main()
4914 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4915 but doesn't restart at beginning of obd_exports_timed list
4918 Severity : enhancement
4920 Description: improve by-nid export eviction on the MDS and OST
4921 Details : allow multiple exports with the same NID to be evicted at one
4922 time without re-searching the exports list.
4925 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4927 Description: MDS may oops in groups_free()
4928 Details : in rare race conditions a newly allocated group_info struct is
4929 freed again, and this can be NULL. The 2.4 compatibility code
4930 for groups_free() checked for a NULL pointer, but 2.6 did not.
4933 Frequency : common for liblustre clients doing little filesystem IO
4934 Bugzilla : 9352, 7313
4935 Description: server may evict liblustre clients accessing contended locks
4936 Details : if a client is granted a lock or receives a completion AST
4937 with a blocking AST already set it would not reply to the AST
4938 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4941 Frequency : lfs setstripe, only systems with more than 160 OSTs
4943 Description: unable to set striping with a starting offset beyond OST 160
4944 Details : llapi_create_file() incorrectly limited the starting stripe
4945 index to the maximum single-file stripe count.
4948 Frequency : LDAP users only
4950 Description: lconf did not handle in-kernel recovery with LDAP properly
4951 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4953 Severity : enhancement
4955 Description: bind OST threads to NUMA nodes to improve performance
4956 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4957 node and do their allocations there to localize memory access
4959 Severity : enhancement
4961 Description: llmount can determine client NID directly from Myrinet (GM)
4962 Details : the client NID code from gmnalnid was moved directly into
4963 llmount, removing the need to use this or specifying the
4964 client NID explicitly when mounting GM clients with zeroconf
4967 Frequency : if client is started with down MDS
4969 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4970 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4971 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4972 ptlrpc_import_delayed_req() because we didn't check if the
4973 request was interrupted, and we also didn't break out of the
4974 event loop if there was no timeout
4979 Description: data loss during non-page-aligned writes to a single file from
4980 both multiple nodes and multiple threads on one node at same time
4981 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4982 inconsistency led to false short-reads, that were cached and later
4983 used by ->prepare_write() to fill in partially written page,
4984 leading to data loss.
4987 Frequency : always, if lconf --abort_recovery used
4989 Description: lconf --abort_recovery fails with 'Operation not supported'
4990 Details : lconf was attempting to abort recovery on the MDT device and not
4993 Severity : enhancement
4995 Description: remove cleanup logs
4996 Details : replace lconf-generated cleanup logs with lustre internal
4997 cleanup routines. Eliminates the need for client-cleanup and
5000 Severity : enhancement
5002 Description: add support for EAs (user and system) on lustre filesystems
5003 Details : it is now possible to store extended attributes in the Lustre
5004 client filesystem, and with the user_xattr mount option it
5005 is possible to allow users to store EAs on their files also
5007 Severity : enhancement
5009 Description: Add possibility (config option) to show minimal available OST free
5011 Details : When compiled with --enable-mindf configure option, statfs(2)
5012 (and so, df) will return least minimal free space available from
5013 all OSTs as amount of free space on FS, instead of summary of
5014 free spaces of all OSTs.
5016 Severity : enhancement
5018 Description: do not expand extent locks acquired on OST-side
5019 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5020 by server: they are not cached anyway.
5023 Frequency : when mmap is used/binaries executed from Lustre
5025 Description: Unmmap pages before throwing them away from read cache.
5026 Details : llap_shrink cache now attempts to unmap pages before discarding
5027 them (if unmapping failed - do not discard). SLES9 kernel has
5028 extra checks that trigger if this unmapping is not done first.
5033 Description: lconf didn't resolve symlinks before checking to see whether a
5034 given mountpoint was already in use
5037 Frequency : when migrating failover services
5038 Bugzilla : 6395, 9514
5039 Description: When migrating a subset of services from a node (e.g. failback
5040 from a failover service node) the remaining services would
5041 time out and evict clients.
5042 Details : lconf --force (implied by --failover) sets the global obd_timeout
5043 to 5 seconds in order to quickly disconnect, but this caused
5044 other RPCs to time out too quickly. Do not change the global
5045 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5047 Severity : enhancement
5048 Frequency : if MDS is started with down OST
5049 Bugzilla : 9439,5706
5050 Description: Allow startup/shutdown of an MDS without depending on the
5051 availability of the OSTs.
5052 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5053 Add appropriate locking and lov-osc refcounts for safe
5054 cleaning. Add osc abort_inflight calls in case the
5055 synchronize never started.
5058 Frequency : occasional (Cray XT3 only)
5060 Description: root not authorized to access files in CRAY_PORTALS environment
5061 Details : The client process capabilities were not honoured on the MDS in
5062 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5063 been accessed by an authorized user then root was able to access
5064 the file on the local client also. The root user capabilities
5065 are now allowed on the MDS, as this environment has secure UID.
5068 Frequency : occasional
5070 Description: ldiskfs "too long searching" message happens too often
5071 Details : A debugging message (otherwise harmless) prints too often on
5072 the OST console. This has been reduced to only happen when
5073 there are fragmentation problems on the filesystem.
5078 Description: Division by zero in statfs when all OSCs are inactive
5079 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5080 lov_statfs divides by value returned by lov_get_stripecnt().
5084 Bugzilla : 9489, 3273
5085 Description: First write from each client to each OST was only 4kB in size,
5086 to initialize client writeback cache, which caused sub-optimal
5087 RPCs and poor layout on disk for the first written file.
5088 Details : Clients now request an initial cache grant at (re)connect time
5089 and so that they can start streaming writes to the cache right
5090 away and always do full-sized RPCs if there is enough data.
5091 If the OST is rebooted the client also re-establishes its grant
5092 so that client cached writes will be honoured under the grant.
5097 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5098 Details : Now I/O RPCs go to different portal number and (presumably) fast
5099 lock requests (and glimses) and other RPCs get their own service
5100 threads pool that should be able to service those RPCs
5103 Severity : enhancement
5105 Description: Ability to exchange lustre version between client and servers and
5106 issue warnings at client side if client is too old. Also for
5107 liblustre clients there is ability to refuse connection of too old
5109 Details : New 'version' field is added to connect data structure that is
5110 filled with version info. That info is later checked by server and
5114 Frequency : rare, liblustre only.
5115 Bugzilla : 9296, 9581
5116 Description: Two simultaneous writes from liblustre at offset within same page
5117 might proceed at the same time overwriting eachother with stale
5119 Details : I/O lock withing llu_file_prwv was released too early, before data
5120 actually was hitting the wire. Extended lock-holding time until
5121 server acknowledges receiving data.
5124 Frequency : extremely rare. Never observed in practice.
5126 Description: avoid generating lustre_handle cookie of 0.
5127 Details : class_handle_hash() generates handle cookies by incrementing
5128 global counter, and can hit 0 occasionaly (this is unlikely, but
5129 not impossible, because initial value of cookie counter is
5130 selected randonly). Value of 0 is used as a sentinel meaning
5131 "unassigned handle" --- avoid it. Also coalesce two critical
5132 sections in this function into one.
5134 Severity : enhancement
5136 Description: allow liblustre clients to delegate truncate locking to OST
5137 Details : To avoid overhead of locking, liblustre client instructs OST to
5138 take extent lock in ost_punch() on client's behalf. New connection
5139 flag is added to handle backward compatibility.
5141 Severity : enhancement
5142 Bugzilla : 4928, 7341, 9758
5143 Description: allow number of OST service threads to be specified
5144 Details : a module parameter allows the number of OST service threads
5145 to be specified via "options ost ost_num_threads={N}" in the
5146 OSS's /etc/modules.conf or /etc/modprobe.conf.
5150 Bugzilla : 6146, 9635, 9895
5151 Description: servers crash with bad pointer in target_handle_connect()
5152 Details : In rare cases when a client is reconnecting it was possible that
5153 the connection request was the last reference for that export.
5154 We would temporarily drop the export reference and get a new
5155 one, but this may have been the last reference and the export
5156 was just destroyed. Get new reference before dropping old one.
5158 Severity : enhancement
5159 Frequency : if client is started with failover MDS
5161 Description: Allow multiple MDS hostnames in the mount command
5162 Details : Try to read the configuration from all specified MDS
5163 hostnames during a client mount in case the "primary"
5166 Severity : enhancement
5168 Description: Stop sending data to evicted clients as soon as possible.
5169 Details : Check if the client we are about to send or are sending data to
5170 was evicted already. (Check is done every second of waiting,
5171 for which l_wait_event interface was extended to allow checking
5172 of exit condition at specified intervals).
5175 Frequency : rare, normally only when NFS exporting is done from client
5177 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5179 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5182 Frequency : always, if srand() is called before liblustre initialization
5184 Description: Liblustre uses system PRNG disturbing its usage by user application
5185 Details : Introduce internal to lustre fast and high-quality PRNG for
5186 lustre usage and make liblustre and some other places in generic
5187 lustre code to use it.
5189 Severity : enhancement
5190 Bugzilla : 9477, 9557, 9870
5191 Description: Verify that the MDS configuration logs are updated when xml is
5192 Details : Check if the .xml configuration logs are newer than the config
5193 logs stored on the MDS and report an error if this is the case.
5194 Request --write-conf, or allow starting with --old_conf.
5196 Severity : enhancement
5198 Description: Handle symlinks in the path when checking if Lustre is mounted.
5199 Details : Resolve intermediate symlinks when checking if a client has
5200 mounted a filesystem to avoid duplicate client mounts.
5205 Description: lconf can hit an error exception but still return success.
5206 Details : The lconf command catches the Command error exception at the top
5207 level script context and will exit with the associated exit
5208 status, but doesn't ensure that this exit status is non-zero.
5213 Description: failure of ptlrpc thread startup can cause oops
5214 Details : Starting a ptlrpc service thread can fail if there are a large
5215 number of threads or the server memory is very fragmented.
5216 Handle this without oopsing.
5219 Frequency : always, only if liblustre and non-default acceptor port was used
5221 Description: liblustre cannot connect to servers with non-default acceptor port
5222 Details : tcpnal_set_default_params() was not called and was therefore
5223 ignoring the environment varaible TCPNAL_PORT, as well as other
5224 TCPNAL_ environment variables
5229 Description: two objects could be created on the same OST for a single file
5230 Details : If an OST is down, in some cases it was possible to create two
5231 objects on a single OST for a single file. No problems other
5232 than potential performance impact and spurious error messages.
5236 Bugzilla : 5681, 9562
5237 Description: Client may oops in ll_unhash_aliases
5238 Details : Client dcache may become inconsistent in race condition.
5239 In some cases "getcwd" can fail if the current directory is
5245 Description: Inode refcounting problems in NFS export code
5246 Details : link_raw functions used to call d_instantiate without obtaining
5247 extra inode reference first.
5251 Bugzilla : 9942, 9903
5252 Description: Referencing freed requests leading to crash, memleaks with NFS.
5253 Details : We used to require that call to ll_revalidate_it was always
5254 followed by ll_lookup_it. Also with revalidate_special() it is
5255 possible to call ll_revalidate_it() twice for the same dentry
5256 even if first occurence returned success. This fix changes semantic
5257 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5258 reference on a request referred from the intent.
5259 ll_intent_release() then releases such a request.
5262 Frequency : rare, normally benchmark loads only
5264 Description: unlinked inodes were kept in memory on the client
5265 Details : If a client is repeatedly creating and unlinking files it
5266 can accumulate a lot of stale inodes in the inode slab cache.
5267 If there is no other client load running this can cause the
5268 client node to run out of memory. Instead flush old inodes
5269 from client cache that have the same inode number as a new inode.
5272 Frequency : SLES9 2.6.5 kernel and long filenames only
5273 Bugzilla : 9969, 10379
5274 Description: utime reports stale NFS file handle
5275 Details : SLES9 uses out-of-dentry names in some cases, which confused
5276 the lustre dentry revalidation. Change it to always use the
5280 Frequency : rare, unless heavy write-truncate concurrency is continuous
5281 Bugzilla : 4180, 6984, 7171, 9963, 9331
5282 Description: OST becomes very slow and/or deadlocked during object unlink
5283 Details : filter_destroy() was holding onto the parent directory lock
5284 while truncating+unlinking objects. For very large objects this
5285 may block other threads for a long time and slow overall OST
5286 responsiveness. It may also be possible to get a lock ordering
5287 deadlock in this case, or run out of journal credits because of
5288 the combined truncate+unlink. Solution is to do object truncate
5289 first in one transaction without parent lock, and then do the
5290 final unlink in a new transaction with the parent lock. This
5291 reduces the lock hold time dramatically.
5294 Frequency : rare, 2.4 kernels only
5296 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5297 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5298 before unmounting their filesystem in order to avoid deadlock.
5299 The kernel_locked() function in 2.4 kernels only checks whether
5300 the kernel lock is held, not whether it is this process that is
5301 holding it as 2.6 kernels do.
5306 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5307 Details : The client ptlrpc code may be trying to reconnect to a down
5308 server before a previous connection attempt has timed out.
5309 Increase the reconnect interval to be longer than the connection
5310 timeout interval to avoid sending duplicate connections to
5314 Frequency : echo_client brw_test command
5316 Description: fix echo_client to work with OST preallocated code
5317 Details : OST preallocation code (5137) didn't take echo_client IO path
5318 into account: echo_client calls filter methods outside of any
5319 OST thread and, hence, there is no per-thread preallocated
5320 pages and buffers to use. Solution: hijack pga pages for IO. As
5321 a byproduct, this avoids unnecessary data copying.
5325 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5326 Description: Client can oops in mdc_commit_close() after open replay
5327 Details : It was possible for the MDS to return an open request with no
5328 transaction number in mds_finish_transno() if the client was
5329 evicted, but without actually returning an error. Clients
5330 would later try to replay that open and may trip an assertion
5331 Simplify the client close codepath, and always return an error
5332 from the MDS in case the open is not successful.
5335 Frequency : rare, 2.6 OSTs only
5337 Description: OST may deadlock under high load on fragmented files
5338 Details : If there was a heavy load and highly-fragmented OST filesystems
5339 it was possible to have all the OST threads deadlock waiting on
5340 allocation of biovecs, because the biovecs were not released
5341 until the entire RPC IO was completed. Instead, release biovecs
5342 as soon as they are complete to ensure forward IO progress.
5344 Severity : enhancement
5346 Description: Support for specifying external journal device at mount
5347 Details : If an OST or MDS device is formatted with an external journal
5348 device, this device major/minor is stored in the ext3 superblock
5349 and may not be valid for failover. Allow detecting and
5350 specifying the external journal at mount time.
5355 Description: Mounting an MDS with pending unlinked files may cause oops
5356 Details : target_finish_recovery() calls mds_postrecov() which returned
5357 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5358 considers this an error and immediately begins cleaning up the
5359 lov, just after starting the mds_lov process
5361 Severity : enhancement
5363 Description: Implement 'lfs df' to report actual free space on per-OST basis
5364 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5365 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5366 usage of objects; '-h' to report in human readable format.
5368 ------------------------------------------------------------------------------
5370 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5377 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5378 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5379 at startup during orphan unlinking. Instead, skip buddy update
5380 before setup, as it will be regenerated after recovery is complete.
5385 Description: If an OST is inactive, its locks might reference stale inodes.
5386 Details : lov_change_cbdata() must iterate over all namespaces, even if
5387 they are inactive to clear inode references from the lock.
5389 Severity : enhancement
5390 Frequency : occasional, if non-standard max_dirty_mb used
5392 Description: Client will block write RPCs if not enough grant
5393 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5394 then the client will block writes while waiting for another RPC
5395 to complete instead of consuming its dirty limit. With change
5396 we get improved performance when max_dirty_mb is small.
5398 Severity : enhancement
5399 Bugzilla : 3389, 6253
5400 Description: Add support for supplementary groups on the MDS.
5401 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5402 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5403 lookups for user supplementary groups into a cache.
5407 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5408 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5409 for liblustre O_EXCL is enforced by MDS.
5413 Description: readdir never returns NULL in liblustre.
5414 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5415 dentry in struct dirent.
5419 Frequency : liblustre only, depends on application IO pattern
5420 Description: liblustre clients evicted if not contacting servers
5421 Details : Don't put liblustre clients into the ping_evictor list, so
5422 they will not be evicted by the pinger ever.
5424 Severity : enhancement
5426 Description: Add ability to evict clients by NID from MDS.
5427 Details : By echoing "nid:$NID" string into
5428 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5429 $NID would be instantly evicted from this MDS and from all active
5430 OSTs connected to it.
5434 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5435 Details : lookup_it_finish() used to query file size from OSTs that was not
5440 Description: service threads change working directory to that of init
5441 Details : Starting lustre service threads may pin the working directory
5442 of the parent thread, making that filesystem busy. Threads
5443 now change to the working directory of init to avoid this.
5447 Frequency : during shutdown only
5448 Description: shutdown with a failed MDS or OST can cause unmount to hang
5449 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5454 Frequency : 2.6 only
5455 Description: chmod/chown may include an extra supplementary group
5456 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5457 supplementary group and if none is specified this is used.
5460 Bugzilla : 5479 (6816)
5462 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5463 Details : If lookup is in progress on a file that is unlinked we might try
5464 to revalidate the inode and fail in revalidate after lookup is
5465 complete and ll_file_open() enqueues the open again but
5466 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5469 Frequency : always, if lconf --abort_recovery used
5471 Description: lconf --abort_recovery fails with 'Operation not supported'
5472 Details : lconf was attempting to abort recovery on the MDT device and not
5475 ------------------------------------------------------------------------------
5477 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5482 Frequency : rare (only unsupported configurations with a node running as an
5484 Bugzilla : 6514, 5137
5485 Description: Mounting a Lustre file system on a node running as an OST could
5487 Details : OSTs now preallocates memory needed to write out data at
5488 startup, instead of when needed, to avoid having to
5489 allocate memory in possibly low memory situations.
5490 Specifically, if the file system is mounted on on OST,
5491 memory pressure could force it to try to write out data,
5492 which it needed to allocate memory to do. Due to the low
5493 memory, it would be unable to do so and the node would
5494 become unresponsive.
5496 Severity : enhancement
5498 Description: Addition of lconf --service command line option
5499 Details : lconf now accepts a '--service <arg>' option, which is
5500 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5502 Severity : enhancement
5504 Description: Failover mode is now the default for OSTs.
5505 Details : By default, OSTs will now run in failover mode. To return to
5506 the old behaviour, add '--failout' to the lmc line for OSTs.
5508 Severity : enhancement
5510 Description: Health checks are now provided for MDS and OSTs
5511 Details : Additional detailed health check information on MSD and OSTs
5512 is now provided through the procfs health_check value.
5515 Frequency : occasional, depends on IO load
5517 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5518 after numerous create/delete cycles
5519 Details : The ext3 inode allocation policy would not allocate new inodes
5520 very well on the OSTs because there are no new directories
5521 being created. Instead we look for groups with free space if
5522 the parent directories are nearly full.
5527 Description: Network or server problems during mount may cause partially
5528 mounted clients instead of returning an error.
5529 Details : The config llog parsing code may overwrite the error return
5530 code during mount error handling, returning success instead
5536 Description: MDS can fail to allocate large reply buffers
5537 Details : After long uptimes the MDS can fail to allocate large reply
5538 buffers (e.g. zconf client mount config records) due to memory
5539 fragmentation or consumption by the buffer cache. Preallocate
5540 some large reply buffers so that these replies can be sent even
5541 under memory pressure.
5545 Frequency : rare (liblustre)
5546 Description: fsx running with liblustre complained that using truncate() to
5547 extend the file doesn't work. This patch corrects that issue.
5548 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5549 ATTR_SIZE and lsm use in llu_setattr_raw.
5553 Frequency : rare, only 2.6 kernels
5554 Description: Unusual file access patterns on the MDS may result in inode
5555 data being lost in very rare circumstances.
5556 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5557 code in the 2.6 kernel caused the inode and/or EA data not to
5558 be read from disk, causing single-file corruption.
5562 Frequency : rare, only 2.6 filesystems using extents
5563 Description: Heavy concurrent write and delete load may cause data corruption.
5564 Details : It was possible under high-load situations to have an extent
5565 metadata block in the block device cache from a just-unlinked
5566 file overwrite a newly-allocated data block. We now unmap any
5567 metadata buffers that alias just-allocated data blocks.
5571 Frequency : filesystems with default stripe_count larger than 77
5572 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5573 Details : lconf specifies an inode size of 4096 bytes when the default
5574 stripe_count is larger than 77. This conflicts with the default
5575 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5576 this case to avoid pinning too much memory for large EAs.
5578 ------------------------------------------------------------------------------
5580 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5585 Frequency : rare (extremely heavy IO load with hundreds of clients)
5587 Description: Client is evicted, gets IO error writing to file
5588 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5589 caused clients to be evicted for AST timeouts. The fixes in
5590 bug 5192 mean we no longer need to have such short AST timeouts
5591 so ldlm_timeout has been increased.
5594 Frequency : occasional during --force or --failover shutdown under load
5595 Bugzilla : 5949, 4834
5596 Description: Server oops/LBUG if stopped with --force or --failover under load
5597 Details : a collection of import/export refcount and cleanup ordering
5598 issues fixed for safer force cleanup
5601 Frequency : only filesystems larger than 120 OSTs
5602 Bugzilla : 5990, 6223
5603 Description: lfs getstripe would oops on a very large filesystem
5604 Details : lov_getconfig used kfree on vmalloc'd memory
5607 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5608 Bugzilla : 6242, 6243
5609 Description: reading from files that had been truncated to a non-zero size
5610 but never opened returned no data
5611 Details : ll_file_read() reads zeros from no-object files to EOF
5616 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5617 Details : There's little harm in aborting MDS/OSS recovery and letting it
5618 try again, so I removed the LASSERT and return an error instead.
5620 Severity : enhancement
5622 Description: New debugging infrastructure for tracking down data corruption
5623 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5624 (b) cover more of the client-side code path, and (c) try to narrow
5625 down where problems occurred
5629 Bugzilla : 3819, 4364, 4397, 6313
5630 Description: Racing close and eviction MDS could cause assertion in mds_close
5631 Details : It was possible to get multiple mfd references during close and
5632 client eviction, leading to one thread referencing a freed mfd.
5634 Severity: : enhancement
5635 Bugzilla : 3262, 6359
5636 Description: Attempts to reconnect to servers are now more aggressive.
5637 Details : This builds on the enhanced upcall-less recovery that was added
5638 in 1.4.2. When trying to reconnect to servers, clients will
5639 now try each server in the failover group every 10 seconds. By
5640 default, clients would previously try one server every 25 seconds.
5645 Description: After recovery, certain operations trigger a failed
5646 assertion on a client.
5647 Details : Failing over an mds, using lconf -d --failover, while a
5648 client was doing a readdir() call would cause the client to
5649 LBUG after recovery completed and the readdir() was resent.
5651 Severity : enhancement
5653 Description: Default groups are now added by lconf
5654 Details : You can now run lconf --group <servicename> without having to
5655 manually add groups with lmc.
5658 Frequency : occasional
5660 Description: Nodes with an elan id of 0 trigger a failed assertion
5663 Frequency : always when accessing e.g. tty/console device nodes
5665 Description: tty and some other devices nodes cannot be used on lustre
5666 Details : file's private_data field is used by device data and lustre
5667 values in there got lost. New field was added to struct file to
5668 store fs-specific private data.
5671 Frequency : when exporting Lustre via NFS
5673 Description: NFSD failed occasionally when looking up a path component
5674 Details : NFSD is looking up ".." which was broken in ext3 directories
5675 that had grown large enough to become hashed.
5678 Frequency : Clusters with multiple interfaces not on the same subnet
5680 Description: Nodes will repeatedly try to reconnect to an interface which it
5681 cannot reach and report an error to the log.
5682 Details : Extra peer list entries will be created by lconf with some peers
5683 unreachable. lconf now validates the peer before adding it.
5686 Frequency : Only if a default stripe is set on the filesystem root.
5688 Description: Setting a default stripe on the filesystem root prevented the
5689 filesystem from being remounted.
5690 Details : The client was sending extra request flags in the root getattr
5691 request and did not allocate a reply buffer for the dir EA.
5694 Frequency : occasional, higher if lots of files are accessed by one client
5695 Bugzilla : 6159, 6097
5696 Description: Client trips assertion regarding lsm mismatch/magic
5697 Details : While revalidating inodes the VFS looks up inodes with ifind()
5698 and in rare cases can find an inode that is being freed.
5699 The ll_test_inode() code will free the lsm during ifind()
5700 when it finds an existing inode and then the VFS later attaches
5701 this free lsm to a new inode.
5705 Bugzilla : 6422, 7030
5706 Description: MDS deadlock between mkdir and client eviction
5707 Details : Creating a new file via mkdir or mknod (starting a transaction
5708 and getting the ns lock) can deadlock with client eviction
5709 (gets ns lock and trying to finish a synchronous transaction).
5712 Frequency : occasional
5713 Description: While starting a server, the fsfilt_ext3 module could not be
5715 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5716 kernels. Previously, lconf would still use the ext3 name
5717 when trying to load modules. Now, it will correctly use
5718 ext3 on 2.4 and ldiskfs on 2.6.
5720 Severity : enhancement
5721 Description: The default stripe count has been changed to 1
5722 Details : The interpretation of the default stripe count (0, to lfs
5723 or lmc) has been changed to mean striping across a single
5724 OST, rather than all available. For general usage we have
5725 found a stripe count of 1 or 2 works best.
5727 Severity : enhancement
5728 Description: Add support for compiling against Cray portals.
5729 Details : Conditional compiling for some areas that are different
5733 Frequency : occasional
5734 Bugzilla : 6409, 6834
5735 Description: Creating files with an explicit stripe count may lead to
5736 a failed assertion on the MDS
5737 Details : If some OSTs are full or unavailable, creating files may
5738 trigger a failed assertion on the MDS. Now, Lustre will
5739 try to use other servers or return an error to the
5743 Frequency : occasional
5745 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5746 caused lock timeout during readahead (since 1.4.2).
5747 Details : Processes doing readahead might match a lock that hasn't been
5748 granted yet if there are overlapping and conflicting lock
5749 requests. The readahead process waits on ungranted lock
5750 (original lock is CBPENDING), while OST waits for that process
5751 to cancel CBPENDING read lock and eventually evicts client.
5753 Severity : enhancement
5755 Description: Initial enabling of flock support for clients
5756 Details : Implements fcntl advisory locking and file status functions.
5757 This feature is provided as an optional mount flag (default
5758 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5759 locking are implemented yet, and those that are are not guaranteed
5760 to be completely correct in production environments.
5761 mount -t lustre -o [flock|noflock] ...
5764 Frequency : occasional
5766 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5767 assertion in the ext3 JBD (journaling) layer.
5768 Details : The b_committed_data struct is protected by the big kernel lock
5769 in 2.4 kernels, serializing journal_commit_transaction() and
5770 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5771 access to this struct. In 2.6 kernels there is finer grained
5772 locking to improve SMP performance of the JBD layer.
5776 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5781 Description: merge and cleanup kernel patches.
5784 -----------------------------------------------------------------------------
5786 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5788 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5789 versions of Lustre. Please update all servers and clients to
5790 version 1.4.2 or later at the same time. You must also run
5791 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5792 to update the configuration logs.
5794 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5795 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5796 - swab llog records of type '0' so we get proper header size/idx (5861)
5797 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5798 - fix rename of one directory over another leaking an inode (5953)
5799 - avoid SetPageDirty on 2.6 (5981)
5800 - don't re-add just-being-destroyed locks to the waiting list (5653)
5801 - when creating new directories, inherit the parent's custom
5802 striping settings if present parent (3048)
5803 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5804 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5805 - don't hold client locks on temporary worklist from l_lru (5666)
5806 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5807 - automatically evict dead clients (5921)
5808 - Update file size properly in create+truncate+fstat case (6196)
5809 - Do not unhash mountpoint dentries, do not allow removal of
5811 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5812 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5813 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5814 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5815 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5816 - Correctly handle reads of files with no objects (6243)
5817 - lctl recover will also mark a device active if deactivate used (5933)
5819 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5820 - allow --write-conf on an MDS with different nettype than client (5619)
5821 - don't write config llogs to MDS for mounts not from that MDS (5617)
5822 - lconf should create multiple TCP connections from a client (5201)
5823 - init scripts are now turned off by default; run chkconfig --on
5824 lustre and chkconfig --on lustrefs to use them
5825 - upcalls are no longer needed for clients to recover to failover
5827 - add --abort-recovery option to lconf to abort recovery on device
5829 - add support for an arbitrary number of OSTs (3026)
5830 - Quota support protocol changes.
5831 - forward compatibility changes to wire structs (6007)
5832 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5833 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5834 - improved "lustre status" script
5835 - initialize blocksize for non-regular files (6062)
5836 - added --disable-server and --disable-client configure options (5782)
5837 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5838 - Vanilla 2.4.29 support
5839 - increase maximum number of obd devices to 520 (6242)
5840 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5841 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5842 SLES 9 2.6.5 kernels
5843 - limit stripes per file to 160 (the maximum EA size) (6093)
5845 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5848 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5849 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5850 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5851 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5852 - small ext3 extents cleanups and fixes (5733)
5853 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5854 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5855 - avoid needless client->OST connect, fix handle mismatch (5317)
5856 - fix DLM error path that led to out-of-sync client, long delays (5779)
5857 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5858 - fix several locking issues related to i_size (5492,5624,5654,5672)
5859 - don't move pending lock onto export if it is already evicted (5683)
5860 - fix kernel oops when creating .foo in unlinked directory (5548)
5861 - fix deadlock in obdfilter statistics vs. object create (5811)
5862 - use time_{before,after} to avoid timer jiffies wrap (5882)
5863 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5864 - Do not leak request if server was not able to process it (5154)
5865 - If mds_open unable to find parent dir, make that negative lookup(5154)
5866 - don't create new directories with extent-mapping (5909, 5936)
5868 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5869 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5870 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5871 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5872 - support for RHEL4 kernel on i686 (5773)
5873 - provide error messages when incompatible logs are encountered (5898)
5875 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5876 * version 1.4.0.10 (1.4.1 release candidate 1)
5878 - don't keep a lock reference when lock is not granted (4238)
5879 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5880 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5881 - drop import inflight refcount on signal_completed_replay error (5255)
5882 - unlock page after async write error during send (3677)
5883 - handle missing objects in filter_preprw_read properly (5265)
5884 - no transno return for symlink open, don't save no-trasno open (3440)
5885 - don't try to complete elan receive that already failed (4012)
5886 - free RPC server reply state on error (5406)
5887 - clean up thread from ptlrpc_start_thread() on error (5160)
5888 - readahead could read extra page into cache that wasn't ejected (5388)
5889 - prevent races in class_attach/setup/cleanup/detach (5260)
5890 - don't dereference de->d_inode after l_dput of de (5458)
5891 - use "int" for stripe value returned from lock_to_stripe (5544)
5892 - mballoc allocation and error-checking fixes in 2.6 (5504)
5893 - block device patches to fix I/O request sizes in 2.6 (5482)
5894 - look up hostnames for IB nals (5602)
5895 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5896 - don't start multiple acceptors for the same port (5277)
5897 - fix incorrect LASSERT in mds_getattr_name (5635)
5898 - export a proc file for general "ping" checking (5628)
5899 - fix "lfs check" to not block when the MDS is down (5628)
5901 - service request history (4965)
5902 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5903 - create an "evict_client" /proc entry on OSTs, like the MDS has
5904 - fix mount usage message, return errors per mount(8) (5168)
5905 - change grep [] to grep "[]" in tests so they work in more UMLs
5906 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5907 - remove ancient LOV_MAGIC_V0 EA support (5047)
5908 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5909 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5910 - allow readahead allocations to fail when low on memory (5383)
5911 - mmap locking landed again, after considerable improvement (2828)
5912 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5914 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5917 - send OST transaction number in read/write reply to free req (4966)
5918 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5919 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5920 - the watchdog thread now runs as interruptible (5246)
5921 - flock/lockf fixes (but it's still disabled, pending 5135)
5922 - don't use EXT3 constants in llite code (5094)
5923 - memory shortage at startup could cause assertion (5176)
5925 - reorganization of lov code
5926 - single portals codebase
5928 - add extents/mballoc support (5025)
5929 - direct I/O reads in the obdfilter (4048)
5930 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5932 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5935 - send OST transaction number in read/write reply to free req (4966)
5936 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5937 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5938 - the watchdog thread now runs as interruptible (5246)
5939 - handle missing objects in filter_preprw_read properly (5265)
5940 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5941 - drop import inflight refcount on signal_completed_replay error (5255)
5942 - unlock page after async write error during send (3677)
5943 - return original error code on reconstructed replies (3761)
5944 - no transno return for symlink open, don't save no-trasno open (3440)
5946 - add pid to ldlm debugging output (4922)
5947 - bump the watchdog timeouts -- we can't handle 30sec yet
5948 - extra debugging for orphan dentry/inode bug (5259)
5950 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5953 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5954 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5955 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5956 - fix NULL dereference in /proc/sys/portals/routes (4827)
5957 - allow failed mdc_close() operations to be interrupted (4561)
5958 - stop precreate on OST before MDS would time out on it (4778)
5959 - don't send partial-page writes before EOF from client (4410)
5960 - discard client grant for sub-page writes on large-page clients (4520)
5961 - don't free dentries not owned by NFS code, check generation (4806)
5962 - fix lsm leak if mds_create_objects() fails (4801)
5963 - limit debug_daemon file size, always print CERROR messages (4789)
5964 - use transno after validating reply (3892)
5965 - process timed out requests if import state changes (3754)
5966 - update mtime on OST during writes, return in glimpse (4829)
5967 - add mkfsoptions to LDAP (4679)
5968 - use ->max_readahead method instead of zapping global ra (5039)
5969 - don't interrupt __l_wait_event() during strace
5971 - add software watchdogs to catch hung threads quickly (4941)
5972 - make lustrefs init script start after nfs is mounted
5973 - fix CWARN/ERROR duplication (4930)
5974 - return async write errors to application if possible (2248)
5975 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5976 - print NAL number in %x format (4645)
5977 - update barely-supported suse-2.4.21-171 series (4842)
5978 - support for sles 9 %post scripts
5979 - support for building 2.6 kernel-source packages
5980 - support for sles km_* packages
5982 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5985 - ignore -ENOENT errors in osc_destroy (3639)
5986 - notify osc create thread that OSC is being cleaned up (4600)
5987 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5988 - reconstruct ost_handle() like mds_handle() (4657)
5989 - create a new thread to do import eviction to avoid deadlock (3969)
5990 - let lconf resolve symlinked-to devices (4629)
5991 - don't unlink "objects" from directory with default EA (4554)
5992 - hold socknal file ref over connect in case target is down (4394)
5993 - allow more than 32000 subdirectories in a single directory (3244)
5994 - fix blocks count for O_DIRECT writes (3751)
5995 - OST returns ENOSPC from object create when no space left (4539)
5996 - don't send truncate RPC if file size isn't changing (4410)
5997 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5998 - bind to privileged port in socknal and tcpnal (3689)
6000 - rate limit CERROR/CWARN console message to avoid overload (4519)
6001 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6002 - basic mmap support (3918)
6003 - kernel patch series update from b1_4 (4711)
6005 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6008 - avoid crash during MDS cleanup with OST shut down (2775)
6009 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6010 - don't use bad inodes on the MDS (3744)
6011 - dynamic object preallocation to improve recovery speed (4236)
6012 - don't hold spinlock over lock dumping or change debug flags (4401)
6013 - don't zero obd_dev when it is force cleaned (3651)
6014 - print grants to console if they go negative (4431)
6015 - "lctl deactivate" will stop automatic recovery attempts (3406)
6016 - look for existing locks in ldlm_handle_enqueue() (3764)
6017 - don't resolve lock handle twice in recovery avoiding race (4401)
6018 - revalidate should check working dir is a directory (4134)
6020 - don't always mark "slow" obdfilter messages as errors (4418)
6022 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6025 - don't close LustreDB during write_conf until it is done (3860)
6026 - fix typo in lconf for_each_profile (3821)
6027 - allow dumping logs from multiple threads at one time (3820)
6028 - don't allow multiple threads in OSC recovery (3812)
6029 - fix debug_size parameters (3864)
6030 - fix mds_postrecov to initialize import for llog ctxt (3121)
6031 - replace config semaphore with spinlock (3306)
6032 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6033 - don't allow enqueue to complete on a destroyed export (3822)
6034 - down write_lock before checking llog header bitmap (3825)
6035 - recover from lock replay timeout (3764)
6036 - up llog sem before sending rpc (3652)
6037 - reduce ns lock hold times when setting kms (3267)
6038 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6039 - fix NULL deref and obd_dev leak on setup error (3312)
6040 - replace some LBUG about llog ops with error handling (3841)
6041 - don't match INVALID dentries from d_lookup and spin (3784)
6042 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6043 - fix invalid assertion in ptlrpc_set_wait (3880)
6045 - add libwrap support for the TCP acceptor (3996)
6046 - add /proc/sys/portals/routes for non-root route listing (3994)
6047 - allow setting MDS UUID in .xml (2580)
6048 - print the stack of a process that LBUGs (4228)
6050 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6053 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6054 - make sure to unset replay flag from failed open requests (3440)
6055 - if default stripe count is 0, use OST count for inode size (3636)
6056 - update parent mtime/ctime on client for create/unlink (2611)
6057 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6058 - free recovery state on server during a forced cleanup (3571)
6059 - unregister_reply for resent reqs (3063)
6060 - loop back devices mounting and status check on 2.6 (3563)
6061 - fix resource-creation race that can provoke i_size == 0 (3513)
6062 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6063 - more debugging for page-accounting assertion (3746)
6064 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6065 - avoid deadlock after precreation failure (3758)
6066 - fix race and lock order deadlock in orphan handling (3450, 3750)
6067 - add validity checks when grabbing inodes from l_ast_data (3599)
6069 - add /proc/.../recovery_status to obdfilter (3428)
6070 - lightweight CDEBUG infrastructure, debug daemon (3668)
6071 - change default OSC RPC parameters to be better on small clusters
6072 - turn off OST read cache for files smaller than 32MB
6073 - install man pages and include them in rpms (3100)
6074 - add new init script for (un)mounting lustre filesystems (2593)
6075 - run chkconfig in %post for init scripts (3701)
6076 - drop scimac NAL (unmaintained)
6078 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6081 - clean kiobufs before and after use (3485)
6082 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6083 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6084 - change init script to fail more clearly if not run as root (1528)
6085 - allow clients to reconnect during replay (1742)
6086 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6087 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6088 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6089 - fix cond_resched (3554)
6090 - only evict unfinished clients after recovery (3515)
6091 - allow bulk resend, prevent data loss (3570)
6092 - dynamic ptlrpc request buffer allocation (2102)
6093 - don't allow unlinking open directory if it isn't empty (2904)
6094 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6095 - remove extraneous obd dereference causing LASSERT failure (3334)
6096 - don't use get_cycles() when creating temp. files on the mds (3156)
6097 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6098 - handle EEXIST for set-stripe, set proper directory name (3336)
6100 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6101 - fix ksocknal_fmb_callback() error messages (2918)
6103 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6106 - don't copy lvb into (possibly NULL) reply on error (2983)
6107 - don't deref dentry after dput, don't free lvb on error (2922)
6108 - use the kms to determine writeback rpc length (2947)
6109 - increment oti_logcookies when osc is inactive (2948)
6110 - update client's i_blocks count via lvb messages (2543)
6111 - handle intent open/close of special files properly (1557)
6112 - mount MDS with errors=remount-ro, like obdfilter (2009)
6113 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6114 - don't use cancelling-locks' kms values (2947)
6115 - use highest lock extent for kms, not last one (2925)
6116 - don't dereference ERR_PTR() dentry in error handling path (3107)
6117 - fix thread race in portals_debug_dumplog() (3122)
6118 - create lprocfs device entries at setup instead of at attach (1519)
6119 - common AST error handler, don't evict client on completion race (3145)
6120 - zero nameidata in detach_mnt in 2.6 (3118)
6121 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6122 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6123 - initialize RPC timeout timer earlier for 2.6 (3219)
6124 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6125 - print nal/nid for unknown nid (3258)
6126 - additional checks for oscc recovery before doing precreate (3284)
6127 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6128 - don't crash in mdc_close for bad permissions on open (3285)
6129 - zero i_rdev for non-device files (3147)
6130 - clear page->private before handing to FS, better assertion (3119)
6131 - tune the read pipeline (3236)
6132 - fix incorrect decref of invalidated dentry (2350)
6133 - provide read-ahead stats and refine rpc in flight stats (3328)
6134 - don't hold journal transaction open across create RPC (3313)
6135 - update atime on MDS at close time (3265)
6136 - close LDAP connection when recovering to avoid server load (3315)
6137 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6138 - don't leak open file on MDS after open resend (3325)
6139 - serialize filter_precreate and filter_destroy_precreated (3329)
6140 - loop device shouldn't call sync_dev() for nul device (3092)
6141 - clear page cache after eviction (2766)
6142 - resynchronize MDS->OST in background (2824)
6143 - refuse to mount the same filesystem twice on same mountpoint (3394)
6144 - allow llmount to create routes for mounting behind routers (3320)
6145 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6146 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6147 - fix rare problem with rename on htree directories (3417)
6149 - allow default OST striping configuration per directory (1414)
6150 - fix compilation for qswnal for 2.6 kernels (3125)
6151 - increase maximum number of MDS request buffers for large systems
6152 - change liblustreapi to be useful for external progs like lfsck (3098)
6153 - increase local configuration timeout for slow disks (3353)
6154 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6156 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6159 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6160 - don't overwrite extent policy data in reply if lock was blocked (2901)
6161 - drop filter export grants atomically with removal from device (2663)
6162 - del obd_self_export from work_list in class_disconnect_exports (2908)
6163 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6164 - swab reply message in mdc_close, other PPC fixes (2464)
6165 - fix destroying of named logs (2325)
6166 - overwrite old logs when running lconf --write_conf (2264)
6167 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6168 - fix race in target_handle_connect (2898)
6169 - mds_reint_create() should take same inode create lock (2926)
6170 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6171 - don't close files for self_export to avoid uninitialized obd (2936)
6172 - allow MDS with the same name as client node (2939)
6173 - hold dentry reference for closed log files for unlink (2325)
6174 - reserve space for all logs during transactions (2059)
6175 - don't evict page beyond end of stripe extent (2925)
6176 - don't oops on a deleted current working directory (2399)
6177 - handle hard links to targets without a parent properly (2517)
6178 - don't dereference NULL lock when racing during eviction (2867)
6179 - don't grow lock extents when lots of conflicting locks (2919)
6181 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - account for cache space usage on clients to avoid data loss (974)
6185 - lfsck support in lustre kernel code (2349)
6186 - reduce journal credits needed for BRW writes (2370)
6187 - orphan handling to avoid losing space on client/server crashes
6188 - ptlrpcd can be blocked, stopping ALL progress (2477)
6189 - use lock value blocks to assist in proper KMS, faster stat (1021)
6190 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6191 - recovery for initial connections (2355)
6192 - fixes for mds_cleanup_orphans (1934)
6193 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6194 - block all file creations until orphan recovery completes (1901)
6195 - client remove rq_connection from request struct (2423)
6196 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6197 - recovery timer race (2670)
6198 - mdc_close recovey bug (2532)
6199 - ptlrpc cleanup bug (2710)
6200 - mds timeout on local locks (2588)
6201 - namespace lock held during RPCs (2431)
6202 - handle interrupted sync write properly (2503)
6203 - don't try to handle a message that hasn't been replied to (2699)
6204 - client assert failure during cleanup after abort recovery (2701)
6205 - leak mdc device after failed mount (2712)
6206 - ptlrpc_check_set allows timedout requests to complete (2714)
6207 - wait for inflight reqs when ptlrpcd finishes (2710)
6208 - make sure unregistered services are removed from the srv_list
6209 - reset bulk XID's when resending them (caught by 1138 test)
6210 - unregister_bulk after timeout
6211 - fix lconf error (2694)
6212 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6213 - readahead locks pages, leaves pending causing memory pressure (2673)
6214 - increase OST request buffers to 4096 on large machines (2729)
6215 - fix up permission of existing directories in simple_mkdir (2661)
6216 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6217 - don't assign transno to errored transactions (2742)
6218 - don't delete objects on OST if given a bogus objid from MDS (2751)
6219 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6220 - if rq_no_resend, then timeout request after recovery (2432)
6221 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6222 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6223 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6224 - MDS recovery shouldn't delete objects that it has given out (2730)
6225 - if enqueue arrives after completion, don't clobber LVB (2819)
6226 - don't unlock pages twice when trigger_group_io returns error (2814)
6227 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6228 - don't write pages to disk if there was an error (1450)
6229 - don't ping imports that have recovery disabled (2676)
6230 - take buffered bytes into account when balancing socknal conn (2817)
6231 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6232 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6233 - remove little-endian swabbing of llog records (1987)
6234 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6235 - retry reposting request buffers if they fail (1191)
6236 - grow extent at grant time to avoid granting a revoked lock (2809)
6237 - lock revoke doesn't evict page if covered by a second lock (2765)
6238 - disable VM readahead to avoid reading outside lock extents (2805)
6240 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6241 - updated kernel patches for hp-2.4.20 kernel (2681)
6243 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6246 - fix truncated write corruption (2366)
6247 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6249 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6250 - ensure len(uuid) < 37 in lmc (1171)
6251 - fix ia64 OOPS in llog_test (2255)
6252 - zero end of page at obdfilter for partial page writes (2648)
6253 - don't leave stale dentries around after renames (bug 2428)
6254 - fix timeouts when evicting a client with a single lock held (2642)
6255 - set deadline for the initial HELLO message to drain (2634)
6256 - print out dotted-quad IP addresses in the socknal (2302)
6258 - additional debugging for MDS client eviction problem (2443)
6259 - fix mkfsoptions support for osts (2603, 2604)
6261 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6264 - add series for the vanilla 2.6.0 kernel
6265 - add series for the vanilla 2.4.24 kernel
6266 - add series for a cray x86/64 UL kernel drop
6267 - fix xattr patches for the vanilla 2.4.19 series
6269 - generate true UUIDs in lmc (1171)
6270 - have portals stack dumping break in UML (2466)
6271 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6272 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6273 - wake callback waiting threads on client eviction (2460)
6274 - Add --ptldebug and --subsystem to lmc (1719)
6275 - update assertion to allow safe interrupt allocation
6276 - set rq_no_resend for cancel requests (2432)
6277 - recalculate ptlrpcd timeout after resend (2494)
6278 - call vfs_rmdir when removing pending directories (2368)
6279 - fix renaming a file to itself (2429)
6280 - lmc creates a default one-stripe lov (2454)
6281 - expand procfs space to handle large clusters (2326)
6282 - increase UML stack to avoid overflow
6283 - update lconf's list of debug and subsystem masks
6284 - fix lfs find --obd (2510)
6285 - /proc tunable for disabling filter read caching (2591)
6286 - stop rpm packages from altering slapd.conf (2301)
6287 - disable nagle in the socknal under 0conf (2578)
6288 - choose mds inode size based on stripe count (2572)
6289 - fix kernel-source rpm problems (2516)
6291 - add --disable-doc to avoid pdf generation (2421)
6292 - update documentation, tests, type-os, comments
6293 - avoid format warnings on ia64
6294 - remove the TOE NAL
6295 - tiny code cleanups by removing unused fields
6297 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6300 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6301 - remove the most blatant lies from BUILDING file (2371)
6302 - change default debug level to reasonable production setting
6303 - reduce client side cache size to reduce cache flush time
6304 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6305 - make TCP zerocopy and pinger support enabled by default (2476)
6306 - sync writes completed after process exits caused crashes (2319)
6307 - maintain correct mount count on the MDS (2356)
6308 - backout 1557, because 2316 wasn't really fixed
6309 - better file I/O statistics gathering in /proc
6310 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6311 - another kernel patch to fix zero-copy TCP function export
6312 - don't take duplicate lock when processing re-sent getattr (2420)
6313 - lctl uses obd_self_export instead of creating new conn (2353)
6314 - MDS/OST recovery case which requires object creation asserted (2425)
6315 - move lfs from /usr/sbin to /usr/bin in packages
6316 - fix race between mds_client_add and mds_client_free (2417)
6317 - use kmalloc instead of slabs in portals (2430)
6318 - don't create duplicate records when a failover MDS is present (2442)
6319 - remove unnecessary mount age check (2332)
6320 - don't remove directory inodes from locks prematurely (2451)
6321 - don't break if MDS service name is the same as hostname (2103)
6322 - fix races in client write RPC generation when cache full (2482)
6324 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6327 - remove now-unused request->rq_obd (278)
6328 - if an allocation fails, print out how much memory we've used (1933)
6329 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6330 - add the "configurable stack size" patch to most series files (1256)
6331 - ability to write large log records, for 100+ OST configs (2306)
6332 - fix NULL deref when filter_prep fails (2314)
6333 - fix operator precedence error in filter_sync
6334 - dynamic allocation of socknal TX descriptors (2315)
6335 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6336 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6337 - ensure that CWARN messages in Portals always get to the syslog
6338 - __init/__exit are not for prototype decls (ldlm_init/exit)
6339 - x86-64 compile warning fixes
6340 - fix gateway LMC keyword conflict (2318)
6341 - fix MDS lock inversions in getattr/reint paths (1844)
6342 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6343 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6344 - DLM race condition prevented some lock evictions (2328)
6345 - ENOMEM detection and retry on socknal sends (2230)
6346 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6347 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6348 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6349 - better error messages when a client is rejected during recovery (1505)
6350 - avoid cancelling locks which were never granted, after failure (2330)
6351 - fix i_sem/journal inversion in mds_client_add (2333)
6352 - fix truncate/getattr lock cycle deadlock (2334)
6353 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6354 - fix two rare exit paths which could leak an l_lock() ref (2321)
6355 - fencepost error in MDS/OST orphan recovery (2226)
6356 - make log record alignment 8 bytes (1988)
6357 - lstripe now fails when requested offset > ost_count (2237)
6358 - ensure that all kernel series have a complete list.h (1607)
6359 - fix crashes in special-file operations (2316)
6360 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6362 - allow configurable automake binary, for testing new versions
6363 - small update to the lfs documentation
6365 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6367 * fix negative export reference count in fsfilt_sync (2312)
6369 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6370 * release candidate 0.9.1
6372 - orphans are moved into the PENDING directory for possible recovery
6373 - replayed opens now open by fid for orphan/rename safety (1042)
6374 - last close of an orphan inode generates a transno (683)
6375 - chdir() and mount() now pin the directory entry (1020)
6376 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6377 - discard very old requests without processing them (1502)
6378 - remove some common, well-understood CERRORs (1505)
6379 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6380 - clear "grant" flags in OST replies until OST grant code lands (1644)
6381 - fix read performance by not clobbering i_blksize on client (1598)
6382 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6383 - make LRU size a /proc tunable, clears locks when reduced (707)
6384 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6385 - server locks take references on exports now (1558)
6386 - build fixes for 2.4.20-rh trees (1663)
6387 - return an error from lov_create if all OSCs are inactive (1751)
6388 - fix import levels when a reconnect happens without a timeout (1597)
6389 - exit early from mds_open if we get a lookup error (1749)
6390 - partial page read at EOF wouldn't wait for disk before sending (1642)
6391 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6392 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6393 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6394 - write/truncate lock inversion (1639)
6395 - Don't auto-load obdclass, portals modules during cleanup (1495)
6396 - fix timestamps from jumping to "now" (1763)
6397 - extra journal assertions (1648)
6398 - add an extra multiunlink test (1771)
6399 - fix read_record/write_record API (1776)
6400 - fix leak of offset_extent, possible incorrect i_size later (1772)
6401 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6402 - Debugging for the kqswnal_get_idle_tx problems (1820)
6403 - Allow recovery to be attempted multiple times (1536)
6404 - Write out MDS last_rcvd file after it is first created (1600)
6405 - Fix tx_descriptor leak in failed transmit situations (1827)
6406 - ext3 journaling fixes for assertion failure after IO error (1871)
6407 - class_export_put() on freed export after completion AST error (1896)
6408 - Fix revalidate looping in VFS (1322)
6409 - Don't access a freed export during MDS_REINT timeout (1521)
6410 - Add open-unlink recovery support on the MDS (1673,1764)
6411 - Return an error if no MDS data was read from last_rcvd (1946)
6412 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6413 - Don't setup a disk device that doesn't match exported UUID (317)
6414 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6415 - avoid committing NULL handle in force close
6416 - local.sh is now a one-stripe LOV configuration
6417 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6418 - fix ext3 htree duplicate directory entry corruption (1516)
6419 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6420 - update to new LOV EA format (2097)
6421 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6423 2003-06-15 Phil Schwan <phil@clusterfs.com>
6426 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6427 - per-import/export recovery handling (958, 931, 959)
6428 - multiple last-rcvd slots, for serving multiple FSes (949)
6429 - connections are again shared between multiple imp/exports (963, 964)
6430 - "umount -f" would hang if any requests needed to be sent (393, 978)
6431 - avoid pinning large req buffer by copying for queued messages (989)
6432 - add "uuid" to "lctl device" command to help upcalls (991)
6433 - "open" RPCs with transnos would confuse recovery counters (1037)
6434 - do proper endian conversion of all wire messages (288, 340, 891)
6435 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6436 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6437 - fix ptlrpc_connection leak in target_handle_connect (1174)
6438 - fix import refcounting bug in OST and MDS cleanup (1134)
6439 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6440 - fix very unlikely obd_types race condition (501)
6441 - remove osc_open hack for echo_client (1187)
6442 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6443 - a failure in read_inode2 leads to deadlock (1139)
6444 - cancel ack-locks as soon as transaction is committed (1072)
6445 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6446 - make sure to commitrw after any preprw to avoid deadlock (1162)
6447 - failing to execute a file in a lustre FS would lock inode (1203)
6448 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6449 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6450 - obd_uuid2tgt would walk off the end of the list (1255)
6451 - on IA64 the osc would give portals incorrect bulk size (1258)
6452 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6453 - fix lock inversion caused by new llite matching code (1282)
6454 - limit the number of dirty pages on a client to 10MB (1286)
6455 - timed out locks were not being corrected cancelled (1289)
6456 - fix O_DIRECT above 4GB on IA-32 (1292)
6457 * major user-visible changes
6458 - fail out/fail over policy now controlled by the upcall (993)
6460 - add OBD_PING to check server availability and failure (954)
6461 - lustre messages are now sent in sending host order (288, 340, 891)
6462 - add eadatalen to MDS getattr reply (340)
6463 - OST read replies may contain second buffer, with per-page status (593)
6465 2003-03-11 Phil Schwan <phil@clusterfs.com>
6468 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6469 - failed open()s could cause deadlock; fixed (867, 869)
6470 - stop cancelling OST locks when files are closed (481)
6471 - overlapping XID spaces caused network corruption (851, 853)
6472 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6473 - setattr_raw added, to do single-RPC, server-side setattrs
6474 - lmc/lconf syntax change for OST UUIDs
6475 - fix crashy race condition between ptlrpc_free_req and osc_close
6476 - don't use request in mdc_enqueue if we hit a timeout (889)
6477 - don't set the inode i_size for regular files from the MDS (896)
6478 - handle out of order completion AST (842)
6479 - don't LBUG if a lock request times out after receiving AST (913)
6480 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6481 - if a bad lock AST arrives, send an error instead of dropping entirely
6482 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6483 - fix leak in bulk IO when only partially completed (899, 900, 926)
6484 - fix O_DIRECT for ia64 (55)
6485 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6486 - C-z after timeout could hang a process forever; fixed (977)
6488 - client-side I/O cache (678, 924, 929, 941, 970)
6490 - READPAGE and SETATTRs which don't take server-side locks get
6493 2003-02-11 Phil Schwan <phil@clusterfs.com>
6496 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6497 - Fix fsfilt_extN_readpage() to read a full page of directory
6498 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6499 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6500 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6501 - Most intent operations were removed, in favour of a new RPC mode
6502 that does a single RPC to the server and bypasses most of the VFS
6503 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6504 - Aggressively cancel local locks on DLM servers
6505 - mds_reint_unlink sends EA to the client if it's the last nlink.
6506 client uses that EA to unlink OST objects.
6507 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6508 - recursive symlinks were fixed (439)
6509 - fixed NULL deref in DEBUG_REQ
6510 - filter_update_lastobjid no longer calls sync, which annoyed extN
6511 - fixed multi-client small-writes to a single file problem (445)
6512 - fixed mtime updates during file writes (607)
6513 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6514 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6515 - fixed Linux OST size reporting problem (444, 656)
6516 - OST now updates object mtime with writes or setattr (607, 619)
6517 - client verifies file size before zeroing page past EOF (445)
6518 - OST now writes last allocated objid to disk with allocation (108)
6519 - LOV on echo now works (409)
6521 - mds_reint_unlink sends a new buffer, with the EA included. this
6522 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6523 set if a regular file was being unlinked, and it was the last link
6524 - use PtlGet from the target for bulk writes (315)
6525 - OST now updates object mtime with writes or setattr (607, 619)
6526 - LDLM now has a grant-time callback to revalidate locked items, if
6528 - Many MDS operations were reorganized to combat race conditions
6530 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6531 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6533 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6536 - Fully reactivate OST imports after reconnection (512, others)
6537 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6538 - More graceful error handling for truncating on dead OST (515)
6539 - Don't error out unless we're actually accessing dead stripes (474)
6540 - Fix garbage sizes when stripes are missing (410)
6541 - LRU counters were broken, causing constant lock purge (433, 432)
6542 - garbage on read from stripes with failed OSTs (441)
6543 - mark OSCs as active before reconnecting during recovery (438)
6544 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6545 - lfind did not preserve OST order in output (443)
6546 - symlinks cause hung clients, incorrect data (439)
6547 - stop dereferencing request after dropping refcount (457)
6548 - don't LASSERT(spin_is_locked) on non-SMP (455)
6549 - fixes for many rename() bugs
6550 - fstat didn't correctly synchronize attributes (399)
6551 - server must handle lock cancellation during blocking AST prep (487)
6552 - bulk descriptors were free()d too soon (511)
6553 - fix paths in lconf, which would load incorrect modules (451, 507)
6554 - fix confusing lconf 'host not found' error message (386)
6555 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6556 - fix race condition in mdc_blocking_ast() for inode access (526)
6557 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6558 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6559 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6560 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6561 - fix lstripe in several ways (kernel side) (527)
6562 - fix request leak in ldlm_cli_enqueue (262)
6563 - incorrect OSC was marked inactive after OST failure
6564 - call mds_fs_cleanup before unmounting filesystem (524)
6565 - fix races between taking ns_lock and ldlm_lock_change_resource
6566 - fix races updating LOV export open file list
6567 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6568 - fix recovery NULL deref in ldlm_cli_cancel_unused
6569 - fix some DLM races by using new hash table for lock handles (419)
6570 - permit the client to specify desired inodes, at replay
6571 - duplicate requests when we queue them for replay reintegration
6572 - fix last_rcvd offset calculation
6573 - sync after each recovered transaction, so we always make progress
6574 - never, not always, ERESTART requests without transnos
6575 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6577 - skip replay if the MDS says that the client is already connected
6578 - don't check for a recovery-enabled export to match lctl's UUID
6579 - don't INC_USE_COUNT for phantom exports
6580 - don't crash when cleaning up phantom exports (567)
6581 - don't double-finish or set replay data for errored mdc_open requests
6582 - abort requests when they time out, so we don't get old replies
6583 - send/receive replies for AST messages again
6584 - if the client says that it doesn't have the lock, cancel it on the
6586 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6587 mark it as destroyed, it will all work out when decref is called
6588 - fix module use counts (22, 581)
6590 - ASTs now expect a reply (server cancels lock on error reply)
6592 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6595 - fix many simultaneous client startup (392)
6596 - fix dentry->d_it clobbering
6597 - credentials weren't being shipped for readdir/getattr operations
6598 - remove invalid assertions triggered during some concurrent MD
6600 - proper Lustre versions added (336, 389)
6601 - fix memory leak for create error case (398)
6602 - fix LOV locking bug that would get cli/srv out of sync
6603 - fix echo client over LOV (409)
6604 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6605 - fix double-O_EXCL intent crash (424)
6606 - avoid sending multiple lock CANCELs (352)
6608 - MDS can do multi-client recovery (modulo bugs in new code)
6610 - many updates, edits, cleanups
6612 2002-11-18 Phil Schwan <phil@clusterfs.com>
6615 - fix null d_it dereference (346)
6616 - fix full OST/dbench hang (333)
6617 - fix permission problem with file removal (286)
6618 - fix removal of OSCs from LOV when they fail
6619 - fix NULL deref during bulk timeout (214)
6620 - fix problems related to multiple filesystems on one MDS (241)
6621 - fixed serious subtle metadata locking bugs
6622 - free locks on clients when inodes are removed due to memory
6624 - fix inode pointer in lock data (285)
6625 - partial support for multiple MDS on a single host (241)
6626 - data locks weren't cancelled at clear_inode time (290, 311)
6627 - intent locks could lead to unbounded lock growth (205)
6628 - added a maximum lock count, an LRU list, and a flusher
6629 - fix multiple rename (365)
6630 - properly abstracted the echo client
6631 - OSC locked 1 byte too many; fixed
6632 - rewrote brw callback code:
6633 - fixed recovery bugs related to LOVs (306)
6634 - fixed too-many-pages-in-one-write crash (191)
6635 - fixed (again) crash in sync_io_timeout (214)
6636 - probably fixed callback-related race (385)
6638 - Add capability to MDS protocol
6639 - LDLM cancellations and callbacks on different portals
6641 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6644 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6645 - defer open object destruction to close time (601981, 138)
6646 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6647 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6648 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6649 - don't call obd_setattr in ll_file_release for destroyed objects
6650 * protocol change to lustre_msg: move |version| and add |flags|
6651 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6652 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6653 * added replay of create, unlink, link and rename operations during
6654 MDS failover; recovery should be much more robust now
6655 * remove failed OSCs from LOVs (only lov_create uses this so far)
6656 * the lustre-HOWTO was brought (more) up to date (582544)
6658 2002-10-23 Phil Schwan <phil@clusterfs.com>
6661 - in-use dentries weren't being reused properly (617851)
6662 - prevent multiple LDLM setup (599178)
6663 - fix LOV size calculations for truncate (617853)
6664 - fix client handling of MDS intent errors (POSIX)
6665 - fix permission bug in lovstripe.c test (624321)
6666 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6667 - truncate past end of file could corrupt data
6668 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6669 - a race in recovery could return ETIMEDOUT to apps (623947)
6670 - building outside the source directory was fixed
6671 * the lustre-HOWTO was brought (more) up to date (582544)
6672 * major progress was made on recovery functionality
6674 2002-10-10 Phil Schwan <phil@clusterfs.com>
6677 - recovery deadlock fix
6678 - rm -rf causes LBUG fix (617817)
6679 - file open by multiple tasks fix (618962)
6680 - directory permissions bugs (602707 and 620007)
6681 - journal_stop fixed with locking (611313)
6682 - O_APPEND failures resolved (618273, perhaps 614459)
6683 - lconf PATH fix (619770)
6684 - IA64 build fix (621450)
6685 - RPC buffer sizes scale with amount of memory
6687 2002-10-01 Phil Schwan <phil@clusterfs.com>
6690 - locks would be cancelled without throwing away data pages,
6691 resulting in inconsistent data (605627)
6692 - inode attributes were not always being refreshed (605627, 612449)
6693 - lconf now continues to cleanup after lctl reports an error
6694 - MDS now enforces user permissions (602707)
6695 - lprocfs cleanup fixed, but not yet enabled (614157)
6696 - fixed infinite server hang, should a client not respond to an AST
6697 - avoid going into recovery if user calls readlink() with a buffer
6698 that's too small (613941)
6699 - AST RPCs no longer require replies (614867) -- this may be changed
6700 - don't crash server if client sends an IOV that's too big (611336)
6701 - fixed lock conversion deadlock (611892)
6702 - fixed the following of symlinks (614622)
6703 * recovery: the server can remove locks from a client that dies, other
6704 clients can make progress
6705 * more extN patch fixes
6706 * compile-time configurable ptlrpc buffer allocations
6708 - collaborative read cache document
6709 - Lustre Lite Performance CDR document-in-progress
6711 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6714 - fix typo in patch-2.4.18
6716 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6719 - clear ptlrpc request each time in handle_incoming_request()
6720 - unlink of files now destroys the object on the OST
6722 2002-09-19 Peter Braam <braam@clusterfs.com>
6724 * add hard link support
6725 * change obdfile creation method
6726 * kernel patch changed
6728 2002-09-19 Peter Braam <braam@clusterfs.com>
6731 - stack overflow bug in extN fixed
6733 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6735 * documentation updates
6736 - add man pages for config tools
6737 - update tests/README to describe testing with new config tools
6738 - finish metadata API descriptions
6739 * bug fixes and cleanups
6740 - statfs workaround for 16TB limit
6741 - LOV stripe allocation improved, can stripe on subset of OSTs
6742 - LOV file size/IO offset was wrong for files > 4GB in size
6743 - object EA data was being dropped, caused files to be unreadable
6744 - memory overflow with non-LOV OST caused memory corruption
6745 - fixed regression tests to work with new config tools, obdfilter
6746 - fixed bug when directory size became larger than 1 block
6747 - fixed bug (for single client case) when PWD was deleted
6748 - invalidate local directory pages when doing intent-based ops
6749 - avoid LDLM oops when lock callback contained bad data
6751 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6753 * documentation updates
6754 * bug fixes and cleanups
6755 - configuration tools
6758 - 64-bit compile warnings
6759 - 64-bit internal statfs data
6761 * test_brw on persistent OST devices
6763 * lprocfs (disabled)
6765 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6767 * documentation updates
6768 * bug fixes and cleanups
6769 * configuration tools
6771 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6774 * many small fixes to 0_5_4
6775 * io/network handling
6776 * thinkos in MDS operations
6778 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6781 * crucial basic fixes to 0.5.3
6782 * IOR, Iozone work over Elan
6785 2002-08-07 Phil Schwan <phil@clusterfs.com>
6786 * version 0_5_3, our first alpha
6787 * we use the new Portals iovs
6788 * documentation updates
6789 * bug fixes and cleanups
6790 * small changes in the DLM wire protocol
6792 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6793 * version 0_5_1 with some initial stability,
6794 * locking on MD and file I/O.
6795 * documentation updates
6796 * several bug fixes since 0.5.0
6797 * small changes in wire protocol
6799 2002-07-18 Phil Schwan <phil@clusterfs.com>
6801 * delivered as Lustre Light Alpha
6802 * fixed a crash after handling invalid MDS requests
6803 * fixed directory pages for architectures with non-4k pages sizes
6805 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6806 * release version v0_4_4
6807 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6809 * Fixes a number of interruption problems with OST operations.
6810 * Update documentation for portals header changes
6811 * Move all wire protocol structs/defines to lustre_idl.h
6812 * Fixes symlink length bug.
6813 * Add tcpdump to repository.
6815 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6816 * release version v0_4_3
6817 * Fixes statfs for inodes on extN.
6818 * Fixes bug in runtests which would delete /etc/hosts.
6819 * Use 64-bit object IDs wherever possible (not into VFS though)
6820 Remove ost_get_info, which is unused by lustre, and out of date.
6822 2002-07-03 Peter Braam <braam@clusterfs.com>
6823 * release version v0_4_2 Fixes a lookup error (type not passed)
6824 * move forward to head of Portals
6825 * move forward to latest Lustre kernel
6827 2002-06-25 Peter Braam <braam@clusterfs.com>
6828 * release version v0_4_1. Hopefully stable on single node use.