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 3.10.0-1160.42.2.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 4.18.0-305.19.1.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 3.10.0-1160.42.2.el7 (RHEL7.9)
30 4.12.14-95.48 (SLES12 SP4)
31 4.12.14-122.83 (SLES12 SP5)
32 4.12.14-197.75 (SLES15 SP1)
33 5.3.18-24.61 (SLES15 SP2)
34 5.3.18-59.19 (SLES15 SP3)
35 4.15.0-48 (Ubuntu 18.04)
36 * Other clients known to build on these kernels at some point (others may also work):
37 3.10.0-862.14.4.el7 (RHEL7.5)
38 3.10.0-957.27.2.el7 (RHEL7.6)
39 4.14.0-49.13.1.el7a (RHEL7.5)
40 4.14.0-115.2.2.el7a (RHEL7.6)
41 4.18.0-80.11.2.el8 (RHEL8.0)
42 4.18.0-147.8.1.el8 (RHEL8.1)
43 4.18.0-193.28.1.el8 (RHEL8.2)
44 4.18.0-240.22.1.el8 (RHEL8.3)
45 4.18.0-305.19.1.el8 (RHEL8.4)
46 4.4.120-92.70 (SLES12 SP2)
47 4.4.180-94.100 (SLES12 SP3)
48 4.4.0-131 (Ubuntu 16.04)
49 5.4.0-37 (Ubuntu 20.04)
50 5.8.0-53 (Ubuntu 20.04.2 HWE)
51 5.11.0-31 (Ubuntu 20.04.3 HWE)
52 5.11.0 (vanilla kernel.org)
53 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
54 * Recommended ZFS version: 2.0.0
55 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
56 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
57 stack overflow. For more information, please refer to bugzilla 17630.
58 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
59 the re-exporting nodes to be mounted with "32bitapi" mount option
61 --------------------------------------------------------------------------------
65 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
66 for currently supported client and server kernel versions.
67 * Server primary kernels built and tested during release cycle:
68 3.10.0-1062.18.1.el7 (RHEL7.7)
69 3.10.0-1127.19.1.el7 (RHEL7.8)
70 3.10.0-1160.6.1.el7 (RHEL7.9)
71 * Other server kernels known to build and work at some point (others may also work):
72 3.10.0-862.14.4.el7 (RHEL7.5)
73 3.10.0-957.27.2.el7 (RHEL7.6)
74 4.18.0-80.11.2.el8 (RHEL8.0)
75 4.18.0-147.8.1.el8 (RHEL8.1)
76 4.18.0-193.28.1.el8 (RHEL8.2)
77 4.18.0-240.1.1.el8 (RHEL8.3)
78 4.4.120-92.70 (SLES12 SP2)
79 4.4.180-94.100 (SLES12 SP3)
80 4.4.0-131 (Ubuntu 16.04)
81 4.15.0-32 (Ubuntu 18.04)
82 5.4.0-48 (Ubuntu 20.04)
83 vanilla linux 5.4.0 (ZFS + ldiskfs)
84 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
85 * Client primary kernels built and tested during release cycle:
86 3.10.0-1062.18.1.el7 (RHEL7.7)
87 3.10.0-1127.19.1.el7 (RHEL7.8)
88 3.10.0-1160.6.1.el7 (RHEL7.9)
89 4.12.14-95.48 (SLES12 SP4)
90 4.12.14-122.20 (SLES12 SP5)
91 4.12.14-197.75 (SLES15 SP1)
92 5.3.18-24.24 (SLES15 SP2)
93 4.15.0-48 (Ubuntu 18.04)
94 * Other clients known to build on these kernels at some point (others may also work):
95 3.10.0-862.14.4.el7 (RHEL7.5)
96 3.10.0-957.27.2.el7 (RHEL7.6)
97 4.14.0-49.13.1.el7a (RHEL7.5)
98 4.14.0-115.2.2.el7a (RHEL7.6)
99 4.18.0-80.11.2.el8 (RHEL8.0)
100 4.18.0-147.8.1.el8 (RHEL8.1)
101 4.18.0-193.28.1.el8 (RHEL8.2)
102 4.18.0-240.1.1.el8 (RHEL8.3)
103 4.4.120-92.70 (SLES12 SP2)
104 4.4.180-94.100 (SLES12 SP3)
105 4.4.0-131 (Ubuntu 16.04)
106 5.4.0-37 (Ubuntu 20.04)
107 5.4.0 (vanilla kernel.org)
108 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
109 * Recommended ZFS version: 2.0.0
110 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
111 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
112 stack overflow. For more information, please refer to bugzilla 17630.
113 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
114 the re-exporting nodes to be mounted with "32bitapi" mount option
116 --------------------------------------------------------------------------------
120 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
121 for currently supported client and server kernel versions.
122 * Server primary kernels built and tested during release cycle:
123 3.10.0-1062.1.1.el7 (RHEL7.7)
124 * Other server kernels known to build and work at some point (others may also work):
125 3.10.0-862.14.4.el7 (RHEL7.5)
126 3.10.0-957.27.2.el7 (RHEL7.6)
127 4.18.0-80.7.1.el8 (RHEL8.0)
128 4.4.120-92.70 (SLES12 SP2)
129 4.4.180-94.100 (SLES12 SP3)
130 4.4.0-131 (Ubuntu 16.04)
131 4.15.0-32 (Ubuntu 18.04)
132 vanilla linux 4.6.7 (ZFS only)
133 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
134 * Client primary unpatched kernels built and tested during release cycle:
135 3.10.0-1062.1.1.el7 (RHEL7.7)
136 4.12.14-95.29 (SLES12 SP4)
137 4.15.0-48 (Ubuntu 18.04)
138 * Other clients known to build on unpatched kernels at some point (others may also work):
139 3.10.0-862.14.4.el7 (RHEL7.5)
140 3.10.0-957.27.2.el7 (RHEL7.6)
141 4.14.0-49.13.1.el7a (RHEL7.5)
142 4.14.0-115.2.2.el7a (RHEL7.6)
143 4.18.0-80.7.1.el8 (RHEL8.0)
144 4.4.120-92.70 (SLES12 SP2)
145 4.4.180-94.100 (SLES12 SP3)
146 4.4.0-131 (Ubuntu 16.04)
147 vanilla linux 4.15.0 (ZFS only)
148 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
149 * Recommended ZFS / SPL version: 0.7.13
150 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
151 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
152 stack overflow. For more information, please refer to bugzilla 17630.
153 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
154 the re-exporting nodes to be mounted with "32bitapi" mount option
155 * Note for Ubuntu 14.04.5:
156 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
157 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
159 --------------------------------------------------------------------------------
163 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
164 for currently supported client and server kernel versions.
165 * Server known to build on the following kernels (others may also work):
166 2.6.32-696.18.7.el6 (RHEL6.9)
167 3.10.0-862.14.4.el7 (RHEL7.5)
168 3.10.0-957.el7 (RHEL7.6)
169 4.4.120-92.70 (SLES12 SP2)
170 4.4.162-94.69 (SLES12 SP3)
171 4.4.0-131 (Ubuntu 16.04)
172 4.15.0-32 (Ubuntu 18.04)
173 vanilla linux 4.6.7 (ZFS only)
174 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
175 * Client known to build on unpatched kernels (others may also work):
176 2.6.32-696.18.7.el6 (RHEL6.9)
177 2.6.32-754.6.3.el6 (RHEL6.10)
178 3.10.0-862.14.4.el7 (RHEL7.5)
179 3.10.0-957.el7 (RHEL7.6)
180 4.14.0-49.13.1.el7a (RHEL7.5)
181 4.14.0-115.2.2.el7a (RHEL7.6)
182 4.4.120-92.70 (SLES12 SP2)
183 4.4.162-94.69 (SLES12 SP3)
184 4.4.0-131 (Ubuntu 16.04)
185 4.15.0-32 (Ubuntu 18.04)
186 vanilla linux 4.15.0 (ZFS only)
187 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
188 * Recommended ZFS / SPL version: 0.7.9
189 * Tested with ZFS / SPL version: 0.7.9
190 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
191 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
192 stack overflow. For more information, please refer to bugzilla 17630.
193 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
194 the re-exporting nodes to be mounted with "32bitapi" mount option
195 * Note for Ubuntu 14.04.5:
196 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
197 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
199 --------------------------------------------------------------------------------
201 2018-04-03 Intel Corporation
203 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
204 for currently supported client and server kernel versions.
205 * Server known to build on patched kernels:
206 2.6.32-431.29.2.el6 (RHEL6.5)
207 2.6.32-504.30.3.el6 (RHEL6.6)
208 2.6.32-573.26.1.el6 (RHEL6.7)
209 2.6.32-642.15.1.el6 (RHEL6.8)
210 2.6.32-696.18.7.el6 (RHEL6.9)
211 3.10.0-693.21.1.el7 (RHEL7.4)
212 3.0.101-0.47.71 (SLES11 SP3)
213 3.0.101-107 (SLES11 SP4)
214 3.12.74-60.64.40 (SLES12 SP1)
215 4.4.114-92.64 (SLES12 SP2)
216 4.4.114-94.11 (SLES12 SP3)
217 3.13.0-101 (Ubuntu 14.04, ZFS only)
218 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
219 4.4.0-116 (Ubuntu 16.04)
220 vanilla linux 4.6.7 (ZFS only)
221 * Client known to build on unpatched kernels:
222 2.6.32-431.29.2.el6 (RHEL6.5)
223 2.6.32-504.30.3.el6 (RHEL6.6)
224 2.6.32-573.26.1.el6 (RHEL6.7)
225 2.6.32-642.15.1.el6 (RHEL6.8)
226 2.6.32-696.18.7.el6 (RHEL6.9)
227 3.10.0-693.21.1.el7 (RHEL7.4)
228 3.0.101-0.47.71 (SLES11 SP3)
229 3.0.101-107 (SLES11 SP4)
230 3.12.74-60.64.40 (SLES12 SP1)
231 4.4.114-92.64 (SLES12 SP2)
232 4.4.114-94.11 (SLES12 SP3)
233 3.13.0-101 (Ubuntu 14.04)
234 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
235 4.4.0-116 (Ubuntu 16.04)
237 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
238 * Recommended ZFS / SPL version: 0.7.6
239 * Tested with ZFS / SPL version: 0.7.6
240 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
241 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
242 stack overflow. For more information, please refer to bugzilla 17630.
243 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
244 the re-exporting nodes to be mounted with "32bitapi" mount option
245 * Note for Ubuntu 14.04.5:
246 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
247 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
249 --------------------------------------------------------------------------------
251 2017-07-05 Intel Corporation
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on patched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.26.1.el6 (RHEL6.7)
259 2.6.32-642.15.1.el6 (RHEL6.8)
260 2.6.32-696.3.1.el6 (RHEL6.9)
261 3.10.0-514.21.1.el7 (RHEL7.3)
262 3.0.101-0.47.71 (SLES11 SP3)
263 3.0.101-100 (SLES11 SP4)
264 3.12.74-60.64.40 (SLES12 SP1)
265 4.4.59-92.17 (SLES12 SP2)
266 vanilla linux 4.6.7 (ZFS only)
267 * Client known to build on unpatched kernels:
268 2.6.32-431.29.2.el6 (RHEL6.5)
269 2.6.32-504.30.3.el6 (RHEL6.6)
270 2.6.32-573.26.1.el6 (RHEL6.7)
271 2.6.32-642.15.1.el6 (RHEL6.8)
272 2.6.32-696.3.1.el6 (RHEL6.9)
273 3.10.0-514.21.1.el7 (RHEL7.3)
274 3.0.101-0.47.71 (SLES11 SP3)
275 3.0.101-100 (SLES11 SP4)
276 3.12.74-60.64.40 (SLES12 SP1)
277 4.4.59-92.17 (SLES12 SP2)
279 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
280 * Recommended ZFS / SPL version: 0.6.5.9
281 * Tested with ZFS / SPL version: 0.6.5.9
282 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
283 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
284 stack overflow. For more information, please refer to bugzilla 17630.
285 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
286 the re-exporting nodes to be mounted with "32bitapi" mount option
288 --------------------------------------------------------------------------------
290 12-07-2016 Intel Corporation
292 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
293 for currently supported client and server kernel versions.
294 * Server known to build on patched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.30.3.el6 (RHEL6.6)
297 2.6.32-573.26.1.el6 (RHEL6.7)
298 2.6.32-642.11.1.el6 (RHEL6.8)
299 3.10.0-514.el7 (RHEL7.3)
300 3.0.101-0.47.71 (SLES11 SP3)
301 3.0.101-80 (SLES11 SP4)
302 3.12.62-60.64.8 (SLES12 SP1)
303 vanilla linux 4.5.7 (ZFS only)
304 * Client known to build on unpatched kernels:
305 2.6.32-431.29.2.el6 (RHEL6.5)
306 2.6.32-504.30.3.el6 (RHEL6.6)
307 2.6.32-573.26.1.el6 (RHEL6.7)
308 2.6.32-642.11.1.el6 (RHEL6.8)
309 3.10.0-514.el7 (RHEL7.3)
310 3.0.101-0.47.71 (SLES11 SP3)
311 3.0.101-80 (SLES11 SP4)
312 3.12.62-60.64.8 (SLES12 SP1)
314 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
315 * Recommended ZFS / SPL version: 0.6.5.8
316 * Tested with ZFS / SPL version: 0.6.5.8
317 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
318 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
319 stack overflow. For more information, please refer to bugzilla 17630.
320 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
321 the re-exporting nodes to be mounted with "32bitapi" mount option
323 --------------------------------------------------------------------------------
325 02-29-2016 Intel Corporation
327 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
328 for currently supported client and server kernel versions.
329 * Server known to build on patched kernels:
330 2.6.32-431.29.2.el6 (RHEL6.5)
331 2.6.32-504.30.3.el6 (RHEL6.6)
332 2.6.32-573.12.1.el6 (RHEL6.7)
333 3.10.0-327.3.1.el7 (RHEL7.2)
334 3.0.101-0.47.71 (SLES11 SP3)
335 3.0.101-68 (SLES11 SP4)
336 vanilla linux 4.2.1 (ZFS only)
337 * Client known to build on unpatched kernels:
338 2.6.32-431.29.2.el6 (RHEL6.5)
339 2.6.32-504.30.3.el6 (RHEL6.6)
340 2.6.32-573.12.1.el6 (RHEL6.7)
341 3.10.0-327.3.1.el7 (RHEL7.2)
342 3.0.101-0.47.71 (SLES11 SP3)
343 3.0.101-68 (SLES11 SP4)
346 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
347 * Recommended ZFS / SPL version: 0.6.4.2
348 * Tested with ZFS / SPL version: 0.6.4.2
349 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
350 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
351 stack overflow. For more information, please refer to bugzilla 17630.
352 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
353 the re-exporting nodes to be mounted with "32bitapi" mount option
355 --------------------------------------------------------------------------------
357 03-10-2015 Intel Corporation
359 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
360 for currently supported client and server kernel versions.
361 * Server known to build on patched kernels:
362 2.6.32-431.29.2.el6 (RHEL6.5)
363 2.6.32-504.8.1.el6 (RHEL6.6)
364 3.0.101-0.46 (SLES11 SP3)
365 * Client known to build on unpatched kernels:
366 2.6.32-431.29.2.el6 (RHEL6.5)
367 2.6.32-504.8.1.el6 (RHEL6.6)
368 3.10.0-123.20.1.el7 (RHEL7)
369 3.0.101-0.46 (SLES11 SP3)
370 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
371 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
372 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
373 stack overflow. For more information, please refer to bugzilla 17630.
374 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
375 the re-exporting nodes to be mounted with "32bitapi" mount option
377 Severity : enhancement
379 Description: control OST-index in IDIF via ROCOMPAT flag.
380 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
381 last_rcvd file. For new formatted OST device, it will be auto set;
382 for the case of upgrading from old OST device, you can enable it
383 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
384 enabled, for new created OST-object, its IDIF-in-LMA will contain
385 the OST-index; for the existing OST-object, the OSD will convert
386 old format IDIF as new format IDIF with OST-index stored in the
387 LMA EA when accessing such OST-object or via OI scrub. Once such
388 flag is enabled, it cannot be reverted back, so the system cannot
389 be downgraded to the orignal incompatible version.
391 --------------------------------------------------------------------------------
393 07-30-2014 Intel Corporation
395 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
396 for currently supported client and server kernel versions.
397 * Server known to build on patched kernels:
398 2.6.32-431.20.3.el6 (RHEL6.5)
399 3.0.101-0.31 (SLES11 SP3)
400 * Client known to build on unpatched kernels:
401 2.6.32-431.20.3.el6 (RHEL6.5)
402 3.0.101-0.31 (SLES11 SP3)
405 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
406 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
407 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
408 stack overflow. For more information, please refer to bugzilla 17630.
409 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
410 the re-exporting nodes to be mounted with "32bitapi" mount option
412 --------------------------------------------------------------------------------
414 10-31-2013 Intel Corporation
416 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
417 * Server builds for kernels:
418 2.6.32-358.18.1.el6 (RHEL6.4)
419 * Client builds for unpatched kernels:
420 2.6.32-358.18.1.el6 (RHEL6.4)
421 3.0.92-0.8 (SLES11 SP3)
423 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
424 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
425 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
426 stack overflow. For more information, please refer to bugzilla 17630.
427 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
428 the re-exporting nodes to be mounted with "32bitapi" mount option
430 --------------------------------------------------------------------------------
432 05-30-2013 Intel Corporation
434 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
435 * Server builds for kernels:
436 2.6.32-358.6.2.el6 (RHEL6.4)
437 * Client builds for unpatched kernels:
438 2.6.32-358.6.2.el6 (RHEL6.4)
439 3.0.74-0.6.6 (SLES11 SP2)
441 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
442 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
443 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
444 stack overflow. For more information, please refer to bugzilla 17630.
445 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
446 the re-exporting nodes to be mounted with "32bitapi" mount option
448 --------------------------------------------------------------------------------
450 9-30-2012 Whamcloud, Inc.
452 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
453 * Server support for kernels:
454 2.6.32-279.5.1.el6 (RHEL6)
455 * Client support for unpatched kernels:
456 2.6.32-279.5.1.el6 (RHEL6)
458 2.6.18-238.19.1.el5 (RHEL5)
459 2.6.16.60-0.69.1 (SLES10)
460 * Recommended e2fsprogs version: 1.41.90.wc4
461 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
462 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
463 stack overflow. For more information, please refer to bugzilla 17630.
464 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
465 the re-exporting nodes to be mounted with "32bitapi" mount option
467 Severity : enhancement
469 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
471 Severity : enhancement
473 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
475 Severity : enhancement
477 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
479 Severity : enhancement
481 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
483 Severity : enhancement
485 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
487 --------------------------------------------------------------------------------
489 04-30-2012 Whamcloud, Inc.
491 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
492 * Server support for kernels:
493 2.6.32-220.4.2.el6 (RHEL6)
494 * Client support for unpatched kernels:
495 2.6.32-220.4.2.el6 (RHEL6)
497 2.6.18-238.19.1.el5 (RHEL5)
498 2.6.16.60-0.69.1 (SLES10)
499 * Recommended e2fsprogs version: 1.41.90.wc4
500 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
501 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
502 stack overflow. For more information, please refer to bugzilla 17630.
503 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
504 the re-exporting nodes to be mounted with "32bitapi" mount option
506 Severity : enhancement
508 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
510 Severity : enhancement
512 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
514 Severity : enhancement
516 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
518 Severity : enhancement
520 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
522 --------------------------------------------------------------------------------
524 09-30-2011 Whamcloud, Inc.
526 * Server support for kernels:
527 2.6.18-238.19.1.el5 (RHEL5)
528 2.6.32-131.6.1.el6 (RHEL6)
529 * Client support for unpatched kernels:
530 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
531 2.6.18-238.19.1.el5 (RHEL5)
532 2.6.32-131.6.1.el6 (RHEL6)
533 2.6.29.4-167.fc11 (FC11)
534 2.6.16.60-0.69.1 (SLES10)
536 * Recommended e2fsprogs version: 1.41.90.wc3
537 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
538 Lustre filesystem with 4K stack may cause a stack overflow. For more
539 information, please refer to bugzilla 17630.
540 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
541 mounted with "32bitapi" mount option
543 Severity : enhancement
544 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
546 Severity : enhancement
547 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
549 Severity : enhancement
550 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
553 Frequency : only with SLES10
555 Description: Use OFED "KMP" provided by Novell
556 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
557 Lustre is now built against this package. That means you need to
558 install the ofed-kmp package from Novell for the patchless client.
559 Note that the ofed-kmp that Novell ships may not exactly match the
560 kernel version but should still be compatible.
562 Severity : enhancement
564 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
565 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
567 --------------------------------------------------------------------------------
569 2010-07-15 Oracle, Inc.
571 * Server support for kernels:
572 2.6.18-164.11.1.el5 (RHEL 5)
573 2.6.18-164.11.1.0.1.el5 (OEL 5)
574 * Client support for unpatched kernels:
575 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
576 2.6.18-164.11.1.el5 (RHEL 5),
577 2.6.18-164.11.1.0.1.el5 (OEL 5)
578 2.6.16.60-0.42.8 (SLES 10),
580 2.6.29.4-167.fc11 (FC11)
581 * Recommended e2fsprogs version: 1.41.10-sun2
583 Severity : enhancement
585 Description: Distribute lfs_migrate script with lustre
586 Details : The lfs_migrate mentioned in the manual is now included in the
587 lustre tree in lustre/scripts and will distributed in lustre RPMs.
589 Severity : enhancement
591 Description: Failover nid list added to import proc file.
593 Severity : enhancement
595 Description: "som_preview" mount option required to use SOM
596 Details : The clients will need to use som_preview mount option
600 Description: Lustre does not do 1MB IOs to HW RAID
601 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
603 Severity : enhancement
605 Description: lfs quota output is not very convenient for awk/sed-parsing
606 Details : Some positions in lfs quota output table could be empty or
607 non-empty which made it hard to parse it with scripts, now
608 a dash is put instead of space where there is not supposed
613 Description: Update to OFED 1.5.1
614 Details : OFED 1.5.1 will be used in environments that don't use
615 vendor supplied OFED.
619 Description: Remove lnet. prefix from lctl params display
620 Details : Also renamed lnet.memused to lnet_memused to
625 Description: Use conf_param -d to remove permanant settings.
629 Description: Add list_param -R
630 Details : lctl list_param now supports recursive mode.
633 Bugzilla : 21829, 22312
634 Description: Fix llobdstat.
635 Details : 1. Now updating the write_bytes counter
636 2. Port script fixes from 1.8.
640 Description: decrease the usage of memory on clients.
641 Details : 1. On clients, recycle dentries and inodes unused.
642 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
647 Description: llobdstat fix and enhancement
648 Details : add a counter to set a limit to how many samples will be returned
649 fix a wildcard in the path to limit to obdfilter stats only
653 Description: lfs check only allowed for root
654 Details : Code cleanups and added a regresion test of non-root use of
659 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
660 Details : Fix hr->hr_index overload.
664 Description: Fix FIEMAP functionality.
668 Description: FC11 patchless client support added
672 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
673 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
677 Description: New acc-sm script and various test-framework enhancments.
681 Description: Previously landed patch was reverted.
685 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
689 Description: Add ldiskfs maxdirsize mount option.
693 Description: Fix mkfs.lustre support for 16TB LUNs.
697 Description: Print mmp_check_interval
698 Details : print mmp_check_interval and make it possible to abort
699 mount operation in case it takes too long.
703 Description: Add cascading_rw to lustre mpi tests.
707 Description: Store SOM data in EA.
708 Details : Also several SOM related cleanups.
712 Description: Add refcount to md_open_data
716 Description: Use refcount on lov_request_set to prevent memory corruption.
720 Description: Statahead: use parent dir's i_mutex for synchronization.
724 Description: loadgen improvements.
728 Description: Enable peer health detection in ksocklnd and ko2iblnd.
732 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
736 Description: Fix race on obd_devcie:md_stats
740 Description: Fix SLES 11 client builds.
741 Details : Resolved uninitalized pointer gcc warnings.
745 Description: Allow non-root access for "lfs check".
746 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
750 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
752 Severity : enhancement
754 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
755 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
759 Description: Sanity test_27p fails.
760 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
761 Sanity tests 27 were cleaned up and improved.
765 Description: Unify req format on client/servers
766 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
767 buffers and swabbers approach..
772 Description: drop join file stuff.
774 Severity : enhancement
776 Description: Add 2.6.27 and SLES11 (patchless client) support
780 Description: Fix in ptlrpc_expire_one_request() to print the signed time
783 Severity : enhancement
785 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
788 Severity : enhancement
790 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
792 Severity : enhancement
794 Description: correctly handle big reply message.
795 Details : send LNet event if reply is bigger than buffer and adjust this buffer
800 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
802 Severity : enhancement
804 Description: Add support for OEL5.
808 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
810 Severity : enhancement
812 Description: "weak-modules" support
813 Details : Implement "weak-modules" support which enables kernel modules
814 to be used with any kernel that implements the same kABI. In
815 order to achieve this modules are now installed in
816 /lib/modules/$(uname -r)/updates/kernel on all distributions.
820 Description: client could not reconnect to OST because of an active request.
821 Details : The client now retries to reconnect to the same server, if a
822 connect request failed with EBUSY or -EAGAIN. Test case 35b
823 conf-sanity.sh has been created.
827 Description: Don't allow make backward step on assignin osc next id.
828 Details : race between allocation next id and ll_sync thread can be cause
829 of set wrong osc next id and can be kill valid ost objects.
833 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
837 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
841 Description: Error handling in osc_statfs_interpret() has been improved.
842 Details : Check in osc_statfs_interpret() for EBADR.
846 Description: Do not send statfs() requests to OSTs disabled by administrator.
847 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
849 Severity : enhancement
851 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
853 Severity : enhancement
855 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
857 Severity : enhancement
859 Description: Enhancement for lfs(1) command to use numeric uid/gid.
863 Description: Improved cleanup procedure for conf-sanity test_46a.
867 Description: Do not start bulk transfer, if client tried to reconnect.
868 Details : Do not start bulk transfer, if client tried to reconnect,
869 but failed due to active rpcs. Check for exp_abort_active_req
870 has been implemented.
874 Description: Prevent inconsistences between linux and lustre mount structures.
875 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
876 Make the sleep interruptible.
878 Severity : enhancement
880 Description: provide server to client comms path
881 Details : server to client communications path via new LDLM_SET_INFO rpc
885 Description: truncate starts GFP_FS allocation under transaction causing deadlock
886 Details : ldiskfs_truncate calls grab_cache_page which may start page
887 allocation under an open transaction. This may lead to
888 calling prune_icache with consequent lustre reentrance.
892 Description: send correctly lsm on open replay
893 Details : MDS is trust to LSM size on replay open, but client can set wrong size
898 Description: enable client interface failover
899 Details : When a child reconnects from another NID, properly update export
900 nid hash position and ldlm reverse import.
902 Severity : enhancement
904 Description: Communicate OST degraded/readonly state via statfs to MDS
905 Details : Flags in the statfs returned from OSTs indicate whether the
906 OST is in a degraded RAID state, or if the filesystem has
907 turned read-only after a filesystem error is detected.
912 Description: don't panic if EPROTO was hit when reading symlink
913 Details : correctly handling request reference in error cases.
915 Severity : enhancement
917 Description: Add LustreNetLink, a kernel-userspace communcation path.
919 Severity : enhancement
921 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
926 Description: access to llog context before init.
927 Details : move handling CATALOGS file at osc layer and forbid access to llog
931 Frequency : rare, only when client is writing to OST on local node
933 Description: Avoid deadlock for local client writes
934 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
935 memory freeing context. This allows OST threads to set the
936 PF_MEMALLOC flag on task structures in order to allocate memory
937 from reserved pools and complete IO.
938 Use GFP_HIGHUSER for OST allocations for non-local client writes,
939 so that the OST threads generate memory pressure and allow
940 inactive pages to be reclaimed.
942 Severity : enhancement
944 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
949 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
950 Details : move ldlm namespace creation in setup phase to avoid grab
951 _lprocfs_lock with cli_sem held.
955 Description: Temporarily disable grant shrink.
956 Details : Disable the feature for debugging.
959 Frequency : only when formatting test filesystems
961 Description: Unable to run several mkfs.lustre on loop devices at the same time
962 Details : mkfs.lustre returns error 256 on the concurrent loop devices
963 formatting. The solution is to proper handle the error.
965 Severity : enhancement
967 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
969 Severity : enhancement
971 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
976 Description: resolve race between obd_disconnect and class_disconnect_exports
977 Details : if obd_disconnect will be called to already disconnected export he
978 forget release one reference and osc module can't unloaded.
980 Severity : enhancement
982 Description: Allow tuning service thread via /proc
983 Details : For each service a new
984 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
985 created that can be used to set min/max thread counts, and get the
986 current number of running threads.
991 Descriptoin: don't return error if have partially created objects for file.
992 Details : lov_update_create_set uses set->set_success as index for created
993 objects, so if some requests will be failed, they will have hole
994 at end of array and qos_shrink_lsm for allocate correct lsm.
996 Severity : enhancement
998 Description: Update OFED support to 1.4.1
1003 Description: Reduce small size read RPC
1004 Details : Set read-ahead limit for every file and only do read-ahead when
1005 available read-ahead pages are bigger than 1M to avoid small size
1008 Severity : enhancement
1010 Description: move AT tunable parameters for more consistent usage
1011 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1013 Severity : enhancement
1015 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1016 Details : allow skip disconnected ost for send statfs request and hide error
1022 Description: fix racy locking of mballoc block bitmaps causing BUG
1023 Details : The locking of the mballoc buddy bitmap and the in-memory
1024 block bitmap was using two different spin locks in some
1025 cases. This made it possible to incorrectly access the
1026 mballoc bitmap while another process was modifying it,
1027 causing a sanity assertion to fail. While no on-disk corruption
1028 was reported, there was some risk of this happening.
1031 Frequency : rare, on llog test 6
1033 Descriptoin: don't allow connect to already connected import
1034 Details : allowing connect to already connected import is hide connecting problem.
1039 Descriptoin: allow kill process which wait statahead result
1040 Details : for some reasons 'ls' can stick in waiting result from statahead,
1041 in this case need way for kill this process.
1043 Severity : enhancement
1045 Description: Add state history info file, enhance import info file
1046 Details : Track import connection state changes in a new osc/mdc proc file;
1047 add overview-type data to the osc/mdc import proc file.
1049 Severity : enhancement
1051 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1053 Severity : enhancement
1055 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1060 Descriptoin: OST grant too much space to client even there are not enough space.
1061 Details : Client will shrink its grant cache to OST if there are no write
1062 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1063 this grant cache if there are already not enough avaible space
1064 (left_space < total_clients * 32M).
1067 Frequency : start MDS on uncleanly shutdowned MDS device
1069 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1070 Details : stay in waiting mds<>ost recovery finished produce random bugs
1071 due race between two ll_sync thread for one lov target. send
1072 ACTIVATE event only if connect realy finished and import have
1076 Frequency : rare, connect and disconnect target at same time
1078 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1079 Details : don't call obd_disconnect under lov_lock. this long time
1080 operation and can block ptlrpcd which answer to connect request.
1085 Descriptoin: don't lose wakeup for imp_recovery_waitq
1086 Details : recover_import_no_retry or invalidate_import and import_close can
1087 both sleep on imp_recovery_waitq, but we was send only one wakeup
1091 Frequency : always with long access acl
1093 Descriptoin: mds can't pack reply with long acl.
1094 Details : mds don't control size of acl but they limited by reint/getattr
1097 Severity : enhancement
1099 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1101 Severity : enhancement
1103 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1106 Frequency : start MDS on uncleanly shutdowned MDS device
1108 Descriptoin: aborting recovery hang on MDS
1109 Details : don't throttle destroy RPCs for the MDT.
1112 Frequency : on remount
1114 Description: external journal device not working after the remount
1115 Details : clear dev_rdonly flag for external journal devices in
1121 Description: shutdown vs evict race
1122 Details : client_disconnect_export vs connect request race.
1123 if client will evicted at this time - we start invalidate
1124 thread without referece to import and import can be freed
1128 Frequency : rare, need acl's on inode.
1130 Description: client can't handle ost additional correctly
1131 Details : if ost was added after client connected to mds client can have
1132 hit lnet_try_match_md ... to big messages to wide striped files.
1133 in this case need teach client to handle config events about add
1134 lov target and update client max ea size at that event.
1136 Severity : enhancement
1138 Description: Changelogs
1139 Details : Changelogs are a lightweight mechanism to track filesystem
1140 metadata and namespace changes. The changelog is recorded
1141 permanently on the MDTs, and is periodically "consumed" / purged
1142 when records are no longer needed.
1144 Severity : enhancement
1146 Description: compact fld format with extents
1147 Details : Store range of seq rather than every seq in FLD. Seq
1148 controller update FLD rather than clients. In Case of CMD, mdt0
1149 has FLD, all other metadata server act as non persistent proxy
1150 for FLD queries and cache fld entries in fld cache.
1155 Description: don't skip ost target if they assigned to file
1156 Details : Drop slow OSCs if we can, but not for requested start idx.
1157 This means "if OSC is slow and it is not the requested
1158 start OST, then it can be skipped, otherwise skip it only
1159 if it is inactive/recovering/out-of-space.
1163 Description: more cleanup in mds_lov
1164 Details : not send LOV EA under replay, we can't know about they size at this
1165 time. Don't allow client connect to mds before any ost connected,
1166 for avoid problems with LOV EA size and returning EIO to client.
1168 Severity : enhancement
1170 Description: Interoperability at server side (Disk interoperability)
1172 Severity : enhancement
1174 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1176 Severity : enhancement
1178 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1180 Severity : enhancement
1182 Description: New client IO stack (CLIO).
1184 Severity : enhancement
1186 Description: Commit on sharing. Eliminate inter-client dependencies between
1187 uncommitted transactions by doing transaction commits.
1188 Thereby clients may recovery independently.
1191 Frequency : Create a symlink file with a very long name
1193 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1194 Details : If there is no extra space in the request for early cancels,
1195 ldlm_req_handles_avail() returns 0 instead of a negative value.
1197 Severity : enhancement
1199 Description: Add /proc entry for import status
1200 Details : The mdc, osc, and mgc import directories now have
1201 an import directory that contains useful import data for debugging
1202 connection problems.
1204 Severity : enhancement
1206 Description: Re-disable certain /proc logging
1207 Details : Enable and disable client's offset_stats, extents_stats and
1208 extents_stats_per_process stats logging on the fly.
1211 Frequency : Only on FC kernels 2.6.22+
1213 Description: oops in statahead
1214 Details : Do not drop reference count for the dentry from VFS when lookup,
1215 VFS will do that by itself.
1217 Severity : enhancement
1219 Description: Generic /proc file permissions
1220 Details : Set /Proc file permissions in a more generic way to enable non-
1221 root users operate on some /proc files.
1225 Description: Hitting mdc_commit_close() ASSERTION
1226 Details : Properly handle request reference release in
1227 ll_release_openhandle().
1231 Description: quota recovery deadlock during mds failover
1232 Details : This patch includes att18982, att18236, att18237 in bz14840.
1234 1. fix osts hang when mds does failover with quotaon
1235 2. prevent watchdog storm when osts threads wait for the
1240 Frequency : only patchless client
1241 Description: add workaround for race between add/remove dentry from hash
1243 Severity : enhancement
1245 Description: Allow OST glimpses to return PW locks
1249 Description: LBUG when llog conf file is full
1250 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1254 Description: Prevent import from entering FULL state when server in recovery
1258 Description: service mount cannot take device name with ":"
1259 Details : Only when device name contains ":/" will mount treat it as
1265 Description: replace ptlrpcd with the statahead thread to interpret the async
1266 statahead RPC callback
1270 Frequency : on recovery
1271 Description: I/O failures after umount during fail back
1272 Details : if client reconnected to restarted server we need join to recovery
1273 instead of find server handler is changed and process self
1274 eviction with cancel all locks.
1276 Severity : enhancement
1278 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1280 Severity : enhancement
1282 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1284 Severity : enhancement
1286 Description: Upcall on Lustre log has been dumped
1287 Details : Allow for a user mode script to be called once a Lustre log has
1288 been dumped. It passes the filename of the dumped log to the
1289 script, the location of the script can be specified via
1290 /proc/sys/lnet/debug_log_upcall.
1295 Description: avoid idr_remove called for id which is not allocated.
1296 Details : Move assigment s_dev for clustered nfs to end of initialization,
1297 to avoid problem with error handling.
1302 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1303 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1304 into CONN_USED_HASH and this prodice warning when put connection
1305 again in unused hash.
1308 Severity : enhancement
1310 Description: File striping can now be set to use an arbitrary pool of OSTs.
1312 Severity : enhancement
1314 Description: Export bytes_read/bytes_write count on OSC/OST.
1318 Description: Early reply size mismatch, MGC loses connection
1319 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1320 the connect flags are properly negotiated.
1325 Description: quotas are not honored with O_DIRECT
1326 Details : all writes with the flag O_DIRECT will use grants which leads to
1327 this problem. Now using OBD_BRW_SYNC to guard this.
1331 Description: add quota statistics
1332 Details : 1. sort out quota proc entries and proc code.
1333 2. add quota statistics
1335 Severity : enhancement
1337 Description: enable quota support for HEAD.
1341 Description: Properly propagate oinfo flags from lov to osc for statfs
1342 Details : restore missing copy oi_flags to lov requests.
1344 Severity : enhancement
1346 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1347 : l_getgroups(8), lst(8), routerstat(8)
1348 Details : included man pages for llobdstat(8), llstat(8),
1349 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1351 Severity : enhancement
1353 Description: configure's --enable-quota should check the
1354 : kernel .config for CONFIG_QUOTA
1355 Details : configure is terminated if --enable-quota is passed but
1356 : no quota support is in kernel
1360 Description: Remove portals compatibility
1361 Details : Remove portals compatibility, not interoperable with releases
1366 Description: Resolve device initialization race
1367 Details : Prevent proc handler from accessing devices added to the
1368 obd_devs array but yet be intialized.
1370 Severity : enhancement
1372 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1374 Severity : enhancement
1376 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1381 Description: Using wrong pointer in osc_brw_prep_request
1382 Details : Access to array[-1] can produce panic if kernel compiled with
1383 CONFIG_PAGE_ALLOC enabled
1387 Description: Client runs out of low memory
1388 Details : Consider only lowmem when counting initial number of llap pages
1392 Description: *optional* service tags registration
1393 Details : if the "service tags" package is installed on a Lustre node
1394 When the filesystem is mounted, a local-node service tag will
1395 be created. See http://inventory.sun.com/ for more information
1396 about the Service Tags asset management system.
1400 Description: Kernel BUG tries to release flock
1401 Details : Lustre does not destroy flock lock before last reference goes
1402 away. So always drop flock locks when client is evicted and
1403 perform unlock regardless of successfulness of speaking to MDS.
1407 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1411 Description: sanity test 65a fails if stripecount of -1 is set
1412 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1417 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1418 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1421 Severity : enhancement
1423 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1427 Description: do not process already freed flock
1428 Details : flock can possibly be freed by another thread before it reaches
1429 to ldlm_flock_completion_ast.
1433 Description: LBUG during stress test
1434 Details : Need properly lock accesses the flock deadlock detection list.
1438 Description: oops in page fault handler
1439 Details : kernel page fault handler can return two special 'pages' in error
1440 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1444 Description: timeout with invalidate import.
1445 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1446 should be handled by ptlrpcd. This produce long age waiting and
1447 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1449 Severity : enhancement
1451 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1455 Description: enable MGS and MDT services start separately
1456 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1457 a co-located MGS without starting the MGS, which is a complement
1458 to 'nosvc' mount option.
1462 Frequency : after recovery
1463 Description: precreate to many object's after del orphan.
1464 Details : del orphan st in oscc last_id == next_id and this triger growing
1465 count of precreated objects. Set flag LOW to skip increase count
1466 of precreated objects.
1470 Frequency : rare, on clear nid stats
1471 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1472 Details : when clean nid stats sometimes try destroy live entry,
1473 and this produce panic in free.
1477 Description: Stack overflow during MDS log replay
1478 ease stack pressure by using a thread dealing llog_process.
1482 Description: wait until IO finished before start new when do lock cancel.
1483 Details : VM protocol want old IO finished before start new, in this case
1484 need wait until PG_writeback is cleared until check dirty flag and
1485 call writepages in lock cancel callback.
1487 Severity : enhancement
1489 Description: using special macro for print time and cleanup in includes.
1493 Description: mds_mfd_close() ASSERTION(rc == 0)
1494 Details : In mds_mfd_close(), we need protect inode's writecount change
1495 within its orphan write semaphore to prevent possible races.
1499 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1503 Frequency : rare, on shutdown ost
1504 Description: don't hit live lock with umount ost.
1505 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1506 use shrink_dcache_sb instead.
1510 Description: don't panic with use echo client
1511 Details : echo client pass NULL as client nid pointer and this produce null
1512 pointer dereference.
1516 Description: fix build on ppc32
1517 Details : compile code with -m64 flag produce wrong object file for ppc32.
1521 Description: add message levels for liblustreapi
1525 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1526 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1527 will result in an -EAGAIN and in the exit call path we call
1528 alloc_rr() which will with increasing aggressiveness attempt to
1529 aquire precreated objects on the minimum number of required OSCs.
1533 Description: Use old size assignment to avoid deadlock
1534 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1535 the scheduling while holding a spinlock. We do not need locking
1536 for size in ll_update_inode() because size is only updated from
1537 the MDS for directories or files without objects, so there is no
1538 other place to do the update, and concurrent access to such inodes
1539 are protected by the inode lock.
1543 Description: resolve "_IOWR redefined" build error on SLES10
1547 Description: dump the memory debugging after all modules are unloaded to
1548 suppress false negative in conf_sanity test 39
1550 Severity : enhancement
1552 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1557 Description: fix SLES kernel versioning
1558 Details : the kernel version for our SLES 10 kernel did not include a "-"
1559 before the "smp" at the end. while this was not a problem in
1560 general, it did mean that software trying to use the kernel
1561 version to try to detect a vendor specific kernel would fail.
1562 this was most evident by the OFED build scripts.
1566 Description: Don't update lov_desc members until making sure they are valid
1567 Details : When updating lov_desc members via proc fs, need fix their
1568 validities before doing the real update.
1572 Description: don't put request into delay list while invalidate in flight.
1573 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1574 import in flight. this produce timeout for invalidate and sometimes
1575 can cause stale data.
1579 Frequency : on ppc only
1580 Description: not convert ost objects for directory because it's not exist.
1581 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1585 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1587 Severity : enhancement
1589 Description: Update to support 2.6.22.14 vanilla kernel.
1593 Frequency : rare, on recovery
1594 Description: read procfs can produce deadlock in some situation
1595 Details : Holding lprocfs lock which send rpc can produce block for destroy
1596 obd objects and this also block reconnect with -EALREADY. This
1597 isn't fix all lprocfs bugs - but make it rare.
1599 Severity : enhancement
1601 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1604 Frequency : frequent on X2 node
1606 Description: mdc_set_open_replay_data LBUG
1607 Details : Set replay data for requests that are eligible for replay.
1611 Description: lustre_mgs: operation 101 on unconnected MGS
1612 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1613 MGC, and late on MGC cannot successfully connect to MGS and a lot
1614 of the error messages complaining that MGS is not connected.
1618 Frequency : on network error
1619 Description: panic with double free request if network error
1620 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1621 but it's true only for synchronus enqueue, for async enqueue
1622 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1625 Severity : enhancement
1627 Description: client-side metadata stat-ahead during readdir(directory readahead)
1628 Details : perform client-side metadata stat-ahead when the client detects
1629 readdir and sequential stat of dir entries therein
1632 Frequency : on start mds
1634 Description: Implement get_info(last_id) in obdfilter.
1637 Frequency : occasional
1639 Description: Correctly check stale fid, not start epoch if ost not support SOM
1640 Details : open with flag O_CREATE need set old fid in op_fid3 because
1641 op_fid2 was overwritten with new generated fid, but mds can answer
1642 with one of these two fids and both is not stale. Setattr
1643 incorrectly started an epoch and assume will be called
1644 done_writing, but without SOM done_writing ever being called.
1647 Frequency : rare, depends on device drivers and load
1649 Description: MDS or OSS nodes crash due to stack overflow
1650 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1651 In some cases, in conjunction with device drivers that use a lot
1652 of stack the MDS (or possibly OSS) service threads could overflow
1653 the stack. One change which was identified to consume additional
1654 stack has been reworked to avoid the extra stack usage.
1657 Frequency : occasional
1659 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1660 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1661 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1662 oscc_has_objects and see OSC has no objests and call
1663 oscc_internal_create to resend create request.
1665 Severity : enhancement
1667 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1669 Severity : enhancement
1671 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1674 Frequency : very rare
1676 Description: Fix replay if there is an un-replied request and open
1677 Details : In some cases, older replay request will revert the
1678 mcd->mcd_last_xid on MDS which is used to record the client's
1679 latest sent request.
1681 Severity : enhancement
1683 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1685 Severity : enhancement
1687 Description: Add rhel5 support to HEAD.
1689 Serverity : enhancement
1691 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1696 Description: Don't allow skipping OSTs if index has been specified.
1697 Details : Don't allow skipping OSTs if index has been specified, make
1698 locking in internal create lots better.
1702 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1703 Details : access to bitfield in structure is always rounded to long
1704 and this produce problem with not atomic change any bit.
1708 Description: Lustre make rpms failed.
1709 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1710 builds Lustre rpms from tarball.
1713 Frequency : rare on shutdown ost
1715 Description: If llog cancel was not send before clean_exports phase, this can
1716 produce deadlock in llog code.
1717 Details : If llog thread has last reference to obd and call class_import_put
1718 this produce deadlock because llog_cleanup_commit_master wait when
1719 last llog_commit_thread exited, but this never success because was
1720 called from llog_commit_thread.
1724 Description: allow userland application know is lost one of stripes.
1725 Details : fill lvb_blocks with error code on ost and return it to
1726 application if error flag found.
1730 Description: NULL lov_tgts causing MDS oops
1731 Details : more safe checks for NULL lov_tgts for avoid oops.
1733 Severity : enhancement
1735 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1739 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1740 Details : If an OST is down the MDS will hang indefinitely in
1741 obd_statfs_rqset() waiting for the statfs data. While for
1742 MDS QOS usage of statfs, it should not stuck in waiting.
1744 Severity : enhancement
1746 Description: remote_acl support
1747 Details : Support ACL-based permission check for remote user.
1748 Support setfacl/getfacl for remote user with the utils
1749 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1750 the system "{s,g}etfacl" utils.
1752 Severity : enhancement
1754 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1756 Severity : enhancement
1758 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1762 Description: make mgs_setparam() handle fsname containing dash
1763 Details : fsname containing a dash does not work with lctl conf_param
1765 Severity : enhancement
1767 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1769 Severity : enhancement
1771 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1775 Frequency : rare, at shutdown
1776 Description: access already free / zero obd_namespace.
1777 Details : if client_disconnect_export was called without force flag set,
1778 and exist connect request in flight, this can produce access to
1779 NULL pointer (or already free pointer) when connect_interpret
1780 store ocd flags in obd_namespace.
1784 Frequency : only at startup
1785 Description: not alloc memory with spinlock held.
1786 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1787 if any spinlock held.
1789 Severity : enhancement
1791 Description: make lustre randomly fail allocating memory
1792 Details : Make lustre randomly failed allocating memory for testing purpose.
1794 Severity : enhancement
1796 Description: lost problems with lov objid file
1797 Details : Fixes some scability and access to not inited memory problems
1798 in work with lov objdid file.
1803 Description: lfs find does not continue on file error
1804 Details : Continue other files processing when a file/dir is absent.
1808 Description: Inconsistent usage of lustre_pack_reply()
1809 Details : Standardize the usage of lustre_pack_reply() such that it
1810 always generate a CERROR on failure.
1813 Frequency : occasional
1815 Description: MDS hang or stay in waiting lock
1816 Details : If client receive lock with CBPENDING flag ldlm need send lock
1817 cancel as separate rpc, to avoid situation when cancel request
1818 can't processed due all i/o threads stay in wait lock.
1822 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1826 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1830 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1834 Description: Update to RHEL4 latest kernel.
1836 Severity : enhancement
1838 Description: Build SLES10 patchless client fails
1839 Details : The configure was broken by run ./configure with
1840 --with-linux-obj=.... argument for patchless client. When the
1841 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1842 can't search header adequately. Use absolute path such as
1843 -I($LINUX)/include instead.
1847 Description: interrupt oig_wait produce painc on resend.
1848 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1849 private set, and this produce situation when rq_ptlrpcd_data not
1850 copyed to new allocated request and triggered LBUG on assert
1851 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1852 ptlrpcd set if request is changed and can be safety changed to use
1857 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1858 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1859 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1860 these macros and replace them with their difinitions instead.
1864 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1865 Details : Update which_patch & target file for SLES10 latest kernel.
1868 Description: Remove client patches from SLES 10 kernel.
1869 Details : This causes SLES 10 clients to behave as patchless clients
1870 even on a Lustre-patched (server) kernel.
1872 Severity : enhancement
1874 Description: self-adjustable client's lru lists
1875 Details : use adaptive algorithm for managing client cached locks lru
1876 lists according to current server load, other client's work
1877 pattern, memory activities, etc. Both, server and client
1878 side namespaces provide number of proc tunables for controlling
1881 Severity : enhancement
1883 Description: light-weight GSS support
1884 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1885 and reduce performance overhead.
1887 Severity : enhancement
1889 Description: Linux keyring support
1890 Details : Support using service of Linux keyring for Lustre GSS internal
1891 context refresh/cache mechanism.
1895 Description: Fix errors in lfs documentation
1896 Details : Fixes man pages
1900 Description: don't use GFP_* in generic Lustre code.
1901 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1905 Description: obdclass is limited by single OBD_ALLOC(idarray)
1906 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1910 Description: warnings with build patchless client with vanila 2.6.19 and up
1911 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1912 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1916 Frequency : Only for SLES
1917 Description: sanity_quota fail test_1
1918 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1919 file, which makes TSTID[2] inunique.
1923 Frequency : Only for SLES9
1924 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1925 Details : Update target/ChangeLog/which_patch .
1929 Frequency : Only for test_10 in sanity-quota.sh
1930 Description: a bug in quota test code
1931 Details : checking a test flag in a wrong place
1935 Frequency : Only for test_13 in sanity-quota.sh
1936 Description: a bug in quota test script
1937 Details : shouldn't check overfull allocated quota
1941 Description: jbd statistics
1942 Details : Port older jbd statistics patch for sles10
1944 Severity : enhancement
1946 Description: Add printing inode info into message about error in writepage.
1951 Description: reply_lock_interpret crash due to race with it and lock cancel.
1952 Details : Do not replay locks that are being cancelled. Do not reference
1953 locks by their address during replay, just by their handle.
1957 Frequency : When flocks are used.
1958 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1959 Details : Flock locks might destroy just granted lock if it could be merged
1960 with another existing flock, this is done in completion handler,
1961 so teach ldlm_cli_enquque_fini that this is a valid case for
1967 Description: Oops in read and write path when failing to allocate lock.
1968 Details : Check if lock allocation failed and return error back.
1972 Description: lstripe command fails for valid OST index
1973 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1974 instead of lov->desc.ld_active_tgt_count.
1978 Description: sanity.sh failed test 103
1979 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1984 Description: df doesn't work properly if diskfs blocksize != 4K
1985 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1989 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1990 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1992 Severity : enhancement
1994 Description: Adaptive timeouts
1995 Details : RPC timeouts adapt to changing server load and network
1996 conditions to reduce resend attempts and improve recovery time.
2000 Description: llapi_file_create() does not allow some changes
2001 Details : add llapi_file_open() that allows specifying the mode and
2002 open flags, and also returns an open file handle.
2006 Description: merge and cleanup kernel patches.
2007 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2011 Description: Add journal checksum support.(Kernel part)
2012 Details : The journal checksum feature adds two new flags i.e
2013 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2014 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2015 indicates that the commit block contains the checksum for
2016 the blocks described by the descriptor blocks. Now commit
2017 record can be sent to disk without waiting for descriptor
2018 blocks to be written to disk. This behavior is controlled
2019 using JBD2_FEATURE_ASYNC_COMMIT flag.
2023 Description: OSS needs mutliple precreate threads
2024 Details : Add ability to start more than one create thread per OSS.
2028 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2029 Details : execute lfs setstripe on client
2033 Description: mds_obd_create error creating tmp object
2034 Details : When the user sets quota on root, llog will be affected and can't
2035 create files and write files.
2038 Frequency : Always on ia64 patchless client, and possibly others.
2040 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2041 Details : This allows the patchless client to be loaded on architectures
2042 without this export.
2046 Description: RedHat Update kernel for RHEL5
2047 Details : Add the kernel config file for RHEL5.
2051 Description: RedHat Update kernel for RHEL5
2052 Details : Modify the kernel config file more closer RHEL5.
2056 Description: Build failure against Centos5 (RHEL5)
2057 Details : Define PAGE_SIZE when it isn't present.
2061 Description: test_7 on sanity-quota.sh can't be used on separate machines
2062 Details : add facet to handle it
2066 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2067 Details : with NFS, the anon dentry's parent was set to itself in
2068 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2069 corresponding dentry other than use rec->ur_name.
2071 Severity : enhancement
2073 Description: lfs setstripe enhancement
2074 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2078 Description: enable data checksumming by default
2079 Details : enable checksum by default, allow --disable-checksum
2080 configure option and "-o nochecksum" mount option. Checksums
2081 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2085 Description: lustre support for RHEL5
2086 Details : Add support for RHEL5.
2090 Description: Client eviction due to failover config
2091 Details : after a connection loss, the lustre client should attempt to
2092 reconnect to the last active server first before trying the
2093 other potential connections.
2097 Description: when mds and osts use different quota unit(32bit and 64bit),
2098 quota will be released repeatly.
2099 Details : void sending multiple quota reqs to mds, which will keep the
2100 status between the reqs.
2104 Description: osts not allocated evenly to files
2105 Details : change the condition to increase offset_idx
2108 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2110 Description: Data corruption for OSTs that are formatted larger than 2TB
2112 Details : When generating the bio request for lustre file writes the
2113 sector number would overflow a temporary variable before being
2114 used for the IO. The data reads correctly from Lustre (which
2115 will overflow in a similar manner) but other file data or
2116 filesystem metadata may be corrupted in some cases.
2120 Description: Tune the kernel for good SCSI performance.
2121 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2122 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2127 Description: rewrite ext2-derived code in obdclass/uuid.c
2128 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2129 from scratch preserving functionality.
2133 Description: "lfs find -obd UUID" prints directories
2134 Details : "lfs find -obd UUID" will return all directory names instead
2135 of just file names. It is incorrect because the directories
2136 do not reside on the OSTs.
2140 Description: use i_size_read and i_size_write in 2.6 port
2141 Details : replace inode->i_size access with i_size_read/write()
2144 Frequency : when removing large files
2146 Description: scheduling issue during removal of large Lustre files
2147 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2148 It causes scheduling issues when removing large files (17TB in the
2152 Frequency : only with liblustre clients on XT3
2154 Description: evictions taking too long
2155 Details : allow llrd to evict clients directly on OSTs
2158 Frequency : only on ppc
2160 Description: /proc/fs/lustre/devices broken on ppc
2161 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2162 We should make sure the type of 'index' is loff_t and then cast
2163 explicitly as needed below. Do not assign an explicitly cast
2167 Frequency : only for rhel5
2169 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2170 Details : Modify the target file & which_kernel.
2172 Severity : enhancement
2174 Description: omit set fsid for export NFS
2175 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2176 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2180 Frequency : Always, for kernels after 2.6.16
2181 Description: Fix warning idr_remove called for id=.. which is not allocated.
2182 Details : Last kernels save old s_dev before kill super and not allow
2183 to restore from callback - restore it before call kill_anon_super.
2187 Description: buffer overruns could theoretically occur
2188 Details : llapi_semantic_traverse() modifies the "path" argument by
2189 appending values to the end of the origin string, and a
2190 overrun may occur. Adding buffer overrun check in liblustreapi.
2194 Description: Fix error on 'ls .' at the top of the Lustre mount.
2195 Details : Don't revalidate dentry if it is a root dentry.
2199 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2200 Details : Modify vm-tunables-rhel4.patch.
2204 Description: Kernel config for 2.6.18-vanilla.
2205 Details : Modify targets/2.6-vanilla.target.in.
2206 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2207 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2208 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2209 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2213 Description: improve handling recoverable errors
2214 Details : if request processig with error which can be recoverable on server
2215 request should be resend, otherwise page released from cache and
2220 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2221 Details : Modify target file & which_patch.
2222 A flaw was found in the IA32 system call emulation provided
2223 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2224 value could be stored in the %RAX register, which could trigger an
2225 out-of-bounds system call table access. An untrusted local user
2226 could exploit this flaw to run code in the kernel
2227 (ie a root privilege escalation). (CVE-2007-4573).
2231 Description: O_DIRECT bypasses client statistics.
2232 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2233 read_ahead_stats were updated. copied the stats section from
2234 osc_send_oap_rpc() into async_internal().
2238 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2242 Description: change order of libsysio includes
2243 Details : '#include sysio.h' should always come before '#include xtio.h'
2247 Description: handle "serious error: objid * already exists" more gracefully
2248 Details : If LAST_ID value on disk is smaller than the objects existing in
2249 the O/0/d* directories, it indicates disk corruption and causes an
2250 LBUG(). If the object is 0-length, then we should use the existing
2251 object. This will help to avoid a full fsck in most cases.
2253 Severity : enhancement
2255 Description: adapt the lustre_config script to support the upgrade case
2256 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2257 server targets to 1.6 in parallel.
2261 Description: To avoid grant space > avaible space when the disk is almost
2262 full. Without this patch you might see the error "grant XXXX >
2263 available" or some LBUG about grant, when the disk is almost
2265 Details : In filter_check_grant, for non_grant cache write, we should
2266 check the left space by if (*left > ungranted + bytes), instead
2267 of (*left > ungranted), because only we are sure the left space
2268 is enough for another "bytes", then the ungrant space should be
2269 increase. In client, we should update cl_avail_grant only there
2270 is OBD_MD_FLGRANT in the reply.
2274 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2275 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2276 described in CVE-2007-4573. This problem could allow untrusted
2277 local users to gain root access.
2280 Frequency : when using O_DIRECT and quotas
2282 Description: Incorrect file ownership on O_DIRECT output files
2283 Details : block usage reported by 'lfs quota' does not take into account
2284 files that have been written with O_DIRECT.
2289 Description: touch file failed when fs is not full
2290 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2291 otherwise we can get ENOSP while fs is not full.
2295 Description: parallel lock callbacks
2296 Details : Instead of sending blocking and completion callbacks as separated
2297 requests, adding them to a set and sending in parallel.
2300 Frequency : only for Cray XT3
2301 Bugzilla : 12829/13455
2302 Description: Changing primary group doesn't change the group lustre assigns to
2304 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2305 overridden with the primary group provided by the group upcall,
2306 whereas the supplied fsgid can be trusted if it is in the list of
2307 supplementary groups returned by the group upcall.
2309 Severity : enhancement
2311 Description: Allow masking D_WARNING, D_ERROR messages from console
2312 Details : Console messages can now be disabled via lnet.printk.
2316 Description: User code with malformed file open parameter crashes client node
2317 Details : Before packing join_file req, all the related reference should be
2318 checked carefully in case some malformed flags cause fake
2319 join_file req on client.
2323 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2324 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2325 mode is equal to the granted mode. However, it should also check
2326 that LDLM_FL_CP_REQD is not set.
2330 Description: Heavy nfs access might result in deadlocks
2331 Details : After ELC code landed, it is now improper to enqueue any mds
2332 locks under och_sem, because enqueue might want to decide to
2333 cancel open locks for same inode we are holding och_sem for.
2337 Description: Client eviction while running blogbench
2338 Details : A lot of unlink operations with concurrent I/O can lead to a
2339 deadlock causing evictions. To address the problem, the number of
2340 oustanding OST_DESTROY requests is now throttled to
2341 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2342 callbacks are processed in priority.
2346 Description: enable ACLs on MDS by default
2347 Details : ACLs must be enabled on MDS by default.
2350 Frequency : PPC/PPC64 only
2352 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2353 Details : fix duplicated definitions between asm-ppc64/types.h and
2354 lustre_types.h on PPC.
2357 Frequency : PPC/PPC64 only
2359 Description: asm-ppc/segment.h does not exist
2360 Details : fix compile issue on PPC.
2364 Description: better handle error messages in extents code
2367 Frequency : RHEL4 only
2369 Description: mkfs is very slow on IA64/RHEL4
2370 Details : A performance regression has been discovered in the MPT Fusion
2371 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2372 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2373 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2375 Severity : enhancement
2377 Description: SNMP support enhancement
2378 Details : Adding total number of sampled request for an MDS node in snmp
2381 Severity : enhancement
2383 Description: Optimize ldlm waiting list processing for PR extent locks
2384 Details : When processing waiting list for read extent lock and meeting read
2385 lock that is same or wider to it that is not contended, skip
2386 processing rest of the list and immediatelly return current
2387 status of conflictness, since we are guaranteed there are no
2388 conflicting locks in the rest of the list.
2392 Description: Time out and refuse to reconnect
2393 Details : When the failover node is the primary node, it is possible
2394 to have two identical connections in imp_conn_list. We must
2395 compare not conn's pointers but NIDs, otherwise we can defeat
2396 connection throttling.
2400 Description: port llog fixes from b1_6 into HEAD
2401 Details : Port llog reference couting and some llog cleanups from b1_6
2402 (bug 10800) into HEAD, for protect from panic and access to already
2403 free llog structures.
2407 Description: Detect stride IO mode in read-ahead
2408 Details : When a client does stride read, read-ahead should detect that and
2409 read-ahead pages according to the detected stride pattern.
2413 Description: data checksumming impacts single node performance
2414 Details : add support for several checksum algorithm. Currently, only CRC32
2415 and Adler-32 are supported. The checksum type can be changed on
2416 the fly via /proc/fs/lustre/osc/*/checksum_type.
2420 Description: use adler32 for page checksums
2421 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2426 Description: build for x2 fails
2427 Details : fix compile issue on Cray systems.
2431 Description: Properly match for duplicate locks
2432 Details : Due to different lock order from skiplists code, we need to
2433 traverse entire list for now
2436 Frequency : only on PPC/SLES10
2438 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2439 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2440 be defined. Add a hack in mkfs_lustre.c to work around this compile
2445 Description: LASSERT on MDS when client holding flock lock dies
2446 Details : ldlm pool logic depends on number of granted locks equal to
2447 number of released locks which is not true for flock locks, so
2448 just exclude such locks from consideration.
2452 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2453 Details : Use fsfilt_sync() for both the whole filesystem sync and
2454 individual file sync to eliminate dangerous inode locking
2455 with I_LOCK that can lead to a deadlock.
2459 Description: performance in 1.6.3
2460 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2461 sure that 1MB requests are not fragmented by the block layer.
2463 Severity : enhancement
2465 Description: organize the server-side client stats on per-nid basis
2466 Details : Change the structure of stats under obdfilter and mds to
2474 The "uuid"s file would list the uuids of _active_ exports.
2475 And the clear entry is to clear all stats and stale nids.
2477 Severity : enhancement
2479 Description: eliminate client locks in face of contention
2480 Details : file contention detection and lockless i/o implementation
2481 for contended files.
2485 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2488 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2490 Description: MDS may lose file striping (and hence file data) in some cases
2491 Details : If there are additional extended attributes stored on the MDS,
2492 in particular ACLs, SELinux, or user attributes (if user_xattr
2493 is specified for the client mount options) then there is a risk
2494 of attribute loss. Additionally, the Lustre file striping
2495 needs to be larger than default (e.g. striped over all OSTs),
2496 and an additional attribute must be stored initially in the
2497 inode and then increase in size enough to be moved to the
2498 external attribute block (e.g. ACL growing in size) for file
2503 Description: skiplist implementation simplification
2504 Details : skiplists are used to group compatible locks on granted list
2505 that was implemented as tracking first and last lock of each
2506 lock group the patch changes that to using doubly linked lists
2510 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2511 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2512 because LASSERT on both the data supplied by a client, and the
2513 data on disk is dangerous and incorrect.
2515 Severity : enhancement
2517 Description: Slow truncate/writes to huge files at high offsets.
2518 Details : Directly associate cached pages to lock that protect those pages,
2519 this allows us to quickly find what pages to write and remove
2520 once lock callback is received.
2524 Description: more ldlm soft lockups
2525 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2526 starve other threads from the resource lock for a long time in
2527 case of long waiting queue, so change the debug level from
2528 D_OTHER to the less frequently used D_INFO.
2530 Severity : enhancement
2532 Description: add -gid, -group, -uid, -user options to lfs find
2536 Description: Hung threads in invalidate_inode_pages2_range
2537 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2538 once one is completed. As a result, some RPCs are stuck in the
2539 queue and are never sent.
2543 Description: filter threads hungs on waiting journal commit
2544 Details : Cleanup filter group llog code, then only filter group llog will
2545 be only created in the MDS/OST syncing process.
2549 Description: Procfs and llog threads access destoryed import sometimes.
2550 Details : Sync the import destoryed process with procfs and llog threads by
2551 the import refcount and semaphore.
2553 Severity : enhancement
2555 Description: openlock cache of b1_6 port to HEAD
2560 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2561 Details : If insertion of an extent fails, then discard the inode
2562 preallocation and free data blocks else it can lead to duplicate
2567 Description: don't always update ctime in ext3_xattr_set_handle()
2568 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2569 In some cases the ctime should not be updated, for example for
2570 2.0->1.8 compatibility it is necessary to delete an xattr and it
2571 should not update the ctime.
2575 Bugzilla : 15713/16362
2576 Description: Assertion in iopen_connect_dentry in 1.6.3
2577 Details : looking up an inode via iopen with the wrong generation number can
2578 populate the dcache with a disconneced dentry while the inode
2579 number is in the process of being reallocated. This causes an
2580 assertion failure in iopen since the inode's dentry list contains
2581 both a connected and disconnected dentry.
2585 Description: assertion failure in ldlm_handle2lock()
2586 Details : fix a race between class_handle_unhash() and class_handle2object()
2587 introduced in lustre 1.6.5 by bug 13622.
2592 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2593 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2598 Description: oops in sd_iostats_seq_show()
2599 Details : unloading/reloading the scsi low level driver triggers a kernel
2600 bug when trying to access the sd iostat file.
2605 Description: Kernel panics during QLogic driver reload
2606 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2607 patch, causing memory corruption.
2612 Description: journal_dev option does not work in b1_6
2613 Details : pass mount option during pre-mount.
2615 Severity : enhancement
2617 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2618 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2619 extent information of a file. It can be used to map logical blocks
2620 in a file to physical blocks in the block device.
2624 Description: LDLM soft lockups - improvement
2625 Details : It is be possible to send the lock handle along with each read
2626 or write request because the client is already doing a lock match
2627 itself so there isn't any reason the OST should have to re-do that
2631 Frequency : only on Cray X2
2633 Description: X2 build failures
2634 Details : fix build failures on Cray X2.
2638 Description: xid & resent requests
2639 Details : Initialize RPC XID from clock at startup (randomly if clock is
2642 Severity : enhancement
2644 Description: Add lustre_start utility to start or stop multiple Lustre servers
2649 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2650 Details : In case of memory pressure, list_del() can be called twice on
2651 req->rq_history_list, causing a kernel oops.
2655 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2656 Details : fix stack overflow in the distributed lock manager by defering
2657 export eviction after a failed AST to the elt thread instead of
2658 handling it in the dlm interpret routine.
2662 Description: Convert some comments to new format.
2663 Details : Update documenting comments to match doxygen conventions.
2667 Description: Grammar fixes.
2668 Details : A couple of trivial spelling fixes.
2672 Description: OSD_COUNTERS-mandatory
2673 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2674 to simplify the code.
2678 Description: simplify cmm_device freeing logic.
2679 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2683 Description: Add lockdep support to dt_object_operations locking interface.
2684 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2685 parameter indicating lock ordering. Update mdd code to use new
2690 Description: Introduce failloc constants for lockless IO tests.
2691 Details : Add two new failloc constants to test lockless IO. Only one of
2692 them in implemented---another is checked in yet to be landed
2697 Description: Add lockdep support for inode mutex.
2698 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2699 in the situations where Lustre has to lock more than one inode
2704 Description: Add optional invariants checking support.
2705 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2706 configure switch. This macro is to be used for consistency and
2707 sanity checks that are too expensive to be left in `production'
2712 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2713 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2718 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2719 Details : Use spin_lock_nested() in (the only) situation where more than
2720 one ldlm_lock is locked simultaneously. Also, fix possible
2721 dead-lock in ldlm_lock_change_resource() by enforcing particular
2726 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2727 Details : Instead of specifying each ldlm_lock call-back through separate
2728 parameter, wrap them into struct ldlm_callback_suite.
2732 Description: Kill join_lru obd method and its callers.
2733 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2734 in memory---a requirement implemented through ->o_join_lru() obd
2735 method in HEAD. Remove this method and its users.
2739 Description: Add asynchronous ldlm ENQUEUE completion handler.
2740 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2741 case that stock ldlm_completion_ast() cannot handle as it waits
2742 until lock is granted. Introduce new ldlm_completion_ast_async()
2743 for this. Also comment ldlm_completion_ast().
2747 Description: ldlm_error <-> errno conversion.
2748 Details : Add functions to map (rather arbitrary) between LDLM error codes
2749 and standard errno values. CLIO needs this to prevent LDLM specific
2750 constants from escaping ldlm and osc.
2754 Description: Kill unused ldlm_handle2lock_ns() function.
2755 Details : Kill unused ldlm_handle2lock_ns() function.
2759 Description: Add lu_ref support to ldlm_lock
2760 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2761 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2762 and ldlm_resource. LDLM interface has to be changed, because code that
2763 releases a reference on a lock, has to "know" what reference this is.
2764 In the most frequent case
2766 lock = ldlm_handle2lock(handle);
2768 LDLM_LOCK_PUT(lock);
2770 no changes are required. When any other reference (received _not_ from
2771 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2772 instead of LDLM_LOCK_PUT().
2774 Arguably, changes are pervasive, and interface requires some discipline
2775 for proper use. On the other hand, it was very instrumental in finding
2776 a few leaked lock references.
2780 Description: Add ldlm_lock_addref_try().
2781 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2782 attempts to addref a lock that might be being canceled
2787 Description: Add ldlm_weigh_callback().
2788 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2789 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2790 hard-coded `number of pages' logic.
2794 Description: Add lockdep annotations to llog code.
2795 Details : Use appropriately tagged _nested() locking calls in the places
2796 where llog takes more than one ->lgh_lock lock.
2800 Description: Add loi_kms_set().
2801 Details : Wrap kms updates into a helper function.
2805 Description: Constify instances of struct lsm_operations.
2806 Details : Constify instances of struct lsm_operations.
2810 Description: lu_conf support.
2811 Details : On a server, a file system object is uniquely identified
2812 by a fid, which is sufficient to locate and load all object
2813 state (inode). On a client, on the other hand, more data are
2814 necessary instantiate an object. Change lu_object_find() and
2815 friends to take additional `lu_conf' argument describing object.
2816 Typically this includes layout information.
2820 Description: lu_context fixes.
2821 Details : Introduce new lu_context functions that are needed on the client
2822 side, where some system threads (ptlrpcd) are shared by multiple
2823 modules, and so cannot be stopped during module shutdown.
2827 Description: Add start and stop methods to lu_device_type_operations.
2828 Details : Introduce two new methods in lu_device_type_operations, that are
2829 invoked when first instance of a given type is created and last one
2830 is destroyed respectively. This is need by CLIO.
2834 Description: Add lu_ref support to struct lu_device.
2835 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2836 track leaked references.
2840 Description: Introduce lu_kmem_descr.
2841 Details : lu_kmem_descr and its companion interface allow to create
2842 and destroy a number of kmem caches at once.
2846 Description: Fix lu_object finalization race.
2847 Details : Fix a race between lu_object_find() finding an object and its
2848 concurrent finalization. This race is (most likely) not possible
2849 on the server, but might happen on the client.
2853 Description: Introduce lu_ref interface.
2854 Details : lu_ref is a debugging module allowing to track references to
2855 a given object. It is quite cpu expensive, and has to be
2856 explicitly enabled with --enable-lu_ref. See usage description
2861 Description: Factor lu_site procfs stats into a separate function.
2862 Details : Separate lu_site stats printing code into a separate function
2863 that can be reused on a client.
2867 Description: Constify instances of struct {lu,dt,md}_device_operations.
2868 Details : Constify instances of struct {lu,dt,md}_device_operations.
2872 Description: Introduce struct md_site and move meta-data specific parts of
2873 struct lu_site here.
2874 Details : Move md-specific fields out of struct lu_site into special struct
2875 md_site, so that lu_site can be used on a client.
2879 Description: Kill mdd_lov_destroy().
2880 Details : Remove unused mdd code.
2884 Description: Add st_block checking to multistat.c.
2885 Details : Add st_block checking to multistat.c.
2889 Description: Add lu_ref support to struct obd_device.
2890 Details : Add lu_ref tracking to obd_device.
2894 Description: Kill obd_set_fail_loc().
2895 Details : Remove unused code.
2899 Description: Add special type for ptlrpc_request interpret functions.
2900 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2901 there. Actual usage will be in CLIO.
2905 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2906 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2907 doesn't work with lockdep.
2911 Description: Add rwv.c test program.
2912 Details : New testing program exercising readv(2) and writev(2) (Qian).
2916 Description: Add sendfile.c test program.
2917 Details : New testing program exercising sendfile(2) (Jay).
2921 Description: Ratelimit a message that can be very frequent.
2922 Details : Ratelimit a memory allocation failure message that can
2927 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2928 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2932 Description: Liblustre build fixes.
2933 Details : Liblustre build fixes.
2937 Description: libcfs: add cfs_{need,cond}_resched() interface.
2938 Details : libcfs: add cfs_{need,cond}_resched() definition and
2939 implementations for Linux, NT, and liblustre.
2941 Severity : enhancement
2943 Description: More exported tunables for mballoc
2944 Details : Add support for tunable preallocation window and new tunables for
2945 large/small requests
2949 Description: Detect corruption of block bitmap and checking for preallocations
2950 Details : Checks validity of on-disk block bitmap. Also it does better
2951 checking of number of applied preallocations. When corruption is
2952 found, it turns filesystem readonly to prevent further corruptions.
2956 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2957 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2958 window to 0. If it does hit cache miss, and read-pattern is still
2959 stride-io mode, does not reset the stride window, but also does not
2960 increase the stride window length in this case.
2964 Frequency : only for big-endian servers
2965 Description: Check if system is big-endian while mounting fs with extents feature
2966 Details : Mounting a filesystem with extents feature will fail on big-endian
2967 systems since ext3-based ldiskfs is not supported on big-endian
2968 systems. This can be over-riden with "bigendian_extents" mount option.
2970 Severity : enhancement
2972 Description: The root squash functionality
2973 Details : A security feature, which is to prevent users from being able
2974 to mount lustre on their desktop, run as root, and delete
2975 all of the files in the filesystem. The goal is accomplished by
2976 remapping user id (UID) and group id (GID) of the root user to
2977 a UID and GID specified by the system administartor via Lustre
2978 configuration management server (MGS). The functionality also
2979 allows to specify sets of clients for which the remapping does
2984 Description: Excessive recovery window
2985 Details : With AT enabled, the recovery window can be excessively long (6000+
2986 seconds). To address this problem, we no longer use
2987 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2988 timeout no longer depends on the service time, it is set to
2989 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2990 time via pb_service_time.
2994 Description: Watchdog triggered on MDS failover
2995 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2996 the OSTs know that the MDS "UUID" can be reused for the same export
2997 from a different NID, so we do not need to wait for the export to be
3001 Frequency : rare, only if using MMP with Linux RAID
3003 Description: MMP doesn't work with Linux RAID
3004 Details : While using HA for Lustre servers with Linux RAID, it is possible
3005 that MMP will not detect multiple mounts. To make this work we
3006 need to unplug the device queue in RAID when the MMP block is being
3007 written. Also while reading the MMP block, we should read it from
3008 disk and not the cached one.
3010 Severity : enhancement
3012 Description: open file using fid
3013 Details : A file can be opened using just its fid, like
3014 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3017 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3020 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3021 Details : add proper locking for creation of jbd_4k slab cache
3025 Description: MMP check in ext3_remount() fails without displaying any error
3026 Details : When multiple mount protection fails during remount, proper error
3029 Severity : enhancement
3031 Description: Allow stripe size to be up to 4G-64k
3032 Details : Fix math logic to allow large stripe sizes.
3036 Description: add check for >8TB ldiskfs filesystems
3037 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3038 Don't allow >8TB ldiskfs filesystems to be mounted without
3039 force_over_8tb mount option
3044 Descriptoin: OSS complains "too many blocks" and crashes
3045 Details : In filter_commitrw_write():
3046 1. lquota_chkquota();
3047 2. obdo_from_inode();
3048 3. lquota_pending_commit();
3049 "2." has changed uid and gid in this case, then caused
3050 lquota_chkquota() and lquota_pending_commit() to use
3051 different uid and gid. That is the root of the bug.
3055 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3056 Details : Detect and print out the 'stripe count = -1' case properly.
3060 Description: Out or order replies might be lost on replay
3061 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3062 request with tid smaller than one currently being added, add it
3063 to the start, not end of the list.
3065 --------------------------------------------------------------------------------
3067 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3069 * Support for kernels:
3070 2.6.5-7.283 (SLES 9),
3071 2.6.9-55.EL (RHEL 4),
3072 2.6.16.46-0.14 (SLES 10),
3073 2.6.18.8 vanilla (kernel.org)
3074 * Client support for unpatched kernels:
3075 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3076 2.6.16 - 2.6.22 vanilla (kernel.org)
3077 * Due to recently discovered recovery problems, we do not recommend
3078 using patchless RHEL 4 clients with this or any earlier release.
3079 * Recommended e2fsprogs version: 1.39.cfs8
3080 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3081 * Starting with this release, the ldiskfs backing filesystem required
3082 by Lustre is now in its own package, lustre-ldiskfs. This package
3083 should be installed. It is versioned separately from Lustre and
3084 may be released separately in future.
3088 Description: block reactivating mgc import until all deactivates complete
3089 Details : Fix race when failing back MDT/MGS to itself (testing)
3092 Frequency : at statup only
3094 Description: mds_lov_synchronize race leads to various problems
3095 Details : simultaneous MDT->OST connections at startup can cause the
3096 sync to abort, leaving the OSC in a bad state.
3098 Severity : enhancement
3100 Description: add optional extra BUILD_VERSION info
3101 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3102 to override the lustre version.
3105 Frequency : 2.6.18 servers only
3107 Description: ll_kern_mount() doesn't release the module reference
3108 Details : The ldiskfs module reference count never drops down to 0
3109 because ll_kern_mount() doesn't release the module reference.
3114 Description: server LBUG when using old ost_num_threads parameter
3115 Details : Accept the old ost_num_threads parameter but warn that it
3116 is deprecated, and fix an off-by-one error that caused an LBUG.
3121 Description: Transient SCSI error results in persistent IO issue
3122 Details : iobuf->dr_error is not reinitialized to 0 between two
3126 Frequency : sometimes when underlying device returns I/O errors
3128 Description: OSTs not going read-only during write failures
3129 Details : OSTs are not remounted read-only when the journal commit threads
3130 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3131 instead of the ext3 wrappers.
3135 Description: poor connect scaling with increasing client count
3136 Details : Don't run filter_grant_sanity_check for more than 100 exports
3137 to improve scaling for large numbers of clients.
3140 Frequency : SLES10 only
3142 Description: sanity-quota.sh quotacheck failed: rc = -22
3143 Details : Quotas cannot be enabled on SLES10.
3146 Frequency : liblustre clients only
3148 Description: getdirentries does not give error when run on compute nodes
3149 Details : getdirentries does not fail when the size specified as an argument
3150 is too small to contain at least one entry
3152 Severity : enhancement
3154 Description: Add LNET router traceability for debug purposes
3155 Details : If a checksum failure occurs with a router as part of the
3156 IO path, the NID of the last router that forwarded the bulk data
3157 is printed so it can be identified.
3162 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3163 Details : Due to a race condition, liblustre clients were occasionally
3164 evicted incorrectly.
3166 Severity : enhancement
3168 Description: lfs setstripe use optional parameters instead of postional
3171 Severity : enhancement
3173 Description: Nanosecond timestamp support for ldiskfs
3174 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3175 resolution timestamps. There is not yet support for this at
3176 the Lustre filesystem level.
3179 Frequency : during server recovery
3181 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3182 Details : request with rq_no_resend flag not awake l_wait_event if they get
3186 Frequency : nfs export on patchless client
3188 Description: connectathon hang when test nfs export over patchless client
3189 Details : Disconnected dentry cannot be found with lookup, so we do not need
3190 to unhash it or make it invalid
3193 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3194 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3195 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3200 Description: open req refcounting wrong on reconnect
3201 Details : If reconnect happened between getting open reply from server and
3202 call to mdc_set_replay_data in ll_file_open, we will schedule
3203 replay for unreferenced request that we are about to free.
3204 Subsequent close will crash in variety of ways.
3205 Check that request is still eligible for replay in
3206 mdc_set_replay_data().
3211 Description: disable writes to filesystem when reading health_check file
3212 Details : the default for reading the health_check proc file has changed
3213 to NOT do a journal transaction and write to disk, because this
3214 can cause reads of the /proc file to hang and block HA state
3215 checking on a healthy but otherwise heavily loaded system. It
3216 is possible to return to the previous behaviour during configure
3217 with --enable-health-write.
3219 Severity : enhancement
3221 Description: 64-bit inode version
3222 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3223 to the inode. This will be required for version-based recovery.
3228 Description: MDS fails to start if a duplicate client export is detected
3229 Details : in some rare cases it was possible for a client to connect to
3230 an MDS multiple times. Upon recovery the MDS would detect this
3231 and fail during startup. Handle this more gracefully.
3233 Severity : enhancement
3235 Description: Add -o localflock option to simulate old noflock
3237 Details : This will achieve local-only flock/fcntl locks
3243 Description: log_commit_thread vs filter_destroy race leads to crash
3244 Details : Take import reference before releasing llog record semaphore
3249 Description: Wrong request locking in request set processing
3250 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3251 imp_delayed_list, in this place should be used imp_lock.
3254 Frequency : when reconnection
3256 Description: Grant Leak when osc reconnect to OST
3257 Details : When osc reconnect ost, OST(filter) should check whether it
3258 should grant more space to client by comparing fed_grant and
3259 cl_avail_grant, and return the granted space to client instead
3260 of "new granted" space, because client will call osc_init_grant
3261 to update the client grant space info.
3264 Frequency : when client reconnect to OST
3266 Description: Grant Leak when osc do resend and replay bulk write
3267 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3268 bulk write request, because the grant info while be sync between
3269 OSC and OST when reconnect, and we should ignore the grant info
3270 these of resend/replay write req.
3275 Description: Grant space more than available left space sometimes.
3276 Details : When then OST is about to be full, if two bulk writing from
3277 different clients came to OST. Accord the avaliable space of the
3278 OST, the first req should be permitted, and the second one
3279 should be denied by ENOSPC. But if the seconde arrived before
3280 the first one is committed. The OST might wrongly permit second
3281 writing, which will cause grant space > available space.
3284 Frequency : when client is evicted
3286 Description: Grant might be wrongly erased when osc is evicted by OST
3287 Details : when the import is evicted by server, it will fork another
3288 thread ptlrpc_invalidate_import_thread to invalidate the
3289 import, where the grant will be set to 0. While the original
3290 thread will update the grant it got when connecting. So if
3291 the former happened latter, the grant will be wrongly errased
3292 because of this race.
3297 Description: Checking Stale with correct fid
3298 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3299 to check whether it is stale, because sometimes, we want the
3300 enqueue happened anyway, and op_data.fid2 will not be initialized.
3302 Severity : enhancement
3304 Description: update patchless client
3305 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3308 Frequency : only with 2.4 kernel
3310 Description: random memory corruption
3311 Details : size of struct ll_inode_info is to big for union inode.u and this
3312 can be cause of random memory corruption.
3317 Description: Memory leak in recovery
3318 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3319 It should also check obd_fail before fsfilt_start, otherwise if
3320 fsfilt_start return -EROFS,(failover mds during mds recovery).
3321 then the req will return with repmsg->transno = 0 and rc = EROFS.
3322 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3323 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3324 be freed no matter whether fsfilt_commit success or not.
3327 Frequency : only with huge count clients
3329 Description: Prevents from taking the superblock lock in llap_from_page for
3331 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3332 for llap_from_page call in ll_removepage prevents from taking the
3333 superblock lock for a soon died page.
3338 Description: Not check open intent error before release open handle
3339 Details : in some rare cases, the open intent error is not checked before
3340 release open handle, which may cause
3341 ASSERTION(open_req->rq_transno != 0), because it tries to release
3342 the failed open handle.
3347 Description: Set cat log bitmap only after create log success.
3348 Details : in some rare cases, the cat log bitmap is set too early. and it
3349 should be set only after create log success.
3353 Description: Accessing a block bevice can re-enable I/O when Lustre is
3354 tearing down a device.
3355 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3360 Description: service threads may hog cpus when there are a lot of requests
3361 Details : Insert cond_resched to give other threads a chance to use some CPU
3366 Description: the cat log was not initialized in recovery
3367 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3368 unlink log on mds will not be initialized until mds post
3369 recovery. And also in mds post recovery, the unlink log will
3370 initialization will be done asynchronausly, so there will be race
3371 between add unlink log and unlink log initialization.
3375 Description: brw_stats were being printed incorrectly
3376 Details : brw_stats were being printed as log2 but all of them were not
3377 recorded as log2. Also remove some code duplication arising from
3378 filter_tally_{read,write}.
3382 Frequency : rare, only in recovery.
3383 Description: ASSERTION(req->rq_type != LI_POISON) failed
3384 Details : imp_lock should be held while iterating over imp_sending_list for
3385 prevent destroy request after get timeout in ptlrpc_queue_wait.
3389 Description: replay-single.sh test 52 fails
3390 Details : A lock's skiplist need to be cleanup when it being unlinked
3391 from its resource list.
3395 Description: Short directio read returns full requested size rather than
3397 Details : Direct I/O operations should return actual amount of bytes
3398 transferred rather than requested size.
3400 Severity : enhancement
3402 Description: metadata RPC reduction (e.g. for rm performance)
3403 Details : decrease the amount of synchronous RPC between clients and servers
3404 by canceling conflicing lock before the operation on the client
3405 and packing thier handles into the main operation RPC to server.
3407 Severity : enhancement
3409 Description: Async OSC create to avoid the blocking unnecessarily.
3410 Details : If an OST has no remaining object, system will block on the
3411 creation when it needs to create a new object on this OST. Now,
3412 ways use pre-created objects when available, instead of blocking on
3413 an empty osc while others are not empty. If we must block, we
3414 block for the shortest possible period of time.
3418 Description: improve handling recoverable errors
3419 Details : if request processig with error which can be recoverable on server
3420 request should be resend, otherwise page released from cache and
3423 Severity : enhancement
3425 Description: refine locking for avoid write wrong info into lov_objid file
3426 Details : fix possible races with add new target and write/update data in
3429 --------------------------------------------------------------------------------
3431 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3436 Frequency : on some architectures
3438 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3439 Details : Uninitialized flags sometimes cause configuration commands to
3443 Frequency : patchless clients only
3445 Description: missing __iget() symbol export
3446 Details : The __iget() symbol export is missing. To avoid the need for
3447 this on patchless clients the deathrow inode reaper is turned
3448 off, and we depend on the VM to clean up old inodes. This
3449 dependency was during via the fix for bug 12181.
3451 --------------------------------------------------------------------------------
3453 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3455 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3456 INTEROPERATE with older versions automatically. In many cases a
3457 special upgrade step is needed. Please read the
3458 user documentation before upgrading any part of a live system.
3459 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3460 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3461 * WARNING: Lustre configuration and startup changes are required with
3462 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3466 Severity : enhancement
3468 Description: MountConf
3469 Details : Lustre configuration is now managed via mkfs and mount
3470 commands instead of lmc and lconf. New obd types (MGS, MGC)
3471 are added for dynamic configuration management. See
3472 https://mail.clusterfs.com/wikis/lustre/MountConf for
3475 Severity : enhancement
3477 Description: dynamic OST addition
3478 Details : OSTs can now be added to a live filesystem
3480 Severity : enhancement
3482 Description: startup order invariance
3483 Details : MDTs and OSTs can be started in any order. Clients only
3484 require the MDT to complete startup.
3486 Severity : enhancement
3488 Description: parallel, asynchronous orphan cleanup
3489 Details : orphan cleanup is now performed in separate threads for each
3490 OST, allowing parallel non-blocking operation.
3492 Severity : enhancement
3494 Description: optimized stripe assignment
3495 Details : stripe assignments are now made based on ost space available,
3496 ost previous usage, and OSS previous usage, in order to try
3497 to optimize storage space and networking resources.
3499 Severity : enhancement
3501 Description: Permanently set tunables
3502 Details : All writable /proc/fs/lustre tunables can now be permanently
3503 set on a per-server basis, at mkfs time or on a live system.
3505 Severity : enhancement
3507 Description: Lustre message v2
3508 Details : Add lustre message format v2.
3510 Severity : enhancement
3512 Description: client OST exclusion list
3513 Details : Clients can be started with a list of OSTs that should be
3514 declared "inactive" for known non-responsive OSTs.
3518 Description: SPEC SFS validation failure on NFS v2 over lustre.
3519 Details : Changes the blocksize for regular files to be 2x RPC size,
3520 and not depend on stripe size.
3522 Severity : enhancement
3524 Description: Multiple MD RPCs in flight.
3525 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3526 To avoid overly-overloading MDS, introduce a limit on number of
3527 MDS RPCs in flight for a single client and add /proc controls
3528 to adjust this limit.
3530 Severity : enhancement
3532 Description: client read/write statistics
3533 Details : Add client read/write call usage stats for performance
3534 analysis of user processes.
3535 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3536 file access. extents_stats shows chunk size distribution.
3537 extents_stats_per_process show chunk size distribution per
3540 Severity : enhancement
3542 Description: mds statistics
3543 Details : Add detailed mds operations statistics in
3544 /proc/fs/lustre/mds/*/stats.
3548 Description: Failure of copying files with lustre special EAs.
3549 Details : Client side always return success for setxattr call for lustre
3550 special xattr (currently only "trusted.lov").
3555 Description: Refcount LNET uuids
3556 Details : The global LNET uuid list grew linearly with every startup;
3557 refcount repeated list entries instead of always adding to
3560 Severity : enhancement
3562 Description: Dynamic service threads
3563 Details : Within a small range, start extra service threads
3564 automatically when the request queue builds up.
3567 Frequency : mixed-endian client/server environments
3569 Description: mixed-endian crashes
3570 Details : The new msg_v2 system had some failures in mixed-endian
3573 Severity : enhancement
3575 Description: Easy OST removal
3576 Details : OSTs can be permanently deactivated with e.g. 'lctl
3577 conf_param lustre-OST0001.osc.active=0'
3579 Severity : enhancement
3581 Description: MGS proc entries
3582 Details : Added basic proc entries for the MGS showing what filesystems
3585 Severity : enhancement
3587 Description: provide MGS failover
3588 Details : Added config lock reacquisition after MGS server failover.
3590 Severity : enhancement
3592 Description: add Linux 2.4 support
3593 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3597 Description: a large application tries to do I/O to the same resource and dies
3598 in the middle of it.
3599 Details : Check the req->rq_arrival time after the call to
3600 ost_brw_lock_get(), but before we do anything about
3601 processing it & sending the BULK transfer request. This
3602 should help move old stale pending locks off the queue as
3603 quickly as obd_timeout.
3606 Frequency : when an incorrect nid is specified during startup
3608 Description: ptlrpc connect to non-existant node causes kernel crash
3609 Details : LNET can't be re-entered from an event callback, which
3610 happened when we expire a message after the export has been
3611 cleaned up. Instead, hand the zombie cleanup off to another
3614 Severity : enhancement
3616 Description: plain/inodebits lock performance improvement
3617 Details : Grouping plain/inodebits in granted list by their request modes
3618 and bits policy, thus improving the performance of search through
3622 Frequency : only if OST filesystem is corrupted
3624 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3625 Details : for a short time RPCs with bulk IO are in the replay list,
3626 but replay of bulk IOs is unimplemented. If the OST filesystem
3627 is corrupted due to disk cache incoherency and then replay is
3628 started it is possible to trip an assertion. Avoid putting
3629 committed RPCs into the replay list at all to avoid this issue.
3632 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3634 Description: System hang on startup
3635 Details : This bug allowed the liblustre (e.g. catamount) client to
3636 return to the app before handling all startup RPCs. This
3637 could leave the node unresponsive to lustre network traffic
3638 and manifested as a server ptllnd timeout.
3640 Severity : enhancement
3642 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3643 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3644 boolean to control whether to print peer debug info when a
3645 client's RPC times out.
3648 Frequency : only for kernels with patches from Lustre below 1.4.3
3650 Description: Remove old rdonly API
3651 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3654 Frequency : only for devices with external journals
3656 Description: Set external device read-only also
3657 Details : During a commanded failover stop, we set the disk device
3658 read-only while the server shuts down. We now also set any
3659 external journal device read-only at the same time.
3662 Frequency : when upgrading from 1.4 while trying to change parameters
3664 Description: The wrong (new) MDC name was used when setting parameters for
3665 upgraded MDT's. Also allows changing of OSC (and MDC)
3666 parameters if --writeconf is specified at tunefs upgrade time.
3669 Frequency : when setting specific ost indicies
3671 Description: QOS code breaks on skipped indicies
3672 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3673 created with --index need not be sequential.
3675 Severity : enhancement
3677 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3678 Details : The uninit_groups feature works in conjunction with the kernel
3679 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3680 up the pass1 processing of e2fsck. This is a read-only feature
3681 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3682 mount filesystems that have had this feature enabled.
3684 Severity : enhancement
3686 Description: Improve multi-block allocation algorithm to avoid fragmentation
3687 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3688 allocation locality and avoid filesystem fragmentation.
3690 ------------------------------------------------------------------------------
3692 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3694 * Support for kernels:
3695 2.4.21-47.0.1.EL (RHEL 3)
3696 2.6.5-7.283 (SLES 9)
3697 2.6.9-42.0.10.EL (RHEL 4)
3698 2.6.12.6 vanilla (kernel.org)
3699 2.6.16.27-0.9 (SLES 10)
3700 * Recommended e2fsprogs version: 1.39.cfs5
3702 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3706 Frequency : occasional, depends on client load and configuration
3707 Bugzilla : 12181, 12203
3708 Description: data loss for recently-modified files
3710 Details : In some cases it is possible that recently written or created
3711 files may not be written to disk in a timely manner (this should
3712 normally be within 30s unless client IO load is very high).
3713 The problem appears as zero-length files or files that are a
3714 multiple of 1MB in size after a client crash or client eviction
3715 that are missing data at the end of the file.
3717 This problem is more likely to be hit on clients where files are
3718 repeatedly created and unlinked in the same directory, clients
3719 have a large amount of RAM, have many CPUs, the filesystem has
3720 many OSTs, the clients are rebooted frequently, and/or the files
3721 are not accessed by other nodes after being written.
3723 The presence of the problem can be detected by looking at
3724 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3725 smaller than the second (nr_unused) then dirty files will not
3726 be flushed automatically to disk. "sync; sleep 10" should be
3727 run several times on the node before unmounting it to update
3728 Lustre (this is also safe to run on nodes without this problem).
3730 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3731 that can cause this same problem, so customers using that kernel
3732 also need to update the kernel in addition to Lustre. In order
3733 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3735 It is normal that files written just before a client crash (less
3736 than 30s) may not yet have been flushed to disk, even for local
3740 Frequency : frequent on thin XT3 nodes
3742 Description: UUID collision on thin XT3 Linux nodes
3743 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3744 randomly, since we relied on an insufficiently-seeded PRNG.
3749 Description: OSS hangs after "All ost request buffers busy"
3750 Details : A deadlock between quota and journal operations caused OSS
3751 hangs after printing "All ost request buffers busy."
3754 Frequency : always on liblustre builds
3756 Description: Cleanup compiler warnings on liblustre
3759 Frequency : always on liblustre builds on XT3
3761 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3765 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3766 > 32000 subdirectories
3767 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3768 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3769 subdirectory count crosses 32000. This will aid e2fsck to
3770 correctly handle more than 32000 subdirectories.
3773 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3775 Description: System hang on startup
3776 Details : This bug allowed the liblustre (e.g. catamount) client to
3777 return to the app before handling all startup RPCs. This
3778 could leave the node unresponsive to lustre network traffic
3779 and manifested as a server ptllnd timeout.
3781 Severity : enhancement
3783 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3784 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3785 boolean to control whether to print peer debug info when a
3786 client's RPC times out.
3791 Description: make O_SYNC working on 2.6 kernels
3792 Details : 2.6 kernels use different method for mark pages for write,
3793 so need add a code to lustre for O_SYNC work.
3798 Description: Failure to close file and release space on NFS
3799 Details : Put inode details into lock acquired in ll_intent_file_open.
3800 Use mdc_intent_lock in ll_intent_open to properly
3801 detect all kind of errors unhandled by mdc_enqueue.
3806 Description: proc file read during shutdown sometimes raced obd removal,
3808 Details : Add lock to prevent obd access after proc file removal.
3811 Frequency : Only for files larger than 4GB on 32-bit clients.
3813 Description: improperly doing page alignment of locks
3814 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3815 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3820 Description: RPCs being resent when they shouldn't be
3821 Details : Some RPCs that should not be resent are being resent. This
3822 can cause inconsistencies in the RPC state machine. Do not
3823 resend such requests.
3826 Frequency : rare, only with NFS export
3828 Description: Crash on NFS re-export node
3829 Details : under very unusual load conditions an assertion is hit in
3830 ll_intent_file_open()
3833 Frequency : only if OST filesystem is corrupted
3835 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3836 Details : for a short time RPCs with bulk IO are in the replay list,
3837 but replay of bulk IOs is unimplemented. If the OST filesystem
3838 is corrupted due to disk cache incoherency and then replay is
3839 started it is possible to trip an assertion. Avoid putting
3840 committed RPCs into the replay list at all to avoid this issue.
3845 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3846 Details : Large single O_DIRECT read and write calls can fail to allocate
3847 a sufficiently large buffer to process the request. In case of
3848 allocation failure the allocation is retried with a smaller
3849 buffer and broken into smaller requests.
3851 Severity : enhancement
3853 Description: Add -o localflock option to simulate old noflock behaviour.
3854 Details : This will achieve local-only flock/fcntl locks coherentness.
3859 Description: versioning check is incomplete
3860 Details : Checking the version difference of client vs. server, report
3861 error if the gap is too big.
3866 Description: add support PG_writeback bit
3867 Details : add support for PG_writeback bit for Lustre, for more carefull
3868 work with page cache in 2.6 kernel. This also fix some deadlocks
3869 and remove hack for work O_SYNC with 2.6 kernel.
3871 Severity : enhancement
3873 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3874 Details : The uninit_groups feature works in conjunction with the kernel
3875 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3876 up the pass1 processing of e2fsck. This is a read-only feature
3877 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3878 mount filesystems that have had this feature enabled.
3880 Severity : enhancement
3882 Description: Improve multi-block allocation algorithm to avoid fragmentation
3883 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3884 allocation locality and avoid filesystem fragmentation.
3886 ------------------------------------------------------------------------------
3888 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3890 * Support for kernels:
3891 2.6.9-42.0.3.EL (RHEL 4)
3892 2.6.5-7.276 (SLES 9)
3893 2.4.21-47.0.1.EL (RHEL 3)
3894 2.6.12.6 vanilla (kernel.org)
3895 2.6.16.21-0.8 (SLES10)
3896 * Recommended e2fsprogs version: 1.39.cfs2-0
3898 * The backwards-compatible /proc/sys/portals symlink has been removed
3899 in this release. Before upgrading, please ensure that you change
3900 any configuration scripts or /etc/sysctl.conf files that access
3901 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3902 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3903 in advance of the upgrade on any system running Lustre 1.4.6 or
3904 newer, since /proc/sys/lnet was added in that version.
3905 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3909 Frequency : only when quota is used
3911 Description: avoid scanning export list for quota master
3912 Details : Change the algorithms to avoid scanning export list in order
3913 to improve the efficiency.
3916 Frequency : MDS failover only, very rarely
3918 Description: "went back in time" messages on mds failover
3919 Details : The greatest transno may be lost when the current operation
3920 finishes with an error (transno==0) and the client's last_rcvd
3921 record is over-written. Save the greatest transno in the
3922 mds_last_transno for this case.
3925 Frequency : always for specific kernels and striping counts
3927 Description: client may get "Matching packet too big" without ACL support
3928 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3929 when trying to access files in certain configurations. The
3930 clients should in fact be denied when mounting because they do
3931 not understand ACLs.
3934 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3936 Description: many clients connecting with IO in progress causes connect timeouts
3937 Details : Avoid synchronous journal commits to avoid delays caused by many
3938 clients connecting/disconnecting when bulk IO is in progress.
3939 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3940 OST_IO_PORTAL to avoid delays behind potentially many pending
3944 Frequency : occasionally with multiple writers to a single file
3946 Description: shared writes to file may result in wrong size reported by stat()
3947 Details : Allow growing of kms when extent lock is cancelled
3950 Frequency : always with random mmap IO to multi-striped file
3952 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3953 Details : Only if the hole is at the end of OST object so that kms is too
3954 small. Fix is to increase kms accordingly in ll_nopage.
3957 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3959 Description: writes to a missing object would leak memory on the OST
3960 Details : If there is an inconsistency between the MDS and OST filesystems,
3961 such that the MDS references an object that doesn't exist, writes
3962 to that object will leak memory due to incorrect cleanup in the
3963 error handling path, eventually running out of memory on the OST.
3968 Description: Creating too long symlink causes lustre errors
3969 Details : Check symlink and name lengths before sending requests to MDS.
3972 Frequency : only if flock is enabled (not on by default)
3974 Description: posix locks not released on fd closure on 2.6.9+
3975 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3976 kernels, this caused such locks not to be released on fd close and
3977 then assertions on fs unmount about still used locks.
3980 Frequency : MDS failover only, very rarely
3982 Description: clients may get ASSERTION(granted_lock != NULL)
3983 Details : When request was taking a long time, and a client was resending
3984 a getattr by name lock request. The were multiple lock requests
3985 with the same client lock handle and
3986 mds_getattr_name->fixup_handle_for_resent_request found one of the
3987 lock handles but later failed with ASSERTION(granted_lock != NULL).
3992 Description: handle->h_buffer_credits > 0, assertion failure
3993 Details : h_buffer_credits is zero after truncate, causing assertion
3994 failure. This patch extends the transaction or creates a new
3998 Frequency : NFS re-export or patchless client
3999 Bugzilla : 11179, 10796
4000 Description: Crash on NFS re-export node (__d_move)
4001 Details : We do not want to hash the dentry if we don't have a lock.
4002 But if this dentry is later used in d_move, we'd hit uninitialised
4003 list head d_hash, so we just do this to init d_hash field but
4004 leave dentry unhashed.
4007 Frequency : NFS re-export or patchless client
4009 Description: NFS exports has problem with symbolic link
4010 Details : lustre client didn't properly install dentry when re-exported
4011 to NFS or running patchless client.
4014 Frequency : NFS re-export or patchless client
4016 Description: Various nfs/patchless fixes.
4017 Details : fixes reuse disconected alias for lookup process - this fixes
4018 warning "find_exported_dentry: npd != pd",
4019 fix permission error with open files at nfs.
4020 fix apply umask when do revalidate.
4023 Frequency : occasional
4025 Description: Crash on NFS re-export node
4026 Details : calling clear_page() on the wrong pointer triggered oops in
4027 generic_mapping_read().
4030 Frequency : rarely, using O_DIRECT IO
4032 Description: unaligned directio crashes client with LASSERT
4033 Details : check for unaligned buffers before trying any requests.
4036 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4038 Description: stale data returned from RAID cache
4039 Details : If only a small amount of IO is done to the RAID device before
4040 reading it again it is possible to get stale data from the RAID
4041 cache instead of reading it from disk.
4044 Frequency : always for sles10 kernel
4046 Description: sles10 support
4047 Details : ll_follow_link: compile fixes and using of nd_set_link
4048 under newer kernels.
4051 Frequency : depends on arch, kernel and compiler version, always on sles10
4054 Description: recursive or deep enough symlinks cause stack overflow
4055 Details : getting rid of large stack-allocated variable in
4059 Frequency : depends on hardware
4061 Description: lustre write performance loss in the SLES10 kernel
4062 Details : the performance loss is caused by using of write barriers in the
4063 ext3 code. The SLES10 kernel turns barrier support on by
4064 default. The fix is to undo that change for ldiskfs.
4066 ------------------------------------------------------------------------------
4068 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4070 * Support for kernels:
4071 2.6.9-42.0.3EL (RHEL 4)
4072 2.6.5-7.276 (SLES 9)
4073 2.4.21-47.0.1.EL (RHEL 3)
4074 2.6.12.6 vanilla (kernel.org)
4078 Frequency : quota enabled and large files being deleted
4080 Description: releasing more than 4GB of quota at once hangs OST
4081 Details : If a user deletes more than 4GB of files on a single OST it
4082 will cause the OST to spin in an infinite loop. Release
4083 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4088 Description: statfs data retrieved from /proc may be stale or zero
4089 Details : When reading per-device statfs data from /proc, in the
4090 {kbytes,files}_{total,free,avail} files, it may appear
4091 as zero or be out of date.
4094 Frequency : systems with MD RAID1 external journal devices
4096 Description: lconf's call to blkid is confused by RAID1 journal devices
4097 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4098 of returning all block devices that match the journal UUID.
4101 Frequency : always, for aggregate stripe size over 4GB
4103 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4104 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4105 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4108 Frequency : always if "lfs find" used on a local file/directory
4110 Description: "lfs find" segfaults if used on a local file/directory
4111 Details : The case where a directory component was not specified wasn't
4112 handled correctly. Handle this properly.
4115 Frequency : always on ppc64
4117 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4118 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4119 perform endianness conversion.
4122 Frequency : rarely (truncate to non-zero file size after write under load)
4123 Bugzilla : 10730, 10687
4124 Description: Files padded with zeros to next 4K multiple
4125 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4126 it is possible that files that are truncated to a non-zero size
4127 immediately after being written are filled with zero bytes beyond
4128 the truncated size. No file data is lost.
4130 Severity : enhancement
4132 Description: Allow recovery/failover for liblustre clients.
4133 Details : liblustre clients were unaware of failover configurations until
4136 Severity : enhancement
4138 Description: user file locks should fail when not mounting with flock option
4139 Details : Set up an error-returning stub in ll_file_operations.lock field
4140 to prevent incorrect behaviour when client is mounted without
4141 flock option. Also, set up properly f_op->flock field for
4145 Frequency : always on ia64
4147 Description: "lfs df" loops on printing out MDS statfs information
4148 Details : The obd_ioctl_data was not initialized and in some systems
4149 this caused a failure during the ioctl that did not return
4150 an error. Initialize the struct and return an error on failure.
4153 Frequency : SLES 9 only
4155 Description: Error of copying files with lustre special EAs as root
4156 Details : Client side always return success for setxattr call for lustre
4157 special xattr (currently only "trusted.lov").
4160 Frequency : rarely on clusters with both ia64+i386 clients
4162 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4163 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4164 clients are doing concurrent writes to the same file it is
4165 possible that the ia64 clients may LASSERT because the OST
4166 extent locks are not PAGE_SIZE aligned. Ensure that grown
4167 locks are always aligned on the request boundary.
4170 Frequency : specific use, occasional
4172 Description: Overwriting in use executable truncates on-disk binary image
4173 Details : If one node attempts to overwrite an executable in use by
4174 another node, we now correctly return ETXTBSY instead of
4175 truncating the file.
4177 Severity : enhancement
4179 Description: Async OSC create to avoid the blocking unnecessarily.
4180 Details : If an OST has no remaining object, system will block on the creating
4181 when need to create a new object on this OST. Now, ways use
4182 pre-created objects when available, instead of blocking on an
4183 empty osc while others are not empty. If we must block, we block
4184 for the shortest possible period of time.
4189 Description: chmod on Lustre root is propagated to other clients
4190 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4191 invalid by the follow_mount time.
4194 Frequency : liblustre clients only
4196 Description: Race in 'instant cancel' lock handling could lead to such locks
4197 never to be granted in case of SMP MDS
4198 Details : Do not destroy not yet granted but cbpending locks in
4202 Frequency : replay/resend of open
4204 Description: non null lock assetion failure in mds_intent_policy
4205 Details : Trying to replay/resend lockless open requests resulted in
4206 mds_open() returning 0 with no lock. Now it sets a flag if
4207 a lock is going to be returned.
4209 Severity : enhancement
4211 Description: Checksum enhancements
4212 Details : New checksum enhancements allow for resending RPCs that failed
4215 Severity : enhancement
4217 Description: Tunables on number of dirty pages in cacche
4218 Details : Allow to set limit on number of dirty pages cached.
4223 Description: client crash on unmount - lock still has references
4224 Details : In some error handling cases it was possible to leak a lock
4225 reference on a client while accessing a file. This was not
4226 harmful to the client during operation, but would cause the
4227 client to crash when the filesystem is unmounted.
4230 Frequency : specific case, rare
4232 Description: ETXTBSY on mds though file not in use by client
4233 Details : ETXTBSY is no longer incorrectly returned when attempting to
4234 chmod or chown a directory that the user previously tried to
4235 execute or a currently-executing binary.
4238 Frequency : extremely rare except on liblustre-based clients
4240 Description: Lustre space not freed when files are deleted
4241 Details : Clean up open-unlinked files after client eviction. Previously
4242 the unlink was skipped and the files remained as orphans.
4247 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4248 Details : In some cases it was possible to send a blocking callback to
4249 a client doing a glimpse, even though that client didn't get
4250 a lock granted. When the glimpse lock is cancelled on the OST
4251 the freed lock is left on the waiting list and corrupted the list.
4254 Frequency : all core dumps
4256 Description: Broke core dumps to lustre
4257 Details : Negative dentry may be unhashed if parent does not have UPDATE
4258 lock, but some callers, e.g. do_coredump, expect dentry to be
4259 hashed after successful create, hash it in ll_create_it.
4261 ------------------------------------------------------------------------------
4263 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4265 * Support for kernels:
4266 2.6.9-42.0.2.EL (RHEL 4)
4267 2.6.5-7.276 (SLES 9)
4268 2.4.21-40.EL (RHEL 3)
4269 2.6.12.6 vanilla (kernel.org)
4273 Frequency : always on RHEL 3
4275 Description: Number of open files grows over time
4276 Details : The number of open files grows over time, whether or not
4277 Lustre is started. This was due to a filp leak introduced
4278 by one of our kernel patches.
4280 ------------------------------------------------------------------------------
4282 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4284 * Support for kernels:
4285 2.6.9-42.EL (RHEL 4)
4286 2.6.5-7.276 (SLES 9)
4287 2.4.21-40.EL (RHEL 3)
4288 2.6.12.6 vanilla (kernel.org)
4293 Bugzilla : 5719, 9635, 9792, 9684
4294 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4295 Details : If a server is under heavy load and cannot reply to new
4296 connection requests before the client resends the (re)connect,
4297 the connection handling code can behave badly if two service
4298 threads are concurrently handing separate (re)connections from
4299 the same client. Add better locking to the connection handling
4300 code, and ensure that only a single connection will be processed
4301 for a given client UUID, even if the lock is dropped.
4303 Severity : enhancement
4305 Description: add TCP zero-copy support to kernel
4306 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4307 sends if the hardware supports scatter-gather and checksumming.
4308 This allows socklnd to do client-write and server-read more
4309 efficiently and reduce CPU utilization from skbuf copying.
4312 Frequency : only if NFS exporting from client
4314 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4315 Details : If setting attributes on a file created under NFS that had
4316 never been opened it would be possible to oops the client
4317 if the file had no objects.
4320 Frequency : always for liblustre
4322 Description: liblustre client does MDS+OSTs setattr RPC for each write
4323 Details : When doing a write from a liblustre client, the client
4324 incorrectly issued an RPC to the MDS and each OST the file was
4325 striped over in order to update the timestamps. When writing
4326 with small chunks and many clients this could overwhelm the MDS
4327 with RPCs. In all cases it would slow down the write because
4328 these RPCs are unnecessary.
4330 Severity : enhancement
4332 Description: allow number of MDS service threads to be changed at module load
4333 Details : It is now possible to change the number of MDS service threads
4334 running. Adding "options mds mds_num_threads={N}" to the MDS's
4335 /etc/modprobe.conf will set the number of threads for the next
4336 time Lustre is restarted (assuming the "mds" module is also
4337 reloaded at that time). The default number of threads will
4338 stay the same, 32 for most systems.
4343 Description: OST crash if filesystem is unformatted or corrupt
4344 Details : If an OST is started on a device that has never been formatted
4345 or if the filesystem is corrupt and cannot even mount then the
4346 error handling cleanup routines would dereference a NULL pointer.
4351 Description: NULL pointer deref in llap_from_page.
4352 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4353 incorrect) mapping if the page was truncated/reclaimed while it was
4354 searched for. Check for this condition and skip such pages when
4355 doing readahead. Introduce extra check to llap_from_page() to
4356 verify page->mapping->host is non-NULL (so page is not anonymous).
4359 Frequency : Sometimes when using sys_sendfile
4361 Description: "page not covered by a lock" warnings from ll_readpage
4362 Details : sendfile called ll_readpage without right page locks present.
4363 Now we introduced ll_file_sendfile that does necessary locking
4364 around call to generic_file_sendfile() much like we do in
4368 Frequency : with certain MDS communication failures at client mount time
4370 Description: NULL pointer deref after failed client mount
4371 Details : a client connection request may delayed by the network layer
4372 and not be sent until after the PTLRPC layer has timed out the
4373 request. If the client fails the mount immediately it will try
4374 to clean up before the network times out the request. Add a
4375 reference from the request import to the obd device and delay
4376 the cleanup until the network drops the request.
4379 Frequency : occasionally during client (re)connect
4381 Description: assertion failure during client (re)connect
4382 Details : processing a client connection request may be delayed by the
4383 client or server longer than the client connect timeout. This
4384 causes the client to resend the connection request. If the
4385 original connection request is replied in this interval, the
4386 client may trip an assertion failure in ptlrpc_connect_interpret()
4387 which thought it would be the only running connect process.
4390 Frequency : only with obd_echo servers and clients that are rebooted
4392 Description: kernel BUG accessing uninitialized data structure
4393 Details : When running an obd_echo server it did not start the ping_evictor
4394 thread, and when a client was evicted an uninitialized data
4395 structure was accessed. Start the ping_evictor in the RPC
4396 service startup instead of the OBD startup.
4398 Severity : enhancement
4399 Bugzilla : 10193 (patchless)
4400 Description: Remove dependency on various unexported kernel interfaces.
4401 Details : No longer need reparent_to_init, exit_mm, exit_files,
4402 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4405 Frequency : rare (only users of deprecated and unsupported LDAP config)
4407 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4408 Details : LDAP apparently contains 'lustreName' attributes instead of
4409 'name'. A simple remapping of the name is sufficient.
4412 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4414 Description: waiting_locks_callback trips kernel BUG if client is evicted
4415 Details : Running with the dump_on_timeout debug flag turned on makes
4416 it possible that the waiting_locks_callback() can try to dump
4417 the Lustre kernel debug logs from an interrupt handler. Defer
4418 this log dumping to the expired_lock_main() thread.
4420 Severity : enhancement
4422 Description: Support NFS exporting on 2.6 kernels.
4423 Details : Implement non-rawops metadata methods for NFS server to use without
4424 changing NFS server code.
4427 Frequency : very rare (synthetic metadata workload only)
4429 Description: two racing renames might cause an MDS thread to deadlock
4430 Details : Running the "racer" program may cause one MDS thread to rename
4431 a file from being the source of a rename to being the target of
4432 a rename at exactly the same time that another thread is doing
4433 so, and the second thread has already enqueued these locks after
4434 doing a lookup of the target and is trying to relock them in
4435 order. Ensure that we don't try to re-lock the same resource.
4438 Frequency : only very large systems with liblustre clients
4440 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4441 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4442 all OSTs in parallel. This allows the eviction of stale liblustre
4443 clients to proceed much faster than if they were done in series,
4444 and also offers similar improvements for other set_info RPCs.
4449 Description: excessive CPU usage during initial read phase on client
4450 Details : During the initial read phase on a client, it would agressively
4451 retry readahead on the file, consuming too much CPU and impacting
4452 performance (since 1.4.5.8). Improve the readahead algorithm
4453 to avoid this, and also improve some other common cases (read
4454 of small files in particular, where "small" is files smaller than
4455 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4460 Description: MDS crash when receiving packet with unknown intent.
4461 Details : Do not LBUG in unknown intent case, just return -EFAULT
4463 Severity : enhancement
4464 Bugzilla : 9293, 9385
4465 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4466 Details : Do not serialize getattr (non-intent version) and statfs.
4469 Frequency : occasional, when OST network is overloaded/intermittent
4471 Description: client evicted by OST after bulk IO timeout
4472 Details : If a client sends a bulk IO request (read or write) the OST
4473 may evict the client if it is unresposive to its data GET/PUT
4474 request. This is incorrect if the network is overloaded (takes
4475 too long to transfer the RPC data) or dropped the OST GET/PUT
4476 request. There is no need to evict the client at all, since
4477 the pinger and/or lock callbacks will handle this, and the
4478 client can restart the bulk request.
4481 Frequency : Always when mmapping file with no objects
4483 Description: client crashes when mmapping file with no objects
4484 Details : Check that we actually have objects in a file before doing any
4485 operations on objects in ll_vm_open, ll_vm_close and
4491 Description: Request leak when working with deleted CWD
4492 Details : Introduce advanced request refcount tracking for requests
4493 referenced from lustre intent.
4495 Severity : Enhancement
4497 Description: Cache open file handles on client.
4498 Details : MDS now will return special lock along with openhandle, if
4499 requested and client is allowed to hold openhandle, even if unused,
4500 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4501 closing files for every read/write openration.
4503 Severity : Enhancement
4505 Description: Cache open negative dentries on client when possible.
4506 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4507 negative dentries on lock revocation.
4512 Description: Remounting a client read-only wasn't possible with a zconf mount
4513 Details : It wasn't possible to remount a client read-only with llmount.
4515 Severity : enhancement
4516 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4517 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4518 MPICH is included to add Lustre-specific ADIO interfaces.
4519 This is based closely on the UFS ADIO layer and only differs
4520 in file creation, in order to allow the OST striping to be set.
4521 This is user-contributed code and not supported by CFS.
4526 Description: extended inode attributes (immutable, append-only) work improperly
4527 when 2.4 and 2.6 kernels are used on client/server or vice versa
4528 Details : Introduce kernel-independent values for these flags.
4530 Severity : enhancement
4533 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4534 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4535 granularity. Now, Fractional MB granularity is be supported,
4536 this is very useful for low memory system.
4538 Severity : enhancement
4540 Description: Getattr by fid
4541 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4542 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4543 have locks and avoids getattr rpc requests for it on every stat.
4546 Frequency : Always, for filesystems larger than 2TB
4548 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4549 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4550 filesystems larger than 2TB to be mounted with mballoc enabled.
4553 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4555 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4556 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4557 filesystems larger than 2TB is mounted then the kernel will
4558 silently corrupt the start of the filesystem. CONFIG_LBD is
4559 enabled for all CFS-supported kernels, but the possibility of
4560 this happening with a modified kernel config exists.
4562 Severity : enhancement
4564 Description: add client O_DIRECT support for 2.6 kernels
4565 Details : It is now possible to do O_DIRECT reads and writes to files
4566 in the Lustre client mountpoint on 2.6 kernel clients.
4568 Severity : enhancement
4570 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4571 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4572 parallel, not waiting for response from every OST before sending
4573 a rpc to the next OST.
4578 Description: setattr vs write race when updating file timestamps
4579 Details : Client processes that update a file timestamp into the past
4580 right after writing to the file (e.g. tar) it is possible that
4581 the updated file modification time can be reset to the current
4582 time due to a race between processing the setattr and write RPC.
4584 Severity : enhancement
4586 Description: Bring 'lfs find' closer in line with regular Linux find.
4587 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4588 -print0 options and obtains all the needed info through the lustre
4591 Severity : enhancement
4593 Description: support up to 1024 configured devices on one node
4594 Details : change obd_dev array from statically allocated to dynamically
4595 allocated structs as they are first used to reduce memory usage
4600 Description: Flush dirty partially truncated pages during truncate
4601 Details : Immediatelly flush partially truncated pages in filter_setattr,
4602 this way we completely avoid having any pages in page cache on OST
4603 and can retire ugly workarounds during writes to flush such pages.
4608 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4609 Details : protect inode from truncation within vfs_unlink() context
4610 just take a reference before calling vfs_unlink() and release it
4611 when parent's i_sem is free.
4616 Description: last_id value checked outside lock on OST caused LASSERT failure
4617 Details : If there were multiple MDS->OST object precreate requests in
4618 flight, it was possible that the OST's last object id was checked
4619 outside a lock and incorrectly tripped an assertion. Move checks
4620 inside locks, and discard old precreate requests.
4623 Frequency : always, if extents are used on OSTs
4625 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4626 Details : OSTs using the extents format would not zero the high 16 bits of
4627 the index physical block number. This is not a problem for any
4628 OST filesystems smaller than 16TB, and no kernels support ext3
4629 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4630 new/modified files) and can be fixed for existing filesystems
4631 with e2fsprogs-1.39-cfs1.
4636 Description: import connection selection may be incorrect if timer wraps
4637 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4638 import connections to be ignored if the 32-bit jiffies counter
4639 wraps. Use a 64-bit jiffies counter.
4642 Frequency : very large clusters immediately after boot
4644 Description: LNET request buffers exhausted under heavy short-term load
4645 Details : If a large number of client requests are generated on a service
4646 that has previously never seen so many requests it is possible
4647 that the request buffer growth cannot keep up with the spike in
4648 demand. Instead of dropping incoming requests, they are held in
4649 the LND until the RPC service can accept more requests.
4652 Frequency : Sometimes during replay
4654 Description: Assertion failure in ll_local_open after replay.
4655 Details : If replay happened on an open request reply before we were able
4656 to set replay handler, reply will become not swabbed tripping the
4657 assertion in ll_local_open. Now we set the handler right after
4658 recognising of open request
4661 Frequency : very rare
4663 Description: kernel reports "badness in vsnprintf"
4664 Details : Reading from the "recovery_status" /proc file in small chunks
4665 may cause a negative length in lprocfs_obd_rd_recovery_status()
4666 call to vsnprintf() (which is otherwise harmless). Exit early
4667 if there is no more space in the output buffer.
4669 Severity : enhancement
4671 Description: clear OBD RPC statistics by writing to them
4672 Details : It is now possible to clear the OBD RPC statistics by writing
4673 to the "stats" file.
4678 Description: Client mtime is not the same on different clients after utimes
4679 Details : In some cases, the client was using the utimes() syscall on
4680 a file cached on another node. The clients now validate the
4681 ctime from the MDS + OSTs to determine which one is right.
4686 Description: Inability to activate failout mode
4687 Details : lconf script incorrectly assumed that in pythong string's numeric
4688 value is used in comparisons.
4691 Frequency : always with multiple stripes per file
4693 Description: Inefficient object allocation for mutli-stripe files
4694 Details : When selecting which OSTs to stripe files over, for files with
4695 a stripe count that divides evenly into the number of OSTs,
4696 the MDS is always picking the same starting OST for each file.
4697 Return the OST selection heuristic to the original design.
4702 Description: mount failures may take full timeout to return an error
4703 Details : Under some heavy load conditions it is possible that a
4704 failed mount can wait for the full obd_timeout interval,
4705 possibly several minutes, before reporting an error.
4706 Instead return an error as soon as the status is known.
4708 Frequency : quota enabled and large files being deleted
4710 Description: releasing more than 4GB of quota at once hangs OST
4711 Details : If a user deletes more than 4GB of files on a single OST it
4712 will cause the OST to spin in an infinite loop. Release
4713 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4718 Description: statfs data retrieved from /proc may be stale or zero
4719 Details : When reading per-device statfs data from /proc, in the
4720 {kbytes,files}_{total,free,avail} files, it may appear
4721 as zero or be out of date.
4724 Frequency : systems with MD RAID1 external journal devices
4726 Description: lconf's call to blkid is confused by RAID1 journal devices
4727 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4728 of returning all block devices that match the journal UUID.
4731 Frequency : always, for aggregate stripe size over 4GB
4733 Description: assertion fails when trying to use 4GB stripe size
4734 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4735 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4738 Frequency : always on ppc64
4740 Description: the first write on an ext3 filesystem with mballoc got stuck
4741 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4742 perform endianness conversion.
4744 ------------------------------------------------------------------------------
4746 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4748 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4749 INTEROPERATE with older versions automatically. Please read the
4750 user documentation before upgrading any part of a live system.
4751 * WARNING: Lustre networking configuration changes are required with
4752 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4755 * Support for kernels:
4756 2.6.9-22.0.2.EL (RHEL 4)
4757 2.6.5-7.244 (SLES 9)
4758 2.6.12.6 vanilla (kernel.org)
4761 Severity : enhancement
4762 Bugzilla : 7981/8208
4763 Description: Introduced Lustre Networking (LNET)
4764 Details : LNET is new networking infrastructure for Lustre, it includes
4765 a reorganized network configuration mode (see the user
4766 documentation for full details) as well as support for routing
4767 between different network fabrics. Lustre Networking Devices
4768 (LNDs) for the supported network fabrics have also been
4769 created for this new infrastructure.
4771 Severity : enhancement
4772 Description: Introduced Access control lists
4773 Details : clients can set ACLs on files and directories in order to have
4774 more fine-grained permissions than the standard Unix UGO+RWX.
4775 The MDS must be started with the "-o acl" mount option.
4777 Severity : enhancement
4778 Description: Introduced filesystem quotas
4779 Details : Administrators may now establish per-user quotas on the
4782 Severity : enhancement
4784 Description: Configuration change for the XT3
4785 The PTLLND is now used to run Lustre over Portals on the XT3
4786 The configure option(s) --with-cray-portals are no longer used.
4787 Rather --with-portals=<path-to-portals-includes> is used to
4788 enable building on the XT3. In addition to enable XT3 specific
4789 features the option --enable-cray-xt3 must be used.
4794 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4795 Details : A race between allocating a new llcd and re-getting the llcd_lock
4796 allowed another thread to grab newly-allocated llcd.
4798 Severity : enhancement
4800 Description: 2.6 OST async journal commit and locking fix to improve performance
4801 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4802 2.6 kernels are now async as they already were in 2.4 kernels so
4803 that they can commit concurrently with the network bulk transfer.
4804 For block-allocated files the filter allocation semaphore is held
4805 to avoid filesystem fragmentation during allocation. BKL lock
4806 removed for 2.6 xattr operations where it is no longer needed.
4811 Description: lconf incorrectly determined whether two IP networks could talk
4812 Details : In some more complicated routing and multiple-network
4813 configurations, lconf will avoid trying to make a network
4814 connection to a disjoint part of the IP space. It was doing the
4815 math incorrectly for one set of cases.
4820 Description: Fix for potential infinite loop processing records in an llog.
4821 Details : If an llog record is corrupted/zeroed, it is possible to loop
4822 forever in llog_process(). Validate the llog record length
4823 and skip the remainder of the block on error.
4826 Frequency : occasional (liblustre only)
4828 Description: liblustre could not open files whose last component is a symlink
4829 Details : sysio_path_walk() would incorrectly pass the open intent to
4830 intermediate path components.
4833 Frequency : rare (liblustre only with non-standard tuning)
4834 Bugzilla : 7201 (7350)
4835 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4836 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4837 after it has been released, only do this for referenced locks
4839 Severity : enhancement
4841 Description: specifying an (invalid) directory default stripe_size of -1
4842 would reset the directory default striping
4843 Details : stripe_size -1 was used internally to signal directory stripe
4844 removal, now use "all default" to signal dir stripe removal
4845 as a directory striping of "all default" is not useful
4848 Frequency : common for large clusters running liblustre clients
4850 Description: doing an ls when liblustre clients are running is slow
4851 Details : sending a glimpse AST to a liblustre client waits for every AST
4852 to time out, as liblustre clients will not respond. Since they
4853 cannot cache data we refresh the OST lock LVB from disk instead.
4855 Severity : enhancement
4857 Description: doing an ls at the same time as file IO can be slow
4858 Details : enqueue and other "small" requests can be blocked behind many
4859 large IO requests. Create a new OST IO portal for non-IO
4860 requests so they can be processed faster.
4863 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4865 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4866 created file with mode 000
4867 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4868 allow an ftruncate() as a non-root user to a file with mode 000.
4869 Lustre now respects this flag to disable mode checks when
4870 truncating a file owned by the user
4873 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4876 Description: Revoking locks from clients that went dead or catatonic might take
4878 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4879 cancellation of such locks instant on servers without waiting for
4880 any reply from clients. Clients drops these locks when cancel
4881 notification from server is received without replying.
4884 Frequency : liblustre-only, when liblustre client dies or becomes busy
4886 Description: Doing ls on Linux clients can take a long time with active
4888 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4889 granting such locks to it in the first place if possible. Locks
4890 are taken by proxy on the OST during the read or write and
4891 dropped immediately afterward. Add connect flags handling, do
4892 not grant locks to liblustre clients for glimpse ASTs.
4894 Severity : enhancement
4896 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4897 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4898 sequential reads and behaves badly with applications doing random
4899 reads. Improve it by reading ahead at least read region, and
4900 avoiding excessive large RPC for small reads.
4902 Severity : enhancement
4904 Description: Creating more than 1000 files for a single job may cause a load
4905 imbalance on the OSTs if there are also a large number of OSTs.
4906 Details : qos_prep_create() uses an OST index reseed value that is an
4907 even multiple of the number of available OSTs so that if the
4908 reseed happens in the middle of the object allocation it will
4909 still utilize the OSTs as uniformly as possible.
4914 Description: OST or MDS may oops in ping_evictor_main()
4915 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4916 but doesn't restart at beginning of obd_exports_timed list
4919 Severity : enhancement
4921 Description: improve by-nid export eviction on the MDS and OST
4922 Details : allow multiple exports with the same NID to be evicted at one
4923 time without re-searching the exports list.
4926 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4928 Description: MDS may oops in groups_free()
4929 Details : in rare race conditions a newly allocated group_info struct is
4930 freed again, and this can be NULL. The 2.4 compatibility code
4931 for groups_free() checked for a NULL pointer, but 2.6 did not.
4934 Frequency : common for liblustre clients doing little filesystem IO
4935 Bugzilla : 9352, 7313
4936 Description: server may evict liblustre clients accessing contended locks
4937 Details : if a client is granted a lock or receives a completion AST
4938 with a blocking AST already set it would not reply to the AST
4939 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4942 Frequency : lfs setstripe, only systems with more than 160 OSTs
4944 Description: unable to set striping with a starting offset beyond OST 160
4945 Details : llapi_create_file() incorrectly limited the starting stripe
4946 index to the maximum single-file stripe count.
4949 Frequency : LDAP users only
4951 Description: lconf did not handle in-kernel recovery with LDAP properly
4952 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4954 Severity : enhancement
4956 Description: bind OST threads to NUMA nodes to improve performance
4957 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4958 node and do their allocations there to localize memory access
4960 Severity : enhancement
4962 Description: llmount can determine client NID directly from Myrinet (GM)
4963 Details : the client NID code from gmnalnid was moved directly into
4964 llmount, removing the need to use this or specifying the
4965 client NID explicitly when mounting GM clients with zeroconf
4968 Frequency : if client is started with down MDS
4970 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4971 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4972 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4973 ptlrpc_import_delayed_req() because we didn't check if the
4974 request was interrupted, and we also didn't break out of the
4975 event loop if there was no timeout
4980 Description: data loss during non-page-aligned writes to a single file from
4981 both multiple nodes and multiple threads on one node at same time
4982 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4983 inconsistency led to false short-reads, that were cached and later
4984 used by ->prepare_write() to fill in partially written page,
4985 leading to data loss.
4988 Frequency : always, if lconf --abort_recovery used
4990 Description: lconf --abort_recovery fails with 'Operation not supported'
4991 Details : lconf was attempting to abort recovery on the MDT device and not
4994 Severity : enhancement
4996 Description: remove cleanup logs
4997 Details : replace lconf-generated cleanup logs with lustre internal
4998 cleanup routines. Eliminates the need for client-cleanup and
5001 Severity : enhancement
5003 Description: add support for EAs (user and system) on lustre filesystems
5004 Details : it is now possible to store extended attributes in the Lustre
5005 client filesystem, and with the user_xattr mount option it
5006 is possible to allow users to store EAs on their files also
5008 Severity : enhancement
5010 Description: Add possibility (config option) to show minimal available OST free
5012 Details : When compiled with --enable-mindf configure option, statfs(2)
5013 (and so, df) will return least minimal free space available from
5014 all OSTs as amount of free space on FS, instead of summary of
5015 free spaces of all OSTs.
5017 Severity : enhancement
5019 Description: do not expand extent locks acquired on OST-side
5020 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5021 by server: they are not cached anyway.
5024 Frequency : when mmap is used/binaries executed from Lustre
5026 Description: Unmmap pages before throwing them away from read cache.
5027 Details : llap_shrink cache now attempts to unmap pages before discarding
5028 them (if unmapping failed - do not discard). SLES9 kernel has
5029 extra checks that trigger if this unmapping is not done first.
5034 Description: lconf didn't resolve symlinks before checking to see whether a
5035 given mountpoint was already in use
5038 Frequency : when migrating failover services
5039 Bugzilla : 6395, 9514
5040 Description: When migrating a subset of services from a node (e.g. failback
5041 from a failover service node) the remaining services would
5042 time out and evict clients.
5043 Details : lconf --force (implied by --failover) sets the global obd_timeout
5044 to 5 seconds in order to quickly disconnect, but this caused
5045 other RPCs to time out too quickly. Do not change the global
5046 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5048 Severity : enhancement
5049 Frequency : if MDS is started with down OST
5050 Bugzilla : 9439,5706
5051 Description: Allow startup/shutdown of an MDS without depending on the
5052 availability of the OSTs.
5053 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5054 Add appropriate locking and lov-osc refcounts for safe
5055 cleaning. Add osc abort_inflight calls in case the
5056 synchronize never started.
5059 Frequency : occasional (Cray XT3 only)
5061 Description: root not authorized to access files in CRAY_PORTALS environment
5062 Details : The client process capabilities were not honoured on the MDS in
5063 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5064 been accessed by an authorized user then root was able to access
5065 the file on the local client also. The root user capabilities
5066 are now allowed on the MDS, as this environment has secure UID.
5069 Frequency : occasional
5071 Description: ldiskfs "too long searching" message happens too often
5072 Details : A debugging message (otherwise harmless) prints too often on
5073 the OST console. This has been reduced to only happen when
5074 there are fragmentation problems on the filesystem.
5079 Description: Division by zero in statfs when all OSCs are inactive
5080 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5081 lov_statfs divides by value returned by lov_get_stripecnt().
5085 Bugzilla : 9489, 3273
5086 Description: First write from each client to each OST was only 4kB in size,
5087 to initialize client writeback cache, which caused sub-optimal
5088 RPCs and poor layout on disk for the first written file.
5089 Details : Clients now request an initial cache grant at (re)connect time
5090 and so that they can start streaming writes to the cache right
5091 away and always do full-sized RPCs if there is enough data.
5092 If the OST is rebooted the client also re-establishes its grant
5093 so that client cached writes will be honoured under the grant.
5098 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5099 Details : Now I/O RPCs go to different portal number and (presumably) fast
5100 lock requests (and glimses) and other RPCs get their own service
5101 threads pool that should be able to service those RPCs
5104 Severity : enhancement
5106 Description: Ability to exchange lustre version between client and servers and
5107 issue warnings at client side if client is too old. Also for
5108 liblustre clients there is ability to refuse connection of too old
5110 Details : New 'version' field is added to connect data structure that is
5111 filled with version info. That info is later checked by server and
5115 Frequency : rare, liblustre only.
5116 Bugzilla : 9296, 9581
5117 Description: Two simultaneous writes from liblustre at offset within same page
5118 might proceed at the same time overwriting eachother with stale
5120 Details : I/O lock withing llu_file_prwv was released too early, before data
5121 actually was hitting the wire. Extended lock-holding time until
5122 server acknowledges receiving data.
5125 Frequency : extremely rare. Never observed in practice.
5127 Description: avoid generating lustre_handle cookie of 0.
5128 Details : class_handle_hash() generates handle cookies by incrementing
5129 global counter, and can hit 0 occasionaly (this is unlikely, but
5130 not impossible, because initial value of cookie counter is
5131 selected randonly). Value of 0 is used as a sentinel meaning
5132 "unassigned handle" --- avoid it. Also coalesce two critical
5133 sections in this function into one.
5135 Severity : enhancement
5137 Description: allow liblustre clients to delegate truncate locking to OST
5138 Details : To avoid overhead of locking, liblustre client instructs OST to
5139 take extent lock in ost_punch() on client's behalf. New connection
5140 flag is added to handle backward compatibility.
5142 Severity : enhancement
5143 Bugzilla : 4928, 7341, 9758
5144 Description: allow number of OST service threads to be specified
5145 Details : a module parameter allows the number of OST service threads
5146 to be specified via "options ost ost_num_threads={N}" in the
5147 OSS's /etc/modules.conf or /etc/modprobe.conf.
5151 Bugzilla : 6146, 9635, 9895
5152 Description: servers crash with bad pointer in target_handle_connect()
5153 Details : In rare cases when a client is reconnecting it was possible that
5154 the connection request was the last reference for that export.
5155 We would temporarily drop the export reference and get a new
5156 one, but this may have been the last reference and the export
5157 was just destroyed. Get new reference before dropping old one.
5159 Severity : enhancement
5160 Frequency : if client is started with failover MDS
5162 Description: Allow multiple MDS hostnames in the mount command
5163 Details : Try to read the configuration from all specified MDS
5164 hostnames during a client mount in case the "primary"
5167 Severity : enhancement
5169 Description: Stop sending data to evicted clients as soon as possible.
5170 Details : Check if the client we are about to send or are sending data to
5171 was evicted already. (Check is done every second of waiting,
5172 for which l_wait_event interface was extended to allow checking
5173 of exit condition at specified intervals).
5176 Frequency : rare, normally only when NFS exporting is done from client
5178 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5180 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5183 Frequency : always, if srand() is called before liblustre initialization
5185 Description: Liblustre uses system PRNG disturbing its usage by user application
5186 Details : Introduce internal to lustre fast and high-quality PRNG for
5187 lustre usage and make liblustre and some other places in generic
5188 lustre code to use it.
5190 Severity : enhancement
5191 Bugzilla : 9477, 9557, 9870
5192 Description: Verify that the MDS configuration logs are updated when xml is
5193 Details : Check if the .xml configuration logs are newer than the config
5194 logs stored on the MDS and report an error if this is the case.
5195 Request --write-conf, or allow starting with --old_conf.
5197 Severity : enhancement
5199 Description: Handle symlinks in the path when checking if Lustre is mounted.
5200 Details : Resolve intermediate symlinks when checking if a client has
5201 mounted a filesystem to avoid duplicate client mounts.
5206 Description: lconf can hit an error exception but still return success.
5207 Details : The lconf command catches the Command error exception at the top
5208 level script context and will exit with the associated exit
5209 status, but doesn't ensure that this exit status is non-zero.
5214 Description: failure of ptlrpc thread startup can cause oops
5215 Details : Starting a ptlrpc service thread can fail if there are a large
5216 number of threads or the server memory is very fragmented.
5217 Handle this without oopsing.
5220 Frequency : always, only if liblustre and non-default acceptor port was used
5222 Description: liblustre cannot connect to servers with non-default acceptor port
5223 Details : tcpnal_set_default_params() was not called and was therefore
5224 ignoring the environment varaible TCPNAL_PORT, as well as other
5225 TCPNAL_ environment variables
5230 Description: two objects could be created on the same OST for a single file
5231 Details : If an OST is down, in some cases it was possible to create two
5232 objects on a single OST for a single file. No problems other
5233 than potential performance impact and spurious error messages.
5237 Bugzilla : 5681, 9562
5238 Description: Client may oops in ll_unhash_aliases
5239 Details : Client dcache may become inconsistent in race condition.
5240 In some cases "getcwd" can fail if the current directory is
5246 Description: Inode refcounting problems in NFS export code
5247 Details : link_raw functions used to call d_instantiate without obtaining
5248 extra inode reference first.
5252 Bugzilla : 9942, 9903
5253 Description: Referencing freed requests leading to crash, memleaks with NFS.
5254 Details : We used to require that call to ll_revalidate_it was always
5255 followed by ll_lookup_it. Also with revalidate_special() it is
5256 possible to call ll_revalidate_it() twice for the same dentry
5257 even if first occurence returned success. This fix changes semantic
5258 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5259 reference on a request referred from the intent.
5260 ll_intent_release() then releases such a request.
5263 Frequency : rare, normally benchmark loads only
5265 Description: unlinked inodes were kept in memory on the client
5266 Details : If a client is repeatedly creating and unlinking files it
5267 can accumulate a lot of stale inodes in the inode slab cache.
5268 If there is no other client load running this can cause the
5269 client node to run out of memory. Instead flush old inodes
5270 from client cache that have the same inode number as a new inode.
5273 Frequency : SLES9 2.6.5 kernel and long filenames only
5274 Bugzilla : 9969, 10379
5275 Description: utime reports stale NFS file handle
5276 Details : SLES9 uses out-of-dentry names in some cases, which confused
5277 the lustre dentry revalidation. Change it to always use the
5281 Frequency : rare, unless heavy write-truncate concurrency is continuous
5282 Bugzilla : 4180, 6984, 7171, 9963, 9331
5283 Description: OST becomes very slow and/or deadlocked during object unlink
5284 Details : filter_destroy() was holding onto the parent directory lock
5285 while truncating+unlinking objects. For very large objects this
5286 may block other threads for a long time and slow overall OST
5287 responsiveness. It may also be possible to get a lock ordering
5288 deadlock in this case, or run out of journal credits because of
5289 the combined truncate+unlink. Solution is to do object truncate
5290 first in one transaction without parent lock, and then do the
5291 final unlink in a new transaction with the parent lock. This
5292 reduces the lock hold time dramatically.
5295 Frequency : rare, 2.4 kernels only
5297 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5298 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5299 before unmounting their filesystem in order to avoid deadlock.
5300 The kernel_locked() function in 2.4 kernels only checks whether
5301 the kernel lock is held, not whether it is this process that is
5302 holding it as 2.6 kernels do.
5307 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5308 Details : The client ptlrpc code may be trying to reconnect to a down
5309 server before a previous connection attempt has timed out.
5310 Increase the reconnect interval to be longer than the connection
5311 timeout interval to avoid sending duplicate connections to
5315 Frequency : echo_client brw_test command
5317 Description: fix echo_client to work with OST preallocated code
5318 Details : OST preallocation code (5137) didn't take echo_client IO path
5319 into account: echo_client calls filter methods outside of any
5320 OST thread and, hence, there is no per-thread preallocated
5321 pages and buffers to use. Solution: hijack pga pages for IO. As
5322 a byproduct, this avoids unnecessary data copying.
5326 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5327 Description: Client can oops in mdc_commit_close() after open replay
5328 Details : It was possible for the MDS to return an open request with no
5329 transaction number in mds_finish_transno() if the client was
5330 evicted, but without actually returning an error. Clients
5331 would later try to replay that open and may trip an assertion
5332 Simplify the client close codepath, and always return an error
5333 from the MDS in case the open is not successful.
5336 Frequency : rare, 2.6 OSTs only
5338 Description: OST may deadlock under high load on fragmented files
5339 Details : If there was a heavy load and highly-fragmented OST filesystems
5340 it was possible to have all the OST threads deadlock waiting on
5341 allocation of biovecs, because the biovecs were not released
5342 until the entire RPC IO was completed. Instead, release biovecs
5343 as soon as they are complete to ensure forward IO progress.
5345 Severity : enhancement
5347 Description: Support for specifying external journal device at mount
5348 Details : If an OST or MDS device is formatted with an external journal
5349 device, this device major/minor is stored in the ext3 superblock
5350 and may not be valid for failover. Allow detecting and
5351 specifying the external journal at mount time.
5356 Description: Mounting an MDS with pending unlinked files may cause oops
5357 Details : target_finish_recovery() calls mds_postrecov() which returned
5358 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5359 considers this an error and immediately begins cleaning up the
5360 lov, just after starting the mds_lov process
5362 Severity : enhancement
5364 Description: Implement 'lfs df' to report actual free space on per-OST basis
5365 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5366 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5367 usage of objects; '-h' to report in human readable format.
5369 ------------------------------------------------------------------------------
5371 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5378 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5379 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5380 at startup during orphan unlinking. Instead, skip buddy update
5381 before setup, as it will be regenerated after recovery is complete.
5386 Description: If an OST is inactive, its locks might reference stale inodes.
5387 Details : lov_change_cbdata() must iterate over all namespaces, even if
5388 they are inactive to clear inode references from the lock.
5390 Severity : enhancement
5391 Frequency : occasional, if non-standard max_dirty_mb used
5393 Description: Client will block write RPCs if not enough grant
5394 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5395 then the client will block writes while waiting for another RPC
5396 to complete instead of consuming its dirty limit. With change
5397 we get improved performance when max_dirty_mb is small.
5399 Severity : enhancement
5400 Bugzilla : 3389, 6253
5401 Description: Add support for supplementary groups on the MDS.
5402 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5403 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5404 lookups for user supplementary groups into a cache.
5408 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5409 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5410 for liblustre O_EXCL is enforced by MDS.
5414 Description: readdir never returns NULL in liblustre.
5415 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5416 dentry in struct dirent.
5420 Frequency : liblustre only, depends on application IO pattern
5421 Description: liblustre clients evicted if not contacting servers
5422 Details : Don't put liblustre clients into the ping_evictor list, so
5423 they will not be evicted by the pinger ever.
5425 Severity : enhancement
5427 Description: Add ability to evict clients by NID from MDS.
5428 Details : By echoing "nid:$NID" string into
5429 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5430 $NID would be instantly evicted from this MDS and from all active
5431 OSTs connected to it.
5435 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5436 Details : lookup_it_finish() used to query file size from OSTs that was not
5441 Description: service threads change working directory to that of init
5442 Details : Starting lustre service threads may pin the working directory
5443 of the parent thread, making that filesystem busy. Threads
5444 now change to the working directory of init to avoid this.
5448 Frequency : during shutdown only
5449 Description: shutdown with a failed MDS or OST can cause unmount to hang
5450 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5455 Frequency : 2.6 only
5456 Description: chmod/chown may include an extra supplementary group
5457 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5458 supplementary group and if none is specified this is used.
5461 Bugzilla : 5479 (6816)
5463 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5464 Details : If lookup is in progress on a file that is unlinked we might try
5465 to revalidate the inode and fail in revalidate after lookup is
5466 complete and ll_file_open() enqueues the open again but
5467 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5470 Frequency : always, if lconf --abort_recovery used
5472 Description: lconf --abort_recovery fails with 'Operation not supported'
5473 Details : lconf was attempting to abort recovery on the MDT device and not
5476 ------------------------------------------------------------------------------
5478 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5483 Frequency : rare (only unsupported configurations with a node running as an
5485 Bugzilla : 6514, 5137
5486 Description: Mounting a Lustre file system on a node running as an OST could
5488 Details : OSTs now preallocates memory needed to write out data at
5489 startup, instead of when needed, to avoid having to
5490 allocate memory in possibly low memory situations.
5491 Specifically, if the file system is mounted on on OST,
5492 memory pressure could force it to try to write out data,
5493 which it needed to allocate memory to do. Due to the low
5494 memory, it would be unable to do so and the node would
5495 become unresponsive.
5497 Severity : enhancement
5499 Description: Addition of lconf --service command line option
5500 Details : lconf now accepts a '--service <arg>' option, which is
5501 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5503 Severity : enhancement
5505 Description: Failover mode is now the default for OSTs.
5506 Details : By default, OSTs will now run in failover mode. To return to
5507 the old behaviour, add '--failout' to the lmc line for OSTs.
5509 Severity : enhancement
5511 Description: Health checks are now provided for MDS and OSTs
5512 Details : Additional detailed health check information on MSD and OSTs
5513 is now provided through the procfs health_check value.
5516 Frequency : occasional, depends on IO load
5518 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5519 after numerous create/delete cycles
5520 Details : The ext3 inode allocation policy would not allocate new inodes
5521 very well on the OSTs because there are no new directories
5522 being created. Instead we look for groups with free space if
5523 the parent directories are nearly full.
5528 Description: Network or server problems during mount may cause partially
5529 mounted clients instead of returning an error.
5530 Details : The config llog parsing code may overwrite the error return
5531 code during mount error handling, returning success instead
5537 Description: MDS can fail to allocate large reply buffers
5538 Details : After long uptimes the MDS can fail to allocate large reply
5539 buffers (e.g. zconf client mount config records) due to memory
5540 fragmentation or consumption by the buffer cache. Preallocate
5541 some large reply buffers so that these replies can be sent even
5542 under memory pressure.
5546 Frequency : rare (liblustre)
5547 Description: fsx running with liblustre complained that using truncate() to
5548 extend the file doesn't work. This patch corrects that issue.
5549 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5550 ATTR_SIZE and lsm use in llu_setattr_raw.
5554 Frequency : rare, only 2.6 kernels
5555 Description: Unusual file access patterns on the MDS may result in inode
5556 data being lost in very rare circumstances.
5557 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5558 code in the 2.6 kernel caused the inode and/or EA data not to
5559 be read from disk, causing single-file corruption.
5563 Frequency : rare, only 2.6 filesystems using extents
5564 Description: Heavy concurrent write and delete load may cause data corruption.
5565 Details : It was possible under high-load situations to have an extent
5566 metadata block in the block device cache from a just-unlinked
5567 file overwrite a newly-allocated data block. We now unmap any
5568 metadata buffers that alias just-allocated data blocks.
5572 Frequency : filesystems with default stripe_count larger than 77
5573 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5574 Details : lconf specifies an inode size of 4096 bytes when the default
5575 stripe_count is larger than 77. This conflicts with the default
5576 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5577 this case to avoid pinning too much memory for large EAs.
5579 ------------------------------------------------------------------------------
5581 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5586 Frequency : rare (extremely heavy IO load with hundreds of clients)
5588 Description: Client is evicted, gets IO error writing to file
5589 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5590 caused clients to be evicted for AST timeouts. The fixes in
5591 bug 5192 mean we no longer need to have such short AST timeouts
5592 so ldlm_timeout has been increased.
5595 Frequency : occasional during --force or --failover shutdown under load
5596 Bugzilla : 5949, 4834
5597 Description: Server oops/LBUG if stopped with --force or --failover under load
5598 Details : a collection of import/export refcount and cleanup ordering
5599 issues fixed for safer force cleanup
5602 Frequency : only filesystems larger than 120 OSTs
5603 Bugzilla : 5990, 6223
5604 Description: lfs getstripe would oops on a very large filesystem
5605 Details : lov_getconfig used kfree on vmalloc'd memory
5608 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5609 Bugzilla : 6242, 6243
5610 Description: reading from files that had been truncated to a non-zero size
5611 but never opened returned no data
5612 Details : ll_file_read() reads zeros from no-object files to EOF
5617 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5618 Details : There's little harm in aborting MDS/OSS recovery and letting it
5619 try again, so I removed the LASSERT and return an error instead.
5621 Severity : enhancement
5623 Description: New debugging infrastructure for tracking down data corruption
5624 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5625 (b) cover more of the client-side code path, and (c) try to narrow
5626 down where problems occurred
5630 Bugzilla : 3819, 4364, 4397, 6313
5631 Description: Racing close and eviction MDS could cause assertion in mds_close
5632 Details : It was possible to get multiple mfd references during close and
5633 client eviction, leading to one thread referencing a freed mfd.
5635 Severity: : enhancement
5636 Bugzilla : 3262, 6359
5637 Description: Attempts to reconnect to servers are now more aggressive.
5638 Details : This builds on the enhanced upcall-less recovery that was added
5639 in 1.4.2. When trying to reconnect to servers, clients will
5640 now try each server in the failover group every 10 seconds. By
5641 default, clients would previously try one server every 25 seconds.
5646 Description: After recovery, certain operations trigger a failed
5647 assertion on a client.
5648 Details : Failing over an mds, using lconf -d --failover, while a
5649 client was doing a readdir() call would cause the client to
5650 LBUG after recovery completed and the readdir() was resent.
5652 Severity : enhancement
5654 Description: Default groups are now added by lconf
5655 Details : You can now run lconf --group <servicename> without having to
5656 manually add groups with lmc.
5659 Frequency : occasional
5661 Description: Nodes with an elan id of 0 trigger a failed assertion
5664 Frequency : always when accessing e.g. tty/console device nodes
5666 Description: tty and some other devices nodes cannot be used on lustre
5667 Details : file's private_data field is used by device data and lustre
5668 values in there got lost. New field was added to struct file to
5669 store fs-specific private data.
5672 Frequency : when exporting Lustre via NFS
5674 Description: NFSD failed occasionally when looking up a path component
5675 Details : NFSD is looking up ".." which was broken in ext3 directories
5676 that had grown large enough to become hashed.
5679 Frequency : Clusters with multiple interfaces not on the same subnet
5681 Description: Nodes will repeatedly try to reconnect to an interface which it
5682 cannot reach and report an error to the log.
5683 Details : Extra peer list entries will be created by lconf with some peers
5684 unreachable. lconf now validates the peer before adding it.
5687 Frequency : Only if a default stripe is set on the filesystem root.
5689 Description: Setting a default stripe on the filesystem root prevented the
5690 filesystem from being remounted.
5691 Details : The client was sending extra request flags in the root getattr
5692 request and did not allocate a reply buffer for the dir EA.
5695 Frequency : occasional, higher if lots of files are accessed by one client
5696 Bugzilla : 6159, 6097
5697 Description: Client trips assertion regarding lsm mismatch/magic
5698 Details : While revalidating inodes the VFS looks up inodes with ifind()
5699 and in rare cases can find an inode that is being freed.
5700 The ll_test_inode() code will free the lsm during ifind()
5701 when it finds an existing inode and then the VFS later attaches
5702 this free lsm to a new inode.
5706 Bugzilla : 6422, 7030
5707 Description: MDS deadlock between mkdir and client eviction
5708 Details : Creating a new file via mkdir or mknod (starting a transaction
5709 and getting the ns lock) can deadlock with client eviction
5710 (gets ns lock and trying to finish a synchronous transaction).
5713 Frequency : occasional
5714 Description: While starting a server, the fsfilt_ext3 module could not be
5716 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5717 kernels. Previously, lconf would still use the ext3 name
5718 when trying to load modules. Now, it will correctly use
5719 ext3 on 2.4 and ldiskfs on 2.6.
5721 Severity : enhancement
5722 Description: The default stripe count has been changed to 1
5723 Details : The interpretation of the default stripe count (0, to lfs
5724 or lmc) has been changed to mean striping across a single
5725 OST, rather than all available. For general usage we have
5726 found a stripe count of 1 or 2 works best.
5728 Severity : enhancement
5729 Description: Add support for compiling against Cray portals.
5730 Details : Conditional compiling for some areas that are different
5734 Frequency : occasional
5735 Bugzilla : 6409, 6834
5736 Description: Creating files with an explicit stripe count may lead to
5737 a failed assertion on the MDS
5738 Details : If some OSTs are full or unavailable, creating files may
5739 trigger a failed assertion on the MDS. Now, Lustre will
5740 try to use other servers or return an error to the
5744 Frequency : occasional
5746 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5747 caused lock timeout during readahead (since 1.4.2).
5748 Details : Processes doing readahead might match a lock that hasn't been
5749 granted yet if there are overlapping and conflicting lock
5750 requests. The readahead process waits on ungranted lock
5751 (original lock is CBPENDING), while OST waits for that process
5752 to cancel CBPENDING read lock and eventually evicts client.
5754 Severity : enhancement
5756 Description: Initial enabling of flock support for clients
5757 Details : Implements fcntl advisory locking and file status functions.
5758 This feature is provided as an optional mount flag (default
5759 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5760 locking are implemented yet, and those that are are not guaranteed
5761 to be completely correct in production environments.
5762 mount -t lustre -o [flock|noflock] ...
5765 Frequency : occasional
5767 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5768 assertion in the ext3 JBD (journaling) layer.
5769 Details : The b_committed_data struct is protected by the big kernel lock
5770 in 2.4 kernels, serializing journal_commit_transaction() and
5771 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5772 access to this struct. In 2.6 kernels there is finer grained
5773 locking to improve SMP performance of the JBD layer.
5777 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5782 Description: merge and cleanup kernel patches.
5785 -----------------------------------------------------------------------------
5787 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5789 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5790 versions of Lustre. Please update all servers and clients to
5791 version 1.4.2 or later at the same time. You must also run
5792 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5793 to update the configuration logs.
5795 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5796 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5797 - swab llog records of type '0' so we get proper header size/idx (5861)
5798 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5799 - fix rename of one directory over another leaking an inode (5953)
5800 - avoid SetPageDirty on 2.6 (5981)
5801 - don't re-add just-being-destroyed locks to the waiting list (5653)
5802 - when creating new directories, inherit the parent's custom
5803 striping settings if present parent (3048)
5804 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5805 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5806 - don't hold client locks on temporary worklist from l_lru (5666)
5807 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5808 - automatically evict dead clients (5921)
5809 - Update file size properly in create+truncate+fstat case (6196)
5810 - Do not unhash mountpoint dentries, do not allow removal of
5812 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5813 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5814 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5815 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5816 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5817 - Correctly handle reads of files with no objects (6243)
5818 - lctl recover will also mark a device active if deactivate used (5933)
5820 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5821 - allow --write-conf on an MDS with different nettype than client (5619)
5822 - don't write config llogs to MDS for mounts not from that MDS (5617)
5823 - lconf should create multiple TCP connections from a client (5201)
5824 - init scripts are now turned off by default; run chkconfig --on
5825 lustre and chkconfig --on lustrefs to use them
5826 - upcalls are no longer needed for clients to recover to failover
5828 - add --abort-recovery option to lconf to abort recovery on device
5830 - add support for an arbitrary number of OSTs (3026)
5831 - Quota support protocol changes.
5832 - forward compatibility changes to wire structs (6007)
5833 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5834 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5835 - improved "lustre status" script
5836 - initialize blocksize for non-regular files (6062)
5837 - added --disable-server and --disable-client configure options (5782)
5838 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5839 - Vanilla 2.4.29 support
5840 - increase maximum number of obd devices to 520 (6242)
5841 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5842 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5843 SLES 9 2.6.5 kernels
5844 - limit stripes per file to 160 (the maximum EA size) (6093)
5846 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5849 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5850 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5851 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5852 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5853 - small ext3 extents cleanups and fixes (5733)
5854 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5855 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5856 - avoid needless client->OST connect, fix handle mismatch (5317)
5857 - fix DLM error path that led to out-of-sync client, long delays (5779)
5858 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5859 - fix several locking issues related to i_size (5492,5624,5654,5672)
5860 - don't move pending lock onto export if it is already evicted (5683)
5861 - fix kernel oops when creating .foo in unlinked directory (5548)
5862 - fix deadlock in obdfilter statistics vs. object create (5811)
5863 - use time_{before,after} to avoid timer jiffies wrap (5882)
5864 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5865 - Do not leak request if server was not able to process it (5154)
5866 - If mds_open unable to find parent dir, make that negative lookup(5154)
5867 - don't create new directories with extent-mapping (5909, 5936)
5869 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5870 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5871 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5872 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5873 - support for RHEL4 kernel on i686 (5773)
5874 - provide error messages when incompatible logs are encountered (5898)
5876 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5877 * version 1.4.0.10 (1.4.1 release candidate 1)
5879 - don't keep a lock reference when lock is not granted (4238)
5880 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5881 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5882 - drop import inflight refcount on signal_completed_replay error (5255)
5883 - unlock page after async write error during send (3677)
5884 - handle missing objects in filter_preprw_read properly (5265)
5885 - no transno return for symlink open, don't save no-trasno open (3440)
5886 - don't try to complete elan receive that already failed (4012)
5887 - free RPC server reply state on error (5406)
5888 - clean up thread from ptlrpc_start_thread() on error (5160)
5889 - readahead could read extra page into cache that wasn't ejected (5388)
5890 - prevent races in class_attach/setup/cleanup/detach (5260)
5891 - don't dereference de->d_inode after l_dput of de (5458)
5892 - use "int" for stripe value returned from lock_to_stripe (5544)
5893 - mballoc allocation and error-checking fixes in 2.6 (5504)
5894 - block device patches to fix I/O request sizes in 2.6 (5482)
5895 - look up hostnames for IB nals (5602)
5896 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5897 - don't start multiple acceptors for the same port (5277)
5898 - fix incorrect LASSERT in mds_getattr_name (5635)
5899 - export a proc file for general "ping" checking (5628)
5900 - fix "lfs check" to not block when the MDS is down (5628)
5902 - service request history (4965)
5903 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5904 - create an "evict_client" /proc entry on OSTs, like the MDS has
5905 - fix mount usage message, return errors per mount(8) (5168)
5906 - change grep [] to grep "[]" in tests so they work in more UMLs
5907 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5908 - remove ancient LOV_MAGIC_V0 EA support (5047)
5909 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5910 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5911 - allow readahead allocations to fail when low on memory (5383)
5912 - mmap locking landed again, after considerable improvement (2828)
5913 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5915 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5918 - send OST transaction number in read/write reply to free req (4966)
5919 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5920 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5921 - the watchdog thread now runs as interruptible (5246)
5922 - flock/lockf fixes (but it's still disabled, pending 5135)
5923 - don't use EXT3 constants in llite code (5094)
5924 - memory shortage at startup could cause assertion (5176)
5926 - reorganization of lov code
5927 - single portals codebase
5929 - add extents/mballoc support (5025)
5930 - direct I/O reads in the obdfilter (4048)
5931 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5933 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5936 - send OST transaction number in read/write reply to free req (4966)
5937 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5938 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5939 - the watchdog thread now runs as interruptible (5246)
5940 - handle missing objects in filter_preprw_read properly (5265)
5941 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5942 - drop import inflight refcount on signal_completed_replay error (5255)
5943 - unlock page after async write error during send (3677)
5944 - return original error code on reconstructed replies (3761)
5945 - no transno return for symlink open, don't save no-trasno open (3440)
5947 - add pid to ldlm debugging output (4922)
5948 - bump the watchdog timeouts -- we can't handle 30sec yet
5949 - extra debugging for orphan dentry/inode bug (5259)
5951 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5954 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5955 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5956 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5957 - fix NULL dereference in /proc/sys/portals/routes (4827)
5958 - allow failed mdc_close() operations to be interrupted (4561)
5959 - stop precreate on OST before MDS would time out on it (4778)
5960 - don't send partial-page writes before EOF from client (4410)
5961 - discard client grant for sub-page writes on large-page clients (4520)
5962 - don't free dentries not owned by NFS code, check generation (4806)
5963 - fix lsm leak if mds_create_objects() fails (4801)
5964 - limit debug_daemon file size, always print CERROR messages (4789)
5965 - use transno after validating reply (3892)
5966 - process timed out requests if import state changes (3754)
5967 - update mtime on OST during writes, return in glimpse (4829)
5968 - add mkfsoptions to LDAP (4679)
5969 - use ->max_readahead method instead of zapping global ra (5039)
5970 - don't interrupt __l_wait_event() during strace
5972 - add software watchdogs to catch hung threads quickly (4941)
5973 - make lustrefs init script start after nfs is mounted
5974 - fix CWARN/ERROR duplication (4930)
5975 - return async write errors to application if possible (2248)
5976 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5977 - print NAL number in %x format (4645)
5978 - update barely-supported suse-2.4.21-171 series (4842)
5979 - support for sles 9 %post scripts
5980 - support for building 2.6 kernel-source packages
5981 - support for sles km_* packages
5983 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5986 - ignore -ENOENT errors in osc_destroy (3639)
5987 - notify osc create thread that OSC is being cleaned up (4600)
5988 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5989 - reconstruct ost_handle() like mds_handle() (4657)
5990 - create a new thread to do import eviction to avoid deadlock (3969)
5991 - let lconf resolve symlinked-to devices (4629)
5992 - don't unlink "objects" from directory with default EA (4554)
5993 - hold socknal file ref over connect in case target is down (4394)
5994 - allow more than 32000 subdirectories in a single directory (3244)
5995 - fix blocks count for O_DIRECT writes (3751)
5996 - OST returns ENOSPC from object create when no space left (4539)
5997 - don't send truncate RPC if file size isn't changing (4410)
5998 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5999 - bind to privileged port in socknal and tcpnal (3689)
6001 - rate limit CERROR/CWARN console message to avoid overload (4519)
6002 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6003 - basic mmap support (3918)
6004 - kernel patch series update from b1_4 (4711)
6006 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6009 - avoid crash during MDS cleanup with OST shut down (2775)
6010 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6011 - don't use bad inodes on the MDS (3744)
6012 - dynamic object preallocation to improve recovery speed (4236)
6013 - don't hold spinlock over lock dumping or change debug flags (4401)
6014 - don't zero obd_dev when it is force cleaned (3651)
6015 - print grants to console if they go negative (4431)
6016 - "lctl deactivate" will stop automatic recovery attempts (3406)
6017 - look for existing locks in ldlm_handle_enqueue() (3764)
6018 - don't resolve lock handle twice in recovery avoiding race (4401)
6019 - revalidate should check working dir is a directory (4134)
6021 - don't always mark "slow" obdfilter messages as errors (4418)
6023 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6026 - don't close LustreDB during write_conf until it is done (3860)
6027 - fix typo in lconf for_each_profile (3821)
6028 - allow dumping logs from multiple threads at one time (3820)
6029 - don't allow multiple threads in OSC recovery (3812)
6030 - fix debug_size parameters (3864)
6031 - fix mds_postrecov to initialize import for llog ctxt (3121)
6032 - replace config semaphore with spinlock (3306)
6033 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6034 - don't allow enqueue to complete on a destroyed export (3822)
6035 - down write_lock before checking llog header bitmap (3825)
6036 - recover from lock replay timeout (3764)
6037 - up llog sem before sending rpc (3652)
6038 - reduce ns lock hold times when setting kms (3267)
6039 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6040 - fix NULL deref and obd_dev leak on setup error (3312)
6041 - replace some LBUG about llog ops with error handling (3841)
6042 - don't match INVALID dentries from d_lookup and spin (3784)
6043 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6044 - fix invalid assertion in ptlrpc_set_wait (3880)
6046 - add libwrap support for the TCP acceptor (3996)
6047 - add /proc/sys/portals/routes for non-root route listing (3994)
6048 - allow setting MDS UUID in .xml (2580)
6049 - print the stack of a process that LBUGs (4228)
6051 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6054 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6055 - make sure to unset replay flag from failed open requests (3440)
6056 - if default stripe count is 0, use OST count for inode size (3636)
6057 - update parent mtime/ctime on client for create/unlink (2611)
6058 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6059 - free recovery state on server during a forced cleanup (3571)
6060 - unregister_reply for resent reqs (3063)
6061 - loop back devices mounting and status check on 2.6 (3563)
6062 - fix resource-creation race that can provoke i_size == 0 (3513)
6063 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6064 - more debugging for page-accounting assertion (3746)
6065 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6066 - avoid deadlock after precreation failure (3758)
6067 - fix race and lock order deadlock in orphan handling (3450, 3750)
6068 - add validity checks when grabbing inodes from l_ast_data (3599)
6070 - add /proc/.../recovery_status to obdfilter (3428)
6071 - lightweight CDEBUG infrastructure, debug daemon (3668)
6072 - change default OSC RPC parameters to be better on small clusters
6073 - turn off OST read cache for files smaller than 32MB
6074 - install man pages and include them in rpms (3100)
6075 - add new init script for (un)mounting lustre filesystems (2593)
6076 - run chkconfig in %post for init scripts (3701)
6077 - drop scimac NAL (unmaintained)
6079 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6082 - clean kiobufs before and after use (3485)
6083 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6084 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6085 - change init script to fail more clearly if not run as root (1528)
6086 - allow clients to reconnect during replay (1742)
6087 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6088 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6089 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6090 - fix cond_resched (3554)
6091 - only evict unfinished clients after recovery (3515)
6092 - allow bulk resend, prevent data loss (3570)
6093 - dynamic ptlrpc request buffer allocation (2102)
6094 - don't allow unlinking open directory if it isn't empty (2904)
6095 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6096 - remove extraneous obd dereference causing LASSERT failure (3334)
6097 - don't use get_cycles() when creating temp. files on the mds (3156)
6098 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6099 - handle EEXIST for set-stripe, set proper directory name (3336)
6101 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6102 - fix ksocknal_fmb_callback() error messages (2918)
6104 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6107 - don't copy lvb into (possibly NULL) reply on error (2983)
6108 - don't deref dentry after dput, don't free lvb on error (2922)
6109 - use the kms to determine writeback rpc length (2947)
6110 - increment oti_logcookies when osc is inactive (2948)
6111 - update client's i_blocks count via lvb messages (2543)
6112 - handle intent open/close of special files properly (1557)
6113 - mount MDS with errors=remount-ro, like obdfilter (2009)
6114 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6115 - don't use cancelling-locks' kms values (2947)
6116 - use highest lock extent for kms, not last one (2925)
6117 - don't dereference ERR_PTR() dentry in error handling path (3107)
6118 - fix thread race in portals_debug_dumplog() (3122)
6119 - create lprocfs device entries at setup instead of at attach (1519)
6120 - common AST error handler, don't evict client on completion race (3145)
6121 - zero nameidata in detach_mnt in 2.6 (3118)
6122 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6123 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6124 - initialize RPC timeout timer earlier for 2.6 (3219)
6125 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6126 - print nal/nid for unknown nid (3258)
6127 - additional checks for oscc recovery before doing precreate (3284)
6128 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6129 - don't crash in mdc_close for bad permissions on open (3285)
6130 - zero i_rdev for non-device files (3147)
6131 - clear page->private before handing to FS, better assertion (3119)
6132 - tune the read pipeline (3236)
6133 - fix incorrect decref of invalidated dentry (2350)
6134 - provide read-ahead stats and refine rpc in flight stats (3328)
6135 - don't hold journal transaction open across create RPC (3313)
6136 - update atime on MDS at close time (3265)
6137 - close LDAP connection when recovering to avoid server load (3315)
6138 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6139 - don't leak open file on MDS after open resend (3325)
6140 - serialize filter_precreate and filter_destroy_precreated (3329)
6141 - loop device shouldn't call sync_dev() for nul device (3092)
6142 - clear page cache after eviction (2766)
6143 - resynchronize MDS->OST in background (2824)
6144 - refuse to mount the same filesystem twice on same mountpoint (3394)
6145 - allow llmount to create routes for mounting behind routers (3320)
6146 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6147 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6148 - fix rare problem with rename on htree directories (3417)
6150 - allow default OST striping configuration per directory (1414)
6151 - fix compilation for qswnal for 2.6 kernels (3125)
6152 - increase maximum number of MDS request buffers for large systems
6153 - change liblustreapi to be useful for external progs like lfsck (3098)
6154 - increase local configuration timeout for slow disks (3353)
6155 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6157 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6160 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6161 - don't overwrite extent policy data in reply if lock was blocked (2901)
6162 - drop filter export grants atomically with removal from device (2663)
6163 - del obd_self_export from work_list in class_disconnect_exports (2908)
6164 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6165 - swab reply message in mdc_close, other PPC fixes (2464)
6166 - fix destroying of named logs (2325)
6167 - overwrite old logs when running lconf --write_conf (2264)
6168 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6169 - fix race in target_handle_connect (2898)
6170 - mds_reint_create() should take same inode create lock (2926)
6171 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6172 - don't close files for self_export to avoid uninitialized obd (2936)
6173 - allow MDS with the same name as client node (2939)
6174 - hold dentry reference for closed log files for unlink (2325)
6175 - reserve space for all logs during transactions (2059)
6176 - don't evict page beyond end of stripe extent (2925)
6177 - don't oops on a deleted current working directory (2399)
6178 - handle hard links to targets without a parent properly (2517)
6179 - don't dereference NULL lock when racing during eviction (2867)
6180 - don't grow lock extents when lots of conflicting locks (2919)
6182 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6185 - account for cache space usage on clients to avoid data loss (974)
6186 - lfsck support in lustre kernel code (2349)
6187 - reduce journal credits needed for BRW writes (2370)
6188 - orphan handling to avoid losing space on client/server crashes
6189 - ptlrpcd can be blocked, stopping ALL progress (2477)
6190 - use lock value blocks to assist in proper KMS, faster stat (1021)
6191 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6192 - recovery for initial connections (2355)
6193 - fixes for mds_cleanup_orphans (1934)
6194 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6195 - block all file creations until orphan recovery completes (1901)
6196 - client remove rq_connection from request struct (2423)
6197 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6198 - recovery timer race (2670)
6199 - mdc_close recovey bug (2532)
6200 - ptlrpc cleanup bug (2710)
6201 - mds timeout on local locks (2588)
6202 - namespace lock held during RPCs (2431)
6203 - handle interrupted sync write properly (2503)
6204 - don't try to handle a message that hasn't been replied to (2699)
6205 - client assert failure during cleanup after abort recovery (2701)
6206 - leak mdc device after failed mount (2712)
6207 - ptlrpc_check_set allows timedout requests to complete (2714)
6208 - wait for inflight reqs when ptlrpcd finishes (2710)
6209 - make sure unregistered services are removed from the srv_list
6210 - reset bulk XID's when resending them (caught by 1138 test)
6211 - unregister_bulk after timeout
6212 - fix lconf error (2694)
6213 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6214 - readahead locks pages, leaves pending causing memory pressure (2673)
6215 - increase OST request buffers to 4096 on large machines (2729)
6216 - fix up permission of existing directories in simple_mkdir (2661)
6217 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6218 - don't assign transno to errored transactions (2742)
6219 - don't delete objects on OST if given a bogus objid from MDS (2751)
6220 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6221 - if rq_no_resend, then timeout request after recovery (2432)
6222 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6223 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6224 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6225 - MDS recovery shouldn't delete objects that it has given out (2730)
6226 - if enqueue arrives after completion, don't clobber LVB (2819)
6227 - don't unlock pages twice when trigger_group_io returns error (2814)
6228 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6229 - don't write pages to disk if there was an error (1450)
6230 - don't ping imports that have recovery disabled (2676)
6231 - take buffered bytes into account when balancing socknal conn (2817)
6232 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6233 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6234 - remove little-endian swabbing of llog records (1987)
6235 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6236 - retry reposting request buffers if they fail (1191)
6237 - grow extent at grant time to avoid granting a revoked lock (2809)
6238 - lock revoke doesn't evict page if covered by a second lock (2765)
6239 - disable VM readahead to avoid reading outside lock extents (2805)
6241 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6242 - updated kernel patches for hp-2.4.20 kernel (2681)
6244 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6247 - fix truncated write corruption (2366)
6248 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6250 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6251 - ensure len(uuid) < 37 in lmc (1171)
6252 - fix ia64 OOPS in llog_test (2255)
6253 - zero end of page at obdfilter for partial page writes (2648)
6254 - don't leave stale dentries around after renames (bug 2428)
6255 - fix timeouts when evicting a client with a single lock held (2642)
6256 - set deadline for the initial HELLO message to drain (2634)
6257 - print out dotted-quad IP addresses in the socknal (2302)
6259 - additional debugging for MDS client eviction problem (2443)
6260 - fix mkfsoptions support for osts (2603, 2604)
6262 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6265 - add series for the vanilla 2.6.0 kernel
6266 - add series for the vanilla 2.4.24 kernel
6267 - add series for a cray x86/64 UL kernel drop
6268 - fix xattr patches for the vanilla 2.4.19 series
6270 - generate true UUIDs in lmc (1171)
6271 - have portals stack dumping break in UML (2466)
6272 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6273 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6274 - wake callback waiting threads on client eviction (2460)
6275 - Add --ptldebug and --subsystem to lmc (1719)
6276 - update assertion to allow safe interrupt allocation
6277 - set rq_no_resend for cancel requests (2432)
6278 - recalculate ptlrpcd timeout after resend (2494)
6279 - call vfs_rmdir when removing pending directories (2368)
6280 - fix renaming a file to itself (2429)
6281 - lmc creates a default one-stripe lov (2454)
6282 - expand procfs space to handle large clusters (2326)
6283 - increase UML stack to avoid overflow
6284 - update lconf's list of debug and subsystem masks
6285 - fix lfs find --obd (2510)
6286 - /proc tunable for disabling filter read caching (2591)
6287 - stop rpm packages from altering slapd.conf (2301)
6288 - disable nagle in the socknal under 0conf (2578)
6289 - choose mds inode size based on stripe count (2572)
6290 - fix kernel-source rpm problems (2516)
6292 - add --disable-doc to avoid pdf generation (2421)
6293 - update documentation, tests, type-os, comments
6294 - avoid format warnings on ia64
6295 - remove the TOE NAL
6296 - tiny code cleanups by removing unused fields
6298 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6301 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6302 - remove the most blatant lies from BUILDING file (2371)
6303 - change default debug level to reasonable production setting
6304 - reduce client side cache size to reduce cache flush time
6305 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6306 - make TCP zerocopy and pinger support enabled by default (2476)
6307 - sync writes completed after process exits caused crashes (2319)
6308 - maintain correct mount count on the MDS (2356)
6309 - backout 1557, because 2316 wasn't really fixed
6310 - better file I/O statistics gathering in /proc
6311 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6312 - another kernel patch to fix zero-copy TCP function export
6313 - don't take duplicate lock when processing re-sent getattr (2420)
6314 - lctl uses obd_self_export instead of creating new conn (2353)
6315 - MDS/OST recovery case which requires object creation asserted (2425)
6316 - move lfs from /usr/sbin to /usr/bin in packages
6317 - fix race between mds_client_add and mds_client_free (2417)
6318 - use kmalloc instead of slabs in portals (2430)
6319 - don't create duplicate records when a failover MDS is present (2442)
6320 - remove unnecessary mount age check (2332)
6321 - don't remove directory inodes from locks prematurely (2451)
6322 - don't break if MDS service name is the same as hostname (2103)
6323 - fix races in client write RPC generation when cache full (2482)
6325 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6328 - remove now-unused request->rq_obd (278)
6329 - if an allocation fails, print out how much memory we've used (1933)
6330 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6331 - add the "configurable stack size" patch to most series files (1256)
6332 - ability to write large log records, for 100+ OST configs (2306)
6333 - fix NULL deref when filter_prep fails (2314)
6334 - fix operator precedence error in filter_sync
6335 - dynamic allocation of socknal TX descriptors (2315)
6336 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6337 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6338 - ensure that CWARN messages in Portals always get to the syslog
6339 - __init/__exit are not for prototype decls (ldlm_init/exit)
6340 - x86-64 compile warning fixes
6341 - fix gateway LMC keyword conflict (2318)
6342 - fix MDS lock inversions in getattr/reint paths (1844)
6343 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6344 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6345 - DLM race condition prevented some lock evictions (2328)
6346 - ENOMEM detection and retry on socknal sends (2230)
6347 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6348 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6349 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6350 - better error messages when a client is rejected during recovery (1505)
6351 - avoid cancelling locks which were never granted, after failure (2330)
6352 - fix i_sem/journal inversion in mds_client_add (2333)
6353 - fix truncate/getattr lock cycle deadlock (2334)
6354 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6355 - fix two rare exit paths which could leak an l_lock() ref (2321)
6356 - fencepost error in MDS/OST orphan recovery (2226)
6357 - make log record alignment 8 bytes (1988)
6358 - lstripe now fails when requested offset > ost_count (2237)
6359 - ensure that all kernel series have a complete list.h (1607)
6360 - fix crashes in special-file operations (2316)
6361 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6363 - allow configurable automake binary, for testing new versions
6364 - small update to the lfs documentation
6366 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6368 * fix negative export reference count in fsfilt_sync (2312)
6370 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6371 * release candidate 0.9.1
6373 - orphans are moved into the PENDING directory for possible recovery
6374 - replayed opens now open by fid for orphan/rename safety (1042)
6375 - last close of an orphan inode generates a transno (683)
6376 - chdir() and mount() now pin the directory entry (1020)
6377 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6378 - discard very old requests without processing them (1502)
6379 - remove some common, well-understood CERRORs (1505)
6380 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6381 - clear "grant" flags in OST replies until OST grant code lands (1644)
6382 - fix read performance by not clobbering i_blksize on client (1598)
6383 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6384 - make LRU size a /proc tunable, clears locks when reduced (707)
6385 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6386 - server locks take references on exports now (1558)
6387 - build fixes for 2.4.20-rh trees (1663)
6388 - return an error from lov_create if all OSCs are inactive (1751)
6389 - fix import levels when a reconnect happens without a timeout (1597)
6390 - exit early from mds_open if we get a lookup error (1749)
6391 - partial page read at EOF wouldn't wait for disk before sending (1642)
6392 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6393 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6394 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6395 - write/truncate lock inversion (1639)
6396 - Don't auto-load obdclass, portals modules during cleanup (1495)
6397 - fix timestamps from jumping to "now" (1763)
6398 - extra journal assertions (1648)
6399 - add an extra multiunlink test (1771)
6400 - fix read_record/write_record API (1776)
6401 - fix leak of offset_extent, possible incorrect i_size later (1772)
6402 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6403 - Debugging for the kqswnal_get_idle_tx problems (1820)
6404 - Allow recovery to be attempted multiple times (1536)
6405 - Write out MDS last_rcvd file after it is first created (1600)
6406 - Fix tx_descriptor leak in failed transmit situations (1827)
6407 - ext3 journaling fixes for assertion failure after IO error (1871)
6408 - class_export_put() on freed export after completion AST error (1896)
6409 - Fix revalidate looping in VFS (1322)
6410 - Don't access a freed export during MDS_REINT timeout (1521)
6411 - Add open-unlink recovery support on the MDS (1673,1764)
6412 - Return an error if no MDS data was read from last_rcvd (1946)
6413 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6414 - Don't setup a disk device that doesn't match exported UUID (317)
6415 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6416 - avoid committing NULL handle in force close
6417 - local.sh is now a one-stripe LOV configuration
6418 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6419 - fix ext3 htree duplicate directory entry corruption (1516)
6420 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6421 - update to new LOV EA format (2097)
6422 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6424 2003-06-15 Phil Schwan <phil@clusterfs.com>
6427 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6428 - per-import/export recovery handling (958, 931, 959)
6429 - multiple last-rcvd slots, for serving multiple FSes (949)
6430 - connections are again shared between multiple imp/exports (963, 964)
6431 - "umount -f" would hang if any requests needed to be sent (393, 978)
6432 - avoid pinning large req buffer by copying for queued messages (989)
6433 - add "uuid" to "lctl device" command to help upcalls (991)
6434 - "open" RPCs with transnos would confuse recovery counters (1037)
6435 - do proper endian conversion of all wire messages (288, 340, 891)
6436 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6437 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6438 - fix ptlrpc_connection leak in target_handle_connect (1174)
6439 - fix import refcounting bug in OST and MDS cleanup (1134)
6440 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6441 - fix very unlikely obd_types race condition (501)
6442 - remove osc_open hack for echo_client (1187)
6443 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6444 - a failure in read_inode2 leads to deadlock (1139)
6445 - cancel ack-locks as soon as transaction is committed (1072)
6446 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6447 - make sure to commitrw after any preprw to avoid deadlock (1162)
6448 - failing to execute a file in a lustre FS would lock inode (1203)
6449 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6450 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6451 - obd_uuid2tgt would walk off the end of the list (1255)
6452 - on IA64 the osc would give portals incorrect bulk size (1258)
6453 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6454 - fix lock inversion caused by new llite matching code (1282)
6455 - limit the number of dirty pages on a client to 10MB (1286)
6456 - timed out locks were not being corrected cancelled (1289)
6457 - fix O_DIRECT above 4GB on IA-32 (1292)
6458 * major user-visible changes
6459 - fail out/fail over policy now controlled by the upcall (993)
6461 - add OBD_PING to check server availability and failure (954)
6462 - lustre messages are now sent in sending host order (288, 340, 891)
6463 - add eadatalen to MDS getattr reply (340)
6464 - OST read replies may contain second buffer, with per-page status (593)
6466 2003-03-11 Phil Schwan <phil@clusterfs.com>
6469 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6470 - failed open()s could cause deadlock; fixed (867, 869)
6471 - stop cancelling OST locks when files are closed (481)
6472 - overlapping XID spaces caused network corruption (851, 853)
6473 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6474 - setattr_raw added, to do single-RPC, server-side setattrs
6475 - lmc/lconf syntax change for OST UUIDs
6476 - fix crashy race condition between ptlrpc_free_req and osc_close
6477 - don't use request in mdc_enqueue if we hit a timeout (889)
6478 - don't set the inode i_size for regular files from the MDS (896)
6479 - handle out of order completion AST (842)
6480 - don't LBUG if a lock request times out after receiving AST (913)
6481 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6482 - if a bad lock AST arrives, send an error instead of dropping entirely
6483 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6484 - fix leak in bulk IO when only partially completed (899, 900, 926)
6485 - fix O_DIRECT for ia64 (55)
6486 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6487 - C-z after timeout could hang a process forever; fixed (977)
6489 - client-side I/O cache (678, 924, 929, 941, 970)
6491 - READPAGE and SETATTRs which don't take server-side locks get
6494 2003-02-11 Phil Schwan <phil@clusterfs.com>
6497 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6498 - Fix fsfilt_extN_readpage() to read a full page of directory
6499 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6500 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6501 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6502 - Most intent operations were removed, in favour of a new RPC mode
6503 that does a single RPC to the server and bypasses most of the VFS
6504 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6505 - Aggressively cancel local locks on DLM servers
6506 - mds_reint_unlink sends EA to the client if it's the last nlink.
6507 client uses that EA to unlink OST objects.
6508 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6509 - recursive symlinks were fixed (439)
6510 - fixed NULL deref in DEBUG_REQ
6511 - filter_update_lastobjid no longer calls sync, which annoyed extN
6512 - fixed multi-client small-writes to a single file problem (445)
6513 - fixed mtime updates during file writes (607)
6514 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6515 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6516 - fixed Linux OST size reporting problem (444, 656)
6517 - OST now updates object mtime with writes or setattr (607, 619)
6518 - client verifies file size before zeroing page past EOF (445)
6519 - OST now writes last allocated objid to disk with allocation (108)
6520 - LOV on echo now works (409)
6522 - mds_reint_unlink sends a new buffer, with the EA included. this
6523 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6524 set if a regular file was being unlinked, and it was the last link
6525 - use PtlGet from the target for bulk writes (315)
6526 - OST now updates object mtime with writes or setattr (607, 619)
6527 - LDLM now has a grant-time callback to revalidate locked items, if
6529 - Many MDS operations were reorganized to combat race conditions
6531 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6532 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6534 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6537 - Fully reactivate OST imports after reconnection (512, others)
6538 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6539 - More graceful error handling for truncating on dead OST (515)
6540 - Don't error out unless we're actually accessing dead stripes (474)
6541 - Fix garbage sizes when stripes are missing (410)
6542 - LRU counters were broken, causing constant lock purge (433, 432)
6543 - garbage on read from stripes with failed OSTs (441)
6544 - mark OSCs as active before reconnecting during recovery (438)
6545 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6546 - lfind did not preserve OST order in output (443)
6547 - symlinks cause hung clients, incorrect data (439)
6548 - stop dereferencing request after dropping refcount (457)
6549 - don't LASSERT(spin_is_locked) on non-SMP (455)
6550 - fixes for many rename() bugs
6551 - fstat didn't correctly synchronize attributes (399)
6552 - server must handle lock cancellation during blocking AST prep (487)
6553 - bulk descriptors were free()d too soon (511)
6554 - fix paths in lconf, which would load incorrect modules (451, 507)
6555 - fix confusing lconf 'host not found' error message (386)
6556 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6557 - fix race condition in mdc_blocking_ast() for inode access (526)
6558 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6559 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6560 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6561 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6562 - fix lstripe in several ways (kernel side) (527)
6563 - fix request leak in ldlm_cli_enqueue (262)
6564 - incorrect OSC was marked inactive after OST failure
6565 - call mds_fs_cleanup before unmounting filesystem (524)
6566 - fix races between taking ns_lock and ldlm_lock_change_resource
6567 - fix races updating LOV export open file list
6568 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6569 - fix recovery NULL deref in ldlm_cli_cancel_unused
6570 - fix some DLM races by using new hash table for lock handles (419)
6571 - permit the client to specify desired inodes, at replay
6572 - duplicate requests when we queue them for replay reintegration
6573 - fix last_rcvd offset calculation
6574 - sync after each recovered transaction, so we always make progress
6575 - never, not always, ERESTART requests without transnos
6576 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6578 - skip replay if the MDS says that the client is already connected
6579 - don't check for a recovery-enabled export to match lctl's UUID
6580 - don't INC_USE_COUNT for phantom exports
6581 - don't crash when cleaning up phantom exports (567)
6582 - don't double-finish or set replay data for errored mdc_open requests
6583 - abort requests when they time out, so we don't get old replies
6584 - send/receive replies for AST messages again
6585 - if the client says that it doesn't have the lock, cancel it on the
6587 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6588 mark it as destroyed, it will all work out when decref is called
6589 - fix module use counts (22, 581)
6591 - ASTs now expect a reply (server cancels lock on error reply)
6593 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6596 - fix many simultaneous client startup (392)
6597 - fix dentry->d_it clobbering
6598 - credentials weren't being shipped for readdir/getattr operations
6599 - remove invalid assertions triggered during some concurrent MD
6601 - proper Lustre versions added (336, 389)
6602 - fix memory leak for create error case (398)
6603 - fix LOV locking bug that would get cli/srv out of sync
6604 - fix echo client over LOV (409)
6605 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6606 - fix double-O_EXCL intent crash (424)
6607 - avoid sending multiple lock CANCELs (352)
6609 - MDS can do multi-client recovery (modulo bugs in new code)
6611 - many updates, edits, cleanups
6613 2002-11-18 Phil Schwan <phil@clusterfs.com>
6616 - fix null d_it dereference (346)
6617 - fix full OST/dbench hang (333)
6618 - fix permission problem with file removal (286)
6619 - fix removal of OSCs from LOV when they fail
6620 - fix NULL deref during bulk timeout (214)
6621 - fix problems related to multiple filesystems on one MDS (241)
6622 - fixed serious subtle metadata locking bugs
6623 - free locks on clients when inodes are removed due to memory
6625 - fix inode pointer in lock data (285)
6626 - partial support for multiple MDS on a single host (241)
6627 - data locks weren't cancelled at clear_inode time (290, 311)
6628 - intent locks could lead to unbounded lock growth (205)
6629 - added a maximum lock count, an LRU list, and a flusher
6630 - fix multiple rename (365)
6631 - properly abstracted the echo client
6632 - OSC locked 1 byte too many; fixed
6633 - rewrote brw callback code:
6634 - fixed recovery bugs related to LOVs (306)
6635 - fixed too-many-pages-in-one-write crash (191)
6636 - fixed (again) crash in sync_io_timeout (214)
6637 - probably fixed callback-related race (385)
6639 - Add capability to MDS protocol
6640 - LDLM cancellations and callbacks on different portals
6642 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6645 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6646 - defer open object destruction to close time (601981, 138)
6647 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6648 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6649 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6650 - don't call obd_setattr in ll_file_release for destroyed objects
6651 * protocol change to lustre_msg: move |version| and add |flags|
6652 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6653 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6654 * added replay of create, unlink, link and rename operations during
6655 MDS failover; recovery should be much more robust now
6656 * remove failed OSCs from LOVs (only lov_create uses this so far)
6657 * the lustre-HOWTO was brought (more) up to date (582544)
6659 2002-10-23 Phil Schwan <phil@clusterfs.com>
6662 - in-use dentries weren't being reused properly (617851)
6663 - prevent multiple LDLM setup (599178)
6664 - fix LOV size calculations for truncate (617853)
6665 - fix client handling of MDS intent errors (POSIX)
6666 - fix permission bug in lovstripe.c test (624321)
6667 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6668 - truncate past end of file could corrupt data
6669 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6670 - a race in recovery could return ETIMEDOUT to apps (623947)
6671 - building outside the source directory was fixed
6672 * the lustre-HOWTO was brought (more) up to date (582544)
6673 * major progress was made on recovery functionality
6675 2002-10-10 Phil Schwan <phil@clusterfs.com>
6678 - recovery deadlock fix
6679 - rm -rf causes LBUG fix (617817)
6680 - file open by multiple tasks fix (618962)
6681 - directory permissions bugs (602707 and 620007)
6682 - journal_stop fixed with locking (611313)
6683 - O_APPEND failures resolved (618273, perhaps 614459)
6684 - lconf PATH fix (619770)
6685 - IA64 build fix (621450)
6686 - RPC buffer sizes scale with amount of memory
6688 2002-10-01 Phil Schwan <phil@clusterfs.com>
6691 - locks would be cancelled without throwing away data pages,
6692 resulting in inconsistent data (605627)
6693 - inode attributes were not always being refreshed (605627, 612449)
6694 - lconf now continues to cleanup after lctl reports an error
6695 - MDS now enforces user permissions (602707)
6696 - lprocfs cleanup fixed, but not yet enabled (614157)
6697 - fixed infinite server hang, should a client not respond to an AST
6698 - avoid going into recovery if user calls readlink() with a buffer
6699 that's too small (613941)
6700 - AST RPCs no longer require replies (614867) -- this may be changed
6701 - don't crash server if client sends an IOV that's too big (611336)
6702 - fixed lock conversion deadlock (611892)
6703 - fixed the following of symlinks (614622)
6704 * recovery: the server can remove locks from a client that dies, other
6705 clients can make progress
6706 * more extN patch fixes
6707 * compile-time configurable ptlrpc buffer allocations
6709 - collaborative read cache document
6710 - Lustre Lite Performance CDR document-in-progress
6712 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6715 - fix typo in patch-2.4.18
6717 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6720 - clear ptlrpc request each time in handle_incoming_request()
6721 - unlink of files now destroys the object on the OST
6723 2002-09-19 Peter Braam <braam@clusterfs.com>
6725 * add hard link support
6726 * change obdfile creation method
6727 * kernel patch changed
6729 2002-09-19 Peter Braam <braam@clusterfs.com>
6732 - stack overflow bug in extN fixed
6734 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6736 * documentation updates
6737 - add man pages for config tools
6738 - update tests/README to describe testing with new config tools
6739 - finish metadata API descriptions
6740 * bug fixes and cleanups
6741 - statfs workaround for 16TB limit
6742 - LOV stripe allocation improved, can stripe on subset of OSTs
6743 - LOV file size/IO offset was wrong for files > 4GB in size
6744 - object EA data was being dropped, caused files to be unreadable
6745 - memory overflow with non-LOV OST caused memory corruption
6746 - fixed regression tests to work with new config tools, obdfilter
6747 - fixed bug when directory size became larger than 1 block
6748 - fixed bug (for single client case) when PWD was deleted
6749 - invalidate local directory pages when doing intent-based ops
6750 - avoid LDLM oops when lock callback contained bad data
6752 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6754 * documentation updates
6755 * bug fixes and cleanups
6756 - configuration tools
6759 - 64-bit compile warnings
6760 - 64-bit internal statfs data
6762 * test_brw on persistent OST devices
6764 * lprocfs (disabled)
6766 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6768 * documentation updates
6769 * bug fixes and cleanups
6770 * configuration tools
6772 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6775 * many small fixes to 0_5_4
6776 * io/network handling
6777 * thinkos in MDS operations
6779 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6782 * crucial basic fixes to 0.5.3
6783 * IOR, Iozone work over Elan
6786 2002-08-07 Phil Schwan <phil@clusterfs.com>
6787 * version 0_5_3, our first alpha
6788 * we use the new Portals iovs
6789 * documentation updates
6790 * bug fixes and cleanups
6791 * small changes in the DLM wire protocol
6793 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6794 * version 0_5_1 with some initial stability,
6795 * locking on MD and file I/O.
6796 * documentation updates
6797 * several bug fixes since 0.5.0
6798 * small changes in wire protocol
6800 2002-07-18 Phil Schwan <phil@clusterfs.com>
6802 * delivered as Lustre Light Alpha
6803 * fixed a crash after handling invalid MDS requests
6804 * fixed directory pages for architectures with non-4k pages sizes
6806 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6807 * release version v0_4_4
6808 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6810 * Fixes a number of interruption problems with OST operations.
6811 * Update documentation for portals header changes
6812 * Move all wire protocol structs/defines to lustre_idl.h
6813 * Fixes symlink length bug.
6814 * Add tcpdump to repository.
6816 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6817 * release version v0_4_3
6818 * Fixes statfs for inodes on extN.
6819 * Fixes bug in runtests which would delete /etc/hosts.
6820 * Use 64-bit object IDs wherever possible (not into VFS though)
6821 Remove ost_get_info, which is unused by lustre, and out of date.
6823 2002-07-03 Peter Braam <braam@clusterfs.com>
6824 * release version v0_4_2 Fixes a lookup error (type not passed)
6825 * move forward to head of Portals
6826 * move forward to latest Lustre kernel
6828 2002-06-25 Peter Braam <braam@clusterfs.com>
6829 * release version v0_4_1. Hopefully stable on single node use.