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 3.10.0-1160.49.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.22.1.el8 (RHEL8.3)
16 4.18.0-305.25.1.el8 (RHEL8.4)
17 4.18.0-348.2.1.el8 (RHEL8.5)
18 4.4.120-92.70 (SLES12 SP2)
19 4.4.180-94.100 (SLES12 SP3)
20 4.4.0-131 (Ubuntu 16.04)
21 4.15.0-32 (Ubuntu 18.04)
22 5.4.0-48 (Ubuntu 20.04)
23 vanilla linux 5.4.0 (ZFS + ldiskfs)
24 vanilla linux 5.4.21 (ZFS + ldiskfs)
25 vanilla linux 5.4.136 (ZFS + ldiskfs)
26 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
27 * Client primary kernels built and tested during release cycle:
28 3.10.0-1062.18.1.el7 (RHEL7.7)
29 3.10.0-1127.19.1.el7 (RHEL7.8)
30 3.10.0-1160.49.1.el7 (RHEL7.9)
31 4.12.14-95.48 (SLES12 SP4)
32 4.12.14-122.91 (SLES12 SP5)
33 4.12.14-197.75 (SLES15 SP1)
34 5.3.18-24.96 (SLES15 SP2)
35 5.3.18-59.27 (SLES15 SP3)
36 4.15.0-48 (Ubuntu 18.04)
37 * Other clients known to build on these kernels at some point (others may also work):
38 3.10.0-862.14.4.el7 (RHEL7.5)
39 3.10.0-957.27.2.el7 (RHEL7.6)
40 4.14.0-49.13.1.el7a (RHEL7.5)
41 4.14.0-115.2.2.el7a (RHEL7.6)
42 4.18.0-80.11.2.el8 (RHEL8.0)
43 4.18.0-147.8.1.el8 (RHEL8.1)
44 4.18.0-193.28.1.el8 (RHEL8.2)
45 4.18.0-240.22.1.el8 (RHEL8.3)
46 4.18.0-305.25.1.el8 (RHEL8.4)
47 4.18.0-348.2.1.el8 (RHEL8.5)
48 4.4.120-92.70 (SLES12 SP2)
49 4.4.180-94.100 (SLES12 SP3)
50 4.4.0-131 (Ubuntu 16.04)
51 5.4.0-37 (Ubuntu 20.04)
52 5.8.0-53 (Ubuntu 20.04.2 HWE)
53 5.11.0-31 (Ubuntu 20.04.3 HWE)
54 5.11.0 (vanilla kernel.org)
55 * Recommended e2fsprogs version: 1.46.2.wc1 or newer
56 * Recommended ZFS version: 2.0.6
57 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
58 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
59 stack overflow. For more information, please refer to bugzilla 17630.
60 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
61 the re-exporting nodes to be mounted with "32bitapi" mount option
63 --------------------------------------------------------------------------------
67 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
68 for currently supported client and server kernel versions.
69 * Server primary kernels built and tested during release cycle:
70 3.10.0-1062.18.1.el7 (RHEL7.7)
71 3.10.0-1127.19.1.el7 (RHEL7.8)
72 3.10.0-1160.6.1.el7 (RHEL7.9)
73 * Other server kernels known to build and work at some point (others may also work):
74 3.10.0-862.14.4.el7 (RHEL7.5)
75 3.10.0-957.27.2.el7 (RHEL7.6)
76 4.18.0-80.11.2.el8 (RHEL8.0)
77 4.18.0-147.8.1.el8 (RHEL8.1)
78 4.18.0-193.28.1.el8 (RHEL8.2)
79 4.18.0-240.1.1.el8 (RHEL8.3)
80 4.4.120-92.70 (SLES12 SP2)
81 4.4.180-94.100 (SLES12 SP3)
82 4.4.0-131 (Ubuntu 16.04)
83 4.15.0-32 (Ubuntu 18.04)
84 5.4.0-48 (Ubuntu 20.04)
85 vanilla linux 5.4.0 (ZFS + ldiskfs)
86 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
87 * Client primary kernels built and tested during release cycle:
88 3.10.0-1062.18.1.el7 (RHEL7.7)
89 3.10.0-1127.19.1.el7 (RHEL7.8)
90 3.10.0-1160.6.1.el7 (RHEL7.9)
91 4.12.14-95.48 (SLES12 SP4)
92 4.12.14-122.20 (SLES12 SP5)
93 4.12.14-197.75 (SLES15 SP1)
94 5.3.18-24.24 (SLES15 SP2)
95 4.15.0-48 (Ubuntu 18.04)
96 * Other clients known to build on these kernels at some point (others may also work):
97 3.10.0-862.14.4.el7 (RHEL7.5)
98 3.10.0-957.27.2.el7 (RHEL7.6)
99 4.14.0-49.13.1.el7a (RHEL7.5)
100 4.14.0-115.2.2.el7a (RHEL7.6)
101 4.18.0-80.11.2.el8 (RHEL8.0)
102 4.18.0-147.8.1.el8 (RHEL8.1)
103 4.18.0-193.28.1.el8 (RHEL8.2)
104 4.18.0-240.1.1.el8 (RHEL8.3)
105 4.4.120-92.70 (SLES12 SP2)
106 4.4.180-94.100 (SLES12 SP3)
107 4.4.0-131 (Ubuntu 16.04)
108 5.4.0-37 (Ubuntu 20.04)
109 5.4.0 (vanilla kernel.org)
110 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
111 * Recommended ZFS version: 2.0.0
112 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
113 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
114 stack overflow. For more information, please refer to bugzilla 17630.
115 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
116 the re-exporting nodes to be mounted with "32bitapi" mount option
118 --------------------------------------------------------------------------------
122 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
123 for currently supported client and server kernel versions.
124 * Server primary kernels built and tested during release cycle:
125 3.10.0-1062.1.1.el7 (RHEL7.7)
126 * Other server kernels known to build and work at some point (others may also work):
127 3.10.0-862.14.4.el7 (RHEL7.5)
128 3.10.0-957.27.2.el7 (RHEL7.6)
129 4.18.0-80.7.1.el8 (RHEL8.0)
130 4.4.120-92.70 (SLES12 SP2)
131 4.4.180-94.100 (SLES12 SP3)
132 4.4.0-131 (Ubuntu 16.04)
133 4.15.0-32 (Ubuntu 18.04)
134 vanilla linux 4.6.7 (ZFS only)
135 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
136 * Client primary unpatched kernels built and tested during release cycle:
137 3.10.0-1062.1.1.el7 (RHEL7.7)
138 4.12.14-95.29 (SLES12 SP4)
139 4.15.0-48 (Ubuntu 18.04)
140 * Other clients known to build on unpatched kernels at some point (others may also work):
141 3.10.0-862.14.4.el7 (RHEL7.5)
142 3.10.0-957.27.2.el7 (RHEL7.6)
143 4.14.0-49.13.1.el7a (RHEL7.5)
144 4.14.0-115.2.2.el7a (RHEL7.6)
145 4.18.0-80.7.1.el8 (RHEL8.0)
146 4.4.120-92.70 (SLES12 SP2)
147 4.4.180-94.100 (SLES12 SP3)
148 4.4.0-131 (Ubuntu 16.04)
149 vanilla linux 4.15.0 (ZFS only)
150 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
151 * Recommended ZFS / SPL version: 0.7.13
152 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
153 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
154 stack overflow. For more information, please refer to bugzilla 17630.
155 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
156 the re-exporting nodes to be mounted with "32bitapi" mount option
157 * Note for Ubuntu 14.04.5:
158 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
159 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
161 --------------------------------------------------------------------------------
165 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
166 for currently supported client and server kernel versions.
167 * Server known to build on the following kernels (others may also work):
168 2.6.32-696.18.7.el6 (RHEL6.9)
169 3.10.0-862.14.4.el7 (RHEL7.5)
170 3.10.0-957.el7 (RHEL7.6)
171 4.4.120-92.70 (SLES12 SP2)
172 4.4.162-94.69 (SLES12 SP3)
173 4.4.0-131 (Ubuntu 16.04)
174 4.15.0-32 (Ubuntu 18.04)
175 vanilla linux 4.6.7 (ZFS only)
176 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
177 * Client known to build on unpatched kernels (others may also work):
178 2.6.32-696.18.7.el6 (RHEL6.9)
179 2.6.32-754.6.3.el6 (RHEL6.10)
180 3.10.0-862.14.4.el7 (RHEL7.5)
181 3.10.0-957.el7 (RHEL7.6)
182 4.14.0-49.13.1.el7a (RHEL7.5)
183 4.14.0-115.2.2.el7a (RHEL7.6)
184 4.4.120-92.70 (SLES12 SP2)
185 4.4.162-94.69 (SLES12 SP3)
186 4.4.0-131 (Ubuntu 16.04)
187 4.15.0-32 (Ubuntu 18.04)
188 vanilla linux 4.15.0 (ZFS only)
189 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
190 * Recommended ZFS / SPL version: 0.7.9
191 * Tested with ZFS / SPL version: 0.7.9
192 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
193 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
194 stack overflow. For more information, please refer to bugzilla 17630.
195 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
196 the re-exporting nodes to be mounted with "32bitapi" mount option
197 * Note for Ubuntu 14.04.5:
198 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
199 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
201 --------------------------------------------------------------------------------
203 2018-04-03 Intel Corporation
205 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
206 for currently supported client and server kernel versions.
207 * Server known to build on patched kernels:
208 2.6.32-431.29.2.el6 (RHEL6.5)
209 2.6.32-504.30.3.el6 (RHEL6.6)
210 2.6.32-573.26.1.el6 (RHEL6.7)
211 2.6.32-642.15.1.el6 (RHEL6.8)
212 2.6.32-696.18.7.el6 (RHEL6.9)
213 3.10.0-693.21.1.el7 (RHEL7.4)
214 3.0.101-0.47.71 (SLES11 SP3)
215 3.0.101-107 (SLES11 SP4)
216 3.12.74-60.64.40 (SLES12 SP1)
217 4.4.114-92.64 (SLES12 SP2)
218 4.4.114-94.11 (SLES12 SP3)
219 3.13.0-101 (Ubuntu 14.04, ZFS only)
220 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
221 4.4.0-116 (Ubuntu 16.04)
222 vanilla linux 4.6.7 (ZFS only)
223 * Client known to build on unpatched kernels:
224 2.6.32-431.29.2.el6 (RHEL6.5)
225 2.6.32-504.30.3.el6 (RHEL6.6)
226 2.6.32-573.26.1.el6 (RHEL6.7)
227 2.6.32-642.15.1.el6 (RHEL6.8)
228 2.6.32-696.18.7.el6 (RHEL6.9)
229 3.10.0-693.21.1.el7 (RHEL7.4)
230 3.0.101-0.47.71 (SLES11 SP3)
231 3.0.101-107 (SLES11 SP4)
232 3.12.74-60.64.40 (SLES12 SP1)
233 4.4.114-92.64 (SLES12 SP2)
234 4.4.114-94.11 (SLES12 SP3)
235 3.13.0-101 (Ubuntu 14.04)
236 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
237 4.4.0-116 (Ubuntu 16.04)
239 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
240 * Recommended ZFS / SPL version: 0.7.6
241 * Tested with ZFS / SPL version: 0.7.6
242 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
243 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
244 stack overflow. For more information, please refer to bugzilla 17630.
245 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
246 the re-exporting nodes to be mounted with "32bitapi" mount option
247 * Note for Ubuntu 14.04.5:
248 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
249 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
251 --------------------------------------------------------------------------------
253 2017-07-05 Intel Corporation
255 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
256 for currently supported client and server kernel versions.
257 * Server known to build on patched kernels:
258 2.6.32-431.29.2.el6 (RHEL6.5)
259 2.6.32-504.30.3.el6 (RHEL6.6)
260 2.6.32-573.26.1.el6 (RHEL6.7)
261 2.6.32-642.15.1.el6 (RHEL6.8)
262 2.6.32-696.3.1.el6 (RHEL6.9)
263 3.10.0-514.21.1.el7 (RHEL7.3)
264 3.0.101-0.47.71 (SLES11 SP3)
265 3.0.101-100 (SLES11 SP4)
266 3.12.74-60.64.40 (SLES12 SP1)
267 4.4.59-92.17 (SLES12 SP2)
268 vanilla linux 4.6.7 (ZFS only)
269 * Client known to build on unpatched kernels:
270 2.6.32-431.29.2.el6 (RHEL6.5)
271 2.6.32-504.30.3.el6 (RHEL6.6)
272 2.6.32-573.26.1.el6 (RHEL6.7)
273 2.6.32-642.15.1.el6 (RHEL6.8)
274 2.6.32-696.3.1.el6 (RHEL6.9)
275 3.10.0-514.21.1.el7 (RHEL7.3)
276 3.0.101-0.47.71 (SLES11 SP3)
277 3.0.101-100 (SLES11 SP4)
278 3.12.74-60.64.40 (SLES12 SP1)
279 4.4.59-92.17 (SLES12 SP2)
281 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
282 * Recommended ZFS / SPL version: 0.6.5.9
283 * Tested with ZFS / SPL version: 0.6.5.9
284 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
285 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
286 stack overflow. For more information, please refer to bugzilla 17630.
287 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
288 the re-exporting nodes to be mounted with "32bitapi" mount option
290 --------------------------------------------------------------------------------
292 12-07-2016 Intel Corporation
294 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
295 for currently supported client and server kernel versions.
296 * Server known to build on patched kernels:
297 2.6.32-431.29.2.el6 (RHEL6.5)
298 2.6.32-504.30.3.el6 (RHEL6.6)
299 2.6.32-573.26.1.el6 (RHEL6.7)
300 2.6.32-642.11.1.el6 (RHEL6.8)
301 3.10.0-514.el7 (RHEL7.3)
302 3.0.101-0.47.71 (SLES11 SP3)
303 3.0.101-80 (SLES11 SP4)
304 3.12.62-60.64.8 (SLES12 SP1)
305 vanilla linux 4.5.7 (ZFS only)
306 * Client known to build on unpatched kernels:
307 2.6.32-431.29.2.el6 (RHEL6.5)
308 2.6.32-504.30.3.el6 (RHEL6.6)
309 2.6.32-573.26.1.el6 (RHEL6.7)
310 2.6.32-642.11.1.el6 (RHEL6.8)
311 3.10.0-514.el7 (RHEL7.3)
312 3.0.101-0.47.71 (SLES11 SP3)
313 3.0.101-80 (SLES11 SP4)
314 3.12.62-60.64.8 (SLES12 SP1)
316 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
317 * Recommended ZFS / SPL version: 0.6.5.8
318 * Tested with ZFS / SPL version: 0.6.5.8
319 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
320 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
321 stack overflow. For more information, please refer to bugzilla 17630.
322 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
323 the re-exporting nodes to be mounted with "32bitapi" mount option
325 --------------------------------------------------------------------------------
327 02-29-2016 Intel Corporation
329 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
330 for currently supported client and server kernel versions.
331 * Server known to build on patched kernels:
332 2.6.32-431.29.2.el6 (RHEL6.5)
333 2.6.32-504.30.3.el6 (RHEL6.6)
334 2.6.32-573.12.1.el6 (RHEL6.7)
335 3.10.0-327.3.1.el7 (RHEL7.2)
336 3.0.101-0.47.71 (SLES11 SP3)
337 3.0.101-68 (SLES11 SP4)
338 vanilla linux 4.2.1 (ZFS only)
339 * Client known to build on unpatched kernels:
340 2.6.32-431.29.2.el6 (RHEL6.5)
341 2.6.32-504.30.3.el6 (RHEL6.6)
342 2.6.32-573.12.1.el6 (RHEL6.7)
343 3.10.0-327.3.1.el7 (RHEL7.2)
344 3.0.101-0.47.71 (SLES11 SP3)
345 3.0.101-68 (SLES11 SP4)
348 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
349 * Recommended ZFS / SPL version: 0.6.4.2
350 * Tested with ZFS / SPL version: 0.6.4.2
351 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
352 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
353 stack overflow. For more information, please refer to bugzilla 17630.
354 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
355 the re-exporting nodes to be mounted with "32bitapi" mount option
357 --------------------------------------------------------------------------------
359 03-10-2015 Intel Corporation
361 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
362 for currently supported client and server kernel versions.
363 * Server known to build on patched kernels:
364 2.6.32-431.29.2.el6 (RHEL6.5)
365 2.6.32-504.8.1.el6 (RHEL6.6)
366 3.0.101-0.46 (SLES11 SP3)
367 * Client known to build on unpatched kernels:
368 2.6.32-431.29.2.el6 (RHEL6.5)
369 2.6.32-504.8.1.el6 (RHEL6.6)
370 3.10.0-123.20.1.el7 (RHEL7)
371 3.0.101-0.46 (SLES11 SP3)
372 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
373 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
374 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
375 stack overflow. For more information, please refer to bugzilla 17630.
376 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
377 the re-exporting nodes to be mounted with "32bitapi" mount option
379 Severity : enhancement
381 Description: control OST-index in IDIF via ROCOMPAT flag.
382 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
383 last_rcvd file. For new formatted OST device, it will be auto set;
384 for the case of upgrading from old OST device, you can enable it
385 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
386 enabled, for new created OST-object, its IDIF-in-LMA will contain
387 the OST-index; for the existing OST-object, the OSD will convert
388 old format IDIF as new format IDIF with OST-index stored in the
389 LMA EA when accessing such OST-object or via OI scrub. Once such
390 flag is enabled, it cannot be reverted back, so the system cannot
391 be downgraded to the orignal incompatible version.
393 --------------------------------------------------------------------------------
395 07-30-2014 Intel Corporation
397 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
398 for currently supported client and server kernel versions.
399 * Server known to build on patched kernels:
400 2.6.32-431.20.3.el6 (RHEL6.5)
401 3.0.101-0.31 (SLES11 SP3)
402 * Client known to build on unpatched kernels:
403 2.6.32-431.20.3.el6 (RHEL6.5)
404 3.0.101-0.31 (SLES11 SP3)
407 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
408 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
409 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
410 stack overflow. For more information, please refer to bugzilla 17630.
411 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
412 the re-exporting nodes to be mounted with "32bitapi" mount option
414 --------------------------------------------------------------------------------
416 10-31-2013 Intel Corporation
418 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
419 * Server builds for kernels:
420 2.6.32-358.18.1.el6 (RHEL6.4)
421 * Client builds for unpatched kernels:
422 2.6.32-358.18.1.el6 (RHEL6.4)
423 3.0.92-0.8 (SLES11 SP3)
425 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
426 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
427 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
428 stack overflow. For more information, please refer to bugzilla 17630.
429 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
430 the re-exporting nodes to be mounted with "32bitapi" mount option
432 --------------------------------------------------------------------------------
434 05-30-2013 Intel Corporation
436 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
437 * Server builds for kernels:
438 2.6.32-358.6.2.el6 (RHEL6.4)
439 * Client builds for unpatched kernels:
440 2.6.32-358.6.2.el6 (RHEL6.4)
441 3.0.74-0.6.6 (SLES11 SP2)
443 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
444 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
445 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
446 stack overflow. For more information, please refer to bugzilla 17630.
447 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
448 the re-exporting nodes to be mounted with "32bitapi" mount option
450 --------------------------------------------------------------------------------
452 9-30-2012 Whamcloud, Inc.
454 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
455 * Server support for kernels:
456 2.6.32-279.5.1.el6 (RHEL6)
457 * Client support for unpatched kernels:
458 2.6.32-279.5.1.el6 (RHEL6)
460 2.6.18-238.19.1.el5 (RHEL5)
461 2.6.16.60-0.69.1 (SLES10)
462 * Recommended e2fsprogs version: 1.41.90.wc4
463 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
464 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
465 stack overflow. For more information, please refer to bugzilla 17630.
466 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
467 the re-exporting nodes to be mounted with "32bitapi" mount option
469 Severity : enhancement
471 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
473 Severity : enhancement
475 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
477 Severity : enhancement
479 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
481 Severity : enhancement
483 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
485 Severity : enhancement
487 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
489 --------------------------------------------------------------------------------
491 04-30-2012 Whamcloud, Inc.
493 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
494 * Server support for kernels:
495 2.6.32-220.4.2.el6 (RHEL6)
496 * Client support for unpatched kernels:
497 2.6.32-220.4.2.el6 (RHEL6)
499 2.6.18-238.19.1.el5 (RHEL5)
500 2.6.16.60-0.69.1 (SLES10)
501 * Recommended e2fsprogs version: 1.41.90.wc4
502 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
503 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
504 stack overflow. For more information, please refer to bugzilla 17630.
505 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
506 the re-exporting nodes to be mounted with "32bitapi" mount option
508 Severity : enhancement
510 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
512 Severity : enhancement
514 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
516 Severity : enhancement
518 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
520 Severity : enhancement
522 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
524 --------------------------------------------------------------------------------
526 09-30-2011 Whamcloud, Inc.
528 * Server support for kernels:
529 2.6.18-238.19.1.el5 (RHEL5)
530 2.6.32-131.6.1.el6 (RHEL6)
531 * Client support for unpatched kernels:
532 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
533 2.6.18-238.19.1.el5 (RHEL5)
534 2.6.32-131.6.1.el6 (RHEL6)
535 2.6.29.4-167.fc11 (FC11)
536 2.6.16.60-0.69.1 (SLES10)
538 * Recommended e2fsprogs version: 1.41.90.wc3
539 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
540 Lustre filesystem with 4K stack may cause a stack overflow. For more
541 information, please refer to bugzilla 17630.
542 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
543 mounted with "32bitapi" mount option
545 Severity : enhancement
546 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
548 Severity : enhancement
549 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
551 Severity : enhancement
552 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
555 Frequency : only with SLES10
557 Description: Use OFED "KMP" provided by Novell
558 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
559 Lustre is now built against this package. That means you need to
560 install the ofed-kmp package from Novell for the patchless client.
561 Note that the ofed-kmp that Novell ships may not exactly match the
562 kernel version but should still be compatible.
564 Severity : enhancement
566 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
567 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
569 --------------------------------------------------------------------------------
571 2010-07-15 Oracle, Inc.
573 * Server support for kernels:
574 2.6.18-164.11.1.el5 (RHEL 5)
575 2.6.18-164.11.1.0.1.el5 (OEL 5)
576 * Client support for unpatched kernels:
577 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
578 2.6.18-164.11.1.el5 (RHEL 5),
579 2.6.18-164.11.1.0.1.el5 (OEL 5)
580 2.6.16.60-0.42.8 (SLES 10),
582 2.6.29.4-167.fc11 (FC11)
583 * Recommended e2fsprogs version: 1.41.10-sun2
585 Severity : enhancement
587 Description: Distribute lfs_migrate script with lustre
588 Details : The lfs_migrate mentioned in the manual is now included in the
589 lustre tree in lustre/scripts and will distributed in lustre RPMs.
591 Severity : enhancement
593 Description: Failover nid list added to import proc file.
595 Severity : enhancement
597 Description: "som_preview" mount option required to use SOM
598 Details : The clients will need to use som_preview mount option
602 Description: Lustre does not do 1MB IOs to HW RAID
603 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
605 Severity : enhancement
607 Description: lfs quota output is not very convenient for awk/sed-parsing
608 Details : Some positions in lfs quota output table could be empty or
609 non-empty which made it hard to parse it with scripts, now
610 a dash is put instead of space where there is not supposed
615 Description: Update to OFED 1.5.1
616 Details : OFED 1.5.1 will be used in environments that don't use
617 vendor supplied OFED.
621 Description: Remove lnet. prefix from lctl params display
622 Details : Also renamed lnet.memused to lnet_memused to
627 Description: Use conf_param -d to remove permanant settings.
631 Description: Add list_param -R
632 Details : lctl list_param now supports recursive mode.
635 Bugzilla : 21829, 22312
636 Description: Fix llobdstat.
637 Details : 1. Now updating the write_bytes counter
638 2. Port script fixes from 1.8.
642 Description: decrease the usage of memory on clients.
643 Details : 1. On clients, recycle dentries and inodes unused.
644 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
649 Description: llobdstat fix and enhancement
650 Details : add a counter to set a limit to how many samples will be returned
651 fix a wildcard in the path to limit to obdfilter stats only
655 Description: lfs check only allowed for root
656 Details : Code cleanups and added a regresion test of non-root use of
661 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
662 Details : Fix hr->hr_index overload.
666 Description: Fix FIEMAP functionality.
670 Description: FC11 patchless client support added
674 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
675 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
679 Description: New acc-sm script and various test-framework enhancments.
683 Description: Previously landed patch was reverted.
687 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
691 Description: Add ldiskfs maxdirsize mount option.
695 Description: Fix mkfs.lustre support for 16TB LUNs.
699 Description: Print mmp_check_interval
700 Details : print mmp_check_interval and make it possible to abort
701 mount operation in case it takes too long.
705 Description: Add cascading_rw to lustre mpi tests.
709 Description: Store SOM data in EA.
710 Details : Also several SOM related cleanups.
714 Description: Add refcount to md_open_data
718 Description: Use refcount on lov_request_set to prevent memory corruption.
722 Description: Statahead: use parent dir's i_mutex for synchronization.
726 Description: loadgen improvements.
730 Description: Enable peer health detection in ksocklnd and ko2iblnd.
734 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
738 Description: Fix race on obd_devcie:md_stats
742 Description: Fix SLES 11 client builds.
743 Details : Resolved uninitalized pointer gcc warnings.
747 Description: Allow non-root access for "lfs check".
748 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
752 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
754 Severity : enhancement
756 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
757 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
761 Description: Sanity test_27p fails.
762 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
763 Sanity tests 27 were cleaned up and improved.
767 Description: Unify req format on client/servers
768 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
769 buffers and swabbers approach..
774 Description: drop join file stuff.
776 Severity : enhancement
778 Description: Add 2.6.27 and SLES11 (patchless client) support
782 Description: Fix in ptlrpc_expire_one_request() to print the signed time
785 Severity : enhancement
787 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
790 Severity : enhancement
792 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
794 Severity : enhancement
796 Description: correctly handle big reply message.
797 Details : send LNet event if reply is bigger than buffer and adjust this buffer
802 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
804 Severity : enhancement
806 Description: Add support for OEL5.
810 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
812 Severity : enhancement
814 Description: "weak-modules" support
815 Details : Implement "weak-modules" support which enables kernel modules
816 to be used with any kernel that implements the same kABI. In
817 order to achieve this modules are now installed in
818 /lib/modules/$(uname -r)/updates/kernel on all distributions.
822 Description: client could not reconnect to OST because of an active request.
823 Details : The client now retries to reconnect to the same server, if a
824 connect request failed with EBUSY or -EAGAIN. Test case 35b
825 conf-sanity.sh has been created.
829 Description: Don't allow make backward step on assignin osc next id.
830 Details : race between allocation next id and ll_sync thread can be cause
831 of set wrong osc next id and can be kill valid ost objects.
835 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
839 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
843 Description: Error handling in osc_statfs_interpret() has been improved.
844 Details : Check in osc_statfs_interpret() for EBADR.
848 Description: Do not send statfs() requests to OSTs disabled by administrator.
849 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
851 Severity : enhancement
853 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
855 Severity : enhancement
857 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
859 Severity : enhancement
861 Description: Enhancement for lfs(1) command to use numeric uid/gid.
865 Description: Improved cleanup procedure for conf-sanity test_46a.
869 Description: Do not start bulk transfer, if client tried to reconnect.
870 Details : Do not start bulk transfer, if client tried to reconnect,
871 but failed due to active rpcs. Check for exp_abort_active_req
872 has been implemented.
876 Description: Prevent inconsistences between linux and lustre mount structures.
877 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
878 Make the sleep interruptible.
880 Severity : enhancement
882 Description: provide server to client comms path
883 Details : server to client communications path via new LDLM_SET_INFO rpc
887 Description: truncate starts GFP_FS allocation under transaction causing deadlock
888 Details : ldiskfs_truncate calls grab_cache_page which may start page
889 allocation under an open transaction. This may lead to
890 calling prune_icache with consequent lustre reentrance.
894 Description: send correctly lsm on open replay
895 Details : MDS is trust to LSM size on replay open, but client can set wrong size
900 Description: enable client interface failover
901 Details : When a child reconnects from another NID, properly update export
902 nid hash position and ldlm reverse import.
904 Severity : enhancement
906 Description: Communicate OST degraded/readonly state via statfs to MDS
907 Details : Flags in the statfs returned from OSTs indicate whether the
908 OST is in a degraded RAID state, or if the filesystem has
909 turned read-only after a filesystem error is detected.
914 Description: don't panic if EPROTO was hit when reading symlink
915 Details : correctly handling request reference in error cases.
917 Severity : enhancement
919 Description: Add LustreNetLink, a kernel-userspace communcation path.
921 Severity : enhancement
923 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
928 Description: access to llog context before init.
929 Details : move handling CATALOGS file at osc layer and forbid access to llog
933 Frequency : rare, only when client is writing to OST on local node
935 Description: Avoid deadlock for local client writes
936 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
937 memory freeing context. This allows OST threads to set the
938 PF_MEMALLOC flag on task structures in order to allocate memory
939 from reserved pools and complete IO.
940 Use GFP_HIGHUSER for OST allocations for non-local client writes,
941 so that the OST threads generate memory pressure and allow
942 inactive pages to be reclaimed.
944 Severity : enhancement
946 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
951 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
952 Details : move ldlm namespace creation in setup phase to avoid grab
953 _lprocfs_lock with cli_sem held.
957 Description: Temporarily disable grant shrink.
958 Details : Disable the feature for debugging.
961 Frequency : only when formatting test filesystems
963 Description: Unable to run several mkfs.lustre on loop devices at the same time
964 Details : mkfs.lustre returns error 256 on the concurrent loop devices
965 formatting. The solution is to proper handle the error.
967 Severity : enhancement
969 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
971 Severity : enhancement
973 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
978 Description: resolve race between obd_disconnect and class_disconnect_exports
979 Details : if obd_disconnect will be called to already disconnected export he
980 forget release one reference and osc module can't unloaded.
982 Severity : enhancement
984 Description: Allow tuning service thread via /proc
985 Details : For each service a new
986 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
987 created that can be used to set min/max thread counts, and get the
988 current number of running threads.
993 Descriptoin: don't return error if have partially created objects for file.
994 Details : lov_update_create_set uses set->set_success as index for created
995 objects, so if some requests will be failed, they will have hole
996 at end of array and qos_shrink_lsm for allocate correct lsm.
998 Severity : enhancement
1000 Description: Update OFED support to 1.4.1
1005 Description: Reduce small size read RPC
1006 Details : Set read-ahead limit for every file and only do read-ahead when
1007 available read-ahead pages are bigger than 1M to avoid small size
1010 Severity : enhancement
1012 Description: move AT tunable parameters for more consistent usage
1013 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1015 Severity : enhancement
1017 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1018 Details : allow skip disconnected ost for send statfs request and hide error
1024 Description: fix racy locking of mballoc block bitmaps causing BUG
1025 Details : The locking of the mballoc buddy bitmap and the in-memory
1026 block bitmap was using two different spin locks in some
1027 cases. This made it possible to incorrectly access the
1028 mballoc bitmap while another process was modifying it,
1029 causing a sanity assertion to fail. While no on-disk corruption
1030 was reported, there was some risk of this happening.
1033 Frequency : rare, on llog test 6
1035 Descriptoin: don't allow connect to already connected import
1036 Details : allowing connect to already connected import is hide connecting problem.
1041 Descriptoin: allow kill process which wait statahead result
1042 Details : for some reasons 'ls' can stick in waiting result from statahead,
1043 in this case need way for kill this process.
1045 Severity : enhancement
1047 Description: Add state history info file, enhance import info file
1048 Details : Track import connection state changes in a new osc/mdc proc file;
1049 add overview-type data to the osc/mdc import proc file.
1051 Severity : enhancement
1053 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1055 Severity : enhancement
1057 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1062 Descriptoin: OST grant too much space to client even there are not enough space.
1063 Details : Client will shrink its grant cache to OST if there are no write
1064 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1065 this grant cache if there are already not enough avaible space
1066 (left_space < total_clients * 32M).
1069 Frequency : start MDS on uncleanly shutdowned MDS device
1071 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1072 Details : stay in waiting mds<>ost recovery finished produce random bugs
1073 due race between two ll_sync thread for one lov target. send
1074 ACTIVATE event only if connect realy finished and import have
1078 Frequency : rare, connect and disconnect target at same time
1080 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1081 Details : don't call obd_disconnect under lov_lock. this long time
1082 operation and can block ptlrpcd which answer to connect request.
1087 Descriptoin: don't lose wakeup for imp_recovery_waitq
1088 Details : recover_import_no_retry or invalidate_import and import_close can
1089 both sleep on imp_recovery_waitq, but we was send only one wakeup
1093 Frequency : always with long access acl
1095 Descriptoin: mds can't pack reply with long acl.
1096 Details : mds don't control size of acl but they limited by reint/getattr
1099 Severity : enhancement
1101 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1103 Severity : enhancement
1105 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1108 Frequency : start MDS on uncleanly shutdowned MDS device
1110 Descriptoin: aborting recovery hang on MDS
1111 Details : don't throttle destroy RPCs for the MDT.
1114 Frequency : on remount
1116 Description: external journal device not working after the remount
1117 Details : clear dev_rdonly flag for external journal devices in
1123 Description: shutdown vs evict race
1124 Details : client_disconnect_export vs connect request race.
1125 if client will evicted at this time - we start invalidate
1126 thread without referece to import and import can be freed
1130 Frequency : rare, need acl's on inode.
1132 Description: client can't handle ost additional correctly
1133 Details : if ost was added after client connected to mds client can have
1134 hit lnet_try_match_md ... to big messages to wide striped files.
1135 in this case need teach client to handle config events about add
1136 lov target and update client max ea size at that event.
1138 Severity : enhancement
1140 Description: Changelogs
1141 Details : Changelogs are a lightweight mechanism to track filesystem
1142 metadata and namespace changes. The changelog is recorded
1143 permanently on the MDTs, and is periodically "consumed" / purged
1144 when records are no longer needed.
1146 Severity : enhancement
1148 Description: compact fld format with extents
1149 Details : Store range of seq rather than every seq in FLD. Seq
1150 controller update FLD rather than clients. In Case of CMD, mdt0
1151 has FLD, all other metadata server act as non persistent proxy
1152 for FLD queries and cache fld entries in fld cache.
1157 Description: don't skip ost target if they assigned to file
1158 Details : Drop slow OSCs if we can, but not for requested start idx.
1159 This means "if OSC is slow and it is not the requested
1160 start OST, then it can be skipped, otherwise skip it only
1161 if it is inactive/recovering/out-of-space.
1165 Description: more cleanup in mds_lov
1166 Details : not send LOV EA under replay, we can't know about they size at this
1167 time. Don't allow client connect to mds before any ost connected,
1168 for avoid problems with LOV EA size and returning EIO to client.
1170 Severity : enhancement
1172 Description: Interoperability at server side (Disk interoperability)
1174 Severity : enhancement
1176 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1178 Severity : enhancement
1180 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1182 Severity : enhancement
1184 Description: New client IO stack (CLIO).
1186 Severity : enhancement
1188 Description: Commit on sharing. Eliminate inter-client dependencies between
1189 uncommitted transactions by doing transaction commits.
1190 Thereby clients may recovery independently.
1193 Frequency : Create a symlink file with a very long name
1195 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1196 Details : If there is no extra space in the request for early cancels,
1197 ldlm_req_handles_avail() returns 0 instead of a negative value.
1199 Severity : enhancement
1201 Description: Add /proc entry for import status
1202 Details : The mdc, osc, and mgc import directories now have
1203 an import directory that contains useful import data for debugging
1204 connection problems.
1206 Severity : enhancement
1208 Description: Re-disable certain /proc logging
1209 Details : Enable and disable client's offset_stats, extents_stats and
1210 extents_stats_per_process stats logging on the fly.
1213 Frequency : Only on FC kernels 2.6.22+
1215 Description: oops in statahead
1216 Details : Do not drop reference count for the dentry from VFS when lookup,
1217 VFS will do that by itself.
1219 Severity : enhancement
1221 Description: Generic /proc file permissions
1222 Details : Set /Proc file permissions in a more generic way to enable non-
1223 root users operate on some /proc files.
1227 Description: Hitting mdc_commit_close() ASSERTION
1228 Details : Properly handle request reference release in
1229 ll_release_openhandle().
1233 Description: quota recovery deadlock during mds failover
1234 Details : This patch includes att18982, att18236, att18237 in bz14840.
1236 1. fix osts hang when mds does failover with quotaon
1237 2. prevent watchdog storm when osts threads wait for the
1242 Frequency : only patchless client
1243 Description: add workaround for race between add/remove dentry from hash
1245 Severity : enhancement
1247 Description: Allow OST glimpses to return PW locks
1251 Description: LBUG when llog conf file is full
1252 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1256 Description: Prevent import from entering FULL state when server in recovery
1260 Description: service mount cannot take device name with ":"
1261 Details : Only when device name contains ":/" will mount treat it as
1267 Description: replace ptlrpcd with the statahead thread to interpret the async
1268 statahead RPC callback
1272 Frequency : on recovery
1273 Description: I/O failures after umount during fail back
1274 Details : if client reconnected to restarted server we need join to recovery
1275 instead of find server handler is changed and process self
1276 eviction with cancel all locks.
1278 Severity : enhancement
1280 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1282 Severity : enhancement
1284 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1286 Severity : enhancement
1288 Description: Upcall on Lustre log has been dumped
1289 Details : Allow for a user mode script to be called once a Lustre log has
1290 been dumped. It passes the filename of the dumped log to the
1291 script, the location of the script can be specified via
1292 /proc/sys/lnet/debug_log_upcall.
1297 Description: avoid idr_remove called for id which is not allocated.
1298 Details : Move assigment s_dev for clustered nfs to end of initialization,
1299 to avoid problem with error handling.
1304 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1305 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1306 into CONN_USED_HASH and this prodice warning when put connection
1307 again in unused hash.
1310 Severity : enhancement
1312 Description: File striping can now be set to use an arbitrary pool of OSTs.
1314 Severity : enhancement
1316 Description: Export bytes_read/bytes_write count on OSC/OST.
1320 Description: Early reply size mismatch, MGC loses connection
1321 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1322 the connect flags are properly negotiated.
1327 Description: quotas are not honored with O_DIRECT
1328 Details : all writes with the flag O_DIRECT will use grants which leads to
1329 this problem. Now using OBD_BRW_SYNC to guard this.
1333 Description: add quota statistics
1334 Details : 1. sort out quota proc entries and proc code.
1335 2. add quota statistics
1337 Severity : enhancement
1339 Description: enable quota support for HEAD.
1343 Description: Properly propagate oinfo flags from lov to osc for statfs
1344 Details : restore missing copy oi_flags to lov requests.
1346 Severity : enhancement
1348 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1349 : l_getgroups(8), lst(8), routerstat(8)
1350 Details : included man pages for llobdstat(8), llstat(8),
1351 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1353 Severity : enhancement
1355 Description: configure's --enable-quota should check the
1356 : kernel .config for CONFIG_QUOTA
1357 Details : configure is terminated if --enable-quota is passed but
1358 : no quota support is in kernel
1362 Description: Remove portals compatibility
1363 Details : Remove portals compatibility, not interoperable with releases
1368 Description: Resolve device initialization race
1369 Details : Prevent proc handler from accessing devices added to the
1370 obd_devs array but yet be intialized.
1372 Severity : enhancement
1374 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1376 Severity : enhancement
1378 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1383 Description: Using wrong pointer in osc_brw_prep_request
1384 Details : Access to array[-1] can produce panic if kernel compiled with
1385 CONFIG_PAGE_ALLOC enabled
1389 Description: Client runs out of low memory
1390 Details : Consider only lowmem when counting initial number of llap pages
1394 Description: *optional* service tags registration
1395 Details : if the "service tags" package is installed on a Lustre node
1396 When the filesystem is mounted, a local-node service tag will
1397 be created. See http://inventory.sun.com/ for more information
1398 about the Service Tags asset management system.
1402 Description: Kernel BUG tries to release flock
1403 Details : Lustre does not destroy flock lock before last reference goes
1404 away. So always drop flock locks when client is evicted and
1405 perform unlock regardless of successfulness of speaking to MDS.
1409 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1413 Description: sanity test 65a fails if stripecount of -1 is set
1414 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1419 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1420 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1423 Severity : enhancement
1425 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1429 Description: do not process already freed flock
1430 Details : flock can possibly be freed by another thread before it reaches
1431 to ldlm_flock_completion_ast.
1435 Description: LBUG during stress test
1436 Details : Need properly lock accesses the flock deadlock detection list.
1440 Description: oops in page fault handler
1441 Details : kernel page fault handler can return two special 'pages' in error
1442 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1446 Description: timeout with invalidate import.
1447 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1448 should be handled by ptlrpcd. This produce long age waiting and
1449 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1451 Severity : enhancement
1453 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1457 Description: enable MGS and MDT services start separately
1458 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1459 a co-located MGS without starting the MGS, which is a complement
1460 to 'nosvc' mount option.
1464 Frequency : after recovery
1465 Description: precreate to many object's after del orphan.
1466 Details : del orphan st in oscc last_id == next_id and this triger growing
1467 count of precreated objects. Set flag LOW to skip increase count
1468 of precreated objects.
1472 Frequency : rare, on clear nid stats
1473 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1474 Details : when clean nid stats sometimes try destroy live entry,
1475 and this produce panic in free.
1479 Description: Stack overflow during MDS log replay
1480 ease stack pressure by using a thread dealing llog_process.
1484 Description: wait until IO finished before start new when do lock cancel.
1485 Details : VM protocol want old IO finished before start new, in this case
1486 need wait until PG_writeback is cleared until check dirty flag and
1487 call writepages in lock cancel callback.
1489 Severity : enhancement
1491 Description: using special macro for print time and cleanup in includes.
1495 Description: mds_mfd_close() ASSERTION(rc == 0)
1496 Details : In mds_mfd_close(), we need protect inode's writecount change
1497 within its orphan write semaphore to prevent possible races.
1501 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1505 Frequency : rare, on shutdown ost
1506 Description: don't hit live lock with umount ost.
1507 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1508 use shrink_dcache_sb instead.
1512 Description: don't panic with use echo client
1513 Details : echo client pass NULL as client nid pointer and this produce null
1514 pointer dereference.
1518 Description: fix build on ppc32
1519 Details : compile code with -m64 flag produce wrong object file for ppc32.
1523 Description: add message levels for liblustreapi
1527 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1528 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1529 will result in an -EAGAIN and in the exit call path we call
1530 alloc_rr() which will with increasing aggressiveness attempt to
1531 aquire precreated objects on the minimum number of required OSCs.
1535 Description: Use old size assignment to avoid deadlock
1536 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1537 the scheduling while holding a spinlock. We do not need locking
1538 for size in ll_update_inode() because size is only updated from
1539 the MDS for directories or files without objects, so there is no
1540 other place to do the update, and concurrent access to such inodes
1541 are protected by the inode lock.
1545 Description: resolve "_IOWR redefined" build error on SLES10
1549 Description: dump the memory debugging after all modules are unloaded to
1550 suppress false negative in conf_sanity test 39
1552 Severity : enhancement
1554 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1559 Description: fix SLES kernel versioning
1560 Details : the kernel version for our SLES 10 kernel did not include a "-"
1561 before the "smp" at the end. while this was not a problem in
1562 general, it did mean that software trying to use the kernel
1563 version to try to detect a vendor specific kernel would fail.
1564 this was most evident by the OFED build scripts.
1568 Description: Don't update lov_desc members until making sure they are valid
1569 Details : When updating lov_desc members via proc fs, need fix their
1570 validities before doing the real update.
1574 Description: don't put request into delay list while invalidate in flight.
1575 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1576 import in flight. this produce timeout for invalidate and sometimes
1577 can cause stale data.
1581 Frequency : on ppc only
1582 Description: not convert ost objects for directory because it's not exist.
1583 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1587 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1589 Severity : enhancement
1591 Description: Update to support 2.6.22.14 vanilla kernel.
1595 Frequency : rare, on recovery
1596 Description: read procfs can produce deadlock in some situation
1597 Details : Holding lprocfs lock which send rpc can produce block for destroy
1598 obd objects and this also block reconnect with -EALREADY. This
1599 isn't fix all lprocfs bugs - but make it rare.
1601 Severity : enhancement
1603 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1606 Frequency : frequent on X2 node
1608 Description: mdc_set_open_replay_data LBUG
1609 Details : Set replay data for requests that are eligible for replay.
1613 Description: lustre_mgs: operation 101 on unconnected MGS
1614 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1615 MGC, and late on MGC cannot successfully connect to MGS and a lot
1616 of the error messages complaining that MGS is not connected.
1620 Frequency : on network error
1621 Description: panic with double free request if network error
1622 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1623 but it's true only for synchronus enqueue, for async enqueue
1624 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1627 Severity : enhancement
1629 Description: client-side metadata stat-ahead during readdir(directory readahead)
1630 Details : perform client-side metadata stat-ahead when the client detects
1631 readdir and sequential stat of dir entries therein
1634 Frequency : on start mds
1636 Description: Implement get_info(last_id) in obdfilter.
1639 Frequency : occasional
1641 Description: Correctly check stale fid, not start epoch if ost not support SOM
1642 Details : open with flag O_CREATE need set old fid in op_fid3 because
1643 op_fid2 was overwritten with new generated fid, but mds can answer
1644 with one of these two fids and both is not stale. Setattr
1645 incorrectly started an epoch and assume will be called
1646 done_writing, but without SOM done_writing ever being called.
1649 Frequency : rare, depends on device drivers and load
1651 Description: MDS or OSS nodes crash due to stack overflow
1652 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1653 In some cases, in conjunction with device drivers that use a lot
1654 of stack the MDS (or possibly OSS) service threads could overflow
1655 the stack. One change which was identified to consume additional
1656 stack has been reworked to avoid the extra stack usage.
1659 Frequency : occasional
1661 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1662 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1663 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1664 oscc_has_objects and see OSC has no objests and call
1665 oscc_internal_create to resend create request.
1667 Severity : enhancement
1669 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1671 Severity : enhancement
1673 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1676 Frequency : very rare
1678 Description: Fix replay if there is an un-replied request and open
1679 Details : In some cases, older replay request will revert the
1680 mcd->mcd_last_xid on MDS which is used to record the client's
1681 latest sent request.
1683 Severity : enhancement
1685 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1687 Severity : enhancement
1689 Description: Add rhel5 support to HEAD.
1691 Serverity : enhancement
1693 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1698 Description: Don't allow skipping OSTs if index has been specified.
1699 Details : Don't allow skipping OSTs if index has been specified, make
1700 locking in internal create lots better.
1704 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1705 Details : access to bitfield in structure is always rounded to long
1706 and this produce problem with not atomic change any bit.
1710 Description: Lustre make rpms failed.
1711 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1712 builds Lustre rpms from tarball.
1715 Frequency : rare on shutdown ost
1717 Description: If llog cancel was not send before clean_exports phase, this can
1718 produce deadlock in llog code.
1719 Details : If llog thread has last reference to obd and call class_import_put
1720 this produce deadlock because llog_cleanup_commit_master wait when
1721 last llog_commit_thread exited, but this never success because was
1722 called from llog_commit_thread.
1726 Description: allow userland application know is lost one of stripes.
1727 Details : fill lvb_blocks with error code on ost and return it to
1728 application if error flag found.
1732 Description: NULL lov_tgts causing MDS oops
1733 Details : more safe checks for NULL lov_tgts for avoid oops.
1735 Severity : enhancement
1737 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1741 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1742 Details : If an OST is down the MDS will hang indefinitely in
1743 obd_statfs_rqset() waiting for the statfs data. While for
1744 MDS QOS usage of statfs, it should not stuck in waiting.
1746 Severity : enhancement
1748 Description: remote_acl support
1749 Details : Support ACL-based permission check for remote user.
1750 Support setfacl/getfacl for remote user with the utils
1751 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1752 the system "{s,g}etfacl" utils.
1754 Severity : enhancement
1756 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1758 Severity : enhancement
1760 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1764 Description: make mgs_setparam() handle fsname containing dash
1765 Details : fsname containing a dash does not work with lctl conf_param
1767 Severity : enhancement
1769 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1771 Severity : enhancement
1773 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1777 Frequency : rare, at shutdown
1778 Description: access already free / zero obd_namespace.
1779 Details : if client_disconnect_export was called without force flag set,
1780 and exist connect request in flight, this can produce access to
1781 NULL pointer (or already free pointer) when connect_interpret
1782 store ocd flags in obd_namespace.
1786 Frequency : only at startup
1787 Description: not alloc memory with spinlock held.
1788 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1789 if any spinlock held.
1791 Severity : enhancement
1793 Description: make lustre randomly fail allocating memory
1794 Details : Make lustre randomly failed allocating memory for testing purpose.
1796 Severity : enhancement
1798 Description: lost problems with lov objid file
1799 Details : Fixes some scability and access to not inited memory problems
1800 in work with lov objdid file.
1805 Description: lfs find does not continue on file error
1806 Details : Continue other files processing when a file/dir is absent.
1810 Description: Inconsistent usage of lustre_pack_reply()
1811 Details : Standardize the usage of lustre_pack_reply() such that it
1812 always generate a CERROR on failure.
1815 Frequency : occasional
1817 Description: MDS hang or stay in waiting lock
1818 Details : If client receive lock with CBPENDING flag ldlm need send lock
1819 cancel as separate rpc, to avoid situation when cancel request
1820 can't processed due all i/o threads stay in wait lock.
1824 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1828 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1832 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1836 Description: Update to RHEL4 latest kernel.
1838 Severity : enhancement
1840 Description: Build SLES10 patchless client fails
1841 Details : The configure was broken by run ./configure with
1842 --with-linux-obj=.... argument for patchless client. When the
1843 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1844 can't search header adequately. Use absolute path such as
1845 -I($LINUX)/include instead.
1849 Description: interrupt oig_wait produce painc on resend.
1850 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1851 private set, and this produce situation when rq_ptlrpcd_data not
1852 copyed to new allocated request and triggered LBUG on assert
1853 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1854 ptlrpcd set if request is changed and can be safety changed to use
1859 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1860 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1861 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1862 these macros and replace them with their difinitions instead.
1866 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1867 Details : Update which_patch & target file for SLES10 latest kernel.
1870 Description: Remove client patches from SLES 10 kernel.
1871 Details : This causes SLES 10 clients to behave as patchless clients
1872 even on a Lustre-patched (server) kernel.
1874 Severity : enhancement
1876 Description: self-adjustable client's lru lists
1877 Details : use adaptive algorithm for managing client cached locks lru
1878 lists according to current server load, other client's work
1879 pattern, memory activities, etc. Both, server and client
1880 side namespaces provide number of proc tunables for controlling
1883 Severity : enhancement
1885 Description: light-weight GSS support
1886 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1887 and reduce performance overhead.
1889 Severity : enhancement
1891 Description: Linux keyring support
1892 Details : Support using service of Linux keyring for Lustre GSS internal
1893 context refresh/cache mechanism.
1897 Description: Fix errors in lfs documentation
1898 Details : Fixes man pages
1902 Description: don't use GFP_* in generic Lustre code.
1903 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1907 Description: obdclass is limited by single OBD_ALLOC(idarray)
1908 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1912 Description: warnings with build patchless client with vanila 2.6.19 and up
1913 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1914 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1918 Frequency : Only for SLES
1919 Description: sanity_quota fail test_1
1920 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1921 file, which makes TSTID[2] inunique.
1925 Frequency : Only for SLES9
1926 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1927 Details : Update target/ChangeLog/which_patch .
1931 Frequency : Only for test_10 in sanity-quota.sh
1932 Description: a bug in quota test code
1933 Details : checking a test flag in a wrong place
1937 Frequency : Only for test_13 in sanity-quota.sh
1938 Description: a bug in quota test script
1939 Details : shouldn't check overfull allocated quota
1943 Description: jbd statistics
1944 Details : Port older jbd statistics patch for sles10
1946 Severity : enhancement
1948 Description: Add printing inode info into message about error in writepage.
1953 Description: reply_lock_interpret crash due to race with it and lock cancel.
1954 Details : Do not replay locks that are being cancelled. Do not reference
1955 locks by their address during replay, just by their handle.
1959 Frequency : When flocks are used.
1960 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1961 Details : Flock locks might destroy just granted lock if it could be merged
1962 with another existing flock, this is done in completion handler,
1963 so teach ldlm_cli_enquque_fini that this is a valid case for
1969 Description: Oops in read and write path when failing to allocate lock.
1970 Details : Check if lock allocation failed and return error back.
1974 Description: lstripe command fails for valid OST index
1975 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1976 instead of lov->desc.ld_active_tgt_count.
1980 Description: sanity.sh failed test 103
1981 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1986 Description: df doesn't work properly if diskfs blocksize != 4K
1987 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1991 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1992 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1994 Severity : enhancement
1996 Description: Adaptive timeouts
1997 Details : RPC timeouts adapt to changing server load and network
1998 conditions to reduce resend attempts and improve recovery time.
2002 Description: llapi_file_create() does not allow some changes
2003 Details : add llapi_file_open() that allows specifying the mode and
2004 open flags, and also returns an open file handle.
2008 Description: merge and cleanup kernel patches.
2009 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2013 Description: Add journal checksum support.(Kernel part)
2014 Details : The journal checksum feature adds two new flags i.e
2015 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2016 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2017 indicates that the commit block contains the checksum for
2018 the blocks described by the descriptor blocks. Now commit
2019 record can be sent to disk without waiting for descriptor
2020 blocks to be written to disk. This behavior is controlled
2021 using JBD2_FEATURE_ASYNC_COMMIT flag.
2025 Description: OSS needs mutliple precreate threads
2026 Details : Add ability to start more than one create thread per OSS.
2030 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2031 Details : execute lfs setstripe on client
2035 Description: mds_obd_create error creating tmp object
2036 Details : When the user sets quota on root, llog will be affected and can't
2037 create files and write files.
2040 Frequency : Always on ia64 patchless client, and possibly others.
2042 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2043 Details : This allows the patchless client to be loaded on architectures
2044 without this export.
2048 Description: RedHat Update kernel for RHEL5
2049 Details : Add the kernel config file for RHEL5.
2053 Description: RedHat Update kernel for RHEL5
2054 Details : Modify the kernel config file more closer RHEL5.
2058 Description: Build failure against Centos5 (RHEL5)
2059 Details : Define PAGE_SIZE when it isn't present.
2063 Description: test_7 on sanity-quota.sh can't be used on separate machines
2064 Details : add facet to handle it
2068 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2069 Details : with NFS, the anon dentry's parent was set to itself in
2070 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2071 corresponding dentry other than use rec->ur_name.
2073 Severity : enhancement
2075 Description: lfs setstripe enhancement
2076 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2080 Description: enable data checksumming by default
2081 Details : enable checksum by default, allow --disable-checksum
2082 configure option and "-o nochecksum" mount option. Checksums
2083 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2087 Description: lustre support for RHEL5
2088 Details : Add support for RHEL5.
2092 Description: Client eviction due to failover config
2093 Details : after a connection loss, the lustre client should attempt to
2094 reconnect to the last active server first before trying the
2095 other potential connections.
2099 Description: when mds and osts use different quota unit(32bit and 64bit),
2100 quota will be released repeatly.
2101 Details : void sending multiple quota reqs to mds, which will keep the
2102 status between the reqs.
2106 Description: osts not allocated evenly to files
2107 Details : change the condition to increase offset_idx
2110 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2112 Description: Data corruption for OSTs that are formatted larger than 2TB
2114 Details : When generating the bio request for lustre file writes the
2115 sector number would overflow a temporary variable before being
2116 used for the IO. The data reads correctly from Lustre (which
2117 will overflow in a similar manner) but other file data or
2118 filesystem metadata may be corrupted in some cases.
2122 Description: Tune the kernel for good SCSI performance.
2123 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2124 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2129 Description: rewrite ext2-derived code in obdclass/uuid.c
2130 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2131 from scratch preserving functionality.
2135 Description: "lfs find -obd UUID" prints directories
2136 Details : "lfs find -obd UUID" will return all directory names instead
2137 of just file names. It is incorrect because the directories
2138 do not reside on the OSTs.
2142 Description: use i_size_read and i_size_write in 2.6 port
2143 Details : replace inode->i_size access with i_size_read/write()
2146 Frequency : when removing large files
2148 Description: scheduling issue during removal of large Lustre files
2149 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2150 It causes scheduling issues when removing large files (17TB in the
2154 Frequency : only with liblustre clients on XT3
2156 Description: evictions taking too long
2157 Details : allow llrd to evict clients directly on OSTs
2160 Frequency : only on ppc
2162 Description: /proc/fs/lustre/devices broken on ppc
2163 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2164 We should make sure the type of 'index' is loff_t and then cast
2165 explicitly as needed below. Do not assign an explicitly cast
2169 Frequency : only for rhel5
2171 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2172 Details : Modify the target file & which_kernel.
2174 Severity : enhancement
2176 Description: omit set fsid for export NFS
2177 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2178 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2182 Frequency : Always, for kernels after 2.6.16
2183 Description: Fix warning idr_remove called for id=.. which is not allocated.
2184 Details : Last kernels save old s_dev before kill super and not allow
2185 to restore from callback - restore it before call kill_anon_super.
2189 Description: buffer overruns could theoretically occur
2190 Details : llapi_semantic_traverse() modifies the "path" argument by
2191 appending values to the end of the origin string, and a
2192 overrun may occur. Adding buffer overrun check in liblustreapi.
2196 Description: Fix error on 'ls .' at the top of the Lustre mount.
2197 Details : Don't revalidate dentry if it is a root dentry.
2201 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2202 Details : Modify vm-tunables-rhel4.patch.
2206 Description: Kernel config for 2.6.18-vanilla.
2207 Details : Modify targets/2.6-vanilla.target.in.
2208 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2209 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2210 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2211 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2215 Description: improve handling recoverable errors
2216 Details : if request processig with error which can be recoverable on server
2217 request should be resend, otherwise page released from cache and
2222 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2223 Details : Modify target file & which_patch.
2224 A flaw was found in the IA32 system call emulation provided
2225 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2226 value could be stored in the %RAX register, which could trigger an
2227 out-of-bounds system call table access. An untrusted local user
2228 could exploit this flaw to run code in the kernel
2229 (ie a root privilege escalation). (CVE-2007-4573).
2233 Description: O_DIRECT bypasses client statistics.
2234 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2235 read_ahead_stats were updated. copied the stats section from
2236 osc_send_oap_rpc() into async_internal().
2240 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2244 Description: change order of libsysio includes
2245 Details : '#include sysio.h' should always come before '#include xtio.h'
2249 Description: handle "serious error: objid * already exists" more gracefully
2250 Details : If LAST_ID value on disk is smaller than the objects existing in
2251 the O/0/d* directories, it indicates disk corruption and causes an
2252 LBUG(). If the object is 0-length, then we should use the existing
2253 object. This will help to avoid a full fsck in most cases.
2255 Severity : enhancement
2257 Description: adapt the lustre_config script to support the upgrade case
2258 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2259 server targets to 1.6 in parallel.
2263 Description: To avoid grant space > avaible space when the disk is almost
2264 full. Without this patch you might see the error "grant XXXX >
2265 available" or some LBUG about grant, when the disk is almost
2267 Details : In filter_check_grant, for non_grant cache write, we should
2268 check the left space by if (*left > ungranted + bytes), instead
2269 of (*left > ungranted), because only we are sure the left space
2270 is enough for another "bytes", then the ungrant space should be
2271 increase. In client, we should update cl_avail_grant only there
2272 is OBD_MD_FLGRANT in the reply.
2276 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2277 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2278 described in CVE-2007-4573. This problem could allow untrusted
2279 local users to gain root access.
2282 Frequency : when using O_DIRECT and quotas
2284 Description: Incorrect file ownership on O_DIRECT output files
2285 Details : block usage reported by 'lfs quota' does not take into account
2286 files that have been written with O_DIRECT.
2291 Description: touch file failed when fs is not full
2292 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2293 otherwise we can get ENOSP while fs is not full.
2297 Description: parallel lock callbacks
2298 Details : Instead of sending blocking and completion callbacks as separated
2299 requests, adding them to a set and sending in parallel.
2302 Frequency : only for Cray XT3
2303 Bugzilla : 12829/13455
2304 Description: Changing primary group doesn't change the group lustre assigns to
2306 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2307 overridden with the primary group provided by the group upcall,
2308 whereas the supplied fsgid can be trusted if it is in the list of
2309 supplementary groups returned by the group upcall.
2311 Severity : enhancement
2313 Description: Allow masking D_WARNING, D_ERROR messages from console
2314 Details : Console messages can now be disabled via lnet.printk.
2318 Description: User code with malformed file open parameter crashes client node
2319 Details : Before packing join_file req, all the related reference should be
2320 checked carefully in case some malformed flags cause fake
2321 join_file req on client.
2325 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2326 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2327 mode is equal to the granted mode. However, it should also check
2328 that LDLM_FL_CP_REQD is not set.
2332 Description: Heavy nfs access might result in deadlocks
2333 Details : After ELC code landed, it is now improper to enqueue any mds
2334 locks under och_sem, because enqueue might want to decide to
2335 cancel open locks for same inode we are holding och_sem for.
2339 Description: Client eviction while running blogbench
2340 Details : A lot of unlink operations with concurrent I/O can lead to a
2341 deadlock causing evictions. To address the problem, the number of
2342 oustanding OST_DESTROY requests is now throttled to
2343 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2344 callbacks are processed in priority.
2348 Description: enable ACLs on MDS by default
2349 Details : ACLs must be enabled on MDS by default.
2352 Frequency : PPC/PPC64 only
2354 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2355 Details : fix duplicated definitions between asm-ppc64/types.h and
2356 lustre_types.h on PPC.
2359 Frequency : PPC/PPC64 only
2361 Description: asm-ppc/segment.h does not exist
2362 Details : fix compile issue on PPC.
2366 Description: better handle error messages in extents code
2369 Frequency : RHEL4 only
2371 Description: mkfs is very slow on IA64/RHEL4
2372 Details : A performance regression has been discovered in the MPT Fusion
2373 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2374 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2375 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2377 Severity : enhancement
2379 Description: SNMP support enhancement
2380 Details : Adding total number of sampled request for an MDS node in snmp
2383 Severity : enhancement
2385 Description: Optimize ldlm waiting list processing for PR extent locks
2386 Details : When processing waiting list for read extent lock and meeting read
2387 lock that is same or wider to it that is not contended, skip
2388 processing rest of the list and immediatelly return current
2389 status of conflictness, since we are guaranteed there are no
2390 conflicting locks in the rest of the list.
2394 Description: Time out and refuse to reconnect
2395 Details : When the failover node is the primary node, it is possible
2396 to have two identical connections in imp_conn_list. We must
2397 compare not conn's pointers but NIDs, otherwise we can defeat
2398 connection throttling.
2402 Description: port llog fixes from b1_6 into HEAD
2403 Details : Port llog reference couting and some llog cleanups from b1_6
2404 (bug 10800) into HEAD, for protect from panic and access to already
2405 free llog structures.
2409 Description: Detect stride IO mode in read-ahead
2410 Details : When a client does stride read, read-ahead should detect that and
2411 read-ahead pages according to the detected stride pattern.
2415 Description: data checksumming impacts single node performance
2416 Details : add support for several checksum algorithm. Currently, only CRC32
2417 and Adler-32 are supported. The checksum type can be changed on
2418 the fly via /proc/fs/lustre/osc/*/checksum_type.
2422 Description: use adler32 for page checksums
2423 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2428 Description: build for x2 fails
2429 Details : fix compile issue on Cray systems.
2433 Description: Properly match for duplicate locks
2434 Details : Due to different lock order from skiplists code, we need to
2435 traverse entire list for now
2438 Frequency : only on PPC/SLES10
2440 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2441 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2442 be defined. Add a hack in mkfs_lustre.c to work around this compile
2447 Description: LASSERT on MDS when client holding flock lock dies
2448 Details : ldlm pool logic depends on number of granted locks equal to
2449 number of released locks which is not true for flock locks, so
2450 just exclude such locks from consideration.
2454 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2455 Details : Use fsfilt_sync() for both the whole filesystem sync and
2456 individual file sync to eliminate dangerous inode locking
2457 with I_LOCK that can lead to a deadlock.
2461 Description: performance in 1.6.3
2462 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2463 sure that 1MB requests are not fragmented by the block layer.
2465 Severity : enhancement
2467 Description: organize the server-side client stats on per-nid basis
2468 Details : Change the structure of stats under obdfilter and mds to
2476 The "uuid"s file would list the uuids of _active_ exports.
2477 And the clear entry is to clear all stats and stale nids.
2479 Severity : enhancement
2481 Description: eliminate client locks in face of contention
2482 Details : file contention detection and lockless i/o implementation
2483 for contended files.
2487 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2490 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2492 Description: MDS may lose file striping (and hence file data) in some cases
2493 Details : If there are additional extended attributes stored on the MDS,
2494 in particular ACLs, SELinux, or user attributes (if user_xattr
2495 is specified for the client mount options) then there is a risk
2496 of attribute loss. Additionally, the Lustre file striping
2497 needs to be larger than default (e.g. striped over all OSTs),
2498 and an additional attribute must be stored initially in the
2499 inode and then increase in size enough to be moved to the
2500 external attribute block (e.g. ACL growing in size) for file
2505 Description: skiplist implementation simplification
2506 Details : skiplists are used to group compatible locks on granted list
2507 that was implemented as tracking first and last lock of each
2508 lock group the patch changes that to using doubly linked lists
2512 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2513 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2514 because LASSERT on both the data supplied by a client, and the
2515 data on disk is dangerous and incorrect.
2517 Severity : enhancement
2519 Description: Slow truncate/writes to huge files at high offsets.
2520 Details : Directly associate cached pages to lock that protect those pages,
2521 this allows us to quickly find what pages to write and remove
2522 once lock callback is received.
2526 Description: more ldlm soft lockups
2527 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2528 starve other threads from the resource lock for a long time in
2529 case of long waiting queue, so change the debug level from
2530 D_OTHER to the less frequently used D_INFO.
2532 Severity : enhancement
2534 Description: add -gid, -group, -uid, -user options to lfs find
2538 Description: Hung threads in invalidate_inode_pages2_range
2539 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2540 once one is completed. As a result, some RPCs are stuck in the
2541 queue and are never sent.
2545 Description: filter threads hungs on waiting journal commit
2546 Details : Cleanup filter group llog code, then only filter group llog will
2547 be only created in the MDS/OST syncing process.
2551 Description: Procfs and llog threads access destoryed import sometimes.
2552 Details : Sync the import destoryed process with procfs and llog threads by
2553 the import refcount and semaphore.
2555 Severity : enhancement
2557 Description: openlock cache of b1_6 port to HEAD
2562 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2563 Details : If insertion of an extent fails, then discard the inode
2564 preallocation and free data blocks else it can lead to duplicate
2569 Description: don't always update ctime in ext3_xattr_set_handle()
2570 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2571 In some cases the ctime should not be updated, for example for
2572 2.0->1.8 compatibility it is necessary to delete an xattr and it
2573 should not update the ctime.
2577 Bugzilla : 15713/16362
2578 Description: Assertion in iopen_connect_dentry in 1.6.3
2579 Details : looking up an inode via iopen with the wrong generation number can
2580 populate the dcache with a disconneced dentry while the inode
2581 number is in the process of being reallocated. This causes an
2582 assertion failure in iopen since the inode's dentry list contains
2583 both a connected and disconnected dentry.
2587 Description: assertion failure in ldlm_handle2lock()
2588 Details : fix a race between class_handle_unhash() and class_handle2object()
2589 introduced in lustre 1.6.5 by bug 13622.
2594 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2595 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2600 Description: oops in sd_iostats_seq_show()
2601 Details : unloading/reloading the scsi low level driver triggers a kernel
2602 bug when trying to access the sd iostat file.
2607 Description: Kernel panics during QLogic driver reload
2608 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2609 patch, causing memory corruption.
2614 Description: journal_dev option does not work in b1_6
2615 Details : pass mount option during pre-mount.
2617 Severity : enhancement
2619 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2620 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2621 extent information of a file. It can be used to map logical blocks
2622 in a file to physical blocks in the block device.
2626 Description: LDLM soft lockups - improvement
2627 Details : It is be possible to send the lock handle along with each read
2628 or write request because the client is already doing a lock match
2629 itself so there isn't any reason the OST should have to re-do that
2633 Frequency : only on Cray X2
2635 Description: X2 build failures
2636 Details : fix build failures on Cray X2.
2640 Description: xid & resent requests
2641 Details : Initialize RPC XID from clock at startup (randomly if clock is
2644 Severity : enhancement
2646 Description: Add lustre_start utility to start or stop multiple Lustre servers
2651 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2652 Details : In case of memory pressure, list_del() can be called twice on
2653 req->rq_history_list, causing a kernel oops.
2657 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2658 Details : fix stack overflow in the distributed lock manager by defering
2659 export eviction after a failed AST to the elt thread instead of
2660 handling it in the dlm interpret routine.
2664 Description: Convert some comments to new format.
2665 Details : Update documenting comments to match doxygen conventions.
2669 Description: Grammar fixes.
2670 Details : A couple of trivial spelling fixes.
2674 Description: OSD_COUNTERS-mandatory
2675 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2676 to simplify the code.
2680 Description: simplify cmm_device freeing logic.
2681 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2685 Description: Add lockdep support to dt_object_operations locking interface.
2686 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2687 parameter indicating lock ordering. Update mdd code to use new
2692 Description: Introduce failloc constants for lockless IO tests.
2693 Details : Add two new failloc constants to test lockless IO. Only one of
2694 them in implemented---another is checked in yet to be landed
2699 Description: Add lockdep support for inode mutex.
2700 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2701 in the situations where Lustre has to lock more than one inode
2706 Description: Add optional invariants checking support.
2707 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2708 configure switch. This macro is to be used for consistency and
2709 sanity checks that are too expensive to be left in `production'
2714 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2715 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2720 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2721 Details : Use spin_lock_nested() in (the only) situation where more than
2722 one ldlm_lock is locked simultaneously. Also, fix possible
2723 dead-lock in ldlm_lock_change_resource() by enforcing particular
2728 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2729 Details : Instead of specifying each ldlm_lock call-back through separate
2730 parameter, wrap them into struct ldlm_callback_suite.
2734 Description: Kill join_lru obd method and its callers.
2735 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2736 in memory---a requirement implemented through ->o_join_lru() obd
2737 method in HEAD. Remove this method and its users.
2741 Description: Add asynchronous ldlm ENQUEUE completion handler.
2742 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2743 case that stock ldlm_completion_ast() cannot handle as it waits
2744 until lock is granted. Introduce new ldlm_completion_ast_async()
2745 for this. Also comment ldlm_completion_ast().
2749 Description: ldlm_error <-> errno conversion.
2750 Details : Add functions to map (rather arbitrary) between LDLM error codes
2751 and standard errno values. CLIO needs this to prevent LDLM specific
2752 constants from escaping ldlm and osc.
2756 Description: Kill unused ldlm_handle2lock_ns() function.
2757 Details : Kill unused ldlm_handle2lock_ns() function.
2761 Description: Add lu_ref support to ldlm_lock
2762 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2763 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2764 and ldlm_resource. LDLM interface has to be changed, because code that
2765 releases a reference on a lock, has to "know" what reference this is.
2766 In the most frequent case
2768 lock = ldlm_handle2lock(handle);
2770 LDLM_LOCK_PUT(lock);
2772 no changes are required. When any other reference (received _not_ from
2773 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2774 instead of LDLM_LOCK_PUT().
2776 Arguably, changes are pervasive, and interface requires some discipline
2777 for proper use. On the other hand, it was very instrumental in finding
2778 a few leaked lock references.
2782 Description: Add ldlm_lock_addref_try().
2783 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2784 attempts to addref a lock that might be being canceled
2789 Description: Add ldlm_weigh_callback().
2790 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2791 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2792 hard-coded `number of pages' logic.
2796 Description: Add lockdep annotations to llog code.
2797 Details : Use appropriately tagged _nested() locking calls in the places
2798 where llog takes more than one ->lgh_lock lock.
2802 Description: Add loi_kms_set().
2803 Details : Wrap kms updates into a helper function.
2807 Description: Constify instances of struct lsm_operations.
2808 Details : Constify instances of struct lsm_operations.
2812 Description: lu_conf support.
2813 Details : On a server, a file system object is uniquely identified
2814 by a fid, which is sufficient to locate and load all object
2815 state (inode). On a client, on the other hand, more data are
2816 necessary instantiate an object. Change lu_object_find() and
2817 friends to take additional `lu_conf' argument describing object.
2818 Typically this includes layout information.
2822 Description: lu_context fixes.
2823 Details : Introduce new lu_context functions that are needed on the client
2824 side, where some system threads (ptlrpcd) are shared by multiple
2825 modules, and so cannot be stopped during module shutdown.
2829 Description: Add start and stop methods to lu_device_type_operations.
2830 Details : Introduce two new methods in lu_device_type_operations, that are
2831 invoked when first instance of a given type is created and last one
2832 is destroyed respectively. This is need by CLIO.
2836 Description: Add lu_ref support to struct lu_device.
2837 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2838 track leaked references.
2842 Description: Introduce lu_kmem_descr.
2843 Details : lu_kmem_descr and its companion interface allow to create
2844 and destroy a number of kmem caches at once.
2848 Description: Fix lu_object finalization race.
2849 Details : Fix a race between lu_object_find() finding an object and its
2850 concurrent finalization. This race is (most likely) not possible
2851 on the server, but might happen on the client.
2855 Description: Introduce lu_ref interface.
2856 Details : lu_ref is a debugging module allowing to track references to
2857 a given object. It is quite cpu expensive, and has to be
2858 explicitly enabled with --enable-lu_ref. See usage description
2863 Description: Factor lu_site procfs stats into a separate function.
2864 Details : Separate lu_site stats printing code into a separate function
2865 that can be reused on a client.
2869 Description: Constify instances of struct {lu,dt,md}_device_operations.
2870 Details : Constify instances of struct {lu,dt,md}_device_operations.
2874 Description: Introduce struct md_site and move meta-data specific parts of
2875 struct lu_site here.
2876 Details : Move md-specific fields out of struct lu_site into special struct
2877 md_site, so that lu_site can be used on a client.
2881 Description: Kill mdd_lov_destroy().
2882 Details : Remove unused mdd code.
2886 Description: Add st_block checking to multistat.c.
2887 Details : Add st_block checking to multistat.c.
2891 Description: Add lu_ref support to struct obd_device.
2892 Details : Add lu_ref tracking to obd_device.
2896 Description: Kill obd_set_fail_loc().
2897 Details : Remove unused code.
2901 Description: Add special type for ptlrpc_request interpret functions.
2902 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2903 there. Actual usage will be in CLIO.
2907 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2908 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2909 doesn't work with lockdep.
2913 Description: Add rwv.c test program.
2914 Details : New testing program exercising readv(2) and writev(2) (Qian).
2918 Description: Add sendfile.c test program.
2919 Details : New testing program exercising sendfile(2) (Jay).
2923 Description: Ratelimit a message that can be very frequent.
2924 Details : Ratelimit a memory allocation failure message that can
2929 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2930 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2934 Description: Liblustre build fixes.
2935 Details : Liblustre build fixes.
2939 Description: libcfs: add cfs_{need,cond}_resched() interface.
2940 Details : libcfs: add cfs_{need,cond}_resched() definition and
2941 implementations for Linux, NT, and liblustre.
2943 Severity : enhancement
2945 Description: More exported tunables for mballoc
2946 Details : Add support for tunable preallocation window and new tunables for
2947 large/small requests
2951 Description: Detect corruption of block bitmap and checking for preallocations
2952 Details : Checks validity of on-disk block bitmap. Also it does better
2953 checking of number of applied preallocations. When corruption is
2954 found, it turns filesystem readonly to prevent further corruptions.
2958 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2959 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2960 window to 0. If it does hit cache miss, and read-pattern is still
2961 stride-io mode, does not reset the stride window, but also does not
2962 increase the stride window length in this case.
2966 Frequency : only for big-endian servers
2967 Description: Check if system is big-endian while mounting fs with extents feature
2968 Details : Mounting a filesystem with extents feature will fail on big-endian
2969 systems since ext3-based ldiskfs is not supported on big-endian
2970 systems. This can be over-riden with "bigendian_extents" mount option.
2972 Severity : enhancement
2974 Description: The root squash functionality
2975 Details : A security feature, which is to prevent users from being able
2976 to mount lustre on their desktop, run as root, and delete
2977 all of the files in the filesystem. The goal is accomplished by
2978 remapping user id (UID) and group id (GID) of the root user to
2979 a UID and GID specified by the system administartor via Lustre
2980 configuration management server (MGS). The functionality also
2981 allows to specify sets of clients for which the remapping does
2986 Description: Excessive recovery window
2987 Details : With AT enabled, the recovery window can be excessively long (6000+
2988 seconds). To address this problem, we no longer use
2989 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2990 timeout no longer depends on the service time, it is set to
2991 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2992 time via pb_service_time.
2996 Description: Watchdog triggered on MDS failover
2997 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2998 the OSTs know that the MDS "UUID" can be reused for the same export
2999 from a different NID, so we do not need to wait for the export to be
3003 Frequency : rare, only if using MMP with Linux RAID
3005 Description: MMP doesn't work with Linux RAID
3006 Details : While using HA for Lustre servers with Linux RAID, it is possible
3007 that MMP will not detect multiple mounts. To make this work we
3008 need to unplug the device queue in RAID when the MMP block is being
3009 written. Also while reading the MMP block, we should read it from
3010 disk and not the cached one.
3012 Severity : enhancement
3014 Description: open file using fid
3015 Details : A file can be opened using just its fid, like
3016 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3019 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3022 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3023 Details : add proper locking for creation of jbd_4k slab cache
3027 Description: MMP check in ext3_remount() fails without displaying any error
3028 Details : When multiple mount protection fails during remount, proper error
3031 Severity : enhancement
3033 Description: Allow stripe size to be up to 4G-64k
3034 Details : Fix math logic to allow large stripe sizes.
3038 Description: add check for >8TB ldiskfs filesystems
3039 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3040 Don't allow >8TB ldiskfs filesystems to be mounted without
3041 force_over_8tb mount option
3046 Descriptoin: OSS complains "too many blocks" and crashes
3047 Details : In filter_commitrw_write():
3048 1. lquota_chkquota();
3049 2. obdo_from_inode();
3050 3. lquota_pending_commit();
3051 "2." has changed uid and gid in this case, then caused
3052 lquota_chkquota() and lquota_pending_commit() to use
3053 different uid and gid. That is the root of the bug.
3057 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3058 Details : Detect and print out the 'stripe count = -1' case properly.
3062 Description: Out or order replies might be lost on replay
3063 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3064 request with tid smaller than one currently being added, add it
3065 to the start, not end of the list.
3067 --------------------------------------------------------------------------------
3069 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3071 * Support for kernels:
3072 2.6.5-7.283 (SLES 9),
3073 2.6.9-55.EL (RHEL 4),
3074 2.6.16.46-0.14 (SLES 10),
3075 2.6.18.8 vanilla (kernel.org)
3076 * Client support for unpatched kernels:
3077 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3078 2.6.16 - 2.6.22 vanilla (kernel.org)
3079 * Due to recently discovered recovery problems, we do not recommend
3080 using patchless RHEL 4 clients with this or any earlier release.
3081 * Recommended e2fsprogs version: 1.39.cfs8
3082 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3083 * Starting with this release, the ldiskfs backing filesystem required
3084 by Lustre is now in its own package, lustre-ldiskfs. This package
3085 should be installed. It is versioned separately from Lustre and
3086 may be released separately in future.
3090 Description: block reactivating mgc import until all deactivates complete
3091 Details : Fix race when failing back MDT/MGS to itself (testing)
3094 Frequency : at statup only
3096 Description: mds_lov_synchronize race leads to various problems
3097 Details : simultaneous MDT->OST connections at startup can cause the
3098 sync to abort, leaving the OSC in a bad state.
3100 Severity : enhancement
3102 Description: add optional extra BUILD_VERSION info
3103 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3104 to override the lustre version.
3107 Frequency : 2.6.18 servers only
3109 Description: ll_kern_mount() doesn't release the module reference
3110 Details : The ldiskfs module reference count never drops down to 0
3111 because ll_kern_mount() doesn't release the module reference.
3116 Description: server LBUG when using old ost_num_threads parameter
3117 Details : Accept the old ost_num_threads parameter but warn that it
3118 is deprecated, and fix an off-by-one error that caused an LBUG.
3123 Description: Transient SCSI error results in persistent IO issue
3124 Details : iobuf->dr_error is not reinitialized to 0 between two
3128 Frequency : sometimes when underlying device returns I/O errors
3130 Description: OSTs not going read-only during write failures
3131 Details : OSTs are not remounted read-only when the journal commit threads
3132 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3133 instead of the ext3 wrappers.
3137 Description: poor connect scaling with increasing client count
3138 Details : Don't run filter_grant_sanity_check for more than 100 exports
3139 to improve scaling for large numbers of clients.
3142 Frequency : SLES10 only
3144 Description: sanity-quota.sh quotacheck failed: rc = -22
3145 Details : Quotas cannot be enabled on SLES10.
3148 Frequency : liblustre clients only
3150 Description: getdirentries does not give error when run on compute nodes
3151 Details : getdirentries does not fail when the size specified as an argument
3152 is too small to contain at least one entry
3154 Severity : enhancement
3156 Description: Add LNET router traceability for debug purposes
3157 Details : If a checksum failure occurs with a router as part of the
3158 IO path, the NID of the last router that forwarded the bulk data
3159 is printed so it can be identified.
3164 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3165 Details : Due to a race condition, liblustre clients were occasionally
3166 evicted incorrectly.
3168 Severity : enhancement
3170 Description: lfs setstripe use optional parameters instead of postional
3173 Severity : enhancement
3175 Description: Nanosecond timestamp support for ldiskfs
3176 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3177 resolution timestamps. There is not yet support for this at
3178 the Lustre filesystem level.
3181 Frequency : during server recovery
3183 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3184 Details : request with rq_no_resend flag not awake l_wait_event if they get
3188 Frequency : nfs export on patchless client
3190 Description: connectathon hang when test nfs export over patchless client
3191 Details : Disconnected dentry cannot be found with lookup, so we do not need
3192 to unhash it or make it invalid
3195 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3196 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3197 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3202 Description: open req refcounting wrong on reconnect
3203 Details : If reconnect happened between getting open reply from server and
3204 call to mdc_set_replay_data in ll_file_open, we will schedule
3205 replay for unreferenced request that we are about to free.
3206 Subsequent close will crash in variety of ways.
3207 Check that request is still eligible for replay in
3208 mdc_set_replay_data().
3213 Description: disable writes to filesystem when reading health_check file
3214 Details : the default for reading the health_check proc file has changed
3215 to NOT do a journal transaction and write to disk, because this
3216 can cause reads of the /proc file to hang and block HA state
3217 checking on a healthy but otherwise heavily loaded system. It
3218 is possible to return to the previous behaviour during configure
3219 with --enable-health-write.
3221 Severity : enhancement
3223 Description: 64-bit inode version
3224 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3225 to the inode. This will be required for version-based recovery.
3230 Description: MDS fails to start if a duplicate client export is detected
3231 Details : in some rare cases it was possible for a client to connect to
3232 an MDS multiple times. Upon recovery the MDS would detect this
3233 and fail during startup. Handle this more gracefully.
3235 Severity : enhancement
3237 Description: Add -o localflock option to simulate old noflock
3239 Details : This will achieve local-only flock/fcntl locks
3245 Description: log_commit_thread vs filter_destroy race leads to crash
3246 Details : Take import reference before releasing llog record semaphore
3251 Description: Wrong request locking in request set processing
3252 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3253 imp_delayed_list, in this place should be used imp_lock.
3256 Frequency : when reconnection
3258 Description: Grant Leak when osc reconnect to OST
3259 Details : When osc reconnect ost, OST(filter) should check whether it
3260 should grant more space to client by comparing fed_grant and
3261 cl_avail_grant, and return the granted space to client instead
3262 of "new granted" space, because client will call osc_init_grant
3263 to update the client grant space info.
3266 Frequency : when client reconnect to OST
3268 Description: Grant Leak when osc do resend and replay bulk write
3269 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3270 bulk write request, because the grant info while be sync between
3271 OSC and OST when reconnect, and we should ignore the grant info
3272 these of resend/replay write req.
3277 Description: Grant space more than available left space sometimes.
3278 Details : When then OST is about to be full, if two bulk writing from
3279 different clients came to OST. Accord the avaliable space of the
3280 OST, the first req should be permitted, and the second one
3281 should be denied by ENOSPC. But if the seconde arrived before
3282 the first one is committed. The OST might wrongly permit second
3283 writing, which will cause grant space > available space.
3286 Frequency : when client is evicted
3288 Description: Grant might be wrongly erased when osc is evicted by OST
3289 Details : when the import is evicted by server, it will fork another
3290 thread ptlrpc_invalidate_import_thread to invalidate the
3291 import, where the grant will be set to 0. While the original
3292 thread will update the grant it got when connecting. So if
3293 the former happened latter, the grant will be wrongly errased
3294 because of this race.
3299 Description: Checking Stale with correct fid
3300 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3301 to check whether it is stale, because sometimes, we want the
3302 enqueue happened anyway, and op_data.fid2 will not be initialized.
3304 Severity : enhancement
3306 Description: update patchless client
3307 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3310 Frequency : only with 2.4 kernel
3312 Description: random memory corruption
3313 Details : size of struct ll_inode_info is to big for union inode.u and this
3314 can be cause of random memory corruption.
3319 Description: Memory leak in recovery
3320 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3321 It should also check obd_fail before fsfilt_start, otherwise if
3322 fsfilt_start return -EROFS,(failover mds during mds recovery).
3323 then the req will return with repmsg->transno = 0 and rc = EROFS.
3324 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3325 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3326 be freed no matter whether fsfilt_commit success or not.
3329 Frequency : only with huge count clients
3331 Description: Prevents from taking the superblock lock in llap_from_page for
3333 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3334 for llap_from_page call in ll_removepage prevents from taking the
3335 superblock lock for a soon died page.
3340 Description: Not check open intent error before release open handle
3341 Details : in some rare cases, the open intent error is not checked before
3342 release open handle, which may cause
3343 ASSERTION(open_req->rq_transno != 0), because it tries to release
3344 the failed open handle.
3349 Description: Set cat log bitmap only after create log success.
3350 Details : in some rare cases, the cat log bitmap is set too early. and it
3351 should be set only after create log success.
3355 Description: Accessing a block bevice can re-enable I/O when Lustre is
3356 tearing down a device.
3357 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3362 Description: service threads may hog cpus when there are a lot of requests
3363 Details : Insert cond_resched to give other threads a chance to use some CPU
3368 Description: the cat log was not initialized in recovery
3369 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3370 unlink log on mds will not be initialized until mds post
3371 recovery. And also in mds post recovery, the unlink log will
3372 initialization will be done asynchronausly, so there will be race
3373 between add unlink log and unlink log initialization.
3377 Description: brw_stats were being printed incorrectly
3378 Details : brw_stats were being printed as log2 but all of them were not
3379 recorded as log2. Also remove some code duplication arising from
3380 filter_tally_{read,write}.
3384 Frequency : rare, only in recovery.
3385 Description: ASSERTION(req->rq_type != LI_POISON) failed
3386 Details : imp_lock should be held while iterating over imp_sending_list for
3387 prevent destroy request after get timeout in ptlrpc_queue_wait.
3391 Description: replay-single.sh test 52 fails
3392 Details : A lock's skiplist need to be cleanup when it being unlinked
3393 from its resource list.
3397 Description: Short directio read returns full requested size rather than
3399 Details : Direct I/O operations should return actual amount of bytes
3400 transferred rather than requested size.
3402 Severity : enhancement
3404 Description: metadata RPC reduction (e.g. for rm performance)
3405 Details : decrease the amount of synchronous RPC between clients and servers
3406 by canceling conflicing lock before the operation on the client
3407 and packing thier handles into the main operation RPC to server.
3409 Severity : enhancement
3411 Description: Async OSC create to avoid the blocking unnecessarily.
3412 Details : If an OST has no remaining object, system will block on the
3413 creation when it needs to create a new object on this OST. Now,
3414 ways use pre-created objects when available, instead of blocking on
3415 an empty osc while others are not empty. If we must block, we
3416 block for the shortest possible period of time.
3420 Description: improve handling recoverable errors
3421 Details : if request processig with error which can be recoverable on server
3422 request should be resend, otherwise page released from cache and
3425 Severity : enhancement
3427 Description: refine locking for avoid write wrong info into lov_objid file
3428 Details : fix possible races with add new target and write/update data in
3431 --------------------------------------------------------------------------------
3433 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3438 Frequency : on some architectures
3440 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3441 Details : Uninitialized flags sometimes cause configuration commands to
3445 Frequency : patchless clients only
3447 Description: missing __iget() symbol export
3448 Details : The __iget() symbol export is missing. To avoid the need for
3449 this on patchless clients the deathrow inode reaper is turned
3450 off, and we depend on the VM to clean up old inodes. This
3451 dependency was during via the fix for bug 12181.
3453 --------------------------------------------------------------------------------
3455 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3457 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3458 INTEROPERATE with older versions automatically. In many cases a
3459 special upgrade step is needed. Please read the
3460 user documentation before upgrading any part of a live system.
3461 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3462 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3463 * WARNING: Lustre configuration and startup changes are required with
3464 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3468 Severity : enhancement
3470 Description: MountConf
3471 Details : Lustre configuration is now managed via mkfs and mount
3472 commands instead of lmc and lconf. New obd types (MGS, MGC)
3473 are added for dynamic configuration management. See
3474 https://mail.clusterfs.com/wikis/lustre/MountConf for
3477 Severity : enhancement
3479 Description: dynamic OST addition
3480 Details : OSTs can now be added to a live filesystem
3482 Severity : enhancement
3484 Description: startup order invariance
3485 Details : MDTs and OSTs can be started in any order. Clients only
3486 require the MDT to complete startup.
3488 Severity : enhancement
3490 Description: parallel, asynchronous orphan cleanup
3491 Details : orphan cleanup is now performed in separate threads for each
3492 OST, allowing parallel non-blocking operation.
3494 Severity : enhancement
3496 Description: optimized stripe assignment
3497 Details : stripe assignments are now made based on ost space available,
3498 ost previous usage, and OSS previous usage, in order to try
3499 to optimize storage space and networking resources.
3501 Severity : enhancement
3503 Description: Permanently set tunables
3504 Details : All writable /proc/fs/lustre tunables can now be permanently
3505 set on a per-server basis, at mkfs time or on a live system.
3507 Severity : enhancement
3509 Description: Lustre message v2
3510 Details : Add lustre message format v2.
3512 Severity : enhancement
3514 Description: client OST exclusion list
3515 Details : Clients can be started with a list of OSTs that should be
3516 declared "inactive" for known non-responsive OSTs.
3520 Description: SPEC SFS validation failure on NFS v2 over lustre.
3521 Details : Changes the blocksize for regular files to be 2x RPC size,
3522 and not depend on stripe size.
3524 Severity : enhancement
3526 Description: Multiple MD RPCs in flight.
3527 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3528 To avoid overly-overloading MDS, introduce a limit on number of
3529 MDS RPCs in flight for a single client and add /proc controls
3530 to adjust this limit.
3532 Severity : enhancement
3534 Description: client read/write statistics
3535 Details : Add client read/write call usage stats for performance
3536 analysis of user processes.
3537 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3538 file access. extents_stats shows chunk size distribution.
3539 extents_stats_per_process show chunk size distribution per
3542 Severity : enhancement
3544 Description: mds statistics
3545 Details : Add detailed mds operations statistics in
3546 /proc/fs/lustre/mds/*/stats.
3550 Description: Failure of copying files with lustre special EAs.
3551 Details : Client side always return success for setxattr call for lustre
3552 special xattr (currently only "trusted.lov").
3557 Description: Refcount LNET uuids
3558 Details : The global LNET uuid list grew linearly with every startup;
3559 refcount repeated list entries instead of always adding to
3562 Severity : enhancement
3564 Description: Dynamic service threads
3565 Details : Within a small range, start extra service threads
3566 automatically when the request queue builds up.
3569 Frequency : mixed-endian client/server environments
3571 Description: mixed-endian crashes
3572 Details : The new msg_v2 system had some failures in mixed-endian
3575 Severity : enhancement
3577 Description: Easy OST removal
3578 Details : OSTs can be permanently deactivated with e.g. 'lctl
3579 conf_param lustre-OST0001.osc.active=0'
3581 Severity : enhancement
3583 Description: MGS proc entries
3584 Details : Added basic proc entries for the MGS showing what filesystems
3587 Severity : enhancement
3589 Description: provide MGS failover
3590 Details : Added config lock reacquisition after MGS server failover.
3592 Severity : enhancement
3594 Description: add Linux 2.4 support
3595 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3599 Description: a large application tries to do I/O to the same resource and dies
3600 in the middle of it.
3601 Details : Check the req->rq_arrival time after the call to
3602 ost_brw_lock_get(), but before we do anything about
3603 processing it & sending the BULK transfer request. This
3604 should help move old stale pending locks off the queue as
3605 quickly as obd_timeout.
3608 Frequency : when an incorrect nid is specified during startup
3610 Description: ptlrpc connect to non-existant node causes kernel crash
3611 Details : LNET can't be re-entered from an event callback, which
3612 happened when we expire a message after the export has been
3613 cleaned up. Instead, hand the zombie cleanup off to another
3616 Severity : enhancement
3618 Description: plain/inodebits lock performance improvement
3619 Details : Grouping plain/inodebits in granted list by their request modes
3620 and bits policy, thus improving the performance of search through
3624 Frequency : only if OST filesystem is corrupted
3626 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3627 Details : for a short time RPCs with bulk IO are in the replay list,
3628 but replay of bulk IOs is unimplemented. If the OST filesystem
3629 is corrupted due to disk cache incoherency and then replay is
3630 started it is possible to trip an assertion. Avoid putting
3631 committed RPCs into the replay list at all to avoid this issue.
3634 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3636 Description: System hang on startup
3637 Details : This bug allowed the liblustre (e.g. catamount) client to
3638 return to the app before handling all startup RPCs. This
3639 could leave the node unresponsive to lustre network traffic
3640 and manifested as a server ptllnd timeout.
3642 Severity : enhancement
3644 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3645 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3646 boolean to control whether to print peer debug info when a
3647 client's RPC times out.
3650 Frequency : only for kernels with patches from Lustre below 1.4.3
3652 Description: Remove old rdonly API
3653 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3656 Frequency : only for devices with external journals
3658 Description: Set external device read-only also
3659 Details : During a commanded failover stop, we set the disk device
3660 read-only while the server shuts down. We now also set any
3661 external journal device read-only at the same time.
3664 Frequency : when upgrading from 1.4 while trying to change parameters
3666 Description: The wrong (new) MDC name was used when setting parameters for
3667 upgraded MDT's. Also allows changing of OSC (and MDC)
3668 parameters if --writeconf is specified at tunefs upgrade time.
3671 Frequency : when setting specific ost indicies
3673 Description: QOS code breaks on skipped indicies
3674 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3675 created with --index need not be sequential.
3677 Severity : enhancement
3679 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3680 Details : The uninit_groups feature works in conjunction with the kernel
3681 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3682 up the pass1 processing of e2fsck. This is a read-only feature
3683 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3684 mount filesystems that have had this feature enabled.
3686 Severity : enhancement
3688 Description: Improve multi-block allocation algorithm to avoid fragmentation
3689 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3690 allocation locality and avoid filesystem fragmentation.
3692 ------------------------------------------------------------------------------
3694 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3696 * Support for kernels:
3697 2.4.21-47.0.1.EL (RHEL 3)
3698 2.6.5-7.283 (SLES 9)
3699 2.6.9-42.0.10.EL (RHEL 4)
3700 2.6.12.6 vanilla (kernel.org)
3701 2.6.16.27-0.9 (SLES 10)
3702 * Recommended e2fsprogs version: 1.39.cfs5
3704 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3708 Frequency : occasional, depends on client load and configuration
3709 Bugzilla : 12181, 12203
3710 Description: data loss for recently-modified files
3712 Details : In some cases it is possible that recently written or created
3713 files may not be written to disk in a timely manner (this should
3714 normally be within 30s unless client IO load is very high).
3715 The problem appears as zero-length files or files that are a
3716 multiple of 1MB in size after a client crash or client eviction
3717 that are missing data at the end of the file.
3719 This problem is more likely to be hit on clients where files are
3720 repeatedly created and unlinked in the same directory, clients
3721 have a large amount of RAM, have many CPUs, the filesystem has
3722 many OSTs, the clients are rebooted frequently, and/or the files
3723 are not accessed by other nodes after being written.
3725 The presence of the problem can be detected by looking at
3726 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3727 smaller than the second (nr_unused) then dirty files will not
3728 be flushed automatically to disk. "sync; sleep 10" should be
3729 run several times on the node before unmounting it to update
3730 Lustre (this is also safe to run on nodes without this problem).
3732 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3733 that can cause this same problem, so customers using that kernel
3734 also need to update the kernel in addition to Lustre. In order
3735 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3737 It is normal that files written just before a client crash (less
3738 than 30s) may not yet have been flushed to disk, even for local
3742 Frequency : frequent on thin XT3 nodes
3744 Description: UUID collision on thin XT3 Linux nodes
3745 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3746 randomly, since we relied on an insufficiently-seeded PRNG.
3751 Description: OSS hangs after "All ost request buffers busy"
3752 Details : A deadlock between quota and journal operations caused OSS
3753 hangs after printing "All ost request buffers busy."
3756 Frequency : always on liblustre builds
3758 Description: Cleanup compiler warnings on liblustre
3761 Frequency : always on liblustre builds on XT3
3763 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3767 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3768 > 32000 subdirectories
3769 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3770 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3771 subdirectory count crosses 32000. This will aid e2fsck to
3772 correctly handle more than 32000 subdirectories.
3775 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3777 Description: System hang on startup
3778 Details : This bug allowed the liblustre (e.g. catamount) client to
3779 return to the app before handling all startup RPCs. This
3780 could leave the node unresponsive to lustre network traffic
3781 and manifested as a server ptllnd timeout.
3783 Severity : enhancement
3785 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3786 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3787 boolean to control whether to print peer debug info when a
3788 client's RPC times out.
3793 Description: make O_SYNC working on 2.6 kernels
3794 Details : 2.6 kernels use different method for mark pages for write,
3795 so need add a code to lustre for O_SYNC work.
3800 Description: Failure to close file and release space on NFS
3801 Details : Put inode details into lock acquired in ll_intent_file_open.
3802 Use mdc_intent_lock in ll_intent_open to properly
3803 detect all kind of errors unhandled by mdc_enqueue.
3808 Description: proc file read during shutdown sometimes raced obd removal,
3810 Details : Add lock to prevent obd access after proc file removal.
3813 Frequency : Only for files larger than 4GB on 32-bit clients.
3815 Description: improperly doing page alignment of locks
3816 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3817 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3822 Description: RPCs being resent when they shouldn't be
3823 Details : Some RPCs that should not be resent are being resent. This
3824 can cause inconsistencies in the RPC state machine. Do not
3825 resend such requests.
3828 Frequency : rare, only with NFS export
3830 Description: Crash on NFS re-export node
3831 Details : under very unusual load conditions an assertion is hit in
3832 ll_intent_file_open()
3835 Frequency : only if OST filesystem is corrupted
3837 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3838 Details : for a short time RPCs with bulk IO are in the replay list,
3839 but replay of bulk IOs is unimplemented. If the OST filesystem
3840 is corrupted due to disk cache incoherency and then replay is
3841 started it is possible to trip an assertion. Avoid putting
3842 committed RPCs into the replay list at all to avoid this issue.
3847 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3848 Details : Large single O_DIRECT read and write calls can fail to allocate
3849 a sufficiently large buffer to process the request. In case of
3850 allocation failure the allocation is retried with a smaller
3851 buffer and broken into smaller requests.
3853 Severity : enhancement
3855 Description: Add -o localflock option to simulate old noflock behaviour.
3856 Details : This will achieve local-only flock/fcntl locks coherentness.
3861 Description: versioning check is incomplete
3862 Details : Checking the version difference of client vs. server, report
3863 error if the gap is too big.
3868 Description: add support PG_writeback bit
3869 Details : add support for PG_writeback bit for Lustre, for more carefull
3870 work with page cache in 2.6 kernel. This also fix some deadlocks
3871 and remove hack for work O_SYNC with 2.6 kernel.
3873 Severity : enhancement
3875 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3876 Details : The uninit_groups feature works in conjunction with the kernel
3877 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3878 up the pass1 processing of e2fsck. This is a read-only feature
3879 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3880 mount filesystems that have had this feature enabled.
3882 Severity : enhancement
3884 Description: Improve multi-block allocation algorithm to avoid fragmentation
3885 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3886 allocation locality and avoid filesystem fragmentation.
3888 ------------------------------------------------------------------------------
3890 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3892 * Support for kernels:
3893 2.6.9-42.0.3.EL (RHEL 4)
3894 2.6.5-7.276 (SLES 9)
3895 2.4.21-47.0.1.EL (RHEL 3)
3896 2.6.12.6 vanilla (kernel.org)
3897 2.6.16.21-0.8 (SLES10)
3898 * Recommended e2fsprogs version: 1.39.cfs2-0
3900 * The backwards-compatible /proc/sys/portals symlink has been removed
3901 in this release. Before upgrading, please ensure that you change
3902 any configuration scripts or /etc/sysctl.conf files that access
3903 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3904 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3905 in advance of the upgrade on any system running Lustre 1.4.6 or
3906 newer, since /proc/sys/lnet was added in that version.
3907 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3911 Frequency : only when quota is used
3913 Description: avoid scanning export list for quota master
3914 Details : Change the algorithms to avoid scanning export list in order
3915 to improve the efficiency.
3918 Frequency : MDS failover only, very rarely
3920 Description: "went back in time" messages on mds failover
3921 Details : The greatest transno may be lost when the current operation
3922 finishes with an error (transno==0) and the client's last_rcvd
3923 record is over-written. Save the greatest transno in the
3924 mds_last_transno for this case.
3927 Frequency : always for specific kernels and striping counts
3929 Description: client may get "Matching packet too big" without ACL support
3930 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3931 when trying to access files in certain configurations. The
3932 clients should in fact be denied when mounting because they do
3933 not understand ACLs.
3936 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3938 Description: many clients connecting with IO in progress causes connect timeouts
3939 Details : Avoid synchronous journal commits to avoid delays caused by many
3940 clients connecting/disconnecting when bulk IO is in progress.
3941 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3942 OST_IO_PORTAL to avoid delays behind potentially many pending
3946 Frequency : occasionally with multiple writers to a single file
3948 Description: shared writes to file may result in wrong size reported by stat()
3949 Details : Allow growing of kms when extent lock is cancelled
3952 Frequency : always with random mmap IO to multi-striped file
3954 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3955 Details : Only if the hole is at the end of OST object so that kms is too
3956 small. Fix is to increase kms accordingly in ll_nopage.
3959 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3961 Description: writes to a missing object would leak memory on the OST
3962 Details : If there is an inconsistency between the MDS and OST filesystems,
3963 such that the MDS references an object that doesn't exist, writes
3964 to that object will leak memory due to incorrect cleanup in the
3965 error handling path, eventually running out of memory on the OST.
3970 Description: Creating too long symlink causes lustre errors
3971 Details : Check symlink and name lengths before sending requests to MDS.
3974 Frequency : only if flock is enabled (not on by default)
3976 Description: posix locks not released on fd closure on 2.6.9+
3977 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3978 kernels, this caused such locks not to be released on fd close and
3979 then assertions on fs unmount about still used locks.
3982 Frequency : MDS failover only, very rarely
3984 Description: clients may get ASSERTION(granted_lock != NULL)
3985 Details : When request was taking a long time, and a client was resending
3986 a getattr by name lock request. The were multiple lock requests
3987 with the same client lock handle and
3988 mds_getattr_name->fixup_handle_for_resent_request found one of the
3989 lock handles but later failed with ASSERTION(granted_lock != NULL).
3994 Description: handle->h_buffer_credits > 0, assertion failure
3995 Details : h_buffer_credits is zero after truncate, causing assertion
3996 failure. This patch extends the transaction or creates a new
4000 Frequency : NFS re-export or patchless client
4001 Bugzilla : 11179, 10796
4002 Description: Crash on NFS re-export node (__d_move)
4003 Details : We do not want to hash the dentry if we don't have a lock.
4004 But if this dentry is later used in d_move, we'd hit uninitialised
4005 list head d_hash, so we just do this to init d_hash field but
4006 leave dentry unhashed.
4009 Frequency : NFS re-export or patchless client
4011 Description: NFS exports has problem with symbolic link
4012 Details : lustre client didn't properly install dentry when re-exported
4013 to NFS or running patchless client.
4016 Frequency : NFS re-export or patchless client
4018 Description: Various nfs/patchless fixes.
4019 Details : fixes reuse disconected alias for lookup process - this fixes
4020 warning "find_exported_dentry: npd != pd",
4021 fix permission error with open files at nfs.
4022 fix apply umask when do revalidate.
4025 Frequency : occasional
4027 Description: Crash on NFS re-export node
4028 Details : calling clear_page() on the wrong pointer triggered oops in
4029 generic_mapping_read().
4032 Frequency : rarely, using O_DIRECT IO
4034 Description: unaligned directio crashes client with LASSERT
4035 Details : check for unaligned buffers before trying any requests.
4038 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4040 Description: stale data returned from RAID cache
4041 Details : If only a small amount of IO is done to the RAID device before
4042 reading it again it is possible to get stale data from the RAID
4043 cache instead of reading it from disk.
4046 Frequency : always for sles10 kernel
4048 Description: sles10 support
4049 Details : ll_follow_link: compile fixes and using of nd_set_link
4050 under newer kernels.
4053 Frequency : depends on arch, kernel and compiler version, always on sles10
4056 Description: recursive or deep enough symlinks cause stack overflow
4057 Details : getting rid of large stack-allocated variable in
4061 Frequency : depends on hardware
4063 Description: lustre write performance loss in the SLES10 kernel
4064 Details : the performance loss is caused by using of write barriers in the
4065 ext3 code. The SLES10 kernel turns barrier support on by
4066 default. The fix is to undo that change for ldiskfs.
4068 ------------------------------------------------------------------------------
4070 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4072 * Support for kernels:
4073 2.6.9-42.0.3EL (RHEL 4)
4074 2.6.5-7.276 (SLES 9)
4075 2.4.21-47.0.1.EL (RHEL 3)
4076 2.6.12.6 vanilla (kernel.org)
4080 Frequency : quota enabled and large files being deleted
4082 Description: releasing more than 4GB of quota at once hangs OST
4083 Details : If a user deletes more than 4GB of files on a single OST it
4084 will cause the OST to spin in an infinite loop. Release
4085 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4090 Description: statfs data retrieved from /proc may be stale or zero
4091 Details : When reading per-device statfs data from /proc, in the
4092 {kbytes,files}_{total,free,avail} files, it may appear
4093 as zero or be out of date.
4096 Frequency : systems with MD RAID1 external journal devices
4098 Description: lconf's call to blkid is confused by RAID1 journal devices
4099 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4100 of returning all block devices that match the journal UUID.
4103 Frequency : always, for aggregate stripe size over 4GB
4105 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4106 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4107 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4110 Frequency : always if "lfs find" used on a local file/directory
4112 Description: "lfs find" segfaults if used on a local file/directory
4113 Details : The case where a directory component was not specified wasn't
4114 handled correctly. Handle this properly.
4117 Frequency : always on ppc64
4119 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4120 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4121 perform endianness conversion.
4124 Frequency : rarely (truncate to non-zero file size after write under load)
4125 Bugzilla : 10730, 10687
4126 Description: Files padded with zeros to next 4K multiple
4127 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4128 it is possible that files that are truncated to a non-zero size
4129 immediately after being written are filled with zero bytes beyond
4130 the truncated size. No file data is lost.
4132 Severity : enhancement
4134 Description: Allow recovery/failover for liblustre clients.
4135 Details : liblustre clients were unaware of failover configurations until
4138 Severity : enhancement
4140 Description: user file locks should fail when not mounting with flock option
4141 Details : Set up an error-returning stub in ll_file_operations.lock field
4142 to prevent incorrect behaviour when client is mounted without
4143 flock option. Also, set up properly f_op->flock field for
4147 Frequency : always on ia64
4149 Description: "lfs df" loops on printing out MDS statfs information
4150 Details : The obd_ioctl_data was not initialized and in some systems
4151 this caused a failure during the ioctl that did not return
4152 an error. Initialize the struct and return an error on failure.
4155 Frequency : SLES 9 only
4157 Description: Error of copying files with lustre special EAs as root
4158 Details : Client side always return success for setxattr call for lustre
4159 special xattr (currently only "trusted.lov").
4162 Frequency : rarely on clusters with both ia64+i386 clients
4164 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4165 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4166 clients are doing concurrent writes to the same file it is
4167 possible that the ia64 clients may LASSERT because the OST
4168 extent locks are not PAGE_SIZE aligned. Ensure that grown
4169 locks are always aligned on the request boundary.
4172 Frequency : specific use, occasional
4174 Description: Overwriting in use executable truncates on-disk binary image
4175 Details : If one node attempts to overwrite an executable in use by
4176 another node, we now correctly return ETXTBSY instead of
4177 truncating the file.
4179 Severity : enhancement
4181 Description: Async OSC create to avoid the blocking unnecessarily.
4182 Details : If an OST has no remaining object, system will block on the creating
4183 when need to create a new object on this OST. Now, ways use
4184 pre-created objects when available, instead of blocking on an
4185 empty osc while others are not empty. If we must block, we block
4186 for the shortest possible period of time.
4191 Description: chmod on Lustre root is propagated to other clients
4192 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4193 invalid by the follow_mount time.
4196 Frequency : liblustre clients only
4198 Description: Race in 'instant cancel' lock handling could lead to such locks
4199 never to be granted in case of SMP MDS
4200 Details : Do not destroy not yet granted but cbpending locks in
4204 Frequency : replay/resend of open
4206 Description: non null lock assetion failure in mds_intent_policy
4207 Details : Trying to replay/resend lockless open requests resulted in
4208 mds_open() returning 0 with no lock. Now it sets a flag if
4209 a lock is going to be returned.
4211 Severity : enhancement
4213 Description: Checksum enhancements
4214 Details : New checksum enhancements allow for resending RPCs that failed
4217 Severity : enhancement
4219 Description: Tunables on number of dirty pages in cacche
4220 Details : Allow to set limit on number of dirty pages cached.
4225 Description: client crash on unmount - lock still has references
4226 Details : In some error handling cases it was possible to leak a lock
4227 reference on a client while accessing a file. This was not
4228 harmful to the client during operation, but would cause the
4229 client to crash when the filesystem is unmounted.
4232 Frequency : specific case, rare
4234 Description: ETXTBSY on mds though file not in use by client
4235 Details : ETXTBSY is no longer incorrectly returned when attempting to
4236 chmod or chown a directory that the user previously tried to
4237 execute or a currently-executing binary.
4240 Frequency : extremely rare except on liblustre-based clients
4242 Description: Lustre space not freed when files are deleted
4243 Details : Clean up open-unlinked files after client eviction. Previously
4244 the unlink was skipped and the files remained as orphans.
4249 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4250 Details : In some cases it was possible to send a blocking callback to
4251 a client doing a glimpse, even though that client didn't get
4252 a lock granted. When the glimpse lock is cancelled on the OST
4253 the freed lock is left on the waiting list and corrupted the list.
4256 Frequency : all core dumps
4258 Description: Broke core dumps to lustre
4259 Details : Negative dentry may be unhashed if parent does not have UPDATE
4260 lock, but some callers, e.g. do_coredump, expect dentry to be
4261 hashed after successful create, hash it in ll_create_it.
4263 ------------------------------------------------------------------------------
4265 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4267 * Support for kernels:
4268 2.6.9-42.0.2.EL (RHEL 4)
4269 2.6.5-7.276 (SLES 9)
4270 2.4.21-40.EL (RHEL 3)
4271 2.6.12.6 vanilla (kernel.org)
4275 Frequency : always on RHEL 3
4277 Description: Number of open files grows over time
4278 Details : The number of open files grows over time, whether or not
4279 Lustre is started. This was due to a filp leak introduced
4280 by one of our kernel patches.
4282 ------------------------------------------------------------------------------
4284 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4286 * Support for kernels:
4287 2.6.9-42.EL (RHEL 4)
4288 2.6.5-7.276 (SLES 9)
4289 2.4.21-40.EL (RHEL 3)
4290 2.6.12.6 vanilla (kernel.org)
4295 Bugzilla : 5719, 9635, 9792, 9684
4296 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4297 Details : If a server is under heavy load and cannot reply to new
4298 connection requests before the client resends the (re)connect,
4299 the connection handling code can behave badly if two service
4300 threads are concurrently handing separate (re)connections from
4301 the same client. Add better locking to the connection handling
4302 code, and ensure that only a single connection will be processed
4303 for a given client UUID, even if the lock is dropped.
4305 Severity : enhancement
4307 Description: add TCP zero-copy support to kernel
4308 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4309 sends if the hardware supports scatter-gather and checksumming.
4310 This allows socklnd to do client-write and server-read more
4311 efficiently and reduce CPU utilization from skbuf copying.
4314 Frequency : only if NFS exporting from client
4316 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4317 Details : If setting attributes on a file created under NFS that had
4318 never been opened it would be possible to oops the client
4319 if the file had no objects.
4322 Frequency : always for liblustre
4324 Description: liblustre client does MDS+OSTs setattr RPC for each write
4325 Details : When doing a write from a liblustre client, the client
4326 incorrectly issued an RPC to the MDS and each OST the file was
4327 striped over in order to update the timestamps. When writing
4328 with small chunks and many clients this could overwhelm the MDS
4329 with RPCs. In all cases it would slow down the write because
4330 these RPCs are unnecessary.
4332 Severity : enhancement
4334 Description: allow number of MDS service threads to be changed at module load
4335 Details : It is now possible to change the number of MDS service threads
4336 running. Adding "options mds mds_num_threads={N}" to the MDS's
4337 /etc/modprobe.conf will set the number of threads for the next
4338 time Lustre is restarted (assuming the "mds" module is also
4339 reloaded at that time). The default number of threads will
4340 stay the same, 32 for most systems.
4345 Description: OST crash if filesystem is unformatted or corrupt
4346 Details : If an OST is started on a device that has never been formatted
4347 or if the filesystem is corrupt and cannot even mount then the
4348 error handling cleanup routines would dereference a NULL pointer.
4353 Description: NULL pointer deref in llap_from_page.
4354 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4355 incorrect) mapping if the page was truncated/reclaimed while it was
4356 searched for. Check for this condition and skip such pages when
4357 doing readahead. Introduce extra check to llap_from_page() to
4358 verify page->mapping->host is non-NULL (so page is not anonymous).
4361 Frequency : Sometimes when using sys_sendfile
4363 Description: "page not covered by a lock" warnings from ll_readpage
4364 Details : sendfile called ll_readpage without right page locks present.
4365 Now we introduced ll_file_sendfile that does necessary locking
4366 around call to generic_file_sendfile() much like we do in
4370 Frequency : with certain MDS communication failures at client mount time
4372 Description: NULL pointer deref after failed client mount
4373 Details : a client connection request may delayed by the network layer
4374 and not be sent until after the PTLRPC layer has timed out the
4375 request. If the client fails the mount immediately it will try
4376 to clean up before the network times out the request. Add a
4377 reference from the request import to the obd device and delay
4378 the cleanup until the network drops the request.
4381 Frequency : occasionally during client (re)connect
4383 Description: assertion failure during client (re)connect
4384 Details : processing a client connection request may be delayed by the
4385 client or server longer than the client connect timeout. This
4386 causes the client to resend the connection request. If the
4387 original connection request is replied in this interval, the
4388 client may trip an assertion failure in ptlrpc_connect_interpret()
4389 which thought it would be the only running connect process.
4392 Frequency : only with obd_echo servers and clients that are rebooted
4394 Description: kernel BUG accessing uninitialized data structure
4395 Details : When running an obd_echo server it did not start the ping_evictor
4396 thread, and when a client was evicted an uninitialized data
4397 structure was accessed. Start the ping_evictor in the RPC
4398 service startup instead of the OBD startup.
4400 Severity : enhancement
4401 Bugzilla : 10193 (patchless)
4402 Description: Remove dependency on various unexported kernel interfaces.
4403 Details : No longer need reparent_to_init, exit_mm, exit_files,
4404 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4407 Frequency : rare (only users of deprecated and unsupported LDAP config)
4409 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4410 Details : LDAP apparently contains 'lustreName' attributes instead of
4411 'name'. A simple remapping of the name is sufficient.
4414 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4416 Description: waiting_locks_callback trips kernel BUG if client is evicted
4417 Details : Running with the dump_on_timeout debug flag turned on makes
4418 it possible that the waiting_locks_callback() can try to dump
4419 the Lustre kernel debug logs from an interrupt handler. Defer
4420 this log dumping to the expired_lock_main() thread.
4422 Severity : enhancement
4424 Description: Support NFS exporting on 2.6 kernels.
4425 Details : Implement non-rawops metadata methods for NFS server to use without
4426 changing NFS server code.
4429 Frequency : very rare (synthetic metadata workload only)
4431 Description: two racing renames might cause an MDS thread to deadlock
4432 Details : Running the "racer" program may cause one MDS thread to rename
4433 a file from being the source of a rename to being the target of
4434 a rename at exactly the same time that another thread is doing
4435 so, and the second thread has already enqueued these locks after
4436 doing a lookup of the target and is trying to relock them in
4437 order. Ensure that we don't try to re-lock the same resource.
4440 Frequency : only very large systems with liblustre clients
4442 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4443 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4444 all OSTs in parallel. This allows the eviction of stale liblustre
4445 clients to proceed much faster than if they were done in series,
4446 and also offers similar improvements for other set_info RPCs.
4451 Description: excessive CPU usage during initial read phase on client
4452 Details : During the initial read phase on a client, it would agressively
4453 retry readahead on the file, consuming too much CPU and impacting
4454 performance (since 1.4.5.8). Improve the readahead algorithm
4455 to avoid this, and also improve some other common cases (read
4456 of small files in particular, where "small" is files smaller than
4457 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4462 Description: MDS crash when receiving packet with unknown intent.
4463 Details : Do not LBUG in unknown intent case, just return -EFAULT
4465 Severity : enhancement
4466 Bugzilla : 9293, 9385
4467 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4468 Details : Do not serialize getattr (non-intent version) and statfs.
4471 Frequency : occasional, when OST network is overloaded/intermittent
4473 Description: client evicted by OST after bulk IO timeout
4474 Details : If a client sends a bulk IO request (read or write) the OST
4475 may evict the client if it is unresposive to its data GET/PUT
4476 request. This is incorrect if the network is overloaded (takes
4477 too long to transfer the RPC data) or dropped the OST GET/PUT
4478 request. There is no need to evict the client at all, since
4479 the pinger and/or lock callbacks will handle this, and the
4480 client can restart the bulk request.
4483 Frequency : Always when mmapping file with no objects
4485 Description: client crashes when mmapping file with no objects
4486 Details : Check that we actually have objects in a file before doing any
4487 operations on objects in ll_vm_open, ll_vm_close and
4493 Description: Request leak when working with deleted CWD
4494 Details : Introduce advanced request refcount tracking for requests
4495 referenced from lustre intent.
4497 Severity : Enhancement
4499 Description: Cache open file handles on client.
4500 Details : MDS now will return special lock along with openhandle, if
4501 requested and client is allowed to hold openhandle, even if unused,
4502 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4503 closing files for every read/write openration.
4505 Severity : Enhancement
4507 Description: Cache open negative dentries on client when possible.
4508 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4509 negative dentries on lock revocation.
4514 Description: Remounting a client read-only wasn't possible with a zconf mount
4515 Details : It wasn't possible to remount a client read-only with llmount.
4517 Severity : enhancement
4518 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4519 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4520 MPICH is included to add Lustre-specific ADIO interfaces.
4521 This is based closely on the UFS ADIO layer and only differs
4522 in file creation, in order to allow the OST striping to be set.
4523 This is user-contributed code and not supported by CFS.
4528 Description: extended inode attributes (immutable, append-only) work improperly
4529 when 2.4 and 2.6 kernels are used on client/server or vice versa
4530 Details : Introduce kernel-independent values for these flags.
4532 Severity : enhancement
4535 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4536 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4537 granularity. Now, Fractional MB granularity is be supported,
4538 this is very useful for low memory system.
4540 Severity : enhancement
4542 Description: Getattr by fid
4543 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4544 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4545 have locks and avoids getattr rpc requests for it on every stat.
4548 Frequency : Always, for filesystems larger than 2TB
4550 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4551 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4552 filesystems larger than 2TB to be mounted with mballoc enabled.
4555 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4557 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4558 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4559 filesystems larger than 2TB is mounted then the kernel will
4560 silently corrupt the start of the filesystem. CONFIG_LBD is
4561 enabled for all CFS-supported kernels, but the possibility of
4562 this happening with a modified kernel config exists.
4564 Severity : enhancement
4566 Description: add client O_DIRECT support for 2.6 kernels
4567 Details : It is now possible to do O_DIRECT reads and writes to files
4568 in the Lustre client mountpoint on 2.6 kernel clients.
4570 Severity : enhancement
4572 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4573 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4574 parallel, not waiting for response from every OST before sending
4575 a rpc to the next OST.
4580 Description: setattr vs write race when updating file timestamps
4581 Details : Client processes that update a file timestamp into the past
4582 right after writing to the file (e.g. tar) it is possible that
4583 the updated file modification time can be reset to the current
4584 time due to a race between processing the setattr and write RPC.
4586 Severity : enhancement
4588 Description: Bring 'lfs find' closer in line with regular Linux find.
4589 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4590 -print0 options and obtains all the needed info through the lustre
4593 Severity : enhancement
4595 Description: support up to 1024 configured devices on one node
4596 Details : change obd_dev array from statically allocated to dynamically
4597 allocated structs as they are first used to reduce memory usage
4602 Description: Flush dirty partially truncated pages during truncate
4603 Details : Immediatelly flush partially truncated pages in filter_setattr,
4604 this way we completely avoid having any pages in page cache on OST
4605 and can retire ugly workarounds during writes to flush such pages.
4610 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4611 Details : protect inode from truncation within vfs_unlink() context
4612 just take a reference before calling vfs_unlink() and release it
4613 when parent's i_sem is free.
4618 Description: last_id value checked outside lock on OST caused LASSERT failure
4619 Details : If there were multiple MDS->OST object precreate requests in
4620 flight, it was possible that the OST's last object id was checked
4621 outside a lock and incorrectly tripped an assertion. Move checks
4622 inside locks, and discard old precreate requests.
4625 Frequency : always, if extents are used on OSTs
4627 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4628 Details : OSTs using the extents format would not zero the high 16 bits of
4629 the index physical block number. This is not a problem for any
4630 OST filesystems smaller than 16TB, and no kernels support ext3
4631 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4632 new/modified files) and can be fixed for existing filesystems
4633 with e2fsprogs-1.39-cfs1.
4638 Description: import connection selection may be incorrect if timer wraps
4639 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4640 import connections to be ignored if the 32-bit jiffies counter
4641 wraps. Use a 64-bit jiffies counter.
4644 Frequency : very large clusters immediately after boot
4646 Description: LNET request buffers exhausted under heavy short-term load
4647 Details : If a large number of client requests are generated on a service
4648 that has previously never seen so many requests it is possible
4649 that the request buffer growth cannot keep up with the spike in
4650 demand. Instead of dropping incoming requests, they are held in
4651 the LND until the RPC service can accept more requests.
4654 Frequency : Sometimes during replay
4656 Description: Assertion failure in ll_local_open after replay.
4657 Details : If replay happened on an open request reply before we were able
4658 to set replay handler, reply will become not swabbed tripping the
4659 assertion in ll_local_open. Now we set the handler right after
4660 recognising of open request
4663 Frequency : very rare
4665 Description: kernel reports "badness in vsnprintf"
4666 Details : Reading from the "recovery_status" /proc file in small chunks
4667 may cause a negative length in lprocfs_obd_rd_recovery_status()
4668 call to vsnprintf() (which is otherwise harmless). Exit early
4669 if there is no more space in the output buffer.
4671 Severity : enhancement
4673 Description: clear OBD RPC statistics by writing to them
4674 Details : It is now possible to clear the OBD RPC statistics by writing
4675 to the "stats" file.
4680 Description: Client mtime is not the same on different clients after utimes
4681 Details : In some cases, the client was using the utimes() syscall on
4682 a file cached on another node. The clients now validate the
4683 ctime from the MDS + OSTs to determine which one is right.
4688 Description: Inability to activate failout mode
4689 Details : lconf script incorrectly assumed that in pythong string's numeric
4690 value is used in comparisons.
4693 Frequency : always with multiple stripes per file
4695 Description: Inefficient object allocation for mutli-stripe files
4696 Details : When selecting which OSTs to stripe files over, for files with
4697 a stripe count that divides evenly into the number of OSTs,
4698 the MDS is always picking the same starting OST for each file.
4699 Return the OST selection heuristic to the original design.
4704 Description: mount failures may take full timeout to return an error
4705 Details : Under some heavy load conditions it is possible that a
4706 failed mount can wait for the full obd_timeout interval,
4707 possibly several minutes, before reporting an error.
4708 Instead return an error as soon as the status is known.
4710 Frequency : quota enabled and large files being deleted
4712 Description: releasing more than 4GB of quota at once hangs OST
4713 Details : If a user deletes more than 4GB of files on a single OST it
4714 will cause the OST to spin in an infinite loop. Release
4715 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4720 Description: statfs data retrieved from /proc may be stale or zero
4721 Details : When reading per-device statfs data from /proc, in the
4722 {kbytes,files}_{total,free,avail} files, it may appear
4723 as zero or be out of date.
4726 Frequency : systems with MD RAID1 external journal devices
4728 Description: lconf's call to blkid is confused by RAID1 journal devices
4729 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4730 of returning all block devices that match the journal UUID.
4733 Frequency : always, for aggregate stripe size over 4GB
4735 Description: assertion fails when trying to use 4GB stripe size
4736 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4737 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4740 Frequency : always on ppc64
4742 Description: the first write on an ext3 filesystem with mballoc got stuck
4743 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4744 perform endianness conversion.
4746 ------------------------------------------------------------------------------
4748 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4750 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4751 INTEROPERATE with older versions automatically. Please read the
4752 user documentation before upgrading any part of a live system.
4753 * WARNING: Lustre networking configuration changes are required with
4754 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4757 * Support for kernels:
4758 2.6.9-22.0.2.EL (RHEL 4)
4759 2.6.5-7.244 (SLES 9)
4760 2.6.12.6 vanilla (kernel.org)
4763 Severity : enhancement
4764 Bugzilla : 7981/8208
4765 Description: Introduced Lustre Networking (LNET)
4766 Details : LNET is new networking infrastructure for Lustre, it includes
4767 a reorganized network configuration mode (see the user
4768 documentation for full details) as well as support for routing
4769 between different network fabrics. Lustre Networking Devices
4770 (LNDs) for the supported network fabrics have also been
4771 created for this new infrastructure.
4773 Severity : enhancement
4774 Description: Introduced Access control lists
4775 Details : clients can set ACLs on files and directories in order to have
4776 more fine-grained permissions than the standard Unix UGO+RWX.
4777 The MDS must be started with the "-o acl" mount option.
4779 Severity : enhancement
4780 Description: Introduced filesystem quotas
4781 Details : Administrators may now establish per-user quotas on the
4784 Severity : enhancement
4786 Description: Configuration change for the XT3
4787 The PTLLND is now used to run Lustre over Portals on the XT3
4788 The configure option(s) --with-cray-portals are no longer used.
4789 Rather --with-portals=<path-to-portals-includes> is used to
4790 enable building on the XT3. In addition to enable XT3 specific
4791 features the option --enable-cray-xt3 must be used.
4796 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4797 Details : A race between allocating a new llcd and re-getting the llcd_lock
4798 allowed another thread to grab newly-allocated llcd.
4800 Severity : enhancement
4802 Description: 2.6 OST async journal commit and locking fix to improve performance
4803 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4804 2.6 kernels are now async as they already were in 2.4 kernels so
4805 that they can commit concurrently with the network bulk transfer.
4806 For block-allocated files the filter allocation semaphore is held
4807 to avoid filesystem fragmentation during allocation. BKL lock
4808 removed for 2.6 xattr operations where it is no longer needed.
4813 Description: lconf incorrectly determined whether two IP networks could talk
4814 Details : In some more complicated routing and multiple-network
4815 configurations, lconf will avoid trying to make a network
4816 connection to a disjoint part of the IP space. It was doing the
4817 math incorrectly for one set of cases.
4822 Description: Fix for potential infinite loop processing records in an llog.
4823 Details : If an llog record is corrupted/zeroed, it is possible to loop
4824 forever in llog_process(). Validate the llog record length
4825 and skip the remainder of the block on error.
4828 Frequency : occasional (liblustre only)
4830 Description: liblustre could not open files whose last component is a symlink
4831 Details : sysio_path_walk() would incorrectly pass the open intent to
4832 intermediate path components.
4835 Frequency : rare (liblustre only with non-standard tuning)
4836 Bugzilla : 7201 (7350)
4837 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4838 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4839 after it has been released, only do this for referenced locks
4841 Severity : enhancement
4843 Description: specifying an (invalid) directory default stripe_size of -1
4844 would reset the directory default striping
4845 Details : stripe_size -1 was used internally to signal directory stripe
4846 removal, now use "all default" to signal dir stripe removal
4847 as a directory striping of "all default" is not useful
4850 Frequency : common for large clusters running liblustre clients
4852 Description: doing an ls when liblustre clients are running is slow
4853 Details : sending a glimpse AST to a liblustre client waits for every AST
4854 to time out, as liblustre clients will not respond. Since they
4855 cannot cache data we refresh the OST lock LVB from disk instead.
4857 Severity : enhancement
4859 Description: doing an ls at the same time as file IO can be slow
4860 Details : enqueue and other "small" requests can be blocked behind many
4861 large IO requests. Create a new OST IO portal for non-IO
4862 requests so they can be processed faster.
4865 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4867 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4868 created file with mode 000
4869 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4870 allow an ftruncate() as a non-root user to a file with mode 000.
4871 Lustre now respects this flag to disable mode checks when
4872 truncating a file owned by the user
4875 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4878 Description: Revoking locks from clients that went dead or catatonic might take
4880 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4881 cancellation of such locks instant on servers without waiting for
4882 any reply from clients. Clients drops these locks when cancel
4883 notification from server is received without replying.
4886 Frequency : liblustre-only, when liblustre client dies or becomes busy
4888 Description: Doing ls on Linux clients can take a long time with active
4890 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4891 granting such locks to it in the first place if possible. Locks
4892 are taken by proxy on the OST during the read or write and
4893 dropped immediately afterward. Add connect flags handling, do
4894 not grant locks to liblustre clients for glimpse ASTs.
4896 Severity : enhancement
4898 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4899 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4900 sequential reads and behaves badly with applications doing random
4901 reads. Improve it by reading ahead at least read region, and
4902 avoiding excessive large RPC for small reads.
4904 Severity : enhancement
4906 Description: Creating more than 1000 files for a single job may cause a load
4907 imbalance on the OSTs if there are also a large number of OSTs.
4908 Details : qos_prep_create() uses an OST index reseed value that is an
4909 even multiple of the number of available OSTs so that if the
4910 reseed happens in the middle of the object allocation it will
4911 still utilize the OSTs as uniformly as possible.
4916 Description: OST or MDS may oops in ping_evictor_main()
4917 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4918 but doesn't restart at beginning of obd_exports_timed list
4921 Severity : enhancement
4923 Description: improve by-nid export eviction on the MDS and OST
4924 Details : allow multiple exports with the same NID to be evicted at one
4925 time without re-searching the exports list.
4928 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4930 Description: MDS may oops in groups_free()
4931 Details : in rare race conditions a newly allocated group_info struct is
4932 freed again, and this can be NULL. The 2.4 compatibility code
4933 for groups_free() checked for a NULL pointer, but 2.6 did not.
4936 Frequency : common for liblustre clients doing little filesystem IO
4937 Bugzilla : 9352, 7313
4938 Description: server may evict liblustre clients accessing contended locks
4939 Details : if a client is granted a lock or receives a completion AST
4940 with a blocking AST already set it would not reply to the AST
4941 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4944 Frequency : lfs setstripe, only systems with more than 160 OSTs
4946 Description: unable to set striping with a starting offset beyond OST 160
4947 Details : llapi_create_file() incorrectly limited the starting stripe
4948 index to the maximum single-file stripe count.
4951 Frequency : LDAP users only
4953 Description: lconf did not handle in-kernel recovery with LDAP properly
4954 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4956 Severity : enhancement
4958 Description: bind OST threads to NUMA nodes to improve performance
4959 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4960 node and do their allocations there to localize memory access
4962 Severity : enhancement
4964 Description: llmount can determine client NID directly from Myrinet (GM)
4965 Details : the client NID code from gmnalnid was moved directly into
4966 llmount, removing the need to use this or specifying the
4967 client NID explicitly when mounting GM clients with zeroconf
4970 Frequency : if client is started with down MDS
4972 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4973 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4974 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4975 ptlrpc_import_delayed_req() because we didn't check if the
4976 request was interrupted, and we also didn't break out of the
4977 event loop if there was no timeout
4982 Description: data loss during non-page-aligned writes to a single file from
4983 both multiple nodes and multiple threads on one node at same time
4984 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4985 inconsistency led to false short-reads, that were cached and later
4986 used by ->prepare_write() to fill in partially written page,
4987 leading to data loss.
4990 Frequency : always, if lconf --abort_recovery used
4992 Description: lconf --abort_recovery fails with 'Operation not supported'
4993 Details : lconf was attempting to abort recovery on the MDT device and not
4996 Severity : enhancement
4998 Description: remove cleanup logs
4999 Details : replace lconf-generated cleanup logs with lustre internal
5000 cleanup routines. Eliminates the need for client-cleanup and
5003 Severity : enhancement
5005 Description: add support for EAs (user and system) on lustre filesystems
5006 Details : it is now possible to store extended attributes in the Lustre
5007 client filesystem, and with the user_xattr mount option it
5008 is possible to allow users to store EAs on their files also
5010 Severity : enhancement
5012 Description: Add possibility (config option) to show minimal available OST free
5014 Details : When compiled with --enable-mindf configure option, statfs(2)
5015 (and so, df) will return least minimal free space available from
5016 all OSTs as amount of free space on FS, instead of summary of
5017 free spaces of all OSTs.
5019 Severity : enhancement
5021 Description: do not expand extent locks acquired on OST-side
5022 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5023 by server: they are not cached anyway.
5026 Frequency : when mmap is used/binaries executed from Lustre
5028 Description: Unmmap pages before throwing them away from read cache.
5029 Details : llap_shrink cache now attempts to unmap pages before discarding
5030 them (if unmapping failed - do not discard). SLES9 kernel has
5031 extra checks that trigger if this unmapping is not done first.
5036 Description: lconf didn't resolve symlinks before checking to see whether a
5037 given mountpoint was already in use
5040 Frequency : when migrating failover services
5041 Bugzilla : 6395, 9514
5042 Description: When migrating a subset of services from a node (e.g. failback
5043 from a failover service node) the remaining services would
5044 time out and evict clients.
5045 Details : lconf --force (implied by --failover) sets the global obd_timeout
5046 to 5 seconds in order to quickly disconnect, but this caused
5047 other RPCs to time out too quickly. Do not change the global
5048 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5050 Severity : enhancement
5051 Frequency : if MDS is started with down OST
5052 Bugzilla : 9439,5706
5053 Description: Allow startup/shutdown of an MDS without depending on the
5054 availability of the OSTs.
5055 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5056 Add appropriate locking and lov-osc refcounts for safe
5057 cleaning. Add osc abort_inflight calls in case the
5058 synchronize never started.
5061 Frequency : occasional (Cray XT3 only)
5063 Description: root not authorized to access files in CRAY_PORTALS environment
5064 Details : The client process capabilities were not honoured on the MDS in
5065 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5066 been accessed by an authorized user then root was able to access
5067 the file on the local client also. The root user capabilities
5068 are now allowed on the MDS, as this environment has secure UID.
5071 Frequency : occasional
5073 Description: ldiskfs "too long searching" message happens too often
5074 Details : A debugging message (otherwise harmless) prints too often on
5075 the OST console. This has been reduced to only happen when
5076 there are fragmentation problems on the filesystem.
5081 Description: Division by zero in statfs when all OSCs are inactive
5082 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5083 lov_statfs divides by value returned by lov_get_stripecnt().
5087 Bugzilla : 9489, 3273
5088 Description: First write from each client to each OST was only 4kB in size,
5089 to initialize client writeback cache, which caused sub-optimal
5090 RPCs and poor layout on disk for the first written file.
5091 Details : Clients now request an initial cache grant at (re)connect time
5092 and so that they can start streaming writes to the cache right
5093 away and always do full-sized RPCs if there is enough data.
5094 If the OST is rebooted the client also re-establishes its grant
5095 so that client cached writes will be honoured under the grant.
5100 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5101 Details : Now I/O RPCs go to different portal number and (presumably) fast
5102 lock requests (and glimses) and other RPCs get their own service
5103 threads pool that should be able to service those RPCs
5106 Severity : enhancement
5108 Description: Ability to exchange lustre version between client and servers and
5109 issue warnings at client side if client is too old. Also for
5110 liblustre clients there is ability to refuse connection of too old
5112 Details : New 'version' field is added to connect data structure that is
5113 filled with version info. That info is later checked by server and
5117 Frequency : rare, liblustre only.
5118 Bugzilla : 9296, 9581
5119 Description: Two simultaneous writes from liblustre at offset within same page
5120 might proceed at the same time overwriting eachother with stale
5122 Details : I/O lock withing llu_file_prwv was released too early, before data
5123 actually was hitting the wire. Extended lock-holding time until
5124 server acknowledges receiving data.
5127 Frequency : extremely rare. Never observed in practice.
5129 Description: avoid generating lustre_handle cookie of 0.
5130 Details : class_handle_hash() generates handle cookies by incrementing
5131 global counter, and can hit 0 occasionaly (this is unlikely, but
5132 not impossible, because initial value of cookie counter is
5133 selected randonly). Value of 0 is used as a sentinel meaning
5134 "unassigned handle" --- avoid it. Also coalesce two critical
5135 sections in this function into one.
5137 Severity : enhancement
5139 Description: allow liblustre clients to delegate truncate locking to OST
5140 Details : To avoid overhead of locking, liblustre client instructs OST to
5141 take extent lock in ost_punch() on client's behalf. New connection
5142 flag is added to handle backward compatibility.
5144 Severity : enhancement
5145 Bugzilla : 4928, 7341, 9758
5146 Description: allow number of OST service threads to be specified
5147 Details : a module parameter allows the number of OST service threads
5148 to be specified via "options ost ost_num_threads={N}" in the
5149 OSS's /etc/modules.conf or /etc/modprobe.conf.
5153 Bugzilla : 6146, 9635, 9895
5154 Description: servers crash with bad pointer in target_handle_connect()
5155 Details : In rare cases when a client is reconnecting it was possible that
5156 the connection request was the last reference for that export.
5157 We would temporarily drop the export reference and get a new
5158 one, but this may have been the last reference and the export
5159 was just destroyed. Get new reference before dropping old one.
5161 Severity : enhancement
5162 Frequency : if client is started with failover MDS
5164 Description: Allow multiple MDS hostnames in the mount command
5165 Details : Try to read the configuration from all specified MDS
5166 hostnames during a client mount in case the "primary"
5169 Severity : enhancement
5171 Description: Stop sending data to evicted clients as soon as possible.
5172 Details : Check if the client we are about to send or are sending data to
5173 was evicted already. (Check is done every second of waiting,
5174 for which l_wait_event interface was extended to allow checking
5175 of exit condition at specified intervals).
5178 Frequency : rare, normally only when NFS exporting is done from client
5180 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5182 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5185 Frequency : always, if srand() is called before liblustre initialization
5187 Description: Liblustre uses system PRNG disturbing its usage by user application
5188 Details : Introduce internal to lustre fast and high-quality PRNG for
5189 lustre usage and make liblustre and some other places in generic
5190 lustre code to use it.
5192 Severity : enhancement
5193 Bugzilla : 9477, 9557, 9870
5194 Description: Verify that the MDS configuration logs are updated when xml is
5195 Details : Check if the .xml configuration logs are newer than the config
5196 logs stored on the MDS and report an error if this is the case.
5197 Request --write-conf, or allow starting with --old_conf.
5199 Severity : enhancement
5201 Description: Handle symlinks in the path when checking if Lustre is mounted.
5202 Details : Resolve intermediate symlinks when checking if a client has
5203 mounted a filesystem to avoid duplicate client mounts.
5208 Description: lconf can hit an error exception but still return success.
5209 Details : The lconf command catches the Command error exception at the top
5210 level script context and will exit with the associated exit
5211 status, but doesn't ensure that this exit status is non-zero.
5216 Description: failure of ptlrpc thread startup can cause oops
5217 Details : Starting a ptlrpc service thread can fail if there are a large
5218 number of threads or the server memory is very fragmented.
5219 Handle this without oopsing.
5222 Frequency : always, only if liblustre and non-default acceptor port was used
5224 Description: liblustre cannot connect to servers with non-default acceptor port
5225 Details : tcpnal_set_default_params() was not called and was therefore
5226 ignoring the environment varaible TCPNAL_PORT, as well as other
5227 TCPNAL_ environment variables
5232 Description: two objects could be created on the same OST for a single file
5233 Details : If an OST is down, in some cases it was possible to create two
5234 objects on a single OST for a single file. No problems other
5235 than potential performance impact and spurious error messages.
5239 Bugzilla : 5681, 9562
5240 Description: Client may oops in ll_unhash_aliases
5241 Details : Client dcache may become inconsistent in race condition.
5242 In some cases "getcwd" can fail if the current directory is
5248 Description: Inode refcounting problems in NFS export code
5249 Details : link_raw functions used to call d_instantiate without obtaining
5250 extra inode reference first.
5254 Bugzilla : 9942, 9903
5255 Description: Referencing freed requests leading to crash, memleaks with NFS.
5256 Details : We used to require that call to ll_revalidate_it was always
5257 followed by ll_lookup_it. Also with revalidate_special() it is
5258 possible to call ll_revalidate_it() twice for the same dentry
5259 even if first occurence returned success. This fix changes semantic
5260 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5261 reference on a request referred from the intent.
5262 ll_intent_release() then releases such a request.
5265 Frequency : rare, normally benchmark loads only
5267 Description: unlinked inodes were kept in memory on the client
5268 Details : If a client is repeatedly creating and unlinking files it
5269 can accumulate a lot of stale inodes in the inode slab cache.
5270 If there is no other client load running this can cause the
5271 client node to run out of memory. Instead flush old inodes
5272 from client cache that have the same inode number as a new inode.
5275 Frequency : SLES9 2.6.5 kernel and long filenames only
5276 Bugzilla : 9969, 10379
5277 Description: utime reports stale NFS file handle
5278 Details : SLES9 uses out-of-dentry names in some cases, which confused
5279 the lustre dentry revalidation. Change it to always use the
5283 Frequency : rare, unless heavy write-truncate concurrency is continuous
5284 Bugzilla : 4180, 6984, 7171, 9963, 9331
5285 Description: OST becomes very slow and/or deadlocked during object unlink
5286 Details : filter_destroy() was holding onto the parent directory lock
5287 while truncating+unlinking objects. For very large objects this
5288 may block other threads for a long time and slow overall OST
5289 responsiveness. It may also be possible to get a lock ordering
5290 deadlock in this case, or run out of journal credits because of
5291 the combined truncate+unlink. Solution is to do object truncate
5292 first in one transaction without parent lock, and then do the
5293 final unlink in a new transaction with the parent lock. This
5294 reduces the lock hold time dramatically.
5297 Frequency : rare, 2.4 kernels only
5299 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5300 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5301 before unmounting their filesystem in order to avoid deadlock.
5302 The kernel_locked() function in 2.4 kernels only checks whether
5303 the kernel lock is held, not whether it is this process that is
5304 holding it as 2.6 kernels do.
5309 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5310 Details : The client ptlrpc code may be trying to reconnect to a down
5311 server before a previous connection attempt has timed out.
5312 Increase the reconnect interval to be longer than the connection
5313 timeout interval to avoid sending duplicate connections to
5317 Frequency : echo_client brw_test command
5319 Description: fix echo_client to work with OST preallocated code
5320 Details : OST preallocation code (5137) didn't take echo_client IO path
5321 into account: echo_client calls filter methods outside of any
5322 OST thread and, hence, there is no per-thread preallocated
5323 pages and buffers to use. Solution: hijack pga pages for IO. As
5324 a byproduct, this avoids unnecessary data copying.
5328 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5329 Description: Client can oops in mdc_commit_close() after open replay
5330 Details : It was possible for the MDS to return an open request with no
5331 transaction number in mds_finish_transno() if the client was
5332 evicted, but without actually returning an error. Clients
5333 would later try to replay that open and may trip an assertion
5334 Simplify the client close codepath, and always return an error
5335 from the MDS in case the open is not successful.
5338 Frequency : rare, 2.6 OSTs only
5340 Description: OST may deadlock under high load on fragmented files
5341 Details : If there was a heavy load and highly-fragmented OST filesystems
5342 it was possible to have all the OST threads deadlock waiting on
5343 allocation of biovecs, because the biovecs were not released
5344 until the entire RPC IO was completed. Instead, release biovecs
5345 as soon as they are complete to ensure forward IO progress.
5347 Severity : enhancement
5349 Description: Support for specifying external journal device at mount
5350 Details : If an OST or MDS device is formatted with an external journal
5351 device, this device major/minor is stored in the ext3 superblock
5352 and may not be valid for failover. Allow detecting and
5353 specifying the external journal at mount time.
5358 Description: Mounting an MDS with pending unlinked files may cause oops
5359 Details : target_finish_recovery() calls mds_postrecov() which returned
5360 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5361 considers this an error and immediately begins cleaning up the
5362 lov, just after starting the mds_lov process
5364 Severity : enhancement
5366 Description: Implement 'lfs df' to report actual free space on per-OST basis
5367 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5368 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5369 usage of objects; '-h' to report in human readable format.
5371 ------------------------------------------------------------------------------
5373 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5380 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5381 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5382 at startup during orphan unlinking. Instead, skip buddy update
5383 before setup, as it will be regenerated after recovery is complete.
5388 Description: If an OST is inactive, its locks might reference stale inodes.
5389 Details : lov_change_cbdata() must iterate over all namespaces, even if
5390 they are inactive to clear inode references from the lock.
5392 Severity : enhancement
5393 Frequency : occasional, if non-standard max_dirty_mb used
5395 Description: Client will block write RPCs if not enough grant
5396 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5397 then the client will block writes while waiting for another RPC
5398 to complete instead of consuming its dirty limit. With change
5399 we get improved performance when max_dirty_mb is small.
5401 Severity : enhancement
5402 Bugzilla : 3389, 6253
5403 Description: Add support for supplementary groups on the MDS.
5404 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5405 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5406 lookups for user supplementary groups into a cache.
5410 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5411 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5412 for liblustre O_EXCL is enforced by MDS.
5416 Description: readdir never returns NULL in liblustre.
5417 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5418 dentry in struct dirent.
5422 Frequency : liblustre only, depends on application IO pattern
5423 Description: liblustre clients evicted if not contacting servers
5424 Details : Don't put liblustre clients into the ping_evictor list, so
5425 they will not be evicted by the pinger ever.
5427 Severity : enhancement
5429 Description: Add ability to evict clients by NID from MDS.
5430 Details : By echoing "nid:$NID" string into
5431 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5432 $NID would be instantly evicted from this MDS and from all active
5433 OSTs connected to it.
5437 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5438 Details : lookup_it_finish() used to query file size from OSTs that was not
5443 Description: service threads change working directory to that of init
5444 Details : Starting lustre service threads may pin the working directory
5445 of the parent thread, making that filesystem busy. Threads
5446 now change to the working directory of init to avoid this.
5450 Frequency : during shutdown only
5451 Description: shutdown with a failed MDS or OST can cause unmount to hang
5452 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5457 Frequency : 2.6 only
5458 Description: chmod/chown may include an extra supplementary group
5459 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5460 supplementary group and if none is specified this is used.
5463 Bugzilla : 5479 (6816)
5465 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5466 Details : If lookup is in progress on a file that is unlinked we might try
5467 to revalidate the inode and fail in revalidate after lookup is
5468 complete and ll_file_open() enqueues the open again but
5469 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5472 Frequency : always, if lconf --abort_recovery used
5474 Description: lconf --abort_recovery fails with 'Operation not supported'
5475 Details : lconf was attempting to abort recovery on the MDT device and not
5478 ------------------------------------------------------------------------------
5480 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5485 Frequency : rare (only unsupported configurations with a node running as an
5487 Bugzilla : 6514, 5137
5488 Description: Mounting a Lustre file system on a node running as an OST could
5490 Details : OSTs now preallocates memory needed to write out data at
5491 startup, instead of when needed, to avoid having to
5492 allocate memory in possibly low memory situations.
5493 Specifically, if the file system is mounted on on OST,
5494 memory pressure could force it to try to write out data,
5495 which it needed to allocate memory to do. Due to the low
5496 memory, it would be unable to do so and the node would
5497 become unresponsive.
5499 Severity : enhancement
5501 Description: Addition of lconf --service command line option
5502 Details : lconf now accepts a '--service <arg>' option, which is
5503 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5505 Severity : enhancement
5507 Description: Failover mode is now the default for OSTs.
5508 Details : By default, OSTs will now run in failover mode. To return to
5509 the old behaviour, add '--failout' to the lmc line for OSTs.
5511 Severity : enhancement
5513 Description: Health checks are now provided for MDS and OSTs
5514 Details : Additional detailed health check information on MSD and OSTs
5515 is now provided through the procfs health_check value.
5518 Frequency : occasional, depends on IO load
5520 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5521 after numerous create/delete cycles
5522 Details : The ext3 inode allocation policy would not allocate new inodes
5523 very well on the OSTs because there are no new directories
5524 being created. Instead we look for groups with free space if
5525 the parent directories are nearly full.
5530 Description: Network or server problems during mount may cause partially
5531 mounted clients instead of returning an error.
5532 Details : The config llog parsing code may overwrite the error return
5533 code during mount error handling, returning success instead
5539 Description: MDS can fail to allocate large reply buffers
5540 Details : After long uptimes the MDS can fail to allocate large reply
5541 buffers (e.g. zconf client mount config records) due to memory
5542 fragmentation or consumption by the buffer cache. Preallocate
5543 some large reply buffers so that these replies can be sent even
5544 under memory pressure.
5548 Frequency : rare (liblustre)
5549 Description: fsx running with liblustre complained that using truncate() to
5550 extend the file doesn't work. This patch corrects that issue.
5551 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5552 ATTR_SIZE and lsm use in llu_setattr_raw.
5556 Frequency : rare, only 2.6 kernels
5557 Description: Unusual file access patterns on the MDS may result in inode
5558 data being lost in very rare circumstances.
5559 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5560 code in the 2.6 kernel caused the inode and/or EA data not to
5561 be read from disk, causing single-file corruption.
5565 Frequency : rare, only 2.6 filesystems using extents
5566 Description: Heavy concurrent write and delete load may cause data corruption.
5567 Details : It was possible under high-load situations to have an extent
5568 metadata block in the block device cache from a just-unlinked
5569 file overwrite a newly-allocated data block. We now unmap any
5570 metadata buffers that alias just-allocated data blocks.
5574 Frequency : filesystems with default stripe_count larger than 77
5575 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5576 Details : lconf specifies an inode size of 4096 bytes when the default
5577 stripe_count is larger than 77. This conflicts with the default
5578 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5579 this case to avoid pinning too much memory for large EAs.
5581 ------------------------------------------------------------------------------
5583 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5588 Frequency : rare (extremely heavy IO load with hundreds of clients)
5590 Description: Client is evicted, gets IO error writing to file
5591 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5592 caused clients to be evicted for AST timeouts. The fixes in
5593 bug 5192 mean we no longer need to have such short AST timeouts
5594 so ldlm_timeout has been increased.
5597 Frequency : occasional during --force or --failover shutdown under load
5598 Bugzilla : 5949, 4834
5599 Description: Server oops/LBUG if stopped with --force or --failover under load
5600 Details : a collection of import/export refcount and cleanup ordering
5601 issues fixed for safer force cleanup
5604 Frequency : only filesystems larger than 120 OSTs
5605 Bugzilla : 5990, 6223
5606 Description: lfs getstripe would oops on a very large filesystem
5607 Details : lov_getconfig used kfree on vmalloc'd memory
5610 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5611 Bugzilla : 6242, 6243
5612 Description: reading from files that had been truncated to a non-zero size
5613 but never opened returned no data
5614 Details : ll_file_read() reads zeros from no-object files to EOF
5619 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5620 Details : There's little harm in aborting MDS/OSS recovery and letting it
5621 try again, so I removed the LASSERT and return an error instead.
5623 Severity : enhancement
5625 Description: New debugging infrastructure for tracking down data corruption
5626 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5627 (b) cover more of the client-side code path, and (c) try to narrow
5628 down where problems occurred
5632 Bugzilla : 3819, 4364, 4397, 6313
5633 Description: Racing close and eviction MDS could cause assertion in mds_close
5634 Details : It was possible to get multiple mfd references during close and
5635 client eviction, leading to one thread referencing a freed mfd.
5637 Severity: : enhancement
5638 Bugzilla : 3262, 6359
5639 Description: Attempts to reconnect to servers are now more aggressive.
5640 Details : This builds on the enhanced upcall-less recovery that was added
5641 in 1.4.2. When trying to reconnect to servers, clients will
5642 now try each server in the failover group every 10 seconds. By
5643 default, clients would previously try one server every 25 seconds.
5648 Description: After recovery, certain operations trigger a failed
5649 assertion on a client.
5650 Details : Failing over an mds, using lconf -d --failover, while a
5651 client was doing a readdir() call would cause the client to
5652 LBUG after recovery completed and the readdir() was resent.
5654 Severity : enhancement
5656 Description: Default groups are now added by lconf
5657 Details : You can now run lconf --group <servicename> without having to
5658 manually add groups with lmc.
5661 Frequency : occasional
5663 Description: Nodes with an elan id of 0 trigger a failed assertion
5666 Frequency : always when accessing e.g. tty/console device nodes
5668 Description: tty and some other devices nodes cannot be used on lustre
5669 Details : file's private_data field is used by device data and lustre
5670 values in there got lost. New field was added to struct file to
5671 store fs-specific private data.
5674 Frequency : when exporting Lustre via NFS
5676 Description: NFSD failed occasionally when looking up a path component
5677 Details : NFSD is looking up ".." which was broken in ext3 directories
5678 that had grown large enough to become hashed.
5681 Frequency : Clusters with multiple interfaces not on the same subnet
5683 Description: Nodes will repeatedly try to reconnect to an interface which it
5684 cannot reach and report an error to the log.
5685 Details : Extra peer list entries will be created by lconf with some peers
5686 unreachable. lconf now validates the peer before adding it.
5689 Frequency : Only if a default stripe is set on the filesystem root.
5691 Description: Setting a default stripe on the filesystem root prevented the
5692 filesystem from being remounted.
5693 Details : The client was sending extra request flags in the root getattr
5694 request and did not allocate a reply buffer for the dir EA.
5697 Frequency : occasional, higher if lots of files are accessed by one client
5698 Bugzilla : 6159, 6097
5699 Description: Client trips assertion regarding lsm mismatch/magic
5700 Details : While revalidating inodes the VFS looks up inodes with ifind()
5701 and in rare cases can find an inode that is being freed.
5702 The ll_test_inode() code will free the lsm during ifind()
5703 when it finds an existing inode and then the VFS later attaches
5704 this free lsm to a new inode.
5708 Bugzilla : 6422, 7030
5709 Description: MDS deadlock between mkdir and client eviction
5710 Details : Creating a new file via mkdir or mknod (starting a transaction
5711 and getting the ns lock) can deadlock with client eviction
5712 (gets ns lock and trying to finish a synchronous transaction).
5715 Frequency : occasional
5716 Description: While starting a server, the fsfilt_ext3 module could not be
5718 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5719 kernels. Previously, lconf would still use the ext3 name
5720 when trying to load modules. Now, it will correctly use
5721 ext3 on 2.4 and ldiskfs on 2.6.
5723 Severity : enhancement
5724 Description: The default stripe count has been changed to 1
5725 Details : The interpretation of the default stripe count (0, to lfs
5726 or lmc) has been changed to mean striping across a single
5727 OST, rather than all available. For general usage we have
5728 found a stripe count of 1 or 2 works best.
5730 Severity : enhancement
5731 Description: Add support for compiling against Cray portals.
5732 Details : Conditional compiling for some areas that are different
5736 Frequency : occasional
5737 Bugzilla : 6409, 6834
5738 Description: Creating files with an explicit stripe count may lead to
5739 a failed assertion on the MDS
5740 Details : If some OSTs are full or unavailable, creating files may
5741 trigger a failed assertion on the MDS. Now, Lustre will
5742 try to use other servers or return an error to the
5746 Frequency : occasional
5748 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5749 caused lock timeout during readahead (since 1.4.2).
5750 Details : Processes doing readahead might match a lock that hasn't been
5751 granted yet if there are overlapping and conflicting lock
5752 requests. The readahead process waits on ungranted lock
5753 (original lock is CBPENDING), while OST waits for that process
5754 to cancel CBPENDING read lock and eventually evicts client.
5756 Severity : enhancement
5758 Description: Initial enabling of flock support for clients
5759 Details : Implements fcntl advisory locking and file status functions.
5760 This feature is provided as an optional mount flag (default
5761 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5762 locking are implemented yet, and those that are are not guaranteed
5763 to be completely correct in production environments.
5764 mount -t lustre -o [flock|noflock] ...
5767 Frequency : occasional
5769 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5770 assertion in the ext3 JBD (journaling) layer.
5771 Details : The b_committed_data struct is protected by the big kernel lock
5772 in 2.4 kernels, serializing journal_commit_transaction() and
5773 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5774 access to this struct. In 2.6 kernels there is finer grained
5775 locking to improve SMP performance of the JBD layer.
5779 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5784 Description: merge and cleanup kernel patches.
5787 -----------------------------------------------------------------------------
5789 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5791 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5792 versions of Lustre. Please update all servers and clients to
5793 version 1.4.2 or later at the same time. You must also run
5794 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5795 to update the configuration logs.
5797 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5798 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5799 - swab llog records of type '0' so we get proper header size/idx (5861)
5800 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5801 - fix rename of one directory over another leaking an inode (5953)
5802 - avoid SetPageDirty on 2.6 (5981)
5803 - don't re-add just-being-destroyed locks to the waiting list (5653)
5804 - when creating new directories, inherit the parent's custom
5805 striping settings if present parent (3048)
5806 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5807 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5808 - don't hold client locks on temporary worklist from l_lru (5666)
5809 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5810 - automatically evict dead clients (5921)
5811 - Update file size properly in create+truncate+fstat case (6196)
5812 - Do not unhash mountpoint dentries, do not allow removal of
5814 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5815 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5816 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5817 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5818 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5819 - Correctly handle reads of files with no objects (6243)
5820 - lctl recover will also mark a device active if deactivate used (5933)
5822 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5823 - allow --write-conf on an MDS with different nettype than client (5619)
5824 - don't write config llogs to MDS for mounts not from that MDS (5617)
5825 - lconf should create multiple TCP connections from a client (5201)
5826 - init scripts are now turned off by default; run chkconfig --on
5827 lustre and chkconfig --on lustrefs to use them
5828 - upcalls are no longer needed for clients to recover to failover
5830 - add --abort-recovery option to lconf to abort recovery on device
5832 - add support for an arbitrary number of OSTs (3026)
5833 - Quota support protocol changes.
5834 - forward compatibility changes to wire structs (6007)
5835 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5836 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5837 - improved "lustre status" script
5838 - initialize blocksize for non-regular files (6062)
5839 - added --disable-server and --disable-client configure options (5782)
5840 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5841 - Vanilla 2.4.29 support
5842 - increase maximum number of obd devices to 520 (6242)
5843 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5844 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5845 SLES 9 2.6.5 kernels
5846 - limit stripes per file to 160 (the maximum EA size) (6093)
5848 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5851 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5852 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5853 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5854 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5855 - small ext3 extents cleanups and fixes (5733)
5856 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5857 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5858 - avoid needless client->OST connect, fix handle mismatch (5317)
5859 - fix DLM error path that led to out-of-sync client, long delays (5779)
5860 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5861 - fix several locking issues related to i_size (5492,5624,5654,5672)
5862 - don't move pending lock onto export if it is already evicted (5683)
5863 - fix kernel oops when creating .foo in unlinked directory (5548)
5864 - fix deadlock in obdfilter statistics vs. object create (5811)
5865 - use time_{before,after} to avoid timer jiffies wrap (5882)
5866 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5867 - Do not leak request if server was not able to process it (5154)
5868 - If mds_open unable to find parent dir, make that negative lookup(5154)
5869 - don't create new directories with extent-mapping (5909, 5936)
5871 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5872 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5873 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5874 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5875 - support for RHEL4 kernel on i686 (5773)
5876 - provide error messages when incompatible logs are encountered (5898)
5878 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5879 * version 1.4.0.10 (1.4.1 release candidate 1)
5881 - don't keep a lock reference when lock is not granted (4238)
5882 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5883 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5884 - drop import inflight refcount on signal_completed_replay error (5255)
5885 - unlock page after async write error during send (3677)
5886 - handle missing objects in filter_preprw_read properly (5265)
5887 - no transno return for symlink open, don't save no-trasno open (3440)
5888 - don't try to complete elan receive that already failed (4012)
5889 - free RPC server reply state on error (5406)
5890 - clean up thread from ptlrpc_start_thread() on error (5160)
5891 - readahead could read extra page into cache that wasn't ejected (5388)
5892 - prevent races in class_attach/setup/cleanup/detach (5260)
5893 - don't dereference de->d_inode after l_dput of de (5458)
5894 - use "int" for stripe value returned from lock_to_stripe (5544)
5895 - mballoc allocation and error-checking fixes in 2.6 (5504)
5896 - block device patches to fix I/O request sizes in 2.6 (5482)
5897 - look up hostnames for IB nals (5602)
5898 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5899 - don't start multiple acceptors for the same port (5277)
5900 - fix incorrect LASSERT in mds_getattr_name (5635)
5901 - export a proc file for general "ping" checking (5628)
5902 - fix "lfs check" to not block when the MDS is down (5628)
5904 - service request history (4965)
5905 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5906 - create an "evict_client" /proc entry on OSTs, like the MDS has
5907 - fix mount usage message, return errors per mount(8) (5168)
5908 - change grep [] to grep "[]" in tests so they work in more UMLs
5909 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5910 - remove ancient LOV_MAGIC_V0 EA support (5047)
5911 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5912 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5913 - allow readahead allocations to fail when low on memory (5383)
5914 - mmap locking landed again, after considerable improvement (2828)
5915 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5917 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5920 - send OST transaction number in read/write reply to free req (4966)
5921 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5922 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5923 - the watchdog thread now runs as interruptible (5246)
5924 - flock/lockf fixes (but it's still disabled, pending 5135)
5925 - don't use EXT3 constants in llite code (5094)
5926 - memory shortage at startup could cause assertion (5176)
5928 - reorganization of lov code
5929 - single portals codebase
5931 - add extents/mballoc support (5025)
5932 - direct I/O reads in the obdfilter (4048)
5933 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5935 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5938 - send OST transaction number in read/write reply to free req (4966)
5939 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5940 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5941 - the watchdog thread now runs as interruptible (5246)
5942 - handle missing objects in filter_preprw_read properly (5265)
5943 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5944 - drop import inflight refcount on signal_completed_replay error (5255)
5945 - unlock page after async write error during send (3677)
5946 - return original error code on reconstructed replies (3761)
5947 - no transno return for symlink open, don't save no-trasno open (3440)
5949 - add pid to ldlm debugging output (4922)
5950 - bump the watchdog timeouts -- we can't handle 30sec yet
5951 - extra debugging for orphan dentry/inode bug (5259)
5953 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5956 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5957 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5958 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5959 - fix NULL dereference in /proc/sys/portals/routes (4827)
5960 - allow failed mdc_close() operations to be interrupted (4561)
5961 - stop precreate on OST before MDS would time out on it (4778)
5962 - don't send partial-page writes before EOF from client (4410)
5963 - discard client grant for sub-page writes on large-page clients (4520)
5964 - don't free dentries not owned by NFS code, check generation (4806)
5965 - fix lsm leak if mds_create_objects() fails (4801)
5966 - limit debug_daemon file size, always print CERROR messages (4789)
5967 - use transno after validating reply (3892)
5968 - process timed out requests if import state changes (3754)
5969 - update mtime on OST during writes, return in glimpse (4829)
5970 - add mkfsoptions to LDAP (4679)
5971 - use ->max_readahead method instead of zapping global ra (5039)
5972 - don't interrupt __l_wait_event() during strace
5974 - add software watchdogs to catch hung threads quickly (4941)
5975 - make lustrefs init script start after nfs is mounted
5976 - fix CWARN/ERROR duplication (4930)
5977 - return async write errors to application if possible (2248)
5978 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5979 - print NAL number in %x format (4645)
5980 - update barely-supported suse-2.4.21-171 series (4842)
5981 - support for sles 9 %post scripts
5982 - support for building 2.6 kernel-source packages
5983 - support for sles km_* packages
5985 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5988 - ignore -ENOENT errors in osc_destroy (3639)
5989 - notify osc create thread that OSC is being cleaned up (4600)
5990 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5991 - reconstruct ost_handle() like mds_handle() (4657)
5992 - create a new thread to do import eviction to avoid deadlock (3969)
5993 - let lconf resolve symlinked-to devices (4629)
5994 - don't unlink "objects" from directory with default EA (4554)
5995 - hold socknal file ref over connect in case target is down (4394)
5996 - allow more than 32000 subdirectories in a single directory (3244)
5997 - fix blocks count for O_DIRECT writes (3751)
5998 - OST returns ENOSPC from object create when no space left (4539)
5999 - don't send truncate RPC if file size isn't changing (4410)
6000 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6001 - bind to privileged port in socknal and tcpnal (3689)
6003 - rate limit CERROR/CWARN console message to avoid overload (4519)
6004 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6005 - basic mmap support (3918)
6006 - kernel patch series update from b1_4 (4711)
6008 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - avoid crash during MDS cleanup with OST shut down (2775)
6012 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6013 - don't use bad inodes on the MDS (3744)
6014 - dynamic object preallocation to improve recovery speed (4236)
6015 - don't hold spinlock over lock dumping or change debug flags (4401)
6016 - don't zero obd_dev when it is force cleaned (3651)
6017 - print grants to console if they go negative (4431)
6018 - "lctl deactivate" will stop automatic recovery attempts (3406)
6019 - look for existing locks in ldlm_handle_enqueue() (3764)
6020 - don't resolve lock handle twice in recovery avoiding race (4401)
6021 - revalidate should check working dir is a directory (4134)
6023 - don't always mark "slow" obdfilter messages as errors (4418)
6025 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6028 - don't close LustreDB during write_conf until it is done (3860)
6029 - fix typo in lconf for_each_profile (3821)
6030 - allow dumping logs from multiple threads at one time (3820)
6031 - don't allow multiple threads in OSC recovery (3812)
6032 - fix debug_size parameters (3864)
6033 - fix mds_postrecov to initialize import for llog ctxt (3121)
6034 - replace config semaphore with spinlock (3306)
6035 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6036 - don't allow enqueue to complete on a destroyed export (3822)
6037 - down write_lock before checking llog header bitmap (3825)
6038 - recover from lock replay timeout (3764)
6039 - up llog sem before sending rpc (3652)
6040 - reduce ns lock hold times when setting kms (3267)
6041 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6042 - fix NULL deref and obd_dev leak on setup error (3312)
6043 - replace some LBUG about llog ops with error handling (3841)
6044 - don't match INVALID dentries from d_lookup and spin (3784)
6045 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6046 - fix invalid assertion in ptlrpc_set_wait (3880)
6048 - add libwrap support for the TCP acceptor (3996)
6049 - add /proc/sys/portals/routes for non-root route listing (3994)
6050 - allow setting MDS UUID in .xml (2580)
6051 - print the stack of a process that LBUGs (4228)
6053 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6057 - make sure to unset replay flag from failed open requests (3440)
6058 - if default stripe count is 0, use OST count for inode size (3636)
6059 - update parent mtime/ctime on client for create/unlink (2611)
6060 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6061 - free recovery state on server during a forced cleanup (3571)
6062 - unregister_reply for resent reqs (3063)
6063 - loop back devices mounting and status check on 2.6 (3563)
6064 - fix resource-creation race that can provoke i_size == 0 (3513)
6065 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6066 - more debugging for page-accounting assertion (3746)
6067 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6068 - avoid deadlock after precreation failure (3758)
6069 - fix race and lock order deadlock in orphan handling (3450, 3750)
6070 - add validity checks when grabbing inodes from l_ast_data (3599)
6072 - add /proc/.../recovery_status to obdfilter (3428)
6073 - lightweight CDEBUG infrastructure, debug daemon (3668)
6074 - change default OSC RPC parameters to be better on small clusters
6075 - turn off OST read cache for files smaller than 32MB
6076 - install man pages and include them in rpms (3100)
6077 - add new init script for (un)mounting lustre filesystems (2593)
6078 - run chkconfig in %post for init scripts (3701)
6079 - drop scimac NAL (unmaintained)
6081 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6084 - clean kiobufs before and after use (3485)
6085 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6086 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6087 - change init script to fail more clearly if not run as root (1528)
6088 - allow clients to reconnect during replay (1742)
6089 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6090 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6091 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6092 - fix cond_resched (3554)
6093 - only evict unfinished clients after recovery (3515)
6094 - allow bulk resend, prevent data loss (3570)
6095 - dynamic ptlrpc request buffer allocation (2102)
6096 - don't allow unlinking open directory if it isn't empty (2904)
6097 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6098 - remove extraneous obd dereference causing LASSERT failure (3334)
6099 - don't use get_cycles() when creating temp. files on the mds (3156)
6100 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6101 - handle EEXIST for set-stripe, set proper directory name (3336)
6103 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6104 - fix ksocknal_fmb_callback() error messages (2918)
6106 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6109 - don't copy lvb into (possibly NULL) reply on error (2983)
6110 - don't deref dentry after dput, don't free lvb on error (2922)
6111 - use the kms to determine writeback rpc length (2947)
6112 - increment oti_logcookies when osc is inactive (2948)
6113 - update client's i_blocks count via lvb messages (2543)
6114 - handle intent open/close of special files properly (1557)
6115 - mount MDS with errors=remount-ro, like obdfilter (2009)
6116 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6117 - don't use cancelling-locks' kms values (2947)
6118 - use highest lock extent for kms, not last one (2925)
6119 - don't dereference ERR_PTR() dentry in error handling path (3107)
6120 - fix thread race in portals_debug_dumplog() (3122)
6121 - create lprocfs device entries at setup instead of at attach (1519)
6122 - common AST error handler, don't evict client on completion race (3145)
6123 - zero nameidata in detach_mnt in 2.6 (3118)
6124 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6125 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6126 - initialize RPC timeout timer earlier for 2.6 (3219)
6127 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6128 - print nal/nid for unknown nid (3258)
6129 - additional checks for oscc recovery before doing precreate (3284)
6130 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6131 - don't crash in mdc_close for bad permissions on open (3285)
6132 - zero i_rdev for non-device files (3147)
6133 - clear page->private before handing to FS, better assertion (3119)
6134 - tune the read pipeline (3236)
6135 - fix incorrect decref of invalidated dentry (2350)
6136 - provide read-ahead stats and refine rpc in flight stats (3328)
6137 - don't hold journal transaction open across create RPC (3313)
6138 - update atime on MDS at close time (3265)
6139 - close LDAP connection when recovering to avoid server load (3315)
6140 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6141 - don't leak open file on MDS after open resend (3325)
6142 - serialize filter_precreate and filter_destroy_precreated (3329)
6143 - loop device shouldn't call sync_dev() for nul device (3092)
6144 - clear page cache after eviction (2766)
6145 - resynchronize MDS->OST in background (2824)
6146 - refuse to mount the same filesystem twice on same mountpoint (3394)
6147 - allow llmount to create routes for mounting behind routers (3320)
6148 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6149 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6150 - fix rare problem with rename on htree directories (3417)
6152 - allow default OST striping configuration per directory (1414)
6153 - fix compilation for qswnal for 2.6 kernels (3125)
6154 - increase maximum number of MDS request buffers for large systems
6155 - change liblustreapi to be useful for external progs like lfsck (3098)
6156 - increase local configuration timeout for slow disks (3353)
6157 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6159 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6162 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6163 - don't overwrite extent policy data in reply if lock was blocked (2901)
6164 - drop filter export grants atomically with removal from device (2663)
6165 - del obd_self_export from work_list in class_disconnect_exports (2908)
6166 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6167 - swab reply message in mdc_close, other PPC fixes (2464)
6168 - fix destroying of named logs (2325)
6169 - overwrite old logs when running lconf --write_conf (2264)
6170 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6171 - fix race in target_handle_connect (2898)
6172 - mds_reint_create() should take same inode create lock (2926)
6173 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6174 - don't close files for self_export to avoid uninitialized obd (2936)
6175 - allow MDS with the same name as client node (2939)
6176 - hold dentry reference for closed log files for unlink (2325)
6177 - reserve space for all logs during transactions (2059)
6178 - don't evict page beyond end of stripe extent (2925)
6179 - don't oops on a deleted current working directory (2399)
6180 - handle hard links to targets without a parent properly (2517)
6181 - don't dereference NULL lock when racing during eviction (2867)
6182 - don't grow lock extents when lots of conflicting locks (2919)
6184 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6187 - account for cache space usage on clients to avoid data loss (974)
6188 - lfsck support in lustre kernel code (2349)
6189 - reduce journal credits needed for BRW writes (2370)
6190 - orphan handling to avoid losing space on client/server crashes
6191 - ptlrpcd can be blocked, stopping ALL progress (2477)
6192 - use lock value blocks to assist in proper KMS, faster stat (1021)
6193 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6194 - recovery for initial connections (2355)
6195 - fixes for mds_cleanup_orphans (1934)
6196 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6197 - block all file creations until orphan recovery completes (1901)
6198 - client remove rq_connection from request struct (2423)
6199 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6200 - recovery timer race (2670)
6201 - mdc_close recovey bug (2532)
6202 - ptlrpc cleanup bug (2710)
6203 - mds timeout on local locks (2588)
6204 - namespace lock held during RPCs (2431)
6205 - handle interrupted sync write properly (2503)
6206 - don't try to handle a message that hasn't been replied to (2699)
6207 - client assert failure during cleanup after abort recovery (2701)
6208 - leak mdc device after failed mount (2712)
6209 - ptlrpc_check_set allows timedout requests to complete (2714)
6210 - wait for inflight reqs when ptlrpcd finishes (2710)
6211 - make sure unregistered services are removed from the srv_list
6212 - reset bulk XID's when resending them (caught by 1138 test)
6213 - unregister_bulk after timeout
6214 - fix lconf error (2694)
6215 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6216 - readahead locks pages, leaves pending causing memory pressure (2673)
6217 - increase OST request buffers to 4096 on large machines (2729)
6218 - fix up permission of existing directories in simple_mkdir (2661)
6219 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6220 - don't assign transno to errored transactions (2742)
6221 - don't delete objects on OST if given a bogus objid from MDS (2751)
6222 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6223 - if rq_no_resend, then timeout request after recovery (2432)
6224 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6225 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6226 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6227 - MDS recovery shouldn't delete objects that it has given out (2730)
6228 - if enqueue arrives after completion, don't clobber LVB (2819)
6229 - don't unlock pages twice when trigger_group_io returns error (2814)
6230 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6231 - don't write pages to disk if there was an error (1450)
6232 - don't ping imports that have recovery disabled (2676)
6233 - take buffered bytes into account when balancing socknal conn (2817)
6234 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6235 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6236 - remove little-endian swabbing of llog records (1987)
6237 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6238 - retry reposting request buffers if they fail (1191)
6239 - grow extent at grant time to avoid granting a revoked lock (2809)
6240 - lock revoke doesn't evict page if covered by a second lock (2765)
6241 - disable VM readahead to avoid reading outside lock extents (2805)
6243 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6244 - updated kernel patches for hp-2.4.20 kernel (2681)
6246 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6249 - fix truncated write corruption (2366)
6250 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6252 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6253 - ensure len(uuid) < 37 in lmc (1171)
6254 - fix ia64 OOPS in llog_test (2255)
6255 - zero end of page at obdfilter for partial page writes (2648)
6256 - don't leave stale dentries around after renames (bug 2428)
6257 - fix timeouts when evicting a client with a single lock held (2642)
6258 - set deadline for the initial HELLO message to drain (2634)
6259 - print out dotted-quad IP addresses in the socknal (2302)
6261 - additional debugging for MDS client eviction problem (2443)
6262 - fix mkfsoptions support for osts (2603, 2604)
6264 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6267 - add series for the vanilla 2.6.0 kernel
6268 - add series for the vanilla 2.4.24 kernel
6269 - add series for a cray x86/64 UL kernel drop
6270 - fix xattr patches for the vanilla 2.4.19 series
6272 - generate true UUIDs in lmc (1171)
6273 - have portals stack dumping break in UML (2466)
6274 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6275 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6276 - wake callback waiting threads on client eviction (2460)
6277 - Add --ptldebug and --subsystem to lmc (1719)
6278 - update assertion to allow safe interrupt allocation
6279 - set rq_no_resend for cancel requests (2432)
6280 - recalculate ptlrpcd timeout after resend (2494)
6281 - call vfs_rmdir when removing pending directories (2368)
6282 - fix renaming a file to itself (2429)
6283 - lmc creates a default one-stripe lov (2454)
6284 - expand procfs space to handle large clusters (2326)
6285 - increase UML stack to avoid overflow
6286 - update lconf's list of debug and subsystem masks
6287 - fix lfs find --obd (2510)
6288 - /proc tunable for disabling filter read caching (2591)
6289 - stop rpm packages from altering slapd.conf (2301)
6290 - disable nagle in the socknal under 0conf (2578)
6291 - choose mds inode size based on stripe count (2572)
6292 - fix kernel-source rpm problems (2516)
6294 - add --disable-doc to avoid pdf generation (2421)
6295 - update documentation, tests, type-os, comments
6296 - avoid format warnings on ia64
6297 - remove the TOE NAL
6298 - tiny code cleanups by removing unused fields
6300 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6303 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6304 - remove the most blatant lies from BUILDING file (2371)
6305 - change default debug level to reasonable production setting
6306 - reduce client side cache size to reduce cache flush time
6307 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6308 - make TCP zerocopy and pinger support enabled by default (2476)
6309 - sync writes completed after process exits caused crashes (2319)
6310 - maintain correct mount count on the MDS (2356)
6311 - backout 1557, because 2316 wasn't really fixed
6312 - better file I/O statistics gathering in /proc
6313 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6314 - another kernel patch to fix zero-copy TCP function export
6315 - don't take duplicate lock when processing re-sent getattr (2420)
6316 - lctl uses obd_self_export instead of creating new conn (2353)
6317 - MDS/OST recovery case which requires object creation asserted (2425)
6318 - move lfs from /usr/sbin to /usr/bin in packages
6319 - fix race between mds_client_add and mds_client_free (2417)
6320 - use kmalloc instead of slabs in portals (2430)
6321 - don't create duplicate records when a failover MDS is present (2442)
6322 - remove unnecessary mount age check (2332)
6323 - don't remove directory inodes from locks prematurely (2451)
6324 - don't break if MDS service name is the same as hostname (2103)
6325 - fix races in client write RPC generation when cache full (2482)
6327 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6330 - remove now-unused request->rq_obd (278)
6331 - if an allocation fails, print out how much memory we've used (1933)
6332 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6333 - add the "configurable stack size" patch to most series files (1256)
6334 - ability to write large log records, for 100+ OST configs (2306)
6335 - fix NULL deref when filter_prep fails (2314)
6336 - fix operator precedence error in filter_sync
6337 - dynamic allocation of socknal TX descriptors (2315)
6338 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6339 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6340 - ensure that CWARN messages in Portals always get to the syslog
6341 - __init/__exit are not for prototype decls (ldlm_init/exit)
6342 - x86-64 compile warning fixes
6343 - fix gateway LMC keyword conflict (2318)
6344 - fix MDS lock inversions in getattr/reint paths (1844)
6345 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6346 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6347 - DLM race condition prevented some lock evictions (2328)
6348 - ENOMEM detection and retry on socknal sends (2230)
6349 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6350 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6351 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6352 - better error messages when a client is rejected during recovery (1505)
6353 - avoid cancelling locks which were never granted, after failure (2330)
6354 - fix i_sem/journal inversion in mds_client_add (2333)
6355 - fix truncate/getattr lock cycle deadlock (2334)
6356 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6357 - fix two rare exit paths which could leak an l_lock() ref (2321)
6358 - fencepost error in MDS/OST orphan recovery (2226)
6359 - make log record alignment 8 bytes (1988)
6360 - lstripe now fails when requested offset > ost_count (2237)
6361 - ensure that all kernel series have a complete list.h (1607)
6362 - fix crashes in special-file operations (2316)
6363 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6365 - allow configurable automake binary, for testing new versions
6366 - small update to the lfs documentation
6368 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6370 * fix negative export reference count in fsfilt_sync (2312)
6372 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6373 * release candidate 0.9.1
6375 - orphans are moved into the PENDING directory for possible recovery
6376 - replayed opens now open by fid for orphan/rename safety (1042)
6377 - last close of an orphan inode generates a transno (683)
6378 - chdir() and mount() now pin the directory entry (1020)
6379 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6380 - discard very old requests without processing them (1502)
6381 - remove some common, well-understood CERRORs (1505)
6382 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6383 - clear "grant" flags in OST replies until OST grant code lands (1644)
6384 - fix read performance by not clobbering i_blksize on client (1598)
6385 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6386 - make LRU size a /proc tunable, clears locks when reduced (707)
6387 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6388 - server locks take references on exports now (1558)
6389 - build fixes for 2.4.20-rh trees (1663)
6390 - return an error from lov_create if all OSCs are inactive (1751)
6391 - fix import levels when a reconnect happens without a timeout (1597)
6392 - exit early from mds_open if we get a lookup error (1749)
6393 - partial page read at EOF wouldn't wait for disk before sending (1642)
6394 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6395 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6396 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6397 - write/truncate lock inversion (1639)
6398 - Don't auto-load obdclass, portals modules during cleanup (1495)
6399 - fix timestamps from jumping to "now" (1763)
6400 - extra journal assertions (1648)
6401 - add an extra multiunlink test (1771)
6402 - fix read_record/write_record API (1776)
6403 - fix leak of offset_extent, possible incorrect i_size later (1772)
6404 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6405 - Debugging for the kqswnal_get_idle_tx problems (1820)
6406 - Allow recovery to be attempted multiple times (1536)
6407 - Write out MDS last_rcvd file after it is first created (1600)
6408 - Fix tx_descriptor leak in failed transmit situations (1827)
6409 - ext3 journaling fixes for assertion failure after IO error (1871)
6410 - class_export_put() on freed export after completion AST error (1896)
6411 - Fix revalidate looping in VFS (1322)
6412 - Don't access a freed export during MDS_REINT timeout (1521)
6413 - Add open-unlink recovery support on the MDS (1673,1764)
6414 - Return an error if no MDS data was read from last_rcvd (1946)
6415 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6416 - Don't setup a disk device that doesn't match exported UUID (317)
6417 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6418 - avoid committing NULL handle in force close
6419 - local.sh is now a one-stripe LOV configuration
6420 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6421 - fix ext3 htree duplicate directory entry corruption (1516)
6422 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6423 - update to new LOV EA format (2097)
6424 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6426 2003-06-15 Phil Schwan <phil@clusterfs.com>
6429 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6430 - per-import/export recovery handling (958, 931, 959)
6431 - multiple last-rcvd slots, for serving multiple FSes (949)
6432 - connections are again shared between multiple imp/exports (963, 964)
6433 - "umount -f" would hang if any requests needed to be sent (393, 978)
6434 - avoid pinning large req buffer by copying for queued messages (989)
6435 - add "uuid" to "lctl device" command to help upcalls (991)
6436 - "open" RPCs with transnos would confuse recovery counters (1037)
6437 - do proper endian conversion of all wire messages (288, 340, 891)
6438 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6439 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6440 - fix ptlrpc_connection leak in target_handle_connect (1174)
6441 - fix import refcounting bug in OST and MDS cleanup (1134)
6442 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6443 - fix very unlikely obd_types race condition (501)
6444 - remove osc_open hack for echo_client (1187)
6445 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6446 - a failure in read_inode2 leads to deadlock (1139)
6447 - cancel ack-locks as soon as transaction is committed (1072)
6448 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6449 - make sure to commitrw after any preprw to avoid deadlock (1162)
6450 - failing to execute a file in a lustre FS would lock inode (1203)
6451 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6452 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6453 - obd_uuid2tgt would walk off the end of the list (1255)
6454 - on IA64 the osc would give portals incorrect bulk size (1258)
6455 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6456 - fix lock inversion caused by new llite matching code (1282)
6457 - limit the number of dirty pages on a client to 10MB (1286)
6458 - timed out locks were not being corrected cancelled (1289)
6459 - fix O_DIRECT above 4GB on IA-32 (1292)
6460 * major user-visible changes
6461 - fail out/fail over policy now controlled by the upcall (993)
6463 - add OBD_PING to check server availability and failure (954)
6464 - lustre messages are now sent in sending host order (288, 340, 891)
6465 - add eadatalen to MDS getattr reply (340)
6466 - OST read replies may contain second buffer, with per-page status (593)
6468 2003-03-11 Phil Schwan <phil@clusterfs.com>
6471 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6472 - failed open()s could cause deadlock; fixed (867, 869)
6473 - stop cancelling OST locks when files are closed (481)
6474 - overlapping XID spaces caused network corruption (851, 853)
6475 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6476 - setattr_raw added, to do single-RPC, server-side setattrs
6477 - lmc/lconf syntax change for OST UUIDs
6478 - fix crashy race condition between ptlrpc_free_req and osc_close
6479 - don't use request in mdc_enqueue if we hit a timeout (889)
6480 - don't set the inode i_size for regular files from the MDS (896)
6481 - handle out of order completion AST (842)
6482 - don't LBUG if a lock request times out after receiving AST (913)
6483 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6484 - if a bad lock AST arrives, send an error instead of dropping entirely
6485 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6486 - fix leak in bulk IO when only partially completed (899, 900, 926)
6487 - fix O_DIRECT for ia64 (55)
6488 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6489 - C-z after timeout could hang a process forever; fixed (977)
6491 - client-side I/O cache (678, 924, 929, 941, 970)
6493 - READPAGE and SETATTRs which don't take server-side locks get
6496 2003-02-11 Phil Schwan <phil@clusterfs.com>
6499 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6500 - Fix fsfilt_extN_readpage() to read a full page of directory
6501 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6502 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6503 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6504 - Most intent operations were removed, in favour of a new RPC mode
6505 that does a single RPC to the server and bypasses most of the VFS
6506 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6507 - Aggressively cancel local locks on DLM servers
6508 - mds_reint_unlink sends EA to the client if it's the last nlink.
6509 client uses that EA to unlink OST objects.
6510 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6511 - recursive symlinks were fixed (439)
6512 - fixed NULL deref in DEBUG_REQ
6513 - filter_update_lastobjid no longer calls sync, which annoyed extN
6514 - fixed multi-client small-writes to a single file problem (445)
6515 - fixed mtime updates during file writes (607)
6516 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6517 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6518 - fixed Linux OST size reporting problem (444, 656)
6519 - OST now updates object mtime with writes or setattr (607, 619)
6520 - client verifies file size before zeroing page past EOF (445)
6521 - OST now writes last allocated objid to disk with allocation (108)
6522 - LOV on echo now works (409)
6524 - mds_reint_unlink sends a new buffer, with the EA included. this
6525 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6526 set if a regular file was being unlinked, and it was the last link
6527 - use PtlGet from the target for bulk writes (315)
6528 - OST now updates object mtime with writes or setattr (607, 619)
6529 - LDLM now has a grant-time callback to revalidate locked items, if
6531 - Many MDS operations were reorganized to combat race conditions
6533 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6534 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6536 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6539 - Fully reactivate OST imports after reconnection (512, others)
6540 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6541 - More graceful error handling for truncating on dead OST (515)
6542 - Don't error out unless we're actually accessing dead stripes (474)
6543 - Fix garbage sizes when stripes are missing (410)
6544 - LRU counters were broken, causing constant lock purge (433, 432)
6545 - garbage on read from stripes with failed OSTs (441)
6546 - mark OSCs as active before reconnecting during recovery (438)
6547 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6548 - lfind did not preserve OST order in output (443)
6549 - symlinks cause hung clients, incorrect data (439)
6550 - stop dereferencing request after dropping refcount (457)
6551 - don't LASSERT(spin_is_locked) on non-SMP (455)
6552 - fixes for many rename() bugs
6553 - fstat didn't correctly synchronize attributes (399)
6554 - server must handle lock cancellation during blocking AST prep (487)
6555 - bulk descriptors were free()d too soon (511)
6556 - fix paths in lconf, which would load incorrect modules (451, 507)
6557 - fix confusing lconf 'host not found' error message (386)
6558 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6559 - fix race condition in mdc_blocking_ast() for inode access (526)
6560 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6561 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6562 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6563 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6564 - fix lstripe in several ways (kernel side) (527)
6565 - fix request leak in ldlm_cli_enqueue (262)
6566 - incorrect OSC was marked inactive after OST failure
6567 - call mds_fs_cleanup before unmounting filesystem (524)
6568 - fix races between taking ns_lock and ldlm_lock_change_resource
6569 - fix races updating LOV export open file list
6570 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6571 - fix recovery NULL deref in ldlm_cli_cancel_unused
6572 - fix some DLM races by using new hash table for lock handles (419)
6573 - permit the client to specify desired inodes, at replay
6574 - duplicate requests when we queue them for replay reintegration
6575 - fix last_rcvd offset calculation
6576 - sync after each recovered transaction, so we always make progress
6577 - never, not always, ERESTART requests without transnos
6578 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6580 - skip replay if the MDS says that the client is already connected
6581 - don't check for a recovery-enabled export to match lctl's UUID
6582 - don't INC_USE_COUNT for phantom exports
6583 - don't crash when cleaning up phantom exports (567)
6584 - don't double-finish or set replay data for errored mdc_open requests
6585 - abort requests when they time out, so we don't get old replies
6586 - send/receive replies for AST messages again
6587 - if the client says that it doesn't have the lock, cancel it on the
6589 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6590 mark it as destroyed, it will all work out when decref is called
6591 - fix module use counts (22, 581)
6593 - ASTs now expect a reply (server cancels lock on error reply)
6595 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6598 - fix many simultaneous client startup (392)
6599 - fix dentry->d_it clobbering
6600 - credentials weren't being shipped for readdir/getattr operations
6601 - remove invalid assertions triggered during some concurrent MD
6603 - proper Lustre versions added (336, 389)
6604 - fix memory leak for create error case (398)
6605 - fix LOV locking bug that would get cli/srv out of sync
6606 - fix echo client over LOV (409)
6607 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6608 - fix double-O_EXCL intent crash (424)
6609 - avoid sending multiple lock CANCELs (352)
6611 - MDS can do multi-client recovery (modulo bugs in new code)
6613 - many updates, edits, cleanups
6615 2002-11-18 Phil Schwan <phil@clusterfs.com>
6618 - fix null d_it dereference (346)
6619 - fix full OST/dbench hang (333)
6620 - fix permission problem with file removal (286)
6621 - fix removal of OSCs from LOV when they fail
6622 - fix NULL deref during bulk timeout (214)
6623 - fix problems related to multiple filesystems on one MDS (241)
6624 - fixed serious subtle metadata locking bugs
6625 - free locks on clients when inodes are removed due to memory
6627 - fix inode pointer in lock data (285)
6628 - partial support for multiple MDS on a single host (241)
6629 - data locks weren't cancelled at clear_inode time (290, 311)
6630 - intent locks could lead to unbounded lock growth (205)
6631 - added a maximum lock count, an LRU list, and a flusher
6632 - fix multiple rename (365)
6633 - properly abstracted the echo client
6634 - OSC locked 1 byte too many; fixed
6635 - rewrote brw callback code:
6636 - fixed recovery bugs related to LOVs (306)
6637 - fixed too-many-pages-in-one-write crash (191)
6638 - fixed (again) crash in sync_io_timeout (214)
6639 - probably fixed callback-related race (385)
6641 - Add capability to MDS protocol
6642 - LDLM cancellations and callbacks on different portals
6644 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6647 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6648 - defer open object destruction to close time (601981, 138)
6649 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6650 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6651 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6652 - don't call obd_setattr in ll_file_release for destroyed objects
6653 * protocol change to lustre_msg: move |version| and add |flags|
6654 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6655 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6656 * added replay of create, unlink, link and rename operations during
6657 MDS failover; recovery should be much more robust now
6658 * remove failed OSCs from LOVs (only lov_create uses this so far)
6659 * the lustre-HOWTO was brought (more) up to date (582544)
6661 2002-10-23 Phil Schwan <phil@clusterfs.com>
6664 - in-use dentries weren't being reused properly (617851)
6665 - prevent multiple LDLM setup (599178)
6666 - fix LOV size calculations for truncate (617853)
6667 - fix client handling of MDS intent errors (POSIX)
6668 - fix permission bug in lovstripe.c test (624321)
6669 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6670 - truncate past end of file could corrupt data
6671 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6672 - a race in recovery could return ETIMEDOUT to apps (623947)
6673 - building outside the source directory was fixed
6674 * the lustre-HOWTO was brought (more) up to date (582544)
6675 * major progress was made on recovery functionality
6677 2002-10-10 Phil Schwan <phil@clusterfs.com>
6680 - recovery deadlock fix
6681 - rm -rf causes LBUG fix (617817)
6682 - file open by multiple tasks fix (618962)
6683 - directory permissions bugs (602707 and 620007)
6684 - journal_stop fixed with locking (611313)
6685 - O_APPEND failures resolved (618273, perhaps 614459)
6686 - lconf PATH fix (619770)
6687 - IA64 build fix (621450)
6688 - RPC buffer sizes scale with amount of memory
6690 2002-10-01 Phil Schwan <phil@clusterfs.com>
6693 - locks would be cancelled without throwing away data pages,
6694 resulting in inconsistent data (605627)
6695 - inode attributes were not always being refreshed (605627, 612449)
6696 - lconf now continues to cleanup after lctl reports an error
6697 - MDS now enforces user permissions (602707)
6698 - lprocfs cleanup fixed, but not yet enabled (614157)
6699 - fixed infinite server hang, should a client not respond to an AST
6700 - avoid going into recovery if user calls readlink() with a buffer
6701 that's too small (613941)
6702 - AST RPCs no longer require replies (614867) -- this may be changed
6703 - don't crash server if client sends an IOV that's too big (611336)
6704 - fixed lock conversion deadlock (611892)
6705 - fixed the following of symlinks (614622)
6706 * recovery: the server can remove locks from a client that dies, other
6707 clients can make progress
6708 * more extN patch fixes
6709 * compile-time configurable ptlrpc buffer allocations
6711 - collaborative read cache document
6712 - Lustre Lite Performance CDR document-in-progress
6714 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6717 - fix typo in patch-2.4.18
6719 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6722 - clear ptlrpc request each time in handle_incoming_request()
6723 - unlink of files now destroys the object on the OST
6725 2002-09-19 Peter Braam <braam@clusterfs.com>
6727 * add hard link support
6728 * change obdfile creation method
6729 * kernel patch changed
6731 2002-09-19 Peter Braam <braam@clusterfs.com>
6734 - stack overflow bug in extN fixed
6736 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6738 * documentation updates
6739 - add man pages for config tools
6740 - update tests/README to describe testing with new config tools
6741 - finish metadata API descriptions
6742 * bug fixes and cleanups
6743 - statfs workaround for 16TB limit
6744 - LOV stripe allocation improved, can stripe on subset of OSTs
6745 - LOV file size/IO offset was wrong for files > 4GB in size
6746 - object EA data was being dropped, caused files to be unreadable
6747 - memory overflow with non-LOV OST caused memory corruption
6748 - fixed regression tests to work with new config tools, obdfilter
6749 - fixed bug when directory size became larger than 1 block
6750 - fixed bug (for single client case) when PWD was deleted
6751 - invalidate local directory pages when doing intent-based ops
6752 - avoid LDLM oops when lock callback contained bad data
6754 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6756 * documentation updates
6757 * bug fixes and cleanups
6758 - configuration tools
6761 - 64-bit compile warnings
6762 - 64-bit internal statfs data
6764 * test_brw on persistent OST devices
6766 * lprocfs (disabled)
6768 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6770 * documentation updates
6771 * bug fixes and cleanups
6772 * configuration tools
6774 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6777 * many small fixes to 0_5_4
6778 * io/network handling
6779 * thinkos in MDS operations
6781 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6784 * crucial basic fixes to 0.5.3
6785 * IOR, Iozone work over Elan
6788 2002-08-07 Phil Schwan <phil@clusterfs.com>
6789 * version 0_5_3, our first alpha
6790 * we use the new Portals iovs
6791 * documentation updates
6792 * bug fixes and cleanups
6793 * small changes in the DLM wire protocol
6795 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6796 * version 0_5_1 with some initial stability,
6797 * locking on MD and file I/O.
6798 * documentation updates
6799 * several bug fixes since 0.5.0
6800 * small changes in wire protocol
6802 2002-07-18 Phil Schwan <phil@clusterfs.com>
6804 * delivered as Lustre Light Alpha
6805 * fixed a crash after handling invalid MDS requests
6806 * fixed directory pages for architectures with non-4k pages sizes
6808 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6809 * release version v0_4_4
6810 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6812 * Fixes a number of interruption problems with OST operations.
6813 * Update documentation for portals header changes
6814 * Move all wire protocol structs/defines to lustre_idl.h
6815 * Fixes symlink length bug.
6816 * Add tcpdump to repository.
6818 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6819 * release version v0_4_3
6820 * Fixes statfs for inodes on extN.
6821 * Fixes bug in runtests which would delete /etc/hosts.
6822 * Use 64-bit object IDs wherever possible (not into VFS though)
6823 Remove ost_get_info, which is unused by lustre, and out of date.
6825 2002-07-03 Peter Braam <braam@clusterfs.com>
6826 * release version v0_4_2 Fixes a lookup error (type not passed)
6827 * move forward to head of Portals
6828 * move forward to latest Lustre kernel
6830 2002-06-25 Peter Braam <braam@clusterfs.com>
6831 * release version v0_4_1. Hopefully stable on single node use.