3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.10.1.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.162-94.69 (SLES12 SP3)
11 4.4.0-142 (Ubuntu 16.04)
12 4.15.0-45 (Ubuntu 18.04)
13 vanilla linux 4.6.7 (ZFS only)
14 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
15 * Client known to build on unpatched kernels (others may also work):
16 2.6.32-696.18.7.el6 (RHEL6.9)
17 2.6.32-754.6.3.el6 (RHEL6.10)
18 3.10.0-862.14.4.el7 (RHEL7.5)
19 3.10.0-957.10.1.el7 (RHEL7.6)
20 4.14.0-49.13.1.el7a (RHEL7.5)
21 4.14.0-115.2.2.el7a (RHEL7.6)
22 4.4.120-92.70 (SLES12 SP2)
23 4.4.162-94.69 (SLES12 SP3)
24 4.12.14-95.13 (SLES12 SP4)
25 4.4.0-142 (Ubuntu 16.04)
26 4.15.0-45 (Ubuntu 18.04)
27 vanilla linux 4.15.0 (ZFS only)
28 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
29 * Recommended ZFS / SPL version: 0.7.13
30 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
31 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
32 stack overflow. For more information, please refer to bugzilla 17630.
33 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
34 the re-exporting nodes to be mounted with "32bitapi" mount option
35 * Note for Ubuntu 14.04.5:
36 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
37 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
42 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
43 for currently supported client and server kernel versions.
44 * Server known to build on the following kernels (others may also work):
45 2.6.32-696.18.7.el6 (RHEL6.9)
46 3.10.0-862.14.4.el7 (RHEL7.5)
47 3.10.0-957.10.1.el7 (RHEL7.6)
48 4.4.120-92.70 (SLES12 SP2)
49 4.4.162-94.69 (SLES12 SP3)
50 4.4.0-142 (Ubuntu 16.04)
51 4.15.0-45 (Ubuntu 18.04)
52 vanilla linux 4.6.7 (ZFS only)
53 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
54 * Client known to build on unpatched kernels (others may also work):
55 2.6.32-696.18.7.el6 (RHEL6.9)
56 2.6.32-754.6.3.el6 (RHEL6.10)
57 3.10.0-862.14.4.el7 (RHEL7.5)
58 3.10.0-957.10.1.el7 (RHEL7.6)
59 4.14.0-49.13.1.el7a (RHEL7.5)
60 4.14.0-115.2.2.el7a (RHEL7.6)
61 4.4.120-92.70 (SLES12 SP2)
62 4.4.162-94.69 (SLES12 SP3)
63 4.12.14-95.13 (SLES12 SP4)
64 4.4.0-142 (Ubuntu 16.04)
65 4.15.0-45 (Ubuntu 18.04)
66 vanilla linux 4.15.0 (ZFS only)
67 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
68 * Recommended ZFS / SPL version: 0.7.13
69 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
70 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
71 stack overflow. For more information, please refer to bugzilla 17630.
72 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
73 the re-exporting nodes to be mounted with "32bitapi" mount option
74 * Note for Ubuntu 14.04.5:
75 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
76 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
80 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
81 for currently supported client and server kernel versions.
82 * Server known to build on the following kernels (others may also work):
83 2.6.32-696.18.7.el6 (RHEL6.9)
84 3.10.0-862.14.4.el7 (RHEL7.5)
85 3.10.0-957.el7 (RHEL7.6)
86 4.4.120-92.70 (SLES12 SP2)
87 4.4.162-94.69 (SLES12 SP3)
88 4.4.0-131 (Ubuntu 16.04)
89 4.15.0-32 (Ubuntu 18.04)
90 vanilla linux 4.6.7 (ZFS only)
91 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
92 * Client known to build on unpatched kernels (others may also work):
93 2.6.32-696.18.7.el6 (RHEL6.9)
94 2.6.32-754.6.3.el6 (RHEL6.10)
95 3.10.0-862.14.4.el7 (RHEL7.5)
96 3.10.0-957.el7 (RHEL7.6)
97 4.14.0-49.13.1.el7a (RHEL7.5)
98 4.14.0-115.2.2.el7a (RHEL7.6)
99 4.4.120-92.70 (SLES12 SP2)
100 4.4.162-94.69 (SLES12 SP3)
101 4.4.0-131 (Ubuntu 16.04)
102 4.15.0-32 (Ubuntu 18.04)
103 vanilla linux 4.15.0 (ZFS only)
104 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
105 * Recommended ZFS / SPL version: 0.7.9
106 * Tested with ZFS / SPL version: 0.7.9
107 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
108 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
109 stack overflow. For more information, please refer to bugzilla 17630.
110 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
111 the re-exporting nodes to be mounted with "32bitapi" mount option
112 * Note for Ubuntu 14.04.5:
113 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
114 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
116 --------------------------------------------------------------------------------
119 2018-04-03 Intel Corporation
121 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
122 for currently supported client and server kernel versions.
123 * Server known to build on patched kernels:
124 2.6.32-431.29.2.el6 (RHEL6.5)
125 2.6.32-504.30.3.el6 (RHEL6.6)
126 2.6.32-573.26.1.el6 (RHEL6.7)
127 2.6.32-642.15.1.el6 (RHEL6.8)
128 2.6.32-696.18.7.el6 (RHEL6.9)
129 3.10.0-693.21.1.el7 (RHEL7.4)
130 3.0.101-0.47.71 (SLES11 SP3)
131 3.0.101-107 (SLES11 SP4)
132 3.12.74-60.64.40 (SLES12 SP1)
133 4.4.114-92.64 (SLES12 SP2)
134 4.4.114-94.11 (SLES12 SP3)
135 3.13.0-101 (Ubuntu 14.04, ZFS only)
136 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
137 4.4.0-116 (Ubuntu 16.04)
138 vanilla linux 4.6.7 (ZFS only)
139 * Client known to build on unpatched kernels:
140 2.6.32-431.29.2.el6 (RHEL6.5)
141 2.6.32-504.30.3.el6 (RHEL6.6)
142 2.6.32-573.26.1.el6 (RHEL6.7)
143 2.6.32-642.15.1.el6 (RHEL6.8)
144 2.6.32-696.18.7.el6 (RHEL6.9)
145 3.10.0-693.21.1.el7 (RHEL7.4)
146 3.0.101-0.47.71 (SLES11 SP3)
147 3.0.101-107 (SLES11 SP4)
148 3.12.74-60.64.40 (SLES12 SP1)
149 4.4.114-92.64 (SLES12 SP2)
150 4.4.114-94.11 (SLES12 SP3)
151 3.13.0-101 (Ubuntu 14.04)
152 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
153 4.4.0-116 (Ubuntu 16.04)
155 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
156 * Recommended ZFS / SPL version: 0.7.6
157 * Tested with ZFS / SPL version: 0.7.6
158 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
159 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
160 stack overflow. For more information, please refer to bugzilla 17630.
161 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
162 the re-exporting nodes to be mounted with "32bitapi" mount option
163 * Note for Ubuntu 14.04.5:
164 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
165 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
167 --------------------------------------------------------------------------------
169 2017-07-05 Intel Corporation
171 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
172 for currently supported client and server kernel versions.
173 * Server known to build on patched kernels:
174 2.6.32-431.29.2.el6 (RHEL6.5)
175 2.6.32-504.30.3.el6 (RHEL6.6)
176 2.6.32-573.26.1.el6 (RHEL6.7)
177 2.6.32-642.15.1.el6 (RHEL6.8)
178 2.6.32-696.3.1.el6 (RHEL6.9)
179 3.10.0-514.21.1.el7 (RHEL7.3)
180 3.0.101-0.47.71 (SLES11 SP3)
181 3.0.101-100 (SLES11 SP4)
182 3.12.74-60.64.40 (SLES12 SP1)
183 4.4.59-92.17 (SLES12 SP2)
184 vanilla linux 4.6.7 (ZFS only)
185 * Client known to build on unpatched kernels:
186 2.6.32-431.29.2.el6 (RHEL6.5)
187 2.6.32-504.30.3.el6 (RHEL6.6)
188 2.6.32-573.26.1.el6 (RHEL6.7)
189 2.6.32-642.15.1.el6 (RHEL6.8)
190 2.6.32-696.3.1.el6 (RHEL6.9)
191 3.10.0-514.21.1.el7 (RHEL7.3)
192 3.0.101-0.47.71 (SLES11 SP3)
193 3.0.101-100 (SLES11 SP4)
194 3.12.74-60.64.40 (SLES12 SP1)
195 4.4.59-92.17 (SLES12 SP2)
197 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
198 * Recommended ZFS / SPL version: 0.6.5.9
199 * Tested with ZFS / SPL version: 0.6.5.9
200 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
201 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
202 stack overflow. For more information, please refer to bugzilla 17630.
203 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
204 the re-exporting nodes to be mounted with "32bitapi" mount option
206 --------------------------------------------------------------------------------
208 12-07-2016 Intel Corporation
210 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
211 for currently supported client and server kernel versions.
212 * Server known to build on patched kernels:
213 2.6.32-431.29.2.el6 (RHEL6.5)
214 2.6.32-504.30.3.el6 (RHEL6.6)
215 2.6.32-573.26.1.el6 (RHEL6.7)
216 2.6.32-642.11.1.el6 (RHEL6.8)
217 3.10.0-514.el7 (RHEL7.3)
218 3.0.101-0.47.71 (SLES11 SP3)
219 3.0.101-80 (SLES11 SP4)
220 3.12.62-60.64.8 (SLES12 SP1)
221 vanilla linux 4.5.7 (ZFS only)
222 * Client known to build on unpatched kernels:
223 2.6.32-431.29.2.el6 (RHEL6.5)
224 2.6.32-504.30.3.el6 (RHEL6.6)
225 2.6.32-573.26.1.el6 (RHEL6.7)
226 2.6.32-642.11.1.el6 (RHEL6.8)
227 3.10.0-514.el7 (RHEL7.3)
228 3.0.101-0.47.71 (SLES11 SP3)
229 3.0.101-80 (SLES11 SP4)
230 3.12.62-60.64.8 (SLES12 SP1)
232 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
233 * Recommended ZFS / SPL version: 0.6.5.8
234 * Tested with ZFS / SPL version: 0.6.5.8
235 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
236 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
237 stack overflow. For more information, please refer to bugzilla 17630.
238 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
239 the re-exporting nodes to be mounted with "32bitapi" mount option
241 --------------------------------------------------------------------------------
243 02-29-2016 Intel Corporation
245 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
246 for currently supported client and server kernel versions.
247 * Server known to build on patched kernels:
248 2.6.32-431.29.2.el6 (RHEL6.5)
249 2.6.32-504.30.3.el6 (RHEL6.6)
250 2.6.32-573.12.1.el6 (RHEL6.7)
251 3.10.0-327.3.1.el7 (RHEL7.2)
252 3.0.101-0.47.71 (SLES11 SP3)
253 3.0.101-68 (SLES11 SP4)
254 vanilla linux 4.2.1 (ZFS only)
255 * Client known to build on unpatched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.12.1.el6 (RHEL6.7)
259 3.10.0-327.3.1.el7 (RHEL7.2)
260 3.0.101-0.47.71 (SLES11 SP3)
261 3.0.101-68 (SLES11 SP4)
264 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
265 * Recommended ZFS / SPL version: 0.6.4.2
266 * Tested with ZFS / SPL version: 0.6.4.2
267 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
268 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
269 stack overflow. For more information, please refer to bugzilla 17630.
270 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
271 the re-exporting nodes to be mounted with "32bitapi" mount option
273 --------------------------------------------------------------------------------
275 03-10-2015 Intel Corporation
277 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
278 for currently supported client and server kernel versions.
279 * Server known to build on patched kernels:
280 2.6.32-431.29.2.el6 (RHEL6.5)
281 2.6.32-504.8.1.el6 (RHEL6.6)
282 3.0.101-0.46 (SLES11 SP3)
283 * Client known to build on unpatched kernels:
284 2.6.32-431.29.2.el6 (RHEL6.5)
285 2.6.32-504.8.1.el6 (RHEL6.6)
286 3.10.0-123.20.1.el7 (RHEL7)
287 3.0.101-0.46 (SLES11 SP3)
288 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
289 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
290 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
291 stack overflow. For more information, please refer to bugzilla 17630.
292 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
293 the re-exporting nodes to be mounted with "32bitapi" mount option
295 Severity : enhancement
297 Description: control OST-index in IDIF via ROCOMPAT flag.
298 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
299 last_rcvd file. For new formatted OST device, it will be auto set;
300 for the case of upgrading from old OST device, you can enable it
301 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
302 enabled, for new created OST-object, its IDIF-in-LMA will contain
303 the OST-index; for the existing OST-object, the OSD will convert
304 old format IDIF as new format IDIF with OST-index stored in the
305 LMA EA when accessing such OST-object or via OI scrub. Once such
306 flag is enabled, it cannot be reverted back, so the system cannot
307 be downgraded to the orignal incompatible version.
309 --------------------------------------------------------------------------------
311 07-30-2014 Intel Corporation
313 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
314 for currently supported client and server kernel versions.
315 * Server known to build on patched kernels:
316 2.6.32-431.20.3.el6 (RHEL6.5)
317 3.0.101-0.31 (SLES11 SP3)
318 * Client known to build on unpatched kernels:
319 2.6.32-431.20.3.el6 (RHEL6.5)
320 3.0.101-0.31 (SLES11 SP3)
323 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
324 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
325 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
326 stack overflow. For more information, please refer to bugzilla 17630.
327 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
328 the re-exporting nodes to be mounted with "32bitapi" mount option
330 --------------------------------------------------------------------------------
332 10-31-2013 Intel Corporation
334 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
335 * Server builds for kernels:
336 2.6.32-358.18.1.el6 (RHEL6.4)
337 * Client builds for unpatched kernels:
338 2.6.32-358.18.1.el6 (RHEL6.4)
339 3.0.92-0.8 (SLES11 SP3)
341 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
342 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
343 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
344 stack overflow. For more information, please refer to bugzilla 17630.
345 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
346 the re-exporting nodes to be mounted with "32bitapi" mount option
348 --------------------------------------------------------------------------------
350 05-30-2013 Intel Corporation
352 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
353 * Server builds for kernels:
354 2.6.32-358.6.2.el6 (RHEL6.4)
355 * Client builds for unpatched kernels:
356 2.6.32-358.6.2.el6 (RHEL6.4)
357 3.0.74-0.6.6 (SLES11 SP2)
359 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
360 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
361 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
362 stack overflow. For more information, please refer to bugzilla 17630.
363 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
364 the re-exporting nodes to be mounted with "32bitapi" mount option
366 --------------------------------------------------------------------------------
368 9-30-2012 Whamcloud, Inc.
370 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
371 * Server support for kernels:
372 2.6.32-279.5.1.el6 (RHEL6)
373 * Client support for unpatched kernels:
374 2.6.32-279.5.1.el6 (RHEL6)
376 2.6.18-238.19.1.el5 (RHEL5)
377 2.6.16.60-0.69.1 (SLES10)
378 * Recommended e2fsprogs version: 1.41.90.wc4
379 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
380 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
381 stack overflow. For more information, please refer to bugzilla 17630.
382 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
383 the re-exporting nodes to be mounted with "32bitapi" mount option
385 Severity : enhancement
387 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
389 Severity : enhancement
391 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
393 Severity : enhancement
395 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
397 Severity : enhancement
399 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
401 Severity : enhancement
403 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
405 --------------------------------------------------------------------------------
407 04-30-2012 Whamcloud, Inc.
409 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
410 * Server support for kernels:
411 2.6.32-220.4.2.el6 (RHEL6)
412 * Client support for unpatched kernels:
413 2.6.32-220.4.2.el6 (RHEL6)
415 2.6.18-238.19.1.el5 (RHEL5)
416 2.6.16.60-0.69.1 (SLES10)
417 * Recommended e2fsprogs version: 1.41.90.wc4
418 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
419 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
420 stack overflow. For more information, please refer to bugzilla 17630.
421 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
422 the re-exporting nodes to be mounted with "32bitapi" mount option
424 Severity : enhancement
426 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
428 Severity : enhancement
430 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
432 Severity : enhancement
434 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
436 Severity : enhancement
438 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
440 --------------------------------------------------------------------------------
442 09-30-2011 Whamcloud, Inc.
444 * Server support for kernels:
445 2.6.18-238.19.1.el5 (RHEL5)
446 2.6.32-131.6.1.el6 (RHEL6)
447 * Client support for unpatched kernels:
448 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
449 2.6.18-238.19.1.el5 (RHEL5)
450 2.6.32-131.6.1.el6 (RHEL6)
451 2.6.29.4-167.fc11 (FC11)
452 2.6.16.60-0.69.1 (SLES10)
454 * Recommended e2fsprogs version: 1.41.90.wc3
455 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
456 Lustre filesystem with 4K stack may cause a stack overflow. For more
457 information, please refer to bugzilla 17630.
458 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
459 mounted with "32bitapi" mount option
461 Severity : enhancement
462 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
464 Severity : enhancement
465 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
467 Severity : enhancement
468 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
471 Frequency : only with SLES10
473 Description: Use OFED "KMP" provided by Novell
474 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
475 Lustre is now built against this package. That means you need to
476 install the ofed-kmp package from Novell for the patchless client.
477 Note that the ofed-kmp that Novell ships may not exactly match the
478 kernel version but should still be compatible.
480 Severity : enhancement
482 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
483 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
485 --------------------------------------------------------------------------------
487 2010-07-15 Oracle, Inc.
489 * Server support for kernels:
490 2.6.18-164.11.1.el5 (RHEL 5)
491 2.6.18-164.11.1.0.1.el5 (OEL 5)
492 * Client support for unpatched kernels:
493 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
494 2.6.18-164.11.1.el5 (RHEL 5),
495 2.6.18-164.11.1.0.1.el5 (OEL 5)
496 2.6.16.60-0.42.8 (SLES 10),
498 2.6.29.4-167.fc11 (FC11)
499 * Recommended e2fsprogs version: 1.41.10-sun2
501 Severity : enhancement
503 Description: Distribute lfs_migrate script with lustre
504 Details : The lfs_migrate mentioned in the manual is now included in the
505 lustre tree in lustre/scripts and will distributed in lustre RPMs.
507 Severity : enhancement
509 Description: Failover nid list added to import proc file.
511 Severity : enhancement
513 Description: "som_preview" mount option required to use SOM
514 Details : The clients will need to use som_preview mount option
518 Description: Lustre does not do 1MB IOs to HW RAID
519 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
521 Severity : enhancement
523 Description: lfs quota output is not very convenient for awk/sed-parsing
524 Details : Some positions in lfs quota output table could be empty or
525 non-empty which made it hard to parse it with scripts, now
526 a dash is put instead of space where there is not supposed
531 Description: Update to OFED 1.5.1
532 Details : OFED 1.5.1 will be used in environments that don't use
533 vendor supplied OFED.
537 Description: Remove lnet. prefix from lctl params display
538 Details : Also renamed lnet.memused to lnet_memused to
543 Description: Use conf_param -d to remove permanant settings.
547 Description: Add list_param -R
548 Details : lctl list_param now supports recursive mode.
551 Bugzilla : 21829, 22312
552 Description: Fix llobdstat.
553 Details : 1. Now updating the write_bytes counter
554 2. Port script fixes from 1.8.
558 Description: decrease the usage of memory on clients.
559 Details : 1. On clients, recycle dentries and inodes unused.
560 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
565 Description: llobdstat fix and enhancement
566 Details : add a counter to set a limit to how many samples will be returned
567 fix a wildcard in the path to limit to obdfilter stats only
571 Description: lfs check only allowed for root
572 Details : Code cleanups and added a regresion test of non-root use of
577 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
578 Details : Fix hr->hr_index overload.
582 Description: Fix FIEMAP functionality.
586 Description: FC11 patchless client support added
590 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
591 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
595 Description: New acc-sm script and various test-framework enhancments.
599 Description: Previously landed patch was reverted.
603 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
607 Description: Add ldiskfs maxdirsize mount option.
611 Description: Fix mkfs.lustre support for 16TB LUNs.
615 Description: Print mmp_check_interval
616 Details : print mmp_check_interval and make it possible to abort
617 mount operation in case it takes too long.
621 Description: Add cascading_rw to lustre mpi tests.
625 Description: Store SOM data in EA.
626 Details : Also several SOM related cleanups.
630 Description: Add refcount to md_open_data
634 Description: Use refcount on lov_request_set to prevent memory corruption.
638 Description: Statahead: use parent dir's i_mutex for synchronization.
642 Description: loadgen improvements.
646 Description: Enable peer health detection in ksocklnd and ko2iblnd.
650 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
654 Description: Fix race on obd_devcie:md_stats
658 Description: Fix SLES 11 client builds.
659 Details : Resolved uninitalized pointer gcc warnings.
663 Description: Allow non-root access for "lfs check".
664 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
668 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
670 Severity : enhancement
672 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
673 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
677 Description: Sanity test_27p fails.
678 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
679 Sanity tests 27 were cleaned up and improved.
683 Description: Unify req format on client/servers
684 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
685 buffers and swabbers approach..
690 Description: drop join file stuff.
692 Severity : enhancement
694 Description: Add 2.6.27 and SLES11 (patchless client) support
698 Description: Fix in ptlrpc_expire_one_request() to print the signed time
701 Severity : enhancement
703 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
706 Severity : enhancement
708 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
710 Severity : enhancement
712 Description: correctly handle big reply message.
713 Details : send LNet event if reply is bigger than buffer and adjust this buffer
718 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
720 Severity : enhancement
722 Description: Add support for OEL5.
726 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
728 Severity : enhancement
730 Description: "weak-modules" support
731 Details : Implement "weak-modules" support which enables kernel modules
732 to be used with any kernel that implements the same kABI. In
733 order to achieve this modules are now installed in
734 /lib/modules/$(uname -r)/updates/kernel on all distributions.
738 Description: client could not reconnect to OST because of an active request.
739 Details : The client now retries to reconnect to the same server, if a
740 connect request failed with EBUSY or -EAGAIN. Test case 35b
741 conf-sanity.sh has been created.
745 Description: Don't allow make backward step on assignin osc next id.
746 Details : race between allocation next id and ll_sync thread can be cause
747 of set wrong osc next id and can be kill valid ost objects.
751 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
755 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
759 Description: Error handling in osc_statfs_interpret() has been improved.
760 Details : Check in osc_statfs_interpret() for EBADR.
764 Description: Do not send statfs() requests to OSTs disabled by administrator.
765 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
767 Severity : enhancement
769 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
771 Severity : enhancement
773 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
775 Severity : enhancement
777 Description: Enhancement for lfs(1) command to use numeric uid/gid.
781 Description: Improved cleanup procedure for conf-sanity test_46a.
785 Description: Do not start bulk transfer, if client tried to reconnect.
786 Details : Do not start bulk transfer, if client tried to reconnect,
787 but failed due to active rpcs. Check for exp_abort_active_req
788 has been implemented.
792 Description: Prevent inconsistences between linux and lustre mount structures.
793 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
794 Make the sleep interruptible.
796 Severity : enhancement
798 Description: provide server to client comms path
799 Details : server to client communications path via new LDLM_SET_INFO rpc
803 Description: truncate starts GFP_FS allocation under transaction causing deadlock
804 Details : ldiskfs_truncate calls grab_cache_page which may start page
805 allocation under an open transaction. This may lead to
806 calling prune_icache with consequent lustre reentrance.
810 Description: send correctly lsm on open replay
811 Details : MDS is trust to LSM size on replay open, but client can set wrong size
816 Description: enable client interface failover
817 Details : When a child reconnects from another NID, properly update export
818 nid hash position and ldlm reverse import.
820 Severity : enhancement
822 Description: Communicate OST degraded/readonly state via statfs to MDS
823 Details : Flags in the statfs returned from OSTs indicate whether the
824 OST is in a degraded RAID state, or if the filesystem has
825 turned read-only after a filesystem error is detected.
830 Description: don't panic if EPROTO was hit when reading symlink
831 Details : correctly handling request reference in error cases.
833 Severity : enhancement
835 Description: Add LustreNetLink, a kernel-userspace communcation path.
837 Severity : enhancement
839 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
844 Description: access to llog context before init.
845 Details : move handling CATALOGS file at osc layer and forbid access to llog
849 Frequency : rare, only when client is writing to OST on local node
851 Description: Avoid deadlock for local client writes
852 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
853 memory freeing context. This allows OST threads to set the
854 PF_MEMALLOC flag on task structures in order to allocate memory
855 from reserved pools and complete IO.
856 Use GFP_HIGHUSER for OST allocations for non-local client writes,
857 so that the OST threads generate memory pressure and allow
858 inactive pages to be reclaimed.
860 Severity : enhancement
862 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
867 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
868 Details : move ldlm namespace creation in setup phase to avoid grab
869 _lprocfs_lock with cli_sem held.
873 Description: Temporarily disable grant shrink.
874 Details : Disable the feature for debugging.
877 Frequency : only when formatting test filesystems
879 Description: Unable to run several mkfs.lustre on loop devices at the same time
880 Details : mkfs.lustre returns error 256 on the concurrent loop devices
881 formatting. The solution is to proper handle the error.
883 Severity : enhancement
885 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
887 Severity : enhancement
889 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
894 Description: resolve race between obd_disconnect and class_disconnect_exports
895 Details : if obd_disconnect will be called to already disconnected export he
896 forget release one reference and osc module can't unloaded.
898 Severity : enhancement
900 Description: Allow tuning service thread via /proc
901 Details : For each service a new
902 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
903 created that can be used to set min/max thread counts, and get the
904 current number of running threads.
909 Descriptoin: don't return error if have partially created objects for file.
910 Details : lov_update_create_set uses set->set_success as index for created
911 objects, so if some requests will be failed, they will have hole
912 at end of array and qos_shrink_lsm for allocate correct lsm.
914 Severity : enhancement
916 Description: Update OFED support to 1.4.1
921 Description: Reduce small size read RPC
922 Details : Set read-ahead limit for every file and only do read-ahead when
923 available read-ahead pages are bigger than 1M to avoid small size
926 Severity : enhancement
928 Description: move AT tunable parameters for more consistent usage
929 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
931 Severity : enhancement
933 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
934 Details : allow skip disconnected ost for send statfs request and hide error
940 Description: fix racy locking of mballoc block bitmaps causing BUG
941 Details : The locking of the mballoc buddy bitmap and the in-memory
942 block bitmap was using two different spin locks in some
943 cases. This made it possible to incorrectly access the
944 mballoc bitmap while another process was modifying it,
945 causing a sanity assertion to fail. While no on-disk corruption
946 was reported, there was some risk of this happening.
949 Frequency : rare, on llog test 6
951 Descriptoin: don't allow connect to already connected import
952 Details : allowing connect to already connected import is hide connecting problem.
957 Descriptoin: allow kill process which wait statahead result
958 Details : for some reasons 'ls' can stick in waiting result from statahead,
959 in this case need way for kill this process.
961 Severity : enhancement
963 Description: Add state history info file, enhance import info file
964 Details : Track import connection state changes in a new osc/mdc proc file;
965 add overview-type data to the osc/mdc import proc file.
967 Severity : enhancement
969 Description: MDS create should not wait for statfs RPC while holding DLM lock.
971 Severity : enhancement
973 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
978 Descriptoin: OST grant too much space to client even there are not enough space.
979 Details : Client will shrink its grant cache to OST if there are no write
980 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
981 this grant cache if there are already not enough avaible space
982 (left_space < total_clients * 32M).
985 Frequency : start MDS on uncleanly shutdowned MDS device
987 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
988 Details : stay in waiting mds<>ost recovery finished produce random bugs
989 due race between two ll_sync thread for one lov target. send
990 ACTIVATE event only if connect realy finished and import have
994 Frequency : rare, connect and disconnect target at same time
996 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
997 Details : don't call obd_disconnect under lov_lock. this long time
998 operation and can block ptlrpcd which answer to connect request.
1003 Descriptoin: don't lose wakeup for imp_recovery_waitq
1004 Details : recover_import_no_retry or invalidate_import and import_close can
1005 both sleep on imp_recovery_waitq, but we was send only one wakeup
1009 Frequency : always with long access acl
1011 Descriptoin: mds can't pack reply with long acl.
1012 Details : mds don't control size of acl but they limited by reint/getattr
1015 Severity : enhancement
1017 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1019 Severity : enhancement
1021 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1024 Frequency : start MDS on uncleanly shutdowned MDS device
1026 Descriptoin: aborting recovery hang on MDS
1027 Details : don't throttle destroy RPCs for the MDT.
1030 Frequency : on remount
1032 Description: external journal device not working after the remount
1033 Details : clear dev_rdonly flag for external journal devices in
1039 Description: shutdown vs evict race
1040 Details : client_disconnect_export vs connect request race.
1041 if client will evicted at this time - we start invalidate
1042 thread without referece to import and import can be freed
1046 Frequency : rare, need acl's on inode.
1048 Description: client can't handle ost additional correctly
1049 Details : if ost was added after client connected to mds client can have
1050 hit lnet_try_match_md ... to big messages to wide striped files.
1051 in this case need teach client to handle config events about add
1052 lov target and update client max ea size at that event.
1054 Severity : enhancement
1056 Description: Changelogs
1057 Details : Changelogs are a lightweight mechanism to track filesystem
1058 metadata and namespace changes. The changelog is recorded
1059 permanently on the MDTs, and is periodically "consumed" / purged
1060 when records are no longer needed.
1062 Severity : enhancement
1064 Description: compact fld format with extents
1065 Details : Store range of seq rather than every seq in FLD. Seq
1066 controller update FLD rather than clients. In Case of CMD, mdt0
1067 has FLD, all other metadata server act as non persistent proxy
1068 for FLD queries and cache fld entries in fld cache.
1073 Description: don't skip ost target if they assigned to file
1074 Details : Drop slow OSCs if we can, but not for requested start idx.
1075 This means "if OSC is slow and it is not the requested
1076 start OST, then it can be skipped, otherwise skip it only
1077 if it is inactive/recovering/out-of-space.
1081 Description: more cleanup in mds_lov
1082 Details : not send LOV EA under replay, we can't know about they size at this
1083 time. Don't allow client connect to mds before any ost connected,
1084 for avoid problems with LOV EA size and returning EIO to client.
1086 Severity : enhancement
1088 Description: Interoperability at server side (Disk interoperability)
1090 Severity : enhancement
1092 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1094 Severity : enhancement
1096 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1098 Severity : enhancement
1100 Description: New client IO stack (CLIO).
1102 Severity : enhancement
1104 Description: Commit on sharing. Eliminate inter-client dependencies between
1105 uncommitted transactions by doing transaction commits.
1106 Thereby clients may recovery independently.
1109 Frequency : Create a symlink file with a very long name
1111 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1112 Details : If there is no extra space in the request for early cancels,
1113 ldlm_req_handles_avail() returns 0 instead of a negative value.
1115 Severity : enhancement
1117 Description: Add /proc entry for import status
1118 Details : The mdc, osc, and mgc import directories now have
1119 an import directory that contains useful import data for debugging
1120 connection problems.
1122 Severity : enhancement
1124 Description: Re-disable certain /proc logging
1125 Details : Enable and disable client's offset_stats, extents_stats and
1126 extents_stats_per_process stats logging on the fly.
1129 Frequency : Only on FC kernels 2.6.22+
1131 Description: oops in statahead
1132 Details : Do not drop reference count for the dentry from VFS when lookup,
1133 VFS will do that by itself.
1135 Severity : enhancement
1137 Description: Generic /proc file permissions
1138 Details : Set /Proc file permissions in a more generic way to enable non-
1139 root users operate on some /proc files.
1143 Description: Hitting mdc_commit_close() ASSERTION
1144 Details : Properly handle request reference release in
1145 ll_release_openhandle().
1149 Description: quota recovery deadlock during mds failover
1150 Details : This patch includes att18982, att18236, att18237 in bz14840.
1152 1. fix osts hang when mds does failover with quotaon
1153 2. prevent watchdog storm when osts threads wait for the
1158 Frequency : only patchless client
1159 Description: add workaround for race between add/remove dentry from hash
1161 Severity : enhancement
1163 Description: Allow OST glimpses to return PW locks
1167 Description: LBUG when llog conf file is full
1168 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1172 Description: Prevent import from entering FULL state when server in recovery
1176 Description: service mount cannot take device name with ":"
1177 Details : Only when device name contains ":/" will mount treat it as
1183 Description: replace ptlrpcd with the statahead thread to interpret the async
1184 statahead RPC callback
1188 Frequency : on recovery
1189 Description: I/O failures after umount during fail back
1190 Details : if client reconnected to restarted server we need join to recovery
1191 instead of find server handler is changed and process self
1192 eviction with cancel all locks.
1194 Severity : enhancement
1196 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1198 Severity : enhancement
1200 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1202 Severity : enhancement
1204 Description: Upcall on Lustre log has been dumped
1205 Details : Allow for a user mode script to be called once a Lustre log has
1206 been dumped. It passes the filename of the dumped log to the
1207 script, the location of the script can be specified via
1208 /proc/sys/lnet/debug_log_upcall.
1213 Description: avoid idr_remove called for id which is not allocated.
1214 Details : Move assigment s_dev for clustered nfs to end of initialization,
1215 to avoid problem with error handling.
1220 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1221 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1222 into CONN_USED_HASH and this prodice warning when put connection
1223 again in unused hash.
1226 Severity : enhancement
1228 Description: File striping can now be set to use an arbitrary pool of OSTs.
1230 Severity : enhancement
1232 Description: Export bytes_read/bytes_write count on OSC/OST.
1236 Description: Early reply size mismatch, MGC loses connection
1237 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1238 the connect flags are properly negotiated.
1243 Description: quotas are not honored with O_DIRECT
1244 Details : all writes with the flag O_DIRECT will use grants which leads to
1245 this problem. Now using OBD_BRW_SYNC to guard this.
1249 Description: add quota statistics
1250 Details : 1. sort out quota proc entries and proc code.
1251 2. add quota statistics
1253 Severity : enhancement
1255 Description: enable quota support for HEAD.
1259 Description: Properly propagate oinfo flags from lov to osc for statfs
1260 Details : restore missing copy oi_flags to lov requests.
1262 Severity : enhancement
1264 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1265 : l_getgroups(8), lst(8), routerstat(8)
1266 Details : included man pages for llobdstat(8), llstat(8),
1267 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1269 Severity : enhancement
1271 Description: configure's --enable-quota should check the
1272 : kernel .config for CONFIG_QUOTA
1273 Details : configure is terminated if --enable-quota is passed but
1274 : no quota support is in kernel
1278 Description: Remove portals compatibility
1279 Details : Remove portals compatibility, not interoperable with releases
1284 Description: Resolve device initialization race
1285 Details : Prevent proc handler from accessing devices added to the
1286 obd_devs array but yet be intialized.
1288 Severity : enhancement
1290 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1292 Severity : enhancement
1294 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1299 Description: Using wrong pointer in osc_brw_prep_request
1300 Details : Access to array[-1] can produce panic if kernel compiled with
1301 CONFIG_PAGE_ALLOC enabled
1305 Description: Client runs out of low memory
1306 Details : Consider only lowmem when counting initial number of llap pages
1310 Description: *optional* service tags registration
1311 Details : if the "service tags" package is installed on a Lustre node
1312 When the filesystem is mounted, a local-node service tag will
1313 be created. See http://inventory.sun.com/ for more information
1314 about the Service Tags asset management system.
1318 Description: Kernel BUG tries to release flock
1319 Details : Lustre does not destroy flock lock before last reference goes
1320 away. So always drop flock locks when client is evicted and
1321 perform unlock regardless of successfulness of speaking to MDS.
1325 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1329 Description: sanity test 65a fails if stripecount of -1 is set
1330 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1335 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1336 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1339 Severity : enhancement
1341 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1345 Description: do not process already freed flock
1346 Details : flock can possibly be freed by another thread before it reaches
1347 to ldlm_flock_completion_ast.
1351 Description: LBUG during stress test
1352 Details : Need properly lock accesses the flock deadlock detection list.
1356 Description: oops in page fault handler
1357 Details : kernel page fault handler can return two special 'pages' in error
1358 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1362 Description: timeout with invalidate import.
1363 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1364 should be handled by ptlrpcd. This produce long age waiting and
1365 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1367 Severity : enhancement
1369 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1373 Description: enable MGS and MDT services start separately
1374 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1375 a co-located MGS without starting the MGS, which is a complement
1376 to 'nosvc' mount option.
1380 Frequency : after recovery
1381 Description: precreate to many object's after del orphan.
1382 Details : del orphan st in oscc last_id == next_id and this triger growing
1383 count of precreated objects. Set flag LOW to skip increase count
1384 of precreated objects.
1388 Frequency : rare, on clear nid stats
1389 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1390 Details : when clean nid stats sometimes try destroy live entry,
1391 and this produce panic in free.
1395 Description: Stack overflow during MDS log replay
1396 ease stack pressure by using a thread dealing llog_process.
1400 Description: wait until IO finished before start new when do lock cancel.
1401 Details : VM protocol want old IO finished before start new, in this case
1402 need wait until PG_writeback is cleared until check dirty flag and
1403 call writepages in lock cancel callback.
1405 Severity : enhancement
1407 Description: using special macro for print time and cleanup in includes.
1411 Description: mds_mfd_close() ASSERTION(rc == 0)
1412 Details : In mds_mfd_close(), we need protect inode's writecount change
1413 within its orphan write semaphore to prevent possible races.
1417 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1421 Frequency : rare, on shutdown ost
1422 Description: don't hit live lock with umount ost.
1423 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1424 use shrink_dcache_sb instead.
1428 Description: don't panic with use echo client
1429 Details : echo client pass NULL as client nid pointer and this produce null
1430 pointer dereference.
1434 Description: fix build on ppc32
1435 Details : compile code with -m64 flag produce wrong object file for ppc32.
1439 Description: add message levels for liblustreapi
1443 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1444 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1445 will result in an -EAGAIN and in the exit call path we call
1446 alloc_rr() which will with increasing aggressiveness attempt to
1447 aquire precreated objects on the minimum number of required OSCs.
1451 Description: Use old size assignment to avoid deadlock
1452 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1453 the scheduling while holding a spinlock. We do not need locking
1454 for size in ll_update_inode() because size is only updated from
1455 the MDS for directories or files without objects, so there is no
1456 other place to do the update, and concurrent access to such inodes
1457 are protected by the inode lock.
1461 Description: resolve "_IOWR redefined" build error on SLES10
1465 Description: dump the memory debugging after all modules are unloaded to
1466 suppress false negative in conf_sanity test 39
1468 Severity : enhancement
1470 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1475 Description: fix SLES kernel versioning
1476 Details : the kernel version for our SLES 10 kernel did not include a "-"
1477 before the "smp" at the end. while this was not a problem in
1478 general, it did mean that software trying to use the kernel
1479 version to try to detect a vendor specific kernel would fail.
1480 this was most evident by the OFED build scripts.
1484 Description: Don't update lov_desc members until making sure they are valid
1485 Details : When updating lov_desc members via proc fs, need fix their
1486 validities before doing the real update.
1490 Description: don't put request into delay list while invalidate in flight.
1491 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1492 import in flight. this produce timeout for invalidate and sometimes
1493 can cause stale data.
1497 Frequency : on ppc only
1498 Description: not convert ost objects for directory because it's not exist.
1499 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1503 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1505 Severity : enhancement
1507 Description: Update to support 2.6.22.14 vanilla kernel.
1511 Frequency : rare, on recovery
1512 Description: read procfs can produce deadlock in some situation
1513 Details : Holding lprocfs lock which send rpc can produce block for destroy
1514 obd objects and this also block reconnect with -EALREADY. This
1515 isn't fix all lprocfs bugs - but make it rare.
1517 Severity : enhancement
1519 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1522 Frequency : frequent on X2 node
1524 Description: mdc_set_open_replay_data LBUG
1525 Details : Set replay data for requests that are eligible for replay.
1529 Description: lustre_mgs: operation 101 on unconnected MGS
1530 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1531 MGC, and late on MGC cannot successfully connect to MGS and a lot
1532 of the error messages complaining that MGS is not connected.
1536 Frequency : on network error
1537 Description: panic with double free request if network error
1538 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1539 but it's true only for synchronus enqueue, for async enqueue
1540 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1543 Severity : enhancement
1545 Description: client-side metadata stat-ahead during readdir(directory readahead)
1546 Details : perform client-side metadata stat-ahead when the client detects
1547 readdir and sequential stat of dir entries therein
1550 Frequency : on start mds
1552 Description: Implement get_info(last_id) in obdfilter.
1555 Frequency : occasional
1557 Description: Correctly check stale fid, not start epoch if ost not support SOM
1558 Details : open with flag O_CREATE need set old fid in op_fid3 because
1559 op_fid2 was overwritten with new generated fid, but mds can answer
1560 with one of these two fids and both is not stale. Setattr
1561 incorrectly started an epoch and assume will be called
1562 done_writing, but without SOM done_writing ever being called.
1565 Frequency : rare, depends on device drivers and load
1567 Description: MDS or OSS nodes crash due to stack overflow
1568 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1569 In some cases, in conjunction with device drivers that use a lot
1570 of stack the MDS (or possibly OSS) service threads could overflow
1571 the stack. One change which was identified to consume additional
1572 stack has been reworked to avoid the extra stack usage.
1575 Frequency : occasional
1577 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1578 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1579 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1580 oscc_has_objects and see OSC has no objests and call
1581 oscc_internal_create to resend create request.
1583 Severity : enhancement
1585 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1587 Severity : enhancement
1589 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1592 Frequency : very rare
1594 Description: Fix replay if there is an un-replied request and open
1595 Details : In some cases, older replay request will revert the
1596 mcd->mcd_last_xid on MDS which is used to record the client's
1597 latest sent request.
1599 Severity : enhancement
1601 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1603 Severity : enhancement
1605 Description: Add rhel5 support to HEAD.
1607 Serverity : enhancement
1609 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1614 Description: Don't allow skipping OSTs if index has been specified.
1615 Details : Don't allow skipping OSTs if index has been specified, make
1616 locking in internal create lots better.
1620 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1621 Details : access to bitfield in structure is always rounded to long
1622 and this produce problem with not atomic change any bit.
1626 Description: Lustre make rpms failed.
1627 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1628 builds Lustre rpms from tarball.
1631 Frequency : rare on shutdown ost
1633 Description: If llog cancel was not send before clean_exports phase, this can
1634 produce deadlock in llog code.
1635 Details : If llog thread has last reference to obd and call class_import_put
1636 this produce deadlock because llog_cleanup_commit_master wait when
1637 last llog_commit_thread exited, but this never success because was
1638 called from llog_commit_thread.
1642 Description: allow userland application know is lost one of stripes.
1643 Details : fill lvb_blocks with error code on ost and return it to
1644 application if error flag found.
1648 Description: NULL lov_tgts causing MDS oops
1649 Details : more safe checks for NULL lov_tgts for avoid oops.
1651 Severity : enhancement
1653 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1657 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1658 Details : If an OST is down the MDS will hang indefinitely in
1659 obd_statfs_rqset() waiting for the statfs data. While for
1660 MDS QOS usage of statfs, it should not stuck in waiting.
1662 Severity : enhancement
1664 Description: remote_acl support
1665 Details : Support ACL-based permission check for remote user.
1666 Support setfacl/getfacl for remote user with the utils
1667 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1668 the system "{s,g}etfacl" utils.
1670 Severity : enhancement
1672 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1674 Severity : enhancement
1676 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1680 Description: make mgs_setparam() handle fsname containing dash
1681 Details : fsname containing a dash does not work with lctl conf_param
1683 Severity : enhancement
1685 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1687 Severity : enhancement
1689 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1693 Frequency : rare, at shutdown
1694 Description: access already free / zero obd_namespace.
1695 Details : if client_disconnect_export was called without force flag set,
1696 and exist connect request in flight, this can produce access to
1697 NULL pointer (or already free pointer) when connect_interpret
1698 store ocd flags in obd_namespace.
1702 Frequency : only at startup
1703 Description: not alloc memory with spinlock held.
1704 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1705 if any spinlock held.
1707 Severity : enhancement
1709 Description: make lustre randomly fail allocating memory
1710 Details : Make lustre randomly failed allocating memory for testing purpose.
1712 Severity : enhancement
1714 Description: lost problems with lov objid file
1715 Details : Fixes some scability and access to not inited memory problems
1716 in work with lov objdid file.
1721 Description: lfs find does not continue on file error
1722 Details : Continue other files processing when a file/dir is absent.
1726 Description: Inconsistent usage of lustre_pack_reply()
1727 Details : Standardize the usage of lustre_pack_reply() such that it
1728 always generate a CERROR on failure.
1731 Frequency : occasional
1733 Description: MDS hang or stay in waiting lock
1734 Details : If client receive lock with CBPENDING flag ldlm need send lock
1735 cancel as separate rpc, to avoid situation when cancel request
1736 can't processed due all i/o threads stay in wait lock.
1740 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1744 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1748 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1752 Description: Update to RHEL4 latest kernel.
1754 Severity : enhancement
1756 Description: Build SLES10 patchless client fails
1757 Details : The configure was broken by run ./configure with
1758 --with-linux-obj=.... argument for patchless client. When the
1759 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1760 can't search header adequately. Use absolute path such as
1761 -I($LINUX)/include instead.
1765 Description: interrupt oig_wait produce painc on resend.
1766 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1767 private set, and this produce situation when rq_ptlrpcd_data not
1768 copyed to new allocated request and triggered LBUG on assert
1769 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1770 ptlrpcd set if request is changed and can be safety changed to use
1775 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1776 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1777 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1778 these macros and replace them with their difinitions instead.
1782 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1783 Details : Update which_patch & target file for SLES10 latest kernel.
1786 Description: Remove client patches from SLES 10 kernel.
1787 Details : This causes SLES 10 clients to behave as patchless clients
1788 even on a Lustre-patched (server) kernel.
1790 Severity : enhancement
1792 Description: self-adjustable client's lru lists
1793 Details : use adaptive algorithm for managing client cached locks lru
1794 lists according to current server load, other client's work
1795 pattern, memory activities, etc. Both, server and client
1796 side namespaces provide number of proc tunables for controlling
1799 Severity : enhancement
1801 Description: light-weight GSS support
1802 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1803 and reduce performance overhead.
1805 Severity : enhancement
1807 Description: Linux keyring support
1808 Details : Support using service of Linux keyring for Lustre GSS internal
1809 context refresh/cache mechanism.
1813 Description: Fix errors in lfs documentation
1814 Details : Fixes man pages
1818 Description: don't use GFP_* in generic Lustre code.
1819 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1823 Description: obdclass is limited by single OBD_ALLOC(idarray)
1824 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1828 Description: warnings with build patchless client with vanila 2.6.19 and up
1829 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1830 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1834 Frequency : Only for SLES
1835 Description: sanity_quota fail test_1
1836 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1837 file, which makes TSTID[2] inunique.
1841 Frequency : Only for SLES9
1842 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1843 Details : Update target/ChangeLog/which_patch .
1847 Frequency : Only for test_10 in sanity-quota.sh
1848 Description: a bug in quota test code
1849 Details : checking a test flag in a wrong place
1853 Frequency : Only for test_13 in sanity-quota.sh
1854 Description: a bug in quota test script
1855 Details : shouldn't check overfull allocated quota
1859 Description: jbd statistics
1860 Details : Port older jbd statistics patch for sles10
1862 Severity : enhancement
1864 Description: Add printing inode info into message about error in writepage.
1869 Description: reply_lock_interpret crash due to race with it and lock cancel.
1870 Details : Do not replay locks that are being cancelled. Do not reference
1871 locks by their address during replay, just by their handle.
1875 Frequency : When flocks are used.
1876 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1877 Details : Flock locks might destroy just granted lock if it could be merged
1878 with another existing flock, this is done in completion handler,
1879 so teach ldlm_cli_enquque_fini that this is a valid case for
1885 Description: Oops in read and write path when failing to allocate lock.
1886 Details : Check if lock allocation failed and return error back.
1890 Description: lstripe command fails for valid OST index
1891 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1892 instead of lov->desc.ld_active_tgt_count.
1896 Description: sanity.sh failed test 103
1897 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1902 Description: df doesn't work properly if diskfs blocksize != 4K
1903 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1907 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1908 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1910 Severity : enhancement
1912 Description: Adaptive timeouts
1913 Details : RPC timeouts adapt to changing server load and network
1914 conditions to reduce resend attempts and improve recovery time.
1918 Description: llapi_file_create() does not allow some changes
1919 Details : add llapi_file_open() that allows specifying the mode and
1920 open flags, and also returns an open file handle.
1924 Description: merge and cleanup kernel patches.
1925 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1929 Description: Add journal checksum support.(Kernel part)
1930 Details : The journal checksum feature adds two new flags i.e
1931 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1932 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1933 indicates that the commit block contains the checksum for
1934 the blocks described by the descriptor blocks. Now commit
1935 record can be sent to disk without waiting for descriptor
1936 blocks to be written to disk. This behavior is controlled
1937 using JBD2_FEATURE_ASYNC_COMMIT flag.
1941 Description: OSS needs mutliple precreate threads
1942 Details : Add ability to start more than one create thread per OSS.
1946 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1947 Details : execute lfs setstripe on client
1951 Description: mds_obd_create error creating tmp object
1952 Details : When the user sets quota on root, llog will be affected and can't
1953 create files and write files.
1956 Frequency : Always on ia64 patchless client, and possibly others.
1958 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1959 Details : This allows the patchless client to be loaded on architectures
1960 without this export.
1964 Description: RedHat Update kernel for RHEL5
1965 Details : Add the kernel config file for RHEL5.
1969 Description: RedHat Update kernel for RHEL5
1970 Details : Modify the kernel config file more closer RHEL5.
1974 Description: Build failure against Centos5 (RHEL5)
1975 Details : Define PAGE_SIZE when it isn't present.
1979 Description: test_7 on sanity-quota.sh can't be used on separate machines
1980 Details : add facet to handle it
1984 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1985 Details : with NFS, the anon dentry's parent was set to itself in
1986 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1987 corresponding dentry other than use rec->ur_name.
1989 Severity : enhancement
1991 Description: lfs setstripe enhancement
1992 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1996 Description: enable data checksumming by default
1997 Details : enable checksum by default, allow --disable-checksum
1998 configure option and "-o nochecksum" mount option. Checksums
1999 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2003 Description: lustre support for RHEL5
2004 Details : Add support for RHEL5.
2008 Description: Client eviction due to failover config
2009 Details : after a connection loss, the lustre client should attempt to
2010 reconnect to the last active server first before trying the
2011 other potential connections.
2015 Description: when mds and osts use different quota unit(32bit and 64bit),
2016 quota will be released repeatly.
2017 Details : void sending multiple quota reqs to mds, which will keep the
2018 status between the reqs.
2022 Description: osts not allocated evenly to files
2023 Details : change the condition to increase offset_idx
2026 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2028 Description: Data corruption for OSTs that are formatted larger than 2TB
2030 Details : When generating the bio request for lustre file writes the
2031 sector number would overflow a temporary variable before being
2032 used for the IO. The data reads correctly from Lustre (which
2033 will overflow in a similar manner) but other file data or
2034 filesystem metadata may be corrupted in some cases.
2038 Description: Tune the kernel for good SCSI performance.
2039 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2040 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2045 Description: rewrite ext2-derived code in obdclass/uuid.c
2046 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2047 from scratch preserving functionality.
2051 Description: "lfs find -obd UUID" prints directories
2052 Details : "lfs find -obd UUID" will return all directory names instead
2053 of just file names. It is incorrect because the directories
2054 do not reside on the OSTs.
2058 Description: use i_size_read and i_size_write in 2.6 port
2059 Details : replace inode->i_size access with i_size_read/write()
2062 Frequency : when removing large files
2064 Description: scheduling issue during removal of large Lustre files
2065 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2066 It causes scheduling issues when removing large files (17TB in the
2070 Frequency : only with liblustre clients on XT3
2072 Description: evictions taking too long
2073 Details : allow llrd to evict clients directly on OSTs
2076 Frequency : only on ppc
2078 Description: /proc/fs/lustre/devices broken on ppc
2079 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2080 We should make sure the type of 'index' is loff_t and then cast
2081 explicitly as needed below. Do not assign an explicitly cast
2085 Frequency : only for rhel5
2087 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2088 Details : Modify the target file & which_kernel.
2090 Severity : enhancement
2092 Description: omit set fsid for export NFS
2093 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2094 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2098 Frequency : Always, for kernels after 2.6.16
2099 Description: Fix warning idr_remove called for id=.. which is not allocated.
2100 Details : Last kernels save old s_dev before kill super and not allow
2101 to restore from callback - restore it before call kill_anon_super.
2105 Description: buffer overruns could theoretically occur
2106 Details : llapi_semantic_traverse() modifies the "path" argument by
2107 appending values to the end of the origin string, and a
2108 overrun may occur. Adding buffer overrun check in liblustreapi.
2112 Description: Fix error on 'ls .' at the top of the Lustre mount.
2113 Details : Don't revalidate dentry if it is a root dentry.
2117 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2118 Details : Modify vm-tunables-rhel4.patch.
2122 Description: Kernel config for 2.6.18-vanilla.
2123 Details : Modify targets/2.6-vanilla.target.in.
2124 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2125 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2126 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2127 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2131 Description: improve handling recoverable errors
2132 Details : if request processig with error which can be recoverable on server
2133 request should be resend, otherwise page released from cache and
2138 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2139 Details : Modify target file & which_patch.
2140 A flaw was found in the IA32 system call emulation provided
2141 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2142 value could be stored in the %RAX register, which could trigger an
2143 out-of-bounds system call table access. An untrusted local user
2144 could exploit this flaw to run code in the kernel
2145 (ie a root privilege escalation). (CVE-2007-4573).
2149 Description: O_DIRECT bypasses client statistics.
2150 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2151 read_ahead_stats were updated. copied the stats section from
2152 osc_send_oap_rpc() into async_internal().
2156 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2160 Description: change order of libsysio includes
2161 Details : '#include sysio.h' should always come before '#include xtio.h'
2165 Description: handle "serious error: objid * already exists" more gracefully
2166 Details : If LAST_ID value on disk is smaller than the objects existing in
2167 the O/0/d* directories, it indicates disk corruption and causes an
2168 LBUG(). If the object is 0-length, then we should use the existing
2169 object. This will help to avoid a full fsck in most cases.
2171 Severity : enhancement
2173 Description: adapt the lustre_config script to support the upgrade case
2174 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2175 server targets to 1.6 in parallel.
2179 Description: To avoid grant space > avaible space when the disk is almost
2180 full. Without this patch you might see the error "grant XXXX >
2181 available" or some LBUG about grant, when the disk is almost
2183 Details : In filter_check_grant, for non_grant cache write, we should
2184 check the left space by if (*left > ungranted + bytes), instead
2185 of (*left > ungranted), because only we are sure the left space
2186 is enough for another "bytes", then the ungrant space should be
2187 increase. In client, we should update cl_avail_grant only there
2188 is OBD_MD_FLGRANT in the reply.
2192 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2193 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2194 described in CVE-2007-4573. This problem could allow untrusted
2195 local users to gain root access.
2198 Frequency : when using O_DIRECT and quotas
2200 Description: Incorrect file ownership on O_DIRECT output files
2201 Details : block usage reported by 'lfs quota' does not take into account
2202 files that have been written with O_DIRECT.
2207 Description: touch file failed when fs is not full
2208 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2209 otherwise we can get ENOSP while fs is not full.
2213 Description: parallel lock callbacks
2214 Details : Instead of sending blocking and completion callbacks as separated
2215 requests, adding them to a set and sending in parallel.
2218 Frequency : only for Cray XT3
2219 Bugzilla : 12829/13455
2220 Description: Changing primary group doesn't change the group lustre assigns to
2222 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2223 overridden with the primary group provided by the group upcall,
2224 whereas the supplied fsgid can be trusted if it is in the list of
2225 supplementary groups returned by the group upcall.
2227 Severity : enhancement
2229 Description: Allow masking D_WARNING, D_ERROR messages from console
2230 Details : Console messages can now be disabled via lnet.printk.
2234 Description: User code with malformed file open parameter crashes client node
2235 Details : Before packing join_file req, all the related reference should be
2236 checked carefully in case some malformed flags cause fake
2237 join_file req on client.
2241 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2242 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2243 mode is equal to the granted mode. However, it should also check
2244 that LDLM_FL_CP_REQD is not set.
2248 Description: Heavy nfs access might result in deadlocks
2249 Details : After ELC code landed, it is now improper to enqueue any mds
2250 locks under och_sem, because enqueue might want to decide to
2251 cancel open locks for same inode we are holding och_sem for.
2255 Description: Client eviction while running blogbench
2256 Details : A lot of unlink operations with concurrent I/O can lead to a
2257 deadlock causing evictions. To address the problem, the number of
2258 oustanding OST_DESTROY requests is now throttled to
2259 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2260 callbacks are processed in priority.
2264 Description: enable ACLs on MDS by default
2265 Details : ACLs must be enabled on MDS by default.
2268 Frequency : PPC/PPC64 only
2270 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2271 Details : fix duplicated definitions between asm-ppc64/types.h and
2272 lustre_types.h on PPC.
2275 Frequency : PPC/PPC64 only
2277 Description: asm-ppc/segment.h does not exist
2278 Details : fix compile issue on PPC.
2282 Description: better handle error messages in extents code
2285 Frequency : RHEL4 only
2287 Description: mkfs is very slow on IA64/RHEL4
2288 Details : A performance regression has been discovered in the MPT Fusion
2289 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2290 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2291 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2293 Severity : enhancement
2295 Description: SNMP support enhancement
2296 Details : Adding total number of sampled request for an MDS node in snmp
2299 Severity : enhancement
2301 Description: Optimize ldlm waiting list processing for PR extent locks
2302 Details : When processing waiting list for read extent lock and meeting read
2303 lock that is same or wider to it that is not contended, skip
2304 processing rest of the list and immediatelly return current
2305 status of conflictness, since we are guaranteed there are no
2306 conflicting locks in the rest of the list.
2310 Description: Time out and refuse to reconnect
2311 Details : When the failover node is the primary node, it is possible
2312 to have two identical connections in imp_conn_list. We must
2313 compare not conn's pointers but NIDs, otherwise we can defeat
2314 connection throttling.
2318 Description: port llog fixes from b1_6 into HEAD
2319 Details : Port llog reference couting and some llog cleanups from b1_6
2320 (bug 10800) into HEAD, for protect from panic and access to already
2321 free llog structures.
2325 Description: Detect stride IO mode in read-ahead
2326 Details : When a client does stride read, read-ahead should detect that and
2327 read-ahead pages according to the detected stride pattern.
2331 Description: data checksumming impacts single node performance
2332 Details : add support for several checksum algorithm. Currently, only CRC32
2333 and Adler-32 are supported. The checksum type can be changed on
2334 the fly via /proc/fs/lustre/osc/*/checksum_type.
2338 Description: use adler32 for page checksums
2339 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2344 Description: build for x2 fails
2345 Details : fix compile issue on Cray systems.
2349 Description: Properly match for duplicate locks
2350 Details : Due to different lock order from skiplists code, we need to
2351 traverse entire list for now
2354 Frequency : only on PPC/SLES10
2356 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2357 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2358 be defined. Add a hack in mkfs_lustre.c to work around this compile
2363 Description: LASSERT on MDS when client holding flock lock dies
2364 Details : ldlm pool logic depends on number of granted locks equal to
2365 number of released locks which is not true for flock locks, so
2366 just exclude such locks from consideration.
2370 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2371 Details : Use fsfilt_sync() for both the whole filesystem sync and
2372 individual file sync to eliminate dangerous inode locking
2373 with I_LOCK that can lead to a deadlock.
2377 Description: performance in 1.6.3
2378 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2379 sure that 1MB requests are not fragmented by the block layer.
2381 Severity : enhancement
2383 Description: organize the server-side client stats on per-nid basis
2384 Details : Change the structure of stats under obdfilter and mds to
2392 The "uuid"s file would list the uuids of _active_ exports.
2393 And the clear entry is to clear all stats and stale nids.
2395 Severity : enhancement
2397 Description: eliminate client locks in face of contention
2398 Details : file contention detection and lockless i/o implementation
2399 for contended files.
2403 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2406 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2408 Description: MDS may lose file striping (and hence file data) in some cases
2409 Details : If there are additional extended attributes stored on the MDS,
2410 in particular ACLs, SELinux, or user attributes (if user_xattr
2411 is specified for the client mount options) then there is a risk
2412 of attribute loss. Additionally, the Lustre file striping
2413 needs to be larger than default (e.g. striped over all OSTs),
2414 and an additional attribute must be stored initially in the
2415 inode and then increase in size enough to be moved to the
2416 external attribute block (e.g. ACL growing in size) for file
2421 Description: skiplist implementation simplification
2422 Details : skiplists are used to group compatible locks on granted list
2423 that was implemented as tracking first and last lock of each
2424 lock group the patch changes that to using doubly linked lists
2428 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2429 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2430 because LASSERT on both the data supplied by a client, and the
2431 data on disk is dangerous and incorrect.
2433 Severity : enhancement
2435 Description: Slow truncate/writes to huge files at high offsets.
2436 Details : Directly associate cached pages to lock that protect those pages,
2437 this allows us to quickly find what pages to write and remove
2438 once lock callback is received.
2442 Description: more ldlm soft lockups
2443 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2444 starve other threads from the resource lock for a long time in
2445 case of long waiting queue, so change the debug level from
2446 D_OTHER to the less frequently used D_INFO.
2448 Severity : enhancement
2450 Description: add -gid, -group, -uid, -user options to lfs find
2454 Description: Hung threads in invalidate_inode_pages2_range
2455 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2456 once one is completed. As a result, some RPCs are stuck in the
2457 queue and are never sent.
2461 Description: filter threads hungs on waiting journal commit
2462 Details : Cleanup filter group llog code, then only filter group llog will
2463 be only created in the MDS/OST syncing process.
2467 Description: Procfs and llog threads access destoryed import sometimes.
2468 Details : Sync the import destoryed process with procfs and llog threads by
2469 the import refcount and semaphore.
2471 Severity : enhancement
2473 Description: openlock cache of b1_6 port to HEAD
2478 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2479 Details : If insertion of an extent fails, then discard the inode
2480 preallocation and free data blocks else it can lead to duplicate
2485 Description: don't always update ctime in ext3_xattr_set_handle()
2486 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2487 In some cases the ctime should not be updated, for example for
2488 2.0->1.8 compatibility it is necessary to delete an xattr and it
2489 should not update the ctime.
2493 Bugzilla : 15713/16362
2494 Description: Assertion in iopen_connect_dentry in 1.6.3
2495 Details : looking up an inode via iopen with the wrong generation number can
2496 populate the dcache with a disconneced dentry while the inode
2497 number is in the process of being reallocated. This causes an
2498 assertion failure in iopen since the inode's dentry list contains
2499 both a connected and disconnected dentry.
2503 Description: assertion failure in ldlm_handle2lock()
2504 Details : fix a race between class_handle_unhash() and class_handle2object()
2505 introduced in lustre 1.6.5 by bug 13622.
2510 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2511 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2516 Description: oops in sd_iostats_seq_show()
2517 Details : unloading/reloading the scsi low level driver triggers a kernel
2518 bug when trying to access the sd iostat file.
2523 Description: Kernel panics during QLogic driver reload
2524 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2525 patch, causing memory corruption.
2530 Description: journal_dev option does not work in b1_6
2531 Details : pass mount option during pre-mount.
2533 Severity : enhancement
2535 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2536 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2537 extent information of a file. It can be used to map logical blocks
2538 in a file to physical blocks in the block device.
2542 Description: LDLM soft lockups - improvement
2543 Details : It is be possible to send the lock handle along with each read
2544 or write request because the client is already doing a lock match
2545 itself so there isn't any reason the OST should have to re-do that
2549 Frequency : only on Cray X2
2551 Description: X2 build failures
2552 Details : fix build failures on Cray X2.
2556 Description: xid & resent requests
2557 Details : Initialize RPC XID from clock at startup (randomly if clock is
2560 Severity : enhancement
2562 Description: Add lustre_start utility to start or stop multiple Lustre servers
2567 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2568 Details : In case of memory pressure, list_del() can be called twice on
2569 req->rq_history_list, causing a kernel oops.
2573 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2574 Details : fix stack overflow in the distributed lock manager by defering
2575 export eviction after a failed AST to the elt thread instead of
2576 handling it in the dlm interpret routine.
2580 Description: Convert some comments to new format.
2581 Details : Update documenting comments to match doxygen conventions.
2585 Description: Grammar fixes.
2586 Details : A couple of trivial spelling fixes.
2590 Description: OSD_COUNTERS-mandatory
2591 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2592 to simplify the code.
2596 Description: simplify cmm_device freeing logic.
2597 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2601 Description: Add lockdep support to dt_object_operations locking interface.
2602 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2603 parameter indicating lock ordering. Update mdd code to use new
2608 Description: Introduce failloc constants for lockless IO tests.
2609 Details : Add two new failloc constants to test lockless IO. Only one of
2610 them in implemented---another is checked in yet to be landed
2615 Description: Add lockdep support for inode mutex.
2616 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2617 in the situations where Lustre has to lock more than one inode
2622 Description: Add optional invariants checking support.
2623 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2624 configure switch. This macro is to be used for consistency and
2625 sanity checks that are too expensive to be left in `production'
2630 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2631 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2636 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2637 Details : Use spin_lock_nested() in (the only) situation where more than
2638 one ldlm_lock is locked simultaneously. Also, fix possible
2639 dead-lock in ldlm_lock_change_resource() by enforcing particular
2644 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2645 Details : Instead of specifying each ldlm_lock call-back through separate
2646 parameter, wrap them into struct ldlm_callback_suite.
2650 Description: Kill join_lru obd method and its callers.
2651 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2652 in memory---a requirement implemented through ->o_join_lru() obd
2653 method in HEAD. Remove this method and its users.
2657 Description: Add asynchronous ldlm ENQUEUE completion handler.
2658 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2659 case that stock ldlm_completion_ast() cannot handle as it waits
2660 until lock is granted. Introduce new ldlm_completion_ast_async()
2661 for this. Also comment ldlm_completion_ast().
2665 Description: ldlm_error <-> errno conversion.
2666 Details : Add functions to map (rather arbitrary) between LDLM error codes
2667 and standard errno values. CLIO needs this to prevent LDLM specific
2668 constants from escaping ldlm and osc.
2672 Description: Kill unused ldlm_handle2lock_ns() function.
2673 Details : Kill unused ldlm_handle2lock_ns() function.
2677 Description: Add lu_ref support to ldlm_lock
2678 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2679 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2680 and ldlm_resource. LDLM interface has to be changed, because code that
2681 releases a reference on a lock, has to "know" what reference this is.
2682 In the most frequent case
2684 lock = ldlm_handle2lock(handle);
2686 LDLM_LOCK_PUT(lock);
2688 no changes are required. When any other reference (received _not_ from
2689 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2690 instead of LDLM_LOCK_PUT().
2692 Arguably, changes are pervasive, and interface requires some discipline
2693 for proper use. On the other hand, it was very instrumental in finding
2694 a few leaked lock references.
2698 Description: Add ldlm_lock_addref_try().
2699 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2700 attempts to addref a lock that might be being canceled
2705 Description: Add ldlm_weigh_callback().
2706 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2707 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2708 hard-coded `number of pages' logic.
2712 Description: Add lockdep annotations to llog code.
2713 Details : Use appropriately tagged _nested() locking calls in the places
2714 where llog takes more than one ->lgh_lock lock.
2718 Description: Add loi_kms_set().
2719 Details : Wrap kms updates into a helper function.
2723 Description: Constify instances of struct lsm_operations.
2724 Details : Constify instances of struct lsm_operations.
2728 Description: lu_conf support.
2729 Details : On a server, a file system object is uniquely identified
2730 by a fid, which is sufficient to locate and load all object
2731 state (inode). On a client, on the other hand, more data are
2732 necessary instantiate an object. Change lu_object_find() and
2733 friends to take additional `lu_conf' argument describing object.
2734 Typically this includes layout information.
2738 Description: lu_context fixes.
2739 Details : Introduce new lu_context functions that are needed on the client
2740 side, where some system threads (ptlrpcd) are shared by multiple
2741 modules, and so cannot be stopped during module shutdown.
2745 Description: Add start and stop methods to lu_device_type_operations.
2746 Details : Introduce two new methods in lu_device_type_operations, that are
2747 invoked when first instance of a given type is created and last one
2748 is destroyed respectively. This is need by CLIO.
2752 Description: Add lu_ref support to struct lu_device.
2753 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2754 track leaked references.
2758 Description: Introduce lu_kmem_descr.
2759 Details : lu_kmem_descr and its companion interface allow to create
2760 and destroy a number of kmem caches at once.
2764 Description: Fix lu_object finalization race.
2765 Details : Fix a race between lu_object_find() finding an object and its
2766 concurrent finalization. This race is (most likely) not possible
2767 on the server, but might happen on the client.
2771 Description: Introduce lu_ref interface.
2772 Details : lu_ref is a debugging module allowing to track references to
2773 a given object. It is quite cpu expensive, and has to be
2774 explicitly enabled with --enable-lu_ref. See usage description
2779 Description: Factor lu_site procfs stats into a separate function.
2780 Details : Separate lu_site stats printing code into a separate function
2781 that can be reused on a client.
2785 Description: Constify instances of struct {lu,dt,md}_device_operations.
2786 Details : Constify instances of struct {lu,dt,md}_device_operations.
2790 Description: Introduce struct md_site and move meta-data specific parts of
2791 struct lu_site here.
2792 Details : Move md-specific fields out of struct lu_site into special struct
2793 md_site, so that lu_site can be used on a client.
2797 Description: Kill mdd_lov_destroy().
2798 Details : Remove unused mdd code.
2802 Description: Add st_block checking to multistat.c.
2803 Details : Add st_block checking to multistat.c.
2807 Description: Add lu_ref support to struct obd_device.
2808 Details : Add lu_ref tracking to obd_device.
2812 Description: Kill obd_set_fail_loc().
2813 Details : Remove unused code.
2817 Description: Add special type for ptlrpc_request interpret functions.
2818 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2819 there. Actual usage will be in CLIO.
2823 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2824 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2825 doesn't work with lockdep.
2829 Description: Add rwv.c test program.
2830 Details : New testing program exercising readv(2) and writev(2) (Qian).
2834 Description: Add sendfile.c test program.
2835 Details : New testing program exercising sendfile(2) (Jay).
2839 Description: Ratelimit a message that can be very frequent.
2840 Details : Ratelimit a memory allocation failure message that can
2845 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2846 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2850 Description: Liblustre build fixes.
2851 Details : Liblustre build fixes.
2855 Description: libcfs: add cfs_{need,cond}_resched() interface.
2856 Details : libcfs: add cfs_{need,cond}_resched() definition and
2857 implementations for Linux, NT, and liblustre.
2859 Severity : enhancement
2861 Description: More exported tunables for mballoc
2862 Details : Add support for tunable preallocation window and new tunables for
2863 large/small requests
2867 Description: Detect corruption of block bitmap and checking for preallocations
2868 Details : Checks validity of on-disk block bitmap. Also it does better
2869 checking of number of applied preallocations. When corruption is
2870 found, it turns filesystem readonly to prevent further corruptions.
2874 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2875 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2876 window to 0. If it does hit cache miss, and read-pattern is still
2877 stride-io mode, does not reset the stride window, but also does not
2878 increase the stride window length in this case.
2882 Frequency : only for big-endian servers
2883 Description: Check if system is big-endian while mounting fs with extents feature
2884 Details : Mounting a filesystem with extents feature will fail on big-endian
2885 systems since ext3-based ldiskfs is not supported on big-endian
2886 systems. This can be over-riden with "bigendian_extents" mount option.
2888 Severity : enhancement
2890 Description: The root squash functionality
2891 Details : A security feature, which is to prevent users from being able
2892 to mount lustre on their desktop, run as root, and delete
2893 all of the files in the filesystem. The goal is accomplished by
2894 remapping user id (UID) and group id (GID) of the root user to
2895 a UID and GID specified by the system administartor via Lustre
2896 configuration management server (MGS). The functionality also
2897 allows to specify sets of clients for which the remapping does
2902 Description: Excessive recovery window
2903 Details : With AT enabled, the recovery window can be excessively long (6000+
2904 seconds). To address this problem, we no longer use
2905 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2906 timeout no longer depends on the service time, it is set to
2907 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2908 time via pb_service_time.
2912 Description: Watchdog triggered on MDS failover
2913 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2914 the OSTs know that the MDS "UUID" can be reused for the same export
2915 from a different NID, so we do not need to wait for the export to be
2919 Frequency : rare, only if using MMP with Linux RAID
2921 Description: MMP doesn't work with Linux RAID
2922 Details : While using HA for Lustre servers with Linux RAID, it is possible
2923 that MMP will not detect multiple mounts. To make this work we
2924 need to unplug the device queue in RAID when the MMP block is being
2925 written. Also while reading the MMP block, we should read it from
2926 disk and not the cached one.
2928 Severity : enhancement
2930 Description: open file using fid
2931 Details : A file can be opened using just its fid, like
2932 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2935 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2938 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2939 Details : add proper locking for creation of jbd_4k slab cache
2943 Description: MMP check in ext3_remount() fails without displaying any error
2944 Details : When multiple mount protection fails during remount, proper error
2947 Severity : enhancement
2949 Description: Allow stripe size to be up to 4G-64k
2950 Details : Fix math logic to allow large stripe sizes.
2954 Description: add check for >8TB ldiskfs filesystems
2955 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2956 Don't allow >8TB ldiskfs filesystems to be mounted without
2957 force_over_8tb mount option
2962 Descriptoin: OSS complains "too many blocks" and crashes
2963 Details : In filter_commitrw_write():
2964 1. lquota_chkquota();
2965 2. obdo_from_inode();
2966 3. lquota_pending_commit();
2967 "2." has changed uid and gid in this case, then caused
2968 lquota_chkquota() and lquota_pending_commit() to use
2969 different uid and gid. That is the root of the bug.
2973 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2974 Details : Detect and print out the 'stripe count = -1' case properly.
2978 Description: Out or order replies might be lost on replay
2979 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2980 request with tid smaller than one currently being added, add it
2981 to the start, not end of the list.
2983 --------------------------------------------------------------------------------
2985 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2987 * Support for kernels:
2988 2.6.5-7.283 (SLES 9),
2989 2.6.9-55.EL (RHEL 4),
2990 2.6.16.46-0.14 (SLES 10),
2991 2.6.18.8 vanilla (kernel.org)
2992 * Client support for unpatched kernels:
2993 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2994 2.6.16 - 2.6.22 vanilla (kernel.org)
2995 * Due to recently discovered recovery problems, we do not recommend
2996 using patchless RHEL 4 clients with this or any earlier release.
2997 * Recommended e2fsprogs version: 1.39.cfs8
2998 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2999 * Starting with this release, the ldiskfs backing filesystem required
3000 by Lustre is now in its own package, lustre-ldiskfs. This package
3001 should be installed. It is versioned separately from Lustre and
3002 may be released separately in future.
3006 Description: block reactivating mgc import until all deactivates complete
3007 Details : Fix race when failing back MDT/MGS to itself (testing)
3010 Frequency : at statup only
3012 Description: mds_lov_synchronize race leads to various problems
3013 Details : simultaneous MDT->OST connections at startup can cause the
3014 sync to abort, leaving the OSC in a bad state.
3016 Severity : enhancement
3018 Description: add optional extra BUILD_VERSION info
3019 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3020 to override the lustre version.
3023 Frequency : 2.6.18 servers only
3025 Description: ll_kern_mount() doesn't release the module reference
3026 Details : The ldiskfs module reference count never drops down to 0
3027 because ll_kern_mount() doesn't release the module reference.
3032 Description: server LBUG when using old ost_num_threads parameter
3033 Details : Accept the old ost_num_threads parameter but warn that it
3034 is deprecated, and fix an off-by-one error that caused an LBUG.
3039 Description: Transient SCSI error results in persistent IO issue
3040 Details : iobuf->dr_error is not reinitialized to 0 between two
3044 Frequency : sometimes when underlying device returns I/O errors
3046 Description: OSTs not going read-only during write failures
3047 Details : OSTs are not remounted read-only when the journal commit threads
3048 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3049 instead of the ext3 wrappers.
3053 Description: poor connect scaling with increasing client count
3054 Details : Don't run filter_grant_sanity_check for more than 100 exports
3055 to improve scaling for large numbers of clients.
3058 Frequency : SLES10 only
3060 Description: sanity-quota.sh quotacheck failed: rc = -22
3061 Details : Quotas cannot be enabled on SLES10.
3064 Frequency : liblustre clients only
3066 Description: getdirentries does not give error when run on compute nodes
3067 Details : getdirentries does not fail when the size specified as an argument
3068 is too small to contain at least one entry
3070 Severity : enhancement
3072 Description: Add LNET router traceability for debug purposes
3073 Details : If a checksum failure occurs with a router as part of the
3074 IO path, the NID of the last router that forwarded the bulk data
3075 is printed so it can be identified.
3080 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3081 Details : Due to a race condition, liblustre clients were occasionally
3082 evicted incorrectly.
3084 Severity : enhancement
3086 Description: lfs setstripe use optional parameters instead of postional
3089 Severity : enhancement
3091 Description: Nanosecond timestamp support for ldiskfs
3092 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3093 resolution timestamps. There is not yet support for this at
3094 the Lustre filesystem level.
3097 Frequency : during server recovery
3099 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3100 Details : request with rq_no_resend flag not awake l_wait_event if they get
3104 Frequency : nfs export on patchless client
3106 Description: connectathon hang when test nfs export over patchless client
3107 Details : Disconnected dentry cannot be found with lookup, so we do not need
3108 to unhash it or make it invalid
3111 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3112 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3113 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3118 Description: open req refcounting wrong on reconnect
3119 Details : If reconnect happened between getting open reply from server and
3120 call to mdc_set_replay_data in ll_file_open, we will schedule
3121 replay for unreferenced request that we are about to free.
3122 Subsequent close will crash in variety of ways.
3123 Check that request is still eligible for replay in
3124 mdc_set_replay_data().
3129 Description: disable writes to filesystem when reading health_check file
3130 Details : the default for reading the health_check proc file has changed
3131 to NOT do a journal transaction and write to disk, because this
3132 can cause reads of the /proc file to hang and block HA state
3133 checking on a healthy but otherwise heavily loaded system. It
3134 is possible to return to the previous behaviour during configure
3135 with --enable-health-write.
3137 Severity : enhancement
3139 Description: 64-bit inode version
3140 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3141 to the inode. This will be required for version-based recovery.
3146 Description: MDS fails to start if a duplicate client export is detected
3147 Details : in some rare cases it was possible for a client to connect to
3148 an MDS multiple times. Upon recovery the MDS would detect this
3149 and fail during startup. Handle this more gracefully.
3151 Severity : enhancement
3153 Description: Add -o localflock option to simulate old noflock
3155 Details : This will achieve local-only flock/fcntl locks
3161 Description: log_commit_thread vs filter_destroy race leads to crash
3162 Details : Take import reference before releasing llog record semaphore
3167 Description: Wrong request locking in request set processing
3168 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3169 imp_delayed_list, in this place should be used imp_lock.
3172 Frequency : when reconnection
3174 Description: Grant Leak when osc reconnect to OST
3175 Details : When osc reconnect ost, OST(filter) should check whether it
3176 should grant more space to client by comparing fed_grant and
3177 cl_avail_grant, and return the granted space to client instead
3178 of "new granted" space, because client will call osc_init_grant
3179 to update the client grant space info.
3182 Frequency : when client reconnect to OST
3184 Description: Grant Leak when osc do resend and replay bulk write
3185 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3186 bulk write request, because the grant info while be sync between
3187 OSC and OST when reconnect, and we should ignore the grant info
3188 these of resend/replay write req.
3193 Description: Grant space more than available left space sometimes.
3194 Details : When then OST is about to be full, if two bulk writing from
3195 different clients came to OST. Accord the avaliable space of the
3196 OST, the first req should be permitted, and the second one
3197 should be denied by ENOSPC. But if the seconde arrived before
3198 the first one is committed. The OST might wrongly permit second
3199 writing, which will cause grant space > available space.
3202 Frequency : when client is evicted
3204 Description: Grant might be wrongly erased when osc is evicted by OST
3205 Details : when the import is evicted by server, it will fork another
3206 thread ptlrpc_invalidate_import_thread to invalidate the
3207 import, where the grant will be set to 0. While the original
3208 thread will update the grant it got when connecting. So if
3209 the former happened latter, the grant will be wrongly errased
3210 because of this race.
3215 Description: Checking Stale with correct fid
3216 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3217 to check whether it is stale, because sometimes, we want the
3218 enqueue happened anyway, and op_data.fid2 will not be initialized.
3220 Severity : enhancement
3222 Description: update patchless client
3223 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3226 Frequency : only with 2.4 kernel
3228 Description: random memory corruption
3229 Details : size of struct ll_inode_info is to big for union inode.u and this
3230 can be cause of random memory corruption.
3235 Description: Memory leak in recovery
3236 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3237 It should also check obd_fail before fsfilt_start, otherwise if
3238 fsfilt_start return -EROFS,(failover mds during mds recovery).
3239 then the req will return with repmsg->transno = 0 and rc = EROFS.
3240 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3241 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3242 be freed no matter whether fsfilt_commit success or not.
3245 Frequency : only with huge count clients
3247 Description: Prevents from taking the superblock lock in llap_from_page for
3249 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3250 for llap_from_page call in ll_removepage prevents from taking the
3251 superblock lock for a soon died page.
3256 Description: Not check open intent error before release open handle
3257 Details : in some rare cases, the open intent error is not checked before
3258 release open handle, which may cause
3259 ASSERTION(open_req->rq_transno != 0), because it tries to release
3260 the failed open handle.
3265 Description: Set cat log bitmap only after create log success.
3266 Details : in some rare cases, the cat log bitmap is set too early. and it
3267 should be set only after create log success.
3271 Description: Accessing a block bevice can re-enable I/O when Lustre is
3272 tearing down a device.
3273 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3278 Description: service threads may hog cpus when there are a lot of requests
3279 Details : Insert cond_resched to give other threads a chance to use some CPU
3284 Description: the cat log was not initialized in recovery
3285 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3286 unlink log on mds will not be initialized until mds post
3287 recovery. And also in mds post recovery, the unlink log will
3288 initialization will be done asynchronausly, so there will be race
3289 between add unlink log and unlink log initialization.
3293 Description: brw_stats were being printed incorrectly
3294 Details : brw_stats were being printed as log2 but all of them were not
3295 recorded as log2. Also remove some code duplication arising from
3296 filter_tally_{read,write}.
3300 Frequency : rare, only in recovery.
3301 Description: ASSERTION(req->rq_type != LI_POISON) failed
3302 Details : imp_lock should be held while iterating over imp_sending_list for
3303 prevent destroy request after get timeout in ptlrpc_queue_wait.
3307 Description: replay-single.sh test 52 fails
3308 Details : A lock's skiplist need to be cleanup when it being unlinked
3309 from its resource list.
3313 Description: Short directio read returns full requested size rather than
3315 Details : Direct I/O operations should return actual amount of bytes
3316 transferred rather than requested size.
3318 Severity : enhancement
3320 Description: metadata RPC reduction (e.g. for rm performance)
3321 Details : decrease the amount of synchronous RPC between clients and servers
3322 by canceling conflicing lock before the operation on the client
3323 and packing thier handles into the main operation RPC to server.
3325 Severity : enhancement
3327 Description: Async OSC create to avoid the blocking unnecessarily.
3328 Details : If an OST has no remaining object, system will block on the
3329 creation when it needs to create a new object on this OST. Now,
3330 ways use pre-created objects when available, instead of blocking on
3331 an empty osc while others are not empty. If we must block, we
3332 block for the shortest possible period of time.
3336 Description: improve handling recoverable errors
3337 Details : if request processig with error which can be recoverable on server
3338 request should be resend, otherwise page released from cache and
3341 Severity : enhancement
3343 Description: refine locking for avoid write wrong info into lov_objid file
3344 Details : fix possible races with add new target and write/update data in
3347 --------------------------------------------------------------------------------
3349 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3354 Frequency : on some architectures
3356 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3357 Details : Uninitialized flags sometimes cause configuration commands to
3361 Frequency : patchless clients only
3363 Description: missing __iget() symbol export
3364 Details : The __iget() symbol export is missing. To avoid the need for
3365 this on patchless clients the deathrow inode reaper is turned
3366 off, and we depend on the VM to clean up old inodes. This
3367 dependency was during via the fix for bug 12181.
3369 --------------------------------------------------------------------------------
3371 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3373 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3374 INTEROPERATE with older versions automatically. In many cases a
3375 special upgrade step is needed. Please read the
3376 user documentation before upgrading any part of a live system.
3377 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3378 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3379 * WARNING: Lustre configuration and startup changes are required with
3380 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3384 Severity : enhancement
3386 Description: MountConf
3387 Details : Lustre configuration is now managed via mkfs and mount
3388 commands instead of lmc and lconf. New obd types (MGS, MGC)
3389 are added for dynamic configuration management. See
3390 https://mail.clusterfs.com/wikis/lustre/MountConf for
3393 Severity : enhancement
3395 Description: dynamic OST addition
3396 Details : OSTs can now be added to a live filesystem
3398 Severity : enhancement
3400 Description: startup order invariance
3401 Details : MDTs and OSTs can be started in any order. Clients only
3402 require the MDT to complete startup.
3404 Severity : enhancement
3406 Description: parallel, asynchronous orphan cleanup
3407 Details : orphan cleanup is now performed in separate threads for each
3408 OST, allowing parallel non-blocking operation.
3410 Severity : enhancement
3412 Description: optimized stripe assignment
3413 Details : stripe assignments are now made based on ost space available,
3414 ost previous usage, and OSS previous usage, in order to try
3415 to optimize storage space and networking resources.
3417 Severity : enhancement
3419 Description: Permanently set tunables
3420 Details : All writable /proc/fs/lustre tunables can now be permanently
3421 set on a per-server basis, at mkfs time or on a live system.
3423 Severity : enhancement
3425 Description: Lustre message v2
3426 Details : Add lustre message format v2.
3428 Severity : enhancement
3430 Description: client OST exclusion list
3431 Details : Clients can be started with a list of OSTs that should be
3432 declared "inactive" for known non-responsive OSTs.
3436 Description: SPEC SFS validation failure on NFS v2 over lustre.
3437 Details : Changes the blocksize for regular files to be 2x RPC size,
3438 and not depend on stripe size.
3440 Severity : enhancement
3442 Description: Multiple MD RPCs in flight.
3443 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3444 To avoid overly-overloading MDS, introduce a limit on number of
3445 MDS RPCs in flight for a single client and add /proc controls
3446 to adjust this limit.
3448 Severity : enhancement
3450 Description: client read/write statistics
3451 Details : Add client read/write call usage stats for performance
3452 analysis of user processes.
3453 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3454 file access. extents_stats shows chunk size distribution.
3455 extents_stats_per_process show chunk size distribution per
3458 Severity : enhancement
3460 Description: mds statistics
3461 Details : Add detailed mds operations statistics in
3462 /proc/fs/lustre/mds/*/stats.
3466 Description: Failure of copying files with lustre special EAs.
3467 Details : Client side always return success for setxattr call for lustre
3468 special xattr (currently only "trusted.lov").
3473 Description: Refcount LNET uuids
3474 Details : The global LNET uuid list grew linearly with every startup;
3475 refcount repeated list entries instead of always adding to
3478 Severity : enhancement
3480 Description: Dynamic service threads
3481 Details : Within a small range, start extra service threads
3482 automatically when the request queue builds up.
3485 Frequency : mixed-endian client/server environments
3487 Description: mixed-endian crashes
3488 Details : The new msg_v2 system had some failures in mixed-endian
3491 Severity : enhancement
3493 Description: Easy OST removal
3494 Details : OSTs can be permanently deactivated with e.g. 'lctl
3495 conf_param lustre-OST0001.osc.active=0'
3497 Severity : enhancement
3499 Description: MGS proc entries
3500 Details : Added basic proc entries for the MGS showing what filesystems
3503 Severity : enhancement
3505 Description: provide MGS failover
3506 Details : Added config lock reacquisition after MGS server failover.
3508 Severity : enhancement
3510 Description: add Linux 2.4 support
3511 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3515 Description: a large application tries to do I/O to the same resource and dies
3516 in the middle of it.
3517 Details : Check the req->rq_arrival time after the call to
3518 ost_brw_lock_get(), but before we do anything about
3519 processing it & sending the BULK transfer request. This
3520 should help move old stale pending locks off the queue as
3521 quickly as obd_timeout.
3524 Frequency : when an incorrect nid is specified during startup
3526 Description: ptlrpc connect to non-existant node causes kernel crash
3527 Details : LNET can't be re-entered from an event callback, which
3528 happened when we expire a message after the export has been
3529 cleaned up. Instead, hand the zombie cleanup off to another
3532 Severity : enhancement
3534 Description: plain/inodebits lock performance improvement
3535 Details : Grouping plain/inodebits in granted list by their request modes
3536 and bits policy, thus improving the performance of search through
3540 Frequency : only if OST filesystem is corrupted
3542 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3543 Details : for a short time RPCs with bulk IO are in the replay list,
3544 but replay of bulk IOs is unimplemented. If the OST filesystem
3545 is corrupted due to disk cache incoherency and then replay is
3546 started it is possible to trip an assertion. Avoid putting
3547 committed RPCs into the replay list at all to avoid this issue.
3550 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3552 Description: System hang on startup
3553 Details : This bug allowed the liblustre (e.g. catamount) client to
3554 return to the app before handling all startup RPCs. This
3555 could leave the node unresponsive to lustre network traffic
3556 and manifested as a server ptllnd timeout.
3558 Severity : enhancement
3560 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3561 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3562 boolean to control whether to print peer debug info when a
3563 client's RPC times out.
3566 Frequency : only for kernels with patches from Lustre below 1.4.3
3568 Description: Remove old rdonly API
3569 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3572 Frequency : only for devices with external journals
3574 Description: Set external device read-only also
3575 Details : During a commanded failover stop, we set the disk device
3576 read-only while the server shuts down. We now also set any
3577 external journal device read-only at the same time.
3580 Frequency : when upgrading from 1.4 while trying to change parameters
3582 Description: The wrong (new) MDC name was used when setting parameters for
3583 upgraded MDT's. Also allows changing of OSC (and MDC)
3584 parameters if --writeconf is specified at tunefs upgrade time.
3587 Frequency : when setting specific ost indicies
3589 Description: QOS code breaks on skipped indicies
3590 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3591 created with --index need not be sequential.
3593 Severity : enhancement
3595 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3596 Details : The uninit_groups feature works in conjunction with the kernel
3597 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3598 up the pass1 processing of e2fsck. This is a read-only feature
3599 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3600 mount filesystems that have had this feature enabled.
3602 Severity : enhancement
3604 Description: Improve multi-block allocation algorithm to avoid fragmentation
3605 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3606 allocation locality and avoid filesystem fragmentation.
3608 ------------------------------------------------------------------------------
3610 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3612 * Support for kernels:
3613 2.4.21-47.0.1.EL (RHEL 3)
3614 2.6.5-7.283 (SLES 9)
3615 2.6.9-42.0.10.EL (RHEL 4)
3616 2.6.12.6 vanilla (kernel.org)
3617 2.6.16.27-0.9 (SLES 10)
3618 * Recommended e2fsprogs version: 1.39.cfs5
3620 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3624 Frequency : occasional, depends on client load and configuration
3625 Bugzilla : 12181, 12203
3626 Description: data loss for recently-modified files
3628 Details : In some cases it is possible that recently written or created
3629 files may not be written to disk in a timely manner (this should
3630 normally be within 30s unless client IO load is very high).
3631 The problem appears as zero-length files or files that are a
3632 multiple of 1MB in size after a client crash or client eviction
3633 that are missing data at the end of the file.
3635 This problem is more likely to be hit on clients where files are
3636 repeatedly created and unlinked in the same directory, clients
3637 have a large amount of RAM, have many CPUs, the filesystem has
3638 many OSTs, the clients are rebooted frequently, and/or the files
3639 are not accessed by other nodes after being written.
3641 The presence of the problem can be detected by looking at
3642 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3643 smaller than the second (nr_unused) then dirty files will not
3644 be flushed automatically to disk. "sync; sleep 10" should be
3645 run several times on the node before unmounting it to update
3646 Lustre (this is also safe to run on nodes without this problem).
3648 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3649 that can cause this same problem, so customers using that kernel
3650 also need to update the kernel in addition to Lustre. In order
3651 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3653 It is normal that files written just before a client crash (less
3654 than 30s) may not yet have been flushed to disk, even for local
3658 Frequency : frequent on thin XT3 nodes
3660 Description: UUID collision on thin XT3 Linux nodes
3661 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3662 randomly, since we relied on an insufficiently-seeded PRNG.
3667 Description: OSS hangs after "All ost request buffers busy"
3668 Details : A deadlock between quota and journal operations caused OSS
3669 hangs after printing "All ost request buffers busy."
3672 Frequency : always on liblustre builds
3674 Description: Cleanup compiler warnings on liblustre
3677 Frequency : always on liblustre builds on XT3
3679 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3683 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3684 > 32000 subdirectories
3685 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3686 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3687 subdirectory count crosses 32000. This will aid e2fsck to
3688 correctly handle more than 32000 subdirectories.
3691 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3693 Description: System hang on startup
3694 Details : This bug allowed the liblustre (e.g. catamount) client to
3695 return to the app before handling all startup RPCs. This
3696 could leave the node unresponsive to lustre network traffic
3697 and manifested as a server ptllnd timeout.
3699 Severity : enhancement
3701 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3702 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3703 boolean to control whether to print peer debug info when a
3704 client's RPC times out.
3709 Description: make O_SYNC working on 2.6 kernels
3710 Details : 2.6 kernels use different method for mark pages for write,
3711 so need add a code to lustre for O_SYNC work.
3716 Description: Failure to close file and release space on NFS
3717 Details : Put inode details into lock acquired in ll_intent_file_open.
3718 Use mdc_intent_lock in ll_intent_open to properly
3719 detect all kind of errors unhandled by mdc_enqueue.
3724 Description: proc file read during shutdown sometimes raced obd removal,
3726 Details : Add lock to prevent obd access after proc file removal.
3729 Frequency : Only for files larger than 4GB on 32-bit clients.
3731 Description: improperly doing page alignment of locks
3732 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3733 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3738 Description: RPCs being resent when they shouldn't be
3739 Details : Some RPCs that should not be resent are being resent. This
3740 can cause inconsistencies in the RPC state machine. Do not
3741 resend such requests.
3744 Frequency : rare, only with NFS export
3746 Description: Crash on NFS re-export node
3747 Details : under very unusual load conditions an assertion is hit in
3748 ll_intent_file_open()
3751 Frequency : only if OST filesystem is corrupted
3753 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3754 Details : for a short time RPCs with bulk IO are in the replay list,
3755 but replay of bulk IOs is unimplemented. If the OST filesystem
3756 is corrupted due to disk cache incoherency and then replay is
3757 started it is possible to trip an assertion. Avoid putting
3758 committed RPCs into the replay list at all to avoid this issue.
3763 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3764 Details : Large single O_DIRECT read and write calls can fail to allocate
3765 a sufficiently large buffer to process the request. In case of
3766 allocation failure the allocation is retried with a smaller
3767 buffer and broken into smaller requests.
3769 Severity : enhancement
3771 Description: Add -o localflock option to simulate old noflock behaviour.
3772 Details : This will achieve local-only flock/fcntl locks coherentness.
3777 Description: versioning check is incomplete
3778 Details : Checking the version difference of client vs. server, report
3779 error if the gap is too big.
3784 Description: add support PG_writeback bit
3785 Details : add support for PG_writeback bit for Lustre, for more carefull
3786 work with page cache in 2.6 kernel. This also fix some deadlocks
3787 and remove hack for work O_SYNC with 2.6 kernel.
3789 Severity : enhancement
3791 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3792 Details : The uninit_groups feature works in conjunction with the kernel
3793 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3794 up the pass1 processing of e2fsck. This is a read-only feature
3795 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3796 mount filesystems that have had this feature enabled.
3798 Severity : enhancement
3800 Description: Improve multi-block allocation algorithm to avoid fragmentation
3801 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3802 allocation locality and avoid filesystem fragmentation.
3804 ------------------------------------------------------------------------------
3806 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3808 * Support for kernels:
3809 2.6.9-42.0.3.EL (RHEL 4)
3810 2.6.5-7.276 (SLES 9)
3811 2.4.21-47.0.1.EL (RHEL 3)
3812 2.6.12.6 vanilla (kernel.org)
3813 2.6.16.21-0.8 (SLES10)
3814 * Recommended e2fsprogs version: 1.39.cfs2-0
3816 * The backwards-compatible /proc/sys/portals symlink has been removed
3817 in this release. Before upgrading, please ensure that you change
3818 any configuration scripts or /etc/sysctl.conf files that access
3819 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3820 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3821 in advance of the upgrade on any system running Lustre 1.4.6 or
3822 newer, since /proc/sys/lnet was added in that version.
3823 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3827 Frequency : only when quota is used
3829 Description: avoid scanning export list for quota master
3830 Details : Change the algorithms to avoid scanning export list in order
3831 to improve the efficiency.
3834 Frequency : MDS failover only, very rarely
3836 Description: "went back in time" messages on mds failover
3837 Details : The greatest transno may be lost when the current operation
3838 finishes with an error (transno==0) and the client's last_rcvd
3839 record is over-written. Save the greatest transno in the
3840 mds_last_transno for this case.
3843 Frequency : always for specific kernels and striping counts
3845 Description: client may get "Matching packet too big" without ACL support
3846 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3847 when trying to access files in certain configurations. The
3848 clients should in fact be denied when mounting because they do
3849 not understand ACLs.
3852 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3854 Description: many clients connecting with IO in progress causes connect timeouts
3855 Details : Avoid synchronous journal commits to avoid delays caused by many
3856 clients connecting/disconnecting when bulk IO is in progress.
3857 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3858 OST_IO_PORTAL to avoid delays behind potentially many pending
3862 Frequency : occasionally with multiple writers to a single file
3864 Description: shared writes to file may result in wrong size reported by stat()
3865 Details : Allow growing of kms when extent lock is cancelled
3868 Frequency : always with random mmap IO to multi-striped file
3870 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3871 Details : Only if the hole is at the end of OST object so that kms is too
3872 small. Fix is to increase kms accordingly in ll_nopage.
3875 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3877 Description: writes to a missing object would leak memory on the OST
3878 Details : If there is an inconsistency between the MDS and OST filesystems,
3879 such that the MDS references an object that doesn't exist, writes
3880 to that object will leak memory due to incorrect cleanup in the
3881 error handling path, eventually running out of memory on the OST.
3886 Description: Creating too long symlink causes lustre errors
3887 Details : Check symlink and name lengths before sending requests to MDS.
3890 Frequency : only if flock is enabled (not on by default)
3892 Description: posix locks not released on fd closure on 2.6.9+
3893 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3894 kernels, this caused such locks not to be released on fd close and
3895 then assertions on fs unmount about still used locks.
3898 Frequency : MDS failover only, very rarely
3900 Description: clients may get ASSERTION(granted_lock != NULL)
3901 Details : When request was taking a long time, and a client was resending
3902 a getattr by name lock request. The were multiple lock requests
3903 with the same client lock handle and
3904 mds_getattr_name->fixup_handle_for_resent_request found one of the
3905 lock handles but later failed with ASSERTION(granted_lock != NULL).
3910 Description: handle->h_buffer_credits > 0, assertion failure
3911 Details : h_buffer_credits is zero after truncate, causing assertion
3912 failure. This patch extends the transaction or creates a new
3916 Frequency : NFS re-export or patchless client
3917 Bugzilla : 11179, 10796
3918 Description: Crash on NFS re-export node (__d_move)
3919 Details : We do not want to hash the dentry if we don't have a lock.
3920 But if this dentry is later used in d_move, we'd hit uninitialised
3921 list head d_hash, so we just do this to init d_hash field but
3922 leave dentry unhashed.
3925 Frequency : NFS re-export or patchless client
3927 Description: NFS exports has problem with symbolic link
3928 Details : lustre client didn't properly install dentry when re-exported
3929 to NFS or running patchless client.
3932 Frequency : NFS re-export or patchless client
3934 Description: Various nfs/patchless fixes.
3935 Details : fixes reuse disconected alias for lookup process - this fixes
3936 warning "find_exported_dentry: npd != pd",
3937 fix permission error with open files at nfs.
3938 fix apply umask when do revalidate.
3941 Frequency : occasional
3943 Description: Crash on NFS re-export node
3944 Details : calling clear_page() on the wrong pointer triggered oops in
3945 generic_mapping_read().
3948 Frequency : rarely, using O_DIRECT IO
3950 Description: unaligned directio crashes client with LASSERT
3951 Details : check for unaligned buffers before trying any requests.
3954 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3956 Description: stale data returned from RAID cache
3957 Details : If only a small amount of IO is done to the RAID device before
3958 reading it again it is possible to get stale data from the RAID
3959 cache instead of reading it from disk.
3962 Frequency : always for sles10 kernel
3964 Description: sles10 support
3965 Details : ll_follow_link: compile fixes and using of nd_set_link
3966 under newer kernels.
3969 Frequency : depends on arch, kernel and compiler version, always on sles10
3972 Description: recursive or deep enough symlinks cause stack overflow
3973 Details : getting rid of large stack-allocated variable in
3977 Frequency : depends on hardware
3979 Description: lustre write performance loss in the SLES10 kernel
3980 Details : the performance loss is caused by using of write barriers in the
3981 ext3 code. The SLES10 kernel turns barrier support on by
3982 default. The fix is to undo that change for ldiskfs.
3984 ------------------------------------------------------------------------------
3986 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3988 * Support for kernels:
3989 2.6.9-42.0.3EL (RHEL 4)
3990 2.6.5-7.276 (SLES 9)
3991 2.4.21-47.0.1.EL (RHEL 3)
3992 2.6.12.6 vanilla (kernel.org)
3996 Frequency : quota enabled and large files being deleted
3998 Description: releasing more than 4GB of quota at once hangs OST
3999 Details : If a user deletes more than 4GB of files on a single OST it
4000 will cause the OST to spin in an infinite loop. Release
4001 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4006 Description: statfs data retrieved from /proc may be stale or zero
4007 Details : When reading per-device statfs data from /proc, in the
4008 {kbytes,files}_{total,free,avail} files, it may appear
4009 as zero or be out of date.
4012 Frequency : systems with MD RAID1 external journal devices
4014 Description: lconf's call to blkid is confused by RAID1 journal devices
4015 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4016 of returning all block devices that match the journal UUID.
4019 Frequency : always, for aggregate stripe size over 4GB
4021 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4022 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4023 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4026 Frequency : always if "lfs find" used on a local file/directory
4028 Description: "lfs find" segfaults if used on a local file/directory
4029 Details : The case where a directory component was not specified wasn't
4030 handled correctly. Handle this properly.
4033 Frequency : always on ppc64
4035 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4036 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4037 perform endianness conversion.
4040 Frequency : rarely (truncate to non-zero file size after write under load)
4041 Bugzilla : 10730, 10687
4042 Description: Files padded with zeros to next 4K multiple
4043 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4044 it is possible that files that are truncated to a non-zero size
4045 immediately after being written are filled with zero bytes beyond
4046 the truncated size. No file data is lost.
4048 Severity : enhancement
4050 Description: Allow recovery/failover for liblustre clients.
4051 Details : liblustre clients were unaware of failover configurations until
4054 Severity : enhancement
4056 Description: user file locks should fail when not mounting with flock option
4057 Details : Set up an error-returning stub in ll_file_operations.lock field
4058 to prevent incorrect behaviour when client is mounted without
4059 flock option. Also, set up properly f_op->flock field for
4063 Frequency : always on ia64
4065 Description: "lfs df" loops on printing out MDS statfs information
4066 Details : The obd_ioctl_data was not initialized and in some systems
4067 this caused a failure during the ioctl that did not return
4068 an error. Initialize the struct and return an error on failure.
4071 Frequency : SLES 9 only
4073 Description: Error of copying files with lustre special EAs as root
4074 Details : Client side always return success for setxattr call for lustre
4075 special xattr (currently only "trusted.lov").
4078 Frequency : rarely on clusters with both ia64+i386 clients
4080 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4081 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4082 clients are doing concurrent writes to the same file it is
4083 possible that the ia64 clients may LASSERT because the OST
4084 extent locks are not PAGE_SIZE aligned. Ensure that grown
4085 locks are always aligned on the request boundary.
4088 Frequency : specific use, occasional
4090 Description: Overwriting in use executable truncates on-disk binary image
4091 Details : If one node attempts to overwrite an executable in use by
4092 another node, we now correctly return ETXTBSY instead of
4093 truncating the file.
4095 Severity : enhancement
4097 Description: Async OSC create to avoid the blocking unnecessarily.
4098 Details : If an OST has no remaining object, system will block on the creating
4099 when need to create a new object on this OST. Now, ways use
4100 pre-created objects when available, instead of blocking on an
4101 empty osc while others are not empty. If we must block, we block
4102 for the shortest possible period of time.
4107 Description: chmod on Lustre root is propagated to other clients
4108 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4109 invalid by the follow_mount time.
4112 Frequency : liblustre clients only
4114 Description: Race in 'instant cancel' lock handling could lead to such locks
4115 never to be granted in case of SMP MDS
4116 Details : Do not destroy not yet granted but cbpending locks in
4120 Frequency : replay/resend of open
4122 Description: non null lock assetion failure in mds_intent_policy
4123 Details : Trying to replay/resend lockless open requests resulted in
4124 mds_open() returning 0 with no lock. Now it sets a flag if
4125 a lock is going to be returned.
4127 Severity : enhancement
4129 Description: Checksum enhancements
4130 Details : New checksum enhancements allow for resending RPCs that failed
4133 Severity : enhancement
4135 Description: Tunables on number of dirty pages in cacche
4136 Details : Allow to set limit on number of dirty pages cached.
4141 Description: client crash on unmount - lock still has references
4142 Details : In some error handling cases it was possible to leak a lock
4143 reference on a client while accessing a file. This was not
4144 harmful to the client during operation, but would cause the
4145 client to crash when the filesystem is unmounted.
4148 Frequency : specific case, rare
4150 Description: ETXTBSY on mds though file not in use by client
4151 Details : ETXTBSY is no longer incorrectly returned when attempting to
4152 chmod or chown a directory that the user previously tried to
4153 execute or a currently-executing binary.
4156 Frequency : extremely rare except on liblustre-based clients
4158 Description: Lustre space not freed when files are deleted
4159 Details : Clean up open-unlinked files after client eviction. Previously
4160 the unlink was skipped and the files remained as orphans.
4165 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4166 Details : In some cases it was possible to send a blocking callback to
4167 a client doing a glimpse, even though that client didn't get
4168 a lock granted. When the glimpse lock is cancelled on the OST
4169 the freed lock is left on the waiting list and corrupted the list.
4172 Frequency : all core dumps
4174 Description: Broke core dumps to lustre
4175 Details : Negative dentry may be unhashed if parent does not have UPDATE
4176 lock, but some callers, e.g. do_coredump, expect dentry to be
4177 hashed after successful create, hash it in ll_create_it.
4179 ------------------------------------------------------------------------------
4181 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4183 * Support for kernels:
4184 2.6.9-42.0.2.EL (RHEL 4)
4185 2.6.5-7.276 (SLES 9)
4186 2.4.21-40.EL (RHEL 3)
4187 2.6.12.6 vanilla (kernel.org)
4191 Frequency : always on RHEL 3
4193 Description: Number of open files grows over time
4194 Details : The number of open files grows over time, whether or not
4195 Lustre is started. This was due to a filp leak introduced
4196 by one of our kernel patches.
4198 ------------------------------------------------------------------------------
4200 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4202 * Support for kernels:
4203 2.6.9-42.EL (RHEL 4)
4204 2.6.5-7.276 (SLES 9)
4205 2.4.21-40.EL (RHEL 3)
4206 2.6.12.6 vanilla (kernel.org)
4211 Bugzilla : 5719, 9635, 9792, 9684
4212 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4213 Details : If a server is under heavy load and cannot reply to new
4214 connection requests before the client resends the (re)connect,
4215 the connection handling code can behave badly if two service
4216 threads are concurrently handing separate (re)connections from
4217 the same client. Add better locking to the connection handling
4218 code, and ensure that only a single connection will be processed
4219 for a given client UUID, even if the lock is dropped.
4221 Severity : enhancement
4223 Description: add TCP zero-copy support to kernel
4224 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4225 sends if the hardware supports scatter-gather and checksumming.
4226 This allows socklnd to do client-write and server-read more
4227 efficiently and reduce CPU utilization from skbuf copying.
4230 Frequency : only if NFS exporting from client
4232 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4233 Details : If setting attributes on a file created under NFS that had
4234 never been opened it would be possible to oops the client
4235 if the file had no objects.
4238 Frequency : always for liblustre
4240 Description: liblustre client does MDS+OSTs setattr RPC for each write
4241 Details : When doing a write from a liblustre client, the client
4242 incorrectly issued an RPC to the MDS and each OST the file was
4243 striped over in order to update the timestamps. When writing
4244 with small chunks and many clients this could overwhelm the MDS
4245 with RPCs. In all cases it would slow down the write because
4246 these RPCs are unnecessary.
4248 Severity : enhancement
4250 Description: allow number of MDS service threads to be changed at module load
4251 Details : It is now possible to change the number of MDS service threads
4252 running. Adding "options mds mds_num_threads={N}" to the MDS's
4253 /etc/modprobe.conf will set the number of threads for the next
4254 time Lustre is restarted (assuming the "mds" module is also
4255 reloaded at that time). The default number of threads will
4256 stay the same, 32 for most systems.
4261 Description: OST crash if filesystem is unformatted or corrupt
4262 Details : If an OST is started on a device that has never been formatted
4263 or if the filesystem is corrupt and cannot even mount then the
4264 error handling cleanup routines would dereference a NULL pointer.
4269 Description: NULL pointer deref in llap_from_page.
4270 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4271 incorrect) mapping if the page was truncated/reclaimed while it was
4272 searched for. Check for this condition and skip such pages when
4273 doing readahead. Introduce extra check to llap_from_page() to
4274 verify page->mapping->host is non-NULL (so page is not anonymous).
4277 Frequency : Sometimes when using sys_sendfile
4279 Description: "page not covered by a lock" warnings from ll_readpage
4280 Details : sendfile called ll_readpage without right page locks present.
4281 Now we introduced ll_file_sendfile that does necessary locking
4282 around call to generic_file_sendfile() much like we do in
4286 Frequency : with certain MDS communication failures at client mount time
4288 Description: NULL pointer deref after failed client mount
4289 Details : a client connection request may delayed by the network layer
4290 and not be sent until after the PTLRPC layer has timed out the
4291 request. If the client fails the mount immediately it will try
4292 to clean up before the network times out the request. Add a
4293 reference from the request import to the obd device and delay
4294 the cleanup until the network drops the request.
4297 Frequency : occasionally during client (re)connect
4299 Description: assertion failure during client (re)connect
4300 Details : processing a client connection request may be delayed by the
4301 client or server longer than the client connect timeout. This
4302 causes the client to resend the connection request. If the
4303 original connection request is replied in this interval, the
4304 client may trip an assertion failure in ptlrpc_connect_interpret()
4305 which thought it would be the only running connect process.
4308 Frequency : only with obd_echo servers and clients that are rebooted
4310 Description: kernel BUG accessing uninitialized data structure
4311 Details : When running an obd_echo server it did not start the ping_evictor
4312 thread, and when a client was evicted an uninitialized data
4313 structure was accessed. Start the ping_evictor in the RPC
4314 service startup instead of the OBD startup.
4316 Severity : enhancement
4317 Bugzilla : 10193 (patchless)
4318 Description: Remove dependency on various unexported kernel interfaces.
4319 Details : No longer need reparent_to_init, exit_mm, exit_files,
4320 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4323 Frequency : rare (only users of deprecated and unsupported LDAP config)
4325 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4326 Details : LDAP apparently contains 'lustreName' attributes instead of
4327 'name'. A simple remapping of the name is sufficient.
4330 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4332 Description: waiting_locks_callback trips kernel BUG if client is evicted
4333 Details : Running with the dump_on_timeout debug flag turned on makes
4334 it possible that the waiting_locks_callback() can try to dump
4335 the Lustre kernel debug logs from an interrupt handler. Defer
4336 this log dumping to the expired_lock_main() thread.
4338 Severity : enhancement
4340 Description: Support NFS exporting on 2.6 kernels.
4341 Details : Implement non-rawops metadata methods for NFS server to use without
4342 changing NFS server code.
4345 Frequency : very rare (synthetic metadata workload only)
4347 Description: two racing renames might cause an MDS thread to deadlock
4348 Details : Running the "racer" program may cause one MDS thread to rename
4349 a file from being the source of a rename to being the target of
4350 a rename at exactly the same time that another thread is doing
4351 so, and the second thread has already enqueued these locks after
4352 doing a lookup of the target and is trying to relock them in
4353 order. Ensure that we don't try to re-lock the same resource.
4356 Frequency : only very large systems with liblustre clients
4358 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4359 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4360 all OSTs in parallel. This allows the eviction of stale liblustre
4361 clients to proceed much faster than if they were done in series,
4362 and also offers similar improvements for other set_info RPCs.
4367 Description: excessive CPU usage during initial read phase on client
4368 Details : During the initial read phase on a client, it would agressively
4369 retry readahead on the file, consuming too much CPU and impacting
4370 performance (since 1.4.5.8). Improve the readahead algorithm
4371 to avoid this, and also improve some other common cases (read
4372 of small files in particular, where "small" is files smaller than
4373 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4378 Description: MDS crash when receiving packet with unknown intent.
4379 Details : Do not LBUG in unknown intent case, just return -EFAULT
4381 Severity : enhancement
4382 Bugzilla : 9293, 9385
4383 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4384 Details : Do not serialize getattr (non-intent version) and statfs.
4387 Frequency : occasional, when OST network is overloaded/intermittent
4389 Description: client evicted by OST after bulk IO timeout
4390 Details : If a client sends a bulk IO request (read or write) the OST
4391 may evict the client if it is unresposive to its data GET/PUT
4392 request. This is incorrect if the network is overloaded (takes
4393 too long to transfer the RPC data) or dropped the OST GET/PUT
4394 request. There is no need to evict the client at all, since
4395 the pinger and/or lock callbacks will handle this, and the
4396 client can restart the bulk request.
4399 Frequency : Always when mmapping file with no objects
4401 Description: client crashes when mmapping file with no objects
4402 Details : Check that we actually have objects in a file before doing any
4403 operations on objects in ll_vm_open, ll_vm_close and
4409 Description: Request leak when working with deleted CWD
4410 Details : Introduce advanced request refcount tracking for requests
4411 referenced from lustre intent.
4413 Severity : Enhancement
4415 Description: Cache open file handles on client.
4416 Details : MDS now will return special lock along with openhandle, if
4417 requested and client is allowed to hold openhandle, even if unused,
4418 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4419 closing files for every read/write openration.
4421 Severity : Enhancement
4423 Description: Cache open negative dentries on client when possible.
4424 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4425 negative dentries on lock revocation.
4430 Description: Remounting a client read-only wasn't possible with a zconf mount
4431 Details : It wasn't possible to remount a client read-only with llmount.
4433 Severity : enhancement
4434 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4435 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4436 MPICH is included to add Lustre-specific ADIO interfaces.
4437 This is based closely on the UFS ADIO layer and only differs
4438 in file creation, in order to allow the OST striping to be set.
4439 This is user-contributed code and not supported by CFS.
4444 Description: extended inode attributes (immutable, append-only) work improperly
4445 when 2.4 and 2.6 kernels are used on client/server or vice versa
4446 Details : Introduce kernel-independent values for these flags.
4448 Severity : enhancement
4451 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4452 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4453 granularity. Now, Fractional MB granularity is be supported,
4454 this is very useful for low memory system.
4456 Severity : enhancement
4458 Description: Getattr by fid
4459 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4460 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4461 have locks and avoids getattr rpc requests for it on every stat.
4464 Frequency : Always, for filesystems larger than 2TB
4466 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4467 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4468 filesystems larger than 2TB to be mounted with mballoc enabled.
4471 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4473 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4474 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4475 filesystems larger than 2TB is mounted then the kernel will
4476 silently corrupt the start of the filesystem. CONFIG_LBD is
4477 enabled for all CFS-supported kernels, but the possibility of
4478 this happening with a modified kernel config exists.
4480 Severity : enhancement
4482 Description: add client O_DIRECT support for 2.6 kernels
4483 Details : It is now possible to do O_DIRECT reads and writes to files
4484 in the Lustre client mountpoint on 2.6 kernel clients.
4486 Severity : enhancement
4488 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4489 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4490 parallel, not waiting for response from every OST before sending
4491 a rpc to the next OST.
4496 Description: setattr vs write race when updating file timestamps
4497 Details : Client processes that update a file timestamp into the past
4498 right after writing to the file (e.g. tar) it is possible that
4499 the updated file modification time can be reset to the current
4500 time due to a race between processing the setattr and write RPC.
4502 Severity : enhancement
4504 Description: Bring 'lfs find' closer in line with regular Linux find.
4505 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4506 -print0 options and obtains all the needed info through the lustre
4509 Severity : enhancement
4511 Description: support up to 1024 configured devices on one node
4512 Details : change obd_dev array from statically allocated to dynamically
4513 allocated structs as they are first used to reduce memory usage
4518 Description: Flush dirty partially truncated pages during truncate
4519 Details : Immediatelly flush partially truncated pages in filter_setattr,
4520 this way we completely avoid having any pages in page cache on OST
4521 and can retire ugly workarounds during writes to flush such pages.
4526 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4527 Details : protect inode from truncation within vfs_unlink() context
4528 just take a reference before calling vfs_unlink() and release it
4529 when parent's i_sem is free.
4534 Description: last_id value checked outside lock on OST caused LASSERT failure
4535 Details : If there were multiple MDS->OST object precreate requests in
4536 flight, it was possible that the OST's last object id was checked
4537 outside a lock and incorrectly tripped an assertion. Move checks
4538 inside locks, and discard old precreate requests.
4541 Frequency : always, if extents are used on OSTs
4543 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4544 Details : OSTs using the extents format would not zero the high 16 bits of
4545 the index physical block number. This is not a problem for any
4546 OST filesystems smaller than 16TB, and no kernels support ext3
4547 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4548 new/modified files) and can be fixed for existing filesystems
4549 with e2fsprogs-1.39-cfs1.
4554 Description: import connection selection may be incorrect if timer wraps
4555 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4556 import connections to be ignored if the 32-bit jiffies counter
4557 wraps. Use a 64-bit jiffies counter.
4560 Frequency : very large clusters immediately after boot
4562 Description: LNET request buffers exhausted under heavy short-term load
4563 Details : If a large number of client requests are generated on a service
4564 that has previously never seen so many requests it is possible
4565 that the request buffer growth cannot keep up with the spike in
4566 demand. Instead of dropping incoming requests, they are held in
4567 the LND until the RPC service can accept more requests.
4570 Frequency : Sometimes during replay
4572 Description: Assertion failure in ll_local_open after replay.
4573 Details : If replay happened on an open request reply before we were able
4574 to set replay handler, reply will become not swabbed tripping the
4575 assertion in ll_local_open. Now we set the handler right after
4576 recognising of open request
4579 Frequency : very rare
4581 Description: kernel reports "badness in vsnprintf"
4582 Details : Reading from the "recovery_status" /proc file in small chunks
4583 may cause a negative length in lprocfs_obd_rd_recovery_status()
4584 call to vsnprintf() (which is otherwise harmless). Exit early
4585 if there is no more space in the output buffer.
4587 Severity : enhancement
4589 Description: clear OBD RPC statistics by writing to them
4590 Details : It is now possible to clear the OBD RPC statistics by writing
4591 to the "stats" file.
4596 Description: Client mtime is not the same on different clients after utimes
4597 Details : In some cases, the client was using the utimes() syscall on
4598 a file cached on another node. The clients now validate the
4599 ctime from the MDS + OSTs to determine which one is right.
4604 Description: Inability to activate failout mode
4605 Details : lconf script incorrectly assumed that in pythong string's numeric
4606 value is used in comparisons.
4609 Frequency : always with multiple stripes per file
4611 Description: Inefficient object allocation for mutli-stripe files
4612 Details : When selecting which OSTs to stripe files over, for files with
4613 a stripe count that divides evenly into the number of OSTs,
4614 the MDS is always picking the same starting OST for each file.
4615 Return the OST selection heuristic to the original design.
4620 Description: mount failures may take full timeout to return an error
4621 Details : Under some heavy load conditions it is possible that a
4622 failed mount can wait for the full obd_timeout interval,
4623 possibly several minutes, before reporting an error.
4624 Instead return an error as soon as the status is known.
4626 Frequency : quota enabled and large files being deleted
4628 Description: releasing more than 4GB of quota at once hangs OST
4629 Details : If a user deletes more than 4GB of files on a single OST it
4630 will cause the OST to spin in an infinite loop. Release
4631 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4636 Description: statfs data retrieved from /proc may be stale or zero
4637 Details : When reading per-device statfs data from /proc, in the
4638 {kbytes,files}_{total,free,avail} files, it may appear
4639 as zero or be out of date.
4642 Frequency : systems with MD RAID1 external journal devices
4644 Description: lconf's call to blkid is confused by RAID1 journal devices
4645 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4646 of returning all block devices that match the journal UUID.
4649 Frequency : always, for aggregate stripe size over 4GB
4651 Description: assertion fails when trying to use 4GB stripe size
4652 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4653 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4656 Frequency : always on ppc64
4658 Description: the first write on an ext3 filesystem with mballoc got stuck
4659 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4660 perform endianness conversion.
4662 ------------------------------------------------------------------------------
4664 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4666 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4667 INTEROPERATE with older versions automatically. Please read the
4668 user documentation before upgrading any part of a live system.
4669 * WARNING: Lustre networking configuration changes are required with
4670 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4673 * Support for kernels:
4674 2.6.9-22.0.2.EL (RHEL 4)
4675 2.6.5-7.244 (SLES 9)
4676 2.6.12.6 vanilla (kernel.org)
4679 Severity : enhancement
4680 Bugzilla : 7981/8208
4681 Description: Introduced Lustre Networking (LNET)
4682 Details : LNET is new networking infrastructure for Lustre, it includes
4683 a reorganized network configuration mode (see the user
4684 documentation for full details) as well as support for routing
4685 between different network fabrics. Lustre Networking Devices
4686 (LNDs) for the supported network fabrics have also been
4687 created for this new infrastructure.
4689 Severity : enhancement
4690 Description: Introduced Access control lists
4691 Details : clients can set ACLs on files and directories in order to have
4692 more fine-grained permissions than the standard Unix UGO+RWX.
4693 The MDS must be started with the "-o acl" mount option.
4695 Severity : enhancement
4696 Description: Introduced filesystem quotas
4697 Details : Administrators may now establish per-user quotas on the
4700 Severity : enhancement
4702 Description: Configuration change for the XT3
4703 The PTLLND is now used to run Lustre over Portals on the XT3
4704 The configure option(s) --with-cray-portals are no longer used.
4705 Rather --with-portals=<path-to-portals-includes> is used to
4706 enable building on the XT3. In addition to enable XT3 specific
4707 features the option --enable-cray-xt3 must be used.
4712 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4713 Details : A race between allocating a new llcd and re-getting the llcd_lock
4714 allowed another thread to grab newly-allocated llcd.
4716 Severity : enhancement
4718 Description: 2.6 OST async journal commit and locking fix to improve performance
4719 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4720 2.6 kernels are now async as they already were in 2.4 kernels so
4721 that they can commit concurrently with the network bulk transfer.
4722 For block-allocated files the filter allocation semaphore is held
4723 to avoid filesystem fragmentation during allocation. BKL lock
4724 removed for 2.6 xattr operations where it is no longer needed.
4729 Description: lconf incorrectly determined whether two IP networks could talk
4730 Details : In some more complicated routing and multiple-network
4731 configurations, lconf will avoid trying to make a network
4732 connection to a disjoint part of the IP space. It was doing the
4733 math incorrectly for one set of cases.
4738 Description: Fix for potential infinite loop processing records in an llog.
4739 Details : If an llog record is corrupted/zeroed, it is possible to loop
4740 forever in llog_process(). Validate the llog record length
4741 and skip the remainder of the block on error.
4744 Frequency : occasional (liblustre only)
4746 Description: liblustre could not open files whose last component is a symlink
4747 Details : sysio_path_walk() would incorrectly pass the open intent to
4748 intermediate path components.
4751 Frequency : rare (liblustre only with non-standard tuning)
4752 Bugzilla : 7201 (7350)
4753 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4754 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4755 after it has been released, only do this for referenced locks
4757 Severity : enhancement
4759 Description: specifying an (invalid) directory default stripe_size of -1
4760 would reset the directory default striping
4761 Details : stripe_size -1 was used internally to signal directory stripe
4762 removal, now use "all default" to signal dir stripe removal
4763 as a directory striping of "all default" is not useful
4766 Frequency : common for large clusters running liblustre clients
4768 Description: doing an ls when liblustre clients are running is slow
4769 Details : sending a glimpse AST to a liblustre client waits for every AST
4770 to time out, as liblustre clients will not respond. Since they
4771 cannot cache data we refresh the OST lock LVB from disk instead.
4773 Severity : enhancement
4775 Description: doing an ls at the same time as file IO can be slow
4776 Details : enqueue and other "small" requests can be blocked behind many
4777 large IO requests. Create a new OST IO portal for non-IO
4778 requests so they can be processed faster.
4781 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4783 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4784 created file with mode 000
4785 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4786 allow an ftruncate() as a non-root user to a file with mode 000.
4787 Lustre now respects this flag to disable mode checks when
4788 truncating a file owned by the user
4791 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4794 Description: Revoking locks from clients that went dead or catatonic might take
4796 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4797 cancellation of such locks instant on servers without waiting for
4798 any reply from clients. Clients drops these locks when cancel
4799 notification from server is received without replying.
4802 Frequency : liblustre-only, when liblustre client dies or becomes busy
4804 Description: Doing ls on Linux clients can take a long time with active
4806 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4807 granting such locks to it in the first place if possible. Locks
4808 are taken by proxy on the OST during the read or write and
4809 dropped immediately afterward. Add connect flags handling, do
4810 not grant locks to liblustre clients for glimpse ASTs.
4812 Severity : enhancement
4814 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4815 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4816 sequential reads and behaves badly with applications doing random
4817 reads. Improve it by reading ahead at least read region, and
4818 avoiding excessive large RPC for small reads.
4820 Severity : enhancement
4822 Description: Creating more than 1000 files for a single job may cause a load
4823 imbalance on the OSTs if there are also a large number of OSTs.
4824 Details : qos_prep_create() uses an OST index reseed value that is an
4825 even multiple of the number of available OSTs so that if the
4826 reseed happens in the middle of the object allocation it will
4827 still utilize the OSTs as uniformly as possible.
4832 Description: OST or MDS may oops in ping_evictor_main()
4833 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4834 but doesn't restart at beginning of obd_exports_timed list
4837 Severity : enhancement
4839 Description: improve by-nid export eviction on the MDS and OST
4840 Details : allow multiple exports with the same NID to be evicted at one
4841 time without re-searching the exports list.
4844 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4846 Description: MDS may oops in groups_free()
4847 Details : in rare race conditions a newly allocated group_info struct is
4848 freed again, and this can be NULL. The 2.4 compatibility code
4849 for groups_free() checked for a NULL pointer, but 2.6 did not.
4852 Frequency : common for liblustre clients doing little filesystem IO
4853 Bugzilla : 9352, 7313
4854 Description: server may evict liblustre clients accessing contended locks
4855 Details : if a client is granted a lock or receives a completion AST
4856 with a blocking AST already set it would not reply to the AST
4857 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4860 Frequency : lfs setstripe, only systems with more than 160 OSTs
4862 Description: unable to set striping with a starting offset beyond OST 160
4863 Details : llapi_create_file() incorrectly limited the starting stripe
4864 index to the maximum single-file stripe count.
4867 Frequency : LDAP users only
4869 Description: lconf did not handle in-kernel recovery with LDAP properly
4870 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4872 Severity : enhancement
4874 Description: bind OST threads to NUMA nodes to improve performance
4875 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4876 node and do their allocations there to localize memory access
4878 Severity : enhancement
4880 Description: llmount can determine client NID directly from Myrinet (GM)
4881 Details : the client NID code from gmnalnid was moved directly into
4882 llmount, removing the need to use this or specifying the
4883 client NID explicitly when mounting GM clients with zeroconf
4886 Frequency : if client is started with down MDS
4888 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4889 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4890 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4891 ptlrpc_import_delayed_req() because we didn't check if the
4892 request was interrupted, and we also didn't break out of the
4893 event loop if there was no timeout
4898 Description: data loss during non-page-aligned writes to a single file from
4899 both multiple nodes and multiple threads on one node at same time
4900 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4901 inconsistency led to false short-reads, that were cached and later
4902 used by ->prepare_write() to fill in partially written page,
4903 leading to data loss.
4906 Frequency : always, if lconf --abort_recovery used
4908 Description: lconf --abort_recovery fails with 'Operation not supported'
4909 Details : lconf was attempting to abort recovery on the MDT device and not
4912 Severity : enhancement
4914 Description: remove cleanup logs
4915 Details : replace lconf-generated cleanup logs with lustre internal
4916 cleanup routines. Eliminates the need for client-cleanup and
4919 Severity : enhancement
4921 Description: add support for EAs (user and system) on lustre filesystems
4922 Details : it is now possible to store extended attributes in the Lustre
4923 client filesystem, and with the user_xattr mount option it
4924 is possible to allow users to store EAs on their files also
4926 Severity : enhancement
4928 Description: Add possibility (config option) to show minimal available OST free
4930 Details : When compiled with --enable-mindf configure option, statfs(2)
4931 (and so, df) will return least minimal free space available from
4932 all OSTs as amount of free space on FS, instead of summary of
4933 free spaces of all OSTs.
4935 Severity : enhancement
4937 Description: do not expand extent locks acquired on OST-side
4938 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4939 by server: they are not cached anyway.
4942 Frequency : when mmap is used/binaries executed from Lustre
4944 Description: Unmmap pages before throwing them away from read cache.
4945 Details : llap_shrink cache now attempts to unmap pages before discarding
4946 them (if unmapping failed - do not discard). SLES9 kernel has
4947 extra checks that trigger if this unmapping is not done first.
4952 Description: lconf didn't resolve symlinks before checking to see whether a
4953 given mountpoint was already in use
4956 Frequency : when migrating failover services
4957 Bugzilla : 6395, 9514
4958 Description: When migrating a subset of services from a node (e.g. failback
4959 from a failover service node) the remaining services would
4960 time out and evict clients.
4961 Details : lconf --force (implied by --failover) sets the global obd_timeout
4962 to 5 seconds in order to quickly disconnect, but this caused
4963 other RPCs to time out too quickly. Do not change the global
4964 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4966 Severity : enhancement
4967 Frequency : if MDS is started with down OST
4968 Bugzilla : 9439,5706
4969 Description: Allow startup/shutdown of an MDS without depending on the
4970 availability of the OSTs.
4971 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4972 Add appropriate locking and lov-osc refcounts for safe
4973 cleaning. Add osc abort_inflight calls in case the
4974 synchronize never started.
4977 Frequency : occasional (Cray XT3 only)
4979 Description: root not authorized to access files in CRAY_PORTALS environment
4980 Details : The client process capabilities were not honoured on the MDS in
4981 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4982 been accessed by an authorized user then root was able to access
4983 the file on the local client also. The root user capabilities
4984 are now allowed on the MDS, as this environment has secure UID.
4987 Frequency : occasional
4989 Description: ldiskfs "too long searching" message happens too often
4990 Details : A debugging message (otherwise harmless) prints too often on
4991 the OST console. This has been reduced to only happen when
4992 there are fragmentation problems on the filesystem.
4997 Description: Division by zero in statfs when all OSCs are inactive
4998 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4999 lov_statfs divides by value returned by lov_get_stripecnt().
5003 Bugzilla : 9489, 3273
5004 Description: First write from each client to each OST was only 4kB in size,
5005 to initialize client writeback cache, which caused sub-optimal
5006 RPCs and poor layout on disk for the first written file.
5007 Details : Clients now request an initial cache grant at (re)connect time
5008 and so that they can start streaming writes to the cache right
5009 away and always do full-sized RPCs if there is enough data.
5010 If the OST is rebooted the client also re-establishes its grant
5011 so that client cached writes will be honoured under the grant.
5016 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5017 Details : Now I/O RPCs go to different portal number and (presumably) fast
5018 lock requests (and glimses) and other RPCs get their own service
5019 threads pool that should be able to service those RPCs
5022 Severity : enhancement
5024 Description: Ability to exchange lustre version between client and servers and
5025 issue warnings at client side if client is too old. Also for
5026 liblustre clients there is ability to refuse connection of too old
5028 Details : New 'version' field is added to connect data structure that is
5029 filled with version info. That info is later checked by server and
5033 Frequency : rare, liblustre only.
5034 Bugzilla : 9296, 9581
5035 Description: Two simultaneous writes from liblustre at offset within same page
5036 might proceed at the same time overwriting eachother with stale
5038 Details : I/O lock withing llu_file_prwv was released too early, before data
5039 actually was hitting the wire. Extended lock-holding time until
5040 server acknowledges receiving data.
5043 Frequency : extremely rare. Never observed in practice.
5045 Description: avoid generating lustre_handle cookie of 0.
5046 Details : class_handle_hash() generates handle cookies by incrementing
5047 global counter, and can hit 0 occasionaly (this is unlikely, but
5048 not impossible, because initial value of cookie counter is
5049 selected randonly). Value of 0 is used as a sentinel meaning
5050 "unassigned handle" --- avoid it. Also coalesce two critical
5051 sections in this function into one.
5053 Severity : enhancement
5055 Description: allow liblustre clients to delegate truncate locking to OST
5056 Details : To avoid overhead of locking, liblustre client instructs OST to
5057 take extent lock in ost_punch() on client's behalf. New connection
5058 flag is added to handle backward compatibility.
5060 Severity : enhancement
5061 Bugzilla : 4928, 7341, 9758
5062 Description: allow number of OST service threads to be specified
5063 Details : a module parameter allows the number of OST service threads
5064 to be specified via "options ost ost_num_threads={N}" in the
5065 OSS's /etc/modules.conf or /etc/modprobe.conf.
5069 Bugzilla : 6146, 9635, 9895
5070 Description: servers crash with bad pointer in target_handle_connect()
5071 Details : In rare cases when a client is reconnecting it was possible that
5072 the connection request was the last reference for that export.
5073 We would temporarily drop the export reference and get a new
5074 one, but this may have been the last reference and the export
5075 was just destroyed. Get new reference before dropping old one.
5077 Severity : enhancement
5078 Frequency : if client is started with failover MDS
5080 Description: Allow multiple MDS hostnames in the mount command
5081 Details : Try to read the configuration from all specified MDS
5082 hostnames during a client mount in case the "primary"
5085 Severity : enhancement
5087 Description: Stop sending data to evicted clients as soon as possible.
5088 Details : Check if the client we are about to send or are sending data to
5089 was evicted already. (Check is done every second of waiting,
5090 for which l_wait_event interface was extended to allow checking
5091 of exit condition at specified intervals).
5094 Frequency : rare, normally only when NFS exporting is done from client
5096 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5098 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5101 Frequency : always, if srand() is called before liblustre initialization
5103 Description: Liblustre uses system PRNG disturbing its usage by user application
5104 Details : Introduce internal to lustre fast and high-quality PRNG for
5105 lustre usage and make liblustre and some other places in generic
5106 lustre code to use it.
5108 Severity : enhancement
5109 Bugzilla : 9477, 9557, 9870
5110 Description: Verify that the MDS configuration logs are updated when xml is
5111 Details : Check if the .xml configuration logs are newer than the config
5112 logs stored on the MDS and report an error if this is the case.
5113 Request --write-conf, or allow starting with --old_conf.
5115 Severity : enhancement
5117 Description: Handle symlinks in the path when checking if Lustre is mounted.
5118 Details : Resolve intermediate symlinks when checking if a client has
5119 mounted a filesystem to avoid duplicate client mounts.
5124 Description: lconf can hit an error exception but still return success.
5125 Details : The lconf command catches the Command error exception at the top
5126 level script context and will exit with the associated exit
5127 status, but doesn't ensure that this exit status is non-zero.
5132 Description: failure of ptlrpc thread startup can cause oops
5133 Details : Starting a ptlrpc service thread can fail if there are a large
5134 number of threads or the server memory is very fragmented.
5135 Handle this without oopsing.
5138 Frequency : always, only if liblustre and non-default acceptor port was used
5140 Description: liblustre cannot connect to servers with non-default acceptor port
5141 Details : tcpnal_set_default_params() was not called and was therefore
5142 ignoring the environment varaible TCPNAL_PORT, as well as other
5143 TCPNAL_ environment variables
5148 Description: two objects could be created on the same OST for a single file
5149 Details : If an OST is down, in some cases it was possible to create two
5150 objects on a single OST for a single file. No problems other
5151 than potential performance impact and spurious error messages.
5155 Bugzilla : 5681, 9562
5156 Description: Client may oops in ll_unhash_aliases
5157 Details : Client dcache may become inconsistent in race condition.
5158 In some cases "getcwd" can fail if the current directory is
5164 Description: Inode refcounting problems in NFS export code
5165 Details : link_raw functions used to call d_instantiate without obtaining
5166 extra inode reference first.
5170 Bugzilla : 9942, 9903
5171 Description: Referencing freed requests leading to crash, memleaks with NFS.
5172 Details : We used to require that call to ll_revalidate_it was always
5173 followed by ll_lookup_it. Also with revalidate_special() it is
5174 possible to call ll_revalidate_it() twice for the same dentry
5175 even if first occurence returned success. This fix changes semantic
5176 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5177 reference on a request referred from the intent.
5178 ll_intent_release() then releases such a request.
5181 Frequency : rare, normally benchmark loads only
5183 Description: unlinked inodes were kept in memory on the client
5184 Details : If a client is repeatedly creating and unlinking files it
5185 can accumulate a lot of stale inodes in the inode slab cache.
5186 If there is no other client load running this can cause the
5187 client node to run out of memory. Instead flush old inodes
5188 from client cache that have the same inode number as a new inode.
5191 Frequency : SLES9 2.6.5 kernel and long filenames only
5192 Bugzilla : 9969, 10379
5193 Description: utime reports stale NFS file handle
5194 Details : SLES9 uses out-of-dentry names in some cases, which confused
5195 the lustre dentry revalidation. Change it to always use the
5199 Frequency : rare, unless heavy write-truncate concurrency is continuous
5200 Bugzilla : 4180, 6984, 7171, 9963, 9331
5201 Description: OST becomes very slow and/or deadlocked during object unlink
5202 Details : filter_destroy() was holding onto the parent directory lock
5203 while truncating+unlinking objects. For very large objects this
5204 may block other threads for a long time and slow overall OST
5205 responsiveness. It may also be possible to get a lock ordering
5206 deadlock in this case, or run out of journal credits because of
5207 the combined truncate+unlink. Solution is to do object truncate
5208 first in one transaction without parent lock, and then do the
5209 final unlink in a new transaction with the parent lock. This
5210 reduces the lock hold time dramatically.
5213 Frequency : rare, 2.4 kernels only
5215 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5216 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5217 before unmounting their filesystem in order to avoid deadlock.
5218 The kernel_locked() function in 2.4 kernels only checks whether
5219 the kernel lock is held, not whether it is this process that is
5220 holding it as 2.6 kernels do.
5225 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5226 Details : The client ptlrpc code may be trying to reconnect to a down
5227 server before a previous connection attempt has timed out.
5228 Increase the reconnect interval to be longer than the connection
5229 timeout interval to avoid sending duplicate connections to
5233 Frequency : echo_client brw_test command
5235 Description: fix echo_client to work with OST preallocated code
5236 Details : OST preallocation code (5137) didn't take echo_client IO path
5237 into account: echo_client calls filter methods outside of any
5238 OST thread and, hence, there is no per-thread preallocated
5239 pages and buffers to use. Solution: hijack pga pages for IO. As
5240 a byproduct, this avoids unnecessary data copying.
5244 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5245 Description: Client can oops in mdc_commit_close() after open replay
5246 Details : It was possible for the MDS to return an open request with no
5247 transaction number in mds_finish_transno() if the client was
5248 evicted, but without actually returning an error. Clients
5249 would later try to replay that open and may trip an assertion
5250 Simplify the client close codepath, and always return an error
5251 from the MDS in case the open is not successful.
5254 Frequency : rare, 2.6 OSTs only
5256 Description: OST may deadlock under high load on fragmented files
5257 Details : If there was a heavy load and highly-fragmented OST filesystems
5258 it was possible to have all the OST threads deadlock waiting on
5259 allocation of biovecs, because the biovecs were not released
5260 until the entire RPC IO was completed. Instead, release biovecs
5261 as soon as they are complete to ensure forward IO progress.
5263 Severity : enhancement
5265 Description: Support for specifying external journal device at mount
5266 Details : If an OST or MDS device is formatted with an external journal
5267 device, this device major/minor is stored in the ext3 superblock
5268 and may not be valid for failover. Allow detecting and
5269 specifying the external journal at mount time.
5274 Description: Mounting an MDS with pending unlinked files may cause oops
5275 Details : target_finish_recovery() calls mds_postrecov() which returned
5276 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5277 considers this an error and immediately begins cleaning up the
5278 lov, just after starting the mds_lov process
5280 Severity : enhancement
5282 Description: Implement 'lfs df' to report actual free space on per-OST basis
5283 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5284 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5285 usage of objects; '-h' to report in human readable format.
5287 ------------------------------------------------------------------------------
5289 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5296 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5297 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5298 at startup during orphan unlinking. Instead, skip buddy update
5299 before setup, as it will be regenerated after recovery is complete.
5304 Description: If an OST is inactive, its locks might reference stale inodes.
5305 Details : lov_change_cbdata() must iterate over all namespaces, even if
5306 they are inactive to clear inode references from the lock.
5308 Severity : enhancement
5309 Frequency : occasional, if non-standard max_dirty_mb used
5311 Description: Client will block write RPCs if not enough grant
5312 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5313 then the client will block writes while waiting for another RPC
5314 to complete instead of consuming its dirty limit. With change
5315 we get improved performance when max_dirty_mb is small.
5317 Severity : enhancement
5318 Bugzilla : 3389, 6253
5319 Description: Add support for supplementary groups on the MDS.
5320 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5321 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5322 lookups for user supplementary groups into a cache.
5326 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5327 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5328 for liblustre O_EXCL is enforced by MDS.
5332 Description: readdir never returns NULL in liblustre.
5333 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5334 dentry in struct dirent.
5338 Frequency : liblustre only, depends on application IO pattern
5339 Description: liblustre clients evicted if not contacting servers
5340 Details : Don't put liblustre clients into the ping_evictor list, so
5341 they will not be evicted by the pinger ever.
5343 Severity : enhancement
5345 Description: Add ability to evict clients by NID from MDS.
5346 Details : By echoing "nid:$NID" string into
5347 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5348 $NID would be instantly evicted from this MDS and from all active
5349 OSTs connected to it.
5353 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5354 Details : lookup_it_finish() used to query file size from OSTs that was not
5359 Description: service threads change working directory to that of init
5360 Details : Starting lustre service threads may pin the working directory
5361 of the parent thread, making that filesystem busy. Threads
5362 now change to the working directory of init to avoid this.
5366 Frequency : during shutdown only
5367 Description: shutdown with a failed MDS or OST can cause unmount to hang
5368 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5373 Frequency : 2.6 only
5374 Description: chmod/chown may include an extra supplementary group
5375 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5376 supplementary group and if none is specified this is used.
5379 Bugzilla : 5479 (6816)
5381 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5382 Details : If lookup is in progress on a file that is unlinked we might try
5383 to revalidate the inode and fail in revalidate after lookup is
5384 complete and ll_file_open() enqueues the open again but
5385 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5388 Frequency : always, if lconf --abort_recovery used
5390 Description: lconf --abort_recovery fails with 'Operation not supported'
5391 Details : lconf was attempting to abort recovery on the MDT device and not
5394 ------------------------------------------------------------------------------
5396 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5401 Frequency : rare (only unsupported configurations with a node running as an
5403 Bugzilla : 6514, 5137
5404 Description: Mounting a Lustre file system on a node running as an OST could
5406 Details : OSTs now preallocates memory needed to write out data at
5407 startup, instead of when needed, to avoid having to
5408 allocate memory in possibly low memory situations.
5409 Specifically, if the file system is mounted on on OST,
5410 memory pressure could force it to try to write out data,
5411 which it needed to allocate memory to do. Due to the low
5412 memory, it would be unable to do so and the node would
5413 become unresponsive.
5415 Severity : enhancement
5417 Description: Addition of lconf --service command line option
5418 Details : lconf now accepts a '--service <arg>' option, which is
5419 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5421 Severity : enhancement
5423 Description: Failover mode is now the default for OSTs.
5424 Details : By default, OSTs will now run in failover mode. To return to
5425 the old behaviour, add '--failout' to the lmc line for OSTs.
5427 Severity : enhancement
5429 Description: Health checks are now provided for MDS and OSTs
5430 Details : Additional detailed health check information on MSD and OSTs
5431 is now provided through the procfs health_check value.
5434 Frequency : occasional, depends on IO load
5436 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5437 after numerous create/delete cycles
5438 Details : The ext3 inode allocation policy would not allocate new inodes
5439 very well on the OSTs because there are no new directories
5440 being created. Instead we look for groups with free space if
5441 the parent directories are nearly full.
5446 Description: Network or server problems during mount may cause partially
5447 mounted clients instead of returning an error.
5448 Details : The config llog parsing code may overwrite the error return
5449 code during mount error handling, returning success instead
5455 Description: MDS can fail to allocate large reply buffers
5456 Details : After long uptimes the MDS can fail to allocate large reply
5457 buffers (e.g. zconf client mount config records) due to memory
5458 fragmentation or consumption by the buffer cache. Preallocate
5459 some large reply buffers so that these replies can be sent even
5460 under memory pressure.
5464 Frequency : rare (liblustre)
5465 Description: fsx running with liblustre complained that using truncate() to
5466 extend the file doesn't work. This patch corrects that issue.
5467 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5468 ATTR_SIZE and lsm use in llu_setattr_raw.
5472 Frequency : rare, only 2.6 kernels
5473 Description: Unusual file access patterns on the MDS may result in inode
5474 data being lost in very rare circumstances.
5475 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5476 code in the 2.6 kernel caused the inode and/or EA data not to
5477 be read from disk, causing single-file corruption.
5481 Frequency : rare, only 2.6 filesystems using extents
5482 Description: Heavy concurrent write and delete load may cause data corruption.
5483 Details : It was possible under high-load situations to have an extent
5484 metadata block in the block device cache from a just-unlinked
5485 file overwrite a newly-allocated data block. We now unmap any
5486 metadata buffers that alias just-allocated data blocks.
5490 Frequency : filesystems with default stripe_count larger than 77
5491 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5492 Details : lconf specifies an inode size of 4096 bytes when the default
5493 stripe_count is larger than 77. This conflicts with the default
5494 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5495 this case to avoid pinning too much memory for large EAs.
5497 ------------------------------------------------------------------------------
5499 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5504 Frequency : rare (extremely heavy IO load with hundreds of clients)
5506 Description: Client is evicted, gets IO error writing to file
5507 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5508 caused clients to be evicted for AST timeouts. The fixes in
5509 bug 5192 mean we no longer need to have such short AST timeouts
5510 so ldlm_timeout has been increased.
5513 Frequency : occasional during --force or --failover shutdown under load
5514 Bugzilla : 5949, 4834
5515 Description: Server oops/LBUG if stopped with --force or --failover under load
5516 Details : a collection of import/export refcount and cleanup ordering
5517 issues fixed for safer force cleanup
5520 Frequency : only filesystems larger than 120 OSTs
5521 Bugzilla : 5990, 6223
5522 Description: lfs getstripe would oops on a very large filesystem
5523 Details : lov_getconfig used kfree on vmalloc'd memory
5526 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5527 Bugzilla : 6242, 6243
5528 Description: reading from files that had been truncated to a non-zero size
5529 but never opened returned no data
5530 Details : ll_file_read() reads zeros from no-object files to EOF
5535 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5536 Details : There's little harm in aborting MDS/OSS recovery and letting it
5537 try again, so I removed the LASSERT and return an error instead.
5539 Severity : enhancement
5541 Description: New debugging infrastructure for tracking down data corruption
5542 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5543 (b) cover more of the client-side code path, and (c) try to narrow
5544 down where problems occurred
5548 Bugzilla : 3819, 4364, 4397, 6313
5549 Description: Racing close and eviction MDS could cause assertion in mds_close
5550 Details : It was possible to get multiple mfd references during close and
5551 client eviction, leading to one thread referencing a freed mfd.
5553 Severity: : enhancement
5554 Bugzilla : 3262, 6359
5555 Description: Attempts to reconnect to servers are now more aggressive.
5556 Details : This builds on the enhanced upcall-less recovery that was added
5557 in 1.4.2. When trying to reconnect to servers, clients will
5558 now try each server in the failover group every 10 seconds. By
5559 default, clients would previously try one server every 25 seconds.
5564 Description: After recovery, certain operations trigger a failed
5565 assertion on a client.
5566 Details : Failing over an mds, using lconf -d --failover, while a
5567 client was doing a readdir() call would cause the client to
5568 LBUG after recovery completed and the readdir() was resent.
5570 Severity : enhancement
5572 Description: Default groups are now added by lconf
5573 Details : You can now run lconf --group <servicename> without having to
5574 manually add groups with lmc.
5577 Frequency : occasional
5579 Description: Nodes with an elan id of 0 trigger a failed assertion
5582 Frequency : always when accessing e.g. tty/console device nodes
5584 Description: tty and some other devices nodes cannot be used on lustre
5585 Details : file's private_data field is used by device data and lustre
5586 values in there got lost. New field was added to struct file to
5587 store fs-specific private data.
5590 Frequency : when exporting Lustre via NFS
5592 Description: NFSD failed occasionally when looking up a path component
5593 Details : NFSD is looking up ".." which was broken in ext3 directories
5594 that had grown large enough to become hashed.
5597 Frequency : Clusters with multiple interfaces not on the same subnet
5599 Description: Nodes will repeatedly try to reconnect to an interface which it
5600 cannot reach and report an error to the log.
5601 Details : Extra peer list entries will be created by lconf with some peers
5602 unreachable. lconf now validates the peer before adding it.
5605 Frequency : Only if a default stripe is set on the filesystem root.
5607 Description: Setting a default stripe on the filesystem root prevented the
5608 filesystem from being remounted.
5609 Details : The client was sending extra request flags in the root getattr
5610 request and did not allocate a reply buffer for the dir EA.
5613 Frequency : occasional, higher if lots of files are accessed by one client
5614 Bugzilla : 6159, 6097
5615 Description: Client trips assertion regarding lsm mismatch/magic
5616 Details : While revalidating inodes the VFS looks up inodes with ifind()
5617 and in rare cases can find an inode that is being freed.
5618 The ll_test_inode() code will free the lsm during ifind()
5619 when it finds an existing inode and then the VFS later attaches
5620 this free lsm to a new inode.
5624 Bugzilla : 6422, 7030
5625 Description: MDS deadlock between mkdir and client eviction
5626 Details : Creating a new file via mkdir or mknod (starting a transaction
5627 and getting the ns lock) can deadlock with client eviction
5628 (gets ns lock and trying to finish a synchronous transaction).
5631 Frequency : occasional
5632 Description: While starting a server, the fsfilt_ext3 module could not be
5634 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5635 kernels. Previously, lconf would still use the ext3 name
5636 when trying to load modules. Now, it will correctly use
5637 ext3 on 2.4 and ldiskfs on 2.6.
5639 Severity : enhancement
5640 Description: The default stripe count has been changed to 1
5641 Details : The interpretation of the default stripe count (0, to lfs
5642 or lmc) has been changed to mean striping across a single
5643 OST, rather than all available. For general usage we have
5644 found a stripe count of 1 or 2 works best.
5646 Severity : enhancement
5647 Description: Add support for compiling against Cray portals.
5648 Details : Conditional compiling for some areas that are different
5652 Frequency : occasional
5653 Bugzilla : 6409, 6834
5654 Description: Creating files with an explicit stripe count may lead to
5655 a failed assertion on the MDS
5656 Details : If some OSTs are full or unavailable, creating files may
5657 trigger a failed assertion on the MDS. Now, Lustre will
5658 try to use other servers or return an error to the
5662 Frequency : occasional
5664 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5665 caused lock timeout during readahead (since 1.4.2).
5666 Details : Processes doing readahead might match a lock that hasn't been
5667 granted yet if there are overlapping and conflicting lock
5668 requests. The readahead process waits on ungranted lock
5669 (original lock is CBPENDING), while OST waits for that process
5670 to cancel CBPENDING read lock and eventually evicts client.
5672 Severity : enhancement
5674 Description: Initial enabling of flock support for clients
5675 Details : Implements fcntl advisory locking and file status functions.
5676 This feature is provided as an optional mount flag (default
5677 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5678 locking are implemented yet, and those that are are not guaranteed
5679 to be completely correct in production environments.
5680 mount -t lustre -o [flock|noflock] ...
5683 Frequency : occasional
5685 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5686 assertion in the ext3 JBD (journaling) layer.
5687 Details : The b_committed_data struct is protected by the big kernel lock
5688 in 2.4 kernels, serializing journal_commit_transaction() and
5689 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5690 access to this struct. In 2.6 kernels there is finer grained
5691 locking to improve SMP performance of the JBD layer.
5695 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5700 Description: merge and cleanup kernel patches.
5703 -----------------------------------------------------------------------------
5705 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5707 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5708 versions of Lustre. Please update all servers and clients to
5709 version 1.4.2 or later at the same time. You must also run
5710 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5711 to update the configuration logs.
5713 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5714 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5715 - swab llog records of type '0' so we get proper header size/idx (5861)
5716 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5717 - fix rename of one directory over another leaking an inode (5953)
5718 - avoid SetPageDirty on 2.6 (5981)
5719 - don't re-add just-being-destroyed locks to the waiting list (5653)
5720 - when creating new directories, inherit the parent's custom
5721 striping settings if present parent (3048)
5722 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5723 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5724 - don't hold client locks on temporary worklist from l_lru (5666)
5725 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5726 - automatically evict dead clients (5921)
5727 - Update file size properly in create+truncate+fstat case (6196)
5728 - Do not unhash mountpoint dentries, do not allow removal of
5730 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5731 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5732 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5733 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5734 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5735 - Correctly handle reads of files with no objects (6243)
5736 - lctl recover will also mark a device active if deactivate used (5933)
5738 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5739 - allow --write-conf on an MDS with different nettype than client (5619)
5740 - don't write config llogs to MDS for mounts not from that MDS (5617)
5741 - lconf should create multiple TCP connections from a client (5201)
5742 - init scripts are now turned off by default; run chkconfig --on
5743 lustre and chkconfig --on lustrefs to use them
5744 - upcalls are no longer needed for clients to recover to failover
5746 - add --abort-recovery option to lconf to abort recovery on device
5748 - add support for an arbitrary number of OSTs (3026)
5749 - Quota support protocol changes.
5750 - forward compatibility changes to wire structs (6007)
5751 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5752 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5753 - improved "lustre status" script
5754 - initialize blocksize for non-regular files (6062)
5755 - added --disable-server and --disable-client configure options (5782)
5756 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5757 - Vanilla 2.4.29 support
5758 - increase maximum number of obd devices to 520 (6242)
5759 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5760 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5761 SLES 9 2.6.5 kernels
5762 - limit stripes per file to 160 (the maximum EA size) (6093)
5764 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5767 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5768 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5769 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5770 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5771 - small ext3 extents cleanups and fixes (5733)
5772 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5773 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5774 - avoid needless client->OST connect, fix handle mismatch (5317)
5775 - fix DLM error path that led to out-of-sync client, long delays (5779)
5776 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5777 - fix several locking issues related to i_size (5492,5624,5654,5672)
5778 - don't move pending lock onto export if it is already evicted (5683)
5779 - fix kernel oops when creating .foo in unlinked directory (5548)
5780 - fix deadlock in obdfilter statistics vs. object create (5811)
5781 - use time_{before,after} to avoid timer jiffies wrap (5882)
5782 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5783 - Do not leak request if server was not able to process it (5154)
5784 - If mds_open unable to find parent dir, make that negative lookup(5154)
5785 - don't create new directories with extent-mapping (5909, 5936)
5787 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5788 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5789 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5790 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5791 - support for RHEL4 kernel on i686 (5773)
5792 - provide error messages when incompatible logs are encountered (5898)
5794 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5795 * version 1.4.0.10 (1.4.1 release candidate 1)
5797 - don't keep a lock reference when lock is not granted (4238)
5798 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5799 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5800 - drop import inflight refcount on signal_completed_replay error (5255)
5801 - unlock page after async write error during send (3677)
5802 - handle missing objects in filter_preprw_read properly (5265)
5803 - no transno return for symlink open, don't save no-trasno open (3440)
5804 - don't try to complete elan receive that already failed (4012)
5805 - free RPC server reply state on error (5406)
5806 - clean up thread from ptlrpc_start_thread() on error (5160)
5807 - readahead could read extra page into cache that wasn't ejected (5388)
5808 - prevent races in class_attach/setup/cleanup/detach (5260)
5809 - don't dereference de->d_inode after l_dput of de (5458)
5810 - use "int" for stripe value returned from lock_to_stripe (5544)
5811 - mballoc allocation and error-checking fixes in 2.6 (5504)
5812 - block device patches to fix I/O request sizes in 2.6 (5482)
5813 - look up hostnames for IB nals (5602)
5814 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5815 - don't start multiple acceptors for the same port (5277)
5816 - fix incorrect LASSERT in mds_getattr_name (5635)
5817 - export a proc file for general "ping" checking (5628)
5818 - fix "lfs check" to not block when the MDS is down (5628)
5820 - service request history (4965)
5821 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5822 - create an "evict_client" /proc entry on OSTs, like the MDS has
5823 - fix mount usage message, return errors per mount(8) (5168)
5824 - change grep [] to grep "[]" in tests so they work in more UMLs
5825 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5826 - remove ancient LOV_MAGIC_V0 EA support (5047)
5827 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5828 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5829 - allow readahead allocations to fail when low on memory (5383)
5830 - mmap locking landed again, after considerable improvement (2828)
5831 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5833 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5836 - send OST transaction number in read/write reply to free req (4966)
5837 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5838 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5839 - the watchdog thread now runs as interruptible (5246)
5840 - flock/lockf fixes (but it's still disabled, pending 5135)
5841 - don't use EXT3 constants in llite code (5094)
5842 - memory shortage at startup could cause assertion (5176)
5844 - reorganization of lov code
5845 - single portals codebase
5847 - add extents/mballoc support (5025)
5848 - direct I/O reads in the obdfilter (4048)
5849 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5851 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5854 - send OST transaction number in read/write reply to free req (4966)
5855 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5856 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5857 - the watchdog thread now runs as interruptible (5246)
5858 - handle missing objects in filter_preprw_read properly (5265)
5859 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5860 - drop import inflight refcount on signal_completed_replay error (5255)
5861 - unlock page after async write error during send (3677)
5862 - return original error code on reconstructed replies (3761)
5863 - no transno return for symlink open, don't save no-trasno open (3440)
5865 - add pid to ldlm debugging output (4922)
5866 - bump the watchdog timeouts -- we can't handle 30sec yet
5867 - extra debugging for orphan dentry/inode bug (5259)
5869 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5872 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5873 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5874 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5875 - fix NULL dereference in /proc/sys/portals/routes (4827)
5876 - allow failed mdc_close() operations to be interrupted (4561)
5877 - stop precreate on OST before MDS would time out on it (4778)
5878 - don't send partial-page writes before EOF from client (4410)
5879 - discard client grant for sub-page writes on large-page clients (4520)
5880 - don't free dentries not owned by NFS code, check generation (4806)
5881 - fix lsm leak if mds_create_objects() fails (4801)
5882 - limit debug_daemon file size, always print CERROR messages (4789)
5883 - use transno after validating reply (3892)
5884 - process timed out requests if import state changes (3754)
5885 - update mtime on OST during writes, return in glimpse (4829)
5886 - add mkfsoptions to LDAP (4679)
5887 - use ->max_readahead method instead of zapping global ra (5039)
5888 - don't interrupt __l_wait_event() during strace
5890 - add software watchdogs to catch hung threads quickly (4941)
5891 - make lustrefs init script start after nfs is mounted
5892 - fix CWARN/ERROR duplication (4930)
5893 - return async write errors to application if possible (2248)
5894 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5895 - print NAL number in %x format (4645)
5896 - update barely-supported suse-2.4.21-171 series (4842)
5897 - support for sles 9 %post scripts
5898 - support for building 2.6 kernel-source packages
5899 - support for sles km_* packages
5901 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5904 - ignore -ENOENT errors in osc_destroy (3639)
5905 - notify osc create thread that OSC is being cleaned up (4600)
5906 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5907 - reconstruct ost_handle() like mds_handle() (4657)
5908 - create a new thread to do import eviction to avoid deadlock (3969)
5909 - let lconf resolve symlinked-to devices (4629)
5910 - don't unlink "objects" from directory with default EA (4554)
5911 - hold socknal file ref over connect in case target is down (4394)
5912 - allow more than 32000 subdirectories in a single directory (3244)
5913 - fix blocks count for O_DIRECT writes (3751)
5914 - OST returns ENOSPC from object create when no space left (4539)
5915 - don't send truncate RPC if file size isn't changing (4410)
5916 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5917 - bind to privileged port in socknal and tcpnal (3689)
5919 - rate limit CERROR/CWARN console message to avoid overload (4519)
5920 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5921 - basic mmap support (3918)
5922 - kernel patch series update from b1_4 (4711)
5924 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5927 - avoid crash during MDS cleanup with OST shut down (2775)
5928 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5929 - don't use bad inodes on the MDS (3744)
5930 - dynamic object preallocation to improve recovery speed (4236)
5931 - don't hold spinlock over lock dumping or change debug flags (4401)
5932 - don't zero obd_dev when it is force cleaned (3651)
5933 - print grants to console if they go negative (4431)
5934 - "lctl deactivate" will stop automatic recovery attempts (3406)
5935 - look for existing locks in ldlm_handle_enqueue() (3764)
5936 - don't resolve lock handle twice in recovery avoiding race (4401)
5937 - revalidate should check working dir is a directory (4134)
5939 - don't always mark "slow" obdfilter messages as errors (4418)
5941 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5944 - don't close LustreDB during write_conf until it is done (3860)
5945 - fix typo in lconf for_each_profile (3821)
5946 - allow dumping logs from multiple threads at one time (3820)
5947 - don't allow multiple threads in OSC recovery (3812)
5948 - fix debug_size parameters (3864)
5949 - fix mds_postrecov to initialize import for llog ctxt (3121)
5950 - replace config semaphore with spinlock (3306)
5951 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5952 - don't allow enqueue to complete on a destroyed export (3822)
5953 - down write_lock before checking llog header bitmap (3825)
5954 - recover from lock replay timeout (3764)
5955 - up llog sem before sending rpc (3652)
5956 - reduce ns lock hold times when setting kms (3267)
5957 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5958 - fix NULL deref and obd_dev leak on setup error (3312)
5959 - replace some LBUG about llog ops with error handling (3841)
5960 - don't match INVALID dentries from d_lookup and spin (3784)
5961 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5962 - fix invalid assertion in ptlrpc_set_wait (3880)
5964 - add libwrap support for the TCP acceptor (3996)
5965 - add /proc/sys/portals/routes for non-root route listing (3994)
5966 - allow setting MDS UUID in .xml (2580)
5967 - print the stack of a process that LBUGs (4228)
5969 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5972 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5973 - make sure to unset replay flag from failed open requests (3440)
5974 - if default stripe count is 0, use OST count for inode size (3636)
5975 - update parent mtime/ctime on client for create/unlink (2611)
5976 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5977 - free recovery state on server during a forced cleanup (3571)
5978 - unregister_reply for resent reqs (3063)
5979 - loop back devices mounting and status check on 2.6 (3563)
5980 - fix resource-creation race that can provoke i_size == 0 (3513)
5981 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5982 - more debugging for page-accounting assertion (3746)
5983 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5984 - avoid deadlock after precreation failure (3758)
5985 - fix race and lock order deadlock in orphan handling (3450, 3750)
5986 - add validity checks when grabbing inodes from l_ast_data (3599)
5988 - add /proc/.../recovery_status to obdfilter (3428)
5989 - lightweight CDEBUG infrastructure, debug daemon (3668)
5990 - change default OSC RPC parameters to be better on small clusters
5991 - turn off OST read cache for files smaller than 32MB
5992 - install man pages and include them in rpms (3100)
5993 - add new init script for (un)mounting lustre filesystems (2593)
5994 - run chkconfig in %post for init scripts (3701)
5995 - drop scimac NAL (unmaintained)
5997 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6000 - clean kiobufs before and after use (3485)
6001 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6002 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6003 - change init script to fail more clearly if not run as root (1528)
6004 - allow clients to reconnect during replay (1742)
6005 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6006 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6007 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6008 - fix cond_resched (3554)
6009 - only evict unfinished clients after recovery (3515)
6010 - allow bulk resend, prevent data loss (3570)
6011 - dynamic ptlrpc request buffer allocation (2102)
6012 - don't allow unlinking open directory if it isn't empty (2904)
6013 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6014 - remove extraneous obd dereference causing LASSERT failure (3334)
6015 - don't use get_cycles() when creating temp. files on the mds (3156)
6016 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6017 - handle EEXIST for set-stripe, set proper directory name (3336)
6019 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6020 - fix ksocknal_fmb_callback() error messages (2918)
6022 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6025 - don't copy lvb into (possibly NULL) reply on error (2983)
6026 - don't deref dentry after dput, don't free lvb on error (2922)
6027 - use the kms to determine writeback rpc length (2947)
6028 - increment oti_logcookies when osc is inactive (2948)
6029 - update client's i_blocks count via lvb messages (2543)
6030 - handle intent open/close of special files properly (1557)
6031 - mount MDS with errors=remount-ro, like obdfilter (2009)
6032 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6033 - don't use cancelling-locks' kms values (2947)
6034 - use highest lock extent for kms, not last one (2925)
6035 - don't dereference ERR_PTR() dentry in error handling path (3107)
6036 - fix thread race in portals_debug_dumplog() (3122)
6037 - create lprocfs device entries at setup instead of at attach (1519)
6038 - common AST error handler, don't evict client on completion race (3145)
6039 - zero nameidata in detach_mnt in 2.6 (3118)
6040 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6041 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6042 - initialize RPC timeout timer earlier for 2.6 (3219)
6043 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6044 - print nal/nid for unknown nid (3258)
6045 - additional checks for oscc recovery before doing precreate (3284)
6046 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6047 - don't crash in mdc_close for bad permissions on open (3285)
6048 - zero i_rdev for non-device files (3147)
6049 - clear page->private before handing to FS, better assertion (3119)
6050 - tune the read pipeline (3236)
6051 - fix incorrect decref of invalidated dentry (2350)
6052 - provide read-ahead stats and refine rpc in flight stats (3328)
6053 - don't hold journal transaction open across create RPC (3313)
6054 - update atime on MDS at close time (3265)
6055 - close LDAP connection when recovering to avoid server load (3315)
6056 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6057 - don't leak open file on MDS after open resend (3325)
6058 - serialize filter_precreate and filter_destroy_precreated (3329)
6059 - loop device shouldn't call sync_dev() for nul device (3092)
6060 - clear page cache after eviction (2766)
6061 - resynchronize MDS->OST in background (2824)
6062 - refuse to mount the same filesystem twice on same mountpoint (3394)
6063 - allow llmount to create routes for mounting behind routers (3320)
6064 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6065 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6066 - fix rare problem with rename on htree directories (3417)
6068 - allow default OST striping configuration per directory (1414)
6069 - fix compilation for qswnal for 2.6 kernels (3125)
6070 - increase maximum number of MDS request buffers for large systems
6071 - change liblustreapi to be useful for external progs like lfsck (3098)
6072 - increase local configuration timeout for slow disks (3353)
6073 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6075 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6078 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6079 - don't overwrite extent policy data in reply if lock was blocked (2901)
6080 - drop filter export grants atomically with removal from device (2663)
6081 - del obd_self_export from work_list in class_disconnect_exports (2908)
6082 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6083 - swab reply message in mdc_close, other PPC fixes (2464)
6084 - fix destroying of named logs (2325)
6085 - overwrite old logs when running lconf --write_conf (2264)
6086 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6087 - fix race in target_handle_connect (2898)
6088 - mds_reint_create() should take same inode create lock (2926)
6089 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6090 - don't close files for self_export to avoid uninitialized obd (2936)
6091 - allow MDS with the same name as client node (2939)
6092 - hold dentry reference for closed log files for unlink (2325)
6093 - reserve space for all logs during transactions (2059)
6094 - don't evict page beyond end of stripe extent (2925)
6095 - don't oops on a deleted current working directory (2399)
6096 - handle hard links to targets without a parent properly (2517)
6097 - don't dereference NULL lock when racing during eviction (2867)
6098 - don't grow lock extents when lots of conflicting locks (2919)
6100 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6103 - account for cache space usage on clients to avoid data loss (974)
6104 - lfsck support in lustre kernel code (2349)
6105 - reduce journal credits needed for BRW writes (2370)
6106 - orphan handling to avoid losing space on client/server crashes
6107 - ptlrpcd can be blocked, stopping ALL progress (2477)
6108 - use lock value blocks to assist in proper KMS, faster stat (1021)
6109 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6110 - recovery for initial connections (2355)
6111 - fixes for mds_cleanup_orphans (1934)
6112 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6113 - block all file creations until orphan recovery completes (1901)
6114 - client remove rq_connection from request struct (2423)
6115 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6116 - recovery timer race (2670)
6117 - mdc_close recovey bug (2532)
6118 - ptlrpc cleanup bug (2710)
6119 - mds timeout on local locks (2588)
6120 - namespace lock held during RPCs (2431)
6121 - handle interrupted sync write properly (2503)
6122 - don't try to handle a message that hasn't been replied to (2699)
6123 - client assert failure during cleanup after abort recovery (2701)
6124 - leak mdc device after failed mount (2712)
6125 - ptlrpc_check_set allows timedout requests to complete (2714)
6126 - wait for inflight reqs when ptlrpcd finishes (2710)
6127 - make sure unregistered services are removed from the srv_list
6128 - reset bulk XID's when resending them (caught by 1138 test)
6129 - unregister_bulk after timeout
6130 - fix lconf error (2694)
6131 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6132 - readahead locks pages, leaves pending causing memory pressure (2673)
6133 - increase OST request buffers to 4096 on large machines (2729)
6134 - fix up permission of existing directories in simple_mkdir (2661)
6135 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6136 - don't assign transno to errored transactions (2742)
6137 - don't delete objects on OST if given a bogus objid from MDS (2751)
6138 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6139 - if rq_no_resend, then timeout request after recovery (2432)
6140 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6141 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6142 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6143 - MDS recovery shouldn't delete objects that it has given out (2730)
6144 - if enqueue arrives after completion, don't clobber LVB (2819)
6145 - don't unlock pages twice when trigger_group_io returns error (2814)
6146 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6147 - don't write pages to disk if there was an error (1450)
6148 - don't ping imports that have recovery disabled (2676)
6149 - take buffered bytes into account when balancing socknal conn (2817)
6150 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6151 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6152 - remove little-endian swabbing of llog records (1987)
6153 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6154 - retry reposting request buffers if they fail (1191)
6155 - grow extent at grant time to avoid granting a revoked lock (2809)
6156 - lock revoke doesn't evict page if covered by a second lock (2765)
6157 - disable VM readahead to avoid reading outside lock extents (2805)
6159 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6160 - updated kernel patches for hp-2.4.20 kernel (2681)
6162 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6165 - fix truncated write corruption (2366)
6166 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6168 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6169 - ensure len(uuid) < 37 in lmc (1171)
6170 - fix ia64 OOPS in llog_test (2255)
6171 - zero end of page at obdfilter for partial page writes (2648)
6172 - don't leave stale dentries around after renames (bug 2428)
6173 - fix timeouts when evicting a client with a single lock held (2642)
6174 - set deadline for the initial HELLO message to drain (2634)
6175 - print out dotted-quad IP addresses in the socknal (2302)
6177 - additional debugging for MDS client eviction problem (2443)
6178 - fix mkfsoptions support for osts (2603, 2604)
6180 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6183 - add series for the vanilla 2.6.0 kernel
6184 - add series for the vanilla 2.4.24 kernel
6185 - add series for a cray x86/64 UL kernel drop
6186 - fix xattr patches for the vanilla 2.4.19 series
6188 - generate true UUIDs in lmc (1171)
6189 - have portals stack dumping break in UML (2466)
6190 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6191 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6192 - wake callback waiting threads on client eviction (2460)
6193 - Add --ptldebug and --subsystem to lmc (1719)
6194 - update assertion to allow safe interrupt allocation
6195 - set rq_no_resend for cancel requests (2432)
6196 - recalculate ptlrpcd timeout after resend (2494)
6197 - call vfs_rmdir when removing pending directories (2368)
6198 - fix renaming a file to itself (2429)
6199 - lmc creates a default one-stripe lov (2454)
6200 - expand procfs space to handle large clusters (2326)
6201 - increase UML stack to avoid overflow
6202 - update lconf's list of debug and subsystem masks
6203 - fix lfs find --obd (2510)
6204 - /proc tunable for disabling filter read caching (2591)
6205 - stop rpm packages from altering slapd.conf (2301)
6206 - disable nagle in the socknal under 0conf (2578)
6207 - choose mds inode size based on stripe count (2572)
6208 - fix kernel-source rpm problems (2516)
6210 - add --disable-doc to avoid pdf generation (2421)
6211 - update documentation, tests, type-os, comments
6212 - avoid format warnings on ia64
6213 - remove the TOE NAL
6214 - tiny code cleanups by removing unused fields
6216 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6219 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6220 - remove the most blatant lies from BUILDING file (2371)
6221 - change default debug level to reasonable production setting
6222 - reduce client side cache size to reduce cache flush time
6223 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6224 - make TCP zerocopy and pinger support enabled by default (2476)
6225 - sync writes completed after process exits caused crashes (2319)
6226 - maintain correct mount count on the MDS (2356)
6227 - backout 1557, because 2316 wasn't really fixed
6228 - better file I/O statistics gathering in /proc
6229 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6230 - another kernel patch to fix zero-copy TCP function export
6231 - don't take duplicate lock when processing re-sent getattr (2420)
6232 - lctl uses obd_self_export instead of creating new conn (2353)
6233 - MDS/OST recovery case which requires object creation asserted (2425)
6234 - move lfs from /usr/sbin to /usr/bin in packages
6235 - fix race between mds_client_add and mds_client_free (2417)
6236 - use kmalloc instead of slabs in portals (2430)
6237 - don't create duplicate records when a failover MDS is present (2442)
6238 - remove unnecessary mount age check (2332)
6239 - don't remove directory inodes from locks prematurely (2451)
6240 - don't break if MDS service name is the same as hostname (2103)
6241 - fix races in client write RPC generation when cache full (2482)
6243 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6246 - remove now-unused request->rq_obd (278)
6247 - if an allocation fails, print out how much memory we've used (1933)
6248 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6249 - add the "configurable stack size" patch to most series files (1256)
6250 - ability to write large log records, for 100+ OST configs (2306)
6251 - fix NULL deref when filter_prep fails (2314)
6252 - fix operator precedence error in filter_sync
6253 - dynamic allocation of socknal TX descriptors (2315)
6254 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6255 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6256 - ensure that CWARN messages in Portals always get to the syslog
6257 - __init/__exit are not for prototype decls (ldlm_init/exit)
6258 - x86-64 compile warning fixes
6259 - fix gateway LMC keyword conflict (2318)
6260 - fix MDS lock inversions in getattr/reint paths (1844)
6261 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6262 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6263 - DLM race condition prevented some lock evictions (2328)
6264 - ENOMEM detection and retry on socknal sends (2230)
6265 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6266 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6267 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6268 - better error messages when a client is rejected during recovery (1505)
6269 - avoid cancelling locks which were never granted, after failure (2330)
6270 - fix i_sem/journal inversion in mds_client_add (2333)
6271 - fix truncate/getattr lock cycle deadlock (2334)
6272 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6273 - fix two rare exit paths which could leak an l_lock() ref (2321)
6274 - fencepost error in MDS/OST orphan recovery (2226)
6275 - make log record alignment 8 bytes (1988)
6276 - lstripe now fails when requested offset > ost_count (2237)
6277 - ensure that all kernel series have a complete list.h (1607)
6278 - fix crashes in special-file operations (2316)
6279 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6281 - allow configurable automake binary, for testing new versions
6282 - small update to the lfs documentation
6284 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6286 * fix negative export reference count in fsfilt_sync (2312)
6288 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6289 * release candidate 0.9.1
6291 - orphans are moved into the PENDING directory for possible recovery
6292 - replayed opens now open by fid for orphan/rename safety (1042)
6293 - last close of an orphan inode generates a transno (683)
6294 - chdir() and mount() now pin the directory entry (1020)
6295 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6296 - discard very old requests without processing them (1502)
6297 - remove some common, well-understood CERRORs (1505)
6298 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6299 - clear "grant" flags in OST replies until OST grant code lands (1644)
6300 - fix read performance by not clobbering i_blksize on client (1598)
6301 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6302 - make LRU size a /proc tunable, clears locks when reduced (707)
6303 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6304 - server locks take references on exports now (1558)
6305 - build fixes for 2.4.20-rh trees (1663)
6306 - return an error from lov_create if all OSCs are inactive (1751)
6307 - fix import levels when a reconnect happens without a timeout (1597)
6308 - exit early from mds_open if we get a lookup error (1749)
6309 - partial page read at EOF wouldn't wait for disk before sending (1642)
6310 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6311 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6312 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6313 - write/truncate lock inversion (1639)
6314 - Don't auto-load obdclass, portals modules during cleanup (1495)
6315 - fix timestamps from jumping to "now" (1763)
6316 - extra journal assertions (1648)
6317 - add an extra multiunlink test (1771)
6318 - fix read_record/write_record API (1776)
6319 - fix leak of offset_extent, possible incorrect i_size later (1772)
6320 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6321 - Debugging for the kqswnal_get_idle_tx problems (1820)
6322 - Allow recovery to be attempted multiple times (1536)
6323 - Write out MDS last_rcvd file after it is first created (1600)
6324 - Fix tx_descriptor leak in failed transmit situations (1827)
6325 - ext3 journaling fixes for assertion failure after IO error (1871)
6326 - class_export_put() on freed export after completion AST error (1896)
6327 - Fix revalidate looping in VFS (1322)
6328 - Don't access a freed export during MDS_REINT timeout (1521)
6329 - Add open-unlink recovery support on the MDS (1673,1764)
6330 - Return an error if no MDS data was read from last_rcvd (1946)
6331 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6332 - Don't setup a disk device that doesn't match exported UUID (317)
6333 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6334 - avoid committing NULL handle in force close
6335 - local.sh is now a one-stripe LOV configuration
6336 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6337 - fix ext3 htree duplicate directory entry corruption (1516)
6338 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6339 - update to new LOV EA format (2097)
6340 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6342 2003-06-15 Phil Schwan <phil@clusterfs.com>
6345 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6346 - per-import/export recovery handling (958, 931, 959)
6347 - multiple last-rcvd slots, for serving multiple FSes (949)
6348 - connections are again shared between multiple imp/exports (963, 964)
6349 - "umount -f" would hang if any requests needed to be sent (393, 978)
6350 - avoid pinning large req buffer by copying for queued messages (989)
6351 - add "uuid" to "lctl device" command to help upcalls (991)
6352 - "open" RPCs with transnos would confuse recovery counters (1037)
6353 - do proper endian conversion of all wire messages (288, 340, 891)
6354 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6355 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6356 - fix ptlrpc_connection leak in target_handle_connect (1174)
6357 - fix import refcounting bug in OST and MDS cleanup (1134)
6358 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6359 - fix very unlikely obd_types race condition (501)
6360 - remove osc_open hack for echo_client (1187)
6361 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6362 - a failure in read_inode2 leads to deadlock (1139)
6363 - cancel ack-locks as soon as transaction is committed (1072)
6364 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6365 - make sure to commitrw after any preprw to avoid deadlock (1162)
6366 - failing to execute a file in a lustre FS would lock inode (1203)
6367 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6368 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6369 - obd_uuid2tgt would walk off the end of the list (1255)
6370 - on IA64 the osc would give portals incorrect bulk size (1258)
6371 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6372 - fix lock inversion caused by new llite matching code (1282)
6373 - limit the number of dirty pages on a client to 10MB (1286)
6374 - timed out locks were not being corrected cancelled (1289)
6375 - fix O_DIRECT above 4GB on IA-32 (1292)
6376 * major user-visible changes
6377 - fail out/fail over policy now controlled by the upcall (993)
6379 - add OBD_PING to check server availability and failure (954)
6380 - lustre messages are now sent in sending host order (288, 340, 891)
6381 - add eadatalen to MDS getattr reply (340)
6382 - OST read replies may contain second buffer, with per-page status (593)
6384 2003-03-11 Phil Schwan <phil@clusterfs.com>
6387 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6388 - failed open()s could cause deadlock; fixed (867, 869)
6389 - stop cancelling OST locks when files are closed (481)
6390 - overlapping XID spaces caused network corruption (851, 853)
6391 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6392 - setattr_raw added, to do single-RPC, server-side setattrs
6393 - lmc/lconf syntax change for OST UUIDs
6394 - fix crashy race condition between ptlrpc_free_req and osc_close
6395 - don't use request in mdc_enqueue if we hit a timeout (889)
6396 - don't set the inode i_size for regular files from the MDS (896)
6397 - handle out of order completion AST (842)
6398 - don't LBUG if a lock request times out after receiving AST (913)
6399 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6400 - if a bad lock AST arrives, send an error instead of dropping entirely
6401 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6402 - fix leak in bulk IO when only partially completed (899, 900, 926)
6403 - fix O_DIRECT for ia64 (55)
6404 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6405 - C-z after timeout could hang a process forever; fixed (977)
6407 - client-side I/O cache (678, 924, 929, 941, 970)
6409 - READPAGE and SETATTRs which don't take server-side locks get
6412 2003-02-11 Phil Schwan <phil@clusterfs.com>
6415 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6416 - Fix fsfilt_extN_readpage() to read a full page of directory
6417 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6418 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6419 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6420 - Most intent operations were removed, in favour of a new RPC mode
6421 that does a single RPC to the server and bypasses most of the VFS
6422 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6423 - Aggressively cancel local locks on DLM servers
6424 - mds_reint_unlink sends EA to the client if it's the last nlink.
6425 client uses that EA to unlink OST objects.
6426 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6427 - recursive symlinks were fixed (439)
6428 - fixed NULL deref in DEBUG_REQ
6429 - filter_update_lastobjid no longer calls sync, which annoyed extN
6430 - fixed multi-client small-writes to a single file problem (445)
6431 - fixed mtime updates during file writes (607)
6432 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6433 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6434 - fixed Linux OST size reporting problem (444, 656)
6435 - OST now updates object mtime with writes or setattr (607, 619)
6436 - client verifies file size before zeroing page past EOF (445)
6437 - OST now writes last allocated objid to disk with allocation (108)
6438 - LOV on echo now works (409)
6440 - mds_reint_unlink sends a new buffer, with the EA included. this
6441 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6442 set if a regular file was being unlinked, and it was the last link
6443 - use PtlGet from the target for bulk writes (315)
6444 - OST now updates object mtime with writes or setattr (607, 619)
6445 - LDLM now has a grant-time callback to revalidate locked items, if
6447 - Many MDS operations were reorganized to combat race conditions
6449 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6450 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6452 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6455 - Fully reactivate OST imports after reconnection (512, others)
6456 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6457 - More graceful error handling for truncating on dead OST (515)
6458 - Don't error out unless we're actually accessing dead stripes (474)
6459 - Fix garbage sizes when stripes are missing (410)
6460 - LRU counters were broken, causing constant lock purge (433, 432)
6461 - garbage on read from stripes with failed OSTs (441)
6462 - mark OSCs as active before reconnecting during recovery (438)
6463 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6464 - lfind did not preserve OST order in output (443)
6465 - symlinks cause hung clients, incorrect data (439)
6466 - stop dereferencing request after dropping refcount (457)
6467 - don't LASSERT(spin_is_locked) on non-SMP (455)
6468 - fixes for many rename() bugs
6469 - fstat didn't correctly synchronize attributes (399)
6470 - server must handle lock cancellation during blocking AST prep (487)
6471 - bulk descriptors were free()d too soon (511)
6472 - fix paths in lconf, which would load incorrect modules (451, 507)
6473 - fix confusing lconf 'host not found' error message (386)
6474 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6475 - fix race condition in mdc_blocking_ast() for inode access (526)
6476 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6477 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6478 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6479 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6480 - fix lstripe in several ways (kernel side) (527)
6481 - fix request leak in ldlm_cli_enqueue (262)
6482 - incorrect OSC was marked inactive after OST failure
6483 - call mds_fs_cleanup before unmounting filesystem (524)
6484 - fix races between taking ns_lock and ldlm_lock_change_resource
6485 - fix races updating LOV export open file list
6486 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6487 - fix recovery NULL deref in ldlm_cli_cancel_unused
6488 - fix some DLM races by using new hash table for lock handles (419)
6489 - permit the client to specify desired inodes, at replay
6490 - duplicate requests when we queue them for replay reintegration
6491 - fix last_rcvd offset calculation
6492 - sync after each recovered transaction, so we always make progress
6493 - never, not always, ERESTART requests without transnos
6494 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6496 - skip replay if the MDS says that the client is already connected
6497 - don't check for a recovery-enabled export to match lctl's UUID
6498 - don't INC_USE_COUNT for phantom exports
6499 - don't crash when cleaning up phantom exports (567)
6500 - don't double-finish or set replay data for errored mdc_open requests
6501 - abort requests when they time out, so we don't get old replies
6502 - send/receive replies for AST messages again
6503 - if the client says that it doesn't have the lock, cancel it on the
6505 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6506 mark it as destroyed, it will all work out when decref is called
6507 - fix module use counts (22, 581)
6509 - ASTs now expect a reply (server cancels lock on error reply)
6511 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6514 - fix many simultaneous client startup (392)
6515 - fix dentry->d_it clobbering
6516 - credentials weren't being shipped for readdir/getattr operations
6517 - remove invalid assertions triggered during some concurrent MD
6519 - proper Lustre versions added (336, 389)
6520 - fix memory leak for create error case (398)
6521 - fix LOV locking bug that would get cli/srv out of sync
6522 - fix echo client over LOV (409)
6523 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6524 - fix double-O_EXCL intent crash (424)
6525 - avoid sending multiple lock CANCELs (352)
6527 - MDS can do multi-client recovery (modulo bugs in new code)
6529 - many updates, edits, cleanups
6531 2002-11-18 Phil Schwan <phil@clusterfs.com>
6534 - fix null d_it dereference (346)
6535 - fix full OST/dbench hang (333)
6536 - fix permission problem with file removal (286)
6537 - fix removal of OSCs from LOV when they fail
6538 - fix NULL deref during bulk timeout (214)
6539 - fix problems related to multiple filesystems on one MDS (241)
6540 - fixed serious subtle metadata locking bugs
6541 - free locks on clients when inodes are removed due to memory
6543 - fix inode pointer in lock data (285)
6544 - partial support for multiple MDS on a single host (241)
6545 - data locks weren't cancelled at clear_inode time (290, 311)
6546 - intent locks could lead to unbounded lock growth (205)
6547 - added a maximum lock count, an LRU list, and a flusher
6548 - fix multiple rename (365)
6549 - properly abstracted the echo client
6550 - OSC locked 1 byte too many; fixed
6551 - rewrote brw callback code:
6552 - fixed recovery bugs related to LOVs (306)
6553 - fixed too-many-pages-in-one-write crash (191)
6554 - fixed (again) crash in sync_io_timeout (214)
6555 - probably fixed callback-related race (385)
6557 - Add capability to MDS protocol
6558 - LDLM cancellations and callbacks on different portals
6560 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6563 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6564 - defer open object destruction to close time (601981, 138)
6565 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6566 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6567 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6568 - don't call obd_setattr in ll_file_release for destroyed objects
6569 * protocol change to lustre_msg: move |version| and add |flags|
6570 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6571 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6572 * added replay of create, unlink, link and rename operations during
6573 MDS failover; recovery should be much more robust now
6574 * remove failed OSCs from LOVs (only lov_create uses this so far)
6575 * the lustre-HOWTO was brought (more) up to date (582544)
6577 2002-10-23 Phil Schwan <phil@clusterfs.com>
6580 - in-use dentries weren't being reused properly (617851)
6581 - prevent multiple LDLM setup (599178)
6582 - fix LOV size calculations for truncate (617853)
6583 - fix client handling of MDS intent errors (POSIX)
6584 - fix permission bug in lovstripe.c test (624321)
6585 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6586 - truncate past end of file could corrupt data
6587 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6588 - a race in recovery could return ETIMEDOUT to apps (623947)
6589 - building outside the source directory was fixed
6590 * the lustre-HOWTO was brought (more) up to date (582544)
6591 * major progress was made on recovery functionality
6593 2002-10-10 Phil Schwan <phil@clusterfs.com>
6596 - recovery deadlock fix
6597 - rm -rf causes LBUG fix (617817)
6598 - file open by multiple tasks fix (618962)
6599 - directory permissions bugs (602707 and 620007)
6600 - journal_stop fixed with locking (611313)
6601 - O_APPEND failures resolved (618273, perhaps 614459)
6602 - lconf PATH fix (619770)
6603 - IA64 build fix (621450)
6604 - RPC buffer sizes scale with amount of memory
6606 2002-10-01 Phil Schwan <phil@clusterfs.com>
6609 - locks would be cancelled without throwing away data pages,
6610 resulting in inconsistent data (605627)
6611 - inode attributes were not always being refreshed (605627, 612449)
6612 - lconf now continues to cleanup after lctl reports an error
6613 - MDS now enforces user permissions (602707)
6614 - lprocfs cleanup fixed, but not yet enabled (614157)
6615 - fixed infinite server hang, should a client not respond to an AST
6616 - avoid going into recovery if user calls readlink() with a buffer
6617 that's too small (613941)
6618 - AST RPCs no longer require replies (614867) -- this may be changed
6619 - don't crash server if client sends an IOV that's too big (611336)
6620 - fixed lock conversion deadlock (611892)
6621 - fixed the following of symlinks (614622)
6622 * recovery: the server can remove locks from a client that dies, other
6623 clients can make progress
6624 * more extN patch fixes
6625 * compile-time configurable ptlrpc buffer allocations
6627 - collaborative read cache document
6628 - Lustre Lite Performance CDR document-in-progress
6630 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6633 - fix typo in patch-2.4.18
6635 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6638 - clear ptlrpc request each time in handle_incoming_request()
6639 - unlink of files now destroys the object on the OST
6641 2002-09-19 Peter Braam <braam@clusterfs.com>
6643 * add hard link support
6644 * change obdfile creation method
6645 * kernel patch changed
6647 2002-09-19 Peter Braam <braam@clusterfs.com>
6650 - stack overflow bug in extN fixed
6652 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6654 * documentation updates
6655 - add man pages for config tools
6656 - update tests/README to describe testing with new config tools
6657 - finish metadata API descriptions
6658 * bug fixes and cleanups
6659 - statfs workaround for 16TB limit
6660 - LOV stripe allocation improved, can stripe on subset of OSTs
6661 - LOV file size/IO offset was wrong for files > 4GB in size
6662 - object EA data was being dropped, caused files to be unreadable
6663 - memory overflow with non-LOV OST caused memory corruption
6664 - fixed regression tests to work with new config tools, obdfilter
6665 - fixed bug when directory size became larger than 1 block
6666 - fixed bug (for single client case) when PWD was deleted
6667 - invalidate local directory pages when doing intent-based ops
6668 - avoid LDLM oops when lock callback contained bad data
6670 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6672 * documentation updates
6673 * bug fixes and cleanups
6674 - configuration tools
6677 - 64-bit compile warnings
6678 - 64-bit internal statfs data
6680 * test_brw on persistent OST devices
6682 * lprocfs (disabled)
6684 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6686 * documentation updates
6687 * bug fixes and cleanups
6688 * configuration tools
6690 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6693 * many small fixes to 0_5_4
6694 * io/network handling
6695 * thinkos in MDS operations
6697 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6700 * crucial basic fixes to 0.5.3
6701 * IOR, Iozone work over Elan
6704 2002-08-07 Phil Schwan <phil@clusterfs.com>
6705 * version 0_5_3, our first alpha
6706 * we use the new Portals iovs
6707 * documentation updates
6708 * bug fixes and cleanups
6709 * small changes in the DLM wire protocol
6711 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6712 * version 0_5_1 with some initial stability,
6713 * locking on MD and file I/O.
6714 * documentation updates
6715 * several bug fixes since 0.5.0
6716 * small changes in wire protocol
6718 2002-07-18 Phil Schwan <phil@clusterfs.com>
6720 * delivered as Lustre Light Alpha
6721 * fixed a crash after handling invalid MDS requests
6722 * fixed directory pages for architectures with non-4k pages sizes
6724 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6725 * release version v0_4_4
6726 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6728 * Fixes a number of interruption problems with OST operations.
6729 * Update documentation for portals header changes
6730 * Move all wire protocol structs/defines to lustre_idl.h
6731 * Fixes symlink length bug.
6732 * Add tcpdump to repository.
6734 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6735 * release version v0_4_3
6736 * Fixes statfs for inodes on extN.
6737 * Fixes bug in runtests which would delete /etc/hosts.
6738 * Use 64-bit object IDs wherever possible (not into VFS though)
6739 Remove ost_get_info, which is unused by lustre, and out of date.
6741 2002-07-03 Peter Braam <braam@clusterfs.com>
6742 * release version v0_4_2 Fixes a lookup error (type not passed)
6743 * move forward to head of Portals
6744 * move forward to latest Lustre kernel
6746 2002-06-25 Peter Braam <braam@clusterfs.com>
6747 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob