3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.9.1.el7 (RHEL7.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 4.18.0-80.11.2.el8 (RHEL8.0)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.100 (SLES12 SP3)
13 4.4.0-131 (Ubuntu 16.04)
14 4.15.0-32 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-1062.9.1.el7 (RHEL7.7)
19 4.12.14-95.29 (SLES12 SP4)
20 4.15.0-48 (Ubuntu 18.04)
21 * Other clients known to build on unpatched kernels at some point (others may also work):
22 3.10.0-862.14.4.el7 (RHEL7.5)
23 3.10.0-957.27.2.el7 (RHEL7.6)
24 4.14.0-49.13.1.el7a (RHEL7.5)
25 4.14.0-115.2.2.el7a (RHEL7.6)
26 4.18.0-80.11.2.el8 (RHEL8.0)
27 4.18.0-147.3.1.el8 (RHEL8.1)
28 4.4.120-92.70 (SLES12 SP2)
29 4.4.180-94.100 (SLES12 SP3)
30 4.4.0-131 (Ubuntu 16.04)
31 vanilla linux 4.15.0 (ZFS only)
32 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
33 * Recommended ZFS / SPL version: 0.7.13
34 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
35 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
36 stack overflow. For more information, please refer to bugzilla 17630.
37 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
38 the re-exporting nodes to be mounted with "32bitapi" mount option
39 * Note for Ubuntu 14.04.5:
40 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
41 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
43 --------------------------------------------------------------------------------
47 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
48 for currently supported client and server kernel versions.
49 * Server primary kernels built and tested during release cycle:
50 3.10.0-1062.1.1.el7 (RHEL7.7)
51 * Other server kernels known to build and work at some point (others may also work):
52 3.10.0-862.14.4.el7 (RHEL7.5)
53 3.10.0-957.27.2.el7 (RHEL7.6)
54 4.18.0-80.7.1.el8 (RHEL8.0)
55 4.4.120-92.70 (SLES12 SP2)
56 4.4.180-94.100 (SLES12 SP3)
57 4.4.0-131 (Ubuntu 16.04)
58 4.15.0-32 (Ubuntu 18.04)
59 vanilla linux 4.6.7 (ZFS only)
60 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
61 * Client primary unpatched kernels built and tested during release cycle:
62 3.10.0-1062.1.1.el7 (RHEL7.7)
63 4.12.14-95.29 (SLES12 SP4)
64 4.15.0-48 (Ubuntu 18.04)
65 * Other clients known to build on unpatched kernels at some point (others may also work):
66 3.10.0-862.14.4.el7 (RHEL7.5)
67 3.10.0-957.27.2.el7 (RHEL7.6)
68 4.14.0-49.13.1.el7a (RHEL7.5)
69 4.14.0-115.2.2.el7a (RHEL7.6)
70 4.18.0-80.7.1.el8 (RHEL8.0)
71 4.4.120-92.70 (SLES12 SP2)
72 4.4.180-94.100 (SLES12 SP3)
73 4.4.0-131 (Ubuntu 16.04)
74 vanilla linux 4.15.0 (ZFS only)
75 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
76 * Recommended ZFS / SPL version: 0.7.13
77 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
78 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
79 stack overflow. For more information, please refer to bugzilla 17630.
80 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
81 the re-exporting nodes to be mounted with "32bitapi" mount option
82 * Note for Ubuntu 14.04.5:
83 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
84 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
86 --------------------------------------------------------------------------------
90 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
91 for currently supported client and server kernel versions.
92 * Server known to build on the following kernels (others may also work):
93 2.6.32-696.18.7.el6 (RHEL6.9)
94 3.10.0-862.14.4.el7 (RHEL7.5)
95 3.10.0-957.el7 (RHEL7.6)
96 4.4.120-92.70 (SLES12 SP2)
97 4.4.162-94.69 (SLES12 SP3)
98 4.4.0-131 (Ubuntu 16.04)
99 4.15.0-32 (Ubuntu 18.04)
100 vanilla linux 4.6.7 (ZFS only)
101 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
102 * Client known to build on unpatched kernels (others may also work):
103 2.6.32-696.18.7.el6 (RHEL6.9)
104 2.6.32-754.6.3.el6 (RHEL6.10)
105 3.10.0-862.14.4.el7 (RHEL7.5)
106 3.10.0-957.el7 (RHEL7.6)
107 4.14.0-49.13.1.el7a (RHEL7.5)
108 4.14.0-115.2.2.el7a (RHEL7.6)
109 4.4.120-92.70 (SLES12 SP2)
110 4.4.162-94.69 (SLES12 SP3)
111 4.4.0-131 (Ubuntu 16.04)
112 4.15.0-32 (Ubuntu 18.04)
113 vanilla linux 4.15.0 (ZFS only)
114 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
115 * Recommended ZFS / SPL version: 0.7.9
116 * Tested with ZFS / SPL version: 0.7.9
117 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
118 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
119 stack overflow. For more information, please refer to bugzilla 17630.
120 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
121 the re-exporting nodes to be mounted with "32bitapi" mount option
122 * Note for Ubuntu 14.04.5:
123 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
124 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
126 --------------------------------------------------------------------------------
128 2018-04-03 Intel Corporation
130 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
131 for currently supported client and server kernel versions.
132 * Server known to build on patched kernels:
133 2.6.32-431.29.2.el6 (RHEL6.5)
134 2.6.32-504.30.3.el6 (RHEL6.6)
135 2.6.32-573.26.1.el6 (RHEL6.7)
136 2.6.32-642.15.1.el6 (RHEL6.8)
137 2.6.32-696.18.7.el6 (RHEL6.9)
138 3.10.0-693.21.1.el7 (RHEL7.4)
139 3.0.101-0.47.71 (SLES11 SP3)
140 3.0.101-107 (SLES11 SP4)
141 3.12.74-60.64.40 (SLES12 SP1)
142 4.4.114-92.64 (SLES12 SP2)
143 4.4.114-94.11 (SLES12 SP3)
144 3.13.0-101 (Ubuntu 14.04, ZFS only)
145 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
146 4.4.0-116 (Ubuntu 16.04)
147 vanilla linux 4.6.7 (ZFS only)
148 * Client known to build on unpatched kernels:
149 2.6.32-431.29.2.el6 (RHEL6.5)
150 2.6.32-504.30.3.el6 (RHEL6.6)
151 2.6.32-573.26.1.el6 (RHEL6.7)
152 2.6.32-642.15.1.el6 (RHEL6.8)
153 2.6.32-696.18.7.el6 (RHEL6.9)
154 3.10.0-693.21.1.el7 (RHEL7.4)
155 3.0.101-0.47.71 (SLES11 SP3)
156 3.0.101-107 (SLES11 SP4)
157 3.12.74-60.64.40 (SLES12 SP1)
158 4.4.114-92.64 (SLES12 SP2)
159 4.4.114-94.11 (SLES12 SP3)
160 3.13.0-101 (Ubuntu 14.04)
161 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
162 4.4.0-116 (Ubuntu 16.04)
164 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
165 * Recommended ZFS / SPL version: 0.7.6
166 * Tested with ZFS / SPL version: 0.7.6
167 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
168 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
169 stack overflow. For more information, please refer to bugzilla 17630.
170 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
171 the re-exporting nodes to be mounted with "32bitapi" mount option
172 * Note for Ubuntu 14.04.5:
173 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
174 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
176 --------------------------------------------------------------------------------
178 2017-07-05 Intel Corporation
180 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
181 for currently supported client and server kernel versions.
182 * Server known to build on patched kernels:
183 2.6.32-431.29.2.el6 (RHEL6.5)
184 2.6.32-504.30.3.el6 (RHEL6.6)
185 2.6.32-573.26.1.el6 (RHEL6.7)
186 2.6.32-642.15.1.el6 (RHEL6.8)
187 2.6.32-696.3.1.el6 (RHEL6.9)
188 3.10.0-514.21.1.el7 (RHEL7.3)
189 3.0.101-0.47.71 (SLES11 SP3)
190 3.0.101-100 (SLES11 SP4)
191 3.12.74-60.64.40 (SLES12 SP1)
192 4.4.59-92.17 (SLES12 SP2)
193 vanilla linux 4.6.7 (ZFS only)
194 * Client known to build on unpatched kernels:
195 2.6.32-431.29.2.el6 (RHEL6.5)
196 2.6.32-504.30.3.el6 (RHEL6.6)
197 2.6.32-573.26.1.el6 (RHEL6.7)
198 2.6.32-642.15.1.el6 (RHEL6.8)
199 2.6.32-696.3.1.el6 (RHEL6.9)
200 3.10.0-514.21.1.el7 (RHEL7.3)
201 3.0.101-0.47.71 (SLES11 SP3)
202 3.0.101-100 (SLES11 SP4)
203 3.12.74-60.64.40 (SLES12 SP1)
204 4.4.59-92.17 (SLES12 SP2)
206 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
207 * Recommended ZFS / SPL version: 0.6.5.9
208 * Tested with ZFS / SPL version: 0.6.5.9
209 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
210 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
211 stack overflow. For more information, please refer to bugzilla 17630.
212 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
213 the re-exporting nodes to be mounted with "32bitapi" mount option
215 --------------------------------------------------------------------------------
217 12-07-2016 Intel Corporation
219 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
220 for currently supported client and server kernel versions.
221 * Server known to build on patched kernels:
222 2.6.32-431.29.2.el6 (RHEL6.5)
223 2.6.32-504.30.3.el6 (RHEL6.6)
224 2.6.32-573.26.1.el6 (RHEL6.7)
225 2.6.32-642.11.1.el6 (RHEL6.8)
226 3.10.0-514.el7 (RHEL7.3)
227 3.0.101-0.47.71 (SLES11 SP3)
228 3.0.101-80 (SLES11 SP4)
229 3.12.62-60.64.8 (SLES12 SP1)
230 vanilla linux 4.5.7 (ZFS only)
231 * Client known to build on unpatched kernels:
232 2.6.32-431.29.2.el6 (RHEL6.5)
233 2.6.32-504.30.3.el6 (RHEL6.6)
234 2.6.32-573.26.1.el6 (RHEL6.7)
235 2.6.32-642.11.1.el6 (RHEL6.8)
236 3.10.0-514.el7 (RHEL7.3)
237 3.0.101-0.47.71 (SLES11 SP3)
238 3.0.101-80 (SLES11 SP4)
239 3.12.62-60.64.8 (SLES12 SP1)
241 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
242 * Recommended ZFS / SPL version: 0.6.5.8
243 * Tested with ZFS / SPL version: 0.6.5.8
244 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
245 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
246 stack overflow. For more information, please refer to bugzilla 17630.
247 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
248 the re-exporting nodes to be mounted with "32bitapi" mount option
250 --------------------------------------------------------------------------------
252 02-29-2016 Intel Corporation
254 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
255 for currently supported client and server kernel versions.
256 * Server known to build on patched kernels:
257 2.6.32-431.29.2.el6 (RHEL6.5)
258 2.6.32-504.30.3.el6 (RHEL6.6)
259 2.6.32-573.12.1.el6 (RHEL6.7)
260 3.10.0-327.3.1.el7 (RHEL7.2)
261 3.0.101-0.47.71 (SLES11 SP3)
262 3.0.101-68 (SLES11 SP4)
263 vanilla linux 4.2.1 (ZFS only)
264 * Client known to build on unpatched kernels:
265 2.6.32-431.29.2.el6 (RHEL6.5)
266 2.6.32-504.30.3.el6 (RHEL6.6)
267 2.6.32-573.12.1.el6 (RHEL6.7)
268 3.10.0-327.3.1.el7 (RHEL7.2)
269 3.0.101-0.47.71 (SLES11 SP3)
270 3.0.101-68 (SLES11 SP4)
273 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
274 * Recommended ZFS / SPL version: 0.6.4.2
275 * Tested with ZFS / SPL version: 0.6.4.2
276 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
277 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
278 stack overflow. For more information, please refer to bugzilla 17630.
279 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
280 the re-exporting nodes to be mounted with "32bitapi" mount option
282 --------------------------------------------------------------------------------
284 03-10-2015 Intel Corporation
286 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
287 for currently supported client and server kernel versions.
288 * Server known to build on patched kernels:
289 2.6.32-431.29.2.el6 (RHEL6.5)
290 2.6.32-504.8.1.el6 (RHEL6.6)
291 3.0.101-0.46 (SLES11 SP3)
292 * Client known to build on unpatched kernels:
293 2.6.32-431.29.2.el6 (RHEL6.5)
294 2.6.32-504.8.1.el6 (RHEL6.6)
295 3.10.0-123.20.1.el7 (RHEL7)
296 3.0.101-0.46 (SLES11 SP3)
297 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
298 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
299 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
300 stack overflow. For more information, please refer to bugzilla 17630.
301 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
302 the re-exporting nodes to be mounted with "32bitapi" mount option
304 Severity : enhancement
306 Description: control OST-index in IDIF via ROCOMPAT flag.
307 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
308 last_rcvd file. For new formatted OST device, it will be auto set;
309 for the case of upgrading from old OST device, you can enable it
310 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
311 enabled, for new created OST-object, its IDIF-in-LMA will contain
312 the OST-index; for the existing OST-object, the OSD will convert
313 old format IDIF as new format IDIF with OST-index stored in the
314 LMA EA when accessing such OST-object or via OI scrub. Once such
315 flag is enabled, it cannot be reverted back, so the system cannot
316 be downgraded to the orignal incompatible version.
318 --------------------------------------------------------------------------------
320 07-30-2014 Intel Corporation
322 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
323 for currently supported client and server kernel versions.
324 * Server known to build on patched kernels:
325 2.6.32-431.20.3.el6 (RHEL6.5)
326 3.0.101-0.31 (SLES11 SP3)
327 * Client known to build on unpatched kernels:
328 2.6.32-431.20.3.el6 (RHEL6.5)
329 3.0.101-0.31 (SLES11 SP3)
332 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
333 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
334 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
335 stack overflow. For more information, please refer to bugzilla 17630.
336 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
337 the re-exporting nodes to be mounted with "32bitapi" mount option
339 --------------------------------------------------------------------------------
341 10-31-2013 Intel Corporation
343 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
344 * Server builds for kernels:
345 2.6.32-358.18.1.el6 (RHEL6.4)
346 * Client builds for unpatched kernels:
347 2.6.32-358.18.1.el6 (RHEL6.4)
348 3.0.92-0.8 (SLES11 SP3)
350 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
351 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
352 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
353 stack overflow. For more information, please refer to bugzilla 17630.
354 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
355 the re-exporting nodes to be mounted with "32bitapi" mount option
357 --------------------------------------------------------------------------------
359 05-30-2013 Intel Corporation
361 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
362 * Server builds for kernels:
363 2.6.32-358.6.2.el6 (RHEL6.4)
364 * Client builds for unpatched kernels:
365 2.6.32-358.6.2.el6 (RHEL6.4)
366 3.0.74-0.6.6 (SLES11 SP2)
368 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
369 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
370 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
371 stack overflow. For more information, please refer to bugzilla 17630.
372 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
373 the re-exporting nodes to be mounted with "32bitapi" mount option
375 --------------------------------------------------------------------------------
377 9-30-2012 Whamcloud, Inc.
379 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
380 * Server support for kernels:
381 2.6.32-279.5.1.el6 (RHEL6)
382 * Client support for unpatched kernels:
383 2.6.32-279.5.1.el6 (RHEL6)
385 2.6.18-238.19.1.el5 (RHEL5)
386 2.6.16.60-0.69.1 (SLES10)
387 * Recommended e2fsprogs version: 1.41.90.wc4
388 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
389 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
390 stack overflow. For more information, please refer to bugzilla 17630.
391 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
392 the re-exporting nodes to be mounted with "32bitapi" mount option
394 Severity : enhancement
396 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
398 Severity : enhancement
400 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
402 Severity : enhancement
404 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
406 Severity : enhancement
408 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
410 Severity : enhancement
412 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
414 --------------------------------------------------------------------------------
416 04-30-2012 Whamcloud, Inc.
418 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
419 * Server support for kernels:
420 2.6.32-220.4.2.el6 (RHEL6)
421 * Client support for unpatched kernels:
422 2.6.32-220.4.2.el6 (RHEL6)
424 2.6.18-238.19.1.el5 (RHEL5)
425 2.6.16.60-0.69.1 (SLES10)
426 * Recommended e2fsprogs version: 1.41.90.wc4
427 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
428 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
429 stack overflow. For more information, please refer to bugzilla 17630.
430 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
431 the re-exporting nodes to be mounted with "32bitapi" mount option
433 Severity : enhancement
435 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
437 Severity : enhancement
439 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
441 Severity : enhancement
443 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
445 Severity : enhancement
447 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
449 --------------------------------------------------------------------------------
451 09-30-2011 Whamcloud, Inc.
453 * Server support for kernels:
454 2.6.18-238.19.1.el5 (RHEL5)
455 2.6.32-131.6.1.el6 (RHEL6)
456 * Client support for unpatched kernels:
457 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
458 2.6.18-238.19.1.el5 (RHEL5)
459 2.6.32-131.6.1.el6 (RHEL6)
460 2.6.29.4-167.fc11 (FC11)
461 2.6.16.60-0.69.1 (SLES10)
463 * Recommended e2fsprogs version: 1.41.90.wc3
464 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
465 Lustre filesystem with 4K stack may cause a stack overflow. For more
466 information, please refer to bugzilla 17630.
467 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
468 mounted with "32bitapi" mount option
470 Severity : enhancement
471 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
473 Severity : enhancement
474 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
476 Severity : enhancement
477 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
480 Frequency : only with SLES10
482 Description: Use OFED "KMP" provided by Novell
483 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
484 Lustre is now built against this package. That means you need to
485 install the ofed-kmp package from Novell for the patchless client.
486 Note that the ofed-kmp that Novell ships may not exactly match the
487 kernel version but should still be compatible.
489 Severity : enhancement
491 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
492 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
494 --------------------------------------------------------------------------------
496 2010-07-15 Oracle, Inc.
498 * Server support for kernels:
499 2.6.18-164.11.1.el5 (RHEL 5)
500 2.6.18-164.11.1.0.1.el5 (OEL 5)
501 * Client support for unpatched kernels:
502 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
503 2.6.18-164.11.1.el5 (RHEL 5),
504 2.6.18-164.11.1.0.1.el5 (OEL 5)
505 2.6.16.60-0.42.8 (SLES 10),
507 2.6.29.4-167.fc11 (FC11)
508 * Recommended e2fsprogs version: 1.41.10-sun2
510 Severity : enhancement
512 Description: Distribute lfs_migrate script with lustre
513 Details : The lfs_migrate mentioned in the manual is now included in the
514 lustre tree in lustre/scripts and will distributed in lustre RPMs.
516 Severity : enhancement
518 Description: Failover nid list added to import proc file.
520 Severity : enhancement
522 Description: "som_preview" mount option required to use SOM
523 Details : The clients will need to use som_preview mount option
527 Description: Lustre does not do 1MB IOs to HW RAID
528 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
530 Severity : enhancement
532 Description: lfs quota output is not very convenient for awk/sed-parsing
533 Details : Some positions in lfs quota output table could be empty or
534 non-empty which made it hard to parse it with scripts, now
535 a dash is put instead of space where there is not supposed
540 Description: Update to OFED 1.5.1
541 Details : OFED 1.5.1 will be used in environments that don't use
542 vendor supplied OFED.
546 Description: Remove lnet. prefix from lctl params display
547 Details : Also renamed lnet.memused to lnet_memused to
552 Description: Use conf_param -d to remove permanant settings.
556 Description: Add list_param -R
557 Details : lctl list_param now supports recursive mode.
560 Bugzilla : 21829, 22312
561 Description: Fix llobdstat.
562 Details : 1. Now updating the write_bytes counter
563 2. Port script fixes from 1.8.
567 Description: decrease the usage of memory on clients.
568 Details : 1. On clients, recycle dentries and inodes unused.
569 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
574 Description: llobdstat fix and enhancement
575 Details : add a counter to set a limit to how many samples will be returned
576 fix a wildcard in the path to limit to obdfilter stats only
580 Description: lfs check only allowed for root
581 Details : Code cleanups and added a regresion test of non-root use of
586 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
587 Details : Fix hr->hr_index overload.
591 Description: Fix FIEMAP functionality.
595 Description: FC11 patchless client support added
599 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
600 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
604 Description: New acc-sm script and various test-framework enhancments.
608 Description: Previously landed patch was reverted.
612 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
616 Description: Add ldiskfs maxdirsize mount option.
620 Description: Fix mkfs.lustre support for 16TB LUNs.
624 Description: Print mmp_check_interval
625 Details : print mmp_check_interval and make it possible to abort
626 mount operation in case it takes too long.
630 Description: Add cascading_rw to lustre mpi tests.
634 Description: Store SOM data in EA.
635 Details : Also several SOM related cleanups.
639 Description: Add refcount to md_open_data
643 Description: Use refcount on lov_request_set to prevent memory corruption.
647 Description: Statahead: use parent dir's i_mutex for synchronization.
651 Description: loadgen improvements.
655 Description: Enable peer health detection in ksocklnd and ko2iblnd.
659 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
663 Description: Fix race on obd_devcie:md_stats
667 Description: Fix SLES 11 client builds.
668 Details : Resolved uninitalized pointer gcc warnings.
672 Description: Allow non-root access for "lfs check".
673 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
677 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
679 Severity : enhancement
681 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
682 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
686 Description: Sanity test_27p fails.
687 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
688 Sanity tests 27 were cleaned up and improved.
692 Description: Unify req format on client/servers
693 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
694 buffers and swabbers approach..
699 Description: drop join file stuff.
701 Severity : enhancement
703 Description: Add 2.6.27 and SLES11 (patchless client) support
707 Description: Fix in ptlrpc_expire_one_request() to print the signed time
710 Severity : enhancement
712 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
715 Severity : enhancement
717 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
719 Severity : enhancement
721 Description: correctly handle big reply message.
722 Details : send LNet event if reply is bigger than buffer and adjust this buffer
727 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
729 Severity : enhancement
731 Description: Add support for OEL5.
735 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
737 Severity : enhancement
739 Description: "weak-modules" support
740 Details : Implement "weak-modules" support which enables kernel modules
741 to be used with any kernel that implements the same kABI. In
742 order to achieve this modules are now installed in
743 /lib/modules/$(uname -r)/updates/kernel on all distributions.
747 Description: client could not reconnect to OST because of an active request.
748 Details : The client now retries to reconnect to the same server, if a
749 connect request failed with EBUSY or -EAGAIN. Test case 35b
750 conf-sanity.sh has been created.
754 Description: Don't allow make backward step on assignin osc next id.
755 Details : race between allocation next id and ll_sync thread can be cause
756 of set wrong osc next id and can be kill valid ost objects.
760 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
764 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
768 Description: Error handling in osc_statfs_interpret() has been improved.
769 Details : Check in osc_statfs_interpret() for EBADR.
773 Description: Do not send statfs() requests to OSTs disabled by administrator.
774 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
776 Severity : enhancement
778 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
780 Severity : enhancement
782 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
784 Severity : enhancement
786 Description: Enhancement for lfs(1) command to use numeric uid/gid.
790 Description: Improved cleanup procedure for conf-sanity test_46a.
794 Description: Do not start bulk transfer, if client tried to reconnect.
795 Details : Do not start bulk transfer, if client tried to reconnect,
796 but failed due to active rpcs. Check for exp_abort_active_req
797 has been implemented.
801 Description: Prevent inconsistences between linux and lustre mount structures.
802 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
803 Make the sleep interruptible.
805 Severity : enhancement
807 Description: provide server to client comms path
808 Details : server to client communications path via new LDLM_SET_INFO rpc
812 Description: truncate starts GFP_FS allocation under transaction causing deadlock
813 Details : ldiskfs_truncate calls grab_cache_page which may start page
814 allocation under an open transaction. This may lead to
815 calling prune_icache with consequent lustre reentrance.
819 Description: send correctly lsm on open replay
820 Details : MDS is trust to LSM size on replay open, but client can set wrong size
825 Description: enable client interface failover
826 Details : When a child reconnects from another NID, properly update export
827 nid hash position and ldlm reverse import.
829 Severity : enhancement
831 Description: Communicate OST degraded/readonly state via statfs to MDS
832 Details : Flags in the statfs returned from OSTs indicate whether the
833 OST is in a degraded RAID state, or if the filesystem has
834 turned read-only after a filesystem error is detected.
839 Description: don't panic if EPROTO was hit when reading symlink
840 Details : correctly handling request reference in error cases.
842 Severity : enhancement
844 Description: Add LustreNetLink, a kernel-userspace communcation path.
846 Severity : enhancement
848 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
853 Description: access to llog context before init.
854 Details : move handling CATALOGS file at osc layer and forbid access to llog
858 Frequency : rare, only when client is writing to OST on local node
860 Description: Avoid deadlock for local client writes
861 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
862 memory freeing context. This allows OST threads to set the
863 PF_MEMALLOC flag on task structures in order to allocate memory
864 from reserved pools and complete IO.
865 Use GFP_HIGHUSER for OST allocations for non-local client writes,
866 so that the OST threads generate memory pressure and allow
867 inactive pages to be reclaimed.
869 Severity : enhancement
871 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
876 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
877 Details : move ldlm namespace creation in setup phase to avoid grab
878 _lprocfs_lock with cli_sem held.
882 Description: Temporarily disable grant shrink.
883 Details : Disable the feature for debugging.
886 Frequency : only when formatting test filesystems
888 Description: Unable to run several mkfs.lustre on loop devices at the same time
889 Details : mkfs.lustre returns error 256 on the concurrent loop devices
890 formatting. The solution is to proper handle the error.
892 Severity : enhancement
894 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
896 Severity : enhancement
898 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
903 Description: resolve race between obd_disconnect and class_disconnect_exports
904 Details : if obd_disconnect will be called to already disconnected export he
905 forget release one reference and osc module can't unloaded.
907 Severity : enhancement
909 Description: Allow tuning service thread via /proc
910 Details : For each service a new
911 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
912 created that can be used to set min/max thread counts, and get the
913 current number of running threads.
918 Descriptoin: don't return error if have partially created objects for file.
919 Details : lov_update_create_set uses set->set_success as index for created
920 objects, so if some requests will be failed, they will have hole
921 at end of array and qos_shrink_lsm for allocate correct lsm.
923 Severity : enhancement
925 Description: Update OFED support to 1.4.1
930 Description: Reduce small size read RPC
931 Details : Set read-ahead limit for every file and only do read-ahead when
932 available read-ahead pages are bigger than 1M to avoid small size
935 Severity : enhancement
937 Description: move AT tunable parameters for more consistent usage
938 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
940 Severity : enhancement
942 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
943 Details : allow skip disconnected ost for send statfs request and hide error
949 Description: fix racy locking of mballoc block bitmaps causing BUG
950 Details : The locking of the mballoc buddy bitmap and the in-memory
951 block bitmap was using two different spin locks in some
952 cases. This made it possible to incorrectly access the
953 mballoc bitmap while another process was modifying it,
954 causing a sanity assertion to fail. While no on-disk corruption
955 was reported, there was some risk of this happening.
958 Frequency : rare, on llog test 6
960 Descriptoin: don't allow connect to already connected import
961 Details : allowing connect to already connected import is hide connecting problem.
966 Descriptoin: allow kill process which wait statahead result
967 Details : for some reasons 'ls' can stick in waiting result from statahead,
968 in this case need way for kill this process.
970 Severity : enhancement
972 Description: Add state history info file, enhance import info file
973 Details : Track import connection state changes in a new osc/mdc proc file;
974 add overview-type data to the osc/mdc import proc file.
976 Severity : enhancement
978 Description: MDS create should not wait for statfs RPC while holding DLM lock.
980 Severity : enhancement
982 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
987 Descriptoin: OST grant too much space to client even there are not enough space.
988 Details : Client will shrink its grant cache to OST if there are no write
989 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
990 this grant cache if there are already not enough avaible space
991 (left_space < total_clients * 32M).
994 Frequency : start MDS on uncleanly shutdowned MDS device
996 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
997 Details : stay in waiting mds<>ost recovery finished produce random bugs
998 due race between two ll_sync thread for one lov target. send
999 ACTIVATE event only if connect realy finished and import have
1003 Frequency : rare, connect and disconnect target at same time
1005 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1006 Details : don't call obd_disconnect under lov_lock. this long time
1007 operation and can block ptlrpcd which answer to connect request.
1012 Descriptoin: don't lose wakeup for imp_recovery_waitq
1013 Details : recover_import_no_retry or invalidate_import and import_close can
1014 both sleep on imp_recovery_waitq, but we was send only one wakeup
1018 Frequency : always with long access acl
1020 Descriptoin: mds can't pack reply with long acl.
1021 Details : mds don't control size of acl but they limited by reint/getattr
1024 Severity : enhancement
1026 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1028 Severity : enhancement
1030 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1033 Frequency : start MDS on uncleanly shutdowned MDS device
1035 Descriptoin: aborting recovery hang on MDS
1036 Details : don't throttle destroy RPCs for the MDT.
1039 Frequency : on remount
1041 Description: external journal device not working after the remount
1042 Details : clear dev_rdonly flag for external journal devices in
1048 Description: shutdown vs evict race
1049 Details : client_disconnect_export vs connect request race.
1050 if client will evicted at this time - we start invalidate
1051 thread without referece to import and import can be freed
1055 Frequency : rare, need acl's on inode.
1057 Description: client can't handle ost additional correctly
1058 Details : if ost was added after client connected to mds client can have
1059 hit lnet_try_match_md ... to big messages to wide striped files.
1060 in this case need teach client to handle config events about add
1061 lov target and update client max ea size at that event.
1063 Severity : enhancement
1065 Description: Changelogs
1066 Details : Changelogs are a lightweight mechanism to track filesystem
1067 metadata and namespace changes. The changelog is recorded
1068 permanently on the MDTs, and is periodically "consumed" / purged
1069 when records are no longer needed.
1071 Severity : enhancement
1073 Description: compact fld format with extents
1074 Details : Store range of seq rather than every seq in FLD. Seq
1075 controller update FLD rather than clients. In Case of CMD, mdt0
1076 has FLD, all other metadata server act as non persistent proxy
1077 for FLD queries and cache fld entries in fld cache.
1082 Description: don't skip ost target if they assigned to file
1083 Details : Drop slow OSCs if we can, but not for requested start idx.
1084 This means "if OSC is slow and it is not the requested
1085 start OST, then it can be skipped, otherwise skip it only
1086 if it is inactive/recovering/out-of-space.
1090 Description: more cleanup in mds_lov
1091 Details : not send LOV EA under replay, we can't know about they size at this
1092 time. Don't allow client connect to mds before any ost connected,
1093 for avoid problems with LOV EA size and returning EIO to client.
1095 Severity : enhancement
1097 Description: Interoperability at server side (Disk interoperability)
1099 Severity : enhancement
1101 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1103 Severity : enhancement
1105 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1107 Severity : enhancement
1109 Description: New client IO stack (CLIO).
1111 Severity : enhancement
1113 Description: Commit on sharing. Eliminate inter-client dependencies between
1114 uncommitted transactions by doing transaction commits.
1115 Thereby clients may recovery independently.
1118 Frequency : Create a symlink file with a very long name
1120 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1121 Details : If there is no extra space in the request for early cancels,
1122 ldlm_req_handles_avail() returns 0 instead of a negative value.
1124 Severity : enhancement
1126 Description: Add /proc entry for import status
1127 Details : The mdc, osc, and mgc import directories now have
1128 an import directory that contains useful import data for debugging
1129 connection problems.
1131 Severity : enhancement
1133 Description: Re-disable certain /proc logging
1134 Details : Enable and disable client's offset_stats, extents_stats and
1135 extents_stats_per_process stats logging on the fly.
1138 Frequency : Only on FC kernels 2.6.22+
1140 Description: oops in statahead
1141 Details : Do not drop reference count for the dentry from VFS when lookup,
1142 VFS will do that by itself.
1144 Severity : enhancement
1146 Description: Generic /proc file permissions
1147 Details : Set /Proc file permissions in a more generic way to enable non-
1148 root users operate on some /proc files.
1152 Description: Hitting mdc_commit_close() ASSERTION
1153 Details : Properly handle request reference release in
1154 ll_release_openhandle().
1158 Description: quota recovery deadlock during mds failover
1159 Details : This patch includes att18982, att18236, att18237 in bz14840.
1161 1. fix osts hang when mds does failover with quotaon
1162 2. prevent watchdog storm when osts threads wait for the
1167 Frequency : only patchless client
1168 Description: add workaround for race between add/remove dentry from hash
1170 Severity : enhancement
1172 Description: Allow OST glimpses to return PW locks
1176 Description: LBUG when llog conf file is full
1177 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1181 Description: Prevent import from entering FULL state when server in recovery
1185 Description: service mount cannot take device name with ":"
1186 Details : Only when device name contains ":/" will mount treat it as
1192 Description: replace ptlrpcd with the statahead thread to interpret the async
1193 statahead RPC callback
1197 Frequency : on recovery
1198 Description: I/O failures after umount during fail back
1199 Details : if client reconnected to restarted server we need join to recovery
1200 instead of find server handler is changed and process self
1201 eviction with cancel all locks.
1203 Severity : enhancement
1205 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1207 Severity : enhancement
1209 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1211 Severity : enhancement
1213 Description: Upcall on Lustre log has been dumped
1214 Details : Allow for a user mode script to be called once a Lustre log has
1215 been dumped. It passes the filename of the dumped log to the
1216 script, the location of the script can be specified via
1217 /proc/sys/lnet/debug_log_upcall.
1222 Description: avoid idr_remove called for id which is not allocated.
1223 Details : Move assigment s_dev for clustered nfs to end of initialization,
1224 to avoid problem with error handling.
1229 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1230 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1231 into CONN_USED_HASH and this prodice warning when put connection
1232 again in unused hash.
1235 Severity : enhancement
1237 Description: File striping can now be set to use an arbitrary pool of OSTs.
1239 Severity : enhancement
1241 Description: Export bytes_read/bytes_write count on OSC/OST.
1245 Description: Early reply size mismatch, MGC loses connection
1246 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1247 the connect flags are properly negotiated.
1252 Description: quotas are not honored with O_DIRECT
1253 Details : all writes with the flag O_DIRECT will use grants which leads to
1254 this problem. Now using OBD_BRW_SYNC to guard this.
1258 Description: add quota statistics
1259 Details : 1. sort out quota proc entries and proc code.
1260 2. add quota statistics
1262 Severity : enhancement
1264 Description: enable quota support for HEAD.
1268 Description: Properly propagate oinfo flags from lov to osc for statfs
1269 Details : restore missing copy oi_flags to lov requests.
1271 Severity : enhancement
1273 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1274 : l_getgroups(8), lst(8), routerstat(8)
1275 Details : included man pages for llobdstat(8), llstat(8),
1276 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1278 Severity : enhancement
1280 Description: configure's --enable-quota should check the
1281 : kernel .config for CONFIG_QUOTA
1282 Details : configure is terminated if --enable-quota is passed but
1283 : no quota support is in kernel
1287 Description: Remove portals compatibility
1288 Details : Remove portals compatibility, not interoperable with releases
1293 Description: Resolve device initialization race
1294 Details : Prevent proc handler from accessing devices added to the
1295 obd_devs array but yet be intialized.
1297 Severity : enhancement
1299 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1301 Severity : enhancement
1303 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1308 Description: Using wrong pointer in osc_brw_prep_request
1309 Details : Access to array[-1] can produce panic if kernel compiled with
1310 CONFIG_PAGE_ALLOC enabled
1314 Description: Client runs out of low memory
1315 Details : Consider only lowmem when counting initial number of llap pages
1319 Description: *optional* service tags registration
1320 Details : if the "service tags" package is installed on a Lustre node
1321 When the filesystem is mounted, a local-node service tag will
1322 be created. See http://inventory.sun.com/ for more information
1323 about the Service Tags asset management system.
1327 Description: Kernel BUG tries to release flock
1328 Details : Lustre does not destroy flock lock before last reference goes
1329 away. So always drop flock locks when client is evicted and
1330 perform unlock regardless of successfulness of speaking to MDS.
1334 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1338 Description: sanity test 65a fails if stripecount of -1 is set
1339 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1344 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1345 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1348 Severity : enhancement
1350 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1354 Description: do not process already freed flock
1355 Details : flock can possibly be freed by another thread before it reaches
1356 to ldlm_flock_completion_ast.
1360 Description: LBUG during stress test
1361 Details : Need properly lock accesses the flock deadlock detection list.
1365 Description: oops in page fault handler
1366 Details : kernel page fault handler can return two special 'pages' in error
1367 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1371 Description: timeout with invalidate import.
1372 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1373 should be handled by ptlrpcd. This produce long age waiting and
1374 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1376 Severity : enhancement
1378 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1382 Description: enable MGS and MDT services start separately
1383 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1384 a co-located MGS without starting the MGS, which is a complement
1385 to 'nosvc' mount option.
1389 Frequency : after recovery
1390 Description: precreate to many object's after del orphan.
1391 Details : del orphan st in oscc last_id == next_id and this triger growing
1392 count of precreated objects. Set flag LOW to skip increase count
1393 of precreated objects.
1397 Frequency : rare, on clear nid stats
1398 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1399 Details : when clean nid stats sometimes try destroy live entry,
1400 and this produce panic in free.
1404 Description: Stack overflow during MDS log replay
1405 ease stack pressure by using a thread dealing llog_process.
1409 Description: wait until IO finished before start new when do lock cancel.
1410 Details : VM protocol want old IO finished before start new, in this case
1411 need wait until PG_writeback is cleared until check dirty flag and
1412 call writepages in lock cancel callback.
1414 Severity : enhancement
1416 Description: using special macro for print time and cleanup in includes.
1420 Description: mds_mfd_close() ASSERTION(rc == 0)
1421 Details : In mds_mfd_close(), we need protect inode's writecount change
1422 within its orphan write semaphore to prevent possible races.
1426 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1430 Frequency : rare, on shutdown ost
1431 Description: don't hit live lock with umount ost.
1432 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1433 use shrink_dcache_sb instead.
1437 Description: don't panic with use echo client
1438 Details : echo client pass NULL as client nid pointer and this produce null
1439 pointer dereference.
1443 Description: fix build on ppc32
1444 Details : compile code with -m64 flag produce wrong object file for ppc32.
1448 Description: add message levels for liblustreapi
1452 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1453 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1454 will result in an -EAGAIN and in the exit call path we call
1455 alloc_rr() which will with increasing aggressiveness attempt to
1456 aquire precreated objects on the minimum number of required OSCs.
1460 Description: Use old size assignment to avoid deadlock
1461 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1462 the scheduling while holding a spinlock. We do not need locking
1463 for size in ll_update_inode() because size is only updated from
1464 the MDS for directories or files without objects, so there is no
1465 other place to do the update, and concurrent access to such inodes
1466 are protected by the inode lock.
1470 Description: resolve "_IOWR redefined" build error on SLES10
1474 Description: dump the memory debugging after all modules are unloaded to
1475 suppress false negative in conf_sanity test 39
1477 Severity : enhancement
1479 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1484 Description: fix SLES kernel versioning
1485 Details : the kernel version for our SLES 10 kernel did not include a "-"
1486 before the "smp" at the end. while this was not a problem in
1487 general, it did mean that software trying to use the kernel
1488 version to try to detect a vendor specific kernel would fail.
1489 this was most evident by the OFED build scripts.
1493 Description: Don't update lov_desc members until making sure they are valid
1494 Details : When updating lov_desc members via proc fs, need fix their
1495 validities before doing the real update.
1499 Description: don't put request into delay list while invalidate in flight.
1500 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1501 import in flight. this produce timeout for invalidate and sometimes
1502 can cause stale data.
1506 Frequency : on ppc only
1507 Description: not convert ost objects for directory because it's not exist.
1508 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1512 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1514 Severity : enhancement
1516 Description: Update to support 2.6.22.14 vanilla kernel.
1520 Frequency : rare, on recovery
1521 Description: read procfs can produce deadlock in some situation
1522 Details : Holding lprocfs lock which send rpc can produce block for destroy
1523 obd objects and this also block reconnect with -EALREADY. This
1524 isn't fix all lprocfs bugs - but make it rare.
1526 Severity : enhancement
1528 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1531 Frequency : frequent on X2 node
1533 Description: mdc_set_open_replay_data LBUG
1534 Details : Set replay data for requests that are eligible for replay.
1538 Description: lustre_mgs: operation 101 on unconnected MGS
1539 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1540 MGC, and late on MGC cannot successfully connect to MGS and a lot
1541 of the error messages complaining that MGS is not connected.
1545 Frequency : on network error
1546 Description: panic with double free request if network error
1547 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1548 but it's true only for synchronus enqueue, for async enqueue
1549 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1552 Severity : enhancement
1554 Description: client-side metadata stat-ahead during readdir(directory readahead)
1555 Details : perform client-side metadata stat-ahead when the client detects
1556 readdir and sequential stat of dir entries therein
1559 Frequency : on start mds
1561 Description: Implement get_info(last_id) in obdfilter.
1564 Frequency : occasional
1566 Description: Correctly check stale fid, not start epoch if ost not support SOM
1567 Details : open with flag O_CREATE need set old fid in op_fid3 because
1568 op_fid2 was overwritten with new generated fid, but mds can answer
1569 with one of these two fids and both is not stale. Setattr
1570 incorrectly started an epoch and assume will be called
1571 done_writing, but without SOM done_writing ever being called.
1574 Frequency : rare, depends on device drivers and load
1576 Description: MDS or OSS nodes crash due to stack overflow
1577 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1578 In some cases, in conjunction with device drivers that use a lot
1579 of stack the MDS (or possibly OSS) service threads could overflow
1580 the stack. One change which was identified to consume additional
1581 stack has been reworked to avoid the extra stack usage.
1584 Frequency : occasional
1586 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1587 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1588 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1589 oscc_has_objects and see OSC has no objests and call
1590 oscc_internal_create to resend create request.
1592 Severity : enhancement
1594 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1596 Severity : enhancement
1598 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1601 Frequency : very rare
1603 Description: Fix replay if there is an un-replied request and open
1604 Details : In some cases, older replay request will revert the
1605 mcd->mcd_last_xid on MDS which is used to record the client's
1606 latest sent request.
1608 Severity : enhancement
1610 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1612 Severity : enhancement
1614 Description: Add rhel5 support to HEAD.
1616 Serverity : enhancement
1618 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1623 Description: Don't allow skipping OSTs if index has been specified.
1624 Details : Don't allow skipping OSTs if index has been specified, make
1625 locking in internal create lots better.
1629 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1630 Details : access to bitfield in structure is always rounded to long
1631 and this produce problem with not atomic change any bit.
1635 Description: Lustre make rpms failed.
1636 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1637 builds Lustre rpms from tarball.
1640 Frequency : rare on shutdown ost
1642 Description: If llog cancel was not send before clean_exports phase, this can
1643 produce deadlock in llog code.
1644 Details : If llog thread has last reference to obd and call class_import_put
1645 this produce deadlock because llog_cleanup_commit_master wait when
1646 last llog_commit_thread exited, but this never success because was
1647 called from llog_commit_thread.
1651 Description: allow userland application know is lost one of stripes.
1652 Details : fill lvb_blocks with error code on ost and return it to
1653 application if error flag found.
1657 Description: NULL lov_tgts causing MDS oops
1658 Details : more safe checks for NULL lov_tgts for avoid oops.
1660 Severity : enhancement
1662 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1666 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1667 Details : If an OST is down the MDS will hang indefinitely in
1668 obd_statfs_rqset() waiting for the statfs data. While for
1669 MDS QOS usage of statfs, it should not stuck in waiting.
1671 Severity : enhancement
1673 Description: remote_acl support
1674 Details : Support ACL-based permission check for remote user.
1675 Support setfacl/getfacl for remote user with the utils
1676 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1677 the system "{s,g}etfacl" utils.
1679 Severity : enhancement
1681 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1683 Severity : enhancement
1685 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1689 Description: make mgs_setparam() handle fsname containing dash
1690 Details : fsname containing a dash does not work with lctl conf_param
1692 Severity : enhancement
1694 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1696 Severity : enhancement
1698 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1702 Frequency : rare, at shutdown
1703 Description: access already free / zero obd_namespace.
1704 Details : if client_disconnect_export was called without force flag set,
1705 and exist connect request in flight, this can produce access to
1706 NULL pointer (or already free pointer) when connect_interpret
1707 store ocd flags in obd_namespace.
1711 Frequency : only at startup
1712 Description: not alloc memory with spinlock held.
1713 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1714 if any spinlock held.
1716 Severity : enhancement
1718 Description: make lustre randomly fail allocating memory
1719 Details : Make lustre randomly failed allocating memory for testing purpose.
1721 Severity : enhancement
1723 Description: lost problems with lov objid file
1724 Details : Fixes some scability and access to not inited memory problems
1725 in work with lov objdid file.
1730 Description: lfs find does not continue on file error
1731 Details : Continue other files processing when a file/dir is absent.
1735 Description: Inconsistent usage of lustre_pack_reply()
1736 Details : Standardize the usage of lustre_pack_reply() such that it
1737 always generate a CERROR on failure.
1740 Frequency : occasional
1742 Description: MDS hang or stay in waiting lock
1743 Details : If client receive lock with CBPENDING flag ldlm need send lock
1744 cancel as separate rpc, to avoid situation when cancel request
1745 can't processed due all i/o threads stay in wait lock.
1749 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1753 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1757 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1761 Description: Update to RHEL4 latest kernel.
1763 Severity : enhancement
1765 Description: Build SLES10 patchless client fails
1766 Details : The configure was broken by run ./configure with
1767 --with-linux-obj=.... argument for patchless client. When the
1768 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1769 can't search header adequately. Use absolute path such as
1770 -I($LINUX)/include instead.
1774 Description: interrupt oig_wait produce painc on resend.
1775 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1776 private set, and this produce situation when rq_ptlrpcd_data not
1777 copyed to new allocated request and triggered LBUG on assert
1778 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1779 ptlrpcd set if request is changed and can be safety changed to use
1784 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1785 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1786 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1787 these macros and replace them with their difinitions instead.
1791 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1792 Details : Update which_patch & target file for SLES10 latest kernel.
1795 Description: Remove client patches from SLES 10 kernel.
1796 Details : This causes SLES 10 clients to behave as patchless clients
1797 even on a Lustre-patched (server) kernel.
1799 Severity : enhancement
1801 Description: self-adjustable client's lru lists
1802 Details : use adaptive algorithm for managing client cached locks lru
1803 lists according to current server load, other client's work
1804 pattern, memory activities, etc. Both, server and client
1805 side namespaces provide number of proc tunables for controlling
1808 Severity : enhancement
1810 Description: light-weight GSS support
1811 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1812 and reduce performance overhead.
1814 Severity : enhancement
1816 Description: Linux keyring support
1817 Details : Support using service of Linux keyring for Lustre GSS internal
1818 context refresh/cache mechanism.
1822 Description: Fix errors in lfs documentation
1823 Details : Fixes man pages
1827 Description: don't use GFP_* in generic Lustre code.
1828 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1832 Description: obdclass is limited by single OBD_ALLOC(idarray)
1833 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1837 Description: warnings with build patchless client with vanila 2.6.19 and up
1838 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1839 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1843 Frequency : Only for SLES
1844 Description: sanity_quota fail test_1
1845 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1846 file, which makes TSTID[2] inunique.
1850 Frequency : Only for SLES9
1851 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1852 Details : Update target/ChangeLog/which_patch .
1856 Frequency : Only for test_10 in sanity-quota.sh
1857 Description: a bug in quota test code
1858 Details : checking a test flag in a wrong place
1862 Frequency : Only for test_13 in sanity-quota.sh
1863 Description: a bug in quota test script
1864 Details : shouldn't check overfull allocated quota
1868 Description: jbd statistics
1869 Details : Port older jbd statistics patch for sles10
1871 Severity : enhancement
1873 Description: Add printing inode info into message about error in writepage.
1878 Description: reply_lock_interpret crash due to race with it and lock cancel.
1879 Details : Do not replay locks that are being cancelled. Do not reference
1880 locks by their address during replay, just by their handle.
1884 Frequency : When flocks are used.
1885 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1886 Details : Flock locks might destroy just granted lock if it could be merged
1887 with another existing flock, this is done in completion handler,
1888 so teach ldlm_cli_enquque_fini that this is a valid case for
1894 Description: Oops in read and write path when failing to allocate lock.
1895 Details : Check if lock allocation failed and return error back.
1899 Description: lstripe command fails for valid OST index
1900 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1901 instead of lov->desc.ld_active_tgt_count.
1905 Description: sanity.sh failed test 103
1906 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1911 Description: df doesn't work properly if diskfs blocksize != 4K
1912 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1916 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1917 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1919 Severity : enhancement
1921 Description: Adaptive timeouts
1922 Details : RPC timeouts adapt to changing server load and network
1923 conditions to reduce resend attempts and improve recovery time.
1927 Description: llapi_file_create() does not allow some changes
1928 Details : add llapi_file_open() that allows specifying the mode and
1929 open flags, and also returns an open file handle.
1933 Description: merge and cleanup kernel patches.
1934 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1938 Description: Add journal checksum support.(Kernel part)
1939 Details : The journal checksum feature adds two new flags i.e
1940 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1941 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1942 indicates that the commit block contains the checksum for
1943 the blocks described by the descriptor blocks. Now commit
1944 record can be sent to disk without waiting for descriptor
1945 blocks to be written to disk. This behavior is controlled
1946 using JBD2_FEATURE_ASYNC_COMMIT flag.
1950 Description: OSS needs mutliple precreate threads
1951 Details : Add ability to start more than one create thread per OSS.
1955 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1956 Details : execute lfs setstripe on client
1960 Description: mds_obd_create error creating tmp object
1961 Details : When the user sets quota on root, llog will be affected and can't
1962 create files and write files.
1965 Frequency : Always on ia64 patchless client, and possibly others.
1967 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1968 Details : This allows the patchless client to be loaded on architectures
1969 without this export.
1973 Description: RedHat Update kernel for RHEL5
1974 Details : Add the kernel config file for RHEL5.
1978 Description: RedHat Update kernel for RHEL5
1979 Details : Modify the kernel config file more closer RHEL5.
1983 Description: Build failure against Centos5 (RHEL5)
1984 Details : Define PAGE_SIZE when it isn't present.
1988 Description: test_7 on sanity-quota.sh can't be used on separate machines
1989 Details : add facet to handle it
1993 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1994 Details : with NFS, the anon dentry's parent was set to itself in
1995 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1996 corresponding dentry other than use rec->ur_name.
1998 Severity : enhancement
2000 Description: lfs setstripe enhancement
2001 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2005 Description: enable data checksumming by default
2006 Details : enable checksum by default, allow --disable-checksum
2007 configure option and "-o nochecksum" mount option. Checksums
2008 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2012 Description: lustre support for RHEL5
2013 Details : Add support for RHEL5.
2017 Description: Client eviction due to failover config
2018 Details : after a connection loss, the lustre client should attempt to
2019 reconnect to the last active server first before trying the
2020 other potential connections.
2024 Description: when mds and osts use different quota unit(32bit and 64bit),
2025 quota will be released repeatly.
2026 Details : void sending multiple quota reqs to mds, which will keep the
2027 status between the reqs.
2031 Description: osts not allocated evenly to files
2032 Details : change the condition to increase offset_idx
2035 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2037 Description: Data corruption for OSTs that are formatted larger than 2TB
2039 Details : When generating the bio request for lustre file writes the
2040 sector number would overflow a temporary variable before being
2041 used for the IO. The data reads correctly from Lustre (which
2042 will overflow in a similar manner) but other file data or
2043 filesystem metadata may be corrupted in some cases.
2047 Description: Tune the kernel for good SCSI performance.
2048 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2049 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2054 Description: rewrite ext2-derived code in obdclass/uuid.c
2055 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2056 from scratch preserving functionality.
2060 Description: "lfs find -obd UUID" prints directories
2061 Details : "lfs find -obd UUID" will return all directory names instead
2062 of just file names. It is incorrect because the directories
2063 do not reside on the OSTs.
2067 Description: use i_size_read and i_size_write in 2.6 port
2068 Details : replace inode->i_size access with i_size_read/write()
2071 Frequency : when removing large files
2073 Description: scheduling issue during removal of large Lustre files
2074 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2075 It causes scheduling issues when removing large files (17TB in the
2079 Frequency : only with liblustre clients on XT3
2081 Description: evictions taking too long
2082 Details : allow llrd to evict clients directly on OSTs
2085 Frequency : only on ppc
2087 Description: /proc/fs/lustre/devices broken on ppc
2088 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2089 We should make sure the type of 'index' is loff_t and then cast
2090 explicitly as needed below. Do not assign an explicitly cast
2094 Frequency : only for rhel5
2096 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2097 Details : Modify the target file & which_kernel.
2099 Severity : enhancement
2101 Description: omit set fsid for export NFS
2102 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2103 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2107 Frequency : Always, for kernels after 2.6.16
2108 Description: Fix warning idr_remove called for id=.. which is not allocated.
2109 Details : Last kernels save old s_dev before kill super and not allow
2110 to restore from callback - restore it before call kill_anon_super.
2114 Description: buffer overruns could theoretically occur
2115 Details : llapi_semantic_traverse() modifies the "path" argument by
2116 appending values to the end of the origin string, and a
2117 overrun may occur. Adding buffer overrun check in liblustreapi.
2121 Description: Fix error on 'ls .' at the top of the Lustre mount.
2122 Details : Don't revalidate dentry if it is a root dentry.
2126 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2127 Details : Modify vm-tunables-rhel4.patch.
2131 Description: Kernel config for 2.6.18-vanilla.
2132 Details : Modify targets/2.6-vanilla.target.in.
2133 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2134 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2135 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2136 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2140 Description: improve handling recoverable errors
2141 Details : if request processig with error which can be recoverable on server
2142 request should be resend, otherwise page released from cache and
2147 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2148 Details : Modify target file & which_patch.
2149 A flaw was found in the IA32 system call emulation provided
2150 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2151 value could be stored in the %RAX register, which could trigger an
2152 out-of-bounds system call table access. An untrusted local user
2153 could exploit this flaw to run code in the kernel
2154 (ie a root privilege escalation). (CVE-2007-4573).
2158 Description: O_DIRECT bypasses client statistics.
2159 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2160 read_ahead_stats were updated. copied the stats section from
2161 osc_send_oap_rpc() into async_internal().
2165 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2169 Description: change order of libsysio includes
2170 Details : '#include sysio.h' should always come before '#include xtio.h'
2174 Description: handle "serious error: objid * already exists" more gracefully
2175 Details : If LAST_ID value on disk is smaller than the objects existing in
2176 the O/0/d* directories, it indicates disk corruption and causes an
2177 LBUG(). If the object is 0-length, then we should use the existing
2178 object. This will help to avoid a full fsck in most cases.
2180 Severity : enhancement
2182 Description: adapt the lustre_config script to support the upgrade case
2183 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2184 server targets to 1.6 in parallel.
2188 Description: To avoid grant space > avaible space when the disk is almost
2189 full. Without this patch you might see the error "grant XXXX >
2190 available" or some LBUG about grant, when the disk is almost
2192 Details : In filter_check_grant, for non_grant cache write, we should
2193 check the left space by if (*left > ungranted + bytes), instead
2194 of (*left > ungranted), because only we are sure the left space
2195 is enough for another "bytes", then the ungrant space should be
2196 increase. In client, we should update cl_avail_grant only there
2197 is OBD_MD_FLGRANT in the reply.
2201 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2202 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2203 described in CVE-2007-4573. This problem could allow untrusted
2204 local users to gain root access.
2207 Frequency : when using O_DIRECT and quotas
2209 Description: Incorrect file ownership on O_DIRECT output files
2210 Details : block usage reported by 'lfs quota' does not take into account
2211 files that have been written with O_DIRECT.
2216 Description: touch file failed when fs is not full
2217 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2218 otherwise we can get ENOSP while fs is not full.
2222 Description: parallel lock callbacks
2223 Details : Instead of sending blocking and completion callbacks as separated
2224 requests, adding them to a set and sending in parallel.
2227 Frequency : only for Cray XT3
2228 Bugzilla : 12829/13455
2229 Description: Changing primary group doesn't change the group lustre assigns to
2231 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2232 overridden with the primary group provided by the group upcall,
2233 whereas the supplied fsgid can be trusted if it is in the list of
2234 supplementary groups returned by the group upcall.
2236 Severity : enhancement
2238 Description: Allow masking D_WARNING, D_ERROR messages from console
2239 Details : Console messages can now be disabled via lnet.printk.
2243 Description: User code with malformed file open parameter crashes client node
2244 Details : Before packing join_file req, all the related reference should be
2245 checked carefully in case some malformed flags cause fake
2246 join_file req on client.
2250 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2251 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2252 mode is equal to the granted mode. However, it should also check
2253 that LDLM_FL_CP_REQD is not set.
2257 Description: Heavy nfs access might result in deadlocks
2258 Details : After ELC code landed, it is now improper to enqueue any mds
2259 locks under och_sem, because enqueue might want to decide to
2260 cancel open locks for same inode we are holding och_sem for.
2264 Description: Client eviction while running blogbench
2265 Details : A lot of unlink operations with concurrent I/O can lead to a
2266 deadlock causing evictions. To address the problem, the number of
2267 oustanding OST_DESTROY requests is now throttled to
2268 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2269 callbacks are processed in priority.
2273 Description: enable ACLs on MDS by default
2274 Details : ACLs must be enabled on MDS by default.
2277 Frequency : PPC/PPC64 only
2279 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2280 Details : fix duplicated definitions between asm-ppc64/types.h and
2281 lustre_types.h on PPC.
2284 Frequency : PPC/PPC64 only
2286 Description: asm-ppc/segment.h does not exist
2287 Details : fix compile issue on PPC.
2291 Description: better handle error messages in extents code
2294 Frequency : RHEL4 only
2296 Description: mkfs is very slow on IA64/RHEL4
2297 Details : A performance regression has been discovered in the MPT Fusion
2298 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2299 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2300 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2302 Severity : enhancement
2304 Description: SNMP support enhancement
2305 Details : Adding total number of sampled request for an MDS node in snmp
2308 Severity : enhancement
2310 Description: Optimize ldlm waiting list processing for PR extent locks
2311 Details : When processing waiting list for read extent lock and meeting read
2312 lock that is same or wider to it that is not contended, skip
2313 processing rest of the list and immediatelly return current
2314 status of conflictness, since we are guaranteed there are no
2315 conflicting locks in the rest of the list.
2319 Description: Time out and refuse to reconnect
2320 Details : When the failover node is the primary node, it is possible
2321 to have two identical connections in imp_conn_list. We must
2322 compare not conn's pointers but NIDs, otherwise we can defeat
2323 connection throttling.
2327 Description: port llog fixes from b1_6 into HEAD
2328 Details : Port llog reference couting and some llog cleanups from b1_6
2329 (bug 10800) into HEAD, for protect from panic and access to already
2330 free llog structures.
2334 Description: Detect stride IO mode in read-ahead
2335 Details : When a client does stride read, read-ahead should detect that and
2336 read-ahead pages according to the detected stride pattern.
2340 Description: data checksumming impacts single node performance
2341 Details : add support for several checksum algorithm. Currently, only CRC32
2342 and Adler-32 are supported. The checksum type can be changed on
2343 the fly via /proc/fs/lustre/osc/*/checksum_type.
2347 Description: use adler32 for page checksums
2348 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2353 Description: build for x2 fails
2354 Details : fix compile issue on Cray systems.
2358 Description: Properly match for duplicate locks
2359 Details : Due to different lock order from skiplists code, we need to
2360 traverse entire list for now
2363 Frequency : only on PPC/SLES10
2365 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2366 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2367 be defined. Add a hack in mkfs_lustre.c to work around this compile
2372 Description: LASSERT on MDS when client holding flock lock dies
2373 Details : ldlm pool logic depends on number of granted locks equal to
2374 number of released locks which is not true for flock locks, so
2375 just exclude such locks from consideration.
2379 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2380 Details : Use fsfilt_sync() for both the whole filesystem sync and
2381 individual file sync to eliminate dangerous inode locking
2382 with I_LOCK that can lead to a deadlock.
2386 Description: performance in 1.6.3
2387 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2388 sure that 1MB requests are not fragmented by the block layer.
2390 Severity : enhancement
2392 Description: organize the server-side client stats on per-nid basis
2393 Details : Change the structure of stats under obdfilter and mds to
2401 The "uuid"s file would list the uuids of _active_ exports.
2402 And the clear entry is to clear all stats and stale nids.
2404 Severity : enhancement
2406 Description: eliminate client locks in face of contention
2407 Details : file contention detection and lockless i/o implementation
2408 for contended files.
2412 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2415 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2417 Description: MDS may lose file striping (and hence file data) in some cases
2418 Details : If there are additional extended attributes stored on the MDS,
2419 in particular ACLs, SELinux, or user attributes (if user_xattr
2420 is specified for the client mount options) then there is a risk
2421 of attribute loss. Additionally, the Lustre file striping
2422 needs to be larger than default (e.g. striped over all OSTs),
2423 and an additional attribute must be stored initially in the
2424 inode and then increase in size enough to be moved to the
2425 external attribute block (e.g. ACL growing in size) for file
2430 Description: skiplist implementation simplification
2431 Details : skiplists are used to group compatible locks on granted list
2432 that was implemented as tracking first and last lock of each
2433 lock group the patch changes that to using doubly linked lists
2437 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2438 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2439 because LASSERT on both the data supplied by a client, and the
2440 data on disk is dangerous and incorrect.
2442 Severity : enhancement
2444 Description: Slow truncate/writes to huge files at high offsets.
2445 Details : Directly associate cached pages to lock that protect those pages,
2446 this allows us to quickly find what pages to write and remove
2447 once lock callback is received.
2451 Description: more ldlm soft lockups
2452 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2453 starve other threads from the resource lock for a long time in
2454 case of long waiting queue, so change the debug level from
2455 D_OTHER to the less frequently used D_INFO.
2457 Severity : enhancement
2459 Description: add -gid, -group, -uid, -user options to lfs find
2463 Description: Hung threads in invalidate_inode_pages2_range
2464 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2465 once one is completed. As a result, some RPCs are stuck in the
2466 queue and are never sent.
2470 Description: filter threads hungs on waiting journal commit
2471 Details : Cleanup filter group llog code, then only filter group llog will
2472 be only created in the MDS/OST syncing process.
2476 Description: Procfs and llog threads access destoryed import sometimes.
2477 Details : Sync the import destoryed process with procfs and llog threads by
2478 the import refcount and semaphore.
2480 Severity : enhancement
2482 Description: openlock cache of b1_6 port to HEAD
2487 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2488 Details : If insertion of an extent fails, then discard the inode
2489 preallocation and free data blocks else it can lead to duplicate
2494 Description: don't always update ctime in ext3_xattr_set_handle()
2495 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2496 In some cases the ctime should not be updated, for example for
2497 2.0->1.8 compatibility it is necessary to delete an xattr and it
2498 should not update the ctime.
2502 Bugzilla : 15713/16362
2503 Description: Assertion in iopen_connect_dentry in 1.6.3
2504 Details : looking up an inode via iopen with the wrong generation number can
2505 populate the dcache with a disconneced dentry while the inode
2506 number is in the process of being reallocated. This causes an
2507 assertion failure in iopen since the inode's dentry list contains
2508 both a connected and disconnected dentry.
2512 Description: assertion failure in ldlm_handle2lock()
2513 Details : fix a race between class_handle_unhash() and class_handle2object()
2514 introduced in lustre 1.6.5 by bug 13622.
2519 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2520 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2525 Description: oops in sd_iostats_seq_show()
2526 Details : unloading/reloading the scsi low level driver triggers a kernel
2527 bug when trying to access the sd iostat file.
2532 Description: Kernel panics during QLogic driver reload
2533 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2534 patch, causing memory corruption.
2539 Description: journal_dev option does not work in b1_6
2540 Details : pass mount option during pre-mount.
2542 Severity : enhancement
2544 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2545 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2546 extent information of a file. It can be used to map logical blocks
2547 in a file to physical blocks in the block device.
2551 Description: LDLM soft lockups - improvement
2552 Details : It is be possible to send the lock handle along with each read
2553 or write request because the client is already doing a lock match
2554 itself so there isn't any reason the OST should have to re-do that
2558 Frequency : only on Cray X2
2560 Description: X2 build failures
2561 Details : fix build failures on Cray X2.
2565 Description: xid & resent requests
2566 Details : Initialize RPC XID from clock at startup (randomly if clock is
2569 Severity : enhancement
2571 Description: Add lustre_start utility to start or stop multiple Lustre servers
2576 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2577 Details : In case of memory pressure, list_del() can be called twice on
2578 req->rq_history_list, causing a kernel oops.
2582 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2583 Details : fix stack overflow in the distributed lock manager by defering
2584 export eviction after a failed AST to the elt thread instead of
2585 handling it in the dlm interpret routine.
2589 Description: Convert some comments to new format.
2590 Details : Update documenting comments to match doxygen conventions.
2594 Description: Grammar fixes.
2595 Details : A couple of trivial spelling fixes.
2599 Description: OSD_COUNTERS-mandatory
2600 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2601 to simplify the code.
2605 Description: simplify cmm_device freeing logic.
2606 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2610 Description: Add lockdep support to dt_object_operations locking interface.
2611 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2612 parameter indicating lock ordering. Update mdd code to use new
2617 Description: Introduce failloc constants for lockless IO tests.
2618 Details : Add two new failloc constants to test lockless IO. Only one of
2619 them in implemented---another is checked in yet to be landed
2624 Description: Add lockdep support for inode mutex.
2625 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2626 in the situations where Lustre has to lock more than one inode
2631 Description: Add optional invariants checking support.
2632 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2633 configure switch. This macro is to be used for consistency and
2634 sanity checks that are too expensive to be left in `production'
2639 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2640 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2645 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2646 Details : Use spin_lock_nested() in (the only) situation where more than
2647 one ldlm_lock is locked simultaneously. Also, fix possible
2648 dead-lock in ldlm_lock_change_resource() by enforcing particular
2653 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2654 Details : Instead of specifying each ldlm_lock call-back through separate
2655 parameter, wrap them into struct ldlm_callback_suite.
2659 Description: Kill join_lru obd method and its callers.
2660 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2661 in memory---a requirement implemented through ->o_join_lru() obd
2662 method in HEAD. Remove this method and its users.
2666 Description: Add asynchronous ldlm ENQUEUE completion handler.
2667 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2668 case that stock ldlm_completion_ast() cannot handle as it waits
2669 until lock is granted. Introduce new ldlm_completion_ast_async()
2670 for this. Also comment ldlm_completion_ast().
2674 Description: ldlm_error <-> errno conversion.
2675 Details : Add functions to map (rather arbitrary) between LDLM error codes
2676 and standard errno values. CLIO needs this to prevent LDLM specific
2677 constants from escaping ldlm and osc.
2681 Description: Kill unused ldlm_handle2lock_ns() function.
2682 Details : Kill unused ldlm_handle2lock_ns() function.
2686 Description: Add lu_ref support to ldlm_lock
2687 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2688 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2689 and ldlm_resource. LDLM interface has to be changed, because code that
2690 releases a reference on a lock, has to "know" what reference this is.
2691 In the most frequent case
2693 lock = ldlm_handle2lock(handle);
2695 LDLM_LOCK_PUT(lock);
2697 no changes are required. When any other reference (received _not_ from
2698 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2699 instead of LDLM_LOCK_PUT().
2701 Arguably, changes are pervasive, and interface requires some discipline
2702 for proper use. On the other hand, it was very instrumental in finding
2703 a few leaked lock references.
2707 Description: Add ldlm_lock_addref_try().
2708 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2709 attempts to addref a lock that might be being canceled
2714 Description: Add ldlm_weigh_callback().
2715 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2716 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2717 hard-coded `number of pages' logic.
2721 Description: Add lockdep annotations to llog code.
2722 Details : Use appropriately tagged _nested() locking calls in the places
2723 where llog takes more than one ->lgh_lock lock.
2727 Description: Add loi_kms_set().
2728 Details : Wrap kms updates into a helper function.
2732 Description: Constify instances of struct lsm_operations.
2733 Details : Constify instances of struct lsm_operations.
2737 Description: lu_conf support.
2738 Details : On a server, a file system object is uniquely identified
2739 by a fid, which is sufficient to locate and load all object
2740 state (inode). On a client, on the other hand, more data are
2741 necessary instantiate an object. Change lu_object_find() and
2742 friends to take additional `lu_conf' argument describing object.
2743 Typically this includes layout information.
2747 Description: lu_context fixes.
2748 Details : Introduce new lu_context functions that are needed on the client
2749 side, where some system threads (ptlrpcd) are shared by multiple
2750 modules, and so cannot be stopped during module shutdown.
2754 Description: Add start and stop methods to lu_device_type_operations.
2755 Details : Introduce two new methods in lu_device_type_operations, that are
2756 invoked when first instance of a given type is created and last one
2757 is destroyed respectively. This is need by CLIO.
2761 Description: Add lu_ref support to struct lu_device.
2762 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2763 track leaked references.
2767 Description: Introduce lu_kmem_descr.
2768 Details : lu_kmem_descr and its companion interface allow to create
2769 and destroy a number of kmem caches at once.
2773 Description: Fix lu_object finalization race.
2774 Details : Fix a race between lu_object_find() finding an object and its
2775 concurrent finalization. This race is (most likely) not possible
2776 on the server, but might happen on the client.
2780 Description: Introduce lu_ref interface.
2781 Details : lu_ref is a debugging module allowing to track references to
2782 a given object. It is quite cpu expensive, and has to be
2783 explicitly enabled with --enable-lu_ref. See usage description
2788 Description: Factor lu_site procfs stats into a separate function.
2789 Details : Separate lu_site stats printing code into a separate function
2790 that can be reused on a client.
2794 Description: Constify instances of struct {lu,dt,md}_device_operations.
2795 Details : Constify instances of struct {lu,dt,md}_device_operations.
2799 Description: Introduce struct md_site and move meta-data specific parts of
2800 struct lu_site here.
2801 Details : Move md-specific fields out of struct lu_site into special struct
2802 md_site, so that lu_site can be used on a client.
2806 Description: Kill mdd_lov_destroy().
2807 Details : Remove unused mdd code.
2811 Description: Add st_block checking to multistat.c.
2812 Details : Add st_block checking to multistat.c.
2816 Description: Add lu_ref support to struct obd_device.
2817 Details : Add lu_ref tracking to obd_device.
2821 Description: Kill obd_set_fail_loc().
2822 Details : Remove unused code.
2826 Description: Add special type for ptlrpc_request interpret functions.
2827 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2828 there. Actual usage will be in CLIO.
2832 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2833 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2834 doesn't work with lockdep.
2838 Description: Add rwv.c test program.
2839 Details : New testing program exercising readv(2) and writev(2) (Qian).
2843 Description: Add sendfile.c test program.
2844 Details : New testing program exercising sendfile(2) (Jay).
2848 Description: Ratelimit a message that can be very frequent.
2849 Details : Ratelimit a memory allocation failure message that can
2854 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2855 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2859 Description: Liblustre build fixes.
2860 Details : Liblustre build fixes.
2864 Description: libcfs: add cfs_{need,cond}_resched() interface.
2865 Details : libcfs: add cfs_{need,cond}_resched() definition and
2866 implementations for Linux, NT, and liblustre.
2868 Severity : enhancement
2870 Description: More exported tunables for mballoc
2871 Details : Add support for tunable preallocation window and new tunables for
2872 large/small requests
2876 Description: Detect corruption of block bitmap and checking for preallocations
2877 Details : Checks validity of on-disk block bitmap. Also it does better
2878 checking of number of applied preallocations. When corruption is
2879 found, it turns filesystem readonly to prevent further corruptions.
2883 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2884 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2885 window to 0. If it does hit cache miss, and read-pattern is still
2886 stride-io mode, does not reset the stride window, but also does not
2887 increase the stride window length in this case.
2891 Frequency : only for big-endian servers
2892 Description: Check if system is big-endian while mounting fs with extents feature
2893 Details : Mounting a filesystem with extents feature will fail on big-endian
2894 systems since ext3-based ldiskfs is not supported on big-endian
2895 systems. This can be over-riden with "bigendian_extents" mount option.
2897 Severity : enhancement
2899 Description: The root squash functionality
2900 Details : A security feature, which is to prevent users from being able
2901 to mount lustre on their desktop, run as root, and delete
2902 all of the files in the filesystem. The goal is accomplished by
2903 remapping user id (UID) and group id (GID) of the root user to
2904 a UID and GID specified by the system administartor via Lustre
2905 configuration management server (MGS). The functionality also
2906 allows to specify sets of clients for which the remapping does
2911 Description: Excessive recovery window
2912 Details : With AT enabled, the recovery window can be excessively long (6000+
2913 seconds). To address this problem, we no longer use
2914 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2915 timeout no longer depends on the service time, it is set to
2916 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2917 time via pb_service_time.
2921 Description: Watchdog triggered on MDS failover
2922 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2923 the OSTs know that the MDS "UUID" can be reused for the same export
2924 from a different NID, so we do not need to wait for the export to be
2928 Frequency : rare, only if using MMP with Linux RAID
2930 Description: MMP doesn't work with Linux RAID
2931 Details : While using HA for Lustre servers with Linux RAID, it is possible
2932 that MMP will not detect multiple mounts. To make this work we
2933 need to unplug the device queue in RAID when the MMP block is being
2934 written. Also while reading the MMP block, we should read it from
2935 disk and not the cached one.
2937 Severity : enhancement
2939 Description: open file using fid
2940 Details : A file can be opened using just its fid, like
2941 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2944 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2947 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2948 Details : add proper locking for creation of jbd_4k slab cache
2952 Description: MMP check in ext3_remount() fails without displaying any error
2953 Details : When multiple mount protection fails during remount, proper error
2956 Severity : enhancement
2958 Description: Allow stripe size to be up to 4G-64k
2959 Details : Fix math logic to allow large stripe sizes.
2963 Description: add check for >8TB ldiskfs filesystems
2964 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2965 Don't allow >8TB ldiskfs filesystems to be mounted without
2966 force_over_8tb mount option
2971 Descriptoin: OSS complains "too many blocks" and crashes
2972 Details : In filter_commitrw_write():
2973 1. lquota_chkquota();
2974 2. obdo_from_inode();
2975 3. lquota_pending_commit();
2976 "2." has changed uid and gid in this case, then caused
2977 lquota_chkquota() and lquota_pending_commit() to use
2978 different uid and gid. That is the root of the bug.
2982 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2983 Details : Detect and print out the 'stripe count = -1' case properly.
2987 Description: Out or order replies might be lost on replay
2988 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2989 request with tid smaller than one currently being added, add it
2990 to the start, not end of the list.
2992 --------------------------------------------------------------------------------
2994 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2996 * Support for kernels:
2997 2.6.5-7.283 (SLES 9),
2998 2.6.9-55.EL (RHEL 4),
2999 2.6.16.46-0.14 (SLES 10),
3000 2.6.18.8 vanilla (kernel.org)
3001 * Client support for unpatched kernels:
3002 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3003 2.6.16 - 2.6.22 vanilla (kernel.org)
3004 * Due to recently discovered recovery problems, we do not recommend
3005 using patchless RHEL 4 clients with this or any earlier release.
3006 * Recommended e2fsprogs version: 1.39.cfs8
3007 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3008 * Starting with this release, the ldiskfs backing filesystem required
3009 by Lustre is now in its own package, lustre-ldiskfs. This package
3010 should be installed. It is versioned separately from Lustre and
3011 may be released separately in future.
3015 Description: block reactivating mgc import until all deactivates complete
3016 Details : Fix race when failing back MDT/MGS to itself (testing)
3019 Frequency : at statup only
3021 Description: mds_lov_synchronize race leads to various problems
3022 Details : simultaneous MDT->OST connections at startup can cause the
3023 sync to abort, leaving the OSC in a bad state.
3025 Severity : enhancement
3027 Description: add optional extra BUILD_VERSION info
3028 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3029 to override the lustre version.
3032 Frequency : 2.6.18 servers only
3034 Description: ll_kern_mount() doesn't release the module reference
3035 Details : The ldiskfs module reference count never drops down to 0
3036 because ll_kern_mount() doesn't release the module reference.
3041 Description: server LBUG when using old ost_num_threads parameter
3042 Details : Accept the old ost_num_threads parameter but warn that it
3043 is deprecated, and fix an off-by-one error that caused an LBUG.
3048 Description: Transient SCSI error results in persistent IO issue
3049 Details : iobuf->dr_error is not reinitialized to 0 between two
3053 Frequency : sometimes when underlying device returns I/O errors
3055 Description: OSTs not going read-only during write failures
3056 Details : OSTs are not remounted read-only when the journal commit threads
3057 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3058 instead of the ext3 wrappers.
3062 Description: poor connect scaling with increasing client count
3063 Details : Don't run filter_grant_sanity_check for more than 100 exports
3064 to improve scaling for large numbers of clients.
3067 Frequency : SLES10 only
3069 Description: sanity-quota.sh quotacheck failed: rc = -22
3070 Details : Quotas cannot be enabled on SLES10.
3073 Frequency : liblustre clients only
3075 Description: getdirentries does not give error when run on compute nodes
3076 Details : getdirentries does not fail when the size specified as an argument
3077 is too small to contain at least one entry
3079 Severity : enhancement
3081 Description: Add LNET router traceability for debug purposes
3082 Details : If a checksum failure occurs with a router as part of the
3083 IO path, the NID of the last router that forwarded the bulk data
3084 is printed so it can be identified.
3089 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3090 Details : Due to a race condition, liblustre clients were occasionally
3091 evicted incorrectly.
3093 Severity : enhancement
3095 Description: lfs setstripe use optional parameters instead of postional
3098 Severity : enhancement
3100 Description: Nanosecond timestamp support for ldiskfs
3101 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3102 resolution timestamps. There is not yet support for this at
3103 the Lustre filesystem level.
3106 Frequency : during server recovery
3108 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3109 Details : request with rq_no_resend flag not awake l_wait_event if they get
3113 Frequency : nfs export on patchless client
3115 Description: connectathon hang when test nfs export over patchless client
3116 Details : Disconnected dentry cannot be found with lookup, so we do not need
3117 to unhash it or make it invalid
3120 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3121 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3122 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3127 Description: open req refcounting wrong on reconnect
3128 Details : If reconnect happened between getting open reply from server and
3129 call to mdc_set_replay_data in ll_file_open, we will schedule
3130 replay for unreferenced request that we are about to free.
3131 Subsequent close will crash in variety of ways.
3132 Check that request is still eligible for replay in
3133 mdc_set_replay_data().
3138 Description: disable writes to filesystem when reading health_check file
3139 Details : the default for reading the health_check proc file has changed
3140 to NOT do a journal transaction and write to disk, because this
3141 can cause reads of the /proc file to hang and block HA state
3142 checking on a healthy but otherwise heavily loaded system. It
3143 is possible to return to the previous behaviour during configure
3144 with --enable-health-write.
3146 Severity : enhancement
3148 Description: 64-bit inode version
3149 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3150 to the inode. This will be required for version-based recovery.
3155 Description: MDS fails to start if a duplicate client export is detected
3156 Details : in some rare cases it was possible for a client to connect to
3157 an MDS multiple times. Upon recovery the MDS would detect this
3158 and fail during startup. Handle this more gracefully.
3160 Severity : enhancement
3162 Description: Add -o localflock option to simulate old noflock
3164 Details : This will achieve local-only flock/fcntl locks
3170 Description: log_commit_thread vs filter_destroy race leads to crash
3171 Details : Take import reference before releasing llog record semaphore
3176 Description: Wrong request locking in request set processing
3177 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3178 imp_delayed_list, in this place should be used imp_lock.
3181 Frequency : when reconnection
3183 Description: Grant Leak when osc reconnect to OST
3184 Details : When osc reconnect ost, OST(filter) should check whether it
3185 should grant more space to client by comparing fed_grant and
3186 cl_avail_grant, and return the granted space to client instead
3187 of "new granted" space, because client will call osc_init_grant
3188 to update the client grant space info.
3191 Frequency : when client reconnect to OST
3193 Description: Grant Leak when osc do resend and replay bulk write
3194 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3195 bulk write request, because the grant info while be sync between
3196 OSC and OST when reconnect, and we should ignore the grant info
3197 these of resend/replay write req.
3202 Description: Grant space more than available left space sometimes.
3203 Details : When then OST is about to be full, if two bulk writing from
3204 different clients came to OST. Accord the avaliable space of the
3205 OST, the first req should be permitted, and the second one
3206 should be denied by ENOSPC. But if the seconde arrived before
3207 the first one is committed. The OST might wrongly permit second
3208 writing, which will cause grant space > available space.
3211 Frequency : when client is evicted
3213 Description: Grant might be wrongly erased when osc is evicted by OST
3214 Details : when the import is evicted by server, it will fork another
3215 thread ptlrpc_invalidate_import_thread to invalidate the
3216 import, where the grant will be set to 0. While the original
3217 thread will update the grant it got when connecting. So if
3218 the former happened latter, the grant will be wrongly errased
3219 because of this race.
3224 Description: Checking Stale with correct fid
3225 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3226 to check whether it is stale, because sometimes, we want the
3227 enqueue happened anyway, and op_data.fid2 will not be initialized.
3229 Severity : enhancement
3231 Description: update patchless client
3232 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3235 Frequency : only with 2.4 kernel
3237 Description: random memory corruption
3238 Details : size of struct ll_inode_info is to big for union inode.u and this
3239 can be cause of random memory corruption.
3244 Description: Memory leak in recovery
3245 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3246 It should also check obd_fail before fsfilt_start, otherwise if
3247 fsfilt_start return -EROFS,(failover mds during mds recovery).
3248 then the req will return with repmsg->transno = 0 and rc = EROFS.
3249 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3250 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3251 be freed no matter whether fsfilt_commit success or not.
3254 Frequency : only with huge count clients
3256 Description: Prevents from taking the superblock lock in llap_from_page for
3258 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3259 for llap_from_page call in ll_removepage prevents from taking the
3260 superblock lock for a soon died page.
3265 Description: Not check open intent error before release open handle
3266 Details : in some rare cases, the open intent error is not checked before
3267 release open handle, which may cause
3268 ASSERTION(open_req->rq_transno != 0), because it tries to release
3269 the failed open handle.
3274 Description: Set cat log bitmap only after create log success.
3275 Details : in some rare cases, the cat log bitmap is set too early. and it
3276 should be set only after create log success.
3280 Description: Accessing a block bevice can re-enable I/O when Lustre is
3281 tearing down a device.
3282 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3287 Description: service threads may hog cpus when there are a lot of requests
3288 Details : Insert cond_resched to give other threads a chance to use some CPU
3293 Description: the cat log was not initialized in recovery
3294 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3295 unlink log on mds will not be initialized until mds post
3296 recovery. And also in mds post recovery, the unlink log will
3297 initialization will be done asynchronausly, so there will be race
3298 between add unlink log and unlink log initialization.
3302 Description: brw_stats were being printed incorrectly
3303 Details : brw_stats were being printed as log2 but all of them were not
3304 recorded as log2. Also remove some code duplication arising from
3305 filter_tally_{read,write}.
3309 Frequency : rare, only in recovery.
3310 Description: ASSERTION(req->rq_type != LI_POISON) failed
3311 Details : imp_lock should be held while iterating over imp_sending_list for
3312 prevent destroy request after get timeout in ptlrpc_queue_wait.
3316 Description: replay-single.sh test 52 fails
3317 Details : A lock's skiplist need to be cleanup when it being unlinked
3318 from its resource list.
3322 Description: Short directio read returns full requested size rather than
3324 Details : Direct I/O operations should return actual amount of bytes
3325 transferred rather than requested size.
3327 Severity : enhancement
3329 Description: metadata RPC reduction (e.g. for rm performance)
3330 Details : decrease the amount of synchronous RPC between clients and servers
3331 by canceling conflicing lock before the operation on the client
3332 and packing thier handles into the main operation RPC to server.
3334 Severity : enhancement
3336 Description: Async OSC create to avoid the blocking unnecessarily.
3337 Details : If an OST has no remaining object, system will block on the
3338 creation when it needs to create a new object on this OST. Now,
3339 ways use pre-created objects when available, instead of blocking on
3340 an empty osc while others are not empty. If we must block, we
3341 block for the shortest possible period of time.
3345 Description: improve handling recoverable errors
3346 Details : if request processig with error which can be recoverable on server
3347 request should be resend, otherwise page released from cache and
3350 Severity : enhancement
3352 Description: refine locking for avoid write wrong info into lov_objid file
3353 Details : fix possible races with add new target and write/update data in
3356 --------------------------------------------------------------------------------
3358 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3363 Frequency : on some architectures
3365 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3366 Details : Uninitialized flags sometimes cause configuration commands to
3370 Frequency : patchless clients only
3372 Description: missing __iget() symbol export
3373 Details : The __iget() symbol export is missing. To avoid the need for
3374 this on patchless clients the deathrow inode reaper is turned
3375 off, and we depend on the VM to clean up old inodes. This
3376 dependency was during via the fix for bug 12181.
3378 --------------------------------------------------------------------------------
3380 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3382 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3383 INTEROPERATE with older versions automatically. In many cases a
3384 special upgrade step is needed. Please read the
3385 user documentation before upgrading any part of a live system.
3386 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3387 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3388 * WARNING: Lustre configuration and startup changes are required with
3389 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3393 Severity : enhancement
3395 Description: MountConf
3396 Details : Lustre configuration is now managed via mkfs and mount
3397 commands instead of lmc and lconf. New obd types (MGS, MGC)
3398 are added for dynamic configuration management. See
3399 https://mail.clusterfs.com/wikis/lustre/MountConf for
3402 Severity : enhancement
3404 Description: dynamic OST addition
3405 Details : OSTs can now be added to a live filesystem
3407 Severity : enhancement
3409 Description: startup order invariance
3410 Details : MDTs and OSTs can be started in any order. Clients only
3411 require the MDT to complete startup.
3413 Severity : enhancement
3415 Description: parallel, asynchronous orphan cleanup
3416 Details : orphan cleanup is now performed in separate threads for each
3417 OST, allowing parallel non-blocking operation.
3419 Severity : enhancement
3421 Description: optimized stripe assignment
3422 Details : stripe assignments are now made based on ost space available,
3423 ost previous usage, and OSS previous usage, in order to try
3424 to optimize storage space and networking resources.
3426 Severity : enhancement
3428 Description: Permanently set tunables
3429 Details : All writable /proc/fs/lustre tunables can now be permanently
3430 set on a per-server basis, at mkfs time or on a live system.
3432 Severity : enhancement
3434 Description: Lustre message v2
3435 Details : Add lustre message format v2.
3437 Severity : enhancement
3439 Description: client OST exclusion list
3440 Details : Clients can be started with a list of OSTs that should be
3441 declared "inactive" for known non-responsive OSTs.
3445 Description: SPEC SFS validation failure on NFS v2 over lustre.
3446 Details : Changes the blocksize for regular files to be 2x RPC size,
3447 and not depend on stripe size.
3449 Severity : enhancement
3451 Description: Multiple MD RPCs in flight.
3452 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3453 To avoid overly-overloading MDS, introduce a limit on number of
3454 MDS RPCs in flight for a single client and add /proc controls
3455 to adjust this limit.
3457 Severity : enhancement
3459 Description: client read/write statistics
3460 Details : Add client read/write call usage stats for performance
3461 analysis of user processes.
3462 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3463 file access. extents_stats shows chunk size distribution.
3464 extents_stats_per_process show chunk size distribution per
3467 Severity : enhancement
3469 Description: mds statistics
3470 Details : Add detailed mds operations statistics in
3471 /proc/fs/lustre/mds/*/stats.
3475 Description: Failure of copying files with lustre special EAs.
3476 Details : Client side always return success for setxattr call for lustre
3477 special xattr (currently only "trusted.lov").
3482 Description: Refcount LNET uuids
3483 Details : The global LNET uuid list grew linearly with every startup;
3484 refcount repeated list entries instead of always adding to
3487 Severity : enhancement
3489 Description: Dynamic service threads
3490 Details : Within a small range, start extra service threads
3491 automatically when the request queue builds up.
3494 Frequency : mixed-endian client/server environments
3496 Description: mixed-endian crashes
3497 Details : The new msg_v2 system had some failures in mixed-endian
3500 Severity : enhancement
3502 Description: Easy OST removal
3503 Details : OSTs can be permanently deactivated with e.g. 'lctl
3504 conf_param lustre-OST0001.osc.active=0'
3506 Severity : enhancement
3508 Description: MGS proc entries
3509 Details : Added basic proc entries for the MGS showing what filesystems
3512 Severity : enhancement
3514 Description: provide MGS failover
3515 Details : Added config lock reacquisition after MGS server failover.
3517 Severity : enhancement
3519 Description: add Linux 2.4 support
3520 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3524 Description: a large application tries to do I/O to the same resource and dies
3525 in the middle of it.
3526 Details : Check the req->rq_arrival time after the call to
3527 ost_brw_lock_get(), but before we do anything about
3528 processing it & sending the BULK transfer request. This
3529 should help move old stale pending locks off the queue as
3530 quickly as obd_timeout.
3533 Frequency : when an incorrect nid is specified during startup
3535 Description: ptlrpc connect to non-existant node causes kernel crash
3536 Details : LNET can't be re-entered from an event callback, which
3537 happened when we expire a message after the export has been
3538 cleaned up. Instead, hand the zombie cleanup off to another
3541 Severity : enhancement
3543 Description: plain/inodebits lock performance improvement
3544 Details : Grouping plain/inodebits in granted list by their request modes
3545 and bits policy, thus improving the performance of search through
3549 Frequency : only if OST filesystem is corrupted
3551 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3552 Details : for a short time RPCs with bulk IO are in the replay list,
3553 but replay of bulk IOs is unimplemented. If the OST filesystem
3554 is corrupted due to disk cache incoherency and then replay is
3555 started it is possible to trip an assertion. Avoid putting
3556 committed RPCs into the replay list at all to avoid this issue.
3559 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3561 Description: System hang on startup
3562 Details : This bug allowed the liblustre (e.g. catamount) client to
3563 return to the app before handling all startup RPCs. This
3564 could leave the node unresponsive to lustre network traffic
3565 and manifested as a server ptllnd timeout.
3567 Severity : enhancement
3569 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3570 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3571 boolean to control whether to print peer debug info when a
3572 client's RPC times out.
3575 Frequency : only for kernels with patches from Lustre below 1.4.3
3577 Description: Remove old rdonly API
3578 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3581 Frequency : only for devices with external journals
3583 Description: Set external device read-only also
3584 Details : During a commanded failover stop, we set the disk device
3585 read-only while the server shuts down. We now also set any
3586 external journal device read-only at the same time.
3589 Frequency : when upgrading from 1.4 while trying to change parameters
3591 Description: The wrong (new) MDC name was used when setting parameters for
3592 upgraded MDT's. Also allows changing of OSC (and MDC)
3593 parameters if --writeconf is specified at tunefs upgrade time.
3596 Frequency : when setting specific ost indicies
3598 Description: QOS code breaks on skipped indicies
3599 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3600 created with --index need not be sequential.
3602 Severity : enhancement
3604 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3605 Details : The uninit_groups feature works in conjunction with the kernel
3606 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3607 up the pass1 processing of e2fsck. This is a read-only feature
3608 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3609 mount filesystems that have had this feature enabled.
3611 Severity : enhancement
3613 Description: Improve multi-block allocation algorithm to avoid fragmentation
3614 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3615 allocation locality and avoid filesystem fragmentation.
3617 ------------------------------------------------------------------------------
3619 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3621 * Support for kernels:
3622 2.4.21-47.0.1.EL (RHEL 3)
3623 2.6.5-7.283 (SLES 9)
3624 2.6.9-42.0.10.EL (RHEL 4)
3625 2.6.12.6 vanilla (kernel.org)
3626 2.6.16.27-0.9 (SLES 10)
3627 * Recommended e2fsprogs version: 1.39.cfs5
3629 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3633 Frequency : occasional, depends on client load and configuration
3634 Bugzilla : 12181, 12203
3635 Description: data loss for recently-modified files
3637 Details : In some cases it is possible that recently written or created
3638 files may not be written to disk in a timely manner (this should
3639 normally be within 30s unless client IO load is very high).
3640 The problem appears as zero-length files or files that are a
3641 multiple of 1MB in size after a client crash or client eviction
3642 that are missing data at the end of the file.
3644 This problem is more likely to be hit on clients where files are
3645 repeatedly created and unlinked in the same directory, clients
3646 have a large amount of RAM, have many CPUs, the filesystem has
3647 many OSTs, the clients are rebooted frequently, and/or the files
3648 are not accessed by other nodes after being written.
3650 The presence of the problem can be detected by looking at
3651 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3652 smaller than the second (nr_unused) then dirty files will not
3653 be flushed automatically to disk. "sync; sleep 10" should be
3654 run several times on the node before unmounting it to update
3655 Lustre (this is also safe to run on nodes without this problem).
3657 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3658 that can cause this same problem, so customers using that kernel
3659 also need to update the kernel in addition to Lustre. In order
3660 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3662 It is normal that files written just before a client crash (less
3663 than 30s) may not yet have been flushed to disk, even for local
3667 Frequency : frequent on thin XT3 nodes
3669 Description: UUID collision on thin XT3 Linux nodes
3670 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3671 randomly, since we relied on an insufficiently-seeded PRNG.
3676 Description: OSS hangs after "All ost request buffers busy"
3677 Details : A deadlock between quota and journal operations caused OSS
3678 hangs after printing "All ost request buffers busy."
3681 Frequency : always on liblustre builds
3683 Description: Cleanup compiler warnings on liblustre
3686 Frequency : always on liblustre builds on XT3
3688 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3692 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3693 > 32000 subdirectories
3694 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3695 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3696 subdirectory count crosses 32000. This will aid e2fsck to
3697 correctly handle more than 32000 subdirectories.
3700 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3702 Description: System hang on startup
3703 Details : This bug allowed the liblustre (e.g. catamount) client to
3704 return to the app before handling all startup RPCs. This
3705 could leave the node unresponsive to lustre network traffic
3706 and manifested as a server ptllnd timeout.
3708 Severity : enhancement
3710 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3711 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3712 boolean to control whether to print peer debug info when a
3713 client's RPC times out.
3718 Description: make O_SYNC working on 2.6 kernels
3719 Details : 2.6 kernels use different method for mark pages for write,
3720 so need add a code to lustre for O_SYNC work.
3725 Description: Failure to close file and release space on NFS
3726 Details : Put inode details into lock acquired in ll_intent_file_open.
3727 Use mdc_intent_lock in ll_intent_open to properly
3728 detect all kind of errors unhandled by mdc_enqueue.
3733 Description: proc file read during shutdown sometimes raced obd removal,
3735 Details : Add lock to prevent obd access after proc file removal.
3738 Frequency : Only for files larger than 4GB on 32-bit clients.
3740 Description: improperly doing page alignment of locks
3741 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3742 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3747 Description: RPCs being resent when they shouldn't be
3748 Details : Some RPCs that should not be resent are being resent. This
3749 can cause inconsistencies in the RPC state machine. Do not
3750 resend such requests.
3753 Frequency : rare, only with NFS export
3755 Description: Crash on NFS re-export node
3756 Details : under very unusual load conditions an assertion is hit in
3757 ll_intent_file_open()
3760 Frequency : only if OST filesystem is corrupted
3762 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3763 Details : for a short time RPCs with bulk IO are in the replay list,
3764 but replay of bulk IOs is unimplemented. If the OST filesystem
3765 is corrupted due to disk cache incoherency and then replay is
3766 started it is possible to trip an assertion. Avoid putting
3767 committed RPCs into the replay list at all to avoid this issue.
3772 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3773 Details : Large single O_DIRECT read and write calls can fail to allocate
3774 a sufficiently large buffer to process the request. In case of
3775 allocation failure the allocation is retried with a smaller
3776 buffer and broken into smaller requests.
3778 Severity : enhancement
3780 Description: Add -o localflock option to simulate old noflock behaviour.
3781 Details : This will achieve local-only flock/fcntl locks coherentness.
3786 Description: versioning check is incomplete
3787 Details : Checking the version difference of client vs. server, report
3788 error if the gap is too big.
3793 Description: add support PG_writeback bit
3794 Details : add support for PG_writeback bit for Lustre, for more carefull
3795 work with page cache in 2.6 kernel. This also fix some deadlocks
3796 and remove hack for work O_SYNC with 2.6 kernel.
3798 Severity : enhancement
3800 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3801 Details : The uninit_groups feature works in conjunction with the kernel
3802 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3803 up the pass1 processing of e2fsck. This is a read-only feature
3804 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3805 mount filesystems that have had this feature enabled.
3807 Severity : enhancement
3809 Description: Improve multi-block allocation algorithm to avoid fragmentation
3810 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3811 allocation locality and avoid filesystem fragmentation.
3813 ------------------------------------------------------------------------------
3815 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3817 * Support for kernels:
3818 2.6.9-42.0.3.EL (RHEL 4)
3819 2.6.5-7.276 (SLES 9)
3820 2.4.21-47.0.1.EL (RHEL 3)
3821 2.6.12.6 vanilla (kernel.org)
3822 2.6.16.21-0.8 (SLES10)
3823 * Recommended e2fsprogs version: 1.39.cfs2-0
3825 * The backwards-compatible /proc/sys/portals symlink has been removed
3826 in this release. Before upgrading, please ensure that you change
3827 any configuration scripts or /etc/sysctl.conf files that access
3828 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3829 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3830 in advance of the upgrade on any system running Lustre 1.4.6 or
3831 newer, since /proc/sys/lnet was added in that version.
3832 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3836 Frequency : only when quota is used
3838 Description: avoid scanning export list for quota master
3839 Details : Change the algorithms to avoid scanning export list in order
3840 to improve the efficiency.
3843 Frequency : MDS failover only, very rarely
3845 Description: "went back in time" messages on mds failover
3846 Details : The greatest transno may be lost when the current operation
3847 finishes with an error (transno==0) and the client's last_rcvd
3848 record is over-written. Save the greatest transno in the
3849 mds_last_transno for this case.
3852 Frequency : always for specific kernels and striping counts
3854 Description: client may get "Matching packet too big" without ACL support
3855 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3856 when trying to access files in certain configurations. The
3857 clients should in fact be denied when mounting because they do
3858 not understand ACLs.
3861 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3863 Description: many clients connecting with IO in progress causes connect timeouts
3864 Details : Avoid synchronous journal commits to avoid delays caused by many
3865 clients connecting/disconnecting when bulk IO is in progress.
3866 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3867 OST_IO_PORTAL to avoid delays behind potentially many pending
3871 Frequency : occasionally with multiple writers to a single file
3873 Description: shared writes to file may result in wrong size reported by stat()
3874 Details : Allow growing of kms when extent lock is cancelled
3877 Frequency : always with random mmap IO to multi-striped file
3879 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3880 Details : Only if the hole is at the end of OST object so that kms is too
3881 small. Fix is to increase kms accordingly in ll_nopage.
3884 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3886 Description: writes to a missing object would leak memory on the OST
3887 Details : If there is an inconsistency between the MDS and OST filesystems,
3888 such that the MDS references an object that doesn't exist, writes
3889 to that object will leak memory due to incorrect cleanup in the
3890 error handling path, eventually running out of memory on the OST.
3895 Description: Creating too long symlink causes lustre errors
3896 Details : Check symlink and name lengths before sending requests to MDS.
3899 Frequency : only if flock is enabled (not on by default)
3901 Description: posix locks not released on fd closure on 2.6.9+
3902 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3903 kernels, this caused such locks not to be released on fd close and
3904 then assertions on fs unmount about still used locks.
3907 Frequency : MDS failover only, very rarely
3909 Description: clients may get ASSERTION(granted_lock != NULL)
3910 Details : When request was taking a long time, and a client was resending
3911 a getattr by name lock request. The were multiple lock requests
3912 with the same client lock handle and
3913 mds_getattr_name->fixup_handle_for_resent_request found one of the
3914 lock handles but later failed with ASSERTION(granted_lock != NULL).
3919 Description: handle->h_buffer_credits > 0, assertion failure
3920 Details : h_buffer_credits is zero after truncate, causing assertion
3921 failure. This patch extends the transaction or creates a new
3925 Frequency : NFS re-export or patchless client
3926 Bugzilla : 11179, 10796
3927 Description: Crash on NFS re-export node (__d_move)
3928 Details : We do not want to hash the dentry if we don't have a lock.
3929 But if this dentry is later used in d_move, we'd hit uninitialised
3930 list head d_hash, so we just do this to init d_hash field but
3931 leave dentry unhashed.
3934 Frequency : NFS re-export or patchless client
3936 Description: NFS exports has problem with symbolic link
3937 Details : lustre client didn't properly install dentry when re-exported
3938 to NFS or running patchless client.
3941 Frequency : NFS re-export or patchless client
3943 Description: Various nfs/patchless fixes.
3944 Details : fixes reuse disconected alias for lookup process - this fixes
3945 warning "find_exported_dentry: npd != pd",
3946 fix permission error with open files at nfs.
3947 fix apply umask when do revalidate.
3950 Frequency : occasional
3952 Description: Crash on NFS re-export node
3953 Details : calling clear_page() on the wrong pointer triggered oops in
3954 generic_mapping_read().
3957 Frequency : rarely, using O_DIRECT IO
3959 Description: unaligned directio crashes client with LASSERT
3960 Details : check for unaligned buffers before trying any requests.
3963 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3965 Description: stale data returned from RAID cache
3966 Details : If only a small amount of IO is done to the RAID device before
3967 reading it again it is possible to get stale data from the RAID
3968 cache instead of reading it from disk.
3971 Frequency : always for sles10 kernel
3973 Description: sles10 support
3974 Details : ll_follow_link: compile fixes and using of nd_set_link
3975 under newer kernels.
3978 Frequency : depends on arch, kernel and compiler version, always on sles10
3981 Description: recursive or deep enough symlinks cause stack overflow
3982 Details : getting rid of large stack-allocated variable in
3986 Frequency : depends on hardware
3988 Description: lustre write performance loss in the SLES10 kernel
3989 Details : the performance loss is caused by using of write barriers in the
3990 ext3 code. The SLES10 kernel turns barrier support on by
3991 default. The fix is to undo that change for ldiskfs.
3993 ------------------------------------------------------------------------------
3995 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3997 * Support for kernels:
3998 2.6.9-42.0.3EL (RHEL 4)
3999 2.6.5-7.276 (SLES 9)
4000 2.4.21-47.0.1.EL (RHEL 3)
4001 2.6.12.6 vanilla (kernel.org)
4005 Frequency : quota enabled and large files being deleted
4007 Description: releasing more than 4GB of quota at once hangs OST
4008 Details : If a user deletes more than 4GB of files on a single OST it
4009 will cause the OST to spin in an infinite loop. Release
4010 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4015 Description: statfs data retrieved from /proc may be stale or zero
4016 Details : When reading per-device statfs data from /proc, in the
4017 {kbytes,files}_{total,free,avail} files, it may appear
4018 as zero or be out of date.
4021 Frequency : systems with MD RAID1 external journal devices
4023 Description: lconf's call to blkid is confused by RAID1 journal devices
4024 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4025 of returning all block devices that match the journal UUID.
4028 Frequency : always, for aggregate stripe size over 4GB
4030 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4031 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4032 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4035 Frequency : always if "lfs find" used on a local file/directory
4037 Description: "lfs find" segfaults if used on a local file/directory
4038 Details : The case where a directory component was not specified wasn't
4039 handled correctly. Handle this properly.
4042 Frequency : always on ppc64
4044 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4045 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4046 perform endianness conversion.
4049 Frequency : rarely (truncate to non-zero file size after write under load)
4050 Bugzilla : 10730, 10687
4051 Description: Files padded with zeros to next 4K multiple
4052 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4053 it is possible that files that are truncated to a non-zero size
4054 immediately after being written are filled with zero bytes beyond
4055 the truncated size. No file data is lost.
4057 Severity : enhancement
4059 Description: Allow recovery/failover for liblustre clients.
4060 Details : liblustre clients were unaware of failover configurations until
4063 Severity : enhancement
4065 Description: user file locks should fail when not mounting with flock option
4066 Details : Set up an error-returning stub in ll_file_operations.lock field
4067 to prevent incorrect behaviour when client is mounted without
4068 flock option. Also, set up properly f_op->flock field for
4072 Frequency : always on ia64
4074 Description: "lfs df" loops on printing out MDS statfs information
4075 Details : The obd_ioctl_data was not initialized and in some systems
4076 this caused a failure during the ioctl that did not return
4077 an error. Initialize the struct and return an error on failure.
4080 Frequency : SLES 9 only
4082 Description: Error of copying files with lustre special EAs as root
4083 Details : Client side always return success for setxattr call for lustre
4084 special xattr (currently only "trusted.lov").
4087 Frequency : rarely on clusters with both ia64+i386 clients
4089 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4090 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4091 clients are doing concurrent writes to the same file it is
4092 possible that the ia64 clients may LASSERT because the OST
4093 extent locks are not PAGE_SIZE aligned. Ensure that grown
4094 locks are always aligned on the request boundary.
4097 Frequency : specific use, occasional
4099 Description: Overwriting in use executable truncates on-disk binary image
4100 Details : If one node attempts to overwrite an executable in use by
4101 another node, we now correctly return ETXTBSY instead of
4102 truncating the file.
4104 Severity : enhancement
4106 Description: Async OSC create to avoid the blocking unnecessarily.
4107 Details : If an OST has no remaining object, system will block on the creating
4108 when need to create a new object on this OST. Now, ways use
4109 pre-created objects when available, instead of blocking on an
4110 empty osc while others are not empty. If we must block, we block
4111 for the shortest possible period of time.
4116 Description: chmod on Lustre root is propagated to other clients
4117 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4118 invalid by the follow_mount time.
4121 Frequency : liblustre clients only
4123 Description: Race in 'instant cancel' lock handling could lead to such locks
4124 never to be granted in case of SMP MDS
4125 Details : Do not destroy not yet granted but cbpending locks in
4129 Frequency : replay/resend of open
4131 Description: non null lock assetion failure in mds_intent_policy
4132 Details : Trying to replay/resend lockless open requests resulted in
4133 mds_open() returning 0 with no lock. Now it sets a flag if
4134 a lock is going to be returned.
4136 Severity : enhancement
4138 Description: Checksum enhancements
4139 Details : New checksum enhancements allow for resending RPCs that failed
4142 Severity : enhancement
4144 Description: Tunables on number of dirty pages in cacche
4145 Details : Allow to set limit on number of dirty pages cached.
4150 Description: client crash on unmount - lock still has references
4151 Details : In some error handling cases it was possible to leak a lock
4152 reference on a client while accessing a file. This was not
4153 harmful to the client during operation, but would cause the
4154 client to crash when the filesystem is unmounted.
4157 Frequency : specific case, rare
4159 Description: ETXTBSY on mds though file not in use by client
4160 Details : ETXTBSY is no longer incorrectly returned when attempting to
4161 chmod or chown a directory that the user previously tried to
4162 execute or a currently-executing binary.
4165 Frequency : extremely rare except on liblustre-based clients
4167 Description: Lustre space not freed when files are deleted
4168 Details : Clean up open-unlinked files after client eviction. Previously
4169 the unlink was skipped and the files remained as orphans.
4174 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4175 Details : In some cases it was possible to send a blocking callback to
4176 a client doing a glimpse, even though that client didn't get
4177 a lock granted. When the glimpse lock is cancelled on the OST
4178 the freed lock is left on the waiting list and corrupted the list.
4181 Frequency : all core dumps
4183 Description: Broke core dumps to lustre
4184 Details : Negative dentry may be unhashed if parent does not have UPDATE
4185 lock, but some callers, e.g. do_coredump, expect dentry to be
4186 hashed after successful create, hash it in ll_create_it.
4188 ------------------------------------------------------------------------------
4190 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4192 * Support for kernels:
4193 2.6.9-42.0.2.EL (RHEL 4)
4194 2.6.5-7.276 (SLES 9)
4195 2.4.21-40.EL (RHEL 3)
4196 2.6.12.6 vanilla (kernel.org)
4200 Frequency : always on RHEL 3
4202 Description: Number of open files grows over time
4203 Details : The number of open files grows over time, whether or not
4204 Lustre is started. This was due to a filp leak introduced
4205 by one of our kernel patches.
4207 ------------------------------------------------------------------------------
4209 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4211 * Support for kernels:
4212 2.6.9-42.EL (RHEL 4)
4213 2.6.5-7.276 (SLES 9)
4214 2.4.21-40.EL (RHEL 3)
4215 2.6.12.6 vanilla (kernel.org)
4220 Bugzilla : 5719, 9635, 9792, 9684
4221 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4222 Details : If a server is under heavy load and cannot reply to new
4223 connection requests before the client resends the (re)connect,
4224 the connection handling code can behave badly if two service
4225 threads are concurrently handing separate (re)connections from
4226 the same client. Add better locking to the connection handling
4227 code, and ensure that only a single connection will be processed
4228 for a given client UUID, even if the lock is dropped.
4230 Severity : enhancement
4232 Description: add TCP zero-copy support to kernel
4233 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4234 sends if the hardware supports scatter-gather and checksumming.
4235 This allows socklnd to do client-write and server-read more
4236 efficiently and reduce CPU utilization from skbuf copying.
4239 Frequency : only if NFS exporting from client
4241 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4242 Details : If setting attributes on a file created under NFS that had
4243 never been opened it would be possible to oops the client
4244 if the file had no objects.
4247 Frequency : always for liblustre
4249 Description: liblustre client does MDS+OSTs setattr RPC for each write
4250 Details : When doing a write from a liblustre client, the client
4251 incorrectly issued an RPC to the MDS and each OST the file was
4252 striped over in order to update the timestamps. When writing
4253 with small chunks and many clients this could overwhelm the MDS
4254 with RPCs. In all cases it would slow down the write because
4255 these RPCs are unnecessary.
4257 Severity : enhancement
4259 Description: allow number of MDS service threads to be changed at module load
4260 Details : It is now possible to change the number of MDS service threads
4261 running. Adding "options mds mds_num_threads={N}" to the MDS's
4262 /etc/modprobe.conf will set the number of threads for the next
4263 time Lustre is restarted (assuming the "mds" module is also
4264 reloaded at that time). The default number of threads will
4265 stay the same, 32 for most systems.
4270 Description: OST crash if filesystem is unformatted or corrupt
4271 Details : If an OST is started on a device that has never been formatted
4272 or if the filesystem is corrupt and cannot even mount then the
4273 error handling cleanup routines would dereference a NULL pointer.
4278 Description: NULL pointer deref in llap_from_page.
4279 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4280 incorrect) mapping if the page was truncated/reclaimed while it was
4281 searched for. Check for this condition and skip such pages when
4282 doing readahead. Introduce extra check to llap_from_page() to
4283 verify page->mapping->host is non-NULL (so page is not anonymous).
4286 Frequency : Sometimes when using sys_sendfile
4288 Description: "page not covered by a lock" warnings from ll_readpage
4289 Details : sendfile called ll_readpage without right page locks present.
4290 Now we introduced ll_file_sendfile that does necessary locking
4291 around call to generic_file_sendfile() much like we do in
4295 Frequency : with certain MDS communication failures at client mount time
4297 Description: NULL pointer deref after failed client mount
4298 Details : a client connection request may delayed by the network layer
4299 and not be sent until after the PTLRPC layer has timed out the
4300 request. If the client fails the mount immediately it will try
4301 to clean up before the network times out the request. Add a
4302 reference from the request import to the obd device and delay
4303 the cleanup until the network drops the request.
4306 Frequency : occasionally during client (re)connect
4308 Description: assertion failure during client (re)connect
4309 Details : processing a client connection request may be delayed by the
4310 client or server longer than the client connect timeout. This
4311 causes the client to resend the connection request. If the
4312 original connection request is replied in this interval, the
4313 client may trip an assertion failure in ptlrpc_connect_interpret()
4314 which thought it would be the only running connect process.
4317 Frequency : only with obd_echo servers and clients that are rebooted
4319 Description: kernel BUG accessing uninitialized data structure
4320 Details : When running an obd_echo server it did not start the ping_evictor
4321 thread, and when a client was evicted an uninitialized data
4322 structure was accessed. Start the ping_evictor in the RPC
4323 service startup instead of the OBD startup.
4325 Severity : enhancement
4326 Bugzilla : 10193 (patchless)
4327 Description: Remove dependency on various unexported kernel interfaces.
4328 Details : No longer need reparent_to_init, exit_mm, exit_files,
4329 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4332 Frequency : rare (only users of deprecated and unsupported LDAP config)
4334 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4335 Details : LDAP apparently contains 'lustreName' attributes instead of
4336 'name'. A simple remapping of the name is sufficient.
4339 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4341 Description: waiting_locks_callback trips kernel BUG if client is evicted
4342 Details : Running with the dump_on_timeout debug flag turned on makes
4343 it possible that the waiting_locks_callback() can try to dump
4344 the Lustre kernel debug logs from an interrupt handler. Defer
4345 this log dumping to the expired_lock_main() thread.
4347 Severity : enhancement
4349 Description: Support NFS exporting on 2.6 kernels.
4350 Details : Implement non-rawops metadata methods for NFS server to use without
4351 changing NFS server code.
4354 Frequency : very rare (synthetic metadata workload only)
4356 Description: two racing renames might cause an MDS thread to deadlock
4357 Details : Running the "racer" program may cause one MDS thread to rename
4358 a file from being the source of a rename to being the target of
4359 a rename at exactly the same time that another thread is doing
4360 so, and the second thread has already enqueued these locks after
4361 doing a lookup of the target and is trying to relock them in
4362 order. Ensure that we don't try to re-lock the same resource.
4365 Frequency : only very large systems with liblustre clients
4367 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4368 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4369 all OSTs in parallel. This allows the eviction of stale liblustre
4370 clients to proceed much faster than if they were done in series,
4371 and also offers similar improvements for other set_info RPCs.
4376 Description: excessive CPU usage during initial read phase on client
4377 Details : During the initial read phase on a client, it would agressively
4378 retry readahead on the file, consuming too much CPU and impacting
4379 performance (since 1.4.5.8). Improve the readahead algorithm
4380 to avoid this, and also improve some other common cases (read
4381 of small files in particular, where "small" is files smaller than
4382 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4387 Description: MDS crash when receiving packet with unknown intent.
4388 Details : Do not LBUG in unknown intent case, just return -EFAULT
4390 Severity : enhancement
4391 Bugzilla : 9293, 9385
4392 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4393 Details : Do not serialize getattr (non-intent version) and statfs.
4396 Frequency : occasional, when OST network is overloaded/intermittent
4398 Description: client evicted by OST after bulk IO timeout
4399 Details : If a client sends a bulk IO request (read or write) the OST
4400 may evict the client if it is unresposive to its data GET/PUT
4401 request. This is incorrect if the network is overloaded (takes
4402 too long to transfer the RPC data) or dropped the OST GET/PUT
4403 request. There is no need to evict the client at all, since
4404 the pinger and/or lock callbacks will handle this, and the
4405 client can restart the bulk request.
4408 Frequency : Always when mmapping file with no objects
4410 Description: client crashes when mmapping file with no objects
4411 Details : Check that we actually have objects in a file before doing any
4412 operations on objects in ll_vm_open, ll_vm_close and
4418 Description: Request leak when working with deleted CWD
4419 Details : Introduce advanced request refcount tracking for requests
4420 referenced from lustre intent.
4422 Severity : Enhancement
4424 Description: Cache open file handles on client.
4425 Details : MDS now will return special lock along with openhandle, if
4426 requested and client is allowed to hold openhandle, even if unused,
4427 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4428 closing files for every read/write openration.
4430 Severity : Enhancement
4432 Description: Cache open negative dentries on client when possible.
4433 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4434 negative dentries on lock revocation.
4439 Description: Remounting a client read-only wasn't possible with a zconf mount
4440 Details : It wasn't possible to remount a client read-only with llmount.
4442 Severity : enhancement
4443 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4444 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4445 MPICH is included to add Lustre-specific ADIO interfaces.
4446 This is based closely on the UFS ADIO layer and only differs
4447 in file creation, in order to allow the OST striping to be set.
4448 This is user-contributed code and not supported by CFS.
4453 Description: extended inode attributes (immutable, append-only) work improperly
4454 when 2.4 and 2.6 kernels are used on client/server or vice versa
4455 Details : Introduce kernel-independent values for these flags.
4457 Severity : enhancement
4460 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4461 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4462 granularity. Now, Fractional MB granularity is be supported,
4463 this is very useful for low memory system.
4465 Severity : enhancement
4467 Description: Getattr by fid
4468 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4469 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4470 have locks and avoids getattr rpc requests for it on every stat.
4473 Frequency : Always, for filesystems larger than 2TB
4475 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4476 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4477 filesystems larger than 2TB to be mounted with mballoc enabled.
4480 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4482 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4483 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4484 filesystems larger than 2TB is mounted then the kernel will
4485 silently corrupt the start of the filesystem. CONFIG_LBD is
4486 enabled for all CFS-supported kernels, but the possibility of
4487 this happening with a modified kernel config exists.
4489 Severity : enhancement
4491 Description: add client O_DIRECT support for 2.6 kernels
4492 Details : It is now possible to do O_DIRECT reads and writes to files
4493 in the Lustre client mountpoint on 2.6 kernel clients.
4495 Severity : enhancement
4497 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4498 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4499 parallel, not waiting for response from every OST before sending
4500 a rpc to the next OST.
4505 Description: setattr vs write race when updating file timestamps
4506 Details : Client processes that update a file timestamp into the past
4507 right after writing to the file (e.g. tar) it is possible that
4508 the updated file modification time can be reset to the current
4509 time due to a race between processing the setattr and write RPC.
4511 Severity : enhancement
4513 Description: Bring 'lfs find' closer in line with regular Linux find.
4514 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4515 -print0 options and obtains all the needed info through the lustre
4518 Severity : enhancement
4520 Description: support up to 1024 configured devices on one node
4521 Details : change obd_dev array from statically allocated to dynamically
4522 allocated structs as they are first used to reduce memory usage
4527 Description: Flush dirty partially truncated pages during truncate
4528 Details : Immediatelly flush partially truncated pages in filter_setattr,
4529 this way we completely avoid having any pages in page cache on OST
4530 and can retire ugly workarounds during writes to flush such pages.
4535 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4536 Details : protect inode from truncation within vfs_unlink() context
4537 just take a reference before calling vfs_unlink() and release it
4538 when parent's i_sem is free.
4543 Description: last_id value checked outside lock on OST caused LASSERT failure
4544 Details : If there were multiple MDS->OST object precreate requests in
4545 flight, it was possible that the OST's last object id was checked
4546 outside a lock and incorrectly tripped an assertion. Move checks
4547 inside locks, and discard old precreate requests.
4550 Frequency : always, if extents are used on OSTs
4552 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4553 Details : OSTs using the extents format would not zero the high 16 bits of
4554 the index physical block number. This is not a problem for any
4555 OST filesystems smaller than 16TB, and no kernels support ext3
4556 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4557 new/modified files) and can be fixed for existing filesystems
4558 with e2fsprogs-1.39-cfs1.
4563 Description: import connection selection may be incorrect if timer wraps
4564 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4565 import connections to be ignored if the 32-bit jiffies counter
4566 wraps. Use a 64-bit jiffies counter.
4569 Frequency : very large clusters immediately after boot
4571 Description: LNET request buffers exhausted under heavy short-term load
4572 Details : If a large number of client requests are generated on a service
4573 that has previously never seen so many requests it is possible
4574 that the request buffer growth cannot keep up with the spike in
4575 demand. Instead of dropping incoming requests, they are held in
4576 the LND until the RPC service can accept more requests.
4579 Frequency : Sometimes during replay
4581 Description: Assertion failure in ll_local_open after replay.
4582 Details : If replay happened on an open request reply before we were able
4583 to set replay handler, reply will become not swabbed tripping the
4584 assertion in ll_local_open. Now we set the handler right after
4585 recognising of open request
4588 Frequency : very rare
4590 Description: kernel reports "badness in vsnprintf"
4591 Details : Reading from the "recovery_status" /proc file in small chunks
4592 may cause a negative length in lprocfs_obd_rd_recovery_status()
4593 call to vsnprintf() (which is otherwise harmless). Exit early
4594 if there is no more space in the output buffer.
4596 Severity : enhancement
4598 Description: clear OBD RPC statistics by writing to them
4599 Details : It is now possible to clear the OBD RPC statistics by writing
4600 to the "stats" file.
4605 Description: Client mtime is not the same on different clients after utimes
4606 Details : In some cases, the client was using the utimes() syscall on
4607 a file cached on another node. The clients now validate the
4608 ctime from the MDS + OSTs to determine which one is right.
4613 Description: Inability to activate failout mode
4614 Details : lconf script incorrectly assumed that in pythong string's numeric
4615 value is used in comparisons.
4618 Frequency : always with multiple stripes per file
4620 Description: Inefficient object allocation for mutli-stripe files
4621 Details : When selecting which OSTs to stripe files over, for files with
4622 a stripe count that divides evenly into the number of OSTs,
4623 the MDS is always picking the same starting OST for each file.
4624 Return the OST selection heuristic to the original design.
4629 Description: mount failures may take full timeout to return an error
4630 Details : Under some heavy load conditions it is possible that a
4631 failed mount can wait for the full obd_timeout interval,
4632 possibly several minutes, before reporting an error.
4633 Instead return an error as soon as the status is known.
4635 Frequency : quota enabled and large files being deleted
4637 Description: releasing more than 4GB of quota at once hangs OST
4638 Details : If a user deletes more than 4GB of files on a single OST it
4639 will cause the OST to spin in an infinite loop. Release
4640 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4645 Description: statfs data retrieved from /proc may be stale or zero
4646 Details : When reading per-device statfs data from /proc, in the
4647 {kbytes,files}_{total,free,avail} files, it may appear
4648 as zero or be out of date.
4651 Frequency : systems with MD RAID1 external journal devices
4653 Description: lconf's call to blkid is confused by RAID1 journal devices
4654 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4655 of returning all block devices that match the journal UUID.
4658 Frequency : always, for aggregate stripe size over 4GB
4660 Description: assertion fails when trying to use 4GB stripe size
4661 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4662 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4665 Frequency : always on ppc64
4667 Description: the first write on an ext3 filesystem with mballoc got stuck
4668 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4669 perform endianness conversion.
4671 ------------------------------------------------------------------------------
4673 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4675 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4676 INTEROPERATE with older versions automatically. Please read the
4677 user documentation before upgrading any part of a live system.
4678 * WARNING: Lustre networking configuration changes are required with
4679 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4682 * Support for kernels:
4683 2.6.9-22.0.2.EL (RHEL 4)
4684 2.6.5-7.244 (SLES 9)
4685 2.6.12.6 vanilla (kernel.org)
4688 Severity : enhancement
4689 Bugzilla : 7981/8208
4690 Description: Introduced Lustre Networking (LNET)
4691 Details : LNET is new networking infrastructure for Lustre, it includes
4692 a reorganized network configuration mode (see the user
4693 documentation for full details) as well as support for routing
4694 between different network fabrics. Lustre Networking Devices
4695 (LNDs) for the supported network fabrics have also been
4696 created for this new infrastructure.
4698 Severity : enhancement
4699 Description: Introduced Access control lists
4700 Details : clients can set ACLs on files and directories in order to have
4701 more fine-grained permissions than the standard Unix UGO+RWX.
4702 The MDS must be started with the "-o acl" mount option.
4704 Severity : enhancement
4705 Description: Introduced filesystem quotas
4706 Details : Administrators may now establish per-user quotas on the
4709 Severity : enhancement
4711 Description: Configuration change for the XT3
4712 The PTLLND is now used to run Lustre over Portals on the XT3
4713 The configure option(s) --with-cray-portals are no longer used.
4714 Rather --with-portals=<path-to-portals-includes> is used to
4715 enable building on the XT3. In addition to enable XT3 specific
4716 features the option --enable-cray-xt3 must be used.
4721 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4722 Details : A race between allocating a new llcd and re-getting the llcd_lock
4723 allowed another thread to grab newly-allocated llcd.
4725 Severity : enhancement
4727 Description: 2.6 OST async journal commit and locking fix to improve performance
4728 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4729 2.6 kernels are now async as they already were in 2.4 kernels so
4730 that they can commit concurrently with the network bulk transfer.
4731 For block-allocated files the filter allocation semaphore is held
4732 to avoid filesystem fragmentation during allocation. BKL lock
4733 removed for 2.6 xattr operations where it is no longer needed.
4738 Description: lconf incorrectly determined whether two IP networks could talk
4739 Details : In some more complicated routing and multiple-network
4740 configurations, lconf will avoid trying to make a network
4741 connection to a disjoint part of the IP space. It was doing the
4742 math incorrectly for one set of cases.
4747 Description: Fix for potential infinite loop processing records in an llog.
4748 Details : If an llog record is corrupted/zeroed, it is possible to loop
4749 forever in llog_process(). Validate the llog record length
4750 and skip the remainder of the block on error.
4753 Frequency : occasional (liblustre only)
4755 Description: liblustre could not open files whose last component is a symlink
4756 Details : sysio_path_walk() would incorrectly pass the open intent to
4757 intermediate path components.
4760 Frequency : rare (liblustre only with non-standard tuning)
4761 Bugzilla : 7201 (7350)
4762 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4763 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4764 after it has been released, only do this for referenced locks
4766 Severity : enhancement
4768 Description: specifying an (invalid) directory default stripe_size of -1
4769 would reset the directory default striping
4770 Details : stripe_size -1 was used internally to signal directory stripe
4771 removal, now use "all default" to signal dir stripe removal
4772 as a directory striping of "all default" is not useful
4775 Frequency : common for large clusters running liblustre clients
4777 Description: doing an ls when liblustre clients are running is slow
4778 Details : sending a glimpse AST to a liblustre client waits for every AST
4779 to time out, as liblustre clients will not respond. Since they
4780 cannot cache data we refresh the OST lock LVB from disk instead.
4782 Severity : enhancement
4784 Description: doing an ls at the same time as file IO can be slow
4785 Details : enqueue and other "small" requests can be blocked behind many
4786 large IO requests. Create a new OST IO portal for non-IO
4787 requests so they can be processed faster.
4790 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4792 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4793 created file with mode 000
4794 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4795 allow an ftruncate() as a non-root user to a file with mode 000.
4796 Lustre now respects this flag to disable mode checks when
4797 truncating a file owned by the user
4800 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4803 Description: Revoking locks from clients that went dead or catatonic might take
4805 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4806 cancellation of such locks instant on servers without waiting for
4807 any reply from clients. Clients drops these locks when cancel
4808 notification from server is received without replying.
4811 Frequency : liblustre-only, when liblustre client dies or becomes busy
4813 Description: Doing ls on Linux clients can take a long time with active
4815 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4816 granting such locks to it in the first place if possible. Locks
4817 are taken by proxy on the OST during the read or write and
4818 dropped immediately afterward. Add connect flags handling, do
4819 not grant locks to liblustre clients for glimpse ASTs.
4821 Severity : enhancement
4823 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4824 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4825 sequential reads and behaves badly with applications doing random
4826 reads. Improve it by reading ahead at least read region, and
4827 avoiding excessive large RPC for small reads.
4829 Severity : enhancement
4831 Description: Creating more than 1000 files for a single job may cause a load
4832 imbalance on the OSTs if there are also a large number of OSTs.
4833 Details : qos_prep_create() uses an OST index reseed value that is an
4834 even multiple of the number of available OSTs so that if the
4835 reseed happens in the middle of the object allocation it will
4836 still utilize the OSTs as uniformly as possible.
4841 Description: OST or MDS may oops in ping_evictor_main()
4842 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4843 but doesn't restart at beginning of obd_exports_timed list
4846 Severity : enhancement
4848 Description: improve by-nid export eviction on the MDS and OST
4849 Details : allow multiple exports with the same NID to be evicted at one
4850 time without re-searching the exports list.
4853 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4855 Description: MDS may oops in groups_free()
4856 Details : in rare race conditions a newly allocated group_info struct is
4857 freed again, and this can be NULL. The 2.4 compatibility code
4858 for groups_free() checked for a NULL pointer, but 2.6 did not.
4861 Frequency : common for liblustre clients doing little filesystem IO
4862 Bugzilla : 9352, 7313
4863 Description: server may evict liblustre clients accessing contended locks
4864 Details : if a client is granted a lock or receives a completion AST
4865 with a blocking AST already set it would not reply to the AST
4866 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4869 Frequency : lfs setstripe, only systems with more than 160 OSTs
4871 Description: unable to set striping with a starting offset beyond OST 160
4872 Details : llapi_create_file() incorrectly limited the starting stripe
4873 index to the maximum single-file stripe count.
4876 Frequency : LDAP users only
4878 Description: lconf did not handle in-kernel recovery with LDAP properly
4879 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4881 Severity : enhancement
4883 Description: bind OST threads to NUMA nodes to improve performance
4884 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4885 node and do their allocations there to localize memory access
4887 Severity : enhancement
4889 Description: llmount can determine client NID directly from Myrinet (GM)
4890 Details : the client NID code from gmnalnid was moved directly into
4891 llmount, removing the need to use this or specifying the
4892 client NID explicitly when mounting GM clients with zeroconf
4895 Frequency : if client is started with down MDS
4897 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4898 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4899 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4900 ptlrpc_import_delayed_req() because we didn't check if the
4901 request was interrupted, and we also didn't break out of the
4902 event loop if there was no timeout
4907 Description: data loss during non-page-aligned writes to a single file from
4908 both multiple nodes and multiple threads on one node at same time
4909 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4910 inconsistency led to false short-reads, that were cached and later
4911 used by ->prepare_write() to fill in partially written page,
4912 leading to data loss.
4915 Frequency : always, if lconf --abort_recovery used
4917 Description: lconf --abort_recovery fails with 'Operation not supported'
4918 Details : lconf was attempting to abort recovery on the MDT device and not
4921 Severity : enhancement
4923 Description: remove cleanup logs
4924 Details : replace lconf-generated cleanup logs with lustre internal
4925 cleanup routines. Eliminates the need for client-cleanup and
4928 Severity : enhancement
4930 Description: add support for EAs (user and system) on lustre filesystems
4931 Details : it is now possible to store extended attributes in the Lustre
4932 client filesystem, and with the user_xattr mount option it
4933 is possible to allow users to store EAs on their files also
4935 Severity : enhancement
4937 Description: Add possibility (config option) to show minimal available OST free
4939 Details : When compiled with --enable-mindf configure option, statfs(2)
4940 (and so, df) will return least minimal free space available from
4941 all OSTs as amount of free space on FS, instead of summary of
4942 free spaces of all OSTs.
4944 Severity : enhancement
4946 Description: do not expand extent locks acquired on OST-side
4947 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4948 by server: they are not cached anyway.
4951 Frequency : when mmap is used/binaries executed from Lustre
4953 Description: Unmmap pages before throwing them away from read cache.
4954 Details : llap_shrink cache now attempts to unmap pages before discarding
4955 them (if unmapping failed - do not discard). SLES9 kernel has
4956 extra checks that trigger if this unmapping is not done first.
4961 Description: lconf didn't resolve symlinks before checking to see whether a
4962 given mountpoint was already in use
4965 Frequency : when migrating failover services
4966 Bugzilla : 6395, 9514
4967 Description: When migrating a subset of services from a node (e.g. failback
4968 from a failover service node) the remaining services would
4969 time out and evict clients.
4970 Details : lconf --force (implied by --failover) sets the global obd_timeout
4971 to 5 seconds in order to quickly disconnect, but this caused
4972 other RPCs to time out too quickly. Do not change the global
4973 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4975 Severity : enhancement
4976 Frequency : if MDS is started with down OST
4977 Bugzilla : 9439,5706
4978 Description: Allow startup/shutdown of an MDS without depending on the
4979 availability of the OSTs.
4980 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4981 Add appropriate locking and lov-osc refcounts for safe
4982 cleaning. Add osc abort_inflight calls in case the
4983 synchronize never started.
4986 Frequency : occasional (Cray XT3 only)
4988 Description: root not authorized to access files in CRAY_PORTALS environment
4989 Details : The client process capabilities were not honoured on the MDS in
4990 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4991 been accessed by an authorized user then root was able to access
4992 the file on the local client also. The root user capabilities
4993 are now allowed on the MDS, as this environment has secure UID.
4996 Frequency : occasional
4998 Description: ldiskfs "too long searching" message happens too often
4999 Details : A debugging message (otherwise harmless) prints too often on
5000 the OST console. This has been reduced to only happen when
5001 there are fragmentation problems on the filesystem.
5006 Description: Division by zero in statfs when all OSCs are inactive
5007 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5008 lov_statfs divides by value returned by lov_get_stripecnt().
5012 Bugzilla : 9489, 3273
5013 Description: First write from each client to each OST was only 4kB in size,
5014 to initialize client writeback cache, which caused sub-optimal
5015 RPCs and poor layout on disk for the first written file.
5016 Details : Clients now request an initial cache grant at (re)connect time
5017 and so that they can start streaming writes to the cache right
5018 away and always do full-sized RPCs if there is enough data.
5019 If the OST is rebooted the client also re-establishes its grant
5020 so that client cached writes will be honoured under the grant.
5025 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5026 Details : Now I/O RPCs go to different portal number and (presumably) fast
5027 lock requests (and glimses) and other RPCs get their own service
5028 threads pool that should be able to service those RPCs
5031 Severity : enhancement
5033 Description: Ability to exchange lustre version between client and servers and
5034 issue warnings at client side if client is too old. Also for
5035 liblustre clients there is ability to refuse connection of too old
5037 Details : New 'version' field is added to connect data structure that is
5038 filled with version info. That info is later checked by server and
5042 Frequency : rare, liblustre only.
5043 Bugzilla : 9296, 9581
5044 Description: Two simultaneous writes from liblustre at offset within same page
5045 might proceed at the same time overwriting eachother with stale
5047 Details : I/O lock withing llu_file_prwv was released too early, before data
5048 actually was hitting the wire. Extended lock-holding time until
5049 server acknowledges receiving data.
5052 Frequency : extremely rare. Never observed in practice.
5054 Description: avoid generating lustre_handle cookie of 0.
5055 Details : class_handle_hash() generates handle cookies by incrementing
5056 global counter, and can hit 0 occasionaly (this is unlikely, but
5057 not impossible, because initial value of cookie counter is
5058 selected randonly). Value of 0 is used as a sentinel meaning
5059 "unassigned handle" --- avoid it. Also coalesce two critical
5060 sections in this function into one.
5062 Severity : enhancement
5064 Description: allow liblustre clients to delegate truncate locking to OST
5065 Details : To avoid overhead of locking, liblustre client instructs OST to
5066 take extent lock in ost_punch() on client's behalf. New connection
5067 flag is added to handle backward compatibility.
5069 Severity : enhancement
5070 Bugzilla : 4928, 7341, 9758
5071 Description: allow number of OST service threads to be specified
5072 Details : a module parameter allows the number of OST service threads
5073 to be specified via "options ost ost_num_threads={N}" in the
5074 OSS's /etc/modules.conf or /etc/modprobe.conf.
5078 Bugzilla : 6146, 9635, 9895
5079 Description: servers crash with bad pointer in target_handle_connect()
5080 Details : In rare cases when a client is reconnecting it was possible that
5081 the connection request was the last reference for that export.
5082 We would temporarily drop the export reference and get a new
5083 one, but this may have been the last reference and the export
5084 was just destroyed. Get new reference before dropping old one.
5086 Severity : enhancement
5087 Frequency : if client is started with failover MDS
5089 Description: Allow multiple MDS hostnames in the mount command
5090 Details : Try to read the configuration from all specified MDS
5091 hostnames during a client mount in case the "primary"
5094 Severity : enhancement
5096 Description: Stop sending data to evicted clients as soon as possible.
5097 Details : Check if the client we are about to send or are sending data to
5098 was evicted already. (Check is done every second of waiting,
5099 for which l_wait_event interface was extended to allow checking
5100 of exit condition at specified intervals).
5103 Frequency : rare, normally only when NFS exporting is done from client
5105 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5107 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5110 Frequency : always, if srand() is called before liblustre initialization
5112 Description: Liblustre uses system PRNG disturbing its usage by user application
5113 Details : Introduce internal to lustre fast and high-quality PRNG for
5114 lustre usage and make liblustre and some other places in generic
5115 lustre code to use it.
5117 Severity : enhancement
5118 Bugzilla : 9477, 9557, 9870
5119 Description: Verify that the MDS configuration logs are updated when xml is
5120 Details : Check if the .xml configuration logs are newer than the config
5121 logs stored on the MDS and report an error if this is the case.
5122 Request --write-conf, or allow starting with --old_conf.
5124 Severity : enhancement
5126 Description: Handle symlinks in the path when checking if Lustre is mounted.
5127 Details : Resolve intermediate symlinks when checking if a client has
5128 mounted a filesystem to avoid duplicate client mounts.
5133 Description: lconf can hit an error exception but still return success.
5134 Details : The lconf command catches the Command error exception at the top
5135 level script context and will exit with the associated exit
5136 status, but doesn't ensure that this exit status is non-zero.
5141 Description: failure of ptlrpc thread startup can cause oops
5142 Details : Starting a ptlrpc service thread can fail if there are a large
5143 number of threads or the server memory is very fragmented.
5144 Handle this without oopsing.
5147 Frequency : always, only if liblustre and non-default acceptor port was used
5149 Description: liblustre cannot connect to servers with non-default acceptor port
5150 Details : tcpnal_set_default_params() was not called and was therefore
5151 ignoring the environment varaible TCPNAL_PORT, as well as other
5152 TCPNAL_ environment variables
5157 Description: two objects could be created on the same OST for a single file
5158 Details : If an OST is down, in some cases it was possible to create two
5159 objects on a single OST for a single file. No problems other
5160 than potential performance impact and spurious error messages.
5164 Bugzilla : 5681, 9562
5165 Description: Client may oops in ll_unhash_aliases
5166 Details : Client dcache may become inconsistent in race condition.
5167 In some cases "getcwd" can fail if the current directory is
5173 Description: Inode refcounting problems in NFS export code
5174 Details : link_raw functions used to call d_instantiate without obtaining
5175 extra inode reference first.
5179 Bugzilla : 9942, 9903
5180 Description: Referencing freed requests leading to crash, memleaks with NFS.
5181 Details : We used to require that call to ll_revalidate_it was always
5182 followed by ll_lookup_it. Also with revalidate_special() it is
5183 possible to call ll_revalidate_it() twice for the same dentry
5184 even if first occurence returned success. This fix changes semantic
5185 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5186 reference on a request referred from the intent.
5187 ll_intent_release() then releases such a request.
5190 Frequency : rare, normally benchmark loads only
5192 Description: unlinked inodes were kept in memory on the client
5193 Details : If a client is repeatedly creating and unlinking files it
5194 can accumulate a lot of stale inodes in the inode slab cache.
5195 If there is no other client load running this can cause the
5196 client node to run out of memory. Instead flush old inodes
5197 from client cache that have the same inode number as a new inode.
5200 Frequency : SLES9 2.6.5 kernel and long filenames only
5201 Bugzilla : 9969, 10379
5202 Description: utime reports stale NFS file handle
5203 Details : SLES9 uses out-of-dentry names in some cases, which confused
5204 the lustre dentry revalidation. Change it to always use the
5208 Frequency : rare, unless heavy write-truncate concurrency is continuous
5209 Bugzilla : 4180, 6984, 7171, 9963, 9331
5210 Description: OST becomes very slow and/or deadlocked during object unlink
5211 Details : filter_destroy() was holding onto the parent directory lock
5212 while truncating+unlinking objects. For very large objects this
5213 may block other threads for a long time and slow overall OST
5214 responsiveness. It may also be possible to get a lock ordering
5215 deadlock in this case, or run out of journal credits because of
5216 the combined truncate+unlink. Solution is to do object truncate
5217 first in one transaction without parent lock, and then do the
5218 final unlink in a new transaction with the parent lock. This
5219 reduces the lock hold time dramatically.
5222 Frequency : rare, 2.4 kernels only
5224 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5225 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5226 before unmounting their filesystem in order to avoid deadlock.
5227 The kernel_locked() function in 2.4 kernels only checks whether
5228 the kernel lock is held, not whether it is this process that is
5229 holding it as 2.6 kernels do.
5234 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5235 Details : The client ptlrpc code may be trying to reconnect to a down
5236 server before a previous connection attempt has timed out.
5237 Increase the reconnect interval to be longer than the connection
5238 timeout interval to avoid sending duplicate connections to
5242 Frequency : echo_client brw_test command
5244 Description: fix echo_client to work with OST preallocated code
5245 Details : OST preallocation code (5137) didn't take echo_client IO path
5246 into account: echo_client calls filter methods outside of any
5247 OST thread and, hence, there is no per-thread preallocated
5248 pages and buffers to use. Solution: hijack pga pages for IO. As
5249 a byproduct, this avoids unnecessary data copying.
5253 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5254 Description: Client can oops in mdc_commit_close() after open replay
5255 Details : It was possible for the MDS to return an open request with no
5256 transaction number in mds_finish_transno() if the client was
5257 evicted, but without actually returning an error. Clients
5258 would later try to replay that open and may trip an assertion
5259 Simplify the client close codepath, and always return an error
5260 from the MDS in case the open is not successful.
5263 Frequency : rare, 2.6 OSTs only
5265 Description: OST may deadlock under high load on fragmented files
5266 Details : If there was a heavy load and highly-fragmented OST filesystems
5267 it was possible to have all the OST threads deadlock waiting on
5268 allocation of biovecs, because the biovecs were not released
5269 until the entire RPC IO was completed. Instead, release biovecs
5270 as soon as they are complete to ensure forward IO progress.
5272 Severity : enhancement
5274 Description: Support for specifying external journal device at mount
5275 Details : If an OST or MDS device is formatted with an external journal
5276 device, this device major/minor is stored in the ext3 superblock
5277 and may not be valid for failover. Allow detecting and
5278 specifying the external journal at mount time.
5283 Description: Mounting an MDS with pending unlinked files may cause oops
5284 Details : target_finish_recovery() calls mds_postrecov() which returned
5285 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5286 considers this an error and immediately begins cleaning up the
5287 lov, just after starting the mds_lov process
5289 Severity : enhancement
5291 Description: Implement 'lfs df' to report actual free space on per-OST basis
5292 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5293 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5294 usage of objects; '-h' to report in human readable format.
5296 ------------------------------------------------------------------------------
5298 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5305 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5306 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5307 at startup during orphan unlinking. Instead, skip buddy update
5308 before setup, as it will be regenerated after recovery is complete.
5313 Description: If an OST is inactive, its locks might reference stale inodes.
5314 Details : lov_change_cbdata() must iterate over all namespaces, even if
5315 they are inactive to clear inode references from the lock.
5317 Severity : enhancement
5318 Frequency : occasional, if non-standard max_dirty_mb used
5320 Description: Client will block write RPCs if not enough grant
5321 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5322 then the client will block writes while waiting for another RPC
5323 to complete instead of consuming its dirty limit. With change
5324 we get improved performance when max_dirty_mb is small.
5326 Severity : enhancement
5327 Bugzilla : 3389, 6253
5328 Description: Add support for supplementary groups on the MDS.
5329 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5330 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5331 lookups for user supplementary groups into a cache.
5335 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5336 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5337 for liblustre O_EXCL is enforced by MDS.
5341 Description: readdir never returns NULL in liblustre.
5342 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5343 dentry in struct dirent.
5347 Frequency : liblustre only, depends on application IO pattern
5348 Description: liblustre clients evicted if not contacting servers
5349 Details : Don't put liblustre clients into the ping_evictor list, so
5350 they will not be evicted by the pinger ever.
5352 Severity : enhancement
5354 Description: Add ability to evict clients by NID from MDS.
5355 Details : By echoing "nid:$NID" string into
5356 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5357 $NID would be instantly evicted from this MDS and from all active
5358 OSTs connected to it.
5362 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5363 Details : lookup_it_finish() used to query file size from OSTs that was not
5368 Description: service threads change working directory to that of init
5369 Details : Starting lustre service threads may pin the working directory
5370 of the parent thread, making that filesystem busy. Threads
5371 now change to the working directory of init to avoid this.
5375 Frequency : during shutdown only
5376 Description: shutdown with a failed MDS or OST can cause unmount to hang
5377 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5382 Frequency : 2.6 only
5383 Description: chmod/chown may include an extra supplementary group
5384 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5385 supplementary group and if none is specified this is used.
5388 Bugzilla : 5479 (6816)
5390 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5391 Details : If lookup is in progress on a file that is unlinked we might try
5392 to revalidate the inode and fail in revalidate after lookup is
5393 complete and ll_file_open() enqueues the open again but
5394 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5397 Frequency : always, if lconf --abort_recovery used
5399 Description: lconf --abort_recovery fails with 'Operation not supported'
5400 Details : lconf was attempting to abort recovery on the MDT device and not
5403 ------------------------------------------------------------------------------
5405 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5410 Frequency : rare (only unsupported configurations with a node running as an
5412 Bugzilla : 6514, 5137
5413 Description: Mounting a Lustre file system on a node running as an OST could
5415 Details : OSTs now preallocates memory needed to write out data at
5416 startup, instead of when needed, to avoid having to
5417 allocate memory in possibly low memory situations.
5418 Specifically, if the file system is mounted on on OST,
5419 memory pressure could force it to try to write out data,
5420 which it needed to allocate memory to do. Due to the low
5421 memory, it would be unable to do so and the node would
5422 become unresponsive.
5424 Severity : enhancement
5426 Description: Addition of lconf --service command line option
5427 Details : lconf now accepts a '--service <arg>' option, which is
5428 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5430 Severity : enhancement
5432 Description: Failover mode is now the default for OSTs.
5433 Details : By default, OSTs will now run in failover mode. To return to
5434 the old behaviour, add '--failout' to the lmc line for OSTs.
5436 Severity : enhancement
5438 Description: Health checks are now provided for MDS and OSTs
5439 Details : Additional detailed health check information on MSD and OSTs
5440 is now provided through the procfs health_check value.
5443 Frequency : occasional, depends on IO load
5445 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5446 after numerous create/delete cycles
5447 Details : The ext3 inode allocation policy would not allocate new inodes
5448 very well on the OSTs because there are no new directories
5449 being created. Instead we look for groups with free space if
5450 the parent directories are nearly full.
5455 Description: Network or server problems during mount may cause partially
5456 mounted clients instead of returning an error.
5457 Details : The config llog parsing code may overwrite the error return
5458 code during mount error handling, returning success instead
5464 Description: MDS can fail to allocate large reply buffers
5465 Details : After long uptimes the MDS can fail to allocate large reply
5466 buffers (e.g. zconf client mount config records) due to memory
5467 fragmentation or consumption by the buffer cache. Preallocate
5468 some large reply buffers so that these replies can be sent even
5469 under memory pressure.
5473 Frequency : rare (liblustre)
5474 Description: fsx running with liblustre complained that using truncate() to
5475 extend the file doesn't work. This patch corrects that issue.
5476 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5477 ATTR_SIZE and lsm use in llu_setattr_raw.
5481 Frequency : rare, only 2.6 kernels
5482 Description: Unusual file access patterns on the MDS may result in inode
5483 data being lost in very rare circumstances.
5484 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5485 code in the 2.6 kernel caused the inode and/or EA data not to
5486 be read from disk, causing single-file corruption.
5490 Frequency : rare, only 2.6 filesystems using extents
5491 Description: Heavy concurrent write and delete load may cause data corruption.
5492 Details : It was possible under high-load situations to have an extent
5493 metadata block in the block device cache from a just-unlinked
5494 file overwrite a newly-allocated data block. We now unmap any
5495 metadata buffers that alias just-allocated data blocks.
5499 Frequency : filesystems with default stripe_count larger than 77
5500 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5501 Details : lconf specifies an inode size of 4096 bytes when the default
5502 stripe_count is larger than 77. This conflicts with the default
5503 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5504 this case to avoid pinning too much memory for large EAs.
5506 ------------------------------------------------------------------------------
5508 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5513 Frequency : rare (extremely heavy IO load with hundreds of clients)
5515 Description: Client is evicted, gets IO error writing to file
5516 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5517 caused clients to be evicted for AST timeouts. The fixes in
5518 bug 5192 mean we no longer need to have such short AST timeouts
5519 so ldlm_timeout has been increased.
5522 Frequency : occasional during --force or --failover shutdown under load
5523 Bugzilla : 5949, 4834
5524 Description: Server oops/LBUG if stopped with --force or --failover under load
5525 Details : a collection of import/export refcount and cleanup ordering
5526 issues fixed for safer force cleanup
5529 Frequency : only filesystems larger than 120 OSTs
5530 Bugzilla : 5990, 6223
5531 Description: lfs getstripe would oops on a very large filesystem
5532 Details : lov_getconfig used kfree on vmalloc'd memory
5535 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5536 Bugzilla : 6242, 6243
5537 Description: reading from files that had been truncated to a non-zero size
5538 but never opened returned no data
5539 Details : ll_file_read() reads zeros from no-object files to EOF
5544 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5545 Details : There's little harm in aborting MDS/OSS recovery and letting it
5546 try again, so I removed the LASSERT and return an error instead.
5548 Severity : enhancement
5550 Description: New debugging infrastructure for tracking down data corruption
5551 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5552 (b) cover more of the client-side code path, and (c) try to narrow
5553 down where problems occurred
5557 Bugzilla : 3819, 4364, 4397, 6313
5558 Description: Racing close and eviction MDS could cause assertion in mds_close
5559 Details : It was possible to get multiple mfd references during close and
5560 client eviction, leading to one thread referencing a freed mfd.
5562 Severity: : enhancement
5563 Bugzilla : 3262, 6359
5564 Description: Attempts to reconnect to servers are now more aggressive.
5565 Details : This builds on the enhanced upcall-less recovery that was added
5566 in 1.4.2. When trying to reconnect to servers, clients will
5567 now try each server in the failover group every 10 seconds. By
5568 default, clients would previously try one server every 25 seconds.
5573 Description: After recovery, certain operations trigger a failed
5574 assertion on a client.
5575 Details : Failing over an mds, using lconf -d --failover, while a
5576 client was doing a readdir() call would cause the client to
5577 LBUG after recovery completed and the readdir() was resent.
5579 Severity : enhancement
5581 Description: Default groups are now added by lconf
5582 Details : You can now run lconf --group <servicename> without having to
5583 manually add groups with lmc.
5586 Frequency : occasional
5588 Description: Nodes with an elan id of 0 trigger a failed assertion
5591 Frequency : always when accessing e.g. tty/console device nodes
5593 Description: tty and some other devices nodes cannot be used on lustre
5594 Details : file's private_data field is used by device data and lustre
5595 values in there got lost. New field was added to struct file to
5596 store fs-specific private data.
5599 Frequency : when exporting Lustre via NFS
5601 Description: NFSD failed occasionally when looking up a path component
5602 Details : NFSD is looking up ".." which was broken in ext3 directories
5603 that had grown large enough to become hashed.
5606 Frequency : Clusters with multiple interfaces not on the same subnet
5608 Description: Nodes will repeatedly try to reconnect to an interface which it
5609 cannot reach and report an error to the log.
5610 Details : Extra peer list entries will be created by lconf with some peers
5611 unreachable. lconf now validates the peer before adding it.
5614 Frequency : Only if a default stripe is set on the filesystem root.
5616 Description: Setting a default stripe on the filesystem root prevented the
5617 filesystem from being remounted.
5618 Details : The client was sending extra request flags in the root getattr
5619 request and did not allocate a reply buffer for the dir EA.
5622 Frequency : occasional, higher if lots of files are accessed by one client
5623 Bugzilla : 6159, 6097
5624 Description: Client trips assertion regarding lsm mismatch/magic
5625 Details : While revalidating inodes the VFS looks up inodes with ifind()
5626 and in rare cases can find an inode that is being freed.
5627 The ll_test_inode() code will free the lsm during ifind()
5628 when it finds an existing inode and then the VFS later attaches
5629 this free lsm to a new inode.
5633 Bugzilla : 6422, 7030
5634 Description: MDS deadlock between mkdir and client eviction
5635 Details : Creating a new file via mkdir or mknod (starting a transaction
5636 and getting the ns lock) can deadlock with client eviction
5637 (gets ns lock and trying to finish a synchronous transaction).
5640 Frequency : occasional
5641 Description: While starting a server, the fsfilt_ext3 module could not be
5643 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5644 kernels. Previously, lconf would still use the ext3 name
5645 when trying to load modules. Now, it will correctly use
5646 ext3 on 2.4 and ldiskfs on 2.6.
5648 Severity : enhancement
5649 Description: The default stripe count has been changed to 1
5650 Details : The interpretation of the default stripe count (0, to lfs
5651 or lmc) has been changed to mean striping across a single
5652 OST, rather than all available. For general usage we have
5653 found a stripe count of 1 or 2 works best.
5655 Severity : enhancement
5656 Description: Add support for compiling against Cray portals.
5657 Details : Conditional compiling for some areas that are different
5661 Frequency : occasional
5662 Bugzilla : 6409, 6834
5663 Description: Creating files with an explicit stripe count may lead to
5664 a failed assertion on the MDS
5665 Details : If some OSTs are full or unavailable, creating files may
5666 trigger a failed assertion on the MDS. Now, Lustre will
5667 try to use other servers or return an error to the
5671 Frequency : occasional
5673 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5674 caused lock timeout during readahead (since 1.4.2).
5675 Details : Processes doing readahead might match a lock that hasn't been
5676 granted yet if there are overlapping and conflicting lock
5677 requests. The readahead process waits on ungranted lock
5678 (original lock is CBPENDING), while OST waits for that process
5679 to cancel CBPENDING read lock and eventually evicts client.
5681 Severity : enhancement
5683 Description: Initial enabling of flock support for clients
5684 Details : Implements fcntl advisory locking and file status functions.
5685 This feature is provided as an optional mount flag (default
5686 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5687 locking are implemented yet, and those that are are not guaranteed
5688 to be completely correct in production environments.
5689 mount -t lustre -o [flock|noflock] ...
5692 Frequency : occasional
5694 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5695 assertion in the ext3 JBD (journaling) layer.
5696 Details : The b_committed_data struct is protected by the big kernel lock
5697 in 2.4 kernels, serializing journal_commit_transaction() and
5698 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5699 access to this struct. In 2.6 kernels there is finer grained
5700 locking to improve SMP performance of the JBD layer.
5704 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5709 Description: merge and cleanup kernel patches.
5712 -----------------------------------------------------------------------------
5714 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5716 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5717 versions of Lustre. Please update all servers and clients to
5718 version 1.4.2 or later at the same time. You must also run
5719 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5720 to update the configuration logs.
5722 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5723 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5724 - swab llog records of type '0' so we get proper header size/idx (5861)
5725 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5726 - fix rename of one directory over another leaking an inode (5953)
5727 - avoid SetPageDirty on 2.6 (5981)
5728 - don't re-add just-being-destroyed locks to the waiting list (5653)
5729 - when creating new directories, inherit the parent's custom
5730 striping settings if present parent (3048)
5731 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5732 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5733 - don't hold client locks on temporary worklist from l_lru (5666)
5734 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5735 - automatically evict dead clients (5921)
5736 - Update file size properly in create+truncate+fstat case (6196)
5737 - Do not unhash mountpoint dentries, do not allow removal of
5739 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5740 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5741 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5742 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5743 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5744 - Correctly handle reads of files with no objects (6243)
5745 - lctl recover will also mark a device active if deactivate used (5933)
5747 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5748 - allow --write-conf on an MDS with different nettype than client (5619)
5749 - don't write config llogs to MDS for mounts not from that MDS (5617)
5750 - lconf should create multiple TCP connections from a client (5201)
5751 - init scripts are now turned off by default; run chkconfig --on
5752 lustre and chkconfig --on lustrefs to use them
5753 - upcalls are no longer needed for clients to recover to failover
5755 - add --abort-recovery option to lconf to abort recovery on device
5757 - add support for an arbitrary number of OSTs (3026)
5758 - Quota support protocol changes.
5759 - forward compatibility changes to wire structs (6007)
5760 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5761 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5762 - improved "lustre status" script
5763 - initialize blocksize for non-regular files (6062)
5764 - added --disable-server and --disable-client configure options (5782)
5765 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5766 - Vanilla 2.4.29 support
5767 - increase maximum number of obd devices to 520 (6242)
5768 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5769 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5770 SLES 9 2.6.5 kernels
5771 - limit stripes per file to 160 (the maximum EA size) (6093)
5773 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5776 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5777 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5778 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5779 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5780 - small ext3 extents cleanups and fixes (5733)
5781 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5782 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5783 - avoid needless client->OST connect, fix handle mismatch (5317)
5784 - fix DLM error path that led to out-of-sync client, long delays (5779)
5785 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5786 - fix several locking issues related to i_size (5492,5624,5654,5672)
5787 - don't move pending lock onto export if it is already evicted (5683)
5788 - fix kernel oops when creating .foo in unlinked directory (5548)
5789 - fix deadlock in obdfilter statistics vs. object create (5811)
5790 - use time_{before,after} to avoid timer jiffies wrap (5882)
5791 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5792 - Do not leak request if server was not able to process it (5154)
5793 - If mds_open unable to find parent dir, make that negative lookup(5154)
5794 - don't create new directories with extent-mapping (5909, 5936)
5796 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5797 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5798 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5799 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5800 - support for RHEL4 kernel on i686 (5773)
5801 - provide error messages when incompatible logs are encountered (5898)
5803 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5804 * version 1.4.0.10 (1.4.1 release candidate 1)
5806 - don't keep a lock reference when lock is not granted (4238)
5807 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5808 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5809 - drop import inflight refcount on signal_completed_replay error (5255)
5810 - unlock page after async write error during send (3677)
5811 - handle missing objects in filter_preprw_read properly (5265)
5812 - no transno return for symlink open, don't save no-trasno open (3440)
5813 - don't try to complete elan receive that already failed (4012)
5814 - free RPC server reply state on error (5406)
5815 - clean up thread from ptlrpc_start_thread() on error (5160)
5816 - readahead could read extra page into cache that wasn't ejected (5388)
5817 - prevent races in class_attach/setup/cleanup/detach (5260)
5818 - don't dereference de->d_inode after l_dput of de (5458)
5819 - use "int" for stripe value returned from lock_to_stripe (5544)
5820 - mballoc allocation and error-checking fixes in 2.6 (5504)
5821 - block device patches to fix I/O request sizes in 2.6 (5482)
5822 - look up hostnames for IB nals (5602)
5823 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5824 - don't start multiple acceptors for the same port (5277)
5825 - fix incorrect LASSERT in mds_getattr_name (5635)
5826 - export a proc file for general "ping" checking (5628)
5827 - fix "lfs check" to not block when the MDS is down (5628)
5829 - service request history (4965)
5830 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5831 - create an "evict_client" /proc entry on OSTs, like the MDS has
5832 - fix mount usage message, return errors per mount(8) (5168)
5833 - change grep [] to grep "[]" in tests so they work in more UMLs
5834 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5835 - remove ancient LOV_MAGIC_V0 EA support (5047)
5836 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5837 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5838 - allow readahead allocations to fail when low on memory (5383)
5839 - mmap locking landed again, after considerable improvement (2828)
5840 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5842 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5845 - send OST transaction number in read/write reply to free req (4966)
5846 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5847 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5848 - the watchdog thread now runs as interruptible (5246)
5849 - flock/lockf fixes (but it's still disabled, pending 5135)
5850 - don't use EXT3 constants in llite code (5094)
5851 - memory shortage at startup could cause assertion (5176)
5853 - reorganization of lov code
5854 - single portals codebase
5856 - add extents/mballoc support (5025)
5857 - direct I/O reads in the obdfilter (4048)
5858 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5860 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5863 - send OST transaction number in read/write reply to free req (4966)
5864 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5865 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5866 - the watchdog thread now runs as interruptible (5246)
5867 - handle missing objects in filter_preprw_read properly (5265)
5868 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5869 - drop import inflight refcount on signal_completed_replay error (5255)
5870 - unlock page after async write error during send (3677)
5871 - return original error code on reconstructed replies (3761)
5872 - no transno return for symlink open, don't save no-trasno open (3440)
5874 - add pid to ldlm debugging output (4922)
5875 - bump the watchdog timeouts -- we can't handle 30sec yet
5876 - extra debugging for orphan dentry/inode bug (5259)
5878 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5881 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5882 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5883 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5884 - fix NULL dereference in /proc/sys/portals/routes (4827)
5885 - allow failed mdc_close() operations to be interrupted (4561)
5886 - stop precreate on OST before MDS would time out on it (4778)
5887 - don't send partial-page writes before EOF from client (4410)
5888 - discard client grant for sub-page writes on large-page clients (4520)
5889 - don't free dentries not owned by NFS code, check generation (4806)
5890 - fix lsm leak if mds_create_objects() fails (4801)
5891 - limit debug_daemon file size, always print CERROR messages (4789)
5892 - use transno after validating reply (3892)
5893 - process timed out requests if import state changes (3754)
5894 - update mtime on OST during writes, return in glimpse (4829)
5895 - add mkfsoptions to LDAP (4679)
5896 - use ->max_readahead method instead of zapping global ra (5039)
5897 - don't interrupt __l_wait_event() during strace
5899 - add software watchdogs to catch hung threads quickly (4941)
5900 - make lustrefs init script start after nfs is mounted
5901 - fix CWARN/ERROR duplication (4930)
5902 - return async write errors to application if possible (2248)
5903 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5904 - print NAL number in %x format (4645)
5905 - update barely-supported suse-2.4.21-171 series (4842)
5906 - support for sles 9 %post scripts
5907 - support for building 2.6 kernel-source packages
5908 - support for sles km_* packages
5910 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5913 - ignore -ENOENT errors in osc_destroy (3639)
5914 - notify osc create thread that OSC is being cleaned up (4600)
5915 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5916 - reconstruct ost_handle() like mds_handle() (4657)
5917 - create a new thread to do import eviction to avoid deadlock (3969)
5918 - let lconf resolve symlinked-to devices (4629)
5919 - don't unlink "objects" from directory with default EA (4554)
5920 - hold socknal file ref over connect in case target is down (4394)
5921 - allow more than 32000 subdirectories in a single directory (3244)
5922 - fix blocks count for O_DIRECT writes (3751)
5923 - OST returns ENOSPC from object create when no space left (4539)
5924 - don't send truncate RPC if file size isn't changing (4410)
5925 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5926 - bind to privileged port in socknal and tcpnal (3689)
5928 - rate limit CERROR/CWARN console message to avoid overload (4519)
5929 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5930 - basic mmap support (3918)
5931 - kernel patch series update from b1_4 (4711)
5933 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5936 - avoid crash during MDS cleanup with OST shut down (2775)
5937 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5938 - don't use bad inodes on the MDS (3744)
5939 - dynamic object preallocation to improve recovery speed (4236)
5940 - don't hold spinlock over lock dumping or change debug flags (4401)
5941 - don't zero obd_dev when it is force cleaned (3651)
5942 - print grants to console if they go negative (4431)
5943 - "lctl deactivate" will stop automatic recovery attempts (3406)
5944 - look for existing locks in ldlm_handle_enqueue() (3764)
5945 - don't resolve lock handle twice in recovery avoiding race (4401)
5946 - revalidate should check working dir is a directory (4134)
5948 - don't always mark "slow" obdfilter messages as errors (4418)
5950 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5953 - don't close LustreDB during write_conf until it is done (3860)
5954 - fix typo in lconf for_each_profile (3821)
5955 - allow dumping logs from multiple threads at one time (3820)
5956 - don't allow multiple threads in OSC recovery (3812)
5957 - fix debug_size parameters (3864)
5958 - fix mds_postrecov to initialize import for llog ctxt (3121)
5959 - replace config semaphore with spinlock (3306)
5960 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5961 - don't allow enqueue to complete on a destroyed export (3822)
5962 - down write_lock before checking llog header bitmap (3825)
5963 - recover from lock replay timeout (3764)
5964 - up llog sem before sending rpc (3652)
5965 - reduce ns lock hold times when setting kms (3267)
5966 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5967 - fix NULL deref and obd_dev leak on setup error (3312)
5968 - replace some LBUG about llog ops with error handling (3841)
5969 - don't match INVALID dentries from d_lookup and spin (3784)
5970 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5971 - fix invalid assertion in ptlrpc_set_wait (3880)
5973 - add libwrap support for the TCP acceptor (3996)
5974 - add /proc/sys/portals/routes for non-root route listing (3994)
5975 - allow setting MDS UUID in .xml (2580)
5976 - print the stack of a process that LBUGs (4228)
5978 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5981 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5982 - make sure to unset replay flag from failed open requests (3440)
5983 - if default stripe count is 0, use OST count for inode size (3636)
5984 - update parent mtime/ctime on client for create/unlink (2611)
5985 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5986 - free recovery state on server during a forced cleanup (3571)
5987 - unregister_reply for resent reqs (3063)
5988 - loop back devices mounting and status check on 2.6 (3563)
5989 - fix resource-creation race that can provoke i_size == 0 (3513)
5990 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5991 - more debugging for page-accounting assertion (3746)
5992 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5993 - avoid deadlock after precreation failure (3758)
5994 - fix race and lock order deadlock in orphan handling (3450, 3750)
5995 - add validity checks when grabbing inodes from l_ast_data (3599)
5997 - add /proc/.../recovery_status to obdfilter (3428)
5998 - lightweight CDEBUG infrastructure, debug daemon (3668)
5999 - change default OSC RPC parameters to be better on small clusters
6000 - turn off OST read cache for files smaller than 32MB
6001 - install man pages and include them in rpms (3100)
6002 - add new init script for (un)mounting lustre filesystems (2593)
6003 - run chkconfig in %post for init scripts (3701)
6004 - drop scimac NAL (unmaintained)
6006 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6009 - clean kiobufs before and after use (3485)
6010 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6011 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6012 - change init script to fail more clearly if not run as root (1528)
6013 - allow clients to reconnect during replay (1742)
6014 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6015 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6016 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6017 - fix cond_resched (3554)
6018 - only evict unfinished clients after recovery (3515)
6019 - allow bulk resend, prevent data loss (3570)
6020 - dynamic ptlrpc request buffer allocation (2102)
6021 - don't allow unlinking open directory if it isn't empty (2904)
6022 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6023 - remove extraneous obd dereference causing LASSERT failure (3334)
6024 - don't use get_cycles() when creating temp. files on the mds (3156)
6025 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6026 - handle EEXIST for set-stripe, set proper directory name (3336)
6028 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6029 - fix ksocknal_fmb_callback() error messages (2918)
6031 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6034 - don't copy lvb into (possibly NULL) reply on error (2983)
6035 - don't deref dentry after dput, don't free lvb on error (2922)
6036 - use the kms to determine writeback rpc length (2947)
6037 - increment oti_logcookies when osc is inactive (2948)
6038 - update client's i_blocks count via lvb messages (2543)
6039 - handle intent open/close of special files properly (1557)
6040 - mount MDS with errors=remount-ro, like obdfilter (2009)
6041 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6042 - don't use cancelling-locks' kms values (2947)
6043 - use highest lock extent for kms, not last one (2925)
6044 - don't dereference ERR_PTR() dentry in error handling path (3107)
6045 - fix thread race in portals_debug_dumplog() (3122)
6046 - create lprocfs device entries at setup instead of at attach (1519)
6047 - common AST error handler, don't evict client on completion race (3145)
6048 - zero nameidata in detach_mnt in 2.6 (3118)
6049 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6050 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6051 - initialize RPC timeout timer earlier for 2.6 (3219)
6052 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6053 - print nal/nid for unknown nid (3258)
6054 - additional checks for oscc recovery before doing precreate (3284)
6055 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6056 - don't crash in mdc_close for bad permissions on open (3285)
6057 - zero i_rdev for non-device files (3147)
6058 - clear page->private before handing to FS, better assertion (3119)
6059 - tune the read pipeline (3236)
6060 - fix incorrect decref of invalidated dentry (2350)
6061 - provide read-ahead stats and refine rpc in flight stats (3328)
6062 - don't hold journal transaction open across create RPC (3313)
6063 - update atime on MDS at close time (3265)
6064 - close LDAP connection when recovering to avoid server load (3315)
6065 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6066 - don't leak open file on MDS after open resend (3325)
6067 - serialize filter_precreate and filter_destroy_precreated (3329)
6068 - loop device shouldn't call sync_dev() for nul device (3092)
6069 - clear page cache after eviction (2766)
6070 - resynchronize MDS->OST in background (2824)
6071 - refuse to mount the same filesystem twice on same mountpoint (3394)
6072 - allow llmount to create routes for mounting behind routers (3320)
6073 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6074 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6075 - fix rare problem with rename on htree directories (3417)
6077 - allow default OST striping configuration per directory (1414)
6078 - fix compilation for qswnal for 2.6 kernels (3125)
6079 - increase maximum number of MDS request buffers for large systems
6080 - change liblustreapi to be useful for external progs like lfsck (3098)
6081 - increase local configuration timeout for slow disks (3353)
6082 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6084 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6087 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6088 - don't overwrite extent policy data in reply if lock was blocked (2901)
6089 - drop filter export grants atomically with removal from device (2663)
6090 - del obd_self_export from work_list in class_disconnect_exports (2908)
6091 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6092 - swab reply message in mdc_close, other PPC fixes (2464)
6093 - fix destroying of named logs (2325)
6094 - overwrite old logs when running lconf --write_conf (2264)
6095 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6096 - fix race in target_handle_connect (2898)
6097 - mds_reint_create() should take same inode create lock (2926)
6098 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6099 - don't close files for self_export to avoid uninitialized obd (2936)
6100 - allow MDS with the same name as client node (2939)
6101 - hold dentry reference for closed log files for unlink (2325)
6102 - reserve space for all logs during transactions (2059)
6103 - don't evict page beyond end of stripe extent (2925)
6104 - don't oops on a deleted current working directory (2399)
6105 - handle hard links to targets without a parent properly (2517)
6106 - don't dereference NULL lock when racing during eviction (2867)
6107 - don't grow lock extents when lots of conflicting locks (2919)
6109 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6112 - account for cache space usage on clients to avoid data loss (974)
6113 - lfsck support in lustre kernel code (2349)
6114 - reduce journal credits needed for BRW writes (2370)
6115 - orphan handling to avoid losing space on client/server crashes
6116 - ptlrpcd can be blocked, stopping ALL progress (2477)
6117 - use lock value blocks to assist in proper KMS, faster stat (1021)
6118 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6119 - recovery for initial connections (2355)
6120 - fixes for mds_cleanup_orphans (1934)
6121 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6122 - block all file creations until orphan recovery completes (1901)
6123 - client remove rq_connection from request struct (2423)
6124 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6125 - recovery timer race (2670)
6126 - mdc_close recovey bug (2532)
6127 - ptlrpc cleanup bug (2710)
6128 - mds timeout on local locks (2588)
6129 - namespace lock held during RPCs (2431)
6130 - handle interrupted sync write properly (2503)
6131 - don't try to handle a message that hasn't been replied to (2699)
6132 - client assert failure during cleanup after abort recovery (2701)
6133 - leak mdc device after failed mount (2712)
6134 - ptlrpc_check_set allows timedout requests to complete (2714)
6135 - wait for inflight reqs when ptlrpcd finishes (2710)
6136 - make sure unregistered services are removed from the srv_list
6137 - reset bulk XID's when resending them (caught by 1138 test)
6138 - unregister_bulk after timeout
6139 - fix lconf error (2694)
6140 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6141 - readahead locks pages, leaves pending causing memory pressure (2673)
6142 - increase OST request buffers to 4096 on large machines (2729)
6143 - fix up permission of existing directories in simple_mkdir (2661)
6144 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6145 - don't assign transno to errored transactions (2742)
6146 - don't delete objects on OST if given a bogus objid from MDS (2751)
6147 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6148 - if rq_no_resend, then timeout request after recovery (2432)
6149 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6150 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6151 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6152 - MDS recovery shouldn't delete objects that it has given out (2730)
6153 - if enqueue arrives after completion, don't clobber LVB (2819)
6154 - don't unlock pages twice when trigger_group_io returns error (2814)
6155 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6156 - don't write pages to disk if there was an error (1450)
6157 - don't ping imports that have recovery disabled (2676)
6158 - take buffered bytes into account when balancing socknal conn (2817)
6159 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6160 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6161 - remove little-endian swabbing of llog records (1987)
6162 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6163 - retry reposting request buffers if they fail (1191)
6164 - grow extent at grant time to avoid granting a revoked lock (2809)
6165 - lock revoke doesn't evict page if covered by a second lock (2765)
6166 - disable VM readahead to avoid reading outside lock extents (2805)
6168 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6169 - updated kernel patches for hp-2.4.20 kernel (2681)
6171 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6174 - fix truncated write corruption (2366)
6175 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6177 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6178 - ensure len(uuid) < 37 in lmc (1171)
6179 - fix ia64 OOPS in llog_test (2255)
6180 - zero end of page at obdfilter for partial page writes (2648)
6181 - don't leave stale dentries around after renames (bug 2428)
6182 - fix timeouts when evicting a client with a single lock held (2642)
6183 - set deadline for the initial HELLO message to drain (2634)
6184 - print out dotted-quad IP addresses in the socknal (2302)
6186 - additional debugging for MDS client eviction problem (2443)
6187 - fix mkfsoptions support for osts (2603, 2604)
6189 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6192 - add series for the vanilla 2.6.0 kernel
6193 - add series for the vanilla 2.4.24 kernel
6194 - add series for a cray x86/64 UL kernel drop
6195 - fix xattr patches for the vanilla 2.4.19 series
6197 - generate true UUIDs in lmc (1171)
6198 - have portals stack dumping break in UML (2466)
6199 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6200 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6201 - wake callback waiting threads on client eviction (2460)
6202 - Add --ptldebug and --subsystem to lmc (1719)
6203 - update assertion to allow safe interrupt allocation
6204 - set rq_no_resend for cancel requests (2432)
6205 - recalculate ptlrpcd timeout after resend (2494)
6206 - call vfs_rmdir when removing pending directories (2368)
6207 - fix renaming a file to itself (2429)
6208 - lmc creates a default one-stripe lov (2454)
6209 - expand procfs space to handle large clusters (2326)
6210 - increase UML stack to avoid overflow
6211 - update lconf's list of debug and subsystem masks
6212 - fix lfs find --obd (2510)
6213 - /proc tunable for disabling filter read caching (2591)
6214 - stop rpm packages from altering slapd.conf (2301)
6215 - disable nagle in the socknal under 0conf (2578)
6216 - choose mds inode size based on stripe count (2572)
6217 - fix kernel-source rpm problems (2516)
6219 - add --disable-doc to avoid pdf generation (2421)
6220 - update documentation, tests, type-os, comments
6221 - avoid format warnings on ia64
6222 - remove the TOE NAL
6223 - tiny code cleanups by removing unused fields
6225 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6228 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6229 - remove the most blatant lies from BUILDING file (2371)
6230 - change default debug level to reasonable production setting
6231 - reduce client side cache size to reduce cache flush time
6232 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6233 - make TCP zerocopy and pinger support enabled by default (2476)
6234 - sync writes completed after process exits caused crashes (2319)
6235 - maintain correct mount count on the MDS (2356)
6236 - backout 1557, because 2316 wasn't really fixed
6237 - better file I/O statistics gathering in /proc
6238 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6239 - another kernel patch to fix zero-copy TCP function export
6240 - don't take duplicate lock when processing re-sent getattr (2420)
6241 - lctl uses obd_self_export instead of creating new conn (2353)
6242 - MDS/OST recovery case which requires object creation asserted (2425)
6243 - move lfs from /usr/sbin to /usr/bin in packages
6244 - fix race between mds_client_add and mds_client_free (2417)
6245 - use kmalloc instead of slabs in portals (2430)
6246 - don't create duplicate records when a failover MDS is present (2442)
6247 - remove unnecessary mount age check (2332)
6248 - don't remove directory inodes from locks prematurely (2451)
6249 - don't break if MDS service name is the same as hostname (2103)
6250 - fix races in client write RPC generation when cache full (2482)
6252 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6255 - remove now-unused request->rq_obd (278)
6256 - if an allocation fails, print out how much memory we've used (1933)
6257 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6258 - add the "configurable stack size" patch to most series files (1256)
6259 - ability to write large log records, for 100+ OST configs (2306)
6260 - fix NULL deref when filter_prep fails (2314)
6261 - fix operator precedence error in filter_sync
6262 - dynamic allocation of socknal TX descriptors (2315)
6263 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6264 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6265 - ensure that CWARN messages in Portals always get to the syslog
6266 - __init/__exit are not for prototype decls (ldlm_init/exit)
6267 - x86-64 compile warning fixes
6268 - fix gateway LMC keyword conflict (2318)
6269 - fix MDS lock inversions in getattr/reint paths (1844)
6270 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6271 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6272 - DLM race condition prevented some lock evictions (2328)
6273 - ENOMEM detection and retry on socknal sends (2230)
6274 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6275 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6276 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6277 - better error messages when a client is rejected during recovery (1505)
6278 - avoid cancelling locks which were never granted, after failure (2330)
6279 - fix i_sem/journal inversion in mds_client_add (2333)
6280 - fix truncate/getattr lock cycle deadlock (2334)
6281 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6282 - fix two rare exit paths which could leak an l_lock() ref (2321)
6283 - fencepost error in MDS/OST orphan recovery (2226)
6284 - make log record alignment 8 bytes (1988)
6285 - lstripe now fails when requested offset > ost_count (2237)
6286 - ensure that all kernel series have a complete list.h (1607)
6287 - fix crashes in special-file operations (2316)
6288 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6290 - allow configurable automake binary, for testing new versions
6291 - small update to the lfs documentation
6293 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6295 * fix negative export reference count in fsfilt_sync (2312)
6297 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6298 * release candidate 0.9.1
6300 - orphans are moved into the PENDING directory for possible recovery
6301 - replayed opens now open by fid for orphan/rename safety (1042)
6302 - last close of an orphan inode generates a transno (683)
6303 - chdir() and mount() now pin the directory entry (1020)
6304 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6305 - discard very old requests without processing them (1502)
6306 - remove some common, well-understood CERRORs (1505)
6307 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6308 - clear "grant" flags in OST replies until OST grant code lands (1644)
6309 - fix read performance by not clobbering i_blksize on client (1598)
6310 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6311 - make LRU size a /proc tunable, clears locks when reduced (707)
6312 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6313 - server locks take references on exports now (1558)
6314 - build fixes for 2.4.20-rh trees (1663)
6315 - return an error from lov_create if all OSCs are inactive (1751)
6316 - fix import levels when a reconnect happens without a timeout (1597)
6317 - exit early from mds_open if we get a lookup error (1749)
6318 - partial page read at EOF wouldn't wait for disk before sending (1642)
6319 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6320 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6321 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6322 - write/truncate lock inversion (1639)
6323 - Don't auto-load obdclass, portals modules during cleanup (1495)
6324 - fix timestamps from jumping to "now" (1763)
6325 - extra journal assertions (1648)
6326 - add an extra multiunlink test (1771)
6327 - fix read_record/write_record API (1776)
6328 - fix leak of offset_extent, possible incorrect i_size later (1772)
6329 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6330 - Debugging for the kqswnal_get_idle_tx problems (1820)
6331 - Allow recovery to be attempted multiple times (1536)
6332 - Write out MDS last_rcvd file after it is first created (1600)
6333 - Fix tx_descriptor leak in failed transmit situations (1827)
6334 - ext3 journaling fixes for assertion failure after IO error (1871)
6335 - class_export_put() on freed export after completion AST error (1896)
6336 - Fix revalidate looping in VFS (1322)
6337 - Don't access a freed export during MDS_REINT timeout (1521)
6338 - Add open-unlink recovery support on the MDS (1673,1764)
6339 - Return an error if no MDS data was read from last_rcvd (1946)
6340 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6341 - Don't setup a disk device that doesn't match exported UUID (317)
6342 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6343 - avoid committing NULL handle in force close
6344 - local.sh is now a one-stripe LOV configuration
6345 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6346 - fix ext3 htree duplicate directory entry corruption (1516)
6347 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6348 - update to new LOV EA format (2097)
6349 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6351 2003-06-15 Phil Schwan <phil@clusterfs.com>
6354 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6355 - per-import/export recovery handling (958, 931, 959)
6356 - multiple last-rcvd slots, for serving multiple FSes (949)
6357 - connections are again shared between multiple imp/exports (963, 964)
6358 - "umount -f" would hang if any requests needed to be sent (393, 978)
6359 - avoid pinning large req buffer by copying for queued messages (989)
6360 - add "uuid" to "lctl device" command to help upcalls (991)
6361 - "open" RPCs with transnos would confuse recovery counters (1037)
6362 - do proper endian conversion of all wire messages (288, 340, 891)
6363 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6364 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6365 - fix ptlrpc_connection leak in target_handle_connect (1174)
6366 - fix import refcounting bug in OST and MDS cleanup (1134)
6367 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6368 - fix very unlikely obd_types race condition (501)
6369 - remove osc_open hack for echo_client (1187)
6370 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6371 - a failure in read_inode2 leads to deadlock (1139)
6372 - cancel ack-locks as soon as transaction is committed (1072)
6373 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6374 - make sure to commitrw after any preprw to avoid deadlock (1162)
6375 - failing to execute a file in a lustre FS would lock inode (1203)
6376 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6377 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6378 - obd_uuid2tgt would walk off the end of the list (1255)
6379 - on IA64 the osc would give portals incorrect bulk size (1258)
6380 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6381 - fix lock inversion caused by new llite matching code (1282)
6382 - limit the number of dirty pages on a client to 10MB (1286)
6383 - timed out locks were not being corrected cancelled (1289)
6384 - fix O_DIRECT above 4GB on IA-32 (1292)
6385 * major user-visible changes
6386 - fail out/fail over policy now controlled by the upcall (993)
6388 - add OBD_PING to check server availability and failure (954)
6389 - lustre messages are now sent in sending host order (288, 340, 891)
6390 - add eadatalen to MDS getattr reply (340)
6391 - OST read replies may contain second buffer, with per-page status (593)
6393 2003-03-11 Phil Schwan <phil@clusterfs.com>
6396 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6397 - failed open()s could cause deadlock; fixed (867, 869)
6398 - stop cancelling OST locks when files are closed (481)
6399 - overlapping XID spaces caused network corruption (851, 853)
6400 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6401 - setattr_raw added, to do single-RPC, server-side setattrs
6402 - lmc/lconf syntax change for OST UUIDs
6403 - fix crashy race condition between ptlrpc_free_req and osc_close
6404 - don't use request in mdc_enqueue if we hit a timeout (889)
6405 - don't set the inode i_size for regular files from the MDS (896)
6406 - handle out of order completion AST (842)
6407 - don't LBUG if a lock request times out after receiving AST (913)
6408 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6409 - if a bad lock AST arrives, send an error instead of dropping entirely
6410 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6411 - fix leak in bulk IO when only partially completed (899, 900, 926)
6412 - fix O_DIRECT for ia64 (55)
6413 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6414 - C-z after timeout could hang a process forever; fixed (977)
6416 - client-side I/O cache (678, 924, 929, 941, 970)
6418 - READPAGE and SETATTRs which don't take server-side locks get
6421 2003-02-11 Phil Schwan <phil@clusterfs.com>
6424 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6425 - Fix fsfilt_extN_readpage() to read a full page of directory
6426 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6427 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6428 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6429 - Most intent operations were removed, in favour of a new RPC mode
6430 that does a single RPC to the server and bypasses most of the VFS
6431 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6432 - Aggressively cancel local locks on DLM servers
6433 - mds_reint_unlink sends EA to the client if it's the last nlink.
6434 client uses that EA to unlink OST objects.
6435 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6436 - recursive symlinks were fixed (439)
6437 - fixed NULL deref in DEBUG_REQ
6438 - filter_update_lastobjid no longer calls sync, which annoyed extN
6439 - fixed multi-client small-writes to a single file problem (445)
6440 - fixed mtime updates during file writes (607)
6441 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6442 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6443 - fixed Linux OST size reporting problem (444, 656)
6444 - OST now updates object mtime with writes or setattr (607, 619)
6445 - client verifies file size before zeroing page past EOF (445)
6446 - OST now writes last allocated objid to disk with allocation (108)
6447 - LOV on echo now works (409)
6449 - mds_reint_unlink sends a new buffer, with the EA included. this
6450 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6451 set if a regular file was being unlinked, and it was the last link
6452 - use PtlGet from the target for bulk writes (315)
6453 - OST now updates object mtime with writes or setattr (607, 619)
6454 - LDLM now has a grant-time callback to revalidate locked items, if
6456 - Many MDS operations were reorganized to combat race conditions
6458 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6459 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6461 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6464 - Fully reactivate OST imports after reconnection (512, others)
6465 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6466 - More graceful error handling for truncating on dead OST (515)
6467 - Don't error out unless we're actually accessing dead stripes (474)
6468 - Fix garbage sizes when stripes are missing (410)
6469 - LRU counters were broken, causing constant lock purge (433, 432)
6470 - garbage on read from stripes with failed OSTs (441)
6471 - mark OSCs as active before reconnecting during recovery (438)
6472 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6473 - lfind did not preserve OST order in output (443)
6474 - symlinks cause hung clients, incorrect data (439)
6475 - stop dereferencing request after dropping refcount (457)
6476 - don't LASSERT(spin_is_locked) on non-SMP (455)
6477 - fixes for many rename() bugs
6478 - fstat didn't correctly synchronize attributes (399)
6479 - server must handle lock cancellation during blocking AST prep (487)
6480 - bulk descriptors were free()d too soon (511)
6481 - fix paths in lconf, which would load incorrect modules (451, 507)
6482 - fix confusing lconf 'host not found' error message (386)
6483 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6484 - fix race condition in mdc_blocking_ast() for inode access (526)
6485 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6486 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6487 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6488 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6489 - fix lstripe in several ways (kernel side) (527)
6490 - fix request leak in ldlm_cli_enqueue (262)
6491 - incorrect OSC was marked inactive after OST failure
6492 - call mds_fs_cleanup before unmounting filesystem (524)
6493 - fix races between taking ns_lock and ldlm_lock_change_resource
6494 - fix races updating LOV export open file list
6495 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6496 - fix recovery NULL deref in ldlm_cli_cancel_unused
6497 - fix some DLM races by using new hash table for lock handles (419)
6498 - permit the client to specify desired inodes, at replay
6499 - duplicate requests when we queue them for replay reintegration
6500 - fix last_rcvd offset calculation
6501 - sync after each recovered transaction, so we always make progress
6502 - never, not always, ERESTART requests without transnos
6503 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6505 - skip replay if the MDS says that the client is already connected
6506 - don't check for a recovery-enabled export to match lctl's UUID
6507 - don't INC_USE_COUNT for phantom exports
6508 - don't crash when cleaning up phantom exports (567)
6509 - don't double-finish or set replay data for errored mdc_open requests
6510 - abort requests when they time out, so we don't get old replies
6511 - send/receive replies for AST messages again
6512 - if the client says that it doesn't have the lock, cancel it on the
6514 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6515 mark it as destroyed, it will all work out when decref is called
6516 - fix module use counts (22, 581)
6518 - ASTs now expect a reply (server cancels lock on error reply)
6520 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6523 - fix many simultaneous client startup (392)
6524 - fix dentry->d_it clobbering
6525 - credentials weren't being shipped for readdir/getattr operations
6526 - remove invalid assertions triggered during some concurrent MD
6528 - proper Lustre versions added (336, 389)
6529 - fix memory leak for create error case (398)
6530 - fix LOV locking bug that would get cli/srv out of sync
6531 - fix echo client over LOV (409)
6532 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6533 - fix double-O_EXCL intent crash (424)
6534 - avoid sending multiple lock CANCELs (352)
6536 - MDS can do multi-client recovery (modulo bugs in new code)
6538 - many updates, edits, cleanups
6540 2002-11-18 Phil Schwan <phil@clusterfs.com>
6543 - fix null d_it dereference (346)
6544 - fix full OST/dbench hang (333)
6545 - fix permission problem with file removal (286)
6546 - fix removal of OSCs from LOV when they fail
6547 - fix NULL deref during bulk timeout (214)
6548 - fix problems related to multiple filesystems on one MDS (241)
6549 - fixed serious subtle metadata locking bugs
6550 - free locks on clients when inodes are removed due to memory
6552 - fix inode pointer in lock data (285)
6553 - partial support for multiple MDS on a single host (241)
6554 - data locks weren't cancelled at clear_inode time (290, 311)
6555 - intent locks could lead to unbounded lock growth (205)
6556 - added a maximum lock count, an LRU list, and a flusher
6557 - fix multiple rename (365)
6558 - properly abstracted the echo client
6559 - OSC locked 1 byte too many; fixed
6560 - rewrote brw callback code:
6561 - fixed recovery bugs related to LOVs (306)
6562 - fixed too-many-pages-in-one-write crash (191)
6563 - fixed (again) crash in sync_io_timeout (214)
6564 - probably fixed callback-related race (385)
6566 - Add capability to MDS protocol
6567 - LDLM cancellations and callbacks on different portals
6569 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6572 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6573 - defer open object destruction to close time (601981, 138)
6574 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6575 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6576 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6577 - don't call obd_setattr in ll_file_release for destroyed objects
6578 * protocol change to lustre_msg: move |version| and add |flags|
6579 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6580 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6581 * added replay of create, unlink, link and rename operations during
6582 MDS failover; recovery should be much more robust now
6583 * remove failed OSCs from LOVs (only lov_create uses this so far)
6584 * the lustre-HOWTO was brought (more) up to date (582544)
6586 2002-10-23 Phil Schwan <phil@clusterfs.com>
6589 - in-use dentries weren't being reused properly (617851)
6590 - prevent multiple LDLM setup (599178)
6591 - fix LOV size calculations for truncate (617853)
6592 - fix client handling of MDS intent errors (POSIX)
6593 - fix permission bug in lovstripe.c test (624321)
6594 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6595 - truncate past end of file could corrupt data
6596 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6597 - a race in recovery could return ETIMEDOUT to apps (623947)
6598 - building outside the source directory was fixed
6599 * the lustre-HOWTO was brought (more) up to date (582544)
6600 * major progress was made on recovery functionality
6602 2002-10-10 Phil Schwan <phil@clusterfs.com>
6605 - recovery deadlock fix
6606 - rm -rf causes LBUG fix (617817)
6607 - file open by multiple tasks fix (618962)
6608 - directory permissions bugs (602707 and 620007)
6609 - journal_stop fixed with locking (611313)
6610 - O_APPEND failures resolved (618273, perhaps 614459)
6611 - lconf PATH fix (619770)
6612 - IA64 build fix (621450)
6613 - RPC buffer sizes scale with amount of memory
6615 2002-10-01 Phil Schwan <phil@clusterfs.com>
6618 - locks would be cancelled without throwing away data pages,
6619 resulting in inconsistent data (605627)
6620 - inode attributes were not always being refreshed (605627, 612449)
6621 - lconf now continues to cleanup after lctl reports an error
6622 - MDS now enforces user permissions (602707)
6623 - lprocfs cleanup fixed, but not yet enabled (614157)
6624 - fixed infinite server hang, should a client not respond to an AST
6625 - avoid going into recovery if user calls readlink() with a buffer
6626 that's too small (613941)
6627 - AST RPCs no longer require replies (614867) -- this may be changed
6628 - don't crash server if client sends an IOV that's too big (611336)
6629 - fixed lock conversion deadlock (611892)
6630 - fixed the following of symlinks (614622)
6631 * recovery: the server can remove locks from a client that dies, other
6632 clients can make progress
6633 * more extN patch fixes
6634 * compile-time configurable ptlrpc buffer allocations
6636 - collaborative read cache document
6637 - Lustre Lite Performance CDR document-in-progress
6639 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6642 - fix typo in patch-2.4.18
6644 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6647 - clear ptlrpc request each time in handle_incoming_request()
6648 - unlink of files now destroys the object on the OST
6650 2002-09-19 Peter Braam <braam@clusterfs.com>
6652 * add hard link support
6653 * change obdfile creation method
6654 * kernel patch changed
6656 2002-09-19 Peter Braam <braam@clusterfs.com>
6659 - stack overflow bug in extN fixed
6661 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6663 * documentation updates
6664 - add man pages for config tools
6665 - update tests/README to describe testing with new config tools
6666 - finish metadata API descriptions
6667 * bug fixes and cleanups
6668 - statfs workaround for 16TB limit
6669 - LOV stripe allocation improved, can stripe on subset of OSTs
6670 - LOV file size/IO offset was wrong for files > 4GB in size
6671 - object EA data was being dropped, caused files to be unreadable
6672 - memory overflow with non-LOV OST caused memory corruption
6673 - fixed regression tests to work with new config tools, obdfilter
6674 - fixed bug when directory size became larger than 1 block
6675 - fixed bug (for single client case) when PWD was deleted
6676 - invalidate local directory pages when doing intent-based ops
6677 - avoid LDLM oops when lock callback contained bad data
6679 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6681 * documentation updates
6682 * bug fixes and cleanups
6683 - configuration tools
6686 - 64-bit compile warnings
6687 - 64-bit internal statfs data
6689 * test_brw on persistent OST devices
6691 * lprocfs (disabled)
6693 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6695 * documentation updates
6696 * bug fixes and cleanups
6697 * configuration tools
6699 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6702 * many small fixes to 0_5_4
6703 * io/network handling
6704 * thinkos in MDS operations
6706 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6709 * crucial basic fixes to 0.5.3
6710 * IOR, Iozone work over Elan
6713 2002-08-07 Phil Schwan <phil@clusterfs.com>
6714 * version 0_5_3, our first alpha
6715 * we use the new Portals iovs
6716 * documentation updates
6717 * bug fixes and cleanups
6718 * small changes in the DLM wire protocol
6720 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6721 * version 0_5_1 with some initial stability,
6722 * locking on MD and file I/O.
6723 * documentation updates
6724 * several bug fixes since 0.5.0
6725 * small changes in wire protocol
6727 2002-07-18 Phil Schwan <phil@clusterfs.com>
6729 * delivered as Lustre Light Alpha
6730 * fixed a crash after handling invalid MDS requests
6731 * fixed directory pages for architectures with non-4k pages sizes
6733 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6734 * release version v0_4_4
6735 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6737 * Fixes a number of interruption problems with OST operations.
6738 * Update documentation for portals header changes
6739 * Move all wire protocol structs/defines to lustre_idl.h
6740 * Fixes symlink length bug.
6741 * Add tcpdump to repository.
6743 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6744 * release version v0_4_3
6745 * Fixes statfs for inodes on extN.
6746 * Fixes bug in runtests which would delete /etc/hosts.
6747 * Use 64-bit object IDs wherever possible (not into VFS though)
6748 Remove ost_get_info, which is unused by lustre, and out of date.
6750 2002-07-03 Peter Braam <braam@clusterfs.com>
6751 * release version v0_4_2 Fixes a lookup error (type not passed)
6752 * move forward to head of Portals
6753 * move forward to latest Lustre kernel
6755 2002-06-25 Peter Braam <braam@clusterfs.com>
6756 * release version v0_4_1. Hopefully stable on single node use.