3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.27.2.el7 (RHEL7.6)
9 3.10.0-1062.el7 (RHEL7.7)
10 4.4.120-92.70 (SLES12 SP2)
11 4.4.180-94.100 (SLES12 SP3)
12 4.4.0-131 (Ubuntu 16.04)
13 4.15.0-32 (Ubuntu 18.04)
14 vanilla linux 4.6.7 (ZFS only)
15 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
16 * Client known to build on unpatched kernels (others may also work):
17 2.6.32-696.18.7.el6 (RHEL6.9)
18 2.6.32-754.6.3.el6 (RHEL6.10)
19 3.10.0-862.14.4.el7 (RHEL7.5)
20 3.10.0-957.27.2.el7 (RHEL7.6)
21 3.10.0-1062.el7 (RHEL7.7)
22 4.14.0-49.13.1.el7a (RHEL7.5)
23 4.14.0-115.2.2.el7a (RHEL7.6)
24 4.18.0-80.4.2.el8 (RHEL8.0)
25 4.4.120-92.70 (SLES12 SP2)
26 4.4.180-94.100 (SLES12 SP3)
27 4.12.14-95.29 (SLES12 SP4)
28 4.4.0-131 (Ubuntu 16.04)
29 4.15.0-48 (Ubuntu 18.04)
30 vanilla linux 4.15.0 (ZFS only)
31 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
32 * Recommended ZFS / SPL version: 0.7.13
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
38 * Note for Ubuntu 14.04.5:
39 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
40 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
42 --------------------------------------------------------------------------------
46 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
47 for currently supported client and server kernel versions.
48 * Server known to build on the following kernels (others may also work):
49 2.6.32-696.18.7.el6 (RHEL6.9)
50 3.10.0-862.14.4.el7 (RHEL7.5)
51 3.10.0-957.el7 (RHEL7.6)
52 4.4.120-92.70 (SLES12 SP2)
53 4.4.162-94.69 (SLES12 SP3)
54 4.4.0-131 (Ubuntu 16.04)
55 4.15.0-32 (Ubuntu 18.04)
56 vanilla linux 4.6.7 (ZFS only)
57 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
58 * Client known to build on unpatched kernels (others may also work):
59 2.6.32-696.18.7.el6 (RHEL6.9)
60 2.6.32-754.6.3.el6 (RHEL6.10)
61 3.10.0-862.14.4.el7 (RHEL7.5)
62 3.10.0-957.el7 (RHEL7.6)
63 4.14.0-49.13.1.el7a (RHEL7.5)
64 4.14.0-115.2.2.el7a (RHEL7.6)
65 4.4.120-92.70 (SLES12 SP2)
66 4.4.162-94.69 (SLES12 SP3)
67 4.4.0-131 (Ubuntu 16.04)
68 4.15.0-32 (Ubuntu 18.04)
69 vanilla linux 4.15.0 (ZFS only)
70 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
71 * Recommended ZFS / SPL version: 0.7.9
72 * Tested with ZFS / SPL version: 0.7.9
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
78 * Note for Ubuntu 14.04.5:
79 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
80 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
82 --------------------------------------------------------------------------------
85 2018-04-03 Intel Corporation
87 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
88 for currently supported client and server kernel versions.
89 * Server known to build on patched kernels:
90 2.6.32-431.29.2.el6 (RHEL6.5)
91 2.6.32-504.30.3.el6 (RHEL6.6)
92 2.6.32-573.26.1.el6 (RHEL6.7)
93 2.6.32-642.15.1.el6 (RHEL6.8)
94 2.6.32-696.18.7.el6 (RHEL6.9)
95 3.10.0-693.21.1.el7 (RHEL7.4)
96 3.0.101-0.47.71 (SLES11 SP3)
97 3.0.101-107 (SLES11 SP4)
98 3.12.74-60.64.40 (SLES12 SP1)
99 4.4.114-92.64 (SLES12 SP2)
100 4.4.114-94.11 (SLES12 SP3)
101 3.13.0-101 (Ubuntu 14.04, ZFS only)
102 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
103 4.4.0-116 (Ubuntu 16.04)
104 vanilla linux 4.6.7 (ZFS only)
105 * Client known to build on unpatched kernels:
106 2.6.32-431.29.2.el6 (RHEL6.5)
107 2.6.32-504.30.3.el6 (RHEL6.6)
108 2.6.32-573.26.1.el6 (RHEL6.7)
109 2.6.32-642.15.1.el6 (RHEL6.8)
110 2.6.32-696.18.7.el6 (RHEL6.9)
111 3.10.0-693.21.1.el7 (RHEL7.4)
112 3.0.101-0.47.71 (SLES11 SP3)
113 3.0.101-107 (SLES11 SP4)
114 3.12.74-60.64.40 (SLES12 SP1)
115 4.4.114-92.64 (SLES12 SP2)
116 4.4.114-94.11 (SLES12 SP3)
117 3.13.0-101 (Ubuntu 14.04)
118 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
119 4.4.0-116 (Ubuntu 16.04)
121 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
122 * Recommended ZFS / SPL version: 0.7.6
123 * Tested with ZFS / SPL version: 0.7.6
124 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
125 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
126 stack overflow. For more information, please refer to bugzilla 17630.
127 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
128 the re-exporting nodes to be mounted with "32bitapi" mount option
129 * Note for Ubuntu 14.04.5:
130 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
131 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
133 --------------------------------------------------------------------------------
135 2017-07-05 Intel Corporation
137 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
138 for currently supported client and server kernel versions.
139 * Server known to build on patched kernels:
140 2.6.32-431.29.2.el6 (RHEL6.5)
141 2.6.32-504.30.3.el6 (RHEL6.6)
142 2.6.32-573.26.1.el6 (RHEL6.7)
143 2.6.32-642.15.1.el6 (RHEL6.8)
144 2.6.32-696.3.1.el6 (RHEL6.9)
145 3.10.0-514.21.1.el7 (RHEL7.3)
146 3.0.101-0.47.71 (SLES11 SP3)
147 3.0.101-100 (SLES11 SP4)
148 3.12.74-60.64.40 (SLES12 SP1)
149 4.4.59-92.17 (SLES12 SP2)
150 vanilla linux 4.6.7 (ZFS only)
151 * Client known to build on unpatched kernels:
152 2.6.32-431.29.2.el6 (RHEL6.5)
153 2.6.32-504.30.3.el6 (RHEL6.6)
154 2.6.32-573.26.1.el6 (RHEL6.7)
155 2.6.32-642.15.1.el6 (RHEL6.8)
156 2.6.32-696.3.1.el6 (RHEL6.9)
157 3.10.0-514.21.1.el7 (RHEL7.3)
158 3.0.101-0.47.71 (SLES11 SP3)
159 3.0.101-100 (SLES11 SP4)
160 3.12.74-60.64.40 (SLES12 SP1)
161 4.4.59-92.17 (SLES12 SP2)
163 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
164 * Recommended ZFS / SPL version: 0.6.5.9
165 * Tested with ZFS / SPL version: 0.6.5.9
166 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
167 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
168 stack overflow. For more information, please refer to bugzilla 17630.
169 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
170 the re-exporting nodes to be mounted with "32bitapi" mount option
172 --------------------------------------------------------------------------------
174 12-07-2016 Intel Corporation
176 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
177 for currently supported client and server kernel versions.
178 * Server known to build on patched kernels:
179 2.6.32-431.29.2.el6 (RHEL6.5)
180 2.6.32-504.30.3.el6 (RHEL6.6)
181 2.6.32-573.26.1.el6 (RHEL6.7)
182 2.6.32-642.11.1.el6 (RHEL6.8)
183 3.10.0-514.el7 (RHEL7.3)
184 3.0.101-0.47.71 (SLES11 SP3)
185 3.0.101-80 (SLES11 SP4)
186 3.12.62-60.64.8 (SLES12 SP1)
187 vanilla linux 4.5.7 (ZFS only)
188 * Client known to build on unpatched kernels:
189 2.6.32-431.29.2.el6 (RHEL6.5)
190 2.6.32-504.30.3.el6 (RHEL6.6)
191 2.6.32-573.26.1.el6 (RHEL6.7)
192 2.6.32-642.11.1.el6 (RHEL6.8)
193 3.10.0-514.el7 (RHEL7.3)
194 3.0.101-0.47.71 (SLES11 SP3)
195 3.0.101-80 (SLES11 SP4)
196 3.12.62-60.64.8 (SLES12 SP1)
198 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
199 * Recommended ZFS / SPL version: 0.6.5.8
200 * Tested with ZFS / SPL version: 0.6.5.8
201 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
202 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
203 stack overflow. For more information, please refer to bugzilla 17630.
204 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
205 the re-exporting nodes to be mounted with "32bitapi" mount option
207 --------------------------------------------------------------------------------
209 02-29-2016 Intel Corporation
211 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
212 for currently supported client and server kernel versions.
213 * Server known to build on patched kernels:
214 2.6.32-431.29.2.el6 (RHEL6.5)
215 2.6.32-504.30.3.el6 (RHEL6.6)
216 2.6.32-573.12.1.el6 (RHEL6.7)
217 3.10.0-327.3.1.el7 (RHEL7.2)
218 3.0.101-0.47.71 (SLES11 SP3)
219 3.0.101-68 (SLES11 SP4)
220 vanilla linux 4.2.1 (ZFS only)
221 * Client known to build on unpatched kernels:
222 2.6.32-431.29.2.el6 (RHEL6.5)
223 2.6.32-504.30.3.el6 (RHEL6.6)
224 2.6.32-573.12.1.el6 (RHEL6.7)
225 3.10.0-327.3.1.el7 (RHEL7.2)
226 3.0.101-0.47.71 (SLES11 SP3)
227 3.0.101-68 (SLES11 SP4)
230 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
231 * Recommended ZFS / SPL version: 0.6.4.2
232 * Tested with ZFS / SPL version: 0.6.4.2
233 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
234 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
235 stack overflow. For more information, please refer to bugzilla 17630.
236 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
237 the re-exporting nodes to be mounted with "32bitapi" mount option
239 --------------------------------------------------------------------------------
241 03-10-2015 Intel Corporation
243 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
244 for currently supported client and server kernel versions.
245 * Server known to build on patched kernels:
246 2.6.32-431.29.2.el6 (RHEL6.5)
247 2.6.32-504.8.1.el6 (RHEL6.6)
248 3.0.101-0.46 (SLES11 SP3)
249 * Client known to build on unpatched kernels:
250 2.6.32-431.29.2.el6 (RHEL6.5)
251 2.6.32-504.8.1.el6 (RHEL6.6)
252 3.10.0-123.20.1.el7 (RHEL7)
253 3.0.101-0.46 (SLES11 SP3)
254 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
255 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
256 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
257 stack overflow. For more information, please refer to bugzilla 17630.
258 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
259 the re-exporting nodes to be mounted with "32bitapi" mount option
261 Severity : enhancement
263 Description: control OST-index in IDIF via ROCOMPAT flag.
264 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
265 last_rcvd file. For new formatted OST device, it will be auto set;
266 for the case of upgrading from old OST device, you can enable it
267 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
268 enabled, for new created OST-object, its IDIF-in-LMA will contain
269 the OST-index; for the existing OST-object, the OSD will convert
270 old format IDIF as new format IDIF with OST-index stored in the
271 LMA EA when accessing such OST-object or via OI scrub. Once such
272 flag is enabled, it cannot be reverted back, so the system cannot
273 be downgraded to the orignal incompatible version.
275 --------------------------------------------------------------------------------
277 07-30-2014 Intel Corporation
279 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
280 for currently supported client and server kernel versions.
281 * Server known to build on patched kernels:
282 2.6.32-431.20.3.el6 (RHEL6.5)
283 3.0.101-0.31 (SLES11 SP3)
284 * Client known to build on unpatched kernels:
285 2.6.32-431.20.3.el6 (RHEL6.5)
286 3.0.101-0.31 (SLES11 SP3)
289 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
290 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
291 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
292 stack overflow. For more information, please refer to bugzilla 17630.
293 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
294 the re-exporting nodes to be mounted with "32bitapi" mount option
296 --------------------------------------------------------------------------------
298 10-31-2013 Intel Corporation
300 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
301 * Server builds for kernels:
302 2.6.32-358.18.1.el6 (RHEL6.4)
303 * Client builds for unpatched kernels:
304 2.6.32-358.18.1.el6 (RHEL6.4)
305 3.0.92-0.8 (SLES11 SP3)
307 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
308 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
309 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
310 stack overflow. For more information, please refer to bugzilla 17630.
311 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
312 the re-exporting nodes to be mounted with "32bitapi" mount option
314 --------------------------------------------------------------------------------
316 05-30-2013 Intel Corporation
318 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
319 * Server builds for kernels:
320 2.6.32-358.6.2.el6 (RHEL6.4)
321 * Client builds for unpatched kernels:
322 2.6.32-358.6.2.el6 (RHEL6.4)
323 3.0.74-0.6.6 (SLES11 SP2)
325 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
326 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
327 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
328 stack overflow. For more information, please refer to bugzilla 17630.
329 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
330 the re-exporting nodes to be mounted with "32bitapi" mount option
332 --------------------------------------------------------------------------------
334 9-30-2012 Whamcloud, Inc.
336 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
337 * Server support for kernels:
338 2.6.32-279.5.1.el6 (RHEL6)
339 * Client support for unpatched kernels:
340 2.6.32-279.5.1.el6 (RHEL6)
342 2.6.18-238.19.1.el5 (RHEL5)
343 2.6.16.60-0.69.1 (SLES10)
344 * Recommended e2fsprogs version: 1.41.90.wc4
345 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
346 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
347 stack overflow. For more information, please refer to bugzilla 17630.
348 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
349 the re-exporting nodes to be mounted with "32bitapi" mount option
351 Severity : enhancement
353 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
355 Severity : enhancement
357 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
359 Severity : enhancement
361 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
363 Severity : enhancement
365 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
367 Severity : enhancement
369 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
371 --------------------------------------------------------------------------------
373 04-30-2012 Whamcloud, Inc.
375 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
376 * Server support for kernels:
377 2.6.32-220.4.2.el6 (RHEL6)
378 * Client support for unpatched kernels:
379 2.6.32-220.4.2.el6 (RHEL6)
381 2.6.18-238.19.1.el5 (RHEL5)
382 2.6.16.60-0.69.1 (SLES10)
383 * Recommended e2fsprogs version: 1.41.90.wc4
384 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
385 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
386 stack overflow. For more information, please refer to bugzilla 17630.
387 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
388 the re-exporting nodes to be mounted with "32bitapi" mount option
390 Severity : enhancement
392 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
394 Severity : enhancement
396 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
398 Severity : enhancement
400 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
402 Severity : enhancement
404 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
406 --------------------------------------------------------------------------------
408 09-30-2011 Whamcloud, Inc.
410 * Server support for kernels:
411 2.6.18-238.19.1.el5 (RHEL5)
412 2.6.32-131.6.1.el6 (RHEL6)
413 * Client support for unpatched kernels:
414 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
415 2.6.18-238.19.1.el5 (RHEL5)
416 2.6.32-131.6.1.el6 (RHEL6)
417 2.6.29.4-167.fc11 (FC11)
418 2.6.16.60-0.69.1 (SLES10)
420 * Recommended e2fsprogs version: 1.41.90.wc3
421 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
422 Lustre filesystem with 4K stack may cause a stack overflow. For more
423 information, please refer to bugzilla 17630.
424 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
425 mounted with "32bitapi" mount option
427 Severity : enhancement
428 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
430 Severity : enhancement
431 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
433 Severity : enhancement
434 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
437 Frequency : only with SLES10
439 Description: Use OFED "KMP" provided by Novell
440 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
441 Lustre is now built against this package. That means you need to
442 install the ofed-kmp package from Novell for the patchless client.
443 Note that the ofed-kmp that Novell ships may not exactly match the
444 kernel version but should still be compatible.
446 Severity : enhancement
448 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
449 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
451 --------------------------------------------------------------------------------
453 2010-07-15 Oracle, Inc.
455 * Server support for kernels:
456 2.6.18-164.11.1.el5 (RHEL 5)
457 2.6.18-164.11.1.0.1.el5 (OEL 5)
458 * Client support for unpatched kernels:
459 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
460 2.6.18-164.11.1.el5 (RHEL 5),
461 2.6.18-164.11.1.0.1.el5 (OEL 5)
462 2.6.16.60-0.42.8 (SLES 10),
464 2.6.29.4-167.fc11 (FC11)
465 * Recommended e2fsprogs version: 1.41.10-sun2
467 Severity : enhancement
469 Description: Distribute lfs_migrate script with lustre
470 Details : The lfs_migrate mentioned in the manual is now included in the
471 lustre tree in lustre/scripts and will distributed in lustre RPMs.
473 Severity : enhancement
475 Description: Failover nid list added to import proc file.
477 Severity : enhancement
479 Description: "som_preview" mount option required to use SOM
480 Details : The clients will need to use som_preview mount option
484 Description: Lustre does not do 1MB IOs to HW RAID
485 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
487 Severity : enhancement
489 Description: lfs quota output is not very convenient for awk/sed-parsing
490 Details : Some positions in lfs quota output table could be empty or
491 non-empty which made it hard to parse it with scripts, now
492 a dash is put instead of space where there is not supposed
497 Description: Update to OFED 1.5.1
498 Details : OFED 1.5.1 will be used in environments that don't use
499 vendor supplied OFED.
503 Description: Remove lnet. prefix from lctl params display
504 Details : Also renamed lnet.memused to lnet_memused to
509 Description: Use conf_param -d to remove permanant settings.
513 Description: Add list_param -R
514 Details : lctl list_param now supports recursive mode.
517 Bugzilla : 21829, 22312
518 Description: Fix llobdstat.
519 Details : 1. Now updating the write_bytes counter
520 2. Port script fixes from 1.8.
524 Description: decrease the usage of memory on clients.
525 Details : 1. On clients, recycle dentries and inodes unused.
526 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
531 Description: llobdstat fix and enhancement
532 Details : add a counter to set a limit to how many samples will be returned
533 fix a wildcard in the path to limit to obdfilter stats only
537 Description: lfs check only allowed for root
538 Details : Code cleanups and added a regresion test of non-root use of
543 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
544 Details : Fix hr->hr_index overload.
548 Description: Fix FIEMAP functionality.
552 Description: FC11 patchless client support added
556 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
557 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
561 Description: New acc-sm script and various test-framework enhancments.
565 Description: Previously landed patch was reverted.
569 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
573 Description: Add ldiskfs maxdirsize mount option.
577 Description: Fix mkfs.lustre support for 16TB LUNs.
581 Description: Print mmp_check_interval
582 Details : print mmp_check_interval and make it possible to abort
583 mount operation in case it takes too long.
587 Description: Add cascading_rw to lustre mpi tests.
591 Description: Store SOM data in EA.
592 Details : Also several SOM related cleanups.
596 Description: Add refcount to md_open_data
600 Description: Use refcount on lov_request_set to prevent memory corruption.
604 Description: Statahead: use parent dir's i_mutex for synchronization.
608 Description: loadgen improvements.
612 Description: Enable peer health detection in ksocklnd and ko2iblnd.
616 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
620 Description: Fix race on obd_devcie:md_stats
624 Description: Fix SLES 11 client builds.
625 Details : Resolved uninitalized pointer gcc warnings.
629 Description: Allow non-root access for "lfs check".
630 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
634 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
636 Severity : enhancement
638 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
639 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
643 Description: Sanity test_27p fails.
644 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
645 Sanity tests 27 were cleaned up and improved.
649 Description: Unify req format on client/servers
650 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
651 buffers and swabbers approach..
656 Description: drop join file stuff.
658 Severity : enhancement
660 Description: Add 2.6.27 and SLES11 (patchless client) support
664 Description: Fix in ptlrpc_expire_one_request() to print the signed time
667 Severity : enhancement
669 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
672 Severity : enhancement
674 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
676 Severity : enhancement
678 Description: correctly handle big reply message.
679 Details : send LNet event if reply is bigger than buffer and adjust this buffer
684 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
686 Severity : enhancement
688 Description: Add support for OEL5.
692 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
694 Severity : enhancement
696 Description: "weak-modules" support
697 Details : Implement "weak-modules" support which enables kernel modules
698 to be used with any kernel that implements the same kABI. In
699 order to achieve this modules are now installed in
700 /lib/modules/$(uname -r)/updates/kernel on all distributions.
704 Description: client could not reconnect to OST because of an active request.
705 Details : The client now retries to reconnect to the same server, if a
706 connect request failed with EBUSY or -EAGAIN. Test case 35b
707 conf-sanity.sh has been created.
711 Description: Don't allow make backward step on assignin osc next id.
712 Details : race between allocation next id and ll_sync thread can be cause
713 of set wrong osc next id and can be kill valid ost objects.
717 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
721 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
725 Description: Error handling in osc_statfs_interpret() has been improved.
726 Details : Check in osc_statfs_interpret() for EBADR.
730 Description: Do not send statfs() requests to OSTs disabled by administrator.
731 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
733 Severity : enhancement
735 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
737 Severity : enhancement
739 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
741 Severity : enhancement
743 Description: Enhancement for lfs(1) command to use numeric uid/gid.
747 Description: Improved cleanup procedure for conf-sanity test_46a.
751 Description: Do not start bulk transfer, if client tried to reconnect.
752 Details : Do not start bulk transfer, if client tried to reconnect,
753 but failed due to active rpcs. Check for exp_abort_active_req
754 has been implemented.
758 Description: Prevent inconsistences between linux and lustre mount structures.
759 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
760 Make the sleep interruptible.
762 Severity : enhancement
764 Description: provide server to client comms path
765 Details : server to client communications path via new LDLM_SET_INFO rpc
769 Description: truncate starts GFP_FS allocation under transaction causing deadlock
770 Details : ldiskfs_truncate calls grab_cache_page which may start page
771 allocation under an open transaction. This may lead to
772 calling prune_icache with consequent lustre reentrance.
776 Description: send correctly lsm on open replay
777 Details : MDS is trust to LSM size on replay open, but client can set wrong size
782 Description: enable client interface failover
783 Details : When a child reconnects from another NID, properly update export
784 nid hash position and ldlm reverse import.
786 Severity : enhancement
788 Description: Communicate OST degraded/readonly state via statfs to MDS
789 Details : Flags in the statfs returned from OSTs indicate whether the
790 OST is in a degraded RAID state, or if the filesystem has
791 turned read-only after a filesystem error is detected.
796 Description: don't panic if EPROTO was hit when reading symlink
797 Details : correctly handling request reference in error cases.
799 Severity : enhancement
801 Description: Add LustreNetLink, a kernel-userspace communcation path.
803 Severity : enhancement
805 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
810 Description: access to llog context before init.
811 Details : move handling CATALOGS file at osc layer and forbid access to llog
815 Frequency : rare, only when client is writing to OST on local node
817 Description: Avoid deadlock for local client writes
818 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
819 memory freeing context. This allows OST threads to set the
820 PF_MEMALLOC flag on task structures in order to allocate memory
821 from reserved pools and complete IO.
822 Use GFP_HIGHUSER for OST allocations for non-local client writes,
823 so that the OST threads generate memory pressure and allow
824 inactive pages to be reclaimed.
826 Severity : enhancement
828 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
833 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
834 Details : move ldlm namespace creation in setup phase to avoid grab
835 _lprocfs_lock with cli_sem held.
839 Description: Temporarily disable grant shrink.
840 Details : Disable the feature for debugging.
843 Frequency : only when formatting test filesystems
845 Description: Unable to run several mkfs.lustre on loop devices at the same time
846 Details : mkfs.lustre returns error 256 on the concurrent loop devices
847 formatting. The solution is to proper handle the error.
849 Severity : enhancement
851 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
853 Severity : enhancement
855 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
860 Description: resolve race between obd_disconnect and class_disconnect_exports
861 Details : if obd_disconnect will be called to already disconnected export he
862 forget release one reference and osc module can't unloaded.
864 Severity : enhancement
866 Description: Allow tuning service thread via /proc
867 Details : For each service a new
868 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
869 created that can be used to set min/max thread counts, and get the
870 current number of running threads.
875 Descriptoin: don't return error if have partially created objects for file.
876 Details : lov_update_create_set uses set->set_success as index for created
877 objects, so if some requests will be failed, they will have hole
878 at end of array and qos_shrink_lsm for allocate correct lsm.
880 Severity : enhancement
882 Description: Update OFED support to 1.4.1
887 Description: Reduce small size read RPC
888 Details : Set read-ahead limit for every file and only do read-ahead when
889 available read-ahead pages are bigger than 1M to avoid small size
892 Severity : enhancement
894 Description: move AT tunable parameters for more consistent usage
895 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
897 Severity : enhancement
899 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
900 Details : allow skip disconnected ost for send statfs request and hide error
906 Description: fix racy locking of mballoc block bitmaps causing BUG
907 Details : The locking of the mballoc buddy bitmap and the in-memory
908 block bitmap was using two different spin locks in some
909 cases. This made it possible to incorrectly access the
910 mballoc bitmap while another process was modifying it,
911 causing a sanity assertion to fail. While no on-disk corruption
912 was reported, there was some risk of this happening.
915 Frequency : rare, on llog test 6
917 Descriptoin: don't allow connect to already connected import
918 Details : allowing connect to already connected import is hide connecting problem.
923 Descriptoin: allow kill process which wait statahead result
924 Details : for some reasons 'ls' can stick in waiting result from statahead,
925 in this case need way for kill this process.
927 Severity : enhancement
929 Description: Add state history info file, enhance import info file
930 Details : Track import connection state changes in a new osc/mdc proc file;
931 add overview-type data to the osc/mdc import proc file.
933 Severity : enhancement
935 Description: MDS create should not wait for statfs RPC while holding DLM lock.
937 Severity : enhancement
939 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
944 Descriptoin: OST grant too much space to client even there are not enough space.
945 Details : Client will shrink its grant cache to OST if there are no write
946 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
947 this grant cache if there are already not enough avaible space
948 (left_space < total_clients * 32M).
951 Frequency : start MDS on uncleanly shutdowned MDS device
953 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
954 Details : stay in waiting mds<>ost recovery finished produce random bugs
955 due race between two ll_sync thread for one lov target. send
956 ACTIVATE event only if connect realy finished and import have
960 Frequency : rare, connect and disconnect target at same time
962 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
963 Details : don't call obd_disconnect under lov_lock. this long time
964 operation and can block ptlrpcd which answer to connect request.
969 Descriptoin: don't lose wakeup for imp_recovery_waitq
970 Details : recover_import_no_retry or invalidate_import and import_close can
971 both sleep on imp_recovery_waitq, but we was send only one wakeup
975 Frequency : always with long access acl
977 Descriptoin: mds can't pack reply with long acl.
978 Details : mds don't control size of acl but they limited by reint/getattr
981 Severity : enhancement
983 Description: Update to SLES10 kernel-2.6.16.60-0.33.
985 Severity : enhancement
987 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
990 Frequency : start MDS on uncleanly shutdowned MDS device
992 Descriptoin: aborting recovery hang on MDS
993 Details : don't throttle destroy RPCs for the MDT.
996 Frequency : on remount
998 Description: external journal device not working after the remount
999 Details : clear dev_rdonly flag for external journal devices in
1005 Description: shutdown vs evict race
1006 Details : client_disconnect_export vs connect request race.
1007 if client will evicted at this time - we start invalidate
1008 thread without referece to import and import can be freed
1012 Frequency : rare, need acl's on inode.
1014 Description: client can't handle ost additional correctly
1015 Details : if ost was added after client connected to mds client can have
1016 hit lnet_try_match_md ... to big messages to wide striped files.
1017 in this case need teach client to handle config events about add
1018 lov target and update client max ea size at that event.
1020 Severity : enhancement
1022 Description: Changelogs
1023 Details : Changelogs are a lightweight mechanism to track filesystem
1024 metadata and namespace changes. The changelog is recorded
1025 permanently on the MDTs, and is periodically "consumed" / purged
1026 when records are no longer needed.
1028 Severity : enhancement
1030 Description: compact fld format with extents
1031 Details : Store range of seq rather than every seq in FLD. Seq
1032 controller update FLD rather than clients. In Case of CMD, mdt0
1033 has FLD, all other metadata server act as non persistent proxy
1034 for FLD queries and cache fld entries in fld cache.
1039 Description: don't skip ost target if they assigned to file
1040 Details : Drop slow OSCs if we can, but not for requested start idx.
1041 This means "if OSC is slow and it is not the requested
1042 start OST, then it can be skipped, otherwise skip it only
1043 if it is inactive/recovering/out-of-space.
1047 Description: more cleanup in mds_lov
1048 Details : not send LOV EA under replay, we can't know about they size at this
1049 time. Don't allow client connect to mds before any ost connected,
1050 for avoid problems with LOV EA size and returning EIO to client.
1052 Severity : enhancement
1054 Description: Interoperability at server side (Disk interoperability)
1056 Severity : enhancement
1058 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1060 Severity : enhancement
1062 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1064 Severity : enhancement
1066 Description: New client IO stack (CLIO).
1068 Severity : enhancement
1070 Description: Commit on sharing. Eliminate inter-client dependencies between
1071 uncommitted transactions by doing transaction commits.
1072 Thereby clients may recovery independently.
1075 Frequency : Create a symlink file with a very long name
1077 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1078 Details : If there is no extra space in the request for early cancels,
1079 ldlm_req_handles_avail() returns 0 instead of a negative value.
1081 Severity : enhancement
1083 Description: Add /proc entry for import status
1084 Details : The mdc, osc, and mgc import directories now have
1085 an import directory that contains useful import data for debugging
1086 connection problems.
1088 Severity : enhancement
1090 Description: Re-disable certain /proc logging
1091 Details : Enable and disable client's offset_stats, extents_stats and
1092 extents_stats_per_process stats logging on the fly.
1095 Frequency : Only on FC kernels 2.6.22+
1097 Description: oops in statahead
1098 Details : Do not drop reference count for the dentry from VFS when lookup,
1099 VFS will do that by itself.
1101 Severity : enhancement
1103 Description: Generic /proc file permissions
1104 Details : Set /Proc file permissions in a more generic way to enable non-
1105 root users operate on some /proc files.
1109 Description: Hitting mdc_commit_close() ASSERTION
1110 Details : Properly handle request reference release in
1111 ll_release_openhandle().
1115 Description: quota recovery deadlock during mds failover
1116 Details : This patch includes att18982, att18236, att18237 in bz14840.
1118 1. fix osts hang when mds does failover with quotaon
1119 2. prevent watchdog storm when osts threads wait for the
1124 Frequency : only patchless client
1125 Description: add workaround for race between add/remove dentry from hash
1127 Severity : enhancement
1129 Description: Allow OST glimpses to return PW locks
1133 Description: LBUG when llog conf file is full
1134 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1138 Description: Prevent import from entering FULL state when server in recovery
1142 Description: service mount cannot take device name with ":"
1143 Details : Only when device name contains ":/" will mount treat it as
1149 Description: replace ptlrpcd with the statahead thread to interpret the async
1150 statahead RPC callback
1154 Frequency : on recovery
1155 Description: I/O failures after umount during fail back
1156 Details : if client reconnected to restarted server we need join to recovery
1157 instead of find server handler is changed and process self
1158 eviction with cancel all locks.
1160 Severity : enhancement
1162 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1164 Severity : enhancement
1166 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1168 Severity : enhancement
1170 Description: Upcall on Lustre log has been dumped
1171 Details : Allow for a user mode script to be called once a Lustre log has
1172 been dumped. It passes the filename of the dumped log to the
1173 script, the location of the script can be specified via
1174 /proc/sys/lnet/debug_log_upcall.
1179 Description: avoid idr_remove called for id which is not allocated.
1180 Details : Move assigment s_dev for clustered nfs to end of initialization,
1181 to avoid problem with error handling.
1186 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1187 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1188 into CONN_USED_HASH and this prodice warning when put connection
1189 again in unused hash.
1192 Severity : enhancement
1194 Description: File striping can now be set to use an arbitrary pool of OSTs.
1196 Severity : enhancement
1198 Description: Export bytes_read/bytes_write count on OSC/OST.
1202 Description: Early reply size mismatch, MGC loses connection
1203 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1204 the connect flags are properly negotiated.
1209 Description: quotas are not honored with O_DIRECT
1210 Details : all writes with the flag O_DIRECT will use grants which leads to
1211 this problem. Now using OBD_BRW_SYNC to guard this.
1215 Description: add quota statistics
1216 Details : 1. sort out quota proc entries and proc code.
1217 2. add quota statistics
1219 Severity : enhancement
1221 Description: enable quota support for HEAD.
1225 Description: Properly propagate oinfo flags from lov to osc for statfs
1226 Details : restore missing copy oi_flags to lov requests.
1228 Severity : enhancement
1230 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1231 : l_getgroups(8), lst(8), routerstat(8)
1232 Details : included man pages for llobdstat(8), llstat(8),
1233 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1235 Severity : enhancement
1237 Description: configure's --enable-quota should check the
1238 : kernel .config for CONFIG_QUOTA
1239 Details : configure is terminated if --enable-quota is passed but
1240 : no quota support is in kernel
1244 Description: Remove portals compatibility
1245 Details : Remove portals compatibility, not interoperable with releases
1250 Description: Resolve device initialization race
1251 Details : Prevent proc handler from accessing devices added to the
1252 obd_devs array but yet be intialized.
1254 Severity : enhancement
1256 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1258 Severity : enhancement
1260 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1265 Description: Using wrong pointer in osc_brw_prep_request
1266 Details : Access to array[-1] can produce panic if kernel compiled with
1267 CONFIG_PAGE_ALLOC enabled
1271 Description: Client runs out of low memory
1272 Details : Consider only lowmem when counting initial number of llap pages
1276 Description: *optional* service tags registration
1277 Details : if the "service tags" package is installed on a Lustre node
1278 When the filesystem is mounted, a local-node service tag will
1279 be created. See http://inventory.sun.com/ for more information
1280 about the Service Tags asset management system.
1284 Description: Kernel BUG tries to release flock
1285 Details : Lustre does not destroy flock lock before last reference goes
1286 away. So always drop flock locks when client is evicted and
1287 perform unlock regardless of successfulness of speaking to MDS.
1291 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1295 Description: sanity test 65a fails if stripecount of -1 is set
1296 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1301 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1302 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1305 Severity : enhancement
1307 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1311 Description: do not process already freed flock
1312 Details : flock can possibly be freed by another thread before it reaches
1313 to ldlm_flock_completion_ast.
1317 Description: LBUG during stress test
1318 Details : Need properly lock accesses the flock deadlock detection list.
1322 Description: oops in page fault handler
1323 Details : kernel page fault handler can return two special 'pages' in error
1324 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1328 Description: timeout with invalidate import.
1329 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1330 should be handled by ptlrpcd. This produce long age waiting and
1331 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1333 Severity : enhancement
1335 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1339 Description: enable MGS and MDT services start separately
1340 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1341 a co-located MGS without starting the MGS, which is a complement
1342 to 'nosvc' mount option.
1346 Frequency : after recovery
1347 Description: precreate to many object's after del orphan.
1348 Details : del orphan st in oscc last_id == next_id and this triger growing
1349 count of precreated objects. Set flag LOW to skip increase count
1350 of precreated objects.
1354 Frequency : rare, on clear nid stats
1355 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1356 Details : when clean nid stats sometimes try destroy live entry,
1357 and this produce panic in free.
1361 Description: Stack overflow during MDS log replay
1362 ease stack pressure by using a thread dealing llog_process.
1366 Description: wait until IO finished before start new when do lock cancel.
1367 Details : VM protocol want old IO finished before start new, in this case
1368 need wait until PG_writeback is cleared until check dirty flag and
1369 call writepages in lock cancel callback.
1371 Severity : enhancement
1373 Description: using special macro for print time and cleanup in includes.
1377 Description: mds_mfd_close() ASSERTION(rc == 0)
1378 Details : In mds_mfd_close(), we need protect inode's writecount change
1379 within its orphan write semaphore to prevent possible races.
1383 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1387 Frequency : rare, on shutdown ost
1388 Description: don't hit live lock with umount ost.
1389 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1390 use shrink_dcache_sb instead.
1394 Description: don't panic with use echo client
1395 Details : echo client pass NULL as client nid pointer and this produce null
1396 pointer dereference.
1400 Description: fix build on ppc32
1401 Details : compile code with -m64 flag produce wrong object file for ppc32.
1405 Description: add message levels for liblustreapi
1409 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1410 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1411 will result in an -EAGAIN and in the exit call path we call
1412 alloc_rr() which will with increasing aggressiveness attempt to
1413 aquire precreated objects on the minimum number of required OSCs.
1417 Description: Use old size assignment to avoid deadlock
1418 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1419 the scheduling while holding a spinlock. We do not need locking
1420 for size in ll_update_inode() because size is only updated from
1421 the MDS for directories or files without objects, so there is no
1422 other place to do the update, and concurrent access to such inodes
1423 are protected by the inode lock.
1427 Description: resolve "_IOWR redefined" build error on SLES10
1431 Description: dump the memory debugging after all modules are unloaded to
1432 suppress false negative in conf_sanity test 39
1434 Severity : enhancement
1436 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1441 Description: fix SLES kernel versioning
1442 Details : the kernel version for our SLES 10 kernel did not include a "-"
1443 before the "smp" at the end. while this was not a problem in
1444 general, it did mean that software trying to use the kernel
1445 version to try to detect a vendor specific kernel would fail.
1446 this was most evident by the OFED build scripts.
1450 Description: Don't update lov_desc members until making sure they are valid
1451 Details : When updating lov_desc members via proc fs, need fix their
1452 validities before doing the real update.
1456 Description: don't put request into delay list while invalidate in flight.
1457 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1458 import in flight. this produce timeout for invalidate and sometimes
1459 can cause stale data.
1463 Frequency : on ppc only
1464 Description: not convert ost objects for directory because it's not exist.
1465 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1469 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1471 Severity : enhancement
1473 Description: Update to support 2.6.22.14 vanilla kernel.
1477 Frequency : rare, on recovery
1478 Description: read procfs can produce deadlock in some situation
1479 Details : Holding lprocfs lock which send rpc can produce block for destroy
1480 obd objects and this also block reconnect with -EALREADY. This
1481 isn't fix all lprocfs bugs - but make it rare.
1483 Severity : enhancement
1485 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1488 Frequency : frequent on X2 node
1490 Description: mdc_set_open_replay_data LBUG
1491 Details : Set replay data for requests that are eligible for replay.
1495 Description: lustre_mgs: operation 101 on unconnected MGS
1496 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1497 MGC, and late on MGC cannot successfully connect to MGS and a lot
1498 of the error messages complaining that MGS is not connected.
1502 Frequency : on network error
1503 Description: panic with double free request if network error
1504 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1505 but it's true only for synchronus enqueue, for async enqueue
1506 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1509 Severity : enhancement
1511 Description: client-side metadata stat-ahead during readdir(directory readahead)
1512 Details : perform client-side metadata stat-ahead when the client detects
1513 readdir and sequential stat of dir entries therein
1516 Frequency : on start mds
1518 Description: Implement get_info(last_id) in obdfilter.
1521 Frequency : occasional
1523 Description: Correctly check stale fid, not start epoch if ost not support SOM
1524 Details : open with flag O_CREATE need set old fid in op_fid3 because
1525 op_fid2 was overwritten with new generated fid, but mds can answer
1526 with one of these two fids and both is not stale. Setattr
1527 incorrectly started an epoch and assume will be called
1528 done_writing, but without SOM done_writing ever being called.
1531 Frequency : rare, depends on device drivers and load
1533 Description: MDS or OSS nodes crash due to stack overflow
1534 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1535 In some cases, in conjunction with device drivers that use a lot
1536 of stack the MDS (or possibly OSS) service threads could overflow
1537 the stack. One change which was identified to consume additional
1538 stack has been reworked to avoid the extra stack usage.
1541 Frequency : occasional
1543 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1544 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1545 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1546 oscc_has_objects and see OSC has no objests and call
1547 oscc_internal_create to resend create request.
1549 Severity : enhancement
1551 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1553 Severity : enhancement
1555 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1558 Frequency : very rare
1560 Description: Fix replay if there is an un-replied request and open
1561 Details : In some cases, older replay request will revert the
1562 mcd->mcd_last_xid on MDS which is used to record the client's
1563 latest sent request.
1565 Severity : enhancement
1567 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1569 Severity : enhancement
1571 Description: Add rhel5 support to HEAD.
1573 Serverity : enhancement
1575 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1580 Description: Don't allow skipping OSTs if index has been specified.
1581 Details : Don't allow skipping OSTs if index has been specified, make
1582 locking in internal create lots better.
1586 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1587 Details : access to bitfield in structure is always rounded to long
1588 and this produce problem with not atomic change any bit.
1592 Description: Lustre make rpms failed.
1593 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1594 builds Lustre rpms from tarball.
1597 Frequency : rare on shutdown ost
1599 Description: If llog cancel was not send before clean_exports phase, this can
1600 produce deadlock in llog code.
1601 Details : If llog thread has last reference to obd and call class_import_put
1602 this produce deadlock because llog_cleanup_commit_master wait when
1603 last llog_commit_thread exited, but this never success because was
1604 called from llog_commit_thread.
1608 Description: allow userland application know is lost one of stripes.
1609 Details : fill lvb_blocks with error code on ost and return it to
1610 application if error flag found.
1614 Description: NULL lov_tgts causing MDS oops
1615 Details : more safe checks for NULL lov_tgts for avoid oops.
1617 Severity : enhancement
1619 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1623 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1624 Details : If an OST is down the MDS will hang indefinitely in
1625 obd_statfs_rqset() waiting for the statfs data. While for
1626 MDS QOS usage of statfs, it should not stuck in waiting.
1628 Severity : enhancement
1630 Description: remote_acl support
1631 Details : Support ACL-based permission check for remote user.
1632 Support setfacl/getfacl for remote user with the utils
1633 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1634 the system "{s,g}etfacl" utils.
1636 Severity : enhancement
1638 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1640 Severity : enhancement
1642 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1646 Description: make mgs_setparam() handle fsname containing dash
1647 Details : fsname containing a dash does not work with lctl conf_param
1649 Severity : enhancement
1651 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1653 Severity : enhancement
1655 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1659 Frequency : rare, at shutdown
1660 Description: access already free / zero obd_namespace.
1661 Details : if client_disconnect_export was called without force flag set,
1662 and exist connect request in flight, this can produce access to
1663 NULL pointer (or already free pointer) when connect_interpret
1664 store ocd flags in obd_namespace.
1668 Frequency : only at startup
1669 Description: not alloc memory with spinlock held.
1670 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1671 if any spinlock held.
1673 Severity : enhancement
1675 Description: make lustre randomly fail allocating memory
1676 Details : Make lustre randomly failed allocating memory for testing purpose.
1678 Severity : enhancement
1680 Description: lost problems with lov objid file
1681 Details : Fixes some scability and access to not inited memory problems
1682 in work with lov objdid file.
1687 Description: lfs find does not continue on file error
1688 Details : Continue other files processing when a file/dir is absent.
1692 Description: Inconsistent usage of lustre_pack_reply()
1693 Details : Standardize the usage of lustre_pack_reply() such that it
1694 always generate a CERROR on failure.
1697 Frequency : occasional
1699 Description: MDS hang or stay in waiting lock
1700 Details : If client receive lock with CBPENDING flag ldlm need send lock
1701 cancel as separate rpc, to avoid situation when cancel request
1702 can't processed due all i/o threads stay in wait lock.
1706 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1710 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1714 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1718 Description: Update to RHEL4 latest kernel.
1720 Severity : enhancement
1722 Description: Build SLES10 patchless client fails
1723 Details : The configure was broken by run ./configure with
1724 --with-linux-obj=.... argument for patchless client. When the
1725 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1726 can't search header adequately. Use absolute path such as
1727 -I($LINUX)/include instead.
1731 Description: interrupt oig_wait produce painc on resend.
1732 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1733 private set, and this produce situation when rq_ptlrpcd_data not
1734 copyed to new allocated request and triggered LBUG on assert
1735 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1736 ptlrpcd set if request is changed and can be safety changed to use
1741 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1742 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1743 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1744 these macros and replace them with their difinitions instead.
1748 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1749 Details : Update which_patch & target file for SLES10 latest kernel.
1752 Description: Remove client patches from SLES 10 kernel.
1753 Details : This causes SLES 10 clients to behave as patchless clients
1754 even on a Lustre-patched (server) kernel.
1756 Severity : enhancement
1758 Description: self-adjustable client's lru lists
1759 Details : use adaptive algorithm for managing client cached locks lru
1760 lists according to current server load, other client's work
1761 pattern, memory activities, etc. Both, server and client
1762 side namespaces provide number of proc tunables for controlling
1765 Severity : enhancement
1767 Description: light-weight GSS support
1768 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1769 and reduce performance overhead.
1771 Severity : enhancement
1773 Description: Linux keyring support
1774 Details : Support using service of Linux keyring for Lustre GSS internal
1775 context refresh/cache mechanism.
1779 Description: Fix errors in lfs documentation
1780 Details : Fixes man pages
1784 Description: don't use GFP_* in generic Lustre code.
1785 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1789 Description: obdclass is limited by single OBD_ALLOC(idarray)
1790 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1794 Description: warnings with build patchless client with vanila 2.6.19 and up
1795 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1796 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1800 Frequency : Only for SLES
1801 Description: sanity_quota fail test_1
1802 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1803 file, which makes TSTID[2] inunique.
1807 Frequency : Only for SLES9
1808 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1809 Details : Update target/ChangeLog/which_patch .
1813 Frequency : Only for test_10 in sanity-quota.sh
1814 Description: a bug in quota test code
1815 Details : checking a test flag in a wrong place
1819 Frequency : Only for test_13 in sanity-quota.sh
1820 Description: a bug in quota test script
1821 Details : shouldn't check overfull allocated quota
1825 Description: jbd statistics
1826 Details : Port older jbd statistics patch for sles10
1828 Severity : enhancement
1830 Description: Add printing inode info into message about error in writepage.
1835 Description: reply_lock_interpret crash due to race with it and lock cancel.
1836 Details : Do not replay locks that are being cancelled. Do not reference
1837 locks by their address during replay, just by their handle.
1841 Frequency : When flocks are used.
1842 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1843 Details : Flock locks might destroy just granted lock if it could be merged
1844 with another existing flock, this is done in completion handler,
1845 so teach ldlm_cli_enquque_fini that this is a valid case for
1851 Description: Oops in read and write path when failing to allocate lock.
1852 Details : Check if lock allocation failed and return error back.
1856 Description: lstripe command fails for valid OST index
1857 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1858 instead of lov->desc.ld_active_tgt_count.
1862 Description: sanity.sh failed test 103
1863 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1868 Description: df doesn't work properly if diskfs blocksize != 4K
1869 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1873 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1874 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1876 Severity : enhancement
1878 Description: Adaptive timeouts
1879 Details : RPC timeouts adapt to changing server load and network
1880 conditions to reduce resend attempts and improve recovery time.
1884 Description: llapi_file_create() does not allow some changes
1885 Details : add llapi_file_open() that allows specifying the mode and
1886 open flags, and also returns an open file handle.
1890 Description: merge and cleanup kernel patches.
1891 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1895 Description: Add journal checksum support.(Kernel part)
1896 Details : The journal checksum feature adds two new flags i.e
1897 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1898 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1899 indicates that the commit block contains the checksum for
1900 the blocks described by the descriptor blocks. Now commit
1901 record can be sent to disk without waiting for descriptor
1902 blocks to be written to disk. This behavior is controlled
1903 using JBD2_FEATURE_ASYNC_COMMIT flag.
1907 Description: OSS needs mutliple precreate threads
1908 Details : Add ability to start more than one create thread per OSS.
1912 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1913 Details : execute lfs setstripe on client
1917 Description: mds_obd_create error creating tmp object
1918 Details : When the user sets quota on root, llog will be affected and can't
1919 create files and write files.
1922 Frequency : Always on ia64 patchless client, and possibly others.
1924 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1925 Details : This allows the patchless client to be loaded on architectures
1926 without this export.
1930 Description: RedHat Update kernel for RHEL5
1931 Details : Add the kernel config file for RHEL5.
1935 Description: RedHat Update kernel for RHEL5
1936 Details : Modify the kernel config file more closer RHEL5.
1940 Description: Build failure against Centos5 (RHEL5)
1941 Details : Define PAGE_SIZE when it isn't present.
1945 Description: test_7 on sanity-quota.sh can't be used on separate machines
1946 Details : add facet to handle it
1950 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1951 Details : with NFS, the anon dentry's parent was set to itself in
1952 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1953 corresponding dentry other than use rec->ur_name.
1955 Severity : enhancement
1957 Description: lfs setstripe enhancement
1958 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1962 Description: enable data checksumming by default
1963 Details : enable checksum by default, allow --disable-checksum
1964 configure option and "-o nochecksum" mount option. Checksums
1965 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1969 Description: lustre support for RHEL5
1970 Details : Add support for RHEL5.
1974 Description: Client eviction due to failover config
1975 Details : after a connection loss, the lustre client should attempt to
1976 reconnect to the last active server first before trying the
1977 other potential connections.
1981 Description: when mds and osts use different quota unit(32bit and 64bit),
1982 quota will be released repeatly.
1983 Details : void sending multiple quota reqs to mds, which will keep the
1984 status between the reqs.
1988 Description: osts not allocated evenly to files
1989 Details : change the condition to increase offset_idx
1992 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1994 Description: Data corruption for OSTs that are formatted larger than 2TB
1996 Details : When generating the bio request for lustre file writes the
1997 sector number would overflow a temporary variable before being
1998 used for the IO. The data reads correctly from Lustre (which
1999 will overflow in a similar manner) but other file data or
2000 filesystem metadata may be corrupted in some cases.
2004 Description: Tune the kernel for good SCSI performance.
2005 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2006 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2011 Description: rewrite ext2-derived code in obdclass/uuid.c
2012 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2013 from scratch preserving functionality.
2017 Description: "lfs find -obd UUID" prints directories
2018 Details : "lfs find -obd UUID" will return all directory names instead
2019 of just file names. It is incorrect because the directories
2020 do not reside on the OSTs.
2024 Description: use i_size_read and i_size_write in 2.6 port
2025 Details : replace inode->i_size access with i_size_read/write()
2028 Frequency : when removing large files
2030 Description: scheduling issue during removal of large Lustre files
2031 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2032 It causes scheduling issues when removing large files (17TB in the
2036 Frequency : only with liblustre clients on XT3
2038 Description: evictions taking too long
2039 Details : allow llrd to evict clients directly on OSTs
2042 Frequency : only on ppc
2044 Description: /proc/fs/lustre/devices broken on ppc
2045 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2046 We should make sure the type of 'index' is loff_t and then cast
2047 explicitly as needed below. Do not assign an explicitly cast
2051 Frequency : only for rhel5
2053 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2054 Details : Modify the target file & which_kernel.
2056 Severity : enhancement
2058 Description: omit set fsid for export NFS
2059 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2060 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2064 Frequency : Always, for kernels after 2.6.16
2065 Description: Fix warning idr_remove called for id=.. which is not allocated.
2066 Details : Last kernels save old s_dev before kill super and not allow
2067 to restore from callback - restore it before call kill_anon_super.
2071 Description: buffer overruns could theoretically occur
2072 Details : llapi_semantic_traverse() modifies the "path" argument by
2073 appending values to the end of the origin string, and a
2074 overrun may occur. Adding buffer overrun check in liblustreapi.
2078 Description: Fix error on 'ls .' at the top of the Lustre mount.
2079 Details : Don't revalidate dentry if it is a root dentry.
2083 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2084 Details : Modify vm-tunables-rhel4.patch.
2088 Description: Kernel config for 2.6.18-vanilla.
2089 Details : Modify targets/2.6-vanilla.target.in.
2090 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2091 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2092 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2093 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2097 Description: improve handling recoverable errors
2098 Details : if request processig with error which can be recoverable on server
2099 request should be resend, otherwise page released from cache and
2104 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2105 Details : Modify target file & which_patch.
2106 A flaw was found in the IA32 system call emulation provided
2107 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2108 value could be stored in the %RAX register, which could trigger an
2109 out-of-bounds system call table access. An untrusted local user
2110 could exploit this flaw to run code in the kernel
2111 (ie a root privilege escalation). (CVE-2007-4573).
2115 Description: O_DIRECT bypasses client statistics.
2116 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2117 read_ahead_stats were updated. copied the stats section from
2118 osc_send_oap_rpc() into async_internal().
2122 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2126 Description: change order of libsysio includes
2127 Details : '#include sysio.h' should always come before '#include xtio.h'
2131 Description: handle "serious error: objid * already exists" more gracefully
2132 Details : If LAST_ID value on disk is smaller than the objects existing in
2133 the O/0/d* directories, it indicates disk corruption and causes an
2134 LBUG(). If the object is 0-length, then we should use the existing
2135 object. This will help to avoid a full fsck in most cases.
2137 Severity : enhancement
2139 Description: adapt the lustre_config script to support the upgrade case
2140 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2141 server targets to 1.6 in parallel.
2145 Description: To avoid grant space > avaible space when the disk is almost
2146 full. Without this patch you might see the error "grant XXXX >
2147 available" or some LBUG about grant, when the disk is almost
2149 Details : In filter_check_grant, for non_grant cache write, we should
2150 check the left space by if (*left > ungranted + bytes), instead
2151 of (*left > ungranted), because only we are sure the left space
2152 is enough for another "bytes", then the ungrant space should be
2153 increase. In client, we should update cl_avail_grant only there
2154 is OBD_MD_FLGRANT in the reply.
2158 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2159 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2160 described in CVE-2007-4573. This problem could allow untrusted
2161 local users to gain root access.
2164 Frequency : when using O_DIRECT and quotas
2166 Description: Incorrect file ownership on O_DIRECT output files
2167 Details : block usage reported by 'lfs quota' does not take into account
2168 files that have been written with O_DIRECT.
2173 Description: touch file failed when fs is not full
2174 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2175 otherwise we can get ENOSP while fs is not full.
2179 Description: parallel lock callbacks
2180 Details : Instead of sending blocking and completion callbacks as separated
2181 requests, adding them to a set and sending in parallel.
2184 Frequency : only for Cray XT3
2185 Bugzilla : 12829/13455
2186 Description: Changing primary group doesn't change the group lustre assigns to
2188 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2189 overridden with the primary group provided by the group upcall,
2190 whereas the supplied fsgid can be trusted if it is in the list of
2191 supplementary groups returned by the group upcall.
2193 Severity : enhancement
2195 Description: Allow masking D_WARNING, D_ERROR messages from console
2196 Details : Console messages can now be disabled via lnet.printk.
2200 Description: User code with malformed file open parameter crashes client node
2201 Details : Before packing join_file req, all the related reference should be
2202 checked carefully in case some malformed flags cause fake
2203 join_file req on client.
2207 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2208 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2209 mode is equal to the granted mode. However, it should also check
2210 that LDLM_FL_CP_REQD is not set.
2214 Description: Heavy nfs access might result in deadlocks
2215 Details : After ELC code landed, it is now improper to enqueue any mds
2216 locks under och_sem, because enqueue might want to decide to
2217 cancel open locks for same inode we are holding och_sem for.
2221 Description: Client eviction while running blogbench
2222 Details : A lot of unlink operations with concurrent I/O can lead to a
2223 deadlock causing evictions. To address the problem, the number of
2224 oustanding OST_DESTROY requests is now throttled to
2225 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2226 callbacks are processed in priority.
2230 Description: enable ACLs on MDS by default
2231 Details : ACLs must be enabled on MDS by default.
2234 Frequency : PPC/PPC64 only
2236 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2237 Details : fix duplicated definitions between asm-ppc64/types.h and
2238 lustre_types.h on PPC.
2241 Frequency : PPC/PPC64 only
2243 Description: asm-ppc/segment.h does not exist
2244 Details : fix compile issue on PPC.
2248 Description: better handle error messages in extents code
2251 Frequency : RHEL4 only
2253 Description: mkfs is very slow on IA64/RHEL4
2254 Details : A performance regression has been discovered in the MPT Fusion
2255 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2256 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2257 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2259 Severity : enhancement
2261 Description: SNMP support enhancement
2262 Details : Adding total number of sampled request for an MDS node in snmp
2265 Severity : enhancement
2267 Description: Optimize ldlm waiting list processing for PR extent locks
2268 Details : When processing waiting list for read extent lock and meeting read
2269 lock that is same or wider to it that is not contended, skip
2270 processing rest of the list and immediatelly return current
2271 status of conflictness, since we are guaranteed there are no
2272 conflicting locks in the rest of the list.
2276 Description: Time out and refuse to reconnect
2277 Details : When the failover node is the primary node, it is possible
2278 to have two identical connections in imp_conn_list. We must
2279 compare not conn's pointers but NIDs, otherwise we can defeat
2280 connection throttling.
2284 Description: port llog fixes from b1_6 into HEAD
2285 Details : Port llog reference couting and some llog cleanups from b1_6
2286 (bug 10800) into HEAD, for protect from panic and access to already
2287 free llog structures.
2291 Description: Detect stride IO mode in read-ahead
2292 Details : When a client does stride read, read-ahead should detect that and
2293 read-ahead pages according to the detected stride pattern.
2297 Description: data checksumming impacts single node performance
2298 Details : add support for several checksum algorithm. Currently, only CRC32
2299 and Adler-32 are supported. The checksum type can be changed on
2300 the fly via /proc/fs/lustre/osc/*/checksum_type.
2304 Description: use adler32 for page checksums
2305 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2310 Description: build for x2 fails
2311 Details : fix compile issue on Cray systems.
2315 Description: Properly match for duplicate locks
2316 Details : Due to different lock order from skiplists code, we need to
2317 traverse entire list for now
2320 Frequency : only on PPC/SLES10
2322 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2323 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2324 be defined. Add a hack in mkfs_lustre.c to work around this compile
2329 Description: LASSERT on MDS when client holding flock lock dies
2330 Details : ldlm pool logic depends on number of granted locks equal to
2331 number of released locks which is not true for flock locks, so
2332 just exclude such locks from consideration.
2336 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2337 Details : Use fsfilt_sync() for both the whole filesystem sync and
2338 individual file sync to eliminate dangerous inode locking
2339 with I_LOCK that can lead to a deadlock.
2343 Description: performance in 1.6.3
2344 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2345 sure that 1MB requests are not fragmented by the block layer.
2347 Severity : enhancement
2349 Description: organize the server-side client stats on per-nid basis
2350 Details : Change the structure of stats under obdfilter and mds to
2358 The "uuid"s file would list the uuids of _active_ exports.
2359 And the clear entry is to clear all stats and stale nids.
2361 Severity : enhancement
2363 Description: eliminate client locks in face of contention
2364 Details : file contention detection and lockless i/o implementation
2365 for contended files.
2369 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2372 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2374 Description: MDS may lose file striping (and hence file data) in some cases
2375 Details : If there are additional extended attributes stored on the MDS,
2376 in particular ACLs, SELinux, or user attributes (if user_xattr
2377 is specified for the client mount options) then there is a risk
2378 of attribute loss. Additionally, the Lustre file striping
2379 needs to be larger than default (e.g. striped over all OSTs),
2380 and an additional attribute must be stored initially in the
2381 inode and then increase in size enough to be moved to the
2382 external attribute block (e.g. ACL growing in size) for file
2387 Description: skiplist implementation simplification
2388 Details : skiplists are used to group compatible locks on granted list
2389 that was implemented as tracking first and last lock of each
2390 lock group the patch changes that to using doubly linked lists
2394 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2395 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2396 because LASSERT on both the data supplied by a client, and the
2397 data on disk is dangerous and incorrect.
2399 Severity : enhancement
2401 Description: Slow truncate/writes to huge files at high offsets.
2402 Details : Directly associate cached pages to lock that protect those pages,
2403 this allows us to quickly find what pages to write and remove
2404 once lock callback is received.
2408 Description: more ldlm soft lockups
2409 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2410 starve other threads from the resource lock for a long time in
2411 case of long waiting queue, so change the debug level from
2412 D_OTHER to the less frequently used D_INFO.
2414 Severity : enhancement
2416 Description: add -gid, -group, -uid, -user options to lfs find
2420 Description: Hung threads in invalidate_inode_pages2_range
2421 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2422 once one is completed. As a result, some RPCs are stuck in the
2423 queue and are never sent.
2427 Description: filter threads hungs on waiting journal commit
2428 Details : Cleanup filter group llog code, then only filter group llog will
2429 be only created in the MDS/OST syncing process.
2433 Description: Procfs and llog threads access destoryed import sometimes.
2434 Details : Sync the import destoryed process with procfs and llog threads by
2435 the import refcount and semaphore.
2437 Severity : enhancement
2439 Description: openlock cache of b1_6 port to HEAD
2444 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2445 Details : If insertion of an extent fails, then discard the inode
2446 preallocation and free data blocks else it can lead to duplicate
2451 Description: don't always update ctime in ext3_xattr_set_handle()
2452 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2453 In some cases the ctime should not be updated, for example for
2454 2.0->1.8 compatibility it is necessary to delete an xattr and it
2455 should not update the ctime.
2459 Bugzilla : 15713/16362
2460 Description: Assertion in iopen_connect_dentry in 1.6.3
2461 Details : looking up an inode via iopen with the wrong generation number can
2462 populate the dcache with a disconneced dentry while the inode
2463 number is in the process of being reallocated. This causes an
2464 assertion failure in iopen since the inode's dentry list contains
2465 both a connected and disconnected dentry.
2469 Description: assertion failure in ldlm_handle2lock()
2470 Details : fix a race between class_handle_unhash() and class_handle2object()
2471 introduced in lustre 1.6.5 by bug 13622.
2476 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2477 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2482 Description: oops in sd_iostats_seq_show()
2483 Details : unloading/reloading the scsi low level driver triggers a kernel
2484 bug when trying to access the sd iostat file.
2489 Description: Kernel panics during QLogic driver reload
2490 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2491 patch, causing memory corruption.
2496 Description: journal_dev option does not work in b1_6
2497 Details : pass mount option during pre-mount.
2499 Severity : enhancement
2501 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2502 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2503 extent information of a file. It can be used to map logical blocks
2504 in a file to physical blocks in the block device.
2508 Description: LDLM soft lockups - improvement
2509 Details : It is be possible to send the lock handle along with each read
2510 or write request because the client is already doing a lock match
2511 itself so there isn't any reason the OST should have to re-do that
2515 Frequency : only on Cray X2
2517 Description: X2 build failures
2518 Details : fix build failures on Cray X2.
2522 Description: xid & resent requests
2523 Details : Initialize RPC XID from clock at startup (randomly if clock is
2526 Severity : enhancement
2528 Description: Add lustre_start utility to start or stop multiple Lustre servers
2533 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2534 Details : In case of memory pressure, list_del() can be called twice on
2535 req->rq_history_list, causing a kernel oops.
2539 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2540 Details : fix stack overflow in the distributed lock manager by defering
2541 export eviction after a failed AST to the elt thread instead of
2542 handling it in the dlm interpret routine.
2546 Description: Convert some comments to new format.
2547 Details : Update documenting comments to match doxygen conventions.
2551 Description: Grammar fixes.
2552 Details : A couple of trivial spelling fixes.
2556 Description: OSD_COUNTERS-mandatory
2557 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2558 to simplify the code.
2562 Description: simplify cmm_device freeing logic.
2563 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2567 Description: Add lockdep support to dt_object_operations locking interface.
2568 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2569 parameter indicating lock ordering. Update mdd code to use new
2574 Description: Introduce failloc constants for lockless IO tests.
2575 Details : Add two new failloc constants to test lockless IO. Only one of
2576 them in implemented---another is checked in yet to be landed
2581 Description: Add lockdep support for inode mutex.
2582 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2583 in the situations where Lustre has to lock more than one inode
2588 Description: Add optional invariants checking support.
2589 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2590 configure switch. This macro is to be used for consistency and
2591 sanity checks that are too expensive to be left in `production'
2596 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2597 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2602 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2603 Details : Use spin_lock_nested() in (the only) situation where more than
2604 one ldlm_lock is locked simultaneously. Also, fix possible
2605 dead-lock in ldlm_lock_change_resource() by enforcing particular
2610 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2611 Details : Instead of specifying each ldlm_lock call-back through separate
2612 parameter, wrap them into struct ldlm_callback_suite.
2616 Description: Kill join_lru obd method and its callers.
2617 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2618 in memory---a requirement implemented through ->o_join_lru() obd
2619 method in HEAD. Remove this method and its users.
2623 Description: Add asynchronous ldlm ENQUEUE completion handler.
2624 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2625 case that stock ldlm_completion_ast() cannot handle as it waits
2626 until lock is granted. Introduce new ldlm_completion_ast_async()
2627 for this. Also comment ldlm_completion_ast().
2631 Description: ldlm_error <-> errno conversion.
2632 Details : Add functions to map (rather arbitrary) between LDLM error codes
2633 and standard errno values. CLIO needs this to prevent LDLM specific
2634 constants from escaping ldlm and osc.
2638 Description: Kill unused ldlm_handle2lock_ns() function.
2639 Details : Kill unused ldlm_handle2lock_ns() function.
2643 Description: Add lu_ref support to ldlm_lock
2644 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2645 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2646 and ldlm_resource. LDLM interface has to be changed, because code that
2647 releases a reference on a lock, has to "know" what reference this is.
2648 In the most frequent case
2650 lock = ldlm_handle2lock(handle);
2652 LDLM_LOCK_PUT(lock);
2654 no changes are required. When any other reference (received _not_ from
2655 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2656 instead of LDLM_LOCK_PUT().
2658 Arguably, changes are pervasive, and interface requires some discipline
2659 for proper use. On the other hand, it was very instrumental in finding
2660 a few leaked lock references.
2664 Description: Add ldlm_lock_addref_try().
2665 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2666 attempts to addref a lock that might be being canceled
2671 Description: Add ldlm_weigh_callback().
2672 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2673 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2674 hard-coded `number of pages' logic.
2678 Description: Add lockdep annotations to llog code.
2679 Details : Use appropriately tagged _nested() locking calls in the places
2680 where llog takes more than one ->lgh_lock lock.
2684 Description: Add loi_kms_set().
2685 Details : Wrap kms updates into a helper function.
2689 Description: Constify instances of struct lsm_operations.
2690 Details : Constify instances of struct lsm_operations.
2694 Description: lu_conf support.
2695 Details : On a server, a file system object is uniquely identified
2696 by a fid, which is sufficient to locate and load all object
2697 state (inode). On a client, on the other hand, more data are
2698 necessary instantiate an object. Change lu_object_find() and
2699 friends to take additional `lu_conf' argument describing object.
2700 Typically this includes layout information.
2704 Description: lu_context fixes.
2705 Details : Introduce new lu_context functions that are needed on the client
2706 side, where some system threads (ptlrpcd) are shared by multiple
2707 modules, and so cannot be stopped during module shutdown.
2711 Description: Add start and stop methods to lu_device_type_operations.
2712 Details : Introduce two new methods in lu_device_type_operations, that are
2713 invoked when first instance of a given type is created and last one
2714 is destroyed respectively. This is need by CLIO.
2718 Description: Add lu_ref support to struct lu_device.
2719 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2720 track leaked references.
2724 Description: Introduce lu_kmem_descr.
2725 Details : lu_kmem_descr and its companion interface allow to create
2726 and destroy a number of kmem caches at once.
2730 Description: Fix lu_object finalization race.
2731 Details : Fix a race between lu_object_find() finding an object and its
2732 concurrent finalization. This race is (most likely) not possible
2733 on the server, but might happen on the client.
2737 Description: Introduce lu_ref interface.
2738 Details : lu_ref is a debugging module allowing to track references to
2739 a given object. It is quite cpu expensive, and has to be
2740 explicitly enabled with --enable-lu_ref. See usage description
2745 Description: Factor lu_site procfs stats into a separate function.
2746 Details : Separate lu_site stats printing code into a separate function
2747 that can be reused on a client.
2751 Description: Constify instances of struct {lu,dt,md}_device_operations.
2752 Details : Constify instances of struct {lu,dt,md}_device_operations.
2756 Description: Introduce struct md_site and move meta-data specific parts of
2757 struct lu_site here.
2758 Details : Move md-specific fields out of struct lu_site into special struct
2759 md_site, so that lu_site can be used on a client.
2763 Description: Kill mdd_lov_destroy().
2764 Details : Remove unused mdd code.
2768 Description: Add st_block checking to multistat.c.
2769 Details : Add st_block checking to multistat.c.
2773 Description: Add lu_ref support to struct obd_device.
2774 Details : Add lu_ref tracking to obd_device.
2778 Description: Kill obd_set_fail_loc().
2779 Details : Remove unused code.
2783 Description: Add special type for ptlrpc_request interpret functions.
2784 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2785 there. Actual usage will be in CLIO.
2789 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2790 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2791 doesn't work with lockdep.
2795 Description: Add rwv.c test program.
2796 Details : New testing program exercising readv(2) and writev(2) (Qian).
2800 Description: Add sendfile.c test program.
2801 Details : New testing program exercising sendfile(2) (Jay).
2805 Description: Ratelimit a message that can be very frequent.
2806 Details : Ratelimit a memory allocation failure message that can
2811 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2812 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2816 Description: Liblustre build fixes.
2817 Details : Liblustre build fixes.
2821 Description: libcfs: add cfs_{need,cond}_resched() interface.
2822 Details : libcfs: add cfs_{need,cond}_resched() definition and
2823 implementations for Linux, NT, and liblustre.
2825 Severity : enhancement
2827 Description: More exported tunables for mballoc
2828 Details : Add support for tunable preallocation window and new tunables for
2829 large/small requests
2833 Description: Detect corruption of block bitmap and checking for preallocations
2834 Details : Checks validity of on-disk block bitmap. Also it does better
2835 checking of number of applied preallocations. When corruption is
2836 found, it turns filesystem readonly to prevent further corruptions.
2840 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2841 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2842 window to 0. If it does hit cache miss, and read-pattern is still
2843 stride-io mode, does not reset the stride window, but also does not
2844 increase the stride window length in this case.
2848 Frequency : only for big-endian servers
2849 Description: Check if system is big-endian while mounting fs with extents feature
2850 Details : Mounting a filesystem with extents feature will fail on big-endian
2851 systems since ext3-based ldiskfs is not supported on big-endian
2852 systems. This can be over-riden with "bigendian_extents" mount option.
2854 Severity : enhancement
2856 Description: The root squash functionality
2857 Details : A security feature, which is to prevent users from being able
2858 to mount lustre on their desktop, run as root, and delete
2859 all of the files in the filesystem. The goal is accomplished by
2860 remapping user id (UID) and group id (GID) of the root user to
2861 a UID and GID specified by the system administartor via Lustre
2862 configuration management server (MGS). The functionality also
2863 allows to specify sets of clients for which the remapping does
2868 Description: Excessive recovery window
2869 Details : With AT enabled, the recovery window can be excessively long (6000+
2870 seconds). To address this problem, we no longer use
2871 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2872 timeout no longer depends on the service time, it is set to
2873 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2874 time via pb_service_time.
2878 Description: Watchdog triggered on MDS failover
2879 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2880 the OSTs know that the MDS "UUID" can be reused for the same export
2881 from a different NID, so we do not need to wait for the export to be
2885 Frequency : rare, only if using MMP with Linux RAID
2887 Description: MMP doesn't work with Linux RAID
2888 Details : While using HA for Lustre servers with Linux RAID, it is possible
2889 that MMP will not detect multiple mounts. To make this work we
2890 need to unplug the device queue in RAID when the MMP block is being
2891 written. Also while reading the MMP block, we should read it from
2892 disk and not the cached one.
2894 Severity : enhancement
2896 Description: open file using fid
2897 Details : A file can be opened using just its fid, like
2898 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2901 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2904 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2905 Details : add proper locking for creation of jbd_4k slab cache
2909 Description: MMP check in ext3_remount() fails without displaying any error
2910 Details : When multiple mount protection fails during remount, proper error
2913 Severity : enhancement
2915 Description: Allow stripe size to be up to 4G-64k
2916 Details : Fix math logic to allow large stripe sizes.
2920 Description: add check for >8TB ldiskfs filesystems
2921 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2922 Don't allow >8TB ldiskfs filesystems to be mounted without
2923 force_over_8tb mount option
2928 Descriptoin: OSS complains "too many blocks" and crashes
2929 Details : In filter_commitrw_write():
2930 1. lquota_chkquota();
2931 2. obdo_from_inode();
2932 3. lquota_pending_commit();
2933 "2." has changed uid and gid in this case, then caused
2934 lquota_chkquota() and lquota_pending_commit() to use
2935 different uid and gid. That is the root of the bug.
2939 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2940 Details : Detect and print out the 'stripe count = -1' case properly.
2944 Description: Out or order replies might be lost on replay
2945 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2946 request with tid smaller than one currently being added, add it
2947 to the start, not end of the list.
2949 --------------------------------------------------------------------------------
2951 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2953 * Support for kernels:
2954 2.6.5-7.283 (SLES 9),
2955 2.6.9-55.EL (RHEL 4),
2956 2.6.16.46-0.14 (SLES 10),
2957 2.6.18.8 vanilla (kernel.org)
2958 * Client support for unpatched kernels:
2959 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2960 2.6.16 - 2.6.22 vanilla (kernel.org)
2961 * Due to recently discovered recovery problems, we do not recommend
2962 using patchless RHEL 4 clients with this or any earlier release.
2963 * Recommended e2fsprogs version: 1.39.cfs8
2964 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2965 * Starting with this release, the ldiskfs backing filesystem required
2966 by Lustre is now in its own package, lustre-ldiskfs. This package
2967 should be installed. It is versioned separately from Lustre and
2968 may be released separately in future.
2972 Description: block reactivating mgc import until all deactivates complete
2973 Details : Fix race when failing back MDT/MGS to itself (testing)
2976 Frequency : at statup only
2978 Description: mds_lov_synchronize race leads to various problems
2979 Details : simultaneous MDT->OST connections at startup can cause the
2980 sync to abort, leaving the OSC in a bad state.
2982 Severity : enhancement
2984 Description: add optional extra BUILD_VERSION info
2985 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2986 to override the lustre version.
2989 Frequency : 2.6.18 servers only
2991 Description: ll_kern_mount() doesn't release the module reference
2992 Details : The ldiskfs module reference count never drops down to 0
2993 because ll_kern_mount() doesn't release the module reference.
2998 Description: server LBUG when using old ost_num_threads parameter
2999 Details : Accept the old ost_num_threads parameter but warn that it
3000 is deprecated, and fix an off-by-one error that caused an LBUG.
3005 Description: Transient SCSI error results in persistent IO issue
3006 Details : iobuf->dr_error is not reinitialized to 0 between two
3010 Frequency : sometimes when underlying device returns I/O errors
3012 Description: OSTs not going read-only during write failures
3013 Details : OSTs are not remounted read-only when the journal commit threads
3014 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3015 instead of the ext3 wrappers.
3019 Description: poor connect scaling with increasing client count
3020 Details : Don't run filter_grant_sanity_check for more than 100 exports
3021 to improve scaling for large numbers of clients.
3024 Frequency : SLES10 only
3026 Description: sanity-quota.sh quotacheck failed: rc = -22
3027 Details : Quotas cannot be enabled on SLES10.
3030 Frequency : liblustre clients only
3032 Description: getdirentries does not give error when run on compute nodes
3033 Details : getdirentries does not fail when the size specified as an argument
3034 is too small to contain at least one entry
3036 Severity : enhancement
3038 Description: Add LNET router traceability for debug purposes
3039 Details : If a checksum failure occurs with a router as part of the
3040 IO path, the NID of the last router that forwarded the bulk data
3041 is printed so it can be identified.
3046 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3047 Details : Due to a race condition, liblustre clients were occasionally
3048 evicted incorrectly.
3050 Severity : enhancement
3052 Description: lfs setstripe use optional parameters instead of postional
3055 Severity : enhancement
3057 Description: Nanosecond timestamp support for ldiskfs
3058 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3059 resolution timestamps. There is not yet support for this at
3060 the Lustre filesystem level.
3063 Frequency : during server recovery
3065 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3066 Details : request with rq_no_resend flag not awake l_wait_event if they get
3070 Frequency : nfs export on patchless client
3072 Description: connectathon hang when test nfs export over patchless client
3073 Details : Disconnected dentry cannot be found with lookup, so we do not need
3074 to unhash it or make it invalid
3077 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3078 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3079 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3084 Description: open req refcounting wrong on reconnect
3085 Details : If reconnect happened between getting open reply from server and
3086 call to mdc_set_replay_data in ll_file_open, we will schedule
3087 replay for unreferenced request that we are about to free.
3088 Subsequent close will crash in variety of ways.
3089 Check that request is still eligible for replay in
3090 mdc_set_replay_data().
3095 Description: disable writes to filesystem when reading health_check file
3096 Details : the default for reading the health_check proc file has changed
3097 to NOT do a journal transaction and write to disk, because this
3098 can cause reads of the /proc file to hang and block HA state
3099 checking on a healthy but otherwise heavily loaded system. It
3100 is possible to return to the previous behaviour during configure
3101 with --enable-health-write.
3103 Severity : enhancement
3105 Description: 64-bit inode version
3106 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3107 to the inode. This will be required for version-based recovery.
3112 Description: MDS fails to start if a duplicate client export is detected
3113 Details : in some rare cases it was possible for a client to connect to
3114 an MDS multiple times. Upon recovery the MDS would detect this
3115 and fail during startup. Handle this more gracefully.
3117 Severity : enhancement
3119 Description: Add -o localflock option to simulate old noflock
3121 Details : This will achieve local-only flock/fcntl locks
3127 Description: log_commit_thread vs filter_destroy race leads to crash
3128 Details : Take import reference before releasing llog record semaphore
3133 Description: Wrong request locking in request set processing
3134 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3135 imp_delayed_list, in this place should be used imp_lock.
3138 Frequency : when reconnection
3140 Description: Grant Leak when osc reconnect to OST
3141 Details : When osc reconnect ost, OST(filter) should check whether it
3142 should grant more space to client by comparing fed_grant and
3143 cl_avail_grant, and return the granted space to client instead
3144 of "new granted" space, because client will call osc_init_grant
3145 to update the client grant space info.
3148 Frequency : when client reconnect to OST
3150 Description: Grant Leak when osc do resend and replay bulk write
3151 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3152 bulk write request, because the grant info while be sync between
3153 OSC and OST when reconnect, and we should ignore the grant info
3154 these of resend/replay write req.
3159 Description: Grant space more than available left space sometimes.
3160 Details : When then OST is about to be full, if two bulk writing from
3161 different clients came to OST. Accord the avaliable space of the
3162 OST, the first req should be permitted, and the second one
3163 should be denied by ENOSPC. But if the seconde arrived before
3164 the first one is committed. The OST might wrongly permit second
3165 writing, which will cause grant space > available space.
3168 Frequency : when client is evicted
3170 Description: Grant might be wrongly erased when osc is evicted by OST
3171 Details : when the import is evicted by server, it will fork another
3172 thread ptlrpc_invalidate_import_thread to invalidate the
3173 import, where the grant will be set to 0. While the original
3174 thread will update the grant it got when connecting. So if
3175 the former happened latter, the grant will be wrongly errased
3176 because of this race.
3181 Description: Checking Stale with correct fid
3182 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3183 to check whether it is stale, because sometimes, we want the
3184 enqueue happened anyway, and op_data.fid2 will not be initialized.
3186 Severity : enhancement
3188 Description: update patchless client
3189 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3192 Frequency : only with 2.4 kernel
3194 Description: random memory corruption
3195 Details : size of struct ll_inode_info is to big for union inode.u and this
3196 can be cause of random memory corruption.
3201 Description: Memory leak in recovery
3202 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3203 It should also check obd_fail before fsfilt_start, otherwise if
3204 fsfilt_start return -EROFS,(failover mds during mds recovery).
3205 then the req will return with repmsg->transno = 0 and rc = EROFS.
3206 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3207 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3208 be freed no matter whether fsfilt_commit success or not.
3211 Frequency : only with huge count clients
3213 Description: Prevents from taking the superblock lock in llap_from_page for
3215 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3216 for llap_from_page call in ll_removepage prevents from taking the
3217 superblock lock for a soon died page.
3222 Description: Not check open intent error before release open handle
3223 Details : in some rare cases, the open intent error is not checked before
3224 release open handle, which may cause
3225 ASSERTION(open_req->rq_transno != 0), because it tries to release
3226 the failed open handle.
3231 Description: Set cat log bitmap only after create log success.
3232 Details : in some rare cases, the cat log bitmap is set too early. and it
3233 should be set only after create log success.
3237 Description: Accessing a block bevice can re-enable I/O when Lustre is
3238 tearing down a device.
3239 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3244 Description: service threads may hog cpus when there are a lot of requests
3245 Details : Insert cond_resched to give other threads a chance to use some CPU
3250 Description: the cat log was not initialized in recovery
3251 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3252 unlink log on mds will not be initialized until mds post
3253 recovery. And also in mds post recovery, the unlink log will
3254 initialization will be done asynchronausly, so there will be race
3255 between add unlink log and unlink log initialization.
3259 Description: brw_stats were being printed incorrectly
3260 Details : brw_stats were being printed as log2 but all of them were not
3261 recorded as log2. Also remove some code duplication arising from
3262 filter_tally_{read,write}.
3266 Frequency : rare, only in recovery.
3267 Description: ASSERTION(req->rq_type != LI_POISON) failed
3268 Details : imp_lock should be held while iterating over imp_sending_list for
3269 prevent destroy request after get timeout in ptlrpc_queue_wait.
3273 Description: replay-single.sh test 52 fails
3274 Details : A lock's skiplist need to be cleanup when it being unlinked
3275 from its resource list.
3279 Description: Short directio read returns full requested size rather than
3281 Details : Direct I/O operations should return actual amount of bytes
3282 transferred rather than requested size.
3284 Severity : enhancement
3286 Description: metadata RPC reduction (e.g. for rm performance)
3287 Details : decrease the amount of synchronous RPC between clients and servers
3288 by canceling conflicing lock before the operation on the client
3289 and packing thier handles into the main operation RPC to server.
3291 Severity : enhancement
3293 Description: Async OSC create to avoid the blocking unnecessarily.
3294 Details : If an OST has no remaining object, system will block on the
3295 creation when it needs to create a new object on this OST. Now,
3296 ways use pre-created objects when available, instead of blocking on
3297 an empty osc while others are not empty. If we must block, we
3298 block for the shortest possible period of time.
3302 Description: improve handling recoverable errors
3303 Details : if request processig with error which can be recoverable on server
3304 request should be resend, otherwise page released from cache and
3307 Severity : enhancement
3309 Description: refine locking for avoid write wrong info into lov_objid file
3310 Details : fix possible races with add new target and write/update data in
3313 --------------------------------------------------------------------------------
3315 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3320 Frequency : on some architectures
3322 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3323 Details : Uninitialized flags sometimes cause configuration commands to
3327 Frequency : patchless clients only
3329 Description: missing __iget() symbol export
3330 Details : The __iget() symbol export is missing. To avoid the need for
3331 this on patchless clients the deathrow inode reaper is turned
3332 off, and we depend on the VM to clean up old inodes. This
3333 dependency was during via the fix for bug 12181.
3335 --------------------------------------------------------------------------------
3337 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3339 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3340 INTEROPERATE with older versions automatically. In many cases a
3341 special upgrade step is needed. Please read the
3342 user documentation before upgrading any part of a live system.
3343 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3344 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3345 * WARNING: Lustre configuration and startup changes are required with
3346 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3350 Severity : enhancement
3352 Description: MountConf
3353 Details : Lustre configuration is now managed via mkfs and mount
3354 commands instead of lmc and lconf. New obd types (MGS, MGC)
3355 are added for dynamic configuration management. See
3356 https://mail.clusterfs.com/wikis/lustre/MountConf for
3359 Severity : enhancement
3361 Description: dynamic OST addition
3362 Details : OSTs can now be added to a live filesystem
3364 Severity : enhancement
3366 Description: startup order invariance
3367 Details : MDTs and OSTs can be started in any order. Clients only
3368 require the MDT to complete startup.
3370 Severity : enhancement
3372 Description: parallel, asynchronous orphan cleanup
3373 Details : orphan cleanup is now performed in separate threads for each
3374 OST, allowing parallel non-blocking operation.
3376 Severity : enhancement
3378 Description: optimized stripe assignment
3379 Details : stripe assignments are now made based on ost space available,
3380 ost previous usage, and OSS previous usage, in order to try
3381 to optimize storage space and networking resources.
3383 Severity : enhancement
3385 Description: Permanently set tunables
3386 Details : All writable /proc/fs/lustre tunables can now be permanently
3387 set on a per-server basis, at mkfs time or on a live system.
3389 Severity : enhancement
3391 Description: Lustre message v2
3392 Details : Add lustre message format v2.
3394 Severity : enhancement
3396 Description: client OST exclusion list
3397 Details : Clients can be started with a list of OSTs that should be
3398 declared "inactive" for known non-responsive OSTs.
3402 Description: SPEC SFS validation failure on NFS v2 over lustre.
3403 Details : Changes the blocksize for regular files to be 2x RPC size,
3404 and not depend on stripe size.
3406 Severity : enhancement
3408 Description: Multiple MD RPCs in flight.
3409 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3410 To avoid overly-overloading MDS, introduce a limit on number of
3411 MDS RPCs in flight for a single client and add /proc controls
3412 to adjust this limit.
3414 Severity : enhancement
3416 Description: client read/write statistics
3417 Details : Add client read/write call usage stats for performance
3418 analysis of user processes.
3419 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3420 file access. extents_stats shows chunk size distribution.
3421 extents_stats_per_process show chunk size distribution per
3424 Severity : enhancement
3426 Description: mds statistics
3427 Details : Add detailed mds operations statistics in
3428 /proc/fs/lustre/mds/*/stats.
3432 Description: Failure of copying files with lustre special EAs.
3433 Details : Client side always return success for setxattr call for lustre
3434 special xattr (currently only "trusted.lov").
3439 Description: Refcount LNET uuids
3440 Details : The global LNET uuid list grew linearly with every startup;
3441 refcount repeated list entries instead of always adding to
3444 Severity : enhancement
3446 Description: Dynamic service threads
3447 Details : Within a small range, start extra service threads
3448 automatically when the request queue builds up.
3451 Frequency : mixed-endian client/server environments
3453 Description: mixed-endian crashes
3454 Details : The new msg_v2 system had some failures in mixed-endian
3457 Severity : enhancement
3459 Description: Easy OST removal
3460 Details : OSTs can be permanently deactivated with e.g. 'lctl
3461 conf_param lustre-OST0001.osc.active=0'
3463 Severity : enhancement
3465 Description: MGS proc entries
3466 Details : Added basic proc entries for the MGS showing what filesystems
3469 Severity : enhancement
3471 Description: provide MGS failover
3472 Details : Added config lock reacquisition after MGS server failover.
3474 Severity : enhancement
3476 Description: add Linux 2.4 support
3477 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3481 Description: a large application tries to do I/O to the same resource and dies
3482 in the middle of it.
3483 Details : Check the req->rq_arrival time after the call to
3484 ost_brw_lock_get(), but before we do anything about
3485 processing it & sending the BULK transfer request. This
3486 should help move old stale pending locks off the queue as
3487 quickly as obd_timeout.
3490 Frequency : when an incorrect nid is specified during startup
3492 Description: ptlrpc connect to non-existant node causes kernel crash
3493 Details : LNET can't be re-entered from an event callback, which
3494 happened when we expire a message after the export has been
3495 cleaned up. Instead, hand the zombie cleanup off to another
3498 Severity : enhancement
3500 Description: plain/inodebits lock performance improvement
3501 Details : Grouping plain/inodebits in granted list by their request modes
3502 and bits policy, thus improving the performance of search through
3506 Frequency : only if OST filesystem is corrupted
3508 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3509 Details : for a short time RPCs with bulk IO are in the replay list,
3510 but replay of bulk IOs is unimplemented. If the OST filesystem
3511 is corrupted due to disk cache incoherency and then replay is
3512 started it is possible to trip an assertion. Avoid putting
3513 committed RPCs into the replay list at all to avoid this issue.
3516 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3518 Description: System hang on startup
3519 Details : This bug allowed the liblustre (e.g. catamount) client to
3520 return to the app before handling all startup RPCs. This
3521 could leave the node unresponsive to lustre network traffic
3522 and manifested as a server ptllnd timeout.
3524 Severity : enhancement
3526 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3527 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3528 boolean to control whether to print peer debug info when a
3529 client's RPC times out.
3532 Frequency : only for kernels with patches from Lustre below 1.4.3
3534 Description: Remove old rdonly API
3535 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3538 Frequency : only for devices with external journals
3540 Description: Set external device read-only also
3541 Details : During a commanded failover stop, we set the disk device
3542 read-only while the server shuts down. We now also set any
3543 external journal device read-only at the same time.
3546 Frequency : when upgrading from 1.4 while trying to change parameters
3548 Description: The wrong (new) MDC name was used when setting parameters for
3549 upgraded MDT's. Also allows changing of OSC (and MDC)
3550 parameters if --writeconf is specified at tunefs upgrade time.
3553 Frequency : when setting specific ost indicies
3555 Description: QOS code breaks on skipped indicies
3556 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3557 created with --index need not be sequential.
3559 Severity : enhancement
3561 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3562 Details : The uninit_groups feature works in conjunction with the kernel
3563 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3564 up the pass1 processing of e2fsck. This is a read-only feature
3565 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3566 mount filesystems that have had this feature enabled.
3568 Severity : enhancement
3570 Description: Improve multi-block allocation algorithm to avoid fragmentation
3571 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3572 allocation locality and avoid filesystem fragmentation.
3574 ------------------------------------------------------------------------------
3576 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3578 * Support for kernels:
3579 2.4.21-47.0.1.EL (RHEL 3)
3580 2.6.5-7.283 (SLES 9)
3581 2.6.9-42.0.10.EL (RHEL 4)
3582 2.6.12.6 vanilla (kernel.org)
3583 2.6.16.27-0.9 (SLES 10)
3584 * Recommended e2fsprogs version: 1.39.cfs5
3586 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3590 Frequency : occasional, depends on client load and configuration
3591 Bugzilla : 12181, 12203
3592 Description: data loss for recently-modified files
3594 Details : In some cases it is possible that recently written or created
3595 files may not be written to disk in a timely manner (this should
3596 normally be within 30s unless client IO load is very high).
3597 The problem appears as zero-length files or files that are a
3598 multiple of 1MB in size after a client crash or client eviction
3599 that are missing data at the end of the file.
3601 This problem is more likely to be hit on clients where files are
3602 repeatedly created and unlinked in the same directory, clients
3603 have a large amount of RAM, have many CPUs, the filesystem has
3604 many OSTs, the clients are rebooted frequently, and/or the files
3605 are not accessed by other nodes after being written.
3607 The presence of the problem can be detected by looking at
3608 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3609 smaller than the second (nr_unused) then dirty files will not
3610 be flushed automatically to disk. "sync; sleep 10" should be
3611 run several times on the node before unmounting it to update
3612 Lustre (this is also safe to run on nodes without this problem).
3614 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3615 that can cause this same problem, so customers using that kernel
3616 also need to update the kernel in addition to Lustre. In order
3617 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3619 It is normal that files written just before a client crash (less
3620 than 30s) may not yet have been flushed to disk, even for local
3624 Frequency : frequent on thin XT3 nodes
3626 Description: UUID collision on thin XT3 Linux nodes
3627 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3628 randomly, since we relied on an insufficiently-seeded PRNG.
3633 Description: OSS hangs after "All ost request buffers busy"
3634 Details : A deadlock between quota and journal operations caused OSS
3635 hangs after printing "All ost request buffers busy."
3638 Frequency : always on liblustre builds
3640 Description: Cleanup compiler warnings on liblustre
3643 Frequency : always on liblustre builds on XT3
3645 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3649 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3650 > 32000 subdirectories
3651 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3652 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3653 subdirectory count crosses 32000. This will aid e2fsck to
3654 correctly handle more than 32000 subdirectories.
3657 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3659 Description: System hang on startup
3660 Details : This bug allowed the liblustre (e.g. catamount) client to
3661 return to the app before handling all startup RPCs. This
3662 could leave the node unresponsive to lustre network traffic
3663 and manifested as a server ptllnd timeout.
3665 Severity : enhancement
3667 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3668 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3669 boolean to control whether to print peer debug info when a
3670 client's RPC times out.
3675 Description: make O_SYNC working on 2.6 kernels
3676 Details : 2.6 kernels use different method for mark pages for write,
3677 so need add a code to lustre for O_SYNC work.
3682 Description: Failure to close file and release space on NFS
3683 Details : Put inode details into lock acquired in ll_intent_file_open.
3684 Use mdc_intent_lock in ll_intent_open to properly
3685 detect all kind of errors unhandled by mdc_enqueue.
3690 Description: proc file read during shutdown sometimes raced obd removal,
3692 Details : Add lock to prevent obd access after proc file removal.
3695 Frequency : Only for files larger than 4GB on 32-bit clients.
3697 Description: improperly doing page alignment of locks
3698 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3699 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3704 Description: RPCs being resent when they shouldn't be
3705 Details : Some RPCs that should not be resent are being resent. This
3706 can cause inconsistencies in the RPC state machine. Do not
3707 resend such requests.
3710 Frequency : rare, only with NFS export
3712 Description: Crash on NFS re-export node
3713 Details : under very unusual load conditions an assertion is hit in
3714 ll_intent_file_open()
3717 Frequency : only if OST filesystem is corrupted
3719 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3720 Details : for a short time RPCs with bulk IO are in the replay list,
3721 but replay of bulk IOs is unimplemented. If the OST filesystem
3722 is corrupted due to disk cache incoherency and then replay is
3723 started it is possible to trip an assertion. Avoid putting
3724 committed RPCs into the replay list at all to avoid this issue.
3729 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3730 Details : Large single O_DIRECT read and write calls can fail to allocate
3731 a sufficiently large buffer to process the request. In case of
3732 allocation failure the allocation is retried with a smaller
3733 buffer and broken into smaller requests.
3735 Severity : enhancement
3737 Description: Add -o localflock option to simulate old noflock behaviour.
3738 Details : This will achieve local-only flock/fcntl locks coherentness.
3743 Description: versioning check is incomplete
3744 Details : Checking the version difference of client vs. server, report
3745 error if the gap is too big.
3750 Description: add support PG_writeback bit
3751 Details : add support for PG_writeback bit for Lustre, for more carefull
3752 work with page cache in 2.6 kernel. This also fix some deadlocks
3753 and remove hack for work O_SYNC with 2.6 kernel.
3755 Severity : enhancement
3757 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3758 Details : The uninit_groups feature works in conjunction with the kernel
3759 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3760 up the pass1 processing of e2fsck. This is a read-only feature
3761 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3762 mount filesystems that have had this feature enabled.
3764 Severity : enhancement
3766 Description: Improve multi-block allocation algorithm to avoid fragmentation
3767 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3768 allocation locality and avoid filesystem fragmentation.
3770 ------------------------------------------------------------------------------
3772 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3774 * Support for kernels:
3775 2.6.9-42.0.3.EL (RHEL 4)
3776 2.6.5-7.276 (SLES 9)
3777 2.4.21-47.0.1.EL (RHEL 3)
3778 2.6.12.6 vanilla (kernel.org)
3779 2.6.16.21-0.8 (SLES10)
3780 * Recommended e2fsprogs version: 1.39.cfs2-0
3782 * The backwards-compatible /proc/sys/portals symlink has been removed
3783 in this release. Before upgrading, please ensure that you change
3784 any configuration scripts or /etc/sysctl.conf files that access
3785 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3786 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3787 in advance of the upgrade on any system running Lustre 1.4.6 or
3788 newer, since /proc/sys/lnet was added in that version.
3789 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3793 Frequency : only when quota is used
3795 Description: avoid scanning export list for quota master
3796 Details : Change the algorithms to avoid scanning export list in order
3797 to improve the efficiency.
3800 Frequency : MDS failover only, very rarely
3802 Description: "went back in time" messages on mds failover
3803 Details : The greatest transno may be lost when the current operation
3804 finishes with an error (transno==0) and the client's last_rcvd
3805 record is over-written. Save the greatest transno in the
3806 mds_last_transno for this case.
3809 Frequency : always for specific kernels and striping counts
3811 Description: client may get "Matching packet too big" without ACL support
3812 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3813 when trying to access files in certain configurations. The
3814 clients should in fact be denied when mounting because they do
3815 not understand ACLs.
3818 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3820 Description: many clients connecting with IO in progress causes connect timeouts
3821 Details : Avoid synchronous journal commits to avoid delays caused by many
3822 clients connecting/disconnecting when bulk IO is in progress.
3823 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3824 OST_IO_PORTAL to avoid delays behind potentially many pending
3828 Frequency : occasionally with multiple writers to a single file
3830 Description: shared writes to file may result in wrong size reported by stat()
3831 Details : Allow growing of kms when extent lock is cancelled
3834 Frequency : always with random mmap IO to multi-striped file
3836 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3837 Details : Only if the hole is at the end of OST object so that kms is too
3838 small. Fix is to increase kms accordingly in ll_nopage.
3841 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3843 Description: writes to a missing object would leak memory on the OST
3844 Details : If there is an inconsistency between the MDS and OST filesystems,
3845 such that the MDS references an object that doesn't exist, writes
3846 to that object will leak memory due to incorrect cleanup in the
3847 error handling path, eventually running out of memory on the OST.
3852 Description: Creating too long symlink causes lustre errors
3853 Details : Check symlink and name lengths before sending requests to MDS.
3856 Frequency : only if flock is enabled (not on by default)
3858 Description: posix locks not released on fd closure on 2.6.9+
3859 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3860 kernels, this caused such locks not to be released on fd close and
3861 then assertions on fs unmount about still used locks.
3864 Frequency : MDS failover only, very rarely
3866 Description: clients may get ASSERTION(granted_lock != NULL)
3867 Details : When request was taking a long time, and a client was resending
3868 a getattr by name lock request. The were multiple lock requests
3869 with the same client lock handle and
3870 mds_getattr_name->fixup_handle_for_resent_request found one of the
3871 lock handles but later failed with ASSERTION(granted_lock != NULL).
3876 Description: handle->h_buffer_credits > 0, assertion failure
3877 Details : h_buffer_credits is zero after truncate, causing assertion
3878 failure. This patch extends the transaction or creates a new
3882 Frequency : NFS re-export or patchless client
3883 Bugzilla : 11179, 10796
3884 Description: Crash on NFS re-export node (__d_move)
3885 Details : We do not want to hash the dentry if we don't have a lock.
3886 But if this dentry is later used in d_move, we'd hit uninitialised
3887 list head d_hash, so we just do this to init d_hash field but
3888 leave dentry unhashed.
3891 Frequency : NFS re-export or patchless client
3893 Description: NFS exports has problem with symbolic link
3894 Details : lustre client didn't properly install dentry when re-exported
3895 to NFS or running patchless client.
3898 Frequency : NFS re-export or patchless client
3900 Description: Various nfs/patchless fixes.
3901 Details : fixes reuse disconected alias for lookup process - this fixes
3902 warning "find_exported_dentry: npd != pd",
3903 fix permission error with open files at nfs.
3904 fix apply umask when do revalidate.
3907 Frequency : occasional
3909 Description: Crash on NFS re-export node
3910 Details : calling clear_page() on the wrong pointer triggered oops in
3911 generic_mapping_read().
3914 Frequency : rarely, using O_DIRECT IO
3916 Description: unaligned directio crashes client with LASSERT
3917 Details : check for unaligned buffers before trying any requests.
3920 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3922 Description: stale data returned from RAID cache
3923 Details : If only a small amount of IO is done to the RAID device before
3924 reading it again it is possible to get stale data from the RAID
3925 cache instead of reading it from disk.
3928 Frequency : always for sles10 kernel
3930 Description: sles10 support
3931 Details : ll_follow_link: compile fixes and using of nd_set_link
3932 under newer kernels.
3935 Frequency : depends on arch, kernel and compiler version, always on sles10
3938 Description: recursive or deep enough symlinks cause stack overflow
3939 Details : getting rid of large stack-allocated variable in
3943 Frequency : depends on hardware
3945 Description: lustre write performance loss in the SLES10 kernel
3946 Details : the performance loss is caused by using of write barriers in the
3947 ext3 code. The SLES10 kernel turns barrier support on by
3948 default. The fix is to undo that change for ldiskfs.
3950 ------------------------------------------------------------------------------
3952 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3954 * Support for kernels:
3955 2.6.9-42.0.3EL (RHEL 4)
3956 2.6.5-7.276 (SLES 9)
3957 2.4.21-47.0.1.EL (RHEL 3)
3958 2.6.12.6 vanilla (kernel.org)
3962 Frequency : quota enabled and large files being deleted
3964 Description: releasing more than 4GB of quota at once hangs OST
3965 Details : If a user deletes more than 4GB of files on a single OST it
3966 will cause the OST to spin in an infinite loop. Release
3967 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3972 Description: statfs data retrieved from /proc may be stale or zero
3973 Details : When reading per-device statfs data from /proc, in the
3974 {kbytes,files}_{total,free,avail} files, it may appear
3975 as zero or be out of date.
3978 Frequency : systems with MD RAID1 external journal devices
3980 Description: lconf's call to blkid is confused by RAID1 journal devices
3981 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3982 of returning all block devices that match the journal UUID.
3985 Frequency : always, for aggregate stripe size over 4GB
3987 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3988 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3989 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3992 Frequency : always if "lfs find" used on a local file/directory
3994 Description: "lfs find" segfaults if used on a local file/directory
3995 Details : The case where a directory component was not specified wasn't
3996 handled correctly. Handle this properly.
3999 Frequency : always on ppc64
4001 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4002 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4003 perform endianness conversion.
4006 Frequency : rarely (truncate to non-zero file size after write under load)
4007 Bugzilla : 10730, 10687
4008 Description: Files padded with zeros to next 4K multiple
4009 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4010 it is possible that files that are truncated to a non-zero size
4011 immediately after being written are filled with zero bytes beyond
4012 the truncated size. No file data is lost.
4014 Severity : enhancement
4016 Description: Allow recovery/failover for liblustre clients.
4017 Details : liblustre clients were unaware of failover configurations until
4020 Severity : enhancement
4022 Description: user file locks should fail when not mounting with flock option
4023 Details : Set up an error-returning stub in ll_file_operations.lock field
4024 to prevent incorrect behaviour when client is mounted without
4025 flock option. Also, set up properly f_op->flock field for
4029 Frequency : always on ia64
4031 Description: "lfs df" loops on printing out MDS statfs information
4032 Details : The obd_ioctl_data was not initialized and in some systems
4033 this caused a failure during the ioctl that did not return
4034 an error. Initialize the struct and return an error on failure.
4037 Frequency : SLES 9 only
4039 Description: Error of copying files with lustre special EAs as root
4040 Details : Client side always return success for setxattr call for lustre
4041 special xattr (currently only "trusted.lov").
4044 Frequency : rarely on clusters with both ia64+i386 clients
4046 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4047 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4048 clients are doing concurrent writes to the same file it is
4049 possible that the ia64 clients may LASSERT because the OST
4050 extent locks are not PAGE_SIZE aligned. Ensure that grown
4051 locks are always aligned on the request boundary.
4054 Frequency : specific use, occasional
4056 Description: Overwriting in use executable truncates on-disk binary image
4057 Details : If one node attempts to overwrite an executable in use by
4058 another node, we now correctly return ETXTBSY instead of
4059 truncating the file.
4061 Severity : enhancement
4063 Description: Async OSC create to avoid the blocking unnecessarily.
4064 Details : If an OST has no remaining object, system will block on the creating
4065 when need to create a new object on this OST. Now, ways use
4066 pre-created objects when available, instead of blocking on an
4067 empty osc while others are not empty. If we must block, we block
4068 for the shortest possible period of time.
4073 Description: chmod on Lustre root is propagated to other clients
4074 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4075 invalid by the follow_mount time.
4078 Frequency : liblustre clients only
4080 Description: Race in 'instant cancel' lock handling could lead to such locks
4081 never to be granted in case of SMP MDS
4082 Details : Do not destroy not yet granted but cbpending locks in
4086 Frequency : replay/resend of open
4088 Description: non null lock assetion failure in mds_intent_policy
4089 Details : Trying to replay/resend lockless open requests resulted in
4090 mds_open() returning 0 with no lock. Now it sets a flag if
4091 a lock is going to be returned.
4093 Severity : enhancement
4095 Description: Checksum enhancements
4096 Details : New checksum enhancements allow for resending RPCs that failed
4099 Severity : enhancement
4101 Description: Tunables on number of dirty pages in cacche
4102 Details : Allow to set limit on number of dirty pages cached.
4107 Description: client crash on unmount - lock still has references
4108 Details : In some error handling cases it was possible to leak a lock
4109 reference on a client while accessing a file. This was not
4110 harmful to the client during operation, but would cause the
4111 client to crash when the filesystem is unmounted.
4114 Frequency : specific case, rare
4116 Description: ETXTBSY on mds though file not in use by client
4117 Details : ETXTBSY is no longer incorrectly returned when attempting to
4118 chmod or chown a directory that the user previously tried to
4119 execute or a currently-executing binary.
4122 Frequency : extremely rare except on liblustre-based clients
4124 Description: Lustre space not freed when files are deleted
4125 Details : Clean up open-unlinked files after client eviction. Previously
4126 the unlink was skipped and the files remained as orphans.
4131 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4132 Details : In some cases it was possible to send a blocking callback to
4133 a client doing a glimpse, even though that client didn't get
4134 a lock granted. When the glimpse lock is cancelled on the OST
4135 the freed lock is left on the waiting list and corrupted the list.
4138 Frequency : all core dumps
4140 Description: Broke core dumps to lustre
4141 Details : Negative dentry may be unhashed if parent does not have UPDATE
4142 lock, but some callers, e.g. do_coredump, expect dentry to be
4143 hashed after successful create, hash it in ll_create_it.
4145 ------------------------------------------------------------------------------
4147 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4149 * Support for kernels:
4150 2.6.9-42.0.2.EL (RHEL 4)
4151 2.6.5-7.276 (SLES 9)
4152 2.4.21-40.EL (RHEL 3)
4153 2.6.12.6 vanilla (kernel.org)
4157 Frequency : always on RHEL 3
4159 Description: Number of open files grows over time
4160 Details : The number of open files grows over time, whether or not
4161 Lustre is started. This was due to a filp leak introduced
4162 by one of our kernel patches.
4164 ------------------------------------------------------------------------------
4166 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4168 * Support for kernels:
4169 2.6.9-42.EL (RHEL 4)
4170 2.6.5-7.276 (SLES 9)
4171 2.4.21-40.EL (RHEL 3)
4172 2.6.12.6 vanilla (kernel.org)
4177 Bugzilla : 5719, 9635, 9792, 9684
4178 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4179 Details : If a server is under heavy load and cannot reply to new
4180 connection requests before the client resends the (re)connect,
4181 the connection handling code can behave badly if two service
4182 threads are concurrently handing separate (re)connections from
4183 the same client. Add better locking to the connection handling
4184 code, and ensure that only a single connection will be processed
4185 for a given client UUID, even if the lock is dropped.
4187 Severity : enhancement
4189 Description: add TCP zero-copy support to kernel
4190 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4191 sends if the hardware supports scatter-gather and checksumming.
4192 This allows socklnd to do client-write and server-read more
4193 efficiently and reduce CPU utilization from skbuf copying.
4196 Frequency : only if NFS exporting from client
4198 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4199 Details : If setting attributes on a file created under NFS that had
4200 never been opened it would be possible to oops the client
4201 if the file had no objects.
4204 Frequency : always for liblustre
4206 Description: liblustre client does MDS+OSTs setattr RPC for each write
4207 Details : When doing a write from a liblustre client, the client
4208 incorrectly issued an RPC to the MDS and each OST the file was
4209 striped over in order to update the timestamps. When writing
4210 with small chunks and many clients this could overwhelm the MDS
4211 with RPCs. In all cases it would slow down the write because
4212 these RPCs are unnecessary.
4214 Severity : enhancement
4216 Description: allow number of MDS service threads to be changed at module load
4217 Details : It is now possible to change the number of MDS service threads
4218 running. Adding "options mds mds_num_threads={N}" to the MDS's
4219 /etc/modprobe.conf will set the number of threads for the next
4220 time Lustre is restarted (assuming the "mds" module is also
4221 reloaded at that time). The default number of threads will
4222 stay the same, 32 for most systems.
4227 Description: OST crash if filesystem is unformatted or corrupt
4228 Details : If an OST is started on a device that has never been formatted
4229 or if the filesystem is corrupt and cannot even mount then the
4230 error handling cleanup routines would dereference a NULL pointer.
4235 Description: NULL pointer deref in llap_from_page.
4236 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4237 incorrect) mapping if the page was truncated/reclaimed while it was
4238 searched for. Check for this condition and skip such pages when
4239 doing readahead. Introduce extra check to llap_from_page() to
4240 verify page->mapping->host is non-NULL (so page is not anonymous).
4243 Frequency : Sometimes when using sys_sendfile
4245 Description: "page not covered by a lock" warnings from ll_readpage
4246 Details : sendfile called ll_readpage without right page locks present.
4247 Now we introduced ll_file_sendfile that does necessary locking
4248 around call to generic_file_sendfile() much like we do in
4252 Frequency : with certain MDS communication failures at client mount time
4254 Description: NULL pointer deref after failed client mount
4255 Details : a client connection request may delayed by the network layer
4256 and not be sent until after the PTLRPC layer has timed out the
4257 request. If the client fails the mount immediately it will try
4258 to clean up before the network times out the request. Add a
4259 reference from the request import to the obd device and delay
4260 the cleanup until the network drops the request.
4263 Frequency : occasionally during client (re)connect
4265 Description: assertion failure during client (re)connect
4266 Details : processing a client connection request may be delayed by the
4267 client or server longer than the client connect timeout. This
4268 causes the client to resend the connection request. If the
4269 original connection request is replied in this interval, the
4270 client may trip an assertion failure in ptlrpc_connect_interpret()
4271 which thought it would be the only running connect process.
4274 Frequency : only with obd_echo servers and clients that are rebooted
4276 Description: kernel BUG accessing uninitialized data structure
4277 Details : When running an obd_echo server it did not start the ping_evictor
4278 thread, and when a client was evicted an uninitialized data
4279 structure was accessed. Start the ping_evictor in the RPC
4280 service startup instead of the OBD startup.
4282 Severity : enhancement
4283 Bugzilla : 10193 (patchless)
4284 Description: Remove dependency on various unexported kernel interfaces.
4285 Details : No longer need reparent_to_init, exit_mm, exit_files,
4286 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4289 Frequency : rare (only users of deprecated and unsupported LDAP config)
4291 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4292 Details : LDAP apparently contains 'lustreName' attributes instead of
4293 'name'. A simple remapping of the name is sufficient.
4296 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4298 Description: waiting_locks_callback trips kernel BUG if client is evicted
4299 Details : Running with the dump_on_timeout debug flag turned on makes
4300 it possible that the waiting_locks_callback() can try to dump
4301 the Lustre kernel debug logs from an interrupt handler. Defer
4302 this log dumping to the expired_lock_main() thread.
4304 Severity : enhancement
4306 Description: Support NFS exporting on 2.6 kernels.
4307 Details : Implement non-rawops metadata methods for NFS server to use without
4308 changing NFS server code.
4311 Frequency : very rare (synthetic metadata workload only)
4313 Description: two racing renames might cause an MDS thread to deadlock
4314 Details : Running the "racer" program may cause one MDS thread to rename
4315 a file from being the source of a rename to being the target of
4316 a rename at exactly the same time that another thread is doing
4317 so, and the second thread has already enqueued these locks after
4318 doing a lookup of the target and is trying to relock them in
4319 order. Ensure that we don't try to re-lock the same resource.
4322 Frequency : only very large systems with liblustre clients
4324 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4325 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4326 all OSTs in parallel. This allows the eviction of stale liblustre
4327 clients to proceed much faster than if they were done in series,
4328 and also offers similar improvements for other set_info RPCs.
4333 Description: excessive CPU usage during initial read phase on client
4334 Details : During the initial read phase on a client, it would agressively
4335 retry readahead on the file, consuming too much CPU and impacting
4336 performance (since 1.4.5.8). Improve the readahead algorithm
4337 to avoid this, and also improve some other common cases (read
4338 of small files in particular, where "small" is files smaller than
4339 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4344 Description: MDS crash when receiving packet with unknown intent.
4345 Details : Do not LBUG in unknown intent case, just return -EFAULT
4347 Severity : enhancement
4348 Bugzilla : 9293, 9385
4349 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4350 Details : Do not serialize getattr (non-intent version) and statfs.
4353 Frequency : occasional, when OST network is overloaded/intermittent
4355 Description: client evicted by OST after bulk IO timeout
4356 Details : If a client sends a bulk IO request (read or write) the OST
4357 may evict the client if it is unresposive to its data GET/PUT
4358 request. This is incorrect if the network is overloaded (takes
4359 too long to transfer the RPC data) or dropped the OST GET/PUT
4360 request. There is no need to evict the client at all, since
4361 the pinger and/or lock callbacks will handle this, and the
4362 client can restart the bulk request.
4365 Frequency : Always when mmapping file with no objects
4367 Description: client crashes when mmapping file with no objects
4368 Details : Check that we actually have objects in a file before doing any
4369 operations on objects in ll_vm_open, ll_vm_close and
4375 Description: Request leak when working with deleted CWD
4376 Details : Introduce advanced request refcount tracking for requests
4377 referenced from lustre intent.
4379 Severity : Enhancement
4381 Description: Cache open file handles on client.
4382 Details : MDS now will return special lock along with openhandle, if
4383 requested and client is allowed to hold openhandle, even if unused,
4384 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4385 closing files for every read/write openration.
4387 Severity : Enhancement
4389 Description: Cache open negative dentries on client when possible.
4390 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4391 negative dentries on lock revocation.
4396 Description: Remounting a client read-only wasn't possible with a zconf mount
4397 Details : It wasn't possible to remount a client read-only with llmount.
4399 Severity : enhancement
4400 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4401 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4402 MPICH is included to add Lustre-specific ADIO interfaces.
4403 This is based closely on the UFS ADIO layer and only differs
4404 in file creation, in order to allow the OST striping to be set.
4405 This is user-contributed code and not supported by CFS.
4410 Description: extended inode attributes (immutable, append-only) work improperly
4411 when 2.4 and 2.6 kernels are used on client/server or vice versa
4412 Details : Introduce kernel-independent values for these flags.
4414 Severity : enhancement
4417 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4418 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4419 granularity. Now, Fractional MB granularity is be supported,
4420 this is very useful for low memory system.
4422 Severity : enhancement
4424 Description: Getattr by fid
4425 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4426 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4427 have locks and avoids getattr rpc requests for it on every stat.
4430 Frequency : Always, for filesystems larger than 2TB
4432 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4433 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4434 filesystems larger than 2TB to be mounted with mballoc enabled.
4437 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4439 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4440 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4441 filesystems larger than 2TB is mounted then the kernel will
4442 silently corrupt the start of the filesystem. CONFIG_LBD is
4443 enabled for all CFS-supported kernels, but the possibility of
4444 this happening with a modified kernel config exists.
4446 Severity : enhancement
4448 Description: add client O_DIRECT support for 2.6 kernels
4449 Details : It is now possible to do O_DIRECT reads and writes to files
4450 in the Lustre client mountpoint on 2.6 kernel clients.
4452 Severity : enhancement
4454 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4455 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4456 parallel, not waiting for response from every OST before sending
4457 a rpc to the next OST.
4462 Description: setattr vs write race when updating file timestamps
4463 Details : Client processes that update a file timestamp into the past
4464 right after writing to the file (e.g. tar) it is possible that
4465 the updated file modification time can be reset to the current
4466 time due to a race between processing the setattr and write RPC.
4468 Severity : enhancement
4470 Description: Bring 'lfs find' closer in line with regular Linux find.
4471 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4472 -print0 options and obtains all the needed info through the lustre
4475 Severity : enhancement
4477 Description: support up to 1024 configured devices on one node
4478 Details : change obd_dev array from statically allocated to dynamically
4479 allocated structs as they are first used to reduce memory usage
4484 Description: Flush dirty partially truncated pages during truncate
4485 Details : Immediatelly flush partially truncated pages in filter_setattr,
4486 this way we completely avoid having any pages in page cache on OST
4487 and can retire ugly workarounds during writes to flush such pages.
4492 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4493 Details : protect inode from truncation within vfs_unlink() context
4494 just take a reference before calling vfs_unlink() and release it
4495 when parent's i_sem is free.
4500 Description: last_id value checked outside lock on OST caused LASSERT failure
4501 Details : If there were multiple MDS->OST object precreate requests in
4502 flight, it was possible that the OST's last object id was checked
4503 outside a lock and incorrectly tripped an assertion. Move checks
4504 inside locks, and discard old precreate requests.
4507 Frequency : always, if extents are used on OSTs
4509 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4510 Details : OSTs using the extents format would not zero the high 16 bits of
4511 the index physical block number. This is not a problem for any
4512 OST filesystems smaller than 16TB, and no kernels support ext3
4513 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4514 new/modified files) and can be fixed for existing filesystems
4515 with e2fsprogs-1.39-cfs1.
4520 Description: import connection selection may be incorrect if timer wraps
4521 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4522 import connections to be ignored if the 32-bit jiffies counter
4523 wraps. Use a 64-bit jiffies counter.
4526 Frequency : very large clusters immediately after boot
4528 Description: LNET request buffers exhausted under heavy short-term load
4529 Details : If a large number of client requests are generated on a service
4530 that has previously never seen so many requests it is possible
4531 that the request buffer growth cannot keep up with the spike in
4532 demand. Instead of dropping incoming requests, they are held in
4533 the LND until the RPC service can accept more requests.
4536 Frequency : Sometimes during replay
4538 Description: Assertion failure in ll_local_open after replay.
4539 Details : If replay happened on an open request reply before we were able
4540 to set replay handler, reply will become not swabbed tripping the
4541 assertion in ll_local_open. Now we set the handler right after
4542 recognising of open request
4545 Frequency : very rare
4547 Description: kernel reports "badness in vsnprintf"
4548 Details : Reading from the "recovery_status" /proc file in small chunks
4549 may cause a negative length in lprocfs_obd_rd_recovery_status()
4550 call to vsnprintf() (which is otherwise harmless). Exit early
4551 if there is no more space in the output buffer.
4553 Severity : enhancement
4555 Description: clear OBD RPC statistics by writing to them
4556 Details : It is now possible to clear the OBD RPC statistics by writing
4557 to the "stats" file.
4562 Description: Client mtime is not the same on different clients after utimes
4563 Details : In some cases, the client was using the utimes() syscall on
4564 a file cached on another node. The clients now validate the
4565 ctime from the MDS + OSTs to determine which one is right.
4570 Description: Inability to activate failout mode
4571 Details : lconf script incorrectly assumed that in pythong string's numeric
4572 value is used in comparisons.
4575 Frequency : always with multiple stripes per file
4577 Description: Inefficient object allocation for mutli-stripe files
4578 Details : When selecting which OSTs to stripe files over, for files with
4579 a stripe count that divides evenly into the number of OSTs,
4580 the MDS is always picking the same starting OST for each file.
4581 Return the OST selection heuristic to the original design.
4586 Description: mount failures may take full timeout to return an error
4587 Details : Under some heavy load conditions it is possible that a
4588 failed mount can wait for the full obd_timeout interval,
4589 possibly several minutes, before reporting an error.
4590 Instead return an error as soon as the status is known.
4592 Frequency : quota enabled and large files being deleted
4594 Description: releasing more than 4GB of quota at once hangs OST
4595 Details : If a user deletes more than 4GB of files on a single OST it
4596 will cause the OST to spin in an infinite loop. Release
4597 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4602 Description: statfs data retrieved from /proc may be stale or zero
4603 Details : When reading per-device statfs data from /proc, in the
4604 {kbytes,files}_{total,free,avail} files, it may appear
4605 as zero or be out of date.
4608 Frequency : systems with MD RAID1 external journal devices
4610 Description: lconf's call to blkid is confused by RAID1 journal devices
4611 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4612 of returning all block devices that match the journal UUID.
4615 Frequency : always, for aggregate stripe size over 4GB
4617 Description: assertion fails when trying to use 4GB stripe size
4618 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4619 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4622 Frequency : always on ppc64
4624 Description: the first write on an ext3 filesystem with mballoc got stuck
4625 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4626 perform endianness conversion.
4628 ------------------------------------------------------------------------------
4630 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4632 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4633 INTEROPERATE with older versions automatically. Please read the
4634 user documentation before upgrading any part of a live system.
4635 * WARNING: Lustre networking configuration changes are required with
4636 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4639 * Support for kernels:
4640 2.6.9-22.0.2.EL (RHEL 4)
4641 2.6.5-7.244 (SLES 9)
4642 2.6.12.6 vanilla (kernel.org)
4645 Severity : enhancement
4646 Bugzilla : 7981/8208
4647 Description: Introduced Lustre Networking (LNET)
4648 Details : LNET is new networking infrastructure for Lustre, it includes
4649 a reorganized network configuration mode (see the user
4650 documentation for full details) as well as support for routing
4651 between different network fabrics. Lustre Networking Devices
4652 (LNDs) for the supported network fabrics have also been
4653 created for this new infrastructure.
4655 Severity : enhancement
4656 Description: Introduced Access control lists
4657 Details : clients can set ACLs on files and directories in order to have
4658 more fine-grained permissions than the standard Unix UGO+RWX.
4659 The MDS must be started with the "-o acl" mount option.
4661 Severity : enhancement
4662 Description: Introduced filesystem quotas
4663 Details : Administrators may now establish per-user quotas on the
4666 Severity : enhancement
4668 Description: Configuration change for the XT3
4669 The PTLLND is now used to run Lustre over Portals on the XT3
4670 The configure option(s) --with-cray-portals are no longer used.
4671 Rather --with-portals=<path-to-portals-includes> is used to
4672 enable building on the XT3. In addition to enable XT3 specific
4673 features the option --enable-cray-xt3 must be used.
4678 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4679 Details : A race between allocating a new llcd and re-getting the llcd_lock
4680 allowed another thread to grab newly-allocated llcd.
4682 Severity : enhancement
4684 Description: 2.6 OST async journal commit and locking fix to improve performance
4685 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4686 2.6 kernels are now async as they already were in 2.4 kernels so
4687 that they can commit concurrently with the network bulk transfer.
4688 For block-allocated files the filter allocation semaphore is held
4689 to avoid filesystem fragmentation during allocation. BKL lock
4690 removed for 2.6 xattr operations where it is no longer needed.
4695 Description: lconf incorrectly determined whether two IP networks could talk
4696 Details : In some more complicated routing and multiple-network
4697 configurations, lconf will avoid trying to make a network
4698 connection to a disjoint part of the IP space. It was doing the
4699 math incorrectly for one set of cases.
4704 Description: Fix for potential infinite loop processing records in an llog.
4705 Details : If an llog record is corrupted/zeroed, it is possible to loop
4706 forever in llog_process(). Validate the llog record length
4707 and skip the remainder of the block on error.
4710 Frequency : occasional (liblustre only)
4712 Description: liblustre could not open files whose last component is a symlink
4713 Details : sysio_path_walk() would incorrectly pass the open intent to
4714 intermediate path components.
4717 Frequency : rare (liblustre only with non-standard tuning)
4718 Bugzilla : 7201 (7350)
4719 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4720 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4721 after it has been released, only do this for referenced locks
4723 Severity : enhancement
4725 Description: specifying an (invalid) directory default stripe_size of -1
4726 would reset the directory default striping
4727 Details : stripe_size -1 was used internally to signal directory stripe
4728 removal, now use "all default" to signal dir stripe removal
4729 as a directory striping of "all default" is not useful
4732 Frequency : common for large clusters running liblustre clients
4734 Description: doing an ls when liblustre clients are running is slow
4735 Details : sending a glimpse AST to a liblustre client waits for every AST
4736 to time out, as liblustre clients will not respond. Since they
4737 cannot cache data we refresh the OST lock LVB from disk instead.
4739 Severity : enhancement
4741 Description: doing an ls at the same time as file IO can be slow
4742 Details : enqueue and other "small" requests can be blocked behind many
4743 large IO requests. Create a new OST IO portal for non-IO
4744 requests so they can be processed faster.
4747 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4749 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4750 created file with mode 000
4751 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4752 allow an ftruncate() as a non-root user to a file with mode 000.
4753 Lustre now respects this flag to disable mode checks when
4754 truncating a file owned by the user
4757 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4760 Description: Revoking locks from clients that went dead or catatonic might take
4762 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4763 cancellation of such locks instant on servers without waiting for
4764 any reply from clients. Clients drops these locks when cancel
4765 notification from server is received without replying.
4768 Frequency : liblustre-only, when liblustre client dies or becomes busy
4770 Description: Doing ls on Linux clients can take a long time with active
4772 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4773 granting such locks to it in the first place if possible. Locks
4774 are taken by proxy on the OST during the read or write and
4775 dropped immediately afterward. Add connect flags handling, do
4776 not grant locks to liblustre clients for glimpse ASTs.
4778 Severity : enhancement
4780 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4781 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4782 sequential reads and behaves badly with applications doing random
4783 reads. Improve it by reading ahead at least read region, and
4784 avoiding excessive large RPC for small reads.
4786 Severity : enhancement
4788 Description: Creating more than 1000 files for a single job may cause a load
4789 imbalance on the OSTs if there are also a large number of OSTs.
4790 Details : qos_prep_create() uses an OST index reseed value that is an
4791 even multiple of the number of available OSTs so that if the
4792 reseed happens in the middle of the object allocation it will
4793 still utilize the OSTs as uniformly as possible.
4798 Description: OST or MDS may oops in ping_evictor_main()
4799 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4800 but doesn't restart at beginning of obd_exports_timed list
4803 Severity : enhancement
4805 Description: improve by-nid export eviction on the MDS and OST
4806 Details : allow multiple exports with the same NID to be evicted at one
4807 time without re-searching the exports list.
4810 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4812 Description: MDS may oops in groups_free()
4813 Details : in rare race conditions a newly allocated group_info struct is
4814 freed again, and this can be NULL. The 2.4 compatibility code
4815 for groups_free() checked for a NULL pointer, but 2.6 did not.
4818 Frequency : common for liblustre clients doing little filesystem IO
4819 Bugzilla : 9352, 7313
4820 Description: server may evict liblustre clients accessing contended locks
4821 Details : if a client is granted a lock or receives a completion AST
4822 with a blocking AST already set it would not reply to the AST
4823 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4826 Frequency : lfs setstripe, only systems with more than 160 OSTs
4828 Description: unable to set striping with a starting offset beyond OST 160
4829 Details : llapi_create_file() incorrectly limited the starting stripe
4830 index to the maximum single-file stripe count.
4833 Frequency : LDAP users only
4835 Description: lconf did not handle in-kernel recovery with LDAP properly
4836 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4838 Severity : enhancement
4840 Description: bind OST threads to NUMA nodes to improve performance
4841 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4842 node and do their allocations there to localize memory access
4844 Severity : enhancement
4846 Description: llmount can determine client NID directly from Myrinet (GM)
4847 Details : the client NID code from gmnalnid was moved directly into
4848 llmount, removing the need to use this or specifying the
4849 client NID explicitly when mounting GM clients with zeroconf
4852 Frequency : if client is started with down MDS
4854 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4855 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4856 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4857 ptlrpc_import_delayed_req() because we didn't check if the
4858 request was interrupted, and we also didn't break out of the
4859 event loop if there was no timeout
4864 Description: data loss during non-page-aligned writes to a single file from
4865 both multiple nodes and multiple threads on one node at same time
4866 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4867 inconsistency led to false short-reads, that were cached and later
4868 used by ->prepare_write() to fill in partially written page,
4869 leading to data loss.
4872 Frequency : always, if lconf --abort_recovery used
4874 Description: lconf --abort_recovery fails with 'Operation not supported'
4875 Details : lconf was attempting to abort recovery on the MDT device and not
4878 Severity : enhancement
4880 Description: remove cleanup logs
4881 Details : replace lconf-generated cleanup logs with lustre internal
4882 cleanup routines. Eliminates the need for client-cleanup and
4885 Severity : enhancement
4887 Description: add support for EAs (user and system) on lustre filesystems
4888 Details : it is now possible to store extended attributes in the Lustre
4889 client filesystem, and with the user_xattr mount option it
4890 is possible to allow users to store EAs on their files also
4892 Severity : enhancement
4894 Description: Add possibility (config option) to show minimal available OST free
4896 Details : When compiled with --enable-mindf configure option, statfs(2)
4897 (and so, df) will return least minimal free space available from
4898 all OSTs as amount of free space on FS, instead of summary of
4899 free spaces of all OSTs.
4901 Severity : enhancement
4903 Description: do not expand extent locks acquired on OST-side
4904 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4905 by server: they are not cached anyway.
4908 Frequency : when mmap is used/binaries executed from Lustre
4910 Description: Unmmap pages before throwing them away from read cache.
4911 Details : llap_shrink cache now attempts to unmap pages before discarding
4912 them (if unmapping failed - do not discard). SLES9 kernel has
4913 extra checks that trigger if this unmapping is not done first.
4918 Description: lconf didn't resolve symlinks before checking to see whether a
4919 given mountpoint was already in use
4922 Frequency : when migrating failover services
4923 Bugzilla : 6395, 9514
4924 Description: When migrating a subset of services from a node (e.g. failback
4925 from a failover service node) the remaining services would
4926 time out and evict clients.
4927 Details : lconf --force (implied by --failover) sets the global obd_timeout
4928 to 5 seconds in order to quickly disconnect, but this caused
4929 other RPCs to time out too quickly. Do not change the global
4930 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4932 Severity : enhancement
4933 Frequency : if MDS is started with down OST
4934 Bugzilla : 9439,5706
4935 Description: Allow startup/shutdown of an MDS without depending on the
4936 availability of the OSTs.
4937 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4938 Add appropriate locking and lov-osc refcounts for safe
4939 cleaning. Add osc abort_inflight calls in case the
4940 synchronize never started.
4943 Frequency : occasional (Cray XT3 only)
4945 Description: root not authorized to access files in CRAY_PORTALS environment
4946 Details : The client process capabilities were not honoured on the MDS in
4947 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4948 been accessed by an authorized user then root was able to access
4949 the file on the local client also. The root user capabilities
4950 are now allowed on the MDS, as this environment has secure UID.
4953 Frequency : occasional
4955 Description: ldiskfs "too long searching" message happens too often
4956 Details : A debugging message (otherwise harmless) prints too often on
4957 the OST console. This has been reduced to only happen when
4958 there are fragmentation problems on the filesystem.
4963 Description: Division by zero in statfs when all OSCs are inactive
4964 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4965 lov_statfs divides by value returned by lov_get_stripecnt().
4969 Bugzilla : 9489, 3273
4970 Description: First write from each client to each OST was only 4kB in size,
4971 to initialize client writeback cache, which caused sub-optimal
4972 RPCs and poor layout on disk for the first written file.
4973 Details : Clients now request an initial cache grant at (re)connect time
4974 and so that they can start streaming writes to the cache right
4975 away and always do full-sized RPCs if there is enough data.
4976 If the OST is rebooted the client also re-establishes its grant
4977 so that client cached writes will be honoured under the grant.
4982 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4983 Details : Now I/O RPCs go to different portal number and (presumably) fast
4984 lock requests (and glimses) and other RPCs get their own service
4985 threads pool that should be able to service those RPCs
4988 Severity : enhancement
4990 Description: Ability to exchange lustre version between client and servers and
4991 issue warnings at client side if client is too old. Also for
4992 liblustre clients there is ability to refuse connection of too old
4994 Details : New 'version' field is added to connect data structure that is
4995 filled with version info. That info is later checked by server and
4999 Frequency : rare, liblustre only.
5000 Bugzilla : 9296, 9581
5001 Description: Two simultaneous writes from liblustre at offset within same page
5002 might proceed at the same time overwriting eachother with stale
5004 Details : I/O lock withing llu_file_prwv was released too early, before data
5005 actually was hitting the wire. Extended lock-holding time until
5006 server acknowledges receiving data.
5009 Frequency : extremely rare. Never observed in practice.
5011 Description: avoid generating lustre_handle cookie of 0.
5012 Details : class_handle_hash() generates handle cookies by incrementing
5013 global counter, and can hit 0 occasionaly (this is unlikely, but
5014 not impossible, because initial value of cookie counter is
5015 selected randonly). Value of 0 is used as a sentinel meaning
5016 "unassigned handle" --- avoid it. Also coalesce two critical
5017 sections in this function into one.
5019 Severity : enhancement
5021 Description: allow liblustre clients to delegate truncate locking to OST
5022 Details : To avoid overhead of locking, liblustre client instructs OST to
5023 take extent lock in ost_punch() on client's behalf. New connection
5024 flag is added to handle backward compatibility.
5026 Severity : enhancement
5027 Bugzilla : 4928, 7341, 9758
5028 Description: allow number of OST service threads to be specified
5029 Details : a module parameter allows the number of OST service threads
5030 to be specified via "options ost ost_num_threads={N}" in the
5031 OSS's /etc/modules.conf or /etc/modprobe.conf.
5035 Bugzilla : 6146, 9635, 9895
5036 Description: servers crash with bad pointer in target_handle_connect()
5037 Details : In rare cases when a client is reconnecting it was possible that
5038 the connection request was the last reference for that export.
5039 We would temporarily drop the export reference and get a new
5040 one, but this may have been the last reference and the export
5041 was just destroyed. Get new reference before dropping old one.
5043 Severity : enhancement
5044 Frequency : if client is started with failover MDS
5046 Description: Allow multiple MDS hostnames in the mount command
5047 Details : Try to read the configuration from all specified MDS
5048 hostnames during a client mount in case the "primary"
5051 Severity : enhancement
5053 Description: Stop sending data to evicted clients as soon as possible.
5054 Details : Check if the client we are about to send or are sending data to
5055 was evicted already. (Check is done every second of waiting,
5056 for which l_wait_event interface was extended to allow checking
5057 of exit condition at specified intervals).
5060 Frequency : rare, normally only when NFS exporting is done from client
5062 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5064 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5067 Frequency : always, if srand() is called before liblustre initialization
5069 Description: Liblustre uses system PRNG disturbing its usage by user application
5070 Details : Introduce internal to lustre fast and high-quality PRNG for
5071 lustre usage and make liblustre and some other places in generic
5072 lustre code to use it.
5074 Severity : enhancement
5075 Bugzilla : 9477, 9557, 9870
5076 Description: Verify that the MDS configuration logs are updated when xml is
5077 Details : Check if the .xml configuration logs are newer than the config
5078 logs stored on the MDS and report an error if this is the case.
5079 Request --write-conf, or allow starting with --old_conf.
5081 Severity : enhancement
5083 Description: Handle symlinks in the path when checking if Lustre is mounted.
5084 Details : Resolve intermediate symlinks when checking if a client has
5085 mounted a filesystem to avoid duplicate client mounts.
5090 Description: lconf can hit an error exception but still return success.
5091 Details : The lconf command catches the Command error exception at the top
5092 level script context and will exit with the associated exit
5093 status, but doesn't ensure that this exit status is non-zero.
5098 Description: failure of ptlrpc thread startup can cause oops
5099 Details : Starting a ptlrpc service thread can fail if there are a large
5100 number of threads or the server memory is very fragmented.
5101 Handle this without oopsing.
5104 Frequency : always, only if liblustre and non-default acceptor port was used
5106 Description: liblustre cannot connect to servers with non-default acceptor port
5107 Details : tcpnal_set_default_params() was not called and was therefore
5108 ignoring the environment varaible TCPNAL_PORT, as well as other
5109 TCPNAL_ environment variables
5114 Description: two objects could be created on the same OST for a single file
5115 Details : If an OST is down, in some cases it was possible to create two
5116 objects on a single OST for a single file. No problems other
5117 than potential performance impact and spurious error messages.
5121 Bugzilla : 5681, 9562
5122 Description: Client may oops in ll_unhash_aliases
5123 Details : Client dcache may become inconsistent in race condition.
5124 In some cases "getcwd" can fail if the current directory is
5130 Description: Inode refcounting problems in NFS export code
5131 Details : link_raw functions used to call d_instantiate without obtaining
5132 extra inode reference first.
5136 Bugzilla : 9942, 9903
5137 Description: Referencing freed requests leading to crash, memleaks with NFS.
5138 Details : We used to require that call to ll_revalidate_it was always
5139 followed by ll_lookup_it. Also with revalidate_special() it is
5140 possible to call ll_revalidate_it() twice for the same dentry
5141 even if first occurence returned success. This fix changes semantic
5142 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5143 reference on a request referred from the intent.
5144 ll_intent_release() then releases such a request.
5147 Frequency : rare, normally benchmark loads only
5149 Description: unlinked inodes were kept in memory on the client
5150 Details : If a client is repeatedly creating and unlinking files it
5151 can accumulate a lot of stale inodes in the inode slab cache.
5152 If there is no other client load running this can cause the
5153 client node to run out of memory. Instead flush old inodes
5154 from client cache that have the same inode number as a new inode.
5157 Frequency : SLES9 2.6.5 kernel and long filenames only
5158 Bugzilla : 9969, 10379
5159 Description: utime reports stale NFS file handle
5160 Details : SLES9 uses out-of-dentry names in some cases, which confused
5161 the lustre dentry revalidation. Change it to always use the
5165 Frequency : rare, unless heavy write-truncate concurrency is continuous
5166 Bugzilla : 4180, 6984, 7171, 9963, 9331
5167 Description: OST becomes very slow and/or deadlocked during object unlink
5168 Details : filter_destroy() was holding onto the parent directory lock
5169 while truncating+unlinking objects. For very large objects this
5170 may block other threads for a long time and slow overall OST
5171 responsiveness. It may also be possible to get a lock ordering
5172 deadlock in this case, or run out of journal credits because of
5173 the combined truncate+unlink. Solution is to do object truncate
5174 first in one transaction without parent lock, and then do the
5175 final unlink in a new transaction with the parent lock. This
5176 reduces the lock hold time dramatically.
5179 Frequency : rare, 2.4 kernels only
5181 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5182 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5183 before unmounting their filesystem in order to avoid deadlock.
5184 The kernel_locked() function in 2.4 kernels only checks whether
5185 the kernel lock is held, not whether it is this process that is
5186 holding it as 2.6 kernels do.
5191 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5192 Details : The client ptlrpc code may be trying to reconnect to a down
5193 server before a previous connection attempt has timed out.
5194 Increase the reconnect interval to be longer than the connection
5195 timeout interval to avoid sending duplicate connections to
5199 Frequency : echo_client brw_test command
5201 Description: fix echo_client to work with OST preallocated code
5202 Details : OST preallocation code (5137) didn't take echo_client IO path
5203 into account: echo_client calls filter methods outside of any
5204 OST thread and, hence, there is no per-thread preallocated
5205 pages and buffers to use. Solution: hijack pga pages for IO. As
5206 a byproduct, this avoids unnecessary data copying.
5210 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5211 Description: Client can oops in mdc_commit_close() after open replay
5212 Details : It was possible for the MDS to return an open request with no
5213 transaction number in mds_finish_transno() if the client was
5214 evicted, but without actually returning an error. Clients
5215 would later try to replay that open and may trip an assertion
5216 Simplify the client close codepath, and always return an error
5217 from the MDS in case the open is not successful.
5220 Frequency : rare, 2.6 OSTs only
5222 Description: OST may deadlock under high load on fragmented files
5223 Details : If there was a heavy load and highly-fragmented OST filesystems
5224 it was possible to have all the OST threads deadlock waiting on
5225 allocation of biovecs, because the biovecs were not released
5226 until the entire RPC IO was completed. Instead, release biovecs
5227 as soon as they are complete to ensure forward IO progress.
5229 Severity : enhancement
5231 Description: Support for specifying external journal device at mount
5232 Details : If an OST or MDS device is formatted with an external journal
5233 device, this device major/minor is stored in the ext3 superblock
5234 and may not be valid for failover. Allow detecting and
5235 specifying the external journal at mount time.
5240 Description: Mounting an MDS with pending unlinked files may cause oops
5241 Details : target_finish_recovery() calls mds_postrecov() which returned
5242 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5243 considers this an error and immediately begins cleaning up the
5244 lov, just after starting the mds_lov process
5246 Severity : enhancement
5248 Description: Implement 'lfs df' to report actual free space on per-OST basis
5249 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5250 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5251 usage of objects; '-h' to report in human readable format.
5253 ------------------------------------------------------------------------------
5255 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5262 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5263 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5264 at startup during orphan unlinking. Instead, skip buddy update
5265 before setup, as it will be regenerated after recovery is complete.
5270 Description: If an OST is inactive, its locks might reference stale inodes.
5271 Details : lov_change_cbdata() must iterate over all namespaces, even if
5272 they are inactive to clear inode references from the lock.
5274 Severity : enhancement
5275 Frequency : occasional, if non-standard max_dirty_mb used
5277 Description: Client will block write RPCs if not enough grant
5278 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5279 then the client will block writes while waiting for another RPC
5280 to complete instead of consuming its dirty limit. With change
5281 we get improved performance when max_dirty_mb is small.
5283 Severity : enhancement
5284 Bugzilla : 3389, 6253
5285 Description: Add support for supplementary groups on the MDS.
5286 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5287 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5288 lookups for user supplementary groups into a cache.
5292 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5293 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5294 for liblustre O_EXCL is enforced by MDS.
5298 Description: readdir never returns NULL in liblustre.
5299 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5300 dentry in struct dirent.
5304 Frequency : liblustre only, depends on application IO pattern
5305 Description: liblustre clients evicted if not contacting servers
5306 Details : Don't put liblustre clients into the ping_evictor list, so
5307 they will not be evicted by the pinger ever.
5309 Severity : enhancement
5311 Description: Add ability to evict clients by NID from MDS.
5312 Details : By echoing "nid:$NID" string into
5313 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5314 $NID would be instantly evicted from this MDS and from all active
5315 OSTs connected to it.
5319 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5320 Details : lookup_it_finish() used to query file size from OSTs that was not
5325 Description: service threads change working directory to that of init
5326 Details : Starting lustre service threads may pin the working directory
5327 of the parent thread, making that filesystem busy. Threads
5328 now change to the working directory of init to avoid this.
5332 Frequency : during shutdown only
5333 Description: shutdown with a failed MDS or OST can cause unmount to hang
5334 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5339 Frequency : 2.6 only
5340 Description: chmod/chown may include an extra supplementary group
5341 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5342 supplementary group and if none is specified this is used.
5345 Bugzilla : 5479 (6816)
5347 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5348 Details : If lookup is in progress on a file that is unlinked we might try
5349 to revalidate the inode and fail in revalidate after lookup is
5350 complete and ll_file_open() enqueues the open again but
5351 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5354 Frequency : always, if lconf --abort_recovery used
5356 Description: lconf --abort_recovery fails with 'Operation not supported'
5357 Details : lconf was attempting to abort recovery on the MDT device and not
5360 ------------------------------------------------------------------------------
5362 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5367 Frequency : rare (only unsupported configurations with a node running as an
5369 Bugzilla : 6514, 5137
5370 Description: Mounting a Lustre file system on a node running as an OST could
5372 Details : OSTs now preallocates memory needed to write out data at
5373 startup, instead of when needed, to avoid having to
5374 allocate memory in possibly low memory situations.
5375 Specifically, if the file system is mounted on on OST,
5376 memory pressure could force it to try to write out data,
5377 which it needed to allocate memory to do. Due to the low
5378 memory, it would be unable to do so and the node would
5379 become unresponsive.
5381 Severity : enhancement
5383 Description: Addition of lconf --service command line option
5384 Details : lconf now accepts a '--service <arg>' option, which is
5385 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5387 Severity : enhancement
5389 Description: Failover mode is now the default for OSTs.
5390 Details : By default, OSTs will now run in failover mode. To return to
5391 the old behaviour, add '--failout' to the lmc line for OSTs.
5393 Severity : enhancement
5395 Description: Health checks are now provided for MDS and OSTs
5396 Details : Additional detailed health check information on MSD and OSTs
5397 is now provided through the procfs health_check value.
5400 Frequency : occasional, depends on IO load
5402 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5403 after numerous create/delete cycles
5404 Details : The ext3 inode allocation policy would not allocate new inodes
5405 very well on the OSTs because there are no new directories
5406 being created. Instead we look for groups with free space if
5407 the parent directories are nearly full.
5412 Description: Network or server problems during mount may cause partially
5413 mounted clients instead of returning an error.
5414 Details : The config llog parsing code may overwrite the error return
5415 code during mount error handling, returning success instead
5421 Description: MDS can fail to allocate large reply buffers
5422 Details : After long uptimes the MDS can fail to allocate large reply
5423 buffers (e.g. zconf client mount config records) due to memory
5424 fragmentation or consumption by the buffer cache. Preallocate
5425 some large reply buffers so that these replies can be sent even
5426 under memory pressure.
5430 Frequency : rare (liblustre)
5431 Description: fsx running with liblustre complained that using truncate() to
5432 extend the file doesn't work. This patch corrects that issue.
5433 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5434 ATTR_SIZE and lsm use in llu_setattr_raw.
5438 Frequency : rare, only 2.6 kernels
5439 Description: Unusual file access patterns on the MDS may result in inode
5440 data being lost in very rare circumstances.
5441 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5442 code in the 2.6 kernel caused the inode and/or EA data not to
5443 be read from disk, causing single-file corruption.
5447 Frequency : rare, only 2.6 filesystems using extents
5448 Description: Heavy concurrent write and delete load may cause data corruption.
5449 Details : It was possible under high-load situations to have an extent
5450 metadata block in the block device cache from a just-unlinked
5451 file overwrite a newly-allocated data block. We now unmap any
5452 metadata buffers that alias just-allocated data blocks.
5456 Frequency : filesystems with default stripe_count larger than 77
5457 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5458 Details : lconf specifies an inode size of 4096 bytes when the default
5459 stripe_count is larger than 77. This conflicts with the default
5460 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5461 this case to avoid pinning too much memory for large EAs.
5463 ------------------------------------------------------------------------------
5465 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5470 Frequency : rare (extremely heavy IO load with hundreds of clients)
5472 Description: Client is evicted, gets IO error writing to file
5473 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5474 caused clients to be evicted for AST timeouts. The fixes in
5475 bug 5192 mean we no longer need to have such short AST timeouts
5476 so ldlm_timeout has been increased.
5479 Frequency : occasional during --force or --failover shutdown under load
5480 Bugzilla : 5949, 4834
5481 Description: Server oops/LBUG if stopped with --force or --failover under load
5482 Details : a collection of import/export refcount and cleanup ordering
5483 issues fixed for safer force cleanup
5486 Frequency : only filesystems larger than 120 OSTs
5487 Bugzilla : 5990, 6223
5488 Description: lfs getstripe would oops on a very large filesystem
5489 Details : lov_getconfig used kfree on vmalloc'd memory
5492 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5493 Bugzilla : 6242, 6243
5494 Description: reading from files that had been truncated to a non-zero size
5495 but never opened returned no data
5496 Details : ll_file_read() reads zeros from no-object files to EOF
5501 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5502 Details : There's little harm in aborting MDS/OSS recovery and letting it
5503 try again, so I removed the LASSERT and return an error instead.
5505 Severity : enhancement
5507 Description: New debugging infrastructure for tracking down data corruption
5508 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5509 (b) cover more of the client-side code path, and (c) try to narrow
5510 down where problems occurred
5514 Bugzilla : 3819, 4364, 4397, 6313
5515 Description: Racing close and eviction MDS could cause assertion in mds_close
5516 Details : It was possible to get multiple mfd references during close and
5517 client eviction, leading to one thread referencing a freed mfd.
5519 Severity: : enhancement
5520 Bugzilla : 3262, 6359
5521 Description: Attempts to reconnect to servers are now more aggressive.
5522 Details : This builds on the enhanced upcall-less recovery that was added
5523 in 1.4.2. When trying to reconnect to servers, clients will
5524 now try each server in the failover group every 10 seconds. By
5525 default, clients would previously try one server every 25 seconds.
5530 Description: After recovery, certain operations trigger a failed
5531 assertion on a client.
5532 Details : Failing over an mds, using lconf -d --failover, while a
5533 client was doing a readdir() call would cause the client to
5534 LBUG after recovery completed and the readdir() was resent.
5536 Severity : enhancement
5538 Description: Default groups are now added by lconf
5539 Details : You can now run lconf --group <servicename> without having to
5540 manually add groups with lmc.
5543 Frequency : occasional
5545 Description: Nodes with an elan id of 0 trigger a failed assertion
5548 Frequency : always when accessing e.g. tty/console device nodes
5550 Description: tty and some other devices nodes cannot be used on lustre
5551 Details : file's private_data field is used by device data and lustre
5552 values in there got lost. New field was added to struct file to
5553 store fs-specific private data.
5556 Frequency : when exporting Lustre via NFS
5558 Description: NFSD failed occasionally when looking up a path component
5559 Details : NFSD is looking up ".." which was broken in ext3 directories
5560 that had grown large enough to become hashed.
5563 Frequency : Clusters with multiple interfaces not on the same subnet
5565 Description: Nodes will repeatedly try to reconnect to an interface which it
5566 cannot reach and report an error to the log.
5567 Details : Extra peer list entries will be created by lconf with some peers
5568 unreachable. lconf now validates the peer before adding it.
5571 Frequency : Only if a default stripe is set on the filesystem root.
5573 Description: Setting a default stripe on the filesystem root prevented the
5574 filesystem from being remounted.
5575 Details : The client was sending extra request flags in the root getattr
5576 request and did not allocate a reply buffer for the dir EA.
5579 Frequency : occasional, higher if lots of files are accessed by one client
5580 Bugzilla : 6159, 6097
5581 Description: Client trips assertion regarding lsm mismatch/magic
5582 Details : While revalidating inodes the VFS looks up inodes with ifind()
5583 and in rare cases can find an inode that is being freed.
5584 The ll_test_inode() code will free the lsm during ifind()
5585 when it finds an existing inode and then the VFS later attaches
5586 this free lsm to a new inode.
5590 Bugzilla : 6422, 7030
5591 Description: MDS deadlock between mkdir and client eviction
5592 Details : Creating a new file via mkdir or mknod (starting a transaction
5593 and getting the ns lock) can deadlock with client eviction
5594 (gets ns lock and trying to finish a synchronous transaction).
5597 Frequency : occasional
5598 Description: While starting a server, the fsfilt_ext3 module could not be
5600 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5601 kernels. Previously, lconf would still use the ext3 name
5602 when trying to load modules. Now, it will correctly use
5603 ext3 on 2.4 and ldiskfs on 2.6.
5605 Severity : enhancement
5606 Description: The default stripe count has been changed to 1
5607 Details : The interpretation of the default stripe count (0, to lfs
5608 or lmc) has been changed to mean striping across a single
5609 OST, rather than all available. For general usage we have
5610 found a stripe count of 1 or 2 works best.
5612 Severity : enhancement
5613 Description: Add support for compiling against Cray portals.
5614 Details : Conditional compiling for some areas that are different
5618 Frequency : occasional
5619 Bugzilla : 6409, 6834
5620 Description: Creating files with an explicit stripe count may lead to
5621 a failed assertion on the MDS
5622 Details : If some OSTs are full or unavailable, creating files may
5623 trigger a failed assertion on the MDS. Now, Lustre will
5624 try to use other servers or return an error to the
5628 Frequency : occasional
5630 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5631 caused lock timeout during readahead (since 1.4.2).
5632 Details : Processes doing readahead might match a lock that hasn't been
5633 granted yet if there are overlapping and conflicting lock
5634 requests. The readahead process waits on ungranted lock
5635 (original lock is CBPENDING), while OST waits for that process
5636 to cancel CBPENDING read lock and eventually evicts client.
5638 Severity : enhancement
5640 Description: Initial enabling of flock support for clients
5641 Details : Implements fcntl advisory locking and file status functions.
5642 This feature is provided as an optional mount flag (default
5643 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5644 locking are implemented yet, and those that are are not guaranteed
5645 to be completely correct in production environments.
5646 mount -t lustre -o [flock|noflock] ...
5649 Frequency : occasional
5651 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5652 assertion in the ext3 JBD (journaling) layer.
5653 Details : The b_committed_data struct is protected by the big kernel lock
5654 in 2.4 kernels, serializing journal_commit_transaction() and
5655 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5656 access to this struct. In 2.6 kernels there is finer grained
5657 locking to improve SMP performance of the JBD layer.
5661 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5666 Description: merge and cleanup kernel patches.
5669 -----------------------------------------------------------------------------
5671 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5673 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5674 versions of Lustre. Please update all servers and clients to
5675 version 1.4.2 or later at the same time. You must also run
5676 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5677 to update the configuration logs.
5679 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5680 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5681 - swab llog records of type '0' so we get proper header size/idx (5861)
5682 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5683 - fix rename of one directory over another leaking an inode (5953)
5684 - avoid SetPageDirty on 2.6 (5981)
5685 - don't re-add just-being-destroyed locks to the waiting list (5653)
5686 - when creating new directories, inherit the parent's custom
5687 striping settings if present parent (3048)
5688 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5689 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5690 - don't hold client locks on temporary worklist from l_lru (5666)
5691 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5692 - automatically evict dead clients (5921)
5693 - Update file size properly in create+truncate+fstat case (6196)
5694 - Do not unhash mountpoint dentries, do not allow removal of
5696 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5697 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5698 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5699 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5700 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5701 - Correctly handle reads of files with no objects (6243)
5702 - lctl recover will also mark a device active if deactivate used (5933)
5704 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5705 - allow --write-conf on an MDS with different nettype than client (5619)
5706 - don't write config llogs to MDS for mounts not from that MDS (5617)
5707 - lconf should create multiple TCP connections from a client (5201)
5708 - init scripts are now turned off by default; run chkconfig --on
5709 lustre and chkconfig --on lustrefs to use them
5710 - upcalls are no longer needed for clients to recover to failover
5712 - add --abort-recovery option to lconf to abort recovery on device
5714 - add support for an arbitrary number of OSTs (3026)
5715 - Quota support protocol changes.
5716 - forward compatibility changes to wire structs (6007)
5717 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5718 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5719 - improved "lustre status" script
5720 - initialize blocksize for non-regular files (6062)
5721 - added --disable-server and --disable-client configure options (5782)
5722 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5723 - Vanilla 2.4.29 support
5724 - increase maximum number of obd devices to 520 (6242)
5725 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5726 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5727 SLES 9 2.6.5 kernels
5728 - limit stripes per file to 160 (the maximum EA size) (6093)
5730 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5733 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5734 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5735 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5736 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5737 - small ext3 extents cleanups and fixes (5733)
5738 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5739 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5740 - avoid needless client->OST connect, fix handle mismatch (5317)
5741 - fix DLM error path that led to out-of-sync client, long delays (5779)
5742 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5743 - fix several locking issues related to i_size (5492,5624,5654,5672)
5744 - don't move pending lock onto export if it is already evicted (5683)
5745 - fix kernel oops when creating .foo in unlinked directory (5548)
5746 - fix deadlock in obdfilter statistics vs. object create (5811)
5747 - use time_{before,after} to avoid timer jiffies wrap (5882)
5748 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5749 - Do not leak request if server was not able to process it (5154)
5750 - If mds_open unable to find parent dir, make that negative lookup(5154)
5751 - don't create new directories with extent-mapping (5909, 5936)
5753 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5754 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5755 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5756 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5757 - support for RHEL4 kernel on i686 (5773)
5758 - provide error messages when incompatible logs are encountered (5898)
5760 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5761 * version 1.4.0.10 (1.4.1 release candidate 1)
5763 - don't keep a lock reference when lock is not granted (4238)
5764 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5765 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5766 - drop import inflight refcount on signal_completed_replay error (5255)
5767 - unlock page after async write error during send (3677)
5768 - handle missing objects in filter_preprw_read properly (5265)
5769 - no transno return for symlink open, don't save no-trasno open (3440)
5770 - don't try to complete elan receive that already failed (4012)
5771 - free RPC server reply state on error (5406)
5772 - clean up thread from ptlrpc_start_thread() on error (5160)
5773 - readahead could read extra page into cache that wasn't ejected (5388)
5774 - prevent races in class_attach/setup/cleanup/detach (5260)
5775 - don't dereference de->d_inode after l_dput of de (5458)
5776 - use "int" for stripe value returned from lock_to_stripe (5544)
5777 - mballoc allocation and error-checking fixes in 2.6 (5504)
5778 - block device patches to fix I/O request sizes in 2.6 (5482)
5779 - look up hostnames for IB nals (5602)
5780 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5781 - don't start multiple acceptors for the same port (5277)
5782 - fix incorrect LASSERT in mds_getattr_name (5635)
5783 - export a proc file for general "ping" checking (5628)
5784 - fix "lfs check" to not block when the MDS is down (5628)
5786 - service request history (4965)
5787 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5788 - create an "evict_client" /proc entry on OSTs, like the MDS has
5789 - fix mount usage message, return errors per mount(8) (5168)
5790 - change grep [] to grep "[]" in tests so they work in more UMLs
5791 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5792 - remove ancient LOV_MAGIC_V0 EA support (5047)
5793 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5794 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5795 - allow readahead allocations to fail when low on memory (5383)
5796 - mmap locking landed again, after considerable improvement (2828)
5797 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5799 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5802 - send OST transaction number in read/write reply to free req (4966)
5803 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5804 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5805 - the watchdog thread now runs as interruptible (5246)
5806 - flock/lockf fixes (but it's still disabled, pending 5135)
5807 - don't use EXT3 constants in llite code (5094)
5808 - memory shortage at startup could cause assertion (5176)
5810 - reorganization of lov code
5811 - single portals codebase
5813 - add extents/mballoc support (5025)
5814 - direct I/O reads in the obdfilter (4048)
5815 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5817 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5820 - send OST transaction number in read/write reply to free req (4966)
5821 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5822 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5823 - the watchdog thread now runs as interruptible (5246)
5824 - handle missing objects in filter_preprw_read properly (5265)
5825 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5826 - drop import inflight refcount on signal_completed_replay error (5255)
5827 - unlock page after async write error during send (3677)
5828 - return original error code on reconstructed replies (3761)
5829 - no transno return for symlink open, don't save no-trasno open (3440)
5831 - add pid to ldlm debugging output (4922)
5832 - bump the watchdog timeouts -- we can't handle 30sec yet
5833 - extra debugging for orphan dentry/inode bug (5259)
5835 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5838 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5839 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5840 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5841 - fix NULL dereference in /proc/sys/portals/routes (4827)
5842 - allow failed mdc_close() operations to be interrupted (4561)
5843 - stop precreate on OST before MDS would time out on it (4778)
5844 - don't send partial-page writes before EOF from client (4410)
5845 - discard client grant for sub-page writes on large-page clients (4520)
5846 - don't free dentries not owned by NFS code, check generation (4806)
5847 - fix lsm leak if mds_create_objects() fails (4801)
5848 - limit debug_daemon file size, always print CERROR messages (4789)
5849 - use transno after validating reply (3892)
5850 - process timed out requests if import state changes (3754)
5851 - update mtime on OST during writes, return in glimpse (4829)
5852 - add mkfsoptions to LDAP (4679)
5853 - use ->max_readahead method instead of zapping global ra (5039)
5854 - don't interrupt __l_wait_event() during strace
5856 - add software watchdogs to catch hung threads quickly (4941)
5857 - make lustrefs init script start after nfs is mounted
5858 - fix CWARN/ERROR duplication (4930)
5859 - return async write errors to application if possible (2248)
5860 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5861 - print NAL number in %x format (4645)
5862 - update barely-supported suse-2.4.21-171 series (4842)
5863 - support for sles 9 %post scripts
5864 - support for building 2.6 kernel-source packages
5865 - support for sles km_* packages
5867 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5870 - ignore -ENOENT errors in osc_destroy (3639)
5871 - notify osc create thread that OSC is being cleaned up (4600)
5872 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5873 - reconstruct ost_handle() like mds_handle() (4657)
5874 - create a new thread to do import eviction to avoid deadlock (3969)
5875 - let lconf resolve symlinked-to devices (4629)
5876 - don't unlink "objects" from directory with default EA (4554)
5877 - hold socknal file ref over connect in case target is down (4394)
5878 - allow more than 32000 subdirectories in a single directory (3244)
5879 - fix blocks count for O_DIRECT writes (3751)
5880 - OST returns ENOSPC from object create when no space left (4539)
5881 - don't send truncate RPC if file size isn't changing (4410)
5882 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5883 - bind to privileged port in socknal and tcpnal (3689)
5885 - rate limit CERROR/CWARN console message to avoid overload (4519)
5886 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5887 - basic mmap support (3918)
5888 - kernel patch series update from b1_4 (4711)
5890 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5893 - avoid crash during MDS cleanup with OST shut down (2775)
5894 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5895 - don't use bad inodes on the MDS (3744)
5896 - dynamic object preallocation to improve recovery speed (4236)
5897 - don't hold spinlock over lock dumping or change debug flags (4401)
5898 - don't zero obd_dev when it is force cleaned (3651)
5899 - print grants to console if they go negative (4431)
5900 - "lctl deactivate" will stop automatic recovery attempts (3406)
5901 - look for existing locks in ldlm_handle_enqueue() (3764)
5902 - don't resolve lock handle twice in recovery avoiding race (4401)
5903 - revalidate should check working dir is a directory (4134)
5905 - don't always mark "slow" obdfilter messages as errors (4418)
5907 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5910 - don't close LustreDB during write_conf until it is done (3860)
5911 - fix typo in lconf for_each_profile (3821)
5912 - allow dumping logs from multiple threads at one time (3820)
5913 - don't allow multiple threads in OSC recovery (3812)
5914 - fix debug_size parameters (3864)
5915 - fix mds_postrecov to initialize import for llog ctxt (3121)
5916 - replace config semaphore with spinlock (3306)
5917 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5918 - don't allow enqueue to complete on a destroyed export (3822)
5919 - down write_lock before checking llog header bitmap (3825)
5920 - recover from lock replay timeout (3764)
5921 - up llog sem before sending rpc (3652)
5922 - reduce ns lock hold times when setting kms (3267)
5923 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5924 - fix NULL deref and obd_dev leak on setup error (3312)
5925 - replace some LBUG about llog ops with error handling (3841)
5926 - don't match INVALID dentries from d_lookup and spin (3784)
5927 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5928 - fix invalid assertion in ptlrpc_set_wait (3880)
5930 - add libwrap support for the TCP acceptor (3996)
5931 - add /proc/sys/portals/routes for non-root route listing (3994)
5932 - allow setting MDS UUID in .xml (2580)
5933 - print the stack of a process that LBUGs (4228)
5935 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5938 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5939 - make sure to unset replay flag from failed open requests (3440)
5940 - if default stripe count is 0, use OST count for inode size (3636)
5941 - update parent mtime/ctime on client for create/unlink (2611)
5942 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5943 - free recovery state on server during a forced cleanup (3571)
5944 - unregister_reply for resent reqs (3063)
5945 - loop back devices mounting and status check on 2.6 (3563)
5946 - fix resource-creation race that can provoke i_size == 0 (3513)
5947 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5948 - more debugging for page-accounting assertion (3746)
5949 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5950 - avoid deadlock after precreation failure (3758)
5951 - fix race and lock order deadlock in orphan handling (3450, 3750)
5952 - add validity checks when grabbing inodes from l_ast_data (3599)
5954 - add /proc/.../recovery_status to obdfilter (3428)
5955 - lightweight CDEBUG infrastructure, debug daemon (3668)
5956 - change default OSC RPC parameters to be better on small clusters
5957 - turn off OST read cache for files smaller than 32MB
5958 - install man pages and include them in rpms (3100)
5959 - add new init script for (un)mounting lustre filesystems (2593)
5960 - run chkconfig in %post for init scripts (3701)
5961 - drop scimac NAL (unmaintained)
5963 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5966 - clean kiobufs before and after use (3485)
5967 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5968 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5969 - change init script to fail more clearly if not run as root (1528)
5970 - allow clients to reconnect during replay (1742)
5971 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5972 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5973 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5974 - fix cond_resched (3554)
5975 - only evict unfinished clients after recovery (3515)
5976 - allow bulk resend, prevent data loss (3570)
5977 - dynamic ptlrpc request buffer allocation (2102)
5978 - don't allow unlinking open directory if it isn't empty (2904)
5979 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5980 - remove extraneous obd dereference causing LASSERT failure (3334)
5981 - don't use get_cycles() when creating temp. files on the mds (3156)
5982 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5983 - handle EEXIST for set-stripe, set proper directory name (3336)
5985 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5986 - fix ksocknal_fmb_callback() error messages (2918)
5988 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5991 - don't copy lvb into (possibly NULL) reply on error (2983)
5992 - don't deref dentry after dput, don't free lvb on error (2922)
5993 - use the kms to determine writeback rpc length (2947)
5994 - increment oti_logcookies when osc is inactive (2948)
5995 - update client's i_blocks count via lvb messages (2543)
5996 - handle intent open/close of special files properly (1557)
5997 - mount MDS with errors=remount-ro, like obdfilter (2009)
5998 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5999 - don't use cancelling-locks' kms values (2947)
6000 - use highest lock extent for kms, not last one (2925)
6001 - don't dereference ERR_PTR() dentry in error handling path (3107)
6002 - fix thread race in portals_debug_dumplog() (3122)
6003 - create lprocfs device entries at setup instead of at attach (1519)
6004 - common AST error handler, don't evict client on completion race (3145)
6005 - zero nameidata in detach_mnt in 2.6 (3118)
6006 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6007 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6008 - initialize RPC timeout timer earlier for 2.6 (3219)
6009 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6010 - print nal/nid for unknown nid (3258)
6011 - additional checks for oscc recovery before doing precreate (3284)
6012 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6013 - don't crash in mdc_close for bad permissions on open (3285)
6014 - zero i_rdev for non-device files (3147)
6015 - clear page->private before handing to FS, better assertion (3119)
6016 - tune the read pipeline (3236)
6017 - fix incorrect decref of invalidated dentry (2350)
6018 - provide read-ahead stats and refine rpc in flight stats (3328)
6019 - don't hold journal transaction open across create RPC (3313)
6020 - update atime on MDS at close time (3265)
6021 - close LDAP connection when recovering to avoid server load (3315)
6022 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6023 - don't leak open file on MDS after open resend (3325)
6024 - serialize filter_precreate and filter_destroy_precreated (3329)
6025 - loop device shouldn't call sync_dev() for nul device (3092)
6026 - clear page cache after eviction (2766)
6027 - resynchronize MDS->OST in background (2824)
6028 - refuse to mount the same filesystem twice on same mountpoint (3394)
6029 - allow llmount to create routes for mounting behind routers (3320)
6030 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6031 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6032 - fix rare problem with rename on htree directories (3417)
6034 - allow default OST striping configuration per directory (1414)
6035 - fix compilation for qswnal for 2.6 kernels (3125)
6036 - increase maximum number of MDS request buffers for large systems
6037 - change liblustreapi to be useful for external progs like lfsck (3098)
6038 - increase local configuration timeout for slow disks (3353)
6039 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6041 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6044 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6045 - don't overwrite extent policy data in reply if lock was blocked (2901)
6046 - drop filter export grants atomically with removal from device (2663)
6047 - del obd_self_export from work_list in class_disconnect_exports (2908)
6048 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6049 - swab reply message in mdc_close, other PPC fixes (2464)
6050 - fix destroying of named logs (2325)
6051 - overwrite old logs when running lconf --write_conf (2264)
6052 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6053 - fix race in target_handle_connect (2898)
6054 - mds_reint_create() should take same inode create lock (2926)
6055 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6056 - don't close files for self_export to avoid uninitialized obd (2936)
6057 - allow MDS with the same name as client node (2939)
6058 - hold dentry reference for closed log files for unlink (2325)
6059 - reserve space for all logs during transactions (2059)
6060 - don't evict page beyond end of stripe extent (2925)
6061 - don't oops on a deleted current working directory (2399)
6062 - handle hard links to targets without a parent properly (2517)
6063 - don't dereference NULL lock when racing during eviction (2867)
6064 - don't grow lock extents when lots of conflicting locks (2919)
6066 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6069 - account for cache space usage on clients to avoid data loss (974)
6070 - lfsck support in lustre kernel code (2349)
6071 - reduce journal credits needed for BRW writes (2370)
6072 - orphan handling to avoid losing space on client/server crashes
6073 - ptlrpcd can be blocked, stopping ALL progress (2477)
6074 - use lock value blocks to assist in proper KMS, faster stat (1021)
6075 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6076 - recovery for initial connections (2355)
6077 - fixes for mds_cleanup_orphans (1934)
6078 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6079 - block all file creations until orphan recovery completes (1901)
6080 - client remove rq_connection from request struct (2423)
6081 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6082 - recovery timer race (2670)
6083 - mdc_close recovey bug (2532)
6084 - ptlrpc cleanup bug (2710)
6085 - mds timeout on local locks (2588)
6086 - namespace lock held during RPCs (2431)
6087 - handle interrupted sync write properly (2503)
6088 - don't try to handle a message that hasn't been replied to (2699)
6089 - client assert failure during cleanup after abort recovery (2701)
6090 - leak mdc device after failed mount (2712)
6091 - ptlrpc_check_set allows timedout requests to complete (2714)
6092 - wait for inflight reqs when ptlrpcd finishes (2710)
6093 - make sure unregistered services are removed from the srv_list
6094 - reset bulk XID's when resending them (caught by 1138 test)
6095 - unregister_bulk after timeout
6096 - fix lconf error (2694)
6097 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6098 - readahead locks pages, leaves pending causing memory pressure (2673)
6099 - increase OST request buffers to 4096 on large machines (2729)
6100 - fix up permission of existing directories in simple_mkdir (2661)
6101 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6102 - don't assign transno to errored transactions (2742)
6103 - don't delete objects on OST if given a bogus objid from MDS (2751)
6104 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6105 - if rq_no_resend, then timeout request after recovery (2432)
6106 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6107 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6108 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6109 - MDS recovery shouldn't delete objects that it has given out (2730)
6110 - if enqueue arrives after completion, don't clobber LVB (2819)
6111 - don't unlock pages twice when trigger_group_io returns error (2814)
6112 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6113 - don't write pages to disk if there was an error (1450)
6114 - don't ping imports that have recovery disabled (2676)
6115 - take buffered bytes into account when balancing socknal conn (2817)
6116 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6117 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6118 - remove little-endian swabbing of llog records (1987)
6119 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6120 - retry reposting request buffers if they fail (1191)
6121 - grow extent at grant time to avoid granting a revoked lock (2809)
6122 - lock revoke doesn't evict page if covered by a second lock (2765)
6123 - disable VM readahead to avoid reading outside lock extents (2805)
6125 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6126 - updated kernel patches for hp-2.4.20 kernel (2681)
6128 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6131 - fix truncated write corruption (2366)
6132 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6134 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6135 - ensure len(uuid) < 37 in lmc (1171)
6136 - fix ia64 OOPS in llog_test (2255)
6137 - zero end of page at obdfilter for partial page writes (2648)
6138 - don't leave stale dentries around after renames (bug 2428)
6139 - fix timeouts when evicting a client with a single lock held (2642)
6140 - set deadline for the initial HELLO message to drain (2634)
6141 - print out dotted-quad IP addresses in the socknal (2302)
6143 - additional debugging for MDS client eviction problem (2443)
6144 - fix mkfsoptions support for osts (2603, 2604)
6146 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6149 - add series for the vanilla 2.6.0 kernel
6150 - add series for the vanilla 2.4.24 kernel
6151 - add series for a cray x86/64 UL kernel drop
6152 - fix xattr patches for the vanilla 2.4.19 series
6154 - generate true UUIDs in lmc (1171)
6155 - have portals stack dumping break in UML (2466)
6156 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6157 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6158 - wake callback waiting threads on client eviction (2460)
6159 - Add --ptldebug and --subsystem to lmc (1719)
6160 - update assertion to allow safe interrupt allocation
6161 - set rq_no_resend for cancel requests (2432)
6162 - recalculate ptlrpcd timeout after resend (2494)
6163 - call vfs_rmdir when removing pending directories (2368)
6164 - fix renaming a file to itself (2429)
6165 - lmc creates a default one-stripe lov (2454)
6166 - expand procfs space to handle large clusters (2326)
6167 - increase UML stack to avoid overflow
6168 - update lconf's list of debug and subsystem masks
6169 - fix lfs find --obd (2510)
6170 - /proc tunable for disabling filter read caching (2591)
6171 - stop rpm packages from altering slapd.conf (2301)
6172 - disable nagle in the socknal under 0conf (2578)
6173 - choose mds inode size based on stripe count (2572)
6174 - fix kernel-source rpm problems (2516)
6176 - add --disable-doc to avoid pdf generation (2421)
6177 - update documentation, tests, type-os, comments
6178 - avoid format warnings on ia64
6179 - remove the TOE NAL
6180 - tiny code cleanups by removing unused fields
6182 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6185 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6186 - remove the most blatant lies from BUILDING file (2371)
6187 - change default debug level to reasonable production setting
6188 - reduce client side cache size to reduce cache flush time
6189 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6190 - make TCP zerocopy and pinger support enabled by default (2476)
6191 - sync writes completed after process exits caused crashes (2319)
6192 - maintain correct mount count on the MDS (2356)
6193 - backout 1557, because 2316 wasn't really fixed
6194 - better file I/O statistics gathering in /proc
6195 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6196 - another kernel patch to fix zero-copy TCP function export
6197 - don't take duplicate lock when processing re-sent getattr (2420)
6198 - lctl uses obd_self_export instead of creating new conn (2353)
6199 - MDS/OST recovery case which requires object creation asserted (2425)
6200 - move lfs from /usr/sbin to /usr/bin in packages
6201 - fix race between mds_client_add and mds_client_free (2417)
6202 - use kmalloc instead of slabs in portals (2430)
6203 - don't create duplicate records when a failover MDS is present (2442)
6204 - remove unnecessary mount age check (2332)
6205 - don't remove directory inodes from locks prematurely (2451)
6206 - don't break if MDS service name is the same as hostname (2103)
6207 - fix races in client write RPC generation when cache full (2482)
6209 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6212 - remove now-unused request->rq_obd (278)
6213 - if an allocation fails, print out how much memory we've used (1933)
6214 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6215 - add the "configurable stack size" patch to most series files (1256)
6216 - ability to write large log records, for 100+ OST configs (2306)
6217 - fix NULL deref when filter_prep fails (2314)
6218 - fix operator precedence error in filter_sync
6219 - dynamic allocation of socknal TX descriptors (2315)
6220 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6221 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6222 - ensure that CWARN messages in Portals always get to the syslog
6223 - __init/__exit are not for prototype decls (ldlm_init/exit)
6224 - x86-64 compile warning fixes
6225 - fix gateway LMC keyword conflict (2318)
6226 - fix MDS lock inversions in getattr/reint paths (1844)
6227 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6228 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6229 - DLM race condition prevented some lock evictions (2328)
6230 - ENOMEM detection and retry on socknal sends (2230)
6231 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6232 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6233 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6234 - better error messages when a client is rejected during recovery (1505)
6235 - avoid cancelling locks which were never granted, after failure (2330)
6236 - fix i_sem/journal inversion in mds_client_add (2333)
6237 - fix truncate/getattr lock cycle deadlock (2334)
6238 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6239 - fix two rare exit paths which could leak an l_lock() ref (2321)
6240 - fencepost error in MDS/OST orphan recovery (2226)
6241 - make log record alignment 8 bytes (1988)
6242 - lstripe now fails when requested offset > ost_count (2237)
6243 - ensure that all kernel series have a complete list.h (1607)
6244 - fix crashes in special-file operations (2316)
6245 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6247 - allow configurable automake binary, for testing new versions
6248 - small update to the lfs documentation
6250 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6252 * fix negative export reference count in fsfilt_sync (2312)
6254 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6255 * release candidate 0.9.1
6257 - orphans are moved into the PENDING directory for possible recovery
6258 - replayed opens now open by fid for orphan/rename safety (1042)
6259 - last close of an orphan inode generates a transno (683)
6260 - chdir() and mount() now pin the directory entry (1020)
6261 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6262 - discard very old requests without processing them (1502)
6263 - remove some common, well-understood CERRORs (1505)
6264 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6265 - clear "grant" flags in OST replies until OST grant code lands (1644)
6266 - fix read performance by not clobbering i_blksize on client (1598)
6267 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6268 - make LRU size a /proc tunable, clears locks when reduced (707)
6269 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6270 - server locks take references on exports now (1558)
6271 - build fixes for 2.4.20-rh trees (1663)
6272 - return an error from lov_create if all OSCs are inactive (1751)
6273 - fix import levels when a reconnect happens without a timeout (1597)
6274 - exit early from mds_open if we get a lookup error (1749)
6275 - partial page read at EOF wouldn't wait for disk before sending (1642)
6276 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6277 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6278 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6279 - write/truncate lock inversion (1639)
6280 - Don't auto-load obdclass, portals modules during cleanup (1495)
6281 - fix timestamps from jumping to "now" (1763)
6282 - extra journal assertions (1648)
6283 - add an extra multiunlink test (1771)
6284 - fix read_record/write_record API (1776)
6285 - fix leak of offset_extent, possible incorrect i_size later (1772)
6286 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6287 - Debugging for the kqswnal_get_idle_tx problems (1820)
6288 - Allow recovery to be attempted multiple times (1536)
6289 - Write out MDS last_rcvd file after it is first created (1600)
6290 - Fix tx_descriptor leak in failed transmit situations (1827)
6291 - ext3 journaling fixes for assertion failure after IO error (1871)
6292 - class_export_put() on freed export after completion AST error (1896)
6293 - Fix revalidate looping in VFS (1322)
6294 - Don't access a freed export during MDS_REINT timeout (1521)
6295 - Add open-unlink recovery support on the MDS (1673,1764)
6296 - Return an error if no MDS data was read from last_rcvd (1946)
6297 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6298 - Don't setup a disk device that doesn't match exported UUID (317)
6299 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6300 - avoid committing NULL handle in force close
6301 - local.sh is now a one-stripe LOV configuration
6302 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6303 - fix ext3 htree duplicate directory entry corruption (1516)
6304 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6305 - update to new LOV EA format (2097)
6306 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6308 2003-06-15 Phil Schwan <phil@clusterfs.com>
6311 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6312 - per-import/export recovery handling (958, 931, 959)
6313 - multiple last-rcvd slots, for serving multiple FSes (949)
6314 - connections are again shared between multiple imp/exports (963, 964)
6315 - "umount -f" would hang if any requests needed to be sent (393, 978)
6316 - avoid pinning large req buffer by copying for queued messages (989)
6317 - add "uuid" to "lctl device" command to help upcalls (991)
6318 - "open" RPCs with transnos would confuse recovery counters (1037)
6319 - do proper endian conversion of all wire messages (288, 340, 891)
6320 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6321 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6322 - fix ptlrpc_connection leak in target_handle_connect (1174)
6323 - fix import refcounting bug in OST and MDS cleanup (1134)
6324 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6325 - fix very unlikely obd_types race condition (501)
6326 - remove osc_open hack for echo_client (1187)
6327 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6328 - a failure in read_inode2 leads to deadlock (1139)
6329 - cancel ack-locks as soon as transaction is committed (1072)
6330 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6331 - make sure to commitrw after any preprw to avoid deadlock (1162)
6332 - failing to execute a file in a lustre FS would lock inode (1203)
6333 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6334 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6335 - obd_uuid2tgt would walk off the end of the list (1255)
6336 - on IA64 the osc would give portals incorrect bulk size (1258)
6337 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6338 - fix lock inversion caused by new llite matching code (1282)
6339 - limit the number of dirty pages on a client to 10MB (1286)
6340 - timed out locks were not being corrected cancelled (1289)
6341 - fix O_DIRECT above 4GB on IA-32 (1292)
6342 * major user-visible changes
6343 - fail out/fail over policy now controlled by the upcall (993)
6345 - add OBD_PING to check server availability and failure (954)
6346 - lustre messages are now sent in sending host order (288, 340, 891)
6347 - add eadatalen to MDS getattr reply (340)
6348 - OST read replies may contain second buffer, with per-page status (593)
6350 2003-03-11 Phil Schwan <phil@clusterfs.com>
6353 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6354 - failed open()s could cause deadlock; fixed (867, 869)
6355 - stop cancelling OST locks when files are closed (481)
6356 - overlapping XID spaces caused network corruption (851, 853)
6357 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6358 - setattr_raw added, to do single-RPC, server-side setattrs
6359 - lmc/lconf syntax change for OST UUIDs
6360 - fix crashy race condition between ptlrpc_free_req and osc_close
6361 - don't use request in mdc_enqueue if we hit a timeout (889)
6362 - don't set the inode i_size for regular files from the MDS (896)
6363 - handle out of order completion AST (842)
6364 - don't LBUG if a lock request times out after receiving AST (913)
6365 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6366 - if a bad lock AST arrives, send an error instead of dropping entirely
6367 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6368 - fix leak in bulk IO when only partially completed (899, 900, 926)
6369 - fix O_DIRECT for ia64 (55)
6370 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6371 - C-z after timeout could hang a process forever; fixed (977)
6373 - client-side I/O cache (678, 924, 929, 941, 970)
6375 - READPAGE and SETATTRs which don't take server-side locks get
6378 2003-02-11 Phil Schwan <phil@clusterfs.com>
6381 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6382 - Fix fsfilt_extN_readpage() to read a full page of directory
6383 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6384 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6385 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6386 - Most intent operations were removed, in favour of a new RPC mode
6387 that does a single RPC to the server and bypasses most of the VFS
6388 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6389 - Aggressively cancel local locks on DLM servers
6390 - mds_reint_unlink sends EA to the client if it's the last nlink.
6391 client uses that EA to unlink OST objects.
6392 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6393 - recursive symlinks were fixed (439)
6394 - fixed NULL deref in DEBUG_REQ
6395 - filter_update_lastobjid no longer calls sync, which annoyed extN
6396 - fixed multi-client small-writes to a single file problem (445)
6397 - fixed mtime updates during file writes (607)
6398 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6399 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6400 - fixed Linux OST size reporting problem (444, 656)
6401 - OST now updates object mtime with writes or setattr (607, 619)
6402 - client verifies file size before zeroing page past EOF (445)
6403 - OST now writes last allocated objid to disk with allocation (108)
6404 - LOV on echo now works (409)
6406 - mds_reint_unlink sends a new buffer, with the EA included. this
6407 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6408 set if a regular file was being unlinked, and it was the last link
6409 - use PtlGet from the target for bulk writes (315)
6410 - OST now updates object mtime with writes or setattr (607, 619)
6411 - LDLM now has a grant-time callback to revalidate locked items, if
6413 - Many MDS operations were reorganized to combat race conditions
6415 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6416 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6418 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6421 - Fully reactivate OST imports after reconnection (512, others)
6422 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6423 - More graceful error handling for truncating on dead OST (515)
6424 - Don't error out unless we're actually accessing dead stripes (474)
6425 - Fix garbage sizes when stripes are missing (410)
6426 - LRU counters were broken, causing constant lock purge (433, 432)
6427 - garbage on read from stripes with failed OSTs (441)
6428 - mark OSCs as active before reconnecting during recovery (438)
6429 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6430 - lfind did not preserve OST order in output (443)
6431 - symlinks cause hung clients, incorrect data (439)
6432 - stop dereferencing request after dropping refcount (457)
6433 - don't LASSERT(spin_is_locked) on non-SMP (455)
6434 - fixes for many rename() bugs
6435 - fstat didn't correctly synchronize attributes (399)
6436 - server must handle lock cancellation during blocking AST prep (487)
6437 - bulk descriptors were free()d too soon (511)
6438 - fix paths in lconf, which would load incorrect modules (451, 507)
6439 - fix confusing lconf 'host not found' error message (386)
6440 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6441 - fix race condition in mdc_blocking_ast() for inode access (526)
6442 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6443 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6444 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6445 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6446 - fix lstripe in several ways (kernel side) (527)
6447 - fix request leak in ldlm_cli_enqueue (262)
6448 - incorrect OSC was marked inactive after OST failure
6449 - call mds_fs_cleanup before unmounting filesystem (524)
6450 - fix races between taking ns_lock and ldlm_lock_change_resource
6451 - fix races updating LOV export open file list
6452 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6453 - fix recovery NULL deref in ldlm_cli_cancel_unused
6454 - fix some DLM races by using new hash table for lock handles (419)
6455 - permit the client to specify desired inodes, at replay
6456 - duplicate requests when we queue them for replay reintegration
6457 - fix last_rcvd offset calculation
6458 - sync after each recovered transaction, so we always make progress
6459 - never, not always, ERESTART requests without transnos
6460 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6462 - skip replay if the MDS says that the client is already connected
6463 - don't check for a recovery-enabled export to match lctl's UUID
6464 - don't INC_USE_COUNT for phantom exports
6465 - don't crash when cleaning up phantom exports (567)
6466 - don't double-finish or set replay data for errored mdc_open requests
6467 - abort requests when they time out, so we don't get old replies
6468 - send/receive replies for AST messages again
6469 - if the client says that it doesn't have the lock, cancel it on the
6471 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6472 mark it as destroyed, it will all work out when decref is called
6473 - fix module use counts (22, 581)
6475 - ASTs now expect a reply (server cancels lock on error reply)
6477 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6480 - fix many simultaneous client startup (392)
6481 - fix dentry->d_it clobbering
6482 - credentials weren't being shipped for readdir/getattr operations
6483 - remove invalid assertions triggered during some concurrent MD
6485 - proper Lustre versions added (336, 389)
6486 - fix memory leak for create error case (398)
6487 - fix LOV locking bug that would get cli/srv out of sync
6488 - fix echo client over LOV (409)
6489 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6490 - fix double-O_EXCL intent crash (424)
6491 - avoid sending multiple lock CANCELs (352)
6493 - MDS can do multi-client recovery (modulo bugs in new code)
6495 - many updates, edits, cleanups
6497 2002-11-18 Phil Schwan <phil@clusterfs.com>
6500 - fix null d_it dereference (346)
6501 - fix full OST/dbench hang (333)
6502 - fix permission problem with file removal (286)
6503 - fix removal of OSCs from LOV when they fail
6504 - fix NULL deref during bulk timeout (214)
6505 - fix problems related to multiple filesystems on one MDS (241)
6506 - fixed serious subtle metadata locking bugs
6507 - free locks on clients when inodes are removed due to memory
6509 - fix inode pointer in lock data (285)
6510 - partial support for multiple MDS on a single host (241)
6511 - data locks weren't cancelled at clear_inode time (290, 311)
6512 - intent locks could lead to unbounded lock growth (205)
6513 - added a maximum lock count, an LRU list, and a flusher
6514 - fix multiple rename (365)
6515 - properly abstracted the echo client
6516 - OSC locked 1 byte too many; fixed
6517 - rewrote brw callback code:
6518 - fixed recovery bugs related to LOVs (306)
6519 - fixed too-many-pages-in-one-write crash (191)
6520 - fixed (again) crash in sync_io_timeout (214)
6521 - probably fixed callback-related race (385)
6523 - Add capability to MDS protocol
6524 - LDLM cancellations and callbacks on different portals
6526 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6529 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6530 - defer open object destruction to close time (601981, 138)
6531 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6532 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6533 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6534 - don't call obd_setattr in ll_file_release for destroyed objects
6535 * protocol change to lustre_msg: move |version| and add |flags|
6536 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6537 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6538 * added replay of create, unlink, link and rename operations during
6539 MDS failover; recovery should be much more robust now
6540 * remove failed OSCs from LOVs (only lov_create uses this so far)
6541 * the lustre-HOWTO was brought (more) up to date (582544)
6543 2002-10-23 Phil Schwan <phil@clusterfs.com>
6546 - in-use dentries weren't being reused properly (617851)
6547 - prevent multiple LDLM setup (599178)
6548 - fix LOV size calculations for truncate (617853)
6549 - fix client handling of MDS intent errors (POSIX)
6550 - fix permission bug in lovstripe.c test (624321)
6551 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6552 - truncate past end of file could corrupt data
6553 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6554 - a race in recovery could return ETIMEDOUT to apps (623947)
6555 - building outside the source directory was fixed
6556 * the lustre-HOWTO was brought (more) up to date (582544)
6557 * major progress was made on recovery functionality
6559 2002-10-10 Phil Schwan <phil@clusterfs.com>
6562 - recovery deadlock fix
6563 - rm -rf causes LBUG fix (617817)
6564 - file open by multiple tasks fix (618962)
6565 - directory permissions bugs (602707 and 620007)
6566 - journal_stop fixed with locking (611313)
6567 - O_APPEND failures resolved (618273, perhaps 614459)
6568 - lconf PATH fix (619770)
6569 - IA64 build fix (621450)
6570 - RPC buffer sizes scale with amount of memory
6572 2002-10-01 Phil Schwan <phil@clusterfs.com>
6575 - locks would be cancelled without throwing away data pages,
6576 resulting in inconsistent data (605627)
6577 - inode attributes were not always being refreshed (605627, 612449)
6578 - lconf now continues to cleanup after lctl reports an error
6579 - MDS now enforces user permissions (602707)
6580 - lprocfs cleanup fixed, but not yet enabled (614157)
6581 - fixed infinite server hang, should a client not respond to an AST
6582 - avoid going into recovery if user calls readlink() with a buffer
6583 that's too small (613941)
6584 - AST RPCs no longer require replies (614867) -- this may be changed
6585 - don't crash server if client sends an IOV that's too big (611336)
6586 - fixed lock conversion deadlock (611892)
6587 - fixed the following of symlinks (614622)
6588 * recovery: the server can remove locks from a client that dies, other
6589 clients can make progress
6590 * more extN patch fixes
6591 * compile-time configurable ptlrpc buffer allocations
6593 - collaborative read cache document
6594 - Lustre Lite Performance CDR document-in-progress
6596 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6599 - fix typo in patch-2.4.18
6601 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6604 - clear ptlrpc request each time in handle_incoming_request()
6605 - unlink of files now destroys the object on the OST
6607 2002-09-19 Peter Braam <braam@clusterfs.com>
6609 * add hard link support
6610 * change obdfile creation method
6611 * kernel patch changed
6613 2002-09-19 Peter Braam <braam@clusterfs.com>
6616 - stack overflow bug in extN fixed
6618 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6620 * documentation updates
6621 - add man pages for config tools
6622 - update tests/README to describe testing with new config tools
6623 - finish metadata API descriptions
6624 * bug fixes and cleanups
6625 - statfs workaround for 16TB limit
6626 - LOV stripe allocation improved, can stripe on subset of OSTs
6627 - LOV file size/IO offset was wrong for files > 4GB in size
6628 - object EA data was being dropped, caused files to be unreadable
6629 - memory overflow with non-LOV OST caused memory corruption
6630 - fixed regression tests to work with new config tools, obdfilter
6631 - fixed bug when directory size became larger than 1 block
6632 - fixed bug (for single client case) when PWD was deleted
6633 - invalidate local directory pages when doing intent-based ops
6634 - avoid LDLM oops when lock callback contained bad data
6636 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6638 * documentation updates
6639 * bug fixes and cleanups
6640 - configuration tools
6643 - 64-bit compile warnings
6644 - 64-bit internal statfs data
6646 * test_brw on persistent OST devices
6648 * lprocfs (disabled)
6650 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6652 * documentation updates
6653 * bug fixes and cleanups
6654 * configuration tools
6656 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6659 * many small fixes to 0_5_4
6660 * io/network handling
6661 * thinkos in MDS operations
6663 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6666 * crucial basic fixes to 0.5.3
6667 * IOR, Iozone work over Elan
6670 2002-08-07 Phil Schwan <phil@clusterfs.com>
6671 * version 0_5_3, our first alpha
6672 * we use the new Portals iovs
6673 * documentation updates
6674 * bug fixes and cleanups
6675 * small changes in the DLM wire protocol
6677 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6678 * version 0_5_1 with some initial stability,
6679 * locking on MD and file I/O.
6680 * documentation updates
6681 * several bug fixes since 0.5.0
6682 * small changes in wire protocol
6684 2002-07-18 Phil Schwan <phil@clusterfs.com>
6686 * delivered as Lustre Light Alpha
6687 * fixed a crash after handling invalid MDS requests
6688 * fixed directory pages for architectures with non-4k pages sizes
6690 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6691 * release version v0_4_4
6692 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6694 * Fixes a number of interruption problems with OST operations.
6695 * Update documentation for portals header changes
6696 * Move all wire protocol structs/defines to lustre_idl.h
6697 * Fixes symlink length bug.
6698 * Add tcpdump to repository.
6700 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6701 * release version v0_4_3
6702 * Fixes statfs for inodes on extN.
6703 * Fixes bug in runtests which would delete /etc/hosts.
6704 * Use 64-bit object IDs wherever possible (not into VFS though)
6705 Remove ost_get_info, which is unused by lustre, and out of date.
6707 2002-07-03 Peter Braam <braam@clusterfs.com>
6708 * release version v0_4_2 Fixes a lookup error (type not passed)
6709 * move forward to head of Portals
6710 * move forward to latest Lustre kernel
6712 2002-06-25 Peter Braam <braam@clusterfs.com>
6713 * release version v0_4_1. Hopefully stable on single node use.