3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.23.1.el6 (RHEL6.9)
11 3.10.0-693.21.1.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.126-94.22 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.23.1.el6 (RHEL6.9)
24 3.10.0-693.21.1.el7 (RHEL7.4)
25 3.0.101-0.47.71 (SLES11 SP3)
26 3.0.101-107 (SLES11 SP4)
27 3.12.74-60.64.40 (SLES12 SP1)
28 4.4.120-92.70 (SLES12 SP2)
29 4.4.126-94.22 (SLES12 SP3)
31 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
32 * Recommended ZFS / SPL version: 0.7.9
33 * Tested with ZFS / SPL version: 0.7.9
34 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
35 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
36 stack overflow. For more information, please refer to bugzilla 17630.
37 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
38 the re-exporting nodes to be mounted with "32bitapi" mount option
40 01-31-2018 Intel Corporation
42 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
43 for currently supported client and server kernel versions.
44 * Server known to build on patched kernels:
45 2.6.32-431.29.2.el6 (RHEL6.5)
46 2.6.32-504.30.3.el6 (RHEL6.6)
47 2.6.32-573.26.1.el6 (RHEL6.7)
48 2.6.32-642.15.1.el6 (RHEL6.8)
49 2.6.32-696.23.1.el6 (RHEL6.9)
50 3.10.0-693.21.1.el7 (RHEL7.4)
51 3.0.101-0.47.71 (SLES11 SP3)
52 3.0.101-107 (SLES11 SP4)
53 3.12.74-60.64.40 (SLES12 SP1)
54 4.4.120-92.70 (SLES12 SP2)
55 4.4.120-94.17 (SLES12 SP3)
56 vanilla linux 4.6.7 (ZFS only)
57 * Client known to build on unpatched kernels:
58 2.6.32-431.29.2.el6 (RHEL6.5)
59 2.6.32-504.30.3.el6 (RHEL6.6)
60 2.6.32-573.26.1.el6 (RHEL6.7)
61 2.6.32-642.15.1.el6 (RHEL6.8)
62 2.6.32-696.23.1.el6 (RHEL6.9)
63 3.10.0-693.21.1.el7 (RHEL7.4)
64 3.0.101-0.47.71 (SLES11 SP3)
65 3.0.101-107 (SLES11 SP4)
66 3.12.74-60.64.40 (SLES12 SP1)
67 4.4.120-92.70 (SLES12 SP2)
68 4.4.120-94.17 (SLES12 SP3)
69 4.4.0-116 (Ubuntu 16.04)
71 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
72 * Recommended ZFS / SPL version: 0.7.5
73 * Tested with ZFS / SPL version: 0.7.5
74 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
75 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
76 stack overflow. For more information, please refer to bugzilla 17630.
77 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
78 the re-exporting nodes to be mounted with "32bitapi" mount option
80 --------------------------------------------------------------------------------
82 12-11-2017 Intel Corporation
84 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
85 for currently supported client and server kernel versions.
86 * Server known to build on patched kernels:
87 2.6.32-431.29.2.el6 (RHEL6.5)
88 2.6.32-504.30.3.el6 (RHEL6.6)
89 2.6.32-573.26.1.el6 (RHEL6.7)
90 2.6.32-642.15.1.el6 (RHEL6.8)
91 2.6.32-696.20.1.el6 (RHEL6.9)
92 3.10.0-693.17.1.el7 (RHEL7.4)
93 3.0.101-0.47.71 (SLES11 SP3)
94 3.0.101-107 (SLES11 SP4)
95 3.12.74-60.64.40 (SLES12 SP1)
96 4.4.103-92.56 (SLES12 SP2)
97 4.4.103-6.38 (SLES12 SP3)
98 vanilla linux 4.6.7 (ZFS only)
99 * Client known to build on unpatched kernels:
100 2.6.32-431.29.2.el6 (RHEL6.5)
101 2.6.32-504.30.3.el6 (RHEL6.6)
102 2.6.32-573.26.1.el6 (RHEL6.7)
103 2.6.32-642.15.1.el6 (RHEL6.8)
104 2.6.32-696.20.1.el6 (RHEL6.9)
105 3.10.0-693.17.1.el7 (RHEL7.4)
106 3.0.101-0.47.71 (SLES11 SP3)
107 3.0.101-107 (SLES11 SP4)
108 3.12.74-60.64.40 (SLES12 SP1)
109 4.4.103-92.56 (SLES12 SP2)
110 4.4.103-6.38 (SLES12 SP3)
112 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
113 * Recommended ZFS / SPL version: 0.7.3
114 * Tested with ZFS / SPL version: 0.7.3
115 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
116 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
117 stack overflow. For more information, please refer to bugzilla 17630.
118 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
119 the re-exporting nodes to be mounted with "32bitapi" mount option
121 --------------------------------------------------------------------------------
123 10-02-2017 Intel Corporation
125 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
126 for currently supported client and server kernel versions.
127 * Server known to build on patched kernels:
128 2.6.32-431.29.2.el6 (RHEL6.5)
129 2.6.32-504.30.3.el6 (RHEL6.6)
130 2.6.32-573.26.1.el6 (RHEL6.7)
131 2.6.32-642.15.1.el6 (RHEL6.8)
132 2.6.32-696.10.3.el6 (RHEL6.9)
133 3.10.0-693.2.2.el7 (RHEL7.4)
134 3.0.101-0.47.71 (SLES11 SP3)
135 3.0.101-107 (SLES11 SP4)
136 3.12.74-60.64.40 (SLES12 SP1)
137 4.4.74-92.35 (SLES12 SP2)
138 vanilla linux 4.6.7 (ZFS only)
139 * Client known to build on unpatched kernels:
140 2.6.32-431.29.2.el6 (RHEL6.5)
141 2.6.32-504.30.3.el6 (RHEL6.6)
142 2.6.32-573.26.1.el6 (RHEL6.7)
143 2.6.32-642.15.1.el6 (RHEL6.8)
144 2.6.32-696.10.3.el6 (RHEL6.9)
145 3.10.0-693.2.2.el7 (RHEL7.4)
146 3.0.101-0.47.71 (SLES11 SP3)
147 3.0.101-107 (SLES11 SP4)
148 3.12.74-60.64.40 (SLES12 SP1)
149 4.4.74-92.35 (SLES12 SP2)
151 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
152 * Recommended ZFS / SPL version: 0.7.1
153 * Tested with ZFS / SPL version: 0.7.1
154 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
155 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
156 stack overflow. For more information, please refer to bugzilla 17630.
157 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
158 the re-exporting nodes to be mounted with "32bitapi" mount option
160 --------------------------------------------------------------------------------
162 2017-07-05 Intel Corporation
164 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
165 for currently supported client and server kernel versions.
166 * Server known to build on patched kernels:
167 2.6.32-431.29.2.el6 (RHEL6.5)
168 2.6.32-504.30.3.el6 (RHEL6.6)
169 2.6.32-573.26.1.el6 (RHEL6.7)
170 2.6.32-642.15.1.el6 (RHEL6.8)
171 2.6.32-696.3.1.el6 (RHEL6.9)
172 3.10.0-514.21.1.el7 (RHEL7.3)
173 3.0.101-0.47.71 (SLES11 SP3)
174 3.0.101-100 (SLES11 SP4)
175 3.12.74-60.64.40 (SLES12 SP1)
176 4.4.59-92.17 (SLES12 SP2)
177 vanilla linux 4.6.7 (ZFS only)
178 * Client known to build on unpatched 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 2.6.32-642.15.1.el6 (RHEL6.8)
183 2.6.32-696.3.1.el6 (RHEL6.9)
184 3.10.0-514.21.1.el7 (RHEL7.3)
185 3.0.101-0.47.71 (SLES11 SP3)
186 3.0.101-100 (SLES11 SP4)
187 3.12.74-60.64.40 (SLES12 SP1)
188 4.4.59-92.17 (SLES12 SP2)
190 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
191 * Recommended ZFS / SPL version: 0.6.5.9
192 * Tested with ZFS / SPL version: 0.6.5.9
193 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
194 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
195 stack overflow. For more information, please refer to bugzilla 17630.
196 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
197 the re-exporting nodes to be mounted with "32bitapi" mount option
199 --------------------------------------------------------------------------------
201 12-07-2016 Intel Corporation
203 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
204 for currently supported client and server kernel versions.
205 * Server known to build on patched kernels:
206 2.6.32-431.29.2.el6 (RHEL6.5)
207 2.6.32-504.30.3.el6 (RHEL6.6)
208 2.6.32-573.26.1.el6 (RHEL6.7)
209 2.6.32-642.11.1.el6 (RHEL6.8)
210 3.10.0-514.el7 (RHEL7.3)
211 3.0.101-0.47.71 (SLES11 SP3)
212 3.0.101-80 (SLES11 SP4)
213 3.12.62-60.64.8 (SLES12 SP1)
214 vanilla linux 4.5.7 (ZFS only)
215 * Client known to build on unpatched kernels:
216 2.6.32-431.29.2.el6 (RHEL6.5)
217 2.6.32-504.30.3.el6 (RHEL6.6)
218 2.6.32-573.26.1.el6 (RHEL6.7)
219 2.6.32-642.11.1.el6 (RHEL6.8)
220 3.10.0-514.el7 (RHEL7.3)
221 3.0.101-0.47.71 (SLES11 SP3)
222 3.0.101-80 (SLES11 SP4)
223 3.12.62-60.64.8 (SLES12 SP1)
225 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
226 * Recommended ZFS / SPL version: 0.6.5.8
227 * Tested with ZFS / SPL version: 0.6.5.8
228 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
229 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
230 stack overflow. For more information, please refer to bugzilla 17630.
231 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
232 the re-exporting nodes to be mounted with "32bitapi" mount option
234 --------------------------------------------------------------------------------
236 02-29-2016 Intel Corporation
238 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
239 for currently supported client and server kernel versions.
240 * Server known to build on patched kernels:
241 2.6.32-431.29.2.el6 (RHEL6.5)
242 2.6.32-504.30.3.el6 (RHEL6.6)
243 2.6.32-573.12.1.el6 (RHEL6.7)
244 3.10.0-327.3.1.el7 (RHEL7.2)
245 3.0.101-0.47.71 (SLES11 SP3)
246 3.0.101-68 (SLES11 SP4)
247 vanilla linux 4.2.1 (ZFS only)
248 * Client known to build on unpatched kernels:
249 2.6.32-431.29.2.el6 (RHEL6.5)
250 2.6.32-504.30.3.el6 (RHEL6.6)
251 2.6.32-573.12.1.el6 (RHEL6.7)
252 3.10.0-327.3.1.el7 (RHEL7.2)
253 3.0.101-0.47.71 (SLES11 SP3)
254 3.0.101-68 (SLES11 SP4)
257 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
258 * Recommended ZFS / SPL version: 0.6.4.2
259 * Tested with ZFS / SPL version: 0.6.4.2
260 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
261 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
262 stack overflow. For more information, please refer to bugzilla 17630.
263 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
264 the re-exporting nodes to be mounted with "32bitapi" mount option
266 --------------------------------------------------------------------------------
268 03-10-2015 Intel Corporation
270 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
271 for currently supported client and server kernel versions.
272 * Server known to build on patched kernels:
273 2.6.32-431.29.2.el6 (RHEL6.5)
274 2.6.32-504.8.1.el6 (RHEL6.6)
275 3.0.101-0.46 (SLES11 SP3)
276 * Client known to build on unpatched kernels:
277 2.6.32-431.29.2.el6 (RHEL6.5)
278 2.6.32-504.8.1.el6 (RHEL6.6)
279 3.10.0-123.20.1.el7 (RHEL7)
280 3.0.101-0.46 (SLES11 SP3)
281 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
282 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
283 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
284 stack overflow. For more information, please refer to bugzilla 17630.
285 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
286 the re-exporting nodes to be mounted with "32bitapi" mount option
288 Severity : enhancement
290 Description: control OST-index in IDIF via ROCOMPAT flag.
291 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
292 last_rcvd file. For new formatted OST device, it will be auto set;
293 for the case of upgrading from old OST device, you can enable it
294 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
295 enabled, for new created OST-object, its IDIF-in-LMA will contain
296 the OST-index; for the existing OST-object, the OSD will convert
297 old format IDIF as new format IDIF with OST-index stored in the
298 LMA EA when accessing such OST-object or via OI scrub. Once such
299 flag is enabled, it cannot be reverted back, so the system cannot
300 be downgraded to the orignal incompatible version.
302 --------------------------------------------------------------------------------
304 07-30-2014 Intel Corporation
306 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
307 for currently supported client and server kernel versions.
308 * Server known to build on patched kernels:
309 2.6.32-431.20.3.el6 (RHEL6.5)
310 3.0.101-0.31 (SLES11 SP3)
311 * Client known to build on unpatched kernels:
312 2.6.32-431.20.3.el6 (RHEL6.5)
313 3.0.101-0.31 (SLES11 SP3)
316 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
317 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
318 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
319 stack overflow. For more information, please refer to bugzilla 17630.
320 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
321 the re-exporting nodes to be mounted with "32bitapi" mount option
323 --------------------------------------------------------------------------------
325 10-31-2013 Intel Corporation
327 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
328 * Server builds for kernels:
329 2.6.32-358.18.1.el6 (RHEL6.4)
330 * Client builds for unpatched kernels:
331 2.6.32-358.18.1.el6 (RHEL6.4)
332 3.0.92-0.8 (SLES11 SP3)
334 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
335 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
336 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
337 stack overflow. For more information, please refer to bugzilla 17630.
338 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
339 the re-exporting nodes to be mounted with "32bitapi" mount option
341 --------------------------------------------------------------------------------
343 05-30-2013 Intel Corporation
345 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
346 * Server builds for kernels:
347 2.6.32-358.6.2.el6 (RHEL6.4)
348 * Client builds for unpatched kernels:
349 2.6.32-358.6.2.el6 (RHEL6.4)
350 3.0.74-0.6.6 (SLES11 SP2)
352 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
353 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
354 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
355 stack overflow. For more information, please refer to bugzilla 17630.
356 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
357 the re-exporting nodes to be mounted with "32bitapi" mount option
359 --------------------------------------------------------------------------------
361 9-30-2012 Whamcloud, Inc.
363 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
364 * Server support for kernels:
365 2.6.32-279.5.1.el6 (RHEL6)
366 * Client support for unpatched kernels:
367 2.6.32-279.5.1.el6 (RHEL6)
369 2.6.18-238.19.1.el5 (RHEL5)
370 2.6.16.60-0.69.1 (SLES10)
371 * Recommended e2fsprogs version: 1.41.90.wc4
372 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
373 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
374 stack overflow. For more information, please refer to bugzilla 17630.
375 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
376 the re-exporting nodes to be mounted with "32bitapi" mount option
378 Severity : enhancement
380 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
382 Severity : enhancement
384 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
386 Severity : enhancement
388 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
390 Severity : enhancement
392 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
394 Severity : enhancement
396 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
398 --------------------------------------------------------------------------------
400 04-30-2012 Whamcloud, Inc.
402 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
403 * Server support for kernels:
404 2.6.32-220.4.2.el6 (RHEL6)
405 * Client support for unpatched kernels:
406 2.6.32-220.4.2.el6 (RHEL6)
408 2.6.18-238.19.1.el5 (RHEL5)
409 2.6.16.60-0.69.1 (SLES10)
410 * Recommended e2fsprogs version: 1.41.90.wc4
411 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
412 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
413 stack overflow. For more information, please refer to bugzilla 17630.
414 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
415 the re-exporting nodes to be mounted with "32bitapi" mount option
417 Severity : enhancement
419 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
421 Severity : enhancement
423 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
425 Severity : enhancement
427 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
429 Severity : enhancement
431 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
433 --------------------------------------------------------------------------------
435 09-30-2011 Whamcloud, Inc.
437 * Server support for kernels:
438 2.6.18-238.19.1.el5 (RHEL5)
439 2.6.32-131.6.1.el6 (RHEL6)
440 * Client support for unpatched kernels:
441 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
442 2.6.18-238.19.1.el5 (RHEL5)
443 2.6.32-131.6.1.el6 (RHEL6)
444 2.6.29.4-167.fc11 (FC11)
445 2.6.16.60-0.69.1 (SLES10)
447 * Recommended e2fsprogs version: 1.41.90.wc3
448 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
449 Lustre filesystem with 4K stack may cause a stack overflow. For more
450 information, please refer to bugzilla 17630.
451 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
452 mounted with "32bitapi" mount option
454 Severity : enhancement
455 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
457 Severity : enhancement
458 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
460 Severity : enhancement
461 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
464 Frequency : only with SLES10
466 Description: Use OFED "KMP" provided by Novell
467 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
468 Lustre is now built against this package. That means you need to
469 install the ofed-kmp package from Novell for the patchless client.
470 Note that the ofed-kmp that Novell ships may not exactly match the
471 kernel version but should still be compatible.
473 Severity : enhancement
475 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
476 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
478 --------------------------------------------------------------------------------
480 2010-07-15 Oracle, Inc.
482 * Server support for kernels:
483 2.6.18-164.11.1.el5 (RHEL 5)
484 2.6.18-164.11.1.0.1.el5 (OEL 5)
485 * Client support for unpatched kernels:
486 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
487 2.6.18-164.11.1.el5 (RHEL 5),
488 2.6.18-164.11.1.0.1.el5 (OEL 5)
489 2.6.16.60-0.42.8 (SLES 10),
491 2.6.29.4-167.fc11 (FC11)
492 * Recommended e2fsprogs version: 1.41.10-sun2
494 Severity : enhancement
496 Description: Distribute lfs_migrate script with lustre
497 Details : The lfs_migrate mentioned in the manual is now included in the
498 lustre tree in lustre/scripts and will distributed in lustre RPMs.
500 Severity : enhancement
502 Description: Failover nid list added to import proc file.
504 Severity : enhancement
506 Description: "som_preview" mount option required to use SOM
507 Details : The clients will need to use som_preview mount option
511 Description: Lustre does not do 1MB IOs to HW RAID
512 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
514 Severity : enhancement
516 Description: lfs quota output is not very convenient for awk/sed-parsing
517 Details : Some positions in lfs quota output table could be empty or
518 non-empty which made it hard to parse it with scripts, now
519 a dash is put instead of space where there is not supposed
524 Description: Update to OFED 1.5.1
525 Details : OFED 1.5.1 will be used in environments that don't use
526 vendor supplied OFED.
530 Description: Remove lnet. prefix from lctl params display
531 Details : Also renamed lnet.memused to lnet_memused to
536 Description: Use conf_param -d to remove permanant settings.
540 Description: Add list_param -R
541 Details : lctl list_param now supports recursive mode.
544 Bugzilla : 21829, 22312
545 Description: Fix llobdstat.
546 Details : 1. Now updating the write_bytes counter
547 2. Port script fixes from 1.8.
551 Description: decrease the usage of memory on clients.
552 Details : 1. On clients, recycle dentries and inodes unused.
553 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
558 Description: llobdstat fix and enhancement
559 Details : add a counter to set a limit to how many samples will be returned
560 fix a wildcard in the path to limit to obdfilter stats only
564 Description: lfs check only allowed for root
565 Details : Code cleanups and added a regresion test of non-root use of
570 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
571 Details : Fix hr->hr_index overload.
575 Description: Fix FIEMAP functionality.
579 Description: FC11 patchless client support added
583 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
584 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
588 Description: New acc-sm script and various test-framework enhancments.
592 Description: Previously landed patch was reverted.
596 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
600 Description: Add ldiskfs maxdirsize mount option.
604 Description: Fix mkfs.lustre support for 16TB LUNs.
608 Description: Print mmp_check_interval
609 Details : print mmp_check_interval and make it possible to abort
610 mount operation in case it takes too long.
614 Description: Add cascading_rw to lustre mpi tests.
618 Description: Store SOM data in EA.
619 Details : Also several SOM related cleanups.
623 Description: Add refcount to md_open_data
627 Description: Use refcount on lov_request_set to prevent memory corruption.
631 Description: Statahead: use parent dir's i_mutex for synchronization.
635 Description: loadgen improvements.
639 Description: Enable peer health detection in ksocklnd and ko2iblnd.
643 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
647 Description: Fix race on obd_devcie:md_stats
651 Description: Fix SLES 11 client builds.
652 Details : Resolved uninitalized pointer gcc warnings.
656 Description: Allow non-root access for "lfs check".
657 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
661 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
663 Severity : enhancement
665 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
666 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
670 Description: Sanity test_27p fails.
671 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
672 Sanity tests 27 were cleaned up and improved.
676 Description: Unify req format on client/servers
677 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
678 buffers and swabbers approach..
683 Description: drop join file stuff.
685 Severity : enhancement
687 Description: Add 2.6.27 and SLES11 (patchless client) support
691 Description: Fix in ptlrpc_expire_one_request() to print the signed time
694 Severity : enhancement
696 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
699 Severity : enhancement
701 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
703 Severity : enhancement
705 Description: correctly handle big reply message.
706 Details : send LNet event if reply is bigger than buffer and adjust this buffer
711 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
713 Severity : enhancement
715 Description: Add support for OEL5.
719 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
721 Severity : enhancement
723 Description: "weak-modules" support
724 Details : Implement "weak-modules" support which enables kernel modules
725 to be used with any kernel that implements the same kABI. In
726 order to achieve this modules are now installed in
727 /lib/modules/$(uname -r)/updates/kernel on all distributions.
731 Description: client could not reconnect to OST because of an active request.
732 Details : The client now retries to reconnect to the same server, if a
733 connect request failed with EBUSY or -EAGAIN. Test case 35b
734 conf-sanity.sh has been created.
738 Description: Don't allow make backward step on assignin osc next id.
739 Details : race between allocation next id and ll_sync thread can be cause
740 of set wrong osc next id and can be kill valid ost objects.
744 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
748 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
752 Description: Error handling in osc_statfs_interpret() has been improved.
753 Details : Check in osc_statfs_interpret() for EBADR.
757 Description: Do not send statfs() requests to OSTs disabled by administrator.
758 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
760 Severity : enhancement
762 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
764 Severity : enhancement
766 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
768 Severity : enhancement
770 Description: Enhancement for lfs(1) command to use numeric uid/gid.
774 Description: Improved cleanup procedure for conf-sanity test_46a.
778 Description: Do not start bulk transfer, if client tried to reconnect.
779 Details : Do not start bulk transfer, if client tried to reconnect,
780 but failed due to active rpcs. Check for exp_abort_active_req
781 has been implemented.
785 Description: Prevent inconsistences between linux and lustre mount structures.
786 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
787 Make the sleep interruptible.
789 Severity : enhancement
791 Description: provide server to client comms path
792 Details : server to client communications path via new LDLM_SET_INFO rpc
796 Description: truncate starts GFP_FS allocation under transaction causing deadlock
797 Details : ldiskfs_truncate calls grab_cache_page which may start page
798 allocation under an open transaction. This may lead to
799 calling prune_icache with consequent lustre reentrance.
803 Description: send correctly lsm on open replay
804 Details : MDS is trust to LSM size on replay open, but client can set wrong size
809 Description: enable client interface failover
810 Details : When a child reconnects from another NID, properly update export
811 nid hash position and ldlm reverse import.
813 Severity : enhancement
815 Description: Communicate OST degraded/readonly state via statfs to MDS
816 Details : Flags in the statfs returned from OSTs indicate whether the
817 OST is in a degraded RAID state, or if the filesystem has
818 turned read-only after a filesystem error is detected.
823 Description: don't panic if EPROTO was hit when reading symlink
824 Details : correctly handling request reference in error cases.
826 Severity : enhancement
828 Description: Add LustreNetLink, a kernel-userspace communcation path.
830 Severity : enhancement
832 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
837 Description: access to llog context before init.
838 Details : move handling CATALOGS file at osc layer and forbid access to llog
842 Frequency : rare, only when client is writing to OST on local node
844 Description: Avoid deadlock for local client writes
845 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
846 memory freeing context. This allows OST threads to set the
847 PF_MEMALLOC flag on task structures in order to allocate memory
848 from reserved pools and complete IO.
849 Use GFP_HIGHUSER for OST allocations for non-local client writes,
850 so that the OST threads generate memory pressure and allow
851 inactive pages to be reclaimed.
853 Severity : enhancement
855 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
860 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
861 Details : move ldlm namespace creation in setup phase to avoid grab
862 _lprocfs_lock with cli_sem held.
866 Description: Temporarily disable grant shrink.
867 Details : Disable the feature for debugging.
870 Frequency : only when formatting test filesystems
872 Description: Unable to run several mkfs.lustre on loop devices at the same time
873 Details : mkfs.lustre returns error 256 on the concurrent loop devices
874 formatting. The solution is to proper handle the error.
876 Severity : enhancement
878 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
880 Severity : enhancement
882 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
887 Description: resolve race between obd_disconnect and class_disconnect_exports
888 Details : if obd_disconnect will be called to already disconnected export he
889 forget release one reference and osc module can't unloaded.
891 Severity : enhancement
893 Description: Allow tuning service thread via /proc
894 Details : For each service a new
895 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
896 created that can be used to set min/max thread counts, and get the
897 current number of running threads.
902 Descriptoin: don't return error if have partially created objects for file.
903 Details : lov_update_create_set uses set->set_success as index for created
904 objects, so if some requests will be failed, they will have hole
905 at end of array and qos_shrink_lsm for allocate correct lsm.
907 Severity : enhancement
909 Description: Update OFED support to 1.4.1
914 Description: Reduce small size read RPC
915 Details : Set read-ahead limit for every file and only do read-ahead when
916 available read-ahead pages are bigger than 1M to avoid small size
919 Severity : enhancement
921 Description: move AT tunable parameters for more consistent usage
922 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
924 Severity : enhancement
926 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
927 Details : allow skip disconnected ost for send statfs request and hide error
933 Description: fix racy locking of mballoc block bitmaps causing BUG
934 Details : The locking of the mballoc buddy bitmap and the in-memory
935 block bitmap was using two different spin locks in some
936 cases. This made it possible to incorrectly access the
937 mballoc bitmap while another process was modifying it,
938 causing a sanity assertion to fail. While no on-disk corruption
939 was reported, there was some risk of this happening.
942 Frequency : rare, on llog test 6
944 Descriptoin: don't allow connect to already connected import
945 Details : allowing connect to already connected import is hide connecting problem.
950 Descriptoin: allow kill process which wait statahead result
951 Details : for some reasons 'ls' can stick in waiting result from statahead,
952 in this case need way for kill this process.
954 Severity : enhancement
956 Description: Add state history info file, enhance import info file
957 Details : Track import connection state changes in a new osc/mdc proc file;
958 add overview-type data to the osc/mdc import proc file.
960 Severity : enhancement
962 Description: MDS create should not wait for statfs RPC while holding DLM lock.
964 Severity : enhancement
966 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
971 Descriptoin: OST grant too much space to client even there are not enough space.
972 Details : Client will shrink its grant cache to OST if there are no write
973 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
974 this grant cache if there are already not enough avaible space
975 (left_space < total_clients * 32M).
978 Frequency : start MDS on uncleanly shutdowned MDS device
980 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
981 Details : stay in waiting mds<>ost recovery finished produce random bugs
982 due race between two ll_sync thread for one lov target. send
983 ACTIVATE event only if connect realy finished and import have
987 Frequency : rare, connect and disconnect target at same time
989 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
990 Details : don't call obd_disconnect under lov_lock. this long time
991 operation and can block ptlrpcd which answer to connect request.
996 Descriptoin: don't lose wakeup for imp_recovery_waitq
997 Details : recover_import_no_retry or invalidate_import and import_close can
998 both sleep on imp_recovery_waitq, but we was send only one wakeup
1002 Frequency : always with long access acl
1004 Descriptoin: mds can't pack reply with long acl.
1005 Details : mds don't control size of acl but they limited by reint/getattr
1008 Severity : enhancement
1010 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1012 Severity : enhancement
1014 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1017 Frequency : start MDS on uncleanly shutdowned MDS device
1019 Descriptoin: aborting recovery hang on MDS
1020 Details : don't throttle destroy RPCs for the MDT.
1023 Frequency : on remount
1025 Description: external journal device not working after the remount
1026 Details : clear dev_rdonly flag for external journal devices in
1032 Description: shutdown vs evict race
1033 Details : client_disconnect_export vs connect request race.
1034 if client will evicted at this time - we start invalidate
1035 thread without referece to import and import can be freed
1039 Frequency : rare, need acl's on inode.
1041 Description: client can't handle ost additional correctly
1042 Details : if ost was added after client connected to mds client can have
1043 hit lnet_try_match_md ... to big messages to wide striped files.
1044 in this case need teach client to handle config events about add
1045 lov target and update client max ea size at that event.
1047 Severity : enhancement
1049 Description: Changelogs
1050 Details : Changelogs are a lightweight mechanism to track filesystem
1051 metadata and namespace changes. The changelog is recorded
1052 permanently on the MDTs, and is periodically "consumed" / purged
1053 when records are no longer needed.
1055 Severity : enhancement
1057 Description: compact fld format with extents
1058 Details : Store range of seq rather than every seq in FLD. Seq
1059 controller update FLD rather than clients. In Case of CMD, mdt0
1060 has FLD, all other metadata server act as non persistent proxy
1061 for FLD queries and cache fld entries in fld cache.
1066 Description: don't skip ost target if they assigned to file
1067 Details : Drop slow OSCs if we can, but not for requested start idx.
1068 This means "if OSC is slow and it is not the requested
1069 start OST, then it can be skipped, otherwise skip it only
1070 if it is inactive/recovering/out-of-space.
1074 Description: more cleanup in mds_lov
1075 Details : not send LOV EA under replay, we can't know about they size at this
1076 time. Don't allow client connect to mds before any ost connected,
1077 for avoid problems with LOV EA size and returning EIO to client.
1079 Severity : enhancement
1081 Description: Interoperability at server side (Disk interoperability)
1083 Severity : enhancement
1085 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1087 Severity : enhancement
1089 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1091 Severity : enhancement
1093 Description: New client IO stack (CLIO).
1095 Severity : enhancement
1097 Description: Commit on sharing. Eliminate inter-client dependencies between
1098 uncommitted transactions by doing transaction commits.
1099 Thereby clients may recovery independently.
1102 Frequency : Create a symlink file with a very long name
1104 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1105 Details : If there is no extra space in the request for early cancels,
1106 ldlm_req_handles_avail() returns 0 instead of a negative value.
1108 Severity : enhancement
1110 Description: Add /proc entry for import status
1111 Details : The mdc, osc, and mgc import directories now have
1112 an import directory that contains useful import data for debugging
1113 connection problems.
1115 Severity : enhancement
1117 Description: Re-disable certain /proc logging
1118 Details : Enable and disable client's offset_stats, extents_stats and
1119 extents_stats_per_process stats logging on the fly.
1122 Frequency : Only on FC kernels 2.6.22+
1124 Description: oops in statahead
1125 Details : Do not drop reference count for the dentry from VFS when lookup,
1126 VFS will do that by itself.
1128 Severity : enhancement
1130 Description: Generic /proc file permissions
1131 Details : Set /Proc file permissions in a more generic way to enable non-
1132 root users operate on some /proc files.
1136 Description: Hitting mdc_commit_close() ASSERTION
1137 Details : Properly handle request reference release in
1138 ll_release_openhandle().
1142 Description: quota recovery deadlock during mds failover
1143 Details : This patch includes att18982, att18236, att18237 in bz14840.
1145 1. fix osts hang when mds does failover with quotaon
1146 2. prevent watchdog storm when osts threads wait for the
1151 Frequency : only patchless client
1152 Description: add workaround for race between add/remove dentry from hash
1154 Severity : enhancement
1156 Description: Allow OST glimpses to return PW locks
1160 Description: LBUG when llog conf file is full
1161 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1165 Description: Prevent import from entering FULL state when server in recovery
1169 Description: service mount cannot take device name with ":"
1170 Details : Only when device name contains ":/" will mount treat it as
1176 Description: replace ptlrpcd with the statahead thread to interpret the async
1177 statahead RPC callback
1181 Frequency : on recovery
1182 Description: I/O failures after umount during fail back
1183 Details : if client reconnected to restarted server we need join to recovery
1184 instead of find server handler is changed and process self
1185 eviction with cancel all locks.
1187 Severity : enhancement
1189 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1191 Severity : enhancement
1193 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1195 Severity : enhancement
1197 Description: Upcall on Lustre log has been dumped
1198 Details : Allow for a user mode script to be called once a Lustre log has
1199 been dumped. It passes the filename of the dumped log to the
1200 script, the location of the script can be specified via
1201 /proc/sys/lnet/debug_log_upcall.
1206 Description: avoid idr_remove called for id which is not allocated.
1207 Details : Move assigment s_dev for clustered nfs to end of initialization,
1208 to avoid problem with error handling.
1213 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1214 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1215 into CONN_USED_HASH and this prodice warning when put connection
1216 again in unused hash.
1219 Severity : enhancement
1221 Description: File striping can now be set to use an arbitrary pool of OSTs.
1223 Severity : enhancement
1225 Description: Export bytes_read/bytes_write count on OSC/OST.
1229 Description: Early reply size mismatch, MGC loses connection
1230 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1231 the connect flags are properly negotiated.
1236 Description: quotas are not honored with O_DIRECT
1237 Details : all writes with the flag O_DIRECT will use grants which leads to
1238 this problem. Now using OBD_BRW_SYNC to guard this.
1242 Description: add quota statistics
1243 Details : 1. sort out quota proc entries and proc code.
1244 2. add quota statistics
1246 Severity : enhancement
1248 Description: enable quota support for HEAD.
1252 Description: Properly propagate oinfo flags from lov to osc for statfs
1253 Details : restore missing copy oi_flags to lov requests.
1255 Severity : enhancement
1257 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1258 : l_getgroups(8), lst(8), routerstat(8)
1259 Details : included man pages for llobdstat(8), llstat(8),
1260 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1262 Severity : enhancement
1264 Description: configure's --enable-quota should check the
1265 : kernel .config for CONFIG_QUOTA
1266 Details : configure is terminated if --enable-quota is passed but
1267 : no quota support is in kernel
1271 Description: Remove portals compatibility
1272 Details : Remove portals compatibility, not interoperable with releases
1277 Description: Resolve device initialization race
1278 Details : Prevent proc handler from accessing devices added to the
1279 obd_devs array but yet be intialized.
1281 Severity : enhancement
1283 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1285 Severity : enhancement
1287 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1292 Description: Using wrong pointer in osc_brw_prep_request
1293 Details : Access to array[-1] can produce panic if kernel compiled with
1294 CONFIG_PAGE_ALLOC enabled
1298 Description: Client runs out of low memory
1299 Details : Consider only lowmem when counting initial number of llap pages
1303 Description: *optional* service tags registration
1304 Details : if the "service tags" package is installed on a Lustre node
1305 When the filesystem is mounted, a local-node service tag will
1306 be created. See http://inventory.sun.com/ for more information
1307 about the Service Tags asset management system.
1311 Description: Kernel BUG tries to release flock
1312 Details : Lustre does not destroy flock lock before last reference goes
1313 away. So always drop flock locks when client is evicted and
1314 perform unlock regardless of successfulness of speaking to MDS.
1318 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1322 Description: sanity test 65a fails if stripecount of -1 is set
1323 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1328 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1329 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1332 Severity : enhancement
1334 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1338 Description: do not process already freed flock
1339 Details : flock can possibly be freed by another thread before it reaches
1340 to ldlm_flock_completion_ast.
1344 Description: LBUG during stress test
1345 Details : Need properly lock accesses the flock deadlock detection list.
1349 Description: oops in page fault handler
1350 Details : kernel page fault handler can return two special 'pages' in error
1351 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1355 Description: timeout with invalidate import.
1356 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1357 should be handled by ptlrpcd. This produce long age waiting and
1358 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1360 Severity : enhancement
1362 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1366 Description: enable MGS and MDT services start separately
1367 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1368 a co-located MGS without starting the MGS, which is a complement
1369 to 'nosvc' mount option.
1373 Frequency : after recovery
1374 Description: precreate to many object's after del orphan.
1375 Details : del orphan st in oscc last_id == next_id and this triger growing
1376 count of precreated objects. Set flag LOW to skip increase count
1377 of precreated objects.
1381 Frequency : rare, on clear nid stats
1382 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1383 Details : when clean nid stats sometimes try destroy live entry,
1384 and this produce panic in free.
1388 Description: Stack overflow during MDS log replay
1389 ease stack pressure by using a thread dealing llog_process.
1393 Description: wait until IO finished before start new when do lock cancel.
1394 Details : VM protocol want old IO finished before start new, in this case
1395 need wait until PG_writeback is cleared until check dirty flag and
1396 call writepages in lock cancel callback.
1398 Severity : enhancement
1400 Description: using special macro for print time and cleanup in includes.
1404 Description: mds_mfd_close() ASSERTION(rc == 0)
1405 Details : In mds_mfd_close(), we need protect inode's writecount change
1406 within its orphan write semaphore to prevent possible races.
1410 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1414 Frequency : rare, on shutdown ost
1415 Description: don't hit live lock with umount ost.
1416 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1417 use shrink_dcache_sb instead.
1421 Description: don't panic with use echo client
1422 Details : echo client pass NULL as client nid pointer and this produce null
1423 pointer dereference.
1427 Description: fix build on ppc32
1428 Details : compile code with -m64 flag produce wrong object file for ppc32.
1432 Description: add message levels for liblustreapi
1436 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1437 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1438 will result in an -EAGAIN and in the exit call path we call
1439 alloc_rr() which will with increasing aggressiveness attempt to
1440 aquire precreated objects on the minimum number of required OSCs.
1444 Description: Use old size assignment to avoid deadlock
1445 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1446 the scheduling while holding a spinlock. We do not need locking
1447 for size in ll_update_inode() because size is only updated from
1448 the MDS for directories or files without objects, so there is no
1449 other place to do the update, and concurrent access to such inodes
1450 are protected by the inode lock.
1454 Description: resolve "_IOWR redefined" build error on SLES10
1458 Description: dump the memory debugging after all modules are unloaded to
1459 suppress false negative in conf_sanity test 39
1461 Severity : enhancement
1463 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1468 Description: fix SLES kernel versioning
1469 Details : the kernel version for our SLES 10 kernel did not include a "-"
1470 before the "smp" at the end. while this was not a problem in
1471 general, it did mean that software trying to use the kernel
1472 version to try to detect a vendor specific kernel would fail.
1473 this was most evident by the OFED build scripts.
1477 Description: Don't update lov_desc members until making sure they are valid
1478 Details : When updating lov_desc members via proc fs, need fix their
1479 validities before doing the real update.
1483 Description: don't put request into delay list while invalidate in flight.
1484 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1485 import in flight. this produce timeout for invalidate and sometimes
1486 can cause stale data.
1490 Frequency : on ppc only
1491 Description: not convert ost objects for directory because it's not exist.
1492 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1496 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1498 Severity : enhancement
1500 Description: Update to support 2.6.22.14 vanilla kernel.
1504 Frequency : rare, on recovery
1505 Description: read procfs can produce deadlock in some situation
1506 Details : Holding lprocfs lock which send rpc can produce block for destroy
1507 obd objects and this also block reconnect with -EALREADY. This
1508 isn't fix all lprocfs bugs - but make it rare.
1510 Severity : enhancement
1512 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1515 Frequency : frequent on X2 node
1517 Description: mdc_set_open_replay_data LBUG
1518 Details : Set replay data for requests that are eligible for replay.
1522 Description: lustre_mgs: operation 101 on unconnected MGS
1523 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1524 MGC, and late on MGC cannot successfully connect to MGS and a lot
1525 of the error messages complaining that MGS is not connected.
1529 Frequency : on network error
1530 Description: panic with double free request if network error
1531 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1532 but it's true only for synchronus enqueue, for async enqueue
1533 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1536 Severity : enhancement
1538 Description: client-side metadata stat-ahead during readdir(directory readahead)
1539 Details : perform client-side metadata stat-ahead when the client detects
1540 readdir and sequential stat of dir entries therein
1543 Frequency : on start mds
1545 Description: Implement get_info(last_id) in obdfilter.
1548 Frequency : occasional
1550 Description: Correctly check stale fid, not start epoch if ost not support SOM
1551 Details : open with flag O_CREATE need set old fid in op_fid3 because
1552 op_fid2 was overwritten with new generated fid, but mds can answer
1553 with one of these two fids and both is not stale. Setattr
1554 incorrectly started an epoch and assume will be called
1555 done_writing, but without SOM done_writing ever being called.
1558 Frequency : rare, depends on device drivers and load
1560 Description: MDS or OSS nodes crash due to stack overflow
1561 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1562 In some cases, in conjunction with device drivers that use a lot
1563 of stack the MDS (or possibly OSS) service threads could overflow
1564 the stack. One change which was identified to consume additional
1565 stack has been reworked to avoid the extra stack usage.
1568 Frequency : occasional
1570 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1571 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1572 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1573 oscc_has_objects and see OSC has no objests and call
1574 oscc_internal_create to resend create request.
1576 Severity : enhancement
1578 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1580 Severity : enhancement
1582 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1585 Frequency : very rare
1587 Description: Fix replay if there is an un-replied request and open
1588 Details : In some cases, older replay request will revert the
1589 mcd->mcd_last_xid on MDS which is used to record the client's
1590 latest sent request.
1592 Severity : enhancement
1594 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1596 Severity : enhancement
1598 Description: Add rhel5 support to HEAD.
1600 Serverity : enhancement
1602 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1607 Description: Don't allow skipping OSTs if index has been specified.
1608 Details : Don't allow skipping OSTs if index has been specified, make
1609 locking in internal create lots better.
1613 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1614 Details : access to bitfield in structure is always rounded to long
1615 and this produce problem with not atomic change any bit.
1619 Description: Lustre make rpms failed.
1620 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1621 builds Lustre rpms from tarball.
1624 Frequency : rare on shutdown ost
1626 Description: If llog cancel was not send before clean_exports phase, this can
1627 produce deadlock in llog code.
1628 Details : If llog thread has last reference to obd and call class_import_put
1629 this produce deadlock because llog_cleanup_commit_master wait when
1630 last llog_commit_thread exited, but this never success because was
1631 called from llog_commit_thread.
1635 Description: allow userland application know is lost one of stripes.
1636 Details : fill lvb_blocks with error code on ost and return it to
1637 application if error flag found.
1641 Description: NULL lov_tgts causing MDS oops
1642 Details : more safe checks for NULL lov_tgts for avoid oops.
1644 Severity : enhancement
1646 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1650 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1651 Details : If an OST is down the MDS will hang indefinitely in
1652 obd_statfs_rqset() waiting for the statfs data. While for
1653 MDS QOS usage of statfs, it should not stuck in waiting.
1655 Severity : enhancement
1657 Description: remote_acl support
1658 Details : Support ACL-based permission check for remote user.
1659 Support setfacl/getfacl for remote user with the utils
1660 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1661 the system "{s,g}etfacl" utils.
1663 Severity : enhancement
1665 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1667 Severity : enhancement
1669 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1673 Description: make mgs_setparam() handle fsname containing dash
1674 Details : fsname containing a dash does not work with lctl conf_param
1676 Severity : enhancement
1678 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1680 Severity : enhancement
1682 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1686 Frequency : rare, at shutdown
1687 Description: access already free / zero obd_namespace.
1688 Details : if client_disconnect_export was called without force flag set,
1689 and exist connect request in flight, this can produce access to
1690 NULL pointer (or already free pointer) when connect_interpret
1691 store ocd flags in obd_namespace.
1695 Frequency : only at startup
1696 Description: not alloc memory with spinlock held.
1697 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1698 if any spinlock held.
1700 Severity : enhancement
1702 Description: make lustre randomly fail allocating memory
1703 Details : Make lustre randomly failed allocating memory for testing purpose.
1705 Severity : enhancement
1707 Description: lost problems with lov objid file
1708 Details : Fixes some scability and access to not inited memory problems
1709 in work with lov objdid file.
1714 Description: lfs find does not continue on file error
1715 Details : Continue other files processing when a file/dir is absent.
1719 Description: Inconsistent usage of lustre_pack_reply()
1720 Details : Standardize the usage of lustre_pack_reply() such that it
1721 always generate a CERROR on failure.
1724 Frequency : occasional
1726 Description: MDS hang or stay in waiting lock
1727 Details : If client receive lock with CBPENDING flag ldlm need send lock
1728 cancel as separate rpc, to avoid situation when cancel request
1729 can't processed due all i/o threads stay in wait lock.
1733 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1737 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1741 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1745 Description: Update to RHEL4 latest kernel.
1747 Severity : enhancement
1749 Description: Build SLES10 patchless client fails
1750 Details : The configure was broken by run ./configure with
1751 --with-linux-obj=.... argument for patchless client. When the
1752 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1753 can't search header adequately. Use absolute path such as
1754 -I($LINUX)/include instead.
1758 Description: interrupt oig_wait produce painc on resend.
1759 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1760 private set, and this produce situation when rq_ptlrpcd_data not
1761 copyed to new allocated request and triggered LBUG on assert
1762 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1763 ptlrpcd set if request is changed and can be safety changed to use
1768 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1769 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1770 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1771 these macros and replace them with their difinitions instead.
1775 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1776 Details : Update which_patch & target file for SLES10 latest kernel.
1779 Description: Remove client patches from SLES 10 kernel.
1780 Details : This causes SLES 10 clients to behave as patchless clients
1781 even on a Lustre-patched (server) kernel.
1783 Severity : enhancement
1785 Description: self-adjustable client's lru lists
1786 Details : use adaptive algorithm for managing client cached locks lru
1787 lists according to current server load, other client's work
1788 pattern, memory activities, etc. Both, server and client
1789 side namespaces provide number of proc tunables for controlling
1792 Severity : enhancement
1794 Description: light-weight GSS support
1795 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1796 and reduce performance overhead.
1798 Severity : enhancement
1800 Description: Linux keyring support
1801 Details : Support using service of Linux keyring for Lustre GSS internal
1802 context refresh/cache mechanism.
1806 Description: Fix errors in lfs documentation
1807 Details : Fixes man pages
1811 Description: don't use GFP_* in generic Lustre code.
1812 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1816 Description: obdclass is limited by single OBD_ALLOC(idarray)
1817 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1821 Description: warnings with build patchless client with vanila 2.6.19 and up
1822 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1823 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1827 Frequency : Only for SLES
1828 Description: sanity_quota fail test_1
1829 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1830 file, which makes TSTID[2] inunique.
1834 Frequency : Only for SLES9
1835 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1836 Details : Update target/ChangeLog/which_patch .
1840 Frequency : Only for test_10 in sanity-quota.sh
1841 Description: a bug in quota test code
1842 Details : checking a test flag in a wrong place
1846 Frequency : Only for test_13 in sanity-quota.sh
1847 Description: a bug in quota test script
1848 Details : shouldn't check overfull allocated quota
1852 Description: jbd statistics
1853 Details : Port older jbd statistics patch for sles10
1855 Severity : enhancement
1857 Description: Add printing inode info into message about error in writepage.
1862 Description: reply_lock_interpret crash due to race with it and lock cancel.
1863 Details : Do not replay locks that are being cancelled. Do not reference
1864 locks by their address during replay, just by their handle.
1868 Frequency : When flocks are used.
1869 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1870 Details : Flock locks might destroy just granted lock if it could be merged
1871 with another existing flock, this is done in completion handler,
1872 so teach ldlm_cli_enquque_fini that this is a valid case for
1878 Description: Oops in read and write path when failing to allocate lock.
1879 Details : Check if lock allocation failed and return error back.
1883 Description: lstripe command fails for valid OST index
1884 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1885 instead of lov->desc.ld_active_tgt_count.
1889 Description: sanity.sh failed test 103
1890 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1895 Description: df doesn't work properly if diskfs blocksize != 4K
1896 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1900 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1901 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1903 Severity : enhancement
1905 Description: Adaptive timeouts
1906 Details : RPC timeouts adapt to changing server load and network
1907 conditions to reduce resend attempts and improve recovery time.
1911 Description: llapi_file_create() does not allow some changes
1912 Details : add llapi_file_open() that allows specifying the mode and
1913 open flags, and also returns an open file handle.
1917 Description: merge and cleanup kernel patches.
1918 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1922 Description: Add journal checksum support.(Kernel part)
1923 Details : The journal checksum feature adds two new flags i.e
1924 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1925 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1926 indicates that the commit block contains the checksum for
1927 the blocks described by the descriptor blocks. Now commit
1928 record can be sent to disk without waiting for descriptor
1929 blocks to be written to disk. This behavior is controlled
1930 using JBD2_FEATURE_ASYNC_COMMIT flag.
1934 Description: OSS needs mutliple precreate threads
1935 Details : Add ability to start more than one create thread per OSS.
1939 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1940 Details : execute lfs setstripe on client
1944 Description: mds_obd_create error creating tmp object
1945 Details : When the user sets quota on root, llog will be affected and can't
1946 create files and write files.
1949 Frequency : Always on ia64 patchless client, and possibly others.
1951 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1952 Details : This allows the patchless client to be loaded on architectures
1953 without this export.
1957 Description: RedHat Update kernel for RHEL5
1958 Details : Add the kernel config file for RHEL5.
1962 Description: RedHat Update kernel for RHEL5
1963 Details : Modify the kernel config file more closer RHEL5.
1967 Description: Build failure against Centos5 (RHEL5)
1968 Details : Define PAGE_SIZE when it isn't present.
1972 Description: test_7 on sanity-quota.sh can't be used on separate machines
1973 Details : add facet to handle it
1977 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1978 Details : with NFS, the anon dentry's parent was set to itself in
1979 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1980 corresponding dentry other than use rec->ur_name.
1982 Severity : enhancement
1984 Description: lfs setstripe enhancement
1985 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1989 Description: enable data checksumming by default
1990 Details : enable checksum by default, allow --disable-checksum
1991 configure option and "-o nochecksum" mount option. Checksums
1992 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1996 Description: lustre support for RHEL5
1997 Details : Add support for RHEL5.
2001 Description: Client eviction due to failover config
2002 Details : after a connection loss, the lustre client should attempt to
2003 reconnect to the last active server first before trying the
2004 other potential connections.
2008 Description: when mds and osts use different quota unit(32bit and 64bit),
2009 quota will be released repeatly.
2010 Details : void sending multiple quota reqs to mds, which will keep the
2011 status between the reqs.
2015 Description: osts not allocated evenly to files
2016 Details : change the condition to increase offset_idx
2019 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2021 Description: Data corruption for OSTs that are formatted larger than 2TB
2023 Details : When generating the bio request for lustre file writes the
2024 sector number would overflow a temporary variable before being
2025 used for the IO. The data reads correctly from Lustre (which
2026 will overflow in a similar manner) but other file data or
2027 filesystem metadata may be corrupted in some cases.
2031 Description: Tune the kernel for good SCSI performance.
2032 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2033 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2038 Description: rewrite ext2-derived code in obdclass/uuid.c
2039 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2040 from scratch preserving functionality.
2044 Description: "lfs find -obd UUID" prints directories
2045 Details : "lfs find -obd UUID" will return all directory names instead
2046 of just file names. It is incorrect because the directories
2047 do not reside on the OSTs.
2051 Description: use i_size_read and i_size_write in 2.6 port
2052 Details : replace inode->i_size access with i_size_read/write()
2055 Frequency : when removing large files
2057 Description: scheduling issue during removal of large Lustre files
2058 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2059 It causes scheduling issues when removing large files (17TB in the
2063 Frequency : only with liblustre clients on XT3
2065 Description: evictions taking too long
2066 Details : allow llrd to evict clients directly on OSTs
2069 Frequency : only on ppc
2071 Description: /proc/fs/lustre/devices broken on ppc
2072 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2073 We should make sure the type of 'index' is loff_t and then cast
2074 explicitly as needed below. Do not assign an explicitly cast
2078 Frequency : only for rhel5
2080 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2081 Details : Modify the target file & which_kernel.
2083 Severity : enhancement
2085 Description: omit set fsid for export NFS
2086 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2087 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2091 Frequency : Always, for kernels after 2.6.16
2092 Description: Fix warning idr_remove called for id=.. which is not allocated.
2093 Details : Last kernels save old s_dev before kill super and not allow
2094 to restore from callback - restore it before call kill_anon_super.
2098 Description: buffer overruns could theoretically occur
2099 Details : llapi_semantic_traverse() modifies the "path" argument by
2100 appending values to the end of the origin string, and a
2101 overrun may occur. Adding buffer overrun check in liblustreapi.
2105 Description: Fix error on 'ls .' at the top of the Lustre mount.
2106 Details : Don't revalidate dentry if it is a root dentry.
2110 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2111 Details : Modify vm-tunables-rhel4.patch.
2115 Description: Kernel config for 2.6.18-vanilla.
2116 Details : Modify targets/2.6-vanilla.target.in.
2117 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2118 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2119 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2120 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2124 Description: improve handling recoverable errors
2125 Details : if request processig with error which can be recoverable on server
2126 request should be resend, otherwise page released from cache and
2131 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2132 Details : Modify target file & which_patch.
2133 A flaw was found in the IA32 system call emulation provided
2134 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2135 value could be stored in the %RAX register, which could trigger an
2136 out-of-bounds system call table access. An untrusted local user
2137 could exploit this flaw to run code in the kernel
2138 (ie a root privilege escalation). (CVE-2007-4573).
2142 Description: O_DIRECT bypasses client statistics.
2143 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2144 read_ahead_stats were updated. copied the stats section from
2145 osc_send_oap_rpc() into async_internal().
2149 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2153 Description: change order of libsysio includes
2154 Details : '#include sysio.h' should always come before '#include xtio.h'
2158 Description: handle "serious error: objid * already exists" more gracefully
2159 Details : If LAST_ID value on disk is smaller than the objects existing in
2160 the O/0/d* directories, it indicates disk corruption and causes an
2161 LBUG(). If the object is 0-length, then we should use the existing
2162 object. This will help to avoid a full fsck in most cases.
2164 Severity : enhancement
2166 Description: adapt the lustre_config script to support the upgrade case
2167 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2168 server targets to 1.6 in parallel.
2172 Description: To avoid grant space > avaible space when the disk is almost
2173 full. Without this patch you might see the error "grant XXXX >
2174 available" or some LBUG about grant, when the disk is almost
2176 Details : In filter_check_grant, for non_grant cache write, we should
2177 check the left space by if (*left > ungranted + bytes), instead
2178 of (*left > ungranted), because only we are sure the left space
2179 is enough for another "bytes", then the ungrant space should be
2180 increase. In client, we should update cl_avail_grant only there
2181 is OBD_MD_FLGRANT in the reply.
2185 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2186 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2187 described in CVE-2007-4573. This problem could allow untrusted
2188 local users to gain root access.
2191 Frequency : when using O_DIRECT and quotas
2193 Description: Incorrect file ownership on O_DIRECT output files
2194 Details : block usage reported by 'lfs quota' does not take into account
2195 files that have been written with O_DIRECT.
2200 Description: touch file failed when fs is not full
2201 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2202 otherwise we can get ENOSP while fs is not full.
2206 Description: parallel lock callbacks
2207 Details : Instead of sending blocking and completion callbacks as separated
2208 requests, adding them to a set and sending in parallel.
2211 Frequency : only for Cray XT3
2212 Bugzilla : 12829/13455
2213 Description: Changing primary group doesn't change the group lustre assigns to
2215 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2216 overridden with the primary group provided by the group upcall,
2217 whereas the supplied fsgid can be trusted if it is in the list of
2218 supplementary groups returned by the group upcall.
2220 Severity : enhancement
2222 Description: Allow masking D_WARNING, D_ERROR messages from console
2223 Details : Console messages can now be disabled via lnet.printk.
2227 Description: User code with malformed file open parameter crashes client node
2228 Details : Before packing join_file req, all the related reference should be
2229 checked carefully in case some malformed flags cause fake
2230 join_file req on client.
2234 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2235 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2236 mode is equal to the granted mode. However, it should also check
2237 that LDLM_FL_CP_REQD is not set.
2241 Description: Heavy nfs access might result in deadlocks
2242 Details : After ELC code landed, it is now improper to enqueue any mds
2243 locks under och_sem, because enqueue might want to decide to
2244 cancel open locks for same inode we are holding och_sem for.
2248 Description: Client eviction while running blogbench
2249 Details : A lot of unlink operations with concurrent I/O can lead to a
2250 deadlock causing evictions. To address the problem, the number of
2251 oustanding OST_DESTROY requests is now throttled to
2252 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2253 callbacks are processed in priority.
2257 Description: enable ACLs on MDS by default
2258 Details : ACLs must be enabled on MDS by default.
2261 Frequency : PPC/PPC64 only
2263 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2264 Details : fix duplicated definitions between asm-ppc64/types.h and
2265 lustre_types.h on PPC.
2268 Frequency : PPC/PPC64 only
2270 Description: asm-ppc/segment.h does not exist
2271 Details : fix compile issue on PPC.
2275 Description: better handle error messages in extents code
2278 Frequency : RHEL4 only
2280 Description: mkfs is very slow on IA64/RHEL4
2281 Details : A performance regression has been discovered in the MPT Fusion
2282 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2283 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2284 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2286 Severity : enhancement
2288 Description: SNMP support enhancement
2289 Details : Adding total number of sampled request for an MDS node in snmp
2292 Severity : enhancement
2294 Description: Optimize ldlm waiting list processing for PR extent locks
2295 Details : When processing waiting list for read extent lock and meeting read
2296 lock that is same or wider to it that is not contended, skip
2297 processing rest of the list and immediatelly return current
2298 status of conflictness, since we are guaranteed there are no
2299 conflicting locks in the rest of the list.
2303 Description: Time out and refuse to reconnect
2304 Details : When the failover node is the primary node, it is possible
2305 to have two identical connections in imp_conn_list. We must
2306 compare not conn's pointers but NIDs, otherwise we can defeat
2307 connection throttling.
2311 Description: port llog fixes from b1_6 into HEAD
2312 Details : Port llog reference couting and some llog cleanups from b1_6
2313 (bug 10800) into HEAD, for protect from panic and access to already
2314 free llog structures.
2318 Description: Detect stride IO mode in read-ahead
2319 Details : When a client does stride read, read-ahead should detect that and
2320 read-ahead pages according to the detected stride pattern.
2324 Description: data checksumming impacts single node performance
2325 Details : add support for several checksum algorithm. Currently, only CRC32
2326 and Adler-32 are supported. The checksum type can be changed on
2327 the fly via /proc/fs/lustre/osc/*/checksum_type.
2331 Description: use adler32 for page checksums
2332 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2337 Description: build for x2 fails
2338 Details : fix compile issue on Cray systems.
2342 Description: Properly match for duplicate locks
2343 Details : Due to different lock order from skiplists code, we need to
2344 traverse entire list for now
2347 Frequency : only on PPC/SLES10
2349 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2350 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2351 be defined. Add a hack in mkfs_lustre.c to work around this compile
2356 Description: LASSERT on MDS when client holding flock lock dies
2357 Details : ldlm pool logic depends on number of granted locks equal to
2358 number of released locks which is not true for flock locks, so
2359 just exclude such locks from consideration.
2363 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2364 Details : Use fsfilt_sync() for both the whole filesystem sync and
2365 individual file sync to eliminate dangerous inode locking
2366 with I_LOCK that can lead to a deadlock.
2370 Description: performance in 1.6.3
2371 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2372 sure that 1MB requests are not fragmented by the block layer.
2374 Severity : enhancement
2376 Description: organize the server-side client stats on per-nid basis
2377 Details : Change the structure of stats under obdfilter and mds to
2385 The "uuid"s file would list the uuids of _active_ exports.
2386 And the clear entry is to clear all stats and stale nids.
2388 Severity : enhancement
2390 Description: eliminate client locks in face of contention
2391 Details : file contention detection and lockless i/o implementation
2392 for contended files.
2396 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2399 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2401 Description: MDS may lose file striping (and hence file data) in some cases
2402 Details : If there are additional extended attributes stored on the MDS,
2403 in particular ACLs, SELinux, or user attributes (if user_xattr
2404 is specified for the client mount options) then there is a risk
2405 of attribute loss. Additionally, the Lustre file striping
2406 needs to be larger than default (e.g. striped over all OSTs),
2407 and an additional attribute must be stored initially in the
2408 inode and then increase in size enough to be moved to the
2409 external attribute block (e.g. ACL growing in size) for file
2414 Description: skiplist implementation simplification
2415 Details : skiplists are used to group compatible locks on granted list
2416 that was implemented as tracking first and last lock of each
2417 lock group the patch changes that to using doubly linked lists
2421 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2422 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2423 because LASSERT on both the data supplied by a client, and the
2424 data on disk is dangerous and incorrect.
2426 Severity : enhancement
2428 Description: Slow truncate/writes to huge files at high offsets.
2429 Details : Directly associate cached pages to lock that protect those pages,
2430 this allows us to quickly find what pages to write and remove
2431 once lock callback is received.
2435 Description: more ldlm soft lockups
2436 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2437 starve other threads from the resource lock for a long time in
2438 case of long waiting queue, so change the debug level from
2439 D_OTHER to the less frequently used D_INFO.
2441 Severity : enhancement
2443 Description: add -gid, -group, -uid, -user options to lfs find
2447 Description: Hung threads in invalidate_inode_pages2_range
2448 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2449 once one is completed. As a result, some RPCs are stuck in the
2450 queue and are never sent.
2454 Description: filter threads hungs on waiting journal commit
2455 Details : Cleanup filter group llog code, then only filter group llog will
2456 be only created in the MDS/OST syncing process.
2460 Description: Procfs and llog threads access destoryed import sometimes.
2461 Details : Sync the import destoryed process with procfs and llog threads by
2462 the import refcount and semaphore.
2464 Severity : enhancement
2466 Description: openlock cache of b1_6 port to HEAD
2471 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2472 Details : If insertion of an extent fails, then discard the inode
2473 preallocation and free data blocks else it can lead to duplicate
2478 Description: don't always update ctime in ext3_xattr_set_handle()
2479 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2480 In some cases the ctime should not be updated, for example for
2481 2.0->1.8 compatibility it is necessary to delete an xattr and it
2482 should not update the ctime.
2486 Bugzilla : 15713/16362
2487 Description: Assertion in iopen_connect_dentry in 1.6.3
2488 Details : looking up an inode via iopen with the wrong generation number can
2489 populate the dcache with a disconneced dentry while the inode
2490 number is in the process of being reallocated. This causes an
2491 assertion failure in iopen since the inode's dentry list contains
2492 both a connected and disconnected dentry.
2496 Description: assertion failure in ldlm_handle2lock()
2497 Details : fix a race between class_handle_unhash() and class_handle2object()
2498 introduced in lustre 1.6.5 by bug 13622.
2503 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2504 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2509 Description: oops in sd_iostats_seq_show()
2510 Details : unloading/reloading the scsi low level driver triggers a kernel
2511 bug when trying to access the sd iostat file.
2516 Description: Kernel panics during QLogic driver reload
2517 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2518 patch, causing memory corruption.
2523 Description: journal_dev option does not work in b1_6
2524 Details : pass mount option during pre-mount.
2526 Severity : enhancement
2528 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2529 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2530 extent information of a file. It can be used to map logical blocks
2531 in a file to physical blocks in the block device.
2535 Description: LDLM soft lockups - improvement
2536 Details : It is be possible to send the lock handle along with each read
2537 or write request because the client is already doing a lock match
2538 itself so there isn't any reason the OST should have to re-do that
2542 Frequency : only on Cray X2
2544 Description: X2 build failures
2545 Details : fix build failures on Cray X2.
2549 Description: xid & resent requests
2550 Details : Initialize RPC XID from clock at startup (randomly if clock is
2553 Severity : enhancement
2555 Description: Add lustre_start utility to start or stop multiple Lustre servers
2560 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2561 Details : In case of memory pressure, list_del() can be called twice on
2562 req->rq_history_list, causing a kernel oops.
2566 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2567 Details : fix stack overflow in the distributed lock manager by defering
2568 export eviction after a failed AST to the elt thread instead of
2569 handling it in the dlm interpret routine.
2573 Description: Convert some comments to new format.
2574 Details : Update documenting comments to match doxygen conventions.
2578 Description: Grammar fixes.
2579 Details : A couple of trivial spelling fixes.
2583 Description: OSD_COUNTERS-mandatory
2584 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2585 to simplify the code.
2589 Description: simplify cmm_device freeing logic.
2590 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2594 Description: Add lockdep support to dt_object_operations locking interface.
2595 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2596 parameter indicating lock ordering. Update mdd code to use new
2601 Description: Introduce failloc constants for lockless IO tests.
2602 Details : Add two new failloc constants to test lockless IO. Only one of
2603 them in implemented---another is checked in yet to be landed
2608 Description: Add lockdep support for inode mutex.
2609 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2610 in the situations where Lustre has to lock more than one inode
2615 Description: Add optional invariants checking support.
2616 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2617 configure switch. This macro is to be used for consistency and
2618 sanity checks that are too expensive to be left in `production'
2623 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2624 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2629 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2630 Details : Use spin_lock_nested() in (the only) situation where more than
2631 one ldlm_lock is locked simultaneously. Also, fix possible
2632 dead-lock in ldlm_lock_change_resource() by enforcing particular
2637 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2638 Details : Instead of specifying each ldlm_lock call-back through separate
2639 parameter, wrap them into struct ldlm_callback_suite.
2643 Description: Kill join_lru obd method and its callers.
2644 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2645 in memory---a requirement implemented through ->o_join_lru() obd
2646 method in HEAD. Remove this method and its users.
2650 Description: Add asynchronous ldlm ENQUEUE completion handler.
2651 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2652 case that stock ldlm_completion_ast() cannot handle as it waits
2653 until lock is granted. Introduce new ldlm_completion_ast_async()
2654 for this. Also comment ldlm_completion_ast().
2658 Description: ldlm_error <-> errno conversion.
2659 Details : Add functions to map (rather arbitrary) between LDLM error codes
2660 and standard errno values. CLIO needs this to prevent LDLM specific
2661 constants from escaping ldlm and osc.
2665 Description: Kill unused ldlm_handle2lock_ns() function.
2666 Details : Kill unused ldlm_handle2lock_ns() function.
2670 Description: Add lu_ref support to ldlm_lock
2671 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2672 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2673 and ldlm_resource. LDLM interface has to be changed, because code that
2674 releases a reference on a lock, has to "know" what reference this is.
2675 In the most frequent case
2677 lock = ldlm_handle2lock(handle);
2679 LDLM_LOCK_PUT(lock);
2681 no changes are required. When any other reference (received _not_ from
2682 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2683 instead of LDLM_LOCK_PUT().
2685 Arguably, changes are pervasive, and interface requires some discipline
2686 for proper use. On the other hand, it was very instrumental in finding
2687 a few leaked lock references.
2691 Description: Add ldlm_lock_addref_try().
2692 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2693 attempts to addref a lock that might be being canceled
2698 Description: Add ldlm_weigh_callback().
2699 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2700 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2701 hard-coded `number of pages' logic.
2705 Description: Add lockdep annotations to llog code.
2706 Details : Use appropriately tagged _nested() locking calls in the places
2707 where llog takes more than one ->lgh_lock lock.
2711 Description: Add loi_kms_set().
2712 Details : Wrap kms updates into a helper function.
2716 Description: Constify instances of struct lsm_operations.
2717 Details : Constify instances of struct lsm_operations.
2721 Description: lu_conf support.
2722 Details : On a server, a file system object is uniquely identified
2723 by a fid, which is sufficient to locate and load all object
2724 state (inode). On a client, on the other hand, more data are
2725 necessary instantiate an object. Change lu_object_find() and
2726 friends to take additional `lu_conf' argument describing object.
2727 Typically this includes layout information.
2731 Description: lu_context fixes.
2732 Details : Introduce new lu_context functions that are needed on the client
2733 side, where some system threads (ptlrpcd) are shared by multiple
2734 modules, and so cannot be stopped during module shutdown.
2738 Description: Add start and stop methods to lu_device_type_operations.
2739 Details : Introduce two new methods in lu_device_type_operations, that are
2740 invoked when first instance of a given type is created and last one
2741 is destroyed respectively. This is need by CLIO.
2745 Description: Add lu_ref support to struct lu_device.
2746 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2747 track leaked references.
2751 Description: Introduce lu_kmem_descr.
2752 Details : lu_kmem_descr and its companion interface allow to create
2753 and destroy a number of kmem caches at once.
2757 Description: Fix lu_object finalization race.
2758 Details : Fix a race between lu_object_find() finding an object and its
2759 concurrent finalization. This race is (most likely) not possible
2760 on the server, but might happen on the client.
2764 Description: Introduce lu_ref interface.
2765 Details : lu_ref is a debugging module allowing to track references to
2766 a given object. It is quite cpu expensive, and has to be
2767 explicitly enabled with --enable-lu_ref. See usage description
2772 Description: Factor lu_site procfs stats into a separate function.
2773 Details : Separate lu_site stats printing code into a separate function
2774 that can be reused on a client.
2778 Description: Constify instances of struct {lu,dt,md}_device_operations.
2779 Details : Constify instances of struct {lu,dt,md}_device_operations.
2783 Description: Introduce struct md_site and move meta-data specific parts of
2784 struct lu_site here.
2785 Details : Move md-specific fields out of struct lu_site into special struct
2786 md_site, so that lu_site can be used on a client.
2790 Description: Kill mdd_lov_destroy().
2791 Details : Remove unused mdd code.
2795 Description: Add st_block checking to multistat.c.
2796 Details : Add st_block checking to multistat.c.
2800 Description: Add lu_ref support to struct obd_device.
2801 Details : Add lu_ref tracking to obd_device.
2805 Description: Kill obd_set_fail_loc().
2806 Details : Remove unused code.
2810 Description: Add special type for ptlrpc_request interpret functions.
2811 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2812 there. Actual usage will be in CLIO.
2816 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2817 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2818 doesn't work with lockdep.
2822 Description: Add rwv.c test program.
2823 Details : New testing program exercising readv(2) and writev(2) (Qian).
2827 Description: Add sendfile.c test program.
2828 Details : New testing program exercising sendfile(2) (Jay).
2832 Description: Ratelimit a message that can be very frequent.
2833 Details : Ratelimit a memory allocation failure message that can
2838 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2839 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2843 Description: Liblustre build fixes.
2844 Details : Liblustre build fixes.
2848 Description: libcfs: add cfs_{need,cond}_resched() interface.
2849 Details : libcfs: add cfs_{need,cond}_resched() definition and
2850 implementations for Linux, NT, and liblustre.
2852 Severity : enhancement
2854 Description: More exported tunables for mballoc
2855 Details : Add support for tunable preallocation window and new tunables for
2856 large/small requests
2860 Description: Detect corruption of block bitmap and checking for preallocations
2861 Details : Checks validity of on-disk block bitmap. Also it does better
2862 checking of number of applied preallocations. When corruption is
2863 found, it turns filesystem readonly to prevent further corruptions.
2867 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2868 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2869 window to 0. If it does hit cache miss, and read-pattern is still
2870 stride-io mode, does not reset the stride window, but also does not
2871 increase the stride window length in this case.
2875 Frequency : only for big-endian servers
2876 Description: Check if system is big-endian while mounting fs with extents feature
2877 Details : Mounting a filesystem with extents feature will fail on big-endian
2878 systems since ext3-based ldiskfs is not supported on big-endian
2879 systems. This can be over-riden with "bigendian_extents" mount option.
2881 Severity : enhancement
2883 Description: The root squash functionality
2884 Details : A security feature, which is to prevent users from being able
2885 to mount lustre on their desktop, run as root, and delete
2886 all of the files in the filesystem. The goal is accomplished by
2887 remapping user id (UID) and group id (GID) of the root user to
2888 a UID and GID specified by the system administartor via Lustre
2889 configuration management server (MGS). The functionality also
2890 allows to specify sets of clients for which the remapping does
2895 Description: Excessive recovery window
2896 Details : With AT enabled, the recovery window can be excessively long (6000+
2897 seconds). To address this problem, we no longer use
2898 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2899 timeout no longer depends on the service time, it is set to
2900 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2901 time via pb_service_time.
2905 Description: Watchdog triggered on MDS failover
2906 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2907 the OSTs know that the MDS "UUID" can be reused for the same export
2908 from a different NID, so we do not need to wait for the export to be
2912 Frequency : rare, only if using MMP with Linux RAID
2914 Description: MMP doesn't work with Linux RAID
2915 Details : While using HA for Lustre servers with Linux RAID, it is possible
2916 that MMP will not detect multiple mounts. To make this work we
2917 need to unplug the device queue in RAID when the MMP block is being
2918 written. Also while reading the MMP block, we should read it from
2919 disk and not the cached one.
2921 Severity : enhancement
2923 Description: open file using fid
2924 Details : A file can be opened using just its fid, like
2925 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2928 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2931 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2932 Details : add proper locking for creation of jbd_4k slab cache
2936 Description: MMP check in ext3_remount() fails without displaying any error
2937 Details : When multiple mount protection fails during remount, proper error
2940 Severity : enhancement
2942 Description: Allow stripe size to be up to 4G-64k
2943 Details : Fix math logic to allow large stripe sizes.
2947 Description: add check for >8TB ldiskfs filesystems
2948 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2949 Don't allow >8TB ldiskfs filesystems to be mounted without
2950 force_over_8tb mount option
2955 Descriptoin: OSS complains "too many blocks" and crashes
2956 Details : In filter_commitrw_write():
2957 1. lquota_chkquota();
2958 2. obdo_from_inode();
2959 3. lquota_pending_commit();
2960 "2." has changed uid and gid in this case, then caused
2961 lquota_chkquota() and lquota_pending_commit() to use
2962 different uid and gid. That is the root of the bug.
2966 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2967 Details : Detect and print out the 'stripe count = -1' case properly.
2971 Description: Out or order replies might be lost on replay
2972 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2973 request with tid smaller than one currently being added, add it
2974 to the start, not end of the list.
2976 --------------------------------------------------------------------------------
2978 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2980 * Support for kernels:
2981 2.6.5-7.283 (SLES 9),
2982 2.6.9-55.EL (RHEL 4),
2983 2.6.16.46-0.14 (SLES 10),
2984 2.6.18.8 vanilla (kernel.org)
2985 * Client support for unpatched kernels:
2986 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2987 2.6.16 - 2.6.22 vanilla (kernel.org)
2988 * Due to recently discovered recovery problems, we do not recommend
2989 using patchless RHEL 4 clients with this or any earlier release.
2990 * Recommended e2fsprogs version: 1.39.cfs8
2991 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2992 * Starting with this release, the ldiskfs backing filesystem required
2993 by Lustre is now in its own package, lustre-ldiskfs. This package
2994 should be installed. It is versioned separately from Lustre and
2995 may be released separately in future.
2999 Description: block reactivating mgc import until all deactivates complete
3000 Details : Fix race when failing back MDT/MGS to itself (testing)
3003 Frequency : at statup only
3005 Description: mds_lov_synchronize race leads to various problems
3006 Details : simultaneous MDT->OST connections at startup can cause the
3007 sync to abort, leaving the OSC in a bad state.
3009 Severity : enhancement
3011 Description: add optional extra BUILD_VERSION info
3012 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3013 to override the lustre version.
3016 Frequency : 2.6.18 servers only
3018 Description: ll_kern_mount() doesn't release the module reference
3019 Details : The ldiskfs module reference count never drops down to 0
3020 because ll_kern_mount() doesn't release the module reference.
3025 Description: server LBUG when using old ost_num_threads parameter
3026 Details : Accept the old ost_num_threads parameter but warn that it
3027 is deprecated, and fix an off-by-one error that caused an LBUG.
3032 Description: Transient SCSI error results in persistent IO issue
3033 Details : iobuf->dr_error is not reinitialized to 0 between two
3037 Frequency : sometimes when underlying device returns I/O errors
3039 Description: OSTs not going read-only during write failures
3040 Details : OSTs are not remounted read-only when the journal commit threads
3041 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3042 instead of the ext3 wrappers.
3046 Description: poor connect scaling with increasing client count
3047 Details : Don't run filter_grant_sanity_check for more than 100 exports
3048 to improve scaling for large numbers of clients.
3051 Frequency : SLES10 only
3053 Description: sanity-quota.sh quotacheck failed: rc = -22
3054 Details : Quotas cannot be enabled on SLES10.
3057 Frequency : liblustre clients only
3059 Description: getdirentries does not give error when run on compute nodes
3060 Details : getdirentries does not fail when the size specified as an argument
3061 is too small to contain at least one entry
3063 Severity : enhancement
3065 Description: Add LNET router traceability for debug purposes
3066 Details : If a checksum failure occurs with a router as part of the
3067 IO path, the NID of the last router that forwarded the bulk data
3068 is printed so it can be identified.
3073 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3074 Details : Due to a race condition, liblustre clients were occasionally
3075 evicted incorrectly.
3077 Severity : enhancement
3079 Description: lfs setstripe use optional parameters instead of postional
3082 Severity : enhancement
3084 Description: Nanosecond timestamp support for ldiskfs
3085 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3086 resolution timestamps. There is not yet support for this at
3087 the Lustre filesystem level.
3090 Frequency : during server recovery
3092 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3093 Details : request with rq_no_resend flag not awake l_wait_event if they get
3097 Frequency : nfs export on patchless client
3099 Description: connectathon hang when test nfs export over patchless client
3100 Details : Disconnected dentry cannot be found with lookup, so we do not need
3101 to unhash it or make it invalid
3104 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3105 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3106 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3111 Description: open req refcounting wrong on reconnect
3112 Details : If reconnect happened between getting open reply from server and
3113 call to mdc_set_replay_data in ll_file_open, we will schedule
3114 replay for unreferenced request that we are about to free.
3115 Subsequent close will crash in variety of ways.
3116 Check that request is still eligible for replay in
3117 mdc_set_replay_data().
3122 Description: disable writes to filesystem when reading health_check file
3123 Details : the default for reading the health_check proc file has changed
3124 to NOT do a journal transaction and write to disk, because this
3125 can cause reads of the /proc file to hang and block HA state
3126 checking on a healthy but otherwise heavily loaded system. It
3127 is possible to return to the previous behaviour during configure
3128 with --enable-health-write.
3130 Severity : enhancement
3132 Description: 64-bit inode version
3133 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3134 to the inode. This will be required for version-based recovery.
3139 Description: MDS fails to start if a duplicate client export is detected
3140 Details : in some rare cases it was possible for a client to connect to
3141 an MDS multiple times. Upon recovery the MDS would detect this
3142 and fail during startup. Handle this more gracefully.
3144 Severity : enhancement
3146 Description: Add -o localflock option to simulate old noflock
3148 Details : This will achieve local-only flock/fcntl locks
3154 Description: log_commit_thread vs filter_destroy race leads to crash
3155 Details : Take import reference before releasing llog record semaphore
3160 Description: Wrong request locking in request set processing
3161 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3162 imp_delayed_list, in this place should be used imp_lock.
3165 Frequency : when reconnection
3167 Description: Grant Leak when osc reconnect to OST
3168 Details : When osc reconnect ost, OST(filter) should check whether it
3169 should grant more space to client by comparing fed_grant and
3170 cl_avail_grant, and return the granted space to client instead
3171 of "new granted" space, because client will call osc_init_grant
3172 to update the client grant space info.
3175 Frequency : when client reconnect to OST
3177 Description: Grant Leak when osc do resend and replay bulk write
3178 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3179 bulk write request, because the grant info while be sync between
3180 OSC and OST when reconnect, and we should ignore the grant info
3181 these of resend/replay write req.
3186 Description: Grant space more than available left space sometimes.
3187 Details : When then OST is about to be full, if two bulk writing from
3188 different clients came to OST. Accord the avaliable space of the
3189 OST, the first req should be permitted, and the second one
3190 should be denied by ENOSPC. But if the seconde arrived before
3191 the first one is committed. The OST might wrongly permit second
3192 writing, which will cause grant space > available space.
3195 Frequency : when client is evicted
3197 Description: Grant might be wrongly erased when osc is evicted by OST
3198 Details : when the import is evicted by server, it will fork another
3199 thread ptlrpc_invalidate_import_thread to invalidate the
3200 import, where the grant will be set to 0. While the original
3201 thread will update the grant it got when connecting. So if
3202 the former happened latter, the grant will be wrongly errased
3203 because of this race.
3208 Description: Checking Stale with correct fid
3209 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3210 to check whether it is stale, because sometimes, we want the
3211 enqueue happened anyway, and op_data.fid2 will not be initialized.
3213 Severity : enhancement
3215 Description: update patchless client
3216 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3219 Frequency : only with 2.4 kernel
3221 Description: random memory corruption
3222 Details : size of struct ll_inode_info is to big for union inode.u and this
3223 can be cause of random memory corruption.
3228 Description: Memory leak in recovery
3229 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3230 It should also check obd_fail before fsfilt_start, otherwise if
3231 fsfilt_start return -EROFS,(failover mds during mds recovery).
3232 then the req will return with repmsg->transno = 0 and rc = EROFS.
3233 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3234 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3235 be freed no matter whether fsfilt_commit success or not.
3238 Frequency : only with huge count clients
3240 Description: Prevents from taking the superblock lock in llap_from_page for
3242 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3243 for llap_from_page call in ll_removepage prevents from taking the
3244 superblock lock for a soon died page.
3249 Description: Not check open intent error before release open handle
3250 Details : in some rare cases, the open intent error is not checked before
3251 release open handle, which may cause
3252 ASSERTION(open_req->rq_transno != 0), because it tries to release
3253 the failed open handle.
3258 Description: Set cat log bitmap only after create log success.
3259 Details : in some rare cases, the cat log bitmap is set too early. and it
3260 should be set only after create log success.
3264 Description: Accessing a block bevice can re-enable I/O when Lustre is
3265 tearing down a device.
3266 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3271 Description: service threads may hog cpus when there are a lot of requests
3272 Details : Insert cond_resched to give other threads a chance to use some CPU
3277 Description: the cat log was not initialized in recovery
3278 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3279 unlink log on mds will not be initialized until mds post
3280 recovery. And also in mds post recovery, the unlink log will
3281 initialization will be done asynchronausly, so there will be race
3282 between add unlink log and unlink log initialization.
3286 Description: brw_stats were being printed incorrectly
3287 Details : brw_stats were being printed as log2 but all of them were not
3288 recorded as log2. Also remove some code duplication arising from
3289 filter_tally_{read,write}.
3293 Frequency : rare, only in recovery.
3294 Description: ASSERTION(req->rq_type != LI_POISON) failed
3295 Details : imp_lock should be held while iterating over imp_sending_list for
3296 prevent destroy request after get timeout in ptlrpc_queue_wait.
3300 Description: replay-single.sh test 52 fails
3301 Details : A lock's skiplist need to be cleanup when it being unlinked
3302 from its resource list.
3306 Description: Short directio read returns full requested size rather than
3308 Details : Direct I/O operations should return actual amount of bytes
3309 transferred rather than requested size.
3311 Severity : enhancement
3313 Description: metadata RPC reduction (e.g. for rm performance)
3314 Details : decrease the amount of synchronous RPC between clients and servers
3315 by canceling conflicing lock before the operation on the client
3316 and packing thier handles into the main operation RPC to server.
3318 Severity : enhancement
3320 Description: Async OSC create to avoid the blocking unnecessarily.
3321 Details : If an OST has no remaining object, system will block on the
3322 creation when it needs to create a new object on this OST. Now,
3323 ways use pre-created objects when available, instead of blocking on
3324 an empty osc while others are not empty. If we must block, we
3325 block for the shortest possible period of time.
3329 Description: improve handling recoverable errors
3330 Details : if request processig with error which can be recoverable on server
3331 request should be resend, otherwise page released from cache and
3334 Severity : enhancement
3336 Description: refine locking for avoid write wrong info into lov_objid file
3337 Details : fix possible races with add new target and write/update data in
3340 --------------------------------------------------------------------------------
3342 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3347 Frequency : on some architectures
3349 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3350 Details : Uninitialized flags sometimes cause configuration commands to
3354 Frequency : patchless clients only
3356 Description: missing __iget() symbol export
3357 Details : The __iget() symbol export is missing. To avoid the need for
3358 this on patchless clients the deathrow inode reaper is turned
3359 off, and we depend on the VM to clean up old inodes. This
3360 dependency was during via the fix for bug 12181.
3362 --------------------------------------------------------------------------------
3364 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3366 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3367 INTEROPERATE with older versions automatically. In many cases a
3368 special upgrade step is needed. Please read the
3369 user documentation before upgrading any part of a live system.
3370 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3371 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3372 * WARNING: Lustre configuration and startup changes are required with
3373 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3377 Severity : enhancement
3379 Description: MountConf
3380 Details : Lustre configuration is now managed via mkfs and mount
3381 commands instead of lmc and lconf. New obd types (MGS, MGC)
3382 are added for dynamic configuration management. See
3383 https://mail.clusterfs.com/wikis/lustre/MountConf for
3386 Severity : enhancement
3388 Description: dynamic OST addition
3389 Details : OSTs can now be added to a live filesystem
3391 Severity : enhancement
3393 Description: startup order invariance
3394 Details : MDTs and OSTs can be started in any order. Clients only
3395 require the MDT to complete startup.
3397 Severity : enhancement
3399 Description: parallel, asynchronous orphan cleanup
3400 Details : orphan cleanup is now performed in separate threads for each
3401 OST, allowing parallel non-blocking operation.
3403 Severity : enhancement
3405 Description: optimized stripe assignment
3406 Details : stripe assignments are now made based on ost space available,
3407 ost previous usage, and OSS previous usage, in order to try
3408 to optimize storage space and networking resources.
3410 Severity : enhancement
3412 Description: Permanently set tunables
3413 Details : All writable /proc/fs/lustre tunables can now be permanently
3414 set on a per-server basis, at mkfs time or on a live system.
3416 Severity : enhancement
3418 Description: Lustre message v2
3419 Details : Add lustre message format v2.
3421 Severity : enhancement
3423 Description: client OST exclusion list
3424 Details : Clients can be started with a list of OSTs that should be
3425 declared "inactive" for known non-responsive OSTs.
3429 Description: SPEC SFS validation failure on NFS v2 over lustre.
3430 Details : Changes the blocksize for regular files to be 2x RPC size,
3431 and not depend on stripe size.
3433 Severity : enhancement
3435 Description: Multiple MD RPCs in flight.
3436 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3437 To avoid overly-overloading MDS, introduce a limit on number of
3438 MDS RPCs in flight for a single client and add /proc controls
3439 to adjust this limit.
3441 Severity : enhancement
3443 Description: client read/write statistics
3444 Details : Add client read/write call usage stats for performance
3445 analysis of user processes.
3446 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3447 file access. extents_stats shows chunk size distribution.
3448 extents_stats_per_process show chunk size distribution per
3451 Severity : enhancement
3453 Description: mds statistics
3454 Details : Add detailed mds operations statistics in
3455 /proc/fs/lustre/mds/*/stats.
3459 Description: Failure of copying files with lustre special EAs.
3460 Details : Client side always return success for setxattr call for lustre
3461 special xattr (currently only "trusted.lov").
3466 Description: Refcount LNET uuids
3467 Details : The global LNET uuid list grew linearly with every startup;
3468 refcount repeated list entries instead of always adding to
3471 Severity : enhancement
3473 Description: Dynamic service threads
3474 Details : Within a small range, start extra service threads
3475 automatically when the request queue builds up.
3478 Frequency : mixed-endian client/server environments
3480 Description: mixed-endian crashes
3481 Details : The new msg_v2 system had some failures in mixed-endian
3484 Severity : enhancement
3486 Description: Easy OST removal
3487 Details : OSTs can be permanently deactivated with e.g. 'lctl
3488 conf_param lustre-OST0001.osc.active=0'
3490 Severity : enhancement
3492 Description: MGS proc entries
3493 Details : Added basic proc entries for the MGS showing what filesystems
3496 Severity : enhancement
3498 Description: provide MGS failover
3499 Details : Added config lock reacquisition after MGS server failover.
3501 Severity : enhancement
3503 Description: add Linux 2.4 support
3504 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3508 Description: a large application tries to do I/O to the same resource and dies
3509 in the middle of it.
3510 Details : Check the req->rq_arrival time after the call to
3511 ost_brw_lock_get(), but before we do anything about
3512 processing it & sending the BULK transfer request. This
3513 should help move old stale pending locks off the queue as
3514 quickly as obd_timeout.
3517 Frequency : when an incorrect nid is specified during startup
3519 Description: ptlrpc connect to non-existant node causes kernel crash
3520 Details : LNET can't be re-entered from an event callback, which
3521 happened when we expire a message after the export has been
3522 cleaned up. Instead, hand the zombie cleanup off to another
3525 Severity : enhancement
3527 Description: plain/inodebits lock performance improvement
3528 Details : Grouping plain/inodebits in granted list by their request modes
3529 and bits policy, thus improving the performance of search through
3533 Frequency : only if OST filesystem is corrupted
3535 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3536 Details : for a short time RPCs with bulk IO are in the replay list,
3537 but replay of bulk IOs is unimplemented. If the OST filesystem
3538 is corrupted due to disk cache incoherency and then replay is
3539 started it is possible to trip an assertion. Avoid putting
3540 committed RPCs into the replay list at all to avoid this issue.
3543 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3545 Description: System hang on startup
3546 Details : This bug allowed the liblustre (e.g. catamount) client to
3547 return to the app before handling all startup RPCs. This
3548 could leave the node unresponsive to lustre network traffic
3549 and manifested as a server ptllnd timeout.
3551 Severity : enhancement
3553 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3554 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3555 boolean to control whether to print peer debug info when a
3556 client's RPC times out.
3559 Frequency : only for kernels with patches from Lustre below 1.4.3
3561 Description: Remove old rdonly API
3562 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3565 Frequency : only for devices with external journals
3567 Description: Set external device read-only also
3568 Details : During a commanded failover stop, we set the disk device
3569 read-only while the server shuts down. We now also set any
3570 external journal device read-only at the same time.
3573 Frequency : when upgrading from 1.4 while trying to change parameters
3575 Description: The wrong (new) MDC name was used when setting parameters for
3576 upgraded MDT's. Also allows changing of OSC (and MDC)
3577 parameters if --writeconf is specified at tunefs upgrade time.
3580 Frequency : when setting specific ost indicies
3582 Description: QOS code breaks on skipped indicies
3583 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3584 created with --index need not be sequential.
3586 Severity : enhancement
3588 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3589 Details : The uninit_groups feature works in conjunction with the kernel
3590 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3591 up the pass1 processing of e2fsck. This is a read-only feature
3592 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3593 mount filesystems that have had this feature enabled.
3595 Severity : enhancement
3597 Description: Improve multi-block allocation algorithm to avoid fragmentation
3598 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3599 allocation locality and avoid filesystem fragmentation.
3601 ------------------------------------------------------------------------------
3603 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3605 * Support for kernels:
3606 2.4.21-47.0.1.EL (RHEL 3)
3607 2.6.5-7.283 (SLES 9)
3608 2.6.9-42.0.10.EL (RHEL 4)
3609 2.6.12.6 vanilla (kernel.org)
3610 2.6.16.27-0.9 (SLES 10)
3611 * Recommended e2fsprogs version: 1.39.cfs5
3613 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3617 Frequency : occasional, depends on client load and configuration
3618 Bugzilla : 12181, 12203
3619 Description: data loss for recently-modified files
3621 Details : In some cases it is possible that recently written or created
3622 files may not be written to disk in a timely manner (this should
3623 normally be within 30s unless client IO load is very high).
3624 The problem appears as zero-length files or files that are a
3625 multiple of 1MB in size after a client crash or client eviction
3626 that are missing data at the end of the file.
3628 This problem is more likely to be hit on clients where files are
3629 repeatedly created and unlinked in the same directory, clients
3630 have a large amount of RAM, have many CPUs, the filesystem has
3631 many OSTs, the clients are rebooted frequently, and/or the files
3632 are not accessed by other nodes after being written.
3634 The presence of the problem can be detected by looking at
3635 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3636 smaller than the second (nr_unused) then dirty files will not
3637 be flushed automatically to disk. "sync; sleep 10" should be
3638 run several times on the node before unmounting it to update
3639 Lustre (this is also safe to run on nodes without this problem).
3641 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3642 that can cause this same problem, so customers using that kernel
3643 also need to update the kernel in addition to Lustre. In order
3644 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3646 It is normal that files written just before a client crash (less
3647 than 30s) may not yet have been flushed to disk, even for local
3651 Frequency : frequent on thin XT3 nodes
3653 Description: UUID collision on thin XT3 Linux nodes
3654 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3655 randomly, since we relied on an insufficiently-seeded PRNG.
3660 Description: OSS hangs after "All ost request buffers busy"
3661 Details : A deadlock between quota and journal operations caused OSS
3662 hangs after printing "All ost request buffers busy."
3665 Frequency : always on liblustre builds
3667 Description: Cleanup compiler warnings on liblustre
3670 Frequency : always on liblustre builds on XT3
3672 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3676 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3677 > 32000 subdirectories
3678 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3679 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3680 subdirectory count crosses 32000. This will aid e2fsck to
3681 correctly handle more than 32000 subdirectories.
3684 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3686 Description: System hang on startup
3687 Details : This bug allowed the liblustre (e.g. catamount) client to
3688 return to the app before handling all startup RPCs. This
3689 could leave the node unresponsive to lustre network traffic
3690 and manifested as a server ptllnd timeout.
3692 Severity : enhancement
3694 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3695 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3696 boolean to control whether to print peer debug info when a
3697 client's RPC times out.
3702 Description: make O_SYNC working on 2.6 kernels
3703 Details : 2.6 kernels use different method for mark pages for write,
3704 so need add a code to lustre for O_SYNC work.
3709 Description: Failure to close file and release space on NFS
3710 Details : Put inode details into lock acquired in ll_intent_file_open.
3711 Use mdc_intent_lock in ll_intent_open to properly
3712 detect all kind of errors unhandled by mdc_enqueue.
3717 Description: proc file read during shutdown sometimes raced obd removal,
3719 Details : Add lock to prevent obd access after proc file removal.
3722 Frequency : Only for files larger than 4GB on 32-bit clients.
3724 Description: improperly doing page alignment of locks
3725 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3726 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3731 Description: RPCs being resent when they shouldn't be
3732 Details : Some RPCs that should not be resent are being resent. This
3733 can cause inconsistencies in the RPC state machine. Do not
3734 resend such requests.
3737 Frequency : rare, only with NFS export
3739 Description: Crash on NFS re-export node
3740 Details : under very unusual load conditions an assertion is hit in
3741 ll_intent_file_open()
3744 Frequency : only if OST filesystem is corrupted
3746 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3747 Details : for a short time RPCs with bulk IO are in the replay list,
3748 but replay of bulk IOs is unimplemented. If the OST filesystem
3749 is corrupted due to disk cache incoherency and then replay is
3750 started it is possible to trip an assertion. Avoid putting
3751 committed RPCs into the replay list at all to avoid this issue.
3756 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3757 Details : Large single O_DIRECT read and write calls can fail to allocate
3758 a sufficiently large buffer to process the request. In case of
3759 allocation failure the allocation is retried with a smaller
3760 buffer and broken into smaller requests.
3762 Severity : enhancement
3764 Description: Add -o localflock option to simulate old noflock behaviour.
3765 Details : This will achieve local-only flock/fcntl locks coherentness.
3770 Description: versioning check is incomplete
3771 Details : Checking the version difference of client vs. server, report
3772 error if the gap is too big.
3777 Description: add support PG_writeback bit
3778 Details : add support for PG_writeback bit for Lustre, for more carefull
3779 work with page cache in 2.6 kernel. This also fix some deadlocks
3780 and remove hack for work O_SYNC with 2.6 kernel.
3782 Severity : enhancement
3784 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3785 Details : The uninit_groups feature works in conjunction with the kernel
3786 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3787 up the pass1 processing of e2fsck. This is a read-only feature
3788 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3789 mount filesystems that have had this feature enabled.
3791 Severity : enhancement
3793 Description: Improve multi-block allocation algorithm to avoid fragmentation
3794 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3795 allocation locality and avoid filesystem fragmentation.
3797 ------------------------------------------------------------------------------
3799 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3801 * Support for kernels:
3802 2.6.9-42.0.3.EL (RHEL 4)
3803 2.6.5-7.276 (SLES 9)
3804 2.4.21-47.0.1.EL (RHEL 3)
3805 2.6.12.6 vanilla (kernel.org)
3806 2.6.16.21-0.8 (SLES10)
3807 * Recommended e2fsprogs version: 1.39.cfs2-0
3809 * The backwards-compatible /proc/sys/portals symlink has been removed
3810 in this release. Before upgrading, please ensure that you change
3811 any configuration scripts or /etc/sysctl.conf files that access
3812 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3813 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3814 in advance of the upgrade on any system running Lustre 1.4.6 or
3815 newer, since /proc/sys/lnet was added in that version.
3816 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3820 Frequency : only when quota is used
3822 Description: avoid scanning export list for quota master
3823 Details : Change the algorithms to avoid scanning export list in order
3824 to improve the efficiency.
3827 Frequency : MDS failover only, very rarely
3829 Description: "went back in time" messages on mds failover
3830 Details : The greatest transno may be lost when the current operation
3831 finishes with an error (transno==0) and the client's last_rcvd
3832 record is over-written. Save the greatest transno in the
3833 mds_last_transno for this case.
3836 Frequency : always for specific kernels and striping counts
3838 Description: client may get "Matching packet too big" without ACL support
3839 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3840 when trying to access files in certain configurations. The
3841 clients should in fact be denied when mounting because they do
3842 not understand ACLs.
3845 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3847 Description: many clients connecting with IO in progress causes connect timeouts
3848 Details : Avoid synchronous journal commits to avoid delays caused by many
3849 clients connecting/disconnecting when bulk IO is in progress.
3850 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3851 OST_IO_PORTAL to avoid delays behind potentially many pending
3855 Frequency : occasionally with multiple writers to a single file
3857 Description: shared writes to file may result in wrong size reported by stat()
3858 Details : Allow growing of kms when extent lock is cancelled
3861 Frequency : always with random mmap IO to multi-striped file
3863 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3864 Details : Only if the hole is at the end of OST object so that kms is too
3865 small. Fix is to increase kms accordingly in ll_nopage.
3868 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3870 Description: writes to a missing object would leak memory on the OST
3871 Details : If there is an inconsistency between the MDS and OST filesystems,
3872 such that the MDS references an object that doesn't exist, writes
3873 to that object will leak memory due to incorrect cleanup in the
3874 error handling path, eventually running out of memory on the OST.
3879 Description: Creating too long symlink causes lustre errors
3880 Details : Check symlink and name lengths before sending requests to MDS.
3883 Frequency : only if flock is enabled (not on by default)
3885 Description: posix locks not released on fd closure on 2.6.9+
3886 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3887 kernels, this caused such locks not to be released on fd close and
3888 then assertions on fs unmount about still used locks.
3891 Frequency : MDS failover only, very rarely
3893 Description: clients may get ASSERTION(granted_lock != NULL)
3894 Details : When request was taking a long time, and a client was resending
3895 a getattr by name lock request. The were multiple lock requests
3896 with the same client lock handle and
3897 mds_getattr_name->fixup_handle_for_resent_request found one of the
3898 lock handles but later failed with ASSERTION(granted_lock != NULL).
3903 Description: handle->h_buffer_credits > 0, assertion failure
3904 Details : h_buffer_credits is zero after truncate, causing assertion
3905 failure. This patch extends the transaction or creates a new
3909 Frequency : NFS re-export or patchless client
3910 Bugzilla : 11179, 10796
3911 Description: Crash on NFS re-export node (__d_move)
3912 Details : We do not want to hash the dentry if we don't have a lock.
3913 But if this dentry is later used in d_move, we'd hit uninitialised
3914 list head d_hash, so we just do this to init d_hash field but
3915 leave dentry unhashed.
3918 Frequency : NFS re-export or patchless client
3920 Description: NFS exports has problem with symbolic link
3921 Details : lustre client didn't properly install dentry when re-exported
3922 to NFS or running patchless client.
3925 Frequency : NFS re-export or patchless client
3927 Description: Various nfs/patchless fixes.
3928 Details : fixes reuse disconected alias for lookup process - this fixes
3929 warning "find_exported_dentry: npd != pd",
3930 fix permission error with open files at nfs.
3931 fix apply umask when do revalidate.
3934 Frequency : occasional
3936 Description: Crash on NFS re-export node
3937 Details : calling clear_page() on the wrong pointer triggered oops in
3938 generic_mapping_read().
3941 Frequency : rarely, using O_DIRECT IO
3943 Description: unaligned directio crashes client with LASSERT
3944 Details : check for unaligned buffers before trying any requests.
3947 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3949 Description: stale data returned from RAID cache
3950 Details : If only a small amount of IO is done to the RAID device before
3951 reading it again it is possible to get stale data from the RAID
3952 cache instead of reading it from disk.
3955 Frequency : always for sles10 kernel
3957 Description: sles10 support
3958 Details : ll_follow_link: compile fixes and using of nd_set_link
3959 under newer kernels.
3962 Frequency : depends on arch, kernel and compiler version, always on sles10
3965 Description: recursive or deep enough symlinks cause stack overflow
3966 Details : getting rid of large stack-allocated variable in
3970 Frequency : depends on hardware
3972 Description: lustre write performance loss in the SLES10 kernel
3973 Details : the performance loss is caused by using of write barriers in the
3974 ext3 code. The SLES10 kernel turns barrier support on by
3975 default. The fix is to undo that change for ldiskfs.
3977 ------------------------------------------------------------------------------
3979 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3981 * Support for kernels:
3982 2.6.9-42.0.3EL (RHEL 4)
3983 2.6.5-7.276 (SLES 9)
3984 2.4.21-47.0.1.EL (RHEL 3)
3985 2.6.12.6 vanilla (kernel.org)
3989 Frequency : quota enabled and large files being deleted
3991 Description: releasing more than 4GB of quota at once hangs OST
3992 Details : If a user deletes more than 4GB of files on a single OST it
3993 will cause the OST to spin in an infinite loop. Release
3994 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3999 Description: statfs data retrieved from /proc may be stale or zero
4000 Details : When reading per-device statfs data from /proc, in the
4001 {kbytes,files}_{total,free,avail} files, it may appear
4002 as zero or be out of date.
4005 Frequency : systems with MD RAID1 external journal devices
4007 Description: lconf's call to blkid is confused by RAID1 journal devices
4008 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4009 of returning all block devices that match the journal UUID.
4012 Frequency : always, for aggregate stripe size over 4GB
4014 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4015 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4016 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4019 Frequency : always if "lfs find" used on a local file/directory
4021 Description: "lfs find" segfaults if used on a local file/directory
4022 Details : The case where a directory component was not specified wasn't
4023 handled correctly. Handle this properly.
4026 Frequency : always on ppc64
4028 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4029 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4030 perform endianness conversion.
4033 Frequency : rarely (truncate to non-zero file size after write under load)
4034 Bugzilla : 10730, 10687
4035 Description: Files padded with zeros to next 4K multiple
4036 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4037 it is possible that files that are truncated to a non-zero size
4038 immediately after being written are filled with zero bytes beyond
4039 the truncated size. No file data is lost.
4041 Severity : enhancement
4043 Description: Allow recovery/failover for liblustre clients.
4044 Details : liblustre clients were unaware of failover configurations until
4047 Severity : enhancement
4049 Description: user file locks should fail when not mounting with flock option
4050 Details : Set up an error-returning stub in ll_file_operations.lock field
4051 to prevent incorrect behaviour when client is mounted without
4052 flock option. Also, set up properly f_op->flock field for
4056 Frequency : always on ia64
4058 Description: "lfs df" loops on printing out MDS statfs information
4059 Details : The obd_ioctl_data was not initialized and in some systems
4060 this caused a failure during the ioctl that did not return
4061 an error. Initialize the struct and return an error on failure.
4064 Frequency : SLES 9 only
4066 Description: Error of copying files with lustre special EAs as root
4067 Details : Client side always return success for setxattr call for lustre
4068 special xattr (currently only "trusted.lov").
4071 Frequency : rarely on clusters with both ia64+i386 clients
4073 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4074 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4075 clients are doing concurrent writes to the same file it is
4076 possible that the ia64 clients may LASSERT because the OST
4077 extent locks are not PAGE_SIZE aligned. Ensure that grown
4078 locks are always aligned on the request boundary.
4081 Frequency : specific use, occasional
4083 Description: Overwriting in use executable truncates on-disk binary image
4084 Details : If one node attempts to overwrite an executable in use by
4085 another node, we now correctly return ETXTBSY instead of
4086 truncating the file.
4088 Severity : enhancement
4090 Description: Async OSC create to avoid the blocking unnecessarily.
4091 Details : If an OST has no remaining object, system will block on the creating
4092 when need to create a new object on this OST. Now, ways use
4093 pre-created objects when available, instead of blocking on an
4094 empty osc while others are not empty. If we must block, we block
4095 for the shortest possible period of time.
4100 Description: chmod on Lustre root is propagated to other clients
4101 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4102 invalid by the follow_mount time.
4105 Frequency : liblustre clients only
4107 Description: Race in 'instant cancel' lock handling could lead to such locks
4108 never to be granted in case of SMP MDS
4109 Details : Do not destroy not yet granted but cbpending locks in
4113 Frequency : replay/resend of open
4115 Description: non null lock assetion failure in mds_intent_policy
4116 Details : Trying to replay/resend lockless open requests resulted in
4117 mds_open() returning 0 with no lock. Now it sets a flag if
4118 a lock is going to be returned.
4120 Severity : enhancement
4122 Description: Checksum enhancements
4123 Details : New checksum enhancements allow for resending RPCs that failed
4126 Severity : enhancement
4128 Description: Tunables on number of dirty pages in cacche
4129 Details : Allow to set limit on number of dirty pages cached.
4134 Description: client crash on unmount - lock still has references
4135 Details : In some error handling cases it was possible to leak a lock
4136 reference on a client while accessing a file. This was not
4137 harmful to the client during operation, but would cause the
4138 client to crash when the filesystem is unmounted.
4141 Frequency : specific case, rare
4143 Description: ETXTBSY on mds though file not in use by client
4144 Details : ETXTBSY is no longer incorrectly returned when attempting to
4145 chmod or chown a directory that the user previously tried to
4146 execute or a currently-executing binary.
4149 Frequency : extremely rare except on liblustre-based clients
4151 Description: Lustre space not freed when files are deleted
4152 Details : Clean up open-unlinked files after client eviction. Previously
4153 the unlink was skipped and the files remained as orphans.
4158 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4159 Details : In some cases it was possible to send a blocking callback to
4160 a client doing a glimpse, even though that client didn't get
4161 a lock granted. When the glimpse lock is cancelled on the OST
4162 the freed lock is left on the waiting list and corrupted the list.
4165 Frequency : all core dumps
4167 Description: Broke core dumps to lustre
4168 Details : Negative dentry may be unhashed if parent does not have UPDATE
4169 lock, but some callers, e.g. do_coredump, expect dentry to be
4170 hashed after successful create, hash it in ll_create_it.
4172 ------------------------------------------------------------------------------
4174 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4176 * Support for kernels:
4177 2.6.9-42.0.2.EL (RHEL 4)
4178 2.6.5-7.276 (SLES 9)
4179 2.4.21-40.EL (RHEL 3)
4180 2.6.12.6 vanilla (kernel.org)
4184 Frequency : always on RHEL 3
4186 Description: Number of open files grows over time
4187 Details : The number of open files grows over time, whether or not
4188 Lustre is started. This was due to a filp leak introduced
4189 by one of our kernel patches.
4191 ------------------------------------------------------------------------------
4193 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4195 * Support for kernels:
4196 2.6.9-42.EL (RHEL 4)
4197 2.6.5-7.276 (SLES 9)
4198 2.4.21-40.EL (RHEL 3)
4199 2.6.12.6 vanilla (kernel.org)
4204 Bugzilla : 5719, 9635, 9792, 9684
4205 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4206 Details : If a server is under heavy load and cannot reply to new
4207 connection requests before the client resends the (re)connect,
4208 the connection handling code can behave badly if two service
4209 threads are concurrently handing separate (re)connections from
4210 the same client. Add better locking to the connection handling
4211 code, and ensure that only a single connection will be processed
4212 for a given client UUID, even if the lock is dropped.
4214 Severity : enhancement
4216 Description: add TCP zero-copy support to kernel
4217 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4218 sends if the hardware supports scatter-gather and checksumming.
4219 This allows socklnd to do client-write and server-read more
4220 efficiently and reduce CPU utilization from skbuf copying.
4223 Frequency : only if NFS exporting from client
4225 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4226 Details : If setting attributes on a file created under NFS that had
4227 never been opened it would be possible to oops the client
4228 if the file had no objects.
4231 Frequency : always for liblustre
4233 Description: liblustre client does MDS+OSTs setattr RPC for each write
4234 Details : When doing a write from a liblustre client, the client
4235 incorrectly issued an RPC to the MDS and each OST the file was
4236 striped over in order to update the timestamps. When writing
4237 with small chunks and many clients this could overwhelm the MDS
4238 with RPCs. In all cases it would slow down the write because
4239 these RPCs are unnecessary.
4241 Severity : enhancement
4243 Description: allow number of MDS service threads to be changed at module load
4244 Details : It is now possible to change the number of MDS service threads
4245 running. Adding "options mds mds_num_threads={N}" to the MDS's
4246 /etc/modprobe.conf will set the number of threads for the next
4247 time Lustre is restarted (assuming the "mds" module is also
4248 reloaded at that time). The default number of threads will
4249 stay the same, 32 for most systems.
4254 Description: OST crash if filesystem is unformatted or corrupt
4255 Details : If an OST is started on a device that has never been formatted
4256 or if the filesystem is corrupt and cannot even mount then the
4257 error handling cleanup routines would dereference a NULL pointer.
4262 Description: NULL pointer deref in llap_from_page.
4263 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4264 incorrect) mapping if the page was truncated/reclaimed while it was
4265 searched for. Check for this condition and skip such pages when
4266 doing readahead. Introduce extra check to llap_from_page() to
4267 verify page->mapping->host is non-NULL (so page is not anonymous).
4270 Frequency : Sometimes when using sys_sendfile
4272 Description: "page not covered by a lock" warnings from ll_readpage
4273 Details : sendfile called ll_readpage without right page locks present.
4274 Now we introduced ll_file_sendfile that does necessary locking
4275 around call to generic_file_sendfile() much like we do in
4279 Frequency : with certain MDS communication failures at client mount time
4281 Description: NULL pointer deref after failed client mount
4282 Details : a client connection request may delayed by the network layer
4283 and not be sent until after the PTLRPC layer has timed out the
4284 request. If the client fails the mount immediately it will try
4285 to clean up before the network times out the request. Add a
4286 reference from the request import to the obd device and delay
4287 the cleanup until the network drops the request.
4290 Frequency : occasionally during client (re)connect
4292 Description: assertion failure during client (re)connect
4293 Details : processing a client connection request may be delayed by the
4294 client or server longer than the client connect timeout. This
4295 causes the client to resend the connection request. If the
4296 original connection request is replied in this interval, the
4297 client may trip an assertion failure in ptlrpc_connect_interpret()
4298 which thought it would be the only running connect process.
4301 Frequency : only with obd_echo servers and clients that are rebooted
4303 Description: kernel BUG accessing uninitialized data structure
4304 Details : When running an obd_echo server it did not start the ping_evictor
4305 thread, and when a client was evicted an uninitialized data
4306 structure was accessed. Start the ping_evictor in the RPC
4307 service startup instead of the OBD startup.
4309 Severity : enhancement
4310 Bugzilla : 10193 (patchless)
4311 Description: Remove dependency on various unexported kernel interfaces.
4312 Details : No longer need reparent_to_init, exit_mm, exit_files,
4313 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4316 Frequency : rare (only users of deprecated and unsupported LDAP config)
4318 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4319 Details : LDAP apparently contains 'lustreName' attributes instead of
4320 'name'. A simple remapping of the name is sufficient.
4323 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4325 Description: waiting_locks_callback trips kernel BUG if client is evicted
4326 Details : Running with the dump_on_timeout debug flag turned on makes
4327 it possible that the waiting_locks_callback() can try to dump
4328 the Lustre kernel debug logs from an interrupt handler. Defer
4329 this log dumping to the expired_lock_main() thread.
4331 Severity : enhancement
4333 Description: Support NFS exporting on 2.6 kernels.
4334 Details : Implement non-rawops metadata methods for NFS server to use without
4335 changing NFS server code.
4338 Frequency : very rare (synthetic metadata workload only)
4340 Description: two racing renames might cause an MDS thread to deadlock
4341 Details : Running the "racer" program may cause one MDS thread to rename
4342 a file from being the source of a rename to being the target of
4343 a rename at exactly the same time that another thread is doing
4344 so, and the second thread has already enqueued these locks after
4345 doing a lookup of the target and is trying to relock them in
4346 order. Ensure that we don't try to re-lock the same resource.
4349 Frequency : only very large systems with liblustre clients
4351 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4352 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4353 all OSTs in parallel. This allows the eviction of stale liblustre
4354 clients to proceed much faster than if they were done in series,
4355 and also offers similar improvements for other set_info RPCs.
4360 Description: excessive CPU usage during initial read phase on client
4361 Details : During the initial read phase on a client, it would agressively
4362 retry readahead on the file, consuming too much CPU and impacting
4363 performance (since 1.4.5.8). Improve the readahead algorithm
4364 to avoid this, and also improve some other common cases (read
4365 of small files in particular, where "small" is files smaller than
4366 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4371 Description: MDS crash when receiving packet with unknown intent.
4372 Details : Do not LBUG in unknown intent case, just return -EFAULT
4374 Severity : enhancement
4375 Bugzilla : 9293, 9385
4376 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4377 Details : Do not serialize getattr (non-intent version) and statfs.
4380 Frequency : occasional, when OST network is overloaded/intermittent
4382 Description: client evicted by OST after bulk IO timeout
4383 Details : If a client sends a bulk IO request (read or write) the OST
4384 may evict the client if it is unresposive to its data GET/PUT
4385 request. This is incorrect if the network is overloaded (takes
4386 too long to transfer the RPC data) or dropped the OST GET/PUT
4387 request. There is no need to evict the client at all, since
4388 the pinger and/or lock callbacks will handle this, and the
4389 client can restart the bulk request.
4392 Frequency : Always when mmapping file with no objects
4394 Description: client crashes when mmapping file with no objects
4395 Details : Check that we actually have objects in a file before doing any
4396 operations on objects in ll_vm_open, ll_vm_close and
4402 Description: Request leak when working with deleted CWD
4403 Details : Introduce advanced request refcount tracking for requests
4404 referenced from lustre intent.
4406 Severity : Enhancement
4408 Description: Cache open file handles on client.
4409 Details : MDS now will return special lock along with openhandle, if
4410 requested and client is allowed to hold openhandle, even if unused,
4411 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4412 closing files for every read/write openration.
4414 Severity : Enhancement
4416 Description: Cache open negative dentries on client when possible.
4417 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4418 negative dentries on lock revocation.
4423 Description: Remounting a client read-only wasn't possible with a zconf mount
4424 Details : It wasn't possible to remount a client read-only with llmount.
4426 Severity : enhancement
4427 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4428 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4429 MPICH is included to add Lustre-specific ADIO interfaces.
4430 This is based closely on the UFS ADIO layer and only differs
4431 in file creation, in order to allow the OST striping to be set.
4432 This is user-contributed code and not supported by CFS.
4437 Description: extended inode attributes (immutable, append-only) work improperly
4438 when 2.4 and 2.6 kernels are used on client/server or vice versa
4439 Details : Introduce kernel-independent values for these flags.
4441 Severity : enhancement
4444 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4445 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4446 granularity. Now, Fractional MB granularity is be supported,
4447 this is very useful for low memory system.
4449 Severity : enhancement
4451 Description: Getattr by fid
4452 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4453 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4454 have locks and avoids getattr rpc requests for it on every stat.
4457 Frequency : Always, for filesystems larger than 2TB
4459 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4460 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4461 filesystems larger than 2TB to be mounted with mballoc enabled.
4464 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4466 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4467 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4468 filesystems larger than 2TB is mounted then the kernel will
4469 silently corrupt the start of the filesystem. CONFIG_LBD is
4470 enabled for all CFS-supported kernels, but the possibility of
4471 this happening with a modified kernel config exists.
4473 Severity : enhancement
4475 Description: add client O_DIRECT support for 2.6 kernels
4476 Details : It is now possible to do O_DIRECT reads and writes to files
4477 in the Lustre client mountpoint on 2.6 kernel clients.
4479 Severity : enhancement
4481 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4482 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4483 parallel, not waiting for response from every OST before sending
4484 a rpc to the next OST.
4489 Description: setattr vs write race when updating file timestamps
4490 Details : Client processes that update a file timestamp into the past
4491 right after writing to the file (e.g. tar) it is possible that
4492 the updated file modification time can be reset to the current
4493 time due to a race between processing the setattr and write RPC.
4495 Severity : enhancement
4497 Description: Bring 'lfs find' closer in line with regular Linux find.
4498 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4499 -print0 options and obtains all the needed info through the lustre
4502 Severity : enhancement
4504 Description: support up to 1024 configured devices on one node
4505 Details : change obd_dev array from statically allocated to dynamically
4506 allocated structs as they are first used to reduce memory usage
4511 Description: Flush dirty partially truncated pages during truncate
4512 Details : Immediatelly flush partially truncated pages in filter_setattr,
4513 this way we completely avoid having any pages in page cache on OST
4514 and can retire ugly workarounds during writes to flush such pages.
4519 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4520 Details : protect inode from truncation within vfs_unlink() context
4521 just take a reference before calling vfs_unlink() and release it
4522 when parent's i_sem is free.
4527 Description: last_id value checked outside lock on OST caused LASSERT failure
4528 Details : If there were multiple MDS->OST object precreate requests in
4529 flight, it was possible that the OST's last object id was checked
4530 outside a lock and incorrectly tripped an assertion. Move checks
4531 inside locks, and discard old precreate requests.
4534 Frequency : always, if extents are used on OSTs
4536 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4537 Details : OSTs using the extents format would not zero the high 16 bits of
4538 the index physical block number. This is not a problem for any
4539 OST filesystems smaller than 16TB, and no kernels support ext3
4540 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4541 new/modified files) and can be fixed for existing filesystems
4542 with e2fsprogs-1.39-cfs1.
4547 Description: import connection selection may be incorrect if timer wraps
4548 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4549 import connections to be ignored if the 32-bit jiffies counter
4550 wraps. Use a 64-bit jiffies counter.
4553 Frequency : very large clusters immediately after boot
4555 Description: LNET request buffers exhausted under heavy short-term load
4556 Details : If a large number of client requests are generated on a service
4557 that has previously never seen so many requests it is possible
4558 that the request buffer growth cannot keep up with the spike in
4559 demand. Instead of dropping incoming requests, they are held in
4560 the LND until the RPC service can accept more requests.
4563 Frequency : Sometimes during replay
4565 Description: Assertion failure in ll_local_open after replay.
4566 Details : If replay happened on an open request reply before we were able
4567 to set replay handler, reply will become not swabbed tripping the
4568 assertion in ll_local_open. Now we set the handler right after
4569 recognising of open request
4572 Frequency : very rare
4574 Description: kernel reports "badness in vsnprintf"
4575 Details : Reading from the "recovery_status" /proc file in small chunks
4576 may cause a negative length in lprocfs_obd_rd_recovery_status()
4577 call to vsnprintf() (which is otherwise harmless). Exit early
4578 if there is no more space in the output buffer.
4580 Severity : enhancement
4582 Description: clear OBD RPC statistics by writing to them
4583 Details : It is now possible to clear the OBD RPC statistics by writing
4584 to the "stats" file.
4589 Description: Client mtime is not the same on different clients after utimes
4590 Details : In some cases, the client was using the utimes() syscall on
4591 a file cached on another node. The clients now validate the
4592 ctime from the MDS + OSTs to determine which one is right.
4597 Description: Inability to activate failout mode
4598 Details : lconf script incorrectly assumed that in pythong string's numeric
4599 value is used in comparisons.
4602 Frequency : always with multiple stripes per file
4604 Description: Inefficient object allocation for mutli-stripe files
4605 Details : When selecting which OSTs to stripe files over, for files with
4606 a stripe count that divides evenly into the number of OSTs,
4607 the MDS is always picking the same starting OST for each file.
4608 Return the OST selection heuristic to the original design.
4613 Description: mount failures may take full timeout to return an error
4614 Details : Under some heavy load conditions it is possible that a
4615 failed mount can wait for the full obd_timeout interval,
4616 possibly several minutes, before reporting an error.
4617 Instead return an error as soon as the status is known.
4619 Frequency : quota enabled and large files being deleted
4621 Description: releasing more than 4GB of quota at once hangs OST
4622 Details : If a user deletes more than 4GB of files on a single OST it
4623 will cause the OST to spin in an infinite loop. Release
4624 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4629 Description: statfs data retrieved from /proc may be stale or zero
4630 Details : When reading per-device statfs data from /proc, in the
4631 {kbytes,files}_{total,free,avail} files, it may appear
4632 as zero or be out of date.
4635 Frequency : systems with MD RAID1 external journal devices
4637 Description: lconf's call to blkid is confused by RAID1 journal devices
4638 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4639 of returning all block devices that match the journal UUID.
4642 Frequency : always, for aggregate stripe size over 4GB
4644 Description: assertion fails when trying to use 4GB stripe size
4645 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4646 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4649 Frequency : always on ppc64
4651 Description: the first write on an ext3 filesystem with mballoc got stuck
4652 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4653 perform endianness conversion.
4655 ------------------------------------------------------------------------------
4657 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4659 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4660 INTEROPERATE with older versions automatically. Please read the
4661 user documentation before upgrading any part of a live system.
4662 * WARNING: Lustre networking configuration changes are required with
4663 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4666 * Support for kernels:
4667 2.6.9-22.0.2.EL (RHEL 4)
4668 2.6.5-7.244 (SLES 9)
4669 2.6.12.6 vanilla (kernel.org)
4672 Severity : enhancement
4673 Bugzilla : 7981/8208
4674 Description: Introduced Lustre Networking (LNET)
4675 Details : LNET is new networking infrastructure for Lustre, it includes
4676 a reorganized network configuration mode (see the user
4677 documentation for full details) as well as support for routing
4678 between different network fabrics. Lustre Networking Devices
4679 (LNDs) for the supported network fabrics have also been
4680 created for this new infrastructure.
4682 Severity : enhancement
4683 Description: Introduced Access control lists
4684 Details : clients can set ACLs on files and directories in order to have
4685 more fine-grained permissions than the standard Unix UGO+RWX.
4686 The MDS must be started with the "-o acl" mount option.
4688 Severity : enhancement
4689 Description: Introduced filesystem quotas
4690 Details : Administrators may now establish per-user quotas on the
4693 Severity : enhancement
4695 Description: Configuration change for the XT3
4696 The PTLLND is now used to run Lustre over Portals on the XT3
4697 The configure option(s) --with-cray-portals are no longer used.
4698 Rather --with-portals=<path-to-portals-includes> is used to
4699 enable building on the XT3. In addition to enable XT3 specific
4700 features the option --enable-cray-xt3 must be used.
4705 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4706 Details : A race between allocating a new llcd and re-getting the llcd_lock
4707 allowed another thread to grab newly-allocated llcd.
4709 Severity : enhancement
4711 Description: 2.6 OST async journal commit and locking fix to improve performance
4712 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4713 2.6 kernels are now async as they already were in 2.4 kernels so
4714 that they can commit concurrently with the network bulk transfer.
4715 For block-allocated files the filter allocation semaphore is held
4716 to avoid filesystem fragmentation during allocation. BKL lock
4717 removed for 2.6 xattr operations where it is no longer needed.
4722 Description: lconf incorrectly determined whether two IP networks could talk
4723 Details : In some more complicated routing and multiple-network
4724 configurations, lconf will avoid trying to make a network
4725 connection to a disjoint part of the IP space. It was doing the
4726 math incorrectly for one set of cases.
4731 Description: Fix for potential infinite loop processing records in an llog.
4732 Details : If an llog record is corrupted/zeroed, it is possible to loop
4733 forever in llog_process(). Validate the llog record length
4734 and skip the remainder of the block on error.
4737 Frequency : occasional (liblustre only)
4739 Description: liblustre could not open files whose last component is a symlink
4740 Details : sysio_path_walk() would incorrectly pass the open intent to
4741 intermediate path components.
4744 Frequency : rare (liblustre only with non-standard tuning)
4745 Bugzilla : 7201 (7350)
4746 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4747 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4748 after it has been released, only do this for referenced locks
4750 Severity : enhancement
4752 Description: specifying an (invalid) directory default stripe_size of -1
4753 would reset the directory default striping
4754 Details : stripe_size -1 was used internally to signal directory stripe
4755 removal, now use "all default" to signal dir stripe removal
4756 as a directory striping of "all default" is not useful
4759 Frequency : common for large clusters running liblustre clients
4761 Description: doing an ls when liblustre clients are running is slow
4762 Details : sending a glimpse AST to a liblustre client waits for every AST
4763 to time out, as liblustre clients will not respond. Since they
4764 cannot cache data we refresh the OST lock LVB from disk instead.
4766 Severity : enhancement
4768 Description: doing an ls at the same time as file IO can be slow
4769 Details : enqueue and other "small" requests can be blocked behind many
4770 large IO requests. Create a new OST IO portal for non-IO
4771 requests so they can be processed faster.
4774 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4776 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4777 created file with mode 000
4778 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4779 allow an ftruncate() as a non-root user to a file with mode 000.
4780 Lustre now respects this flag to disable mode checks when
4781 truncating a file owned by the user
4784 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4787 Description: Revoking locks from clients that went dead or catatonic might take
4789 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4790 cancellation of such locks instant on servers without waiting for
4791 any reply from clients. Clients drops these locks when cancel
4792 notification from server is received without replying.
4795 Frequency : liblustre-only, when liblustre client dies or becomes busy
4797 Description: Doing ls on Linux clients can take a long time with active
4799 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4800 granting such locks to it in the first place if possible. Locks
4801 are taken by proxy on the OST during the read or write and
4802 dropped immediately afterward. Add connect flags handling, do
4803 not grant locks to liblustre clients for glimpse ASTs.
4805 Severity : enhancement
4807 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4808 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4809 sequential reads and behaves badly with applications doing random
4810 reads. Improve it by reading ahead at least read region, and
4811 avoiding excessive large RPC for small reads.
4813 Severity : enhancement
4815 Description: Creating more than 1000 files for a single job may cause a load
4816 imbalance on the OSTs if there are also a large number of OSTs.
4817 Details : qos_prep_create() uses an OST index reseed value that is an
4818 even multiple of the number of available OSTs so that if the
4819 reseed happens in the middle of the object allocation it will
4820 still utilize the OSTs as uniformly as possible.
4825 Description: OST or MDS may oops in ping_evictor_main()
4826 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4827 but doesn't restart at beginning of obd_exports_timed list
4830 Severity : enhancement
4832 Description: improve by-nid export eviction on the MDS and OST
4833 Details : allow multiple exports with the same NID to be evicted at one
4834 time without re-searching the exports list.
4837 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4839 Description: MDS may oops in groups_free()
4840 Details : in rare race conditions a newly allocated group_info struct is
4841 freed again, and this can be NULL. The 2.4 compatibility code
4842 for groups_free() checked for a NULL pointer, but 2.6 did not.
4845 Frequency : common for liblustre clients doing little filesystem IO
4846 Bugzilla : 9352, 7313
4847 Description: server may evict liblustre clients accessing contended locks
4848 Details : if a client is granted a lock or receives a completion AST
4849 with a blocking AST already set it would not reply to the AST
4850 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4853 Frequency : lfs setstripe, only systems with more than 160 OSTs
4855 Description: unable to set striping with a starting offset beyond OST 160
4856 Details : llapi_create_file() incorrectly limited the starting stripe
4857 index to the maximum single-file stripe count.
4860 Frequency : LDAP users only
4862 Description: lconf did not handle in-kernel recovery with LDAP properly
4863 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4865 Severity : enhancement
4867 Description: bind OST threads to NUMA nodes to improve performance
4868 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4869 node and do their allocations there to localize memory access
4871 Severity : enhancement
4873 Description: llmount can determine client NID directly from Myrinet (GM)
4874 Details : the client NID code from gmnalnid was moved directly into
4875 llmount, removing the need to use this or specifying the
4876 client NID explicitly when mounting GM clients with zeroconf
4879 Frequency : if client is started with down MDS
4881 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4882 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4883 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4884 ptlrpc_import_delayed_req() because we didn't check if the
4885 request was interrupted, and we also didn't break out of the
4886 event loop if there was no timeout
4891 Description: data loss during non-page-aligned writes to a single file from
4892 both multiple nodes and multiple threads on one node at same time
4893 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4894 inconsistency led to false short-reads, that were cached and later
4895 used by ->prepare_write() to fill in partially written page,
4896 leading to data loss.
4899 Frequency : always, if lconf --abort_recovery used
4901 Description: lconf --abort_recovery fails with 'Operation not supported'
4902 Details : lconf was attempting to abort recovery on the MDT device and not
4905 Severity : enhancement
4907 Description: remove cleanup logs
4908 Details : replace lconf-generated cleanup logs with lustre internal
4909 cleanup routines. Eliminates the need for client-cleanup and
4912 Severity : enhancement
4914 Description: add support for EAs (user and system) on lustre filesystems
4915 Details : it is now possible to store extended attributes in the Lustre
4916 client filesystem, and with the user_xattr mount option it
4917 is possible to allow users to store EAs on their files also
4919 Severity : enhancement
4921 Description: Add possibility (config option) to show minimal available OST free
4923 Details : When compiled with --enable-mindf configure option, statfs(2)
4924 (and so, df) will return least minimal free space available from
4925 all OSTs as amount of free space on FS, instead of summary of
4926 free spaces of all OSTs.
4928 Severity : enhancement
4930 Description: do not expand extent locks acquired on OST-side
4931 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4932 by server: they are not cached anyway.
4935 Frequency : when mmap is used/binaries executed from Lustre
4937 Description: Unmmap pages before throwing them away from read cache.
4938 Details : llap_shrink cache now attempts to unmap pages before discarding
4939 them (if unmapping failed - do not discard). SLES9 kernel has
4940 extra checks that trigger if this unmapping is not done first.
4945 Description: lconf didn't resolve symlinks before checking to see whether a
4946 given mountpoint was already in use
4949 Frequency : when migrating failover services
4950 Bugzilla : 6395, 9514
4951 Description: When migrating a subset of services from a node (e.g. failback
4952 from a failover service node) the remaining services would
4953 time out and evict clients.
4954 Details : lconf --force (implied by --failover) sets the global obd_timeout
4955 to 5 seconds in order to quickly disconnect, but this caused
4956 other RPCs to time out too quickly. Do not change the global
4957 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4959 Severity : enhancement
4960 Frequency : if MDS is started with down OST
4961 Bugzilla : 9439,5706
4962 Description: Allow startup/shutdown of an MDS without depending on the
4963 availability of the OSTs.
4964 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4965 Add appropriate locking and lov-osc refcounts for safe
4966 cleaning. Add osc abort_inflight calls in case the
4967 synchronize never started.
4970 Frequency : occasional (Cray XT3 only)
4972 Description: root not authorized to access files in CRAY_PORTALS environment
4973 Details : The client process capabilities were not honoured on the MDS in
4974 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4975 been accessed by an authorized user then root was able to access
4976 the file on the local client also. The root user capabilities
4977 are now allowed on the MDS, as this environment has secure UID.
4980 Frequency : occasional
4982 Description: ldiskfs "too long searching" message happens too often
4983 Details : A debugging message (otherwise harmless) prints too often on
4984 the OST console. This has been reduced to only happen when
4985 there are fragmentation problems on the filesystem.
4990 Description: Division by zero in statfs when all OSCs are inactive
4991 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4992 lov_statfs divides by value returned by lov_get_stripecnt().
4996 Bugzilla : 9489, 3273
4997 Description: First write from each client to each OST was only 4kB in size,
4998 to initialize client writeback cache, which caused sub-optimal
4999 RPCs and poor layout on disk for the first written file.
5000 Details : Clients now request an initial cache grant at (re)connect time
5001 and so that they can start streaming writes to the cache right
5002 away and always do full-sized RPCs if there is enough data.
5003 If the OST is rebooted the client also re-establishes its grant
5004 so that client cached writes will be honoured under the grant.
5009 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5010 Details : Now I/O RPCs go to different portal number and (presumably) fast
5011 lock requests (and glimses) and other RPCs get their own service
5012 threads pool that should be able to service those RPCs
5015 Severity : enhancement
5017 Description: Ability to exchange lustre version between client and servers and
5018 issue warnings at client side if client is too old. Also for
5019 liblustre clients there is ability to refuse connection of too old
5021 Details : New 'version' field is added to connect data structure that is
5022 filled with version info. That info is later checked by server and
5026 Frequency : rare, liblustre only.
5027 Bugzilla : 9296, 9581
5028 Description: Two simultaneous writes from liblustre at offset within same page
5029 might proceed at the same time overwriting eachother with stale
5031 Details : I/O lock withing llu_file_prwv was released too early, before data
5032 actually was hitting the wire. Extended lock-holding time until
5033 server acknowledges receiving data.
5036 Frequency : extremely rare. Never observed in practice.
5038 Description: avoid generating lustre_handle cookie of 0.
5039 Details : class_handle_hash() generates handle cookies by incrementing
5040 global counter, and can hit 0 occasionaly (this is unlikely, but
5041 not impossible, because initial value of cookie counter is
5042 selected randonly). Value of 0 is used as a sentinel meaning
5043 "unassigned handle" --- avoid it. Also coalesce two critical
5044 sections in this function into one.
5046 Severity : enhancement
5048 Description: allow liblustre clients to delegate truncate locking to OST
5049 Details : To avoid overhead of locking, liblustre client instructs OST to
5050 take extent lock in ost_punch() on client's behalf. New connection
5051 flag is added to handle backward compatibility.
5053 Severity : enhancement
5054 Bugzilla : 4928, 7341, 9758
5055 Description: allow number of OST service threads to be specified
5056 Details : a module parameter allows the number of OST service threads
5057 to be specified via "options ost ost_num_threads={N}" in the
5058 OSS's /etc/modules.conf or /etc/modprobe.conf.
5062 Bugzilla : 6146, 9635, 9895
5063 Description: servers crash with bad pointer in target_handle_connect()
5064 Details : In rare cases when a client is reconnecting it was possible that
5065 the connection request was the last reference for that export.
5066 We would temporarily drop the export reference and get a new
5067 one, but this may have been the last reference and the export
5068 was just destroyed. Get new reference before dropping old one.
5070 Severity : enhancement
5071 Frequency : if client is started with failover MDS
5073 Description: Allow multiple MDS hostnames in the mount command
5074 Details : Try to read the configuration from all specified MDS
5075 hostnames during a client mount in case the "primary"
5078 Severity : enhancement
5080 Description: Stop sending data to evicted clients as soon as possible.
5081 Details : Check if the client we are about to send or are sending data to
5082 was evicted already. (Check is done every second of waiting,
5083 for which l_wait_event interface was extended to allow checking
5084 of exit condition at specified intervals).
5087 Frequency : rare, normally only when NFS exporting is done from client
5089 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5091 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5094 Frequency : always, if srand() is called before liblustre initialization
5096 Description: Liblustre uses system PRNG disturbing its usage by user application
5097 Details : Introduce internal to lustre fast and high-quality PRNG for
5098 lustre usage and make liblustre and some other places in generic
5099 lustre code to use it.
5101 Severity : enhancement
5102 Bugzilla : 9477, 9557, 9870
5103 Description: Verify that the MDS configuration logs are updated when xml is
5104 Details : Check if the .xml configuration logs are newer than the config
5105 logs stored on the MDS and report an error if this is the case.
5106 Request --write-conf, or allow starting with --old_conf.
5108 Severity : enhancement
5110 Description: Handle symlinks in the path when checking if Lustre is mounted.
5111 Details : Resolve intermediate symlinks when checking if a client has
5112 mounted a filesystem to avoid duplicate client mounts.
5117 Description: lconf can hit an error exception but still return success.
5118 Details : The lconf command catches the Command error exception at the top
5119 level script context and will exit with the associated exit
5120 status, but doesn't ensure that this exit status is non-zero.
5125 Description: failure of ptlrpc thread startup can cause oops
5126 Details : Starting a ptlrpc service thread can fail if there are a large
5127 number of threads or the server memory is very fragmented.
5128 Handle this without oopsing.
5131 Frequency : always, only if liblustre and non-default acceptor port was used
5133 Description: liblustre cannot connect to servers with non-default acceptor port
5134 Details : tcpnal_set_default_params() was not called and was therefore
5135 ignoring the environment varaible TCPNAL_PORT, as well as other
5136 TCPNAL_ environment variables
5141 Description: two objects could be created on the same OST for a single file
5142 Details : If an OST is down, in some cases it was possible to create two
5143 objects on a single OST for a single file. No problems other
5144 than potential performance impact and spurious error messages.
5148 Bugzilla : 5681, 9562
5149 Description: Client may oops in ll_unhash_aliases
5150 Details : Client dcache may become inconsistent in race condition.
5151 In some cases "getcwd" can fail if the current directory is
5157 Description: Inode refcounting problems in NFS export code
5158 Details : link_raw functions used to call d_instantiate without obtaining
5159 extra inode reference first.
5163 Bugzilla : 9942, 9903
5164 Description: Referencing freed requests leading to crash, memleaks with NFS.
5165 Details : We used to require that call to ll_revalidate_it was always
5166 followed by ll_lookup_it. Also with revalidate_special() it is
5167 possible to call ll_revalidate_it() twice for the same dentry
5168 even if first occurence returned success. This fix changes semantic
5169 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5170 reference on a request referred from the intent.
5171 ll_intent_release() then releases such a request.
5174 Frequency : rare, normally benchmark loads only
5176 Description: unlinked inodes were kept in memory on the client
5177 Details : If a client is repeatedly creating and unlinking files it
5178 can accumulate a lot of stale inodes in the inode slab cache.
5179 If there is no other client load running this can cause the
5180 client node to run out of memory. Instead flush old inodes
5181 from client cache that have the same inode number as a new inode.
5184 Frequency : SLES9 2.6.5 kernel and long filenames only
5185 Bugzilla : 9969, 10379
5186 Description: utime reports stale NFS file handle
5187 Details : SLES9 uses out-of-dentry names in some cases, which confused
5188 the lustre dentry revalidation. Change it to always use the
5192 Frequency : rare, unless heavy write-truncate concurrency is continuous
5193 Bugzilla : 4180, 6984, 7171, 9963, 9331
5194 Description: OST becomes very slow and/or deadlocked during object unlink
5195 Details : filter_destroy() was holding onto the parent directory lock
5196 while truncating+unlinking objects. For very large objects this
5197 may block other threads for a long time and slow overall OST
5198 responsiveness. It may also be possible to get a lock ordering
5199 deadlock in this case, or run out of journal credits because of
5200 the combined truncate+unlink. Solution is to do object truncate
5201 first in one transaction without parent lock, and then do the
5202 final unlink in a new transaction with the parent lock. This
5203 reduces the lock hold time dramatically.
5206 Frequency : rare, 2.4 kernels only
5208 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5209 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5210 before unmounting their filesystem in order to avoid deadlock.
5211 The kernel_locked() function in 2.4 kernels only checks whether
5212 the kernel lock is held, not whether it is this process that is
5213 holding it as 2.6 kernels do.
5218 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5219 Details : The client ptlrpc code may be trying to reconnect to a down
5220 server before a previous connection attempt has timed out.
5221 Increase the reconnect interval to be longer than the connection
5222 timeout interval to avoid sending duplicate connections to
5226 Frequency : echo_client brw_test command
5228 Description: fix echo_client to work with OST preallocated code
5229 Details : OST preallocation code (5137) didn't take echo_client IO path
5230 into account: echo_client calls filter methods outside of any
5231 OST thread and, hence, there is no per-thread preallocated
5232 pages and buffers to use. Solution: hijack pga pages for IO. As
5233 a byproduct, this avoids unnecessary data copying.
5237 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5238 Description: Client can oops in mdc_commit_close() after open replay
5239 Details : It was possible for the MDS to return an open request with no
5240 transaction number in mds_finish_transno() if the client was
5241 evicted, but without actually returning an error. Clients
5242 would later try to replay that open and may trip an assertion
5243 Simplify the client close codepath, and always return an error
5244 from the MDS in case the open is not successful.
5247 Frequency : rare, 2.6 OSTs only
5249 Description: OST may deadlock under high load on fragmented files
5250 Details : If there was a heavy load and highly-fragmented OST filesystems
5251 it was possible to have all the OST threads deadlock waiting on
5252 allocation of biovecs, because the biovecs were not released
5253 until the entire RPC IO was completed. Instead, release biovecs
5254 as soon as they are complete to ensure forward IO progress.
5256 Severity : enhancement
5258 Description: Support for specifying external journal device at mount
5259 Details : If an OST or MDS device is formatted with an external journal
5260 device, this device major/minor is stored in the ext3 superblock
5261 and may not be valid for failover. Allow detecting and
5262 specifying the external journal at mount time.
5267 Description: Mounting an MDS with pending unlinked files may cause oops
5268 Details : target_finish_recovery() calls mds_postrecov() which returned
5269 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5270 considers this an error and immediately begins cleaning up the
5271 lov, just after starting the mds_lov process
5273 Severity : enhancement
5275 Description: Implement 'lfs df' to report actual free space on per-OST basis
5276 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5277 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5278 usage of objects; '-h' to report in human readable format.
5280 ------------------------------------------------------------------------------
5282 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5289 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5290 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5291 at startup during orphan unlinking. Instead, skip buddy update
5292 before setup, as it will be regenerated after recovery is complete.
5297 Description: If an OST is inactive, its locks might reference stale inodes.
5298 Details : lov_change_cbdata() must iterate over all namespaces, even if
5299 they are inactive to clear inode references from the lock.
5301 Severity : enhancement
5302 Frequency : occasional, if non-standard max_dirty_mb used
5304 Description: Client will block write RPCs if not enough grant
5305 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5306 then the client will block writes while waiting for another RPC
5307 to complete instead of consuming its dirty limit. With change
5308 we get improved performance when max_dirty_mb is small.
5310 Severity : enhancement
5311 Bugzilla : 3389, 6253
5312 Description: Add support for supplementary groups on the MDS.
5313 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5314 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5315 lookups for user supplementary groups into a cache.
5319 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5320 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5321 for liblustre O_EXCL is enforced by MDS.
5325 Description: readdir never returns NULL in liblustre.
5326 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5327 dentry in struct dirent.
5331 Frequency : liblustre only, depends on application IO pattern
5332 Description: liblustre clients evicted if not contacting servers
5333 Details : Don't put liblustre clients into the ping_evictor list, so
5334 they will not be evicted by the pinger ever.
5336 Severity : enhancement
5338 Description: Add ability to evict clients by NID from MDS.
5339 Details : By echoing "nid:$NID" string into
5340 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5341 $NID would be instantly evicted from this MDS and from all active
5342 OSTs connected to it.
5346 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5347 Details : lookup_it_finish() used to query file size from OSTs that was not
5352 Description: service threads change working directory to that of init
5353 Details : Starting lustre service threads may pin the working directory
5354 of the parent thread, making that filesystem busy. Threads
5355 now change to the working directory of init to avoid this.
5359 Frequency : during shutdown only
5360 Description: shutdown with a failed MDS or OST can cause unmount to hang
5361 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5366 Frequency : 2.6 only
5367 Description: chmod/chown may include an extra supplementary group
5368 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5369 supplementary group and if none is specified this is used.
5372 Bugzilla : 5479 (6816)
5374 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5375 Details : If lookup is in progress on a file that is unlinked we might try
5376 to revalidate the inode and fail in revalidate after lookup is
5377 complete and ll_file_open() enqueues the open again but
5378 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5381 Frequency : always, if lconf --abort_recovery used
5383 Description: lconf --abort_recovery fails with 'Operation not supported'
5384 Details : lconf was attempting to abort recovery on the MDT device and not
5387 ------------------------------------------------------------------------------
5389 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5394 Frequency : rare (only unsupported configurations with a node running as an
5396 Bugzilla : 6514, 5137
5397 Description: Mounting a Lustre file system on a node running as an OST could
5399 Details : OSTs now preallocates memory needed to write out data at
5400 startup, instead of when needed, to avoid having to
5401 allocate memory in possibly low memory situations.
5402 Specifically, if the file system is mounted on on OST,
5403 memory pressure could force it to try to write out data,
5404 which it needed to allocate memory to do. Due to the low
5405 memory, it would be unable to do so and the node would
5406 become unresponsive.
5408 Severity : enhancement
5410 Description: Addition of lconf --service command line option
5411 Details : lconf now accepts a '--service <arg>' option, which is
5412 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5414 Severity : enhancement
5416 Description: Failover mode is now the default for OSTs.
5417 Details : By default, OSTs will now run in failover mode. To return to
5418 the old behaviour, add '--failout' to the lmc line for OSTs.
5420 Severity : enhancement
5422 Description: Health checks are now provided for MDS and OSTs
5423 Details : Additional detailed health check information on MSD and OSTs
5424 is now provided through the procfs health_check value.
5427 Frequency : occasional, depends on IO load
5429 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5430 after numerous create/delete cycles
5431 Details : The ext3 inode allocation policy would not allocate new inodes
5432 very well on the OSTs because there are no new directories
5433 being created. Instead we look for groups with free space if
5434 the parent directories are nearly full.
5439 Description: Network or server problems during mount may cause partially
5440 mounted clients instead of returning an error.
5441 Details : The config llog parsing code may overwrite the error return
5442 code during mount error handling, returning success instead
5448 Description: MDS can fail to allocate large reply buffers
5449 Details : After long uptimes the MDS can fail to allocate large reply
5450 buffers (e.g. zconf client mount config records) due to memory
5451 fragmentation or consumption by the buffer cache. Preallocate
5452 some large reply buffers so that these replies can be sent even
5453 under memory pressure.
5457 Frequency : rare (liblustre)
5458 Description: fsx running with liblustre complained that using truncate() to
5459 extend the file doesn't work. This patch corrects that issue.
5460 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5461 ATTR_SIZE and lsm use in llu_setattr_raw.
5465 Frequency : rare, only 2.6 kernels
5466 Description: Unusual file access patterns on the MDS may result in inode
5467 data being lost in very rare circumstances.
5468 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5469 code in the 2.6 kernel caused the inode and/or EA data not to
5470 be read from disk, causing single-file corruption.
5474 Frequency : rare, only 2.6 filesystems using extents
5475 Description: Heavy concurrent write and delete load may cause data corruption.
5476 Details : It was possible under high-load situations to have an extent
5477 metadata block in the block device cache from a just-unlinked
5478 file overwrite a newly-allocated data block. We now unmap any
5479 metadata buffers that alias just-allocated data blocks.
5483 Frequency : filesystems with default stripe_count larger than 77
5484 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5485 Details : lconf specifies an inode size of 4096 bytes when the default
5486 stripe_count is larger than 77. This conflicts with the default
5487 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5488 this case to avoid pinning too much memory for large EAs.
5490 ------------------------------------------------------------------------------
5492 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5497 Frequency : rare (extremely heavy IO load with hundreds of clients)
5499 Description: Client is evicted, gets IO error writing to file
5500 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5501 caused clients to be evicted for AST timeouts. The fixes in
5502 bug 5192 mean we no longer need to have such short AST timeouts
5503 so ldlm_timeout has been increased.
5506 Frequency : occasional during --force or --failover shutdown under load
5507 Bugzilla : 5949, 4834
5508 Description: Server oops/LBUG if stopped with --force or --failover under load
5509 Details : a collection of import/export refcount and cleanup ordering
5510 issues fixed for safer force cleanup
5513 Frequency : only filesystems larger than 120 OSTs
5514 Bugzilla : 5990, 6223
5515 Description: lfs getstripe would oops on a very large filesystem
5516 Details : lov_getconfig used kfree on vmalloc'd memory
5519 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5520 Bugzilla : 6242, 6243
5521 Description: reading from files that had been truncated to a non-zero size
5522 but never opened returned no data
5523 Details : ll_file_read() reads zeros from no-object files to EOF
5528 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5529 Details : There's little harm in aborting MDS/OSS recovery and letting it
5530 try again, so I removed the LASSERT and return an error instead.
5532 Severity : enhancement
5534 Description: New debugging infrastructure for tracking down data corruption
5535 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5536 (b) cover more of the client-side code path, and (c) try to narrow
5537 down where problems occurred
5541 Bugzilla : 3819, 4364, 4397, 6313
5542 Description: Racing close and eviction MDS could cause assertion in mds_close
5543 Details : It was possible to get multiple mfd references during close and
5544 client eviction, leading to one thread referencing a freed mfd.
5546 Severity: : enhancement
5547 Bugzilla : 3262, 6359
5548 Description: Attempts to reconnect to servers are now more aggressive.
5549 Details : This builds on the enhanced upcall-less recovery that was added
5550 in 1.4.2. When trying to reconnect to servers, clients will
5551 now try each server in the failover group every 10 seconds. By
5552 default, clients would previously try one server every 25 seconds.
5557 Description: After recovery, certain operations trigger a failed
5558 assertion on a client.
5559 Details : Failing over an mds, using lconf -d --failover, while a
5560 client was doing a readdir() call would cause the client to
5561 LBUG after recovery completed and the readdir() was resent.
5563 Severity : enhancement
5565 Description: Default groups are now added by lconf
5566 Details : You can now run lconf --group <servicename> without having to
5567 manually add groups with lmc.
5570 Frequency : occasional
5572 Description: Nodes with an elan id of 0 trigger a failed assertion
5575 Frequency : always when accessing e.g. tty/console device nodes
5577 Description: tty and some other devices nodes cannot be used on lustre
5578 Details : file's private_data field is used by device data and lustre
5579 values in there got lost. New field was added to struct file to
5580 store fs-specific private data.
5583 Frequency : when exporting Lustre via NFS
5585 Description: NFSD failed occasionally when looking up a path component
5586 Details : NFSD is looking up ".." which was broken in ext3 directories
5587 that had grown large enough to become hashed.
5590 Frequency : Clusters with multiple interfaces not on the same subnet
5592 Description: Nodes will repeatedly try to reconnect to an interface which it
5593 cannot reach and report an error to the log.
5594 Details : Extra peer list entries will be created by lconf with some peers
5595 unreachable. lconf now validates the peer before adding it.
5598 Frequency : Only if a default stripe is set on the filesystem root.
5600 Description: Setting a default stripe on the filesystem root prevented the
5601 filesystem from being remounted.
5602 Details : The client was sending extra request flags in the root getattr
5603 request and did not allocate a reply buffer for the dir EA.
5606 Frequency : occasional, higher if lots of files are accessed by one client
5607 Bugzilla : 6159, 6097
5608 Description: Client trips assertion regarding lsm mismatch/magic
5609 Details : While revalidating inodes the VFS looks up inodes with ifind()
5610 and in rare cases can find an inode that is being freed.
5611 The ll_test_inode() code will free the lsm during ifind()
5612 when it finds an existing inode and then the VFS later attaches
5613 this free lsm to a new inode.
5617 Bugzilla : 6422, 7030
5618 Description: MDS deadlock between mkdir and client eviction
5619 Details : Creating a new file via mkdir or mknod (starting a transaction
5620 and getting the ns lock) can deadlock with client eviction
5621 (gets ns lock and trying to finish a synchronous transaction).
5624 Frequency : occasional
5625 Description: While starting a server, the fsfilt_ext3 module could not be
5627 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5628 kernels. Previously, lconf would still use the ext3 name
5629 when trying to load modules. Now, it will correctly use
5630 ext3 on 2.4 and ldiskfs on 2.6.
5632 Severity : enhancement
5633 Description: The default stripe count has been changed to 1
5634 Details : The interpretation of the default stripe count (0, to lfs
5635 or lmc) has been changed to mean striping across a single
5636 OST, rather than all available. For general usage we have
5637 found a stripe count of 1 or 2 works best.
5639 Severity : enhancement
5640 Description: Add support for compiling against Cray portals.
5641 Details : Conditional compiling for some areas that are different
5645 Frequency : occasional
5646 Bugzilla : 6409, 6834
5647 Description: Creating files with an explicit stripe count may lead to
5648 a failed assertion on the MDS
5649 Details : If some OSTs are full or unavailable, creating files may
5650 trigger a failed assertion on the MDS. Now, Lustre will
5651 try to use other servers or return an error to the
5655 Frequency : occasional
5657 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5658 caused lock timeout during readahead (since 1.4.2).
5659 Details : Processes doing readahead might match a lock that hasn't been
5660 granted yet if there are overlapping and conflicting lock
5661 requests. The readahead process waits on ungranted lock
5662 (original lock is CBPENDING), while OST waits for that process
5663 to cancel CBPENDING read lock and eventually evicts client.
5665 Severity : enhancement
5667 Description: Initial enabling of flock support for clients
5668 Details : Implements fcntl advisory locking and file status functions.
5669 This feature is provided as an optional mount flag (default
5670 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5671 locking are implemented yet, and those that are are not guaranteed
5672 to be completely correct in production environments.
5673 mount -t lustre -o [flock|noflock] ...
5676 Frequency : occasional
5678 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5679 assertion in the ext3 JBD (journaling) layer.
5680 Details : The b_committed_data struct is protected by the big kernel lock
5681 in 2.4 kernels, serializing journal_commit_transaction() and
5682 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5683 access to this struct. In 2.6 kernels there is finer grained
5684 locking to improve SMP performance of the JBD layer.
5688 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5693 Description: merge and cleanup kernel patches.
5696 -----------------------------------------------------------------------------
5698 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5700 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5701 versions of Lustre. Please update all servers and clients to
5702 version 1.4.2 or later at the same time. You must also run
5703 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5704 to update the configuration logs.
5706 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5707 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5708 - swab llog records of type '0' so we get proper header size/idx (5861)
5709 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5710 - fix rename of one directory over another leaking an inode (5953)
5711 - avoid SetPageDirty on 2.6 (5981)
5712 - don't re-add just-being-destroyed locks to the waiting list (5653)
5713 - when creating new directories, inherit the parent's custom
5714 striping settings if present parent (3048)
5715 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5716 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5717 - don't hold client locks on temporary worklist from l_lru (5666)
5718 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5719 - automatically evict dead clients (5921)
5720 - Update file size properly in create+truncate+fstat case (6196)
5721 - Do not unhash mountpoint dentries, do not allow removal of
5723 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5724 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5725 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5726 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5727 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5728 - Correctly handle reads of files with no objects (6243)
5729 - lctl recover will also mark a device active if deactivate used (5933)
5731 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5732 - allow --write-conf on an MDS with different nettype than client (5619)
5733 - don't write config llogs to MDS for mounts not from that MDS (5617)
5734 - lconf should create multiple TCP connections from a client (5201)
5735 - init scripts are now turned off by default; run chkconfig --on
5736 lustre and chkconfig --on lustrefs to use them
5737 - upcalls are no longer needed for clients to recover to failover
5739 - add --abort-recovery option to lconf to abort recovery on device
5741 - add support for an arbitrary number of OSTs (3026)
5742 - Quota support protocol changes.
5743 - forward compatibility changes to wire structs (6007)
5744 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5745 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5746 - improved "lustre status" script
5747 - initialize blocksize for non-regular files (6062)
5748 - added --disable-server and --disable-client configure options (5782)
5749 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5750 - Vanilla 2.4.29 support
5751 - increase maximum number of obd devices to 520 (6242)
5752 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5753 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5754 SLES 9 2.6.5 kernels
5755 - limit stripes per file to 160 (the maximum EA size) (6093)
5757 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5760 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5761 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5762 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5763 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5764 - small ext3 extents cleanups and fixes (5733)
5765 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5766 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5767 - avoid needless client->OST connect, fix handle mismatch (5317)
5768 - fix DLM error path that led to out-of-sync client, long delays (5779)
5769 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5770 - fix several locking issues related to i_size (5492,5624,5654,5672)
5771 - don't move pending lock onto export if it is already evicted (5683)
5772 - fix kernel oops when creating .foo in unlinked directory (5548)
5773 - fix deadlock in obdfilter statistics vs. object create (5811)
5774 - use time_{before,after} to avoid timer jiffies wrap (5882)
5775 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5776 - Do not leak request if server was not able to process it (5154)
5777 - If mds_open unable to find parent dir, make that negative lookup(5154)
5778 - don't create new directories with extent-mapping (5909, 5936)
5780 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5781 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5782 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5783 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5784 - support for RHEL4 kernel on i686 (5773)
5785 - provide error messages when incompatible logs are encountered (5898)
5787 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5788 * version 1.4.0.10 (1.4.1 release candidate 1)
5790 - don't keep a lock reference when lock is not granted (4238)
5791 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5792 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5793 - drop import inflight refcount on signal_completed_replay error (5255)
5794 - unlock page after async write error during send (3677)
5795 - handle missing objects in filter_preprw_read properly (5265)
5796 - no transno return for symlink open, don't save no-trasno open (3440)
5797 - don't try to complete elan receive that already failed (4012)
5798 - free RPC server reply state on error (5406)
5799 - clean up thread from ptlrpc_start_thread() on error (5160)
5800 - readahead could read extra page into cache that wasn't ejected (5388)
5801 - prevent races in class_attach/setup/cleanup/detach (5260)
5802 - don't dereference de->d_inode after l_dput of de (5458)
5803 - use "int" for stripe value returned from lock_to_stripe (5544)
5804 - mballoc allocation and error-checking fixes in 2.6 (5504)
5805 - block device patches to fix I/O request sizes in 2.6 (5482)
5806 - look up hostnames for IB nals (5602)
5807 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5808 - don't start multiple acceptors for the same port (5277)
5809 - fix incorrect LASSERT in mds_getattr_name (5635)
5810 - export a proc file for general "ping" checking (5628)
5811 - fix "lfs check" to not block when the MDS is down (5628)
5813 - service request history (4965)
5814 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5815 - create an "evict_client" /proc entry on OSTs, like the MDS has
5816 - fix mount usage message, return errors per mount(8) (5168)
5817 - change grep [] to grep "[]" in tests so they work in more UMLs
5818 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5819 - remove ancient LOV_MAGIC_V0 EA support (5047)
5820 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5821 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5822 - allow readahead allocations to fail when low on memory (5383)
5823 - mmap locking landed again, after considerable improvement (2828)
5824 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5826 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5829 - send OST transaction number in read/write reply to free req (4966)
5830 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5831 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5832 - the watchdog thread now runs as interruptible (5246)
5833 - flock/lockf fixes (but it's still disabled, pending 5135)
5834 - don't use EXT3 constants in llite code (5094)
5835 - memory shortage at startup could cause assertion (5176)
5837 - reorganization of lov code
5838 - single portals codebase
5840 - add extents/mballoc support (5025)
5841 - direct I/O reads in the obdfilter (4048)
5842 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5844 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5847 - send OST transaction number in read/write reply to free req (4966)
5848 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5849 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5850 - the watchdog thread now runs as interruptible (5246)
5851 - handle missing objects in filter_preprw_read properly (5265)
5852 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5853 - drop import inflight refcount on signal_completed_replay error (5255)
5854 - unlock page after async write error during send (3677)
5855 - return original error code on reconstructed replies (3761)
5856 - no transno return for symlink open, don't save no-trasno open (3440)
5858 - add pid to ldlm debugging output (4922)
5859 - bump the watchdog timeouts -- we can't handle 30sec yet
5860 - extra debugging for orphan dentry/inode bug (5259)
5862 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5865 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5866 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5867 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5868 - fix NULL dereference in /proc/sys/portals/routes (4827)
5869 - allow failed mdc_close() operations to be interrupted (4561)
5870 - stop precreate on OST before MDS would time out on it (4778)
5871 - don't send partial-page writes before EOF from client (4410)
5872 - discard client grant for sub-page writes on large-page clients (4520)
5873 - don't free dentries not owned by NFS code, check generation (4806)
5874 - fix lsm leak if mds_create_objects() fails (4801)
5875 - limit debug_daemon file size, always print CERROR messages (4789)
5876 - use transno after validating reply (3892)
5877 - process timed out requests if import state changes (3754)
5878 - update mtime on OST during writes, return in glimpse (4829)
5879 - add mkfsoptions to LDAP (4679)
5880 - use ->max_readahead method instead of zapping global ra (5039)
5881 - don't interrupt __l_wait_event() during strace
5883 - add software watchdogs to catch hung threads quickly (4941)
5884 - make lustrefs init script start after nfs is mounted
5885 - fix CWARN/ERROR duplication (4930)
5886 - return async write errors to application if possible (2248)
5887 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5888 - print NAL number in %x format (4645)
5889 - update barely-supported suse-2.4.21-171 series (4842)
5890 - support for sles 9 %post scripts
5891 - support for building 2.6 kernel-source packages
5892 - support for sles km_* packages
5894 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5897 - ignore -ENOENT errors in osc_destroy (3639)
5898 - notify osc create thread that OSC is being cleaned up (4600)
5899 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5900 - reconstruct ost_handle() like mds_handle() (4657)
5901 - create a new thread to do import eviction to avoid deadlock (3969)
5902 - let lconf resolve symlinked-to devices (4629)
5903 - don't unlink "objects" from directory with default EA (4554)
5904 - hold socknal file ref over connect in case target is down (4394)
5905 - allow more than 32000 subdirectories in a single directory (3244)
5906 - fix blocks count for O_DIRECT writes (3751)
5907 - OST returns ENOSPC from object create when no space left (4539)
5908 - don't send truncate RPC if file size isn't changing (4410)
5909 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5910 - bind to privileged port in socknal and tcpnal (3689)
5912 - rate limit CERROR/CWARN console message to avoid overload (4519)
5913 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5914 - basic mmap support (3918)
5915 - kernel patch series update from b1_4 (4711)
5917 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5920 - avoid crash during MDS cleanup with OST shut down (2775)
5921 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5922 - don't use bad inodes on the MDS (3744)
5923 - dynamic object preallocation to improve recovery speed (4236)
5924 - don't hold spinlock over lock dumping or change debug flags (4401)
5925 - don't zero obd_dev when it is force cleaned (3651)
5926 - print grants to console if they go negative (4431)
5927 - "lctl deactivate" will stop automatic recovery attempts (3406)
5928 - look for existing locks in ldlm_handle_enqueue() (3764)
5929 - don't resolve lock handle twice in recovery avoiding race (4401)
5930 - revalidate should check working dir is a directory (4134)
5932 - don't always mark "slow" obdfilter messages as errors (4418)
5934 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - don't close LustreDB during write_conf until it is done (3860)
5938 - fix typo in lconf for_each_profile (3821)
5939 - allow dumping logs from multiple threads at one time (3820)
5940 - don't allow multiple threads in OSC recovery (3812)
5941 - fix debug_size parameters (3864)
5942 - fix mds_postrecov to initialize import for llog ctxt (3121)
5943 - replace config semaphore with spinlock (3306)
5944 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5945 - don't allow enqueue to complete on a destroyed export (3822)
5946 - down write_lock before checking llog header bitmap (3825)
5947 - recover from lock replay timeout (3764)
5948 - up llog sem before sending rpc (3652)
5949 - reduce ns lock hold times when setting kms (3267)
5950 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5951 - fix NULL deref and obd_dev leak on setup error (3312)
5952 - replace some LBUG about llog ops with error handling (3841)
5953 - don't match INVALID dentries from d_lookup and spin (3784)
5954 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5955 - fix invalid assertion in ptlrpc_set_wait (3880)
5957 - add libwrap support for the TCP acceptor (3996)
5958 - add /proc/sys/portals/routes for non-root route listing (3994)
5959 - allow setting MDS UUID in .xml (2580)
5960 - print the stack of a process that LBUGs (4228)
5962 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5965 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5966 - make sure to unset replay flag from failed open requests (3440)
5967 - if default stripe count is 0, use OST count for inode size (3636)
5968 - update parent mtime/ctime on client for create/unlink (2611)
5969 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5970 - free recovery state on server during a forced cleanup (3571)
5971 - unregister_reply for resent reqs (3063)
5972 - loop back devices mounting and status check on 2.6 (3563)
5973 - fix resource-creation race that can provoke i_size == 0 (3513)
5974 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5975 - more debugging for page-accounting assertion (3746)
5976 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5977 - avoid deadlock after precreation failure (3758)
5978 - fix race and lock order deadlock in orphan handling (3450, 3750)
5979 - add validity checks when grabbing inodes from l_ast_data (3599)
5981 - add /proc/.../recovery_status to obdfilter (3428)
5982 - lightweight CDEBUG infrastructure, debug daemon (3668)
5983 - change default OSC RPC parameters to be better on small clusters
5984 - turn off OST read cache for files smaller than 32MB
5985 - install man pages and include them in rpms (3100)
5986 - add new init script for (un)mounting lustre filesystems (2593)
5987 - run chkconfig in %post for init scripts (3701)
5988 - drop scimac NAL (unmaintained)
5990 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5993 - clean kiobufs before and after use (3485)
5994 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5995 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5996 - change init script to fail more clearly if not run as root (1528)
5997 - allow clients to reconnect during replay (1742)
5998 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5999 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6000 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6001 - fix cond_resched (3554)
6002 - only evict unfinished clients after recovery (3515)
6003 - allow bulk resend, prevent data loss (3570)
6004 - dynamic ptlrpc request buffer allocation (2102)
6005 - don't allow unlinking open directory if it isn't empty (2904)
6006 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6007 - remove extraneous obd dereference causing LASSERT failure (3334)
6008 - don't use get_cycles() when creating temp. files on the mds (3156)
6009 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6010 - handle EEXIST for set-stripe, set proper directory name (3336)
6012 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6013 - fix ksocknal_fmb_callback() error messages (2918)
6015 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6018 - don't copy lvb into (possibly NULL) reply on error (2983)
6019 - don't deref dentry after dput, don't free lvb on error (2922)
6020 - use the kms to determine writeback rpc length (2947)
6021 - increment oti_logcookies when osc is inactive (2948)
6022 - update client's i_blocks count via lvb messages (2543)
6023 - handle intent open/close of special files properly (1557)
6024 - mount MDS with errors=remount-ro, like obdfilter (2009)
6025 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6026 - don't use cancelling-locks' kms values (2947)
6027 - use highest lock extent for kms, not last one (2925)
6028 - don't dereference ERR_PTR() dentry in error handling path (3107)
6029 - fix thread race in portals_debug_dumplog() (3122)
6030 - create lprocfs device entries at setup instead of at attach (1519)
6031 - common AST error handler, don't evict client on completion race (3145)
6032 - zero nameidata in detach_mnt in 2.6 (3118)
6033 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6034 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6035 - initialize RPC timeout timer earlier for 2.6 (3219)
6036 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6037 - print nal/nid for unknown nid (3258)
6038 - additional checks for oscc recovery before doing precreate (3284)
6039 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6040 - don't crash in mdc_close for bad permissions on open (3285)
6041 - zero i_rdev for non-device files (3147)
6042 - clear page->private before handing to FS, better assertion (3119)
6043 - tune the read pipeline (3236)
6044 - fix incorrect decref of invalidated dentry (2350)
6045 - provide read-ahead stats and refine rpc in flight stats (3328)
6046 - don't hold journal transaction open across create RPC (3313)
6047 - update atime on MDS at close time (3265)
6048 - close LDAP connection when recovering to avoid server load (3315)
6049 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6050 - don't leak open file on MDS after open resend (3325)
6051 - serialize filter_precreate and filter_destroy_precreated (3329)
6052 - loop device shouldn't call sync_dev() for nul device (3092)
6053 - clear page cache after eviction (2766)
6054 - resynchronize MDS->OST in background (2824)
6055 - refuse to mount the same filesystem twice on same mountpoint (3394)
6056 - allow llmount to create routes for mounting behind routers (3320)
6057 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6058 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6059 - fix rare problem with rename on htree directories (3417)
6061 - allow default OST striping configuration per directory (1414)
6062 - fix compilation for qswnal for 2.6 kernels (3125)
6063 - increase maximum number of MDS request buffers for large systems
6064 - change liblustreapi to be useful for external progs like lfsck (3098)
6065 - increase local configuration timeout for slow disks (3353)
6066 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6068 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6071 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6072 - don't overwrite extent policy data in reply if lock was blocked (2901)
6073 - drop filter export grants atomically with removal from device (2663)
6074 - del obd_self_export from work_list in class_disconnect_exports (2908)
6075 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6076 - swab reply message in mdc_close, other PPC fixes (2464)
6077 - fix destroying of named logs (2325)
6078 - overwrite old logs when running lconf --write_conf (2264)
6079 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6080 - fix race in target_handle_connect (2898)
6081 - mds_reint_create() should take same inode create lock (2926)
6082 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6083 - don't close files for self_export to avoid uninitialized obd (2936)
6084 - allow MDS with the same name as client node (2939)
6085 - hold dentry reference for closed log files for unlink (2325)
6086 - reserve space for all logs during transactions (2059)
6087 - don't evict page beyond end of stripe extent (2925)
6088 - don't oops on a deleted current working directory (2399)
6089 - handle hard links to targets without a parent properly (2517)
6090 - don't dereference NULL lock when racing during eviction (2867)
6091 - don't grow lock extents when lots of conflicting locks (2919)
6093 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6096 - account for cache space usage on clients to avoid data loss (974)
6097 - lfsck support in lustre kernel code (2349)
6098 - reduce journal credits needed for BRW writes (2370)
6099 - orphan handling to avoid losing space on client/server crashes
6100 - ptlrpcd can be blocked, stopping ALL progress (2477)
6101 - use lock value blocks to assist in proper KMS, faster stat (1021)
6102 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6103 - recovery for initial connections (2355)
6104 - fixes for mds_cleanup_orphans (1934)
6105 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6106 - block all file creations until orphan recovery completes (1901)
6107 - client remove rq_connection from request struct (2423)
6108 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6109 - recovery timer race (2670)
6110 - mdc_close recovey bug (2532)
6111 - ptlrpc cleanup bug (2710)
6112 - mds timeout on local locks (2588)
6113 - namespace lock held during RPCs (2431)
6114 - handle interrupted sync write properly (2503)
6115 - don't try to handle a message that hasn't been replied to (2699)
6116 - client assert failure during cleanup after abort recovery (2701)
6117 - leak mdc device after failed mount (2712)
6118 - ptlrpc_check_set allows timedout requests to complete (2714)
6119 - wait for inflight reqs when ptlrpcd finishes (2710)
6120 - make sure unregistered services are removed from the srv_list
6121 - reset bulk XID's when resending them (caught by 1138 test)
6122 - unregister_bulk after timeout
6123 - fix lconf error (2694)
6124 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6125 - readahead locks pages, leaves pending causing memory pressure (2673)
6126 - increase OST request buffers to 4096 on large machines (2729)
6127 - fix up permission of existing directories in simple_mkdir (2661)
6128 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6129 - don't assign transno to errored transactions (2742)
6130 - don't delete objects on OST if given a bogus objid from MDS (2751)
6131 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6132 - if rq_no_resend, then timeout request after recovery (2432)
6133 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6134 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6135 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6136 - MDS recovery shouldn't delete objects that it has given out (2730)
6137 - if enqueue arrives after completion, don't clobber LVB (2819)
6138 - don't unlock pages twice when trigger_group_io returns error (2814)
6139 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6140 - don't write pages to disk if there was an error (1450)
6141 - don't ping imports that have recovery disabled (2676)
6142 - take buffered bytes into account when balancing socknal conn (2817)
6143 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6144 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6145 - remove little-endian swabbing of llog records (1987)
6146 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6147 - retry reposting request buffers if they fail (1191)
6148 - grow extent at grant time to avoid granting a revoked lock (2809)
6149 - lock revoke doesn't evict page if covered by a second lock (2765)
6150 - disable VM readahead to avoid reading outside lock extents (2805)
6152 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6153 - updated kernel patches for hp-2.4.20 kernel (2681)
6155 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6158 - fix truncated write corruption (2366)
6159 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6161 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6162 - ensure len(uuid) < 37 in lmc (1171)
6163 - fix ia64 OOPS in llog_test (2255)
6164 - zero end of page at obdfilter for partial page writes (2648)
6165 - don't leave stale dentries around after renames (bug 2428)
6166 - fix timeouts when evicting a client with a single lock held (2642)
6167 - set deadline for the initial HELLO message to drain (2634)
6168 - print out dotted-quad IP addresses in the socknal (2302)
6170 - additional debugging for MDS client eviction problem (2443)
6171 - fix mkfsoptions support for osts (2603, 2604)
6173 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6176 - add series for the vanilla 2.6.0 kernel
6177 - add series for the vanilla 2.4.24 kernel
6178 - add series for a cray x86/64 UL kernel drop
6179 - fix xattr patches for the vanilla 2.4.19 series
6181 - generate true UUIDs in lmc (1171)
6182 - have portals stack dumping break in UML (2466)
6183 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6184 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6185 - wake callback waiting threads on client eviction (2460)
6186 - Add --ptldebug and --subsystem to lmc (1719)
6187 - update assertion to allow safe interrupt allocation
6188 - set rq_no_resend for cancel requests (2432)
6189 - recalculate ptlrpcd timeout after resend (2494)
6190 - call vfs_rmdir when removing pending directories (2368)
6191 - fix renaming a file to itself (2429)
6192 - lmc creates a default one-stripe lov (2454)
6193 - expand procfs space to handle large clusters (2326)
6194 - increase UML stack to avoid overflow
6195 - update lconf's list of debug and subsystem masks
6196 - fix lfs find --obd (2510)
6197 - /proc tunable for disabling filter read caching (2591)
6198 - stop rpm packages from altering slapd.conf (2301)
6199 - disable nagle in the socknal under 0conf (2578)
6200 - choose mds inode size based on stripe count (2572)
6201 - fix kernel-source rpm problems (2516)
6203 - add --disable-doc to avoid pdf generation (2421)
6204 - update documentation, tests, type-os, comments
6205 - avoid format warnings on ia64
6206 - remove the TOE NAL
6207 - tiny code cleanups by removing unused fields
6209 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6212 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6213 - remove the most blatant lies from BUILDING file (2371)
6214 - change default debug level to reasonable production setting
6215 - reduce client side cache size to reduce cache flush time
6216 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6217 - make TCP zerocopy and pinger support enabled by default (2476)
6218 - sync writes completed after process exits caused crashes (2319)
6219 - maintain correct mount count on the MDS (2356)
6220 - backout 1557, because 2316 wasn't really fixed
6221 - better file I/O statistics gathering in /proc
6222 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6223 - another kernel patch to fix zero-copy TCP function export
6224 - don't take duplicate lock when processing re-sent getattr (2420)
6225 - lctl uses obd_self_export instead of creating new conn (2353)
6226 - MDS/OST recovery case which requires object creation asserted (2425)
6227 - move lfs from /usr/sbin to /usr/bin in packages
6228 - fix race between mds_client_add and mds_client_free (2417)
6229 - use kmalloc instead of slabs in portals (2430)
6230 - don't create duplicate records when a failover MDS is present (2442)
6231 - remove unnecessary mount age check (2332)
6232 - don't remove directory inodes from locks prematurely (2451)
6233 - don't break if MDS service name is the same as hostname (2103)
6234 - fix races in client write RPC generation when cache full (2482)
6236 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6239 - remove now-unused request->rq_obd (278)
6240 - if an allocation fails, print out how much memory we've used (1933)
6241 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6242 - add the "configurable stack size" patch to most series files (1256)
6243 - ability to write large log records, for 100+ OST configs (2306)
6244 - fix NULL deref when filter_prep fails (2314)
6245 - fix operator precedence error in filter_sync
6246 - dynamic allocation of socknal TX descriptors (2315)
6247 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6248 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6249 - ensure that CWARN messages in Portals always get to the syslog
6250 - __init/__exit are not for prototype decls (ldlm_init/exit)
6251 - x86-64 compile warning fixes
6252 - fix gateway LMC keyword conflict (2318)
6253 - fix MDS lock inversions in getattr/reint paths (1844)
6254 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6255 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6256 - DLM race condition prevented some lock evictions (2328)
6257 - ENOMEM detection and retry on socknal sends (2230)
6258 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6259 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6260 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6261 - better error messages when a client is rejected during recovery (1505)
6262 - avoid cancelling locks which were never granted, after failure (2330)
6263 - fix i_sem/journal inversion in mds_client_add (2333)
6264 - fix truncate/getattr lock cycle deadlock (2334)
6265 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6266 - fix two rare exit paths which could leak an l_lock() ref (2321)
6267 - fencepost error in MDS/OST orphan recovery (2226)
6268 - make log record alignment 8 bytes (1988)
6269 - lstripe now fails when requested offset > ost_count (2237)
6270 - ensure that all kernel series have a complete list.h (1607)
6271 - fix crashes in special-file operations (2316)
6272 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6274 - allow configurable automake binary, for testing new versions
6275 - small update to the lfs documentation
6277 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6279 * fix negative export reference count in fsfilt_sync (2312)
6281 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6282 * release candidate 0.9.1
6284 - orphans are moved into the PENDING directory for possible recovery
6285 - replayed opens now open by fid for orphan/rename safety (1042)
6286 - last close of an orphan inode generates a transno (683)
6287 - chdir() and mount() now pin the directory entry (1020)
6288 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6289 - discard very old requests without processing them (1502)
6290 - remove some common, well-understood CERRORs (1505)
6291 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6292 - clear "grant" flags in OST replies until OST grant code lands (1644)
6293 - fix read performance by not clobbering i_blksize on client (1598)
6294 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6295 - make LRU size a /proc tunable, clears locks when reduced (707)
6296 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6297 - server locks take references on exports now (1558)
6298 - build fixes for 2.4.20-rh trees (1663)
6299 - return an error from lov_create if all OSCs are inactive (1751)
6300 - fix import levels when a reconnect happens without a timeout (1597)
6301 - exit early from mds_open if we get a lookup error (1749)
6302 - partial page read at EOF wouldn't wait for disk before sending (1642)
6303 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6304 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6305 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6306 - write/truncate lock inversion (1639)
6307 - Don't auto-load obdclass, portals modules during cleanup (1495)
6308 - fix timestamps from jumping to "now" (1763)
6309 - extra journal assertions (1648)
6310 - add an extra multiunlink test (1771)
6311 - fix read_record/write_record API (1776)
6312 - fix leak of offset_extent, possible incorrect i_size later (1772)
6313 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6314 - Debugging for the kqswnal_get_idle_tx problems (1820)
6315 - Allow recovery to be attempted multiple times (1536)
6316 - Write out MDS last_rcvd file after it is first created (1600)
6317 - Fix tx_descriptor leak in failed transmit situations (1827)
6318 - ext3 journaling fixes for assertion failure after IO error (1871)
6319 - class_export_put() on freed export after completion AST error (1896)
6320 - Fix revalidate looping in VFS (1322)
6321 - Don't access a freed export during MDS_REINT timeout (1521)
6322 - Add open-unlink recovery support on the MDS (1673,1764)
6323 - Return an error if no MDS data was read from last_rcvd (1946)
6324 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6325 - Don't setup a disk device that doesn't match exported UUID (317)
6326 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6327 - avoid committing NULL handle in force close
6328 - local.sh is now a one-stripe LOV configuration
6329 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6330 - fix ext3 htree duplicate directory entry corruption (1516)
6331 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6332 - update to new LOV EA format (2097)
6333 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6335 2003-06-15 Phil Schwan <phil@clusterfs.com>
6338 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6339 - per-import/export recovery handling (958, 931, 959)
6340 - multiple last-rcvd slots, for serving multiple FSes (949)
6341 - connections are again shared between multiple imp/exports (963, 964)
6342 - "umount -f" would hang if any requests needed to be sent (393, 978)
6343 - avoid pinning large req buffer by copying for queued messages (989)
6344 - add "uuid" to "lctl device" command to help upcalls (991)
6345 - "open" RPCs with transnos would confuse recovery counters (1037)
6346 - do proper endian conversion of all wire messages (288, 340, 891)
6347 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6348 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6349 - fix ptlrpc_connection leak in target_handle_connect (1174)
6350 - fix import refcounting bug in OST and MDS cleanup (1134)
6351 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6352 - fix very unlikely obd_types race condition (501)
6353 - remove osc_open hack for echo_client (1187)
6354 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6355 - a failure in read_inode2 leads to deadlock (1139)
6356 - cancel ack-locks as soon as transaction is committed (1072)
6357 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6358 - make sure to commitrw after any preprw to avoid deadlock (1162)
6359 - failing to execute a file in a lustre FS would lock inode (1203)
6360 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6361 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6362 - obd_uuid2tgt would walk off the end of the list (1255)
6363 - on IA64 the osc would give portals incorrect bulk size (1258)
6364 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6365 - fix lock inversion caused by new llite matching code (1282)
6366 - limit the number of dirty pages on a client to 10MB (1286)
6367 - timed out locks were not being corrected cancelled (1289)
6368 - fix O_DIRECT above 4GB on IA-32 (1292)
6369 * major user-visible changes
6370 - fail out/fail over policy now controlled by the upcall (993)
6372 - add OBD_PING to check server availability and failure (954)
6373 - lustre messages are now sent in sending host order (288, 340, 891)
6374 - add eadatalen to MDS getattr reply (340)
6375 - OST read replies may contain second buffer, with per-page status (593)
6377 2003-03-11 Phil Schwan <phil@clusterfs.com>
6380 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6381 - failed open()s could cause deadlock; fixed (867, 869)
6382 - stop cancelling OST locks when files are closed (481)
6383 - overlapping XID spaces caused network corruption (851, 853)
6384 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6385 - setattr_raw added, to do single-RPC, server-side setattrs
6386 - lmc/lconf syntax change for OST UUIDs
6387 - fix crashy race condition between ptlrpc_free_req and osc_close
6388 - don't use request in mdc_enqueue if we hit a timeout (889)
6389 - don't set the inode i_size for regular files from the MDS (896)
6390 - handle out of order completion AST (842)
6391 - don't LBUG if a lock request times out after receiving AST (913)
6392 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6393 - if a bad lock AST arrives, send an error instead of dropping entirely
6394 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6395 - fix leak in bulk IO when only partially completed (899, 900, 926)
6396 - fix O_DIRECT for ia64 (55)
6397 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6398 - C-z after timeout could hang a process forever; fixed (977)
6400 - client-side I/O cache (678, 924, 929, 941, 970)
6402 - READPAGE and SETATTRs which don't take server-side locks get
6405 2003-02-11 Phil Schwan <phil@clusterfs.com>
6408 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6409 - Fix fsfilt_extN_readpage() to read a full page of directory
6410 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6411 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6412 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6413 - Most intent operations were removed, in favour of a new RPC mode
6414 that does a single RPC to the server and bypasses most of the VFS
6415 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6416 - Aggressively cancel local locks on DLM servers
6417 - mds_reint_unlink sends EA to the client if it's the last nlink.
6418 client uses that EA to unlink OST objects.
6419 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6420 - recursive symlinks were fixed (439)
6421 - fixed NULL deref in DEBUG_REQ
6422 - filter_update_lastobjid no longer calls sync, which annoyed extN
6423 - fixed multi-client small-writes to a single file problem (445)
6424 - fixed mtime updates during file writes (607)
6425 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6426 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6427 - fixed Linux OST size reporting problem (444, 656)
6428 - OST now updates object mtime with writes or setattr (607, 619)
6429 - client verifies file size before zeroing page past EOF (445)
6430 - OST now writes last allocated objid to disk with allocation (108)
6431 - LOV on echo now works (409)
6433 - mds_reint_unlink sends a new buffer, with the EA included. this
6434 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6435 set if a regular file was being unlinked, and it was the last link
6436 - use PtlGet from the target for bulk writes (315)
6437 - OST now updates object mtime with writes or setattr (607, 619)
6438 - LDLM now has a grant-time callback to revalidate locked items, if
6440 - Many MDS operations were reorganized to combat race conditions
6442 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6443 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6445 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6448 - Fully reactivate OST imports after reconnection (512, others)
6449 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6450 - More graceful error handling for truncating on dead OST (515)
6451 - Don't error out unless we're actually accessing dead stripes (474)
6452 - Fix garbage sizes when stripes are missing (410)
6453 - LRU counters were broken, causing constant lock purge (433, 432)
6454 - garbage on read from stripes with failed OSTs (441)
6455 - mark OSCs as active before reconnecting during recovery (438)
6456 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6457 - lfind did not preserve OST order in output (443)
6458 - symlinks cause hung clients, incorrect data (439)
6459 - stop dereferencing request after dropping refcount (457)
6460 - don't LASSERT(spin_is_locked) on non-SMP (455)
6461 - fixes for many rename() bugs
6462 - fstat didn't correctly synchronize attributes (399)
6463 - server must handle lock cancellation during blocking AST prep (487)
6464 - bulk descriptors were free()d too soon (511)
6465 - fix paths in lconf, which would load incorrect modules (451, 507)
6466 - fix confusing lconf 'host not found' error message (386)
6467 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6468 - fix race condition in mdc_blocking_ast() for inode access (526)
6469 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6470 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6471 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6472 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6473 - fix lstripe in several ways (kernel side) (527)
6474 - fix request leak in ldlm_cli_enqueue (262)
6475 - incorrect OSC was marked inactive after OST failure
6476 - call mds_fs_cleanup before unmounting filesystem (524)
6477 - fix races between taking ns_lock and ldlm_lock_change_resource
6478 - fix races updating LOV export open file list
6479 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6480 - fix recovery NULL deref in ldlm_cli_cancel_unused
6481 - fix some DLM races by using new hash table for lock handles (419)
6482 - permit the client to specify desired inodes, at replay
6483 - duplicate requests when we queue them for replay reintegration
6484 - fix last_rcvd offset calculation
6485 - sync after each recovered transaction, so we always make progress
6486 - never, not always, ERESTART requests without transnos
6487 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6489 - skip replay if the MDS says that the client is already connected
6490 - don't check for a recovery-enabled export to match lctl's UUID
6491 - don't INC_USE_COUNT for phantom exports
6492 - don't crash when cleaning up phantom exports (567)
6493 - don't double-finish or set replay data for errored mdc_open requests
6494 - abort requests when they time out, so we don't get old replies
6495 - send/receive replies for AST messages again
6496 - if the client says that it doesn't have the lock, cancel it on the
6498 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6499 mark it as destroyed, it will all work out when decref is called
6500 - fix module use counts (22, 581)
6502 - ASTs now expect a reply (server cancels lock on error reply)
6504 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6507 - fix many simultaneous client startup (392)
6508 - fix dentry->d_it clobbering
6509 - credentials weren't being shipped for readdir/getattr operations
6510 - remove invalid assertions triggered during some concurrent MD
6512 - proper Lustre versions added (336, 389)
6513 - fix memory leak for create error case (398)
6514 - fix LOV locking bug that would get cli/srv out of sync
6515 - fix echo client over LOV (409)
6516 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6517 - fix double-O_EXCL intent crash (424)
6518 - avoid sending multiple lock CANCELs (352)
6520 - MDS can do multi-client recovery (modulo bugs in new code)
6522 - many updates, edits, cleanups
6524 2002-11-18 Phil Schwan <phil@clusterfs.com>
6527 - fix null d_it dereference (346)
6528 - fix full OST/dbench hang (333)
6529 - fix permission problem with file removal (286)
6530 - fix removal of OSCs from LOV when they fail
6531 - fix NULL deref during bulk timeout (214)
6532 - fix problems related to multiple filesystems on one MDS (241)
6533 - fixed serious subtle metadata locking bugs
6534 - free locks on clients when inodes are removed due to memory
6536 - fix inode pointer in lock data (285)
6537 - partial support for multiple MDS on a single host (241)
6538 - data locks weren't cancelled at clear_inode time (290, 311)
6539 - intent locks could lead to unbounded lock growth (205)
6540 - added a maximum lock count, an LRU list, and a flusher
6541 - fix multiple rename (365)
6542 - properly abstracted the echo client
6543 - OSC locked 1 byte too many; fixed
6544 - rewrote brw callback code:
6545 - fixed recovery bugs related to LOVs (306)
6546 - fixed too-many-pages-in-one-write crash (191)
6547 - fixed (again) crash in sync_io_timeout (214)
6548 - probably fixed callback-related race (385)
6550 - Add capability to MDS protocol
6551 - LDLM cancellations and callbacks on different portals
6553 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6556 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6557 - defer open object destruction to close time (601981, 138)
6558 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6559 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6560 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6561 - don't call obd_setattr in ll_file_release for destroyed objects
6562 * protocol change to lustre_msg: move |version| and add |flags|
6563 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6564 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6565 * added replay of create, unlink, link and rename operations during
6566 MDS failover; recovery should be much more robust now
6567 * remove failed OSCs from LOVs (only lov_create uses this so far)
6568 * the lustre-HOWTO was brought (more) up to date (582544)
6570 2002-10-23 Phil Schwan <phil@clusterfs.com>
6573 - in-use dentries weren't being reused properly (617851)
6574 - prevent multiple LDLM setup (599178)
6575 - fix LOV size calculations for truncate (617853)
6576 - fix client handling of MDS intent errors (POSIX)
6577 - fix permission bug in lovstripe.c test (624321)
6578 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6579 - truncate past end of file could corrupt data
6580 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6581 - a race in recovery could return ETIMEDOUT to apps (623947)
6582 - building outside the source directory was fixed
6583 * the lustre-HOWTO was brought (more) up to date (582544)
6584 * major progress was made on recovery functionality
6586 2002-10-10 Phil Schwan <phil@clusterfs.com>
6589 - recovery deadlock fix
6590 - rm -rf causes LBUG fix (617817)
6591 - file open by multiple tasks fix (618962)
6592 - directory permissions bugs (602707 and 620007)
6593 - journal_stop fixed with locking (611313)
6594 - O_APPEND failures resolved (618273, perhaps 614459)
6595 - lconf PATH fix (619770)
6596 - IA64 build fix (621450)
6597 - RPC buffer sizes scale with amount of memory
6599 2002-10-01 Phil Schwan <phil@clusterfs.com>
6602 - locks would be cancelled without throwing away data pages,
6603 resulting in inconsistent data (605627)
6604 - inode attributes were not always being refreshed (605627, 612449)
6605 - lconf now continues to cleanup after lctl reports an error
6606 - MDS now enforces user permissions (602707)
6607 - lprocfs cleanup fixed, but not yet enabled (614157)
6608 - fixed infinite server hang, should a client not respond to an AST
6609 - avoid going into recovery if user calls readlink() with a buffer
6610 that's too small (613941)
6611 - AST RPCs no longer require replies (614867) -- this may be changed
6612 - don't crash server if client sends an IOV that's too big (611336)
6613 - fixed lock conversion deadlock (611892)
6614 - fixed the following of symlinks (614622)
6615 * recovery: the server can remove locks from a client that dies, other
6616 clients can make progress
6617 * more extN patch fixes
6618 * compile-time configurable ptlrpc buffer allocations
6620 - collaborative read cache document
6621 - Lustre Lite Performance CDR document-in-progress
6623 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6626 - fix typo in patch-2.4.18
6628 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6631 - clear ptlrpc request each time in handle_incoming_request()
6632 - unlink of files now destroys the object on the OST
6634 2002-09-19 Peter Braam <braam@clusterfs.com>
6636 * add hard link support
6637 * change obdfile creation method
6638 * kernel patch changed
6640 2002-09-19 Peter Braam <braam@clusterfs.com>
6643 - stack overflow bug in extN fixed
6645 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6647 * documentation updates
6648 - add man pages for config tools
6649 - update tests/README to describe testing with new config tools
6650 - finish metadata API descriptions
6651 * bug fixes and cleanups
6652 - statfs workaround for 16TB limit
6653 - LOV stripe allocation improved, can stripe on subset of OSTs
6654 - LOV file size/IO offset was wrong for files > 4GB in size
6655 - object EA data was being dropped, caused files to be unreadable
6656 - memory overflow with non-LOV OST caused memory corruption
6657 - fixed regression tests to work with new config tools, obdfilter
6658 - fixed bug when directory size became larger than 1 block
6659 - fixed bug (for single client case) when PWD was deleted
6660 - invalidate local directory pages when doing intent-based ops
6661 - avoid LDLM oops when lock callback contained bad data
6663 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6665 * documentation updates
6666 * bug fixes and cleanups
6667 - configuration tools
6670 - 64-bit compile warnings
6671 - 64-bit internal statfs data
6673 * test_brw on persistent OST devices
6675 * lprocfs (disabled)
6677 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6679 * documentation updates
6680 * bug fixes and cleanups
6681 * configuration tools
6683 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6686 * many small fixes to 0_5_4
6687 * io/network handling
6688 * thinkos in MDS operations
6690 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6693 * crucial basic fixes to 0.5.3
6694 * IOR, Iozone work over Elan
6697 2002-08-07 Phil Schwan <phil@clusterfs.com>
6698 * version 0_5_3, our first alpha
6699 * we use the new Portals iovs
6700 * documentation updates
6701 * bug fixes and cleanups
6702 * small changes in the DLM wire protocol
6704 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6705 * version 0_5_1 with some initial stability,
6706 * locking on MD and file I/O.
6707 * documentation updates
6708 * several bug fixes since 0.5.0
6709 * small changes in wire protocol
6711 2002-07-18 Phil Schwan <phil@clusterfs.com>
6713 * delivered as Lustre Light Alpha
6714 * fixed a crash after handling invalid MDS requests
6715 * fixed directory pages for architectures with non-4k pages sizes
6717 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6718 * release version v0_4_4
6719 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6721 * Fixes a number of interruption problems with OST operations.
6722 * Update documentation for portals header changes
6723 * Move all wire protocol structs/defines to lustre_idl.h
6724 * Fixes symlink length bug.
6725 * Add tcpdump to repository.
6727 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6728 * release version v0_4_3
6729 * Fixes statfs for inodes on extN.
6730 * Fixes bug in runtests which would delete /etc/hosts.
6731 * Use 64-bit object IDs wherever possible (not into VFS though)
6732 Remove ost_get_info, which is unused by lustre, and out of date.
6734 2002-07-03 Peter Braam <braam@clusterfs.com>
6735 * release version v0_4_2 Fixes a lookup error (type not passed)
6736 * move forward to head of Portals
6737 * move forward to latest Lustre kernel
6739 2002-06-25 Peter Braam <braam@clusterfs.com>
6740 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob