1 05-22-2018 Intel Corporation
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 2.6.32-696.30.1.el6 (RHEL6.9)
11 3.10.0-862.9.1.el7 (RHEL7.5)
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 4.4.132-94.33 (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 2.6.32-696.30.1.el6 (RHEL6.9)
24 3.10.0-862.9.1.el7 (RHEL7.5)
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 4.4.133-94.33 (SLES12 SP3)
30 4.4.0-116 (Ubuntu 16.04)
32 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
33 * Recommended ZFS / SPL version: 0.7.9
34 * Tested with ZFS / SPL version: 0.7.9
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
41 01-31-2018 Intel Corporation
43 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
44 for currently supported client and server kernel versions.
45 * Server known to build on patched kernels:
46 2.6.32-431.29.2.el6 (RHEL6.5)
47 2.6.32-504.30.3.el6 (RHEL6.6)
48 2.6.32-573.26.1.el6 (RHEL6.7)
49 2.6.32-642.15.1.el6 (RHEL6.8)
50 2.6.32-696.23.1.el6 (RHEL6.9)
51 3.10.0-693.21.1.el7 (RHEL7.4)
52 3.0.101-0.47.71 (SLES11 SP3)
53 3.0.101-107 (SLES11 SP4)
54 3.12.74-60.64.40 (SLES12 SP1)
55 4.4.120-92.70 (SLES12 SP2)
56 4.4.120-94.17 (SLES12 SP3)
57 vanilla linux 4.6.7 (ZFS only)
58 * Client known to build on unpatched kernels:
59 2.6.32-431.29.2.el6 (RHEL6.5)
60 2.6.32-504.30.3.el6 (RHEL6.6)
61 2.6.32-573.26.1.el6 (RHEL6.7)
62 2.6.32-642.15.1.el6 (RHEL6.8)
63 2.6.32-696.23.1.el6 (RHEL6.9)
64 3.10.0-693.21.1.el7 (RHEL7.4)
65 3.0.101-0.47.71 (SLES11 SP3)
66 3.0.101-107 (SLES11 SP4)
67 3.12.74-60.64.40 (SLES12 SP1)
68 4.4.120-92.70 (SLES12 SP2)
69 4.4.120-94.17 (SLES12 SP3)
70 4.4.0-116 (Ubuntu 16.04)
72 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
73 * Recommended ZFS / SPL version: 0.7.5
74 * Tested with ZFS / SPL version: 0.7.5
75 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
76 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
77 stack overflow. For more information, please refer to bugzilla 17630.
78 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
79 the re-exporting nodes to be mounted with "32bitapi" mount option
81 --------------------------------------------------------------------------------
83 12-11-2017 Intel Corporation
85 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
86 for currently supported client and server kernel versions.
87 * Server known to build on patched kernels:
88 2.6.32-431.29.2.el6 (RHEL6.5)
89 2.6.32-504.30.3.el6 (RHEL6.6)
90 2.6.32-573.26.1.el6 (RHEL6.7)
91 2.6.32-642.15.1.el6 (RHEL6.8)
92 2.6.32-696.20.1.el6 (RHEL6.9)
93 3.10.0-693.17.1.el7 (RHEL7.4)
94 3.0.101-0.47.71 (SLES11 SP3)
95 3.0.101-107 (SLES11 SP4)
96 3.12.74-60.64.40 (SLES12 SP1)
97 4.4.103-92.56 (SLES12 SP2)
98 4.4.103-6.38 (SLES12 SP3)
99 vanilla linux 4.6.7 (ZFS only)
100 * Client known to build on unpatched kernels:
101 2.6.32-431.29.2.el6 (RHEL6.5)
102 2.6.32-504.30.3.el6 (RHEL6.6)
103 2.6.32-573.26.1.el6 (RHEL6.7)
104 2.6.32-642.15.1.el6 (RHEL6.8)
105 2.6.32-696.20.1.el6 (RHEL6.9)
106 3.10.0-693.17.1.el7 (RHEL7.4)
107 3.0.101-0.47.71 (SLES11 SP3)
108 3.0.101-107 (SLES11 SP4)
109 3.12.74-60.64.40 (SLES12 SP1)
110 4.4.103-92.56 (SLES12 SP2)
111 4.4.103-6.38 (SLES12 SP3)
113 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
114 * Recommended ZFS / SPL version: 0.7.3
115 * Tested with ZFS / SPL version: 0.7.3
116 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
117 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
118 stack overflow. For more information, please refer to bugzilla 17630.
119 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
120 the re-exporting nodes to be mounted with "32bitapi" mount option
122 --------------------------------------------------------------------------------
124 10-02-2017 Intel Corporation
126 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
127 for currently supported client and server kernel versions.
128 * Server known to build on patched kernels:
129 2.6.32-431.29.2.el6 (RHEL6.5)
130 2.6.32-504.30.3.el6 (RHEL6.6)
131 2.6.32-573.26.1.el6 (RHEL6.7)
132 2.6.32-642.15.1.el6 (RHEL6.8)
133 2.6.32-696.10.3.el6 (RHEL6.9)
134 3.10.0-693.2.2.el7 (RHEL7.4)
135 3.0.101-0.47.71 (SLES11 SP3)
136 3.0.101-107 (SLES11 SP4)
137 3.12.74-60.64.40 (SLES12 SP1)
138 4.4.74-92.35 (SLES12 SP2)
139 vanilla linux 4.6.7 (ZFS only)
140 * Client known to build on unpatched kernels:
141 2.6.32-431.29.2.el6 (RHEL6.5)
142 2.6.32-504.30.3.el6 (RHEL6.6)
143 2.6.32-573.26.1.el6 (RHEL6.7)
144 2.6.32-642.15.1.el6 (RHEL6.8)
145 2.6.32-696.10.3.el6 (RHEL6.9)
146 3.10.0-693.2.2.el7 (RHEL7.4)
147 3.0.101-0.47.71 (SLES11 SP3)
148 3.0.101-107 (SLES11 SP4)
149 3.12.74-60.64.40 (SLES12 SP1)
150 4.4.74-92.35 (SLES12 SP2)
152 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
153 * Recommended ZFS / SPL version: 0.7.1
154 * Tested with ZFS / SPL version: 0.7.1
155 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
156 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
157 stack overflow. For more information, please refer to bugzilla 17630.
158 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
159 the re-exporting nodes to be mounted with "32bitapi" mount option
161 --------------------------------------------------------------------------------
163 2017-07-05 Intel Corporation
165 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
166 for currently supported client and server kernel versions.
167 * Server known to build on patched kernels:
168 2.6.32-431.29.2.el6 (RHEL6.5)
169 2.6.32-504.30.3.el6 (RHEL6.6)
170 2.6.32-573.26.1.el6 (RHEL6.7)
171 2.6.32-642.15.1.el6 (RHEL6.8)
172 2.6.32-696.3.1.el6 (RHEL6.9)
173 3.10.0-514.21.1.el7 (RHEL7.3)
174 3.0.101-0.47.71 (SLES11 SP3)
175 3.0.101-100 (SLES11 SP4)
176 3.12.74-60.64.40 (SLES12 SP1)
177 4.4.59-92.17 (SLES12 SP2)
178 vanilla linux 4.6.7 (ZFS only)
179 * Client known to build on unpatched kernels:
180 2.6.32-431.29.2.el6 (RHEL6.5)
181 2.6.32-504.30.3.el6 (RHEL6.6)
182 2.6.32-573.26.1.el6 (RHEL6.7)
183 2.6.32-642.15.1.el6 (RHEL6.8)
184 2.6.32-696.3.1.el6 (RHEL6.9)
185 3.10.0-514.21.1.el7 (RHEL7.3)
186 3.0.101-0.47.71 (SLES11 SP3)
187 3.0.101-100 (SLES11 SP4)
188 3.12.74-60.64.40 (SLES12 SP1)
189 4.4.59-92.17 (SLES12 SP2)
191 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
192 * Recommended ZFS / SPL version: 0.6.5.9
193 * Tested with ZFS / SPL version: 0.6.5.9
194 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
195 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
196 stack overflow. For more information, please refer to bugzilla 17630.
197 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
198 the re-exporting nodes to be mounted with "32bitapi" mount option
200 --------------------------------------------------------------------------------
202 12-07-2016 Intel Corporation
204 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
205 for currently supported client and server kernel versions.
206 * Server known to build on patched kernels:
207 2.6.32-431.29.2.el6 (RHEL6.5)
208 2.6.32-504.30.3.el6 (RHEL6.6)
209 2.6.32-573.26.1.el6 (RHEL6.7)
210 2.6.32-642.11.1.el6 (RHEL6.8)
211 3.10.0-514.el7 (RHEL7.3)
212 3.0.101-0.47.71 (SLES11 SP3)
213 3.0.101-80 (SLES11 SP4)
214 3.12.62-60.64.8 (SLES12 SP1)
215 vanilla linux 4.5.7 (ZFS only)
216 * Client known to build on unpatched kernels:
217 2.6.32-431.29.2.el6 (RHEL6.5)
218 2.6.32-504.30.3.el6 (RHEL6.6)
219 2.6.32-573.26.1.el6 (RHEL6.7)
220 2.6.32-642.11.1.el6 (RHEL6.8)
221 3.10.0-514.el7 (RHEL7.3)
222 3.0.101-0.47.71 (SLES11 SP3)
223 3.0.101-80 (SLES11 SP4)
224 3.12.62-60.64.8 (SLES12 SP1)
226 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
227 * Recommended ZFS / SPL version: 0.6.5.8
228 * Tested with ZFS / SPL version: 0.6.5.8
229 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
230 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
231 stack overflow. For more information, please refer to bugzilla 17630.
232 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
233 the re-exporting nodes to be mounted with "32bitapi" mount option
235 --------------------------------------------------------------------------------
237 02-29-2016 Intel Corporation
239 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
240 for currently supported client and server kernel versions.
241 * Server known to build on patched kernels:
242 2.6.32-431.29.2.el6 (RHEL6.5)
243 2.6.32-504.30.3.el6 (RHEL6.6)
244 2.6.32-573.12.1.el6 (RHEL6.7)
245 3.10.0-327.3.1.el7 (RHEL7.2)
246 3.0.101-0.47.71 (SLES11 SP3)
247 3.0.101-68 (SLES11 SP4)
248 vanilla linux 4.2.1 (ZFS only)
249 * Client known to build on unpatched kernels:
250 2.6.32-431.29.2.el6 (RHEL6.5)
251 2.6.32-504.30.3.el6 (RHEL6.6)
252 2.6.32-573.12.1.el6 (RHEL6.7)
253 3.10.0-327.3.1.el7 (RHEL7.2)
254 3.0.101-0.47.71 (SLES11 SP3)
255 3.0.101-68 (SLES11 SP4)
258 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
259 * Recommended ZFS / SPL version: 0.6.4.2
260 * Tested with ZFS / SPL version: 0.6.4.2
261 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
262 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
263 stack overflow. For more information, please refer to bugzilla 17630.
264 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
265 the re-exporting nodes to be mounted with "32bitapi" mount option
267 --------------------------------------------------------------------------------
269 03-10-2015 Intel Corporation
271 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
272 for currently supported client and server kernel versions.
273 * Server known to build on patched kernels:
274 2.6.32-431.29.2.el6 (RHEL6.5)
275 2.6.32-504.8.1.el6 (RHEL6.6)
276 3.0.101-0.46 (SLES11 SP3)
277 * Client known to build on unpatched kernels:
278 2.6.32-431.29.2.el6 (RHEL6.5)
279 2.6.32-504.8.1.el6 (RHEL6.6)
280 3.10.0-123.20.1.el7 (RHEL7)
281 3.0.101-0.46 (SLES11 SP3)
282 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
283 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
284 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
285 stack overflow. For more information, please refer to bugzilla 17630.
286 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
287 the re-exporting nodes to be mounted with "32bitapi" mount option
289 Severity : enhancement
291 Description: control OST-index in IDIF via ROCOMPAT flag.
292 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
293 last_rcvd file. For new formatted OST device, it will be auto set;
294 for the case of upgrading from old OST device, you can enable it
295 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
296 enabled, for new created OST-object, its IDIF-in-LMA will contain
297 the OST-index; for the existing OST-object, the OSD will convert
298 old format IDIF as new format IDIF with OST-index stored in the
299 LMA EA when accessing such OST-object or via OI scrub. Once such
300 flag is enabled, it cannot be reverted back, so the system cannot
301 be downgraded to the orignal incompatible version.
303 --------------------------------------------------------------------------------
305 07-30-2014 Intel Corporation
307 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
308 for currently supported client and server kernel versions.
309 * Server known to build on patched kernels:
310 2.6.32-431.20.3.el6 (RHEL6.5)
311 3.0.101-0.31 (SLES11 SP3)
312 * Client known to build on unpatched kernels:
313 2.6.32-431.20.3.el6 (RHEL6.5)
314 3.0.101-0.31 (SLES11 SP3)
317 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
318 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
319 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
320 stack overflow. For more information, please refer to bugzilla 17630.
321 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
322 the re-exporting nodes to be mounted with "32bitapi" mount option
324 --------------------------------------------------------------------------------
326 10-31-2013 Intel Corporation
328 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
329 * Server builds for kernels:
330 2.6.32-358.18.1.el6 (RHEL6.4)
331 * Client builds for unpatched kernels:
332 2.6.32-358.18.1.el6 (RHEL6.4)
333 3.0.92-0.8 (SLES11 SP3)
335 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
336 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
337 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
338 stack overflow. For more information, please refer to bugzilla 17630.
339 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
340 the re-exporting nodes to be mounted with "32bitapi" mount option
342 --------------------------------------------------------------------------------
344 05-30-2013 Intel Corporation
346 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
347 * Server builds for kernels:
348 2.6.32-358.6.2.el6 (RHEL6.4)
349 * Client builds for unpatched kernels:
350 2.6.32-358.6.2.el6 (RHEL6.4)
351 3.0.74-0.6.6 (SLES11 SP2)
353 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
354 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
355 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
356 stack overflow. For more information, please refer to bugzilla 17630.
357 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
358 the re-exporting nodes to be mounted with "32bitapi" mount option
360 --------------------------------------------------------------------------------
362 9-30-2012 Whamcloud, Inc.
364 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
365 * Server support for kernels:
366 2.6.32-279.5.1.el6 (RHEL6)
367 * Client support for unpatched kernels:
368 2.6.32-279.5.1.el6 (RHEL6)
370 2.6.18-238.19.1.el5 (RHEL5)
371 2.6.16.60-0.69.1 (SLES10)
372 * Recommended e2fsprogs version: 1.41.90.wc4
373 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
374 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
375 stack overflow. For more information, please refer to bugzilla 17630.
376 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
377 the re-exporting nodes to be mounted with "32bitapi" mount option
379 Severity : enhancement
381 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
383 Severity : enhancement
385 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
387 Severity : enhancement
389 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
391 Severity : enhancement
393 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
395 Severity : enhancement
397 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
399 --------------------------------------------------------------------------------
401 04-30-2012 Whamcloud, Inc.
403 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
404 * Server support for kernels:
405 2.6.32-220.4.2.el6 (RHEL6)
406 * Client support for unpatched kernels:
407 2.6.32-220.4.2.el6 (RHEL6)
409 2.6.18-238.19.1.el5 (RHEL5)
410 2.6.16.60-0.69.1 (SLES10)
411 * Recommended e2fsprogs version: 1.41.90.wc4
412 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
413 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
414 stack overflow. For more information, please refer to bugzilla 17630.
415 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
416 the re-exporting nodes to be mounted with "32bitapi" mount option
418 Severity : enhancement
420 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
422 Severity : enhancement
424 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
426 Severity : enhancement
428 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
430 Severity : enhancement
432 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
434 --------------------------------------------------------------------------------
436 09-30-2011 Whamcloud, Inc.
438 * Server support for kernels:
439 2.6.18-238.19.1.el5 (RHEL5)
440 2.6.32-131.6.1.el6 (RHEL6)
441 * Client support for unpatched kernels:
442 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
443 2.6.18-238.19.1.el5 (RHEL5)
444 2.6.32-131.6.1.el6 (RHEL6)
445 2.6.29.4-167.fc11 (FC11)
446 2.6.16.60-0.69.1 (SLES10)
448 * Recommended e2fsprogs version: 1.41.90.wc3
449 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
450 Lustre filesystem with 4K stack may cause a stack overflow. For more
451 information, please refer to bugzilla 17630.
452 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
453 mounted with "32bitapi" mount option
455 Severity : enhancement
456 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
458 Severity : enhancement
459 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
461 Severity : enhancement
462 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
465 Frequency : only with SLES10
467 Description: Use OFED "KMP" provided by Novell
468 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
469 Lustre is now built against this package. That means you need to
470 install the ofed-kmp package from Novell for the patchless client.
471 Note that the ofed-kmp that Novell ships may not exactly match the
472 kernel version but should still be compatible.
474 Severity : enhancement
476 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
477 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
479 --------------------------------------------------------------------------------
481 2010-07-15 Oracle, Inc.
483 * Server support for kernels:
484 2.6.18-164.11.1.el5 (RHEL 5)
485 2.6.18-164.11.1.0.1.el5 (OEL 5)
486 * Client support for unpatched kernels:
487 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
488 2.6.18-164.11.1.el5 (RHEL 5),
489 2.6.18-164.11.1.0.1.el5 (OEL 5)
490 2.6.16.60-0.42.8 (SLES 10),
492 2.6.29.4-167.fc11 (FC11)
493 * Recommended e2fsprogs version: 1.41.10-sun2
495 Severity : enhancement
497 Description: Distribute lfs_migrate script with lustre
498 Details : The lfs_migrate mentioned in the manual is now included in the
499 lustre tree in lustre/scripts and will distributed in lustre RPMs.
501 Severity : enhancement
503 Description: Failover nid list added to import proc file.
505 Severity : enhancement
507 Description: "som_preview" mount option required to use SOM
508 Details : The clients will need to use som_preview mount option
512 Description: Lustre does not do 1MB IOs to HW RAID
513 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
515 Severity : enhancement
517 Description: lfs quota output is not very convenient for awk/sed-parsing
518 Details : Some positions in lfs quota output table could be empty or
519 non-empty which made it hard to parse it with scripts, now
520 a dash is put instead of space where there is not supposed
525 Description: Update to OFED 1.5.1
526 Details : OFED 1.5.1 will be used in environments that don't use
527 vendor supplied OFED.
531 Description: Remove lnet. prefix from lctl params display
532 Details : Also renamed lnet.memused to lnet_memused to
537 Description: Use conf_param -d to remove permanant settings.
541 Description: Add list_param -R
542 Details : lctl list_param now supports recursive mode.
545 Bugzilla : 21829, 22312
546 Description: Fix llobdstat.
547 Details : 1. Now updating the write_bytes counter
548 2. Port script fixes from 1.8.
552 Description: decrease the usage of memory on clients.
553 Details : 1. On clients, recycle dentries and inodes unused.
554 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
559 Description: llobdstat fix and enhancement
560 Details : add a counter to set a limit to how many samples will be returned
561 fix a wildcard in the path to limit to obdfilter stats only
565 Description: lfs check only allowed for root
566 Details : Code cleanups and added a regresion test of non-root use of
571 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
572 Details : Fix hr->hr_index overload.
576 Description: Fix FIEMAP functionality.
580 Description: FC11 patchless client support added
584 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
585 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
589 Description: New acc-sm script and various test-framework enhancments.
593 Description: Previously landed patch was reverted.
597 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
601 Description: Add ldiskfs maxdirsize mount option.
605 Description: Fix mkfs.lustre support for 16TB LUNs.
609 Description: Print mmp_check_interval
610 Details : print mmp_check_interval and make it possible to abort
611 mount operation in case it takes too long.
615 Description: Add cascading_rw to lustre mpi tests.
619 Description: Store SOM data in EA.
620 Details : Also several SOM related cleanups.
624 Description: Add refcount to md_open_data
628 Description: Use refcount on lov_request_set to prevent memory corruption.
632 Description: Statahead: use parent dir's i_mutex for synchronization.
636 Description: loadgen improvements.
640 Description: Enable peer health detection in ksocklnd and ko2iblnd.
644 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
648 Description: Fix race on obd_devcie:md_stats
652 Description: Fix SLES 11 client builds.
653 Details : Resolved uninitalized pointer gcc warnings.
657 Description: Allow non-root access for "lfs check".
658 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
662 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
664 Severity : enhancement
666 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
667 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
671 Description: Sanity test_27p fails.
672 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
673 Sanity tests 27 were cleaned up and improved.
677 Description: Unify req format on client/servers
678 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
679 buffers and swabbers approach..
684 Description: drop join file stuff.
686 Severity : enhancement
688 Description: Add 2.6.27 and SLES11 (patchless client) support
692 Description: Fix in ptlrpc_expire_one_request() to print the signed time
695 Severity : enhancement
697 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
700 Severity : enhancement
702 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
704 Severity : enhancement
706 Description: correctly handle big reply message.
707 Details : send LNet event if reply is bigger than buffer and adjust this buffer
712 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
714 Severity : enhancement
716 Description: Add support for OEL5.
720 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
722 Severity : enhancement
724 Description: "weak-modules" support
725 Details : Implement "weak-modules" support which enables kernel modules
726 to be used with any kernel that implements the same kABI. In
727 order to achieve this modules are now installed in
728 /lib/modules/$(uname -r)/updates/kernel on all distributions.
732 Description: client could not reconnect to OST because of an active request.
733 Details : The client now retries to reconnect to the same server, if a
734 connect request failed with EBUSY or -EAGAIN. Test case 35b
735 conf-sanity.sh has been created.
739 Description: Don't allow make backward step on assignin osc next id.
740 Details : race between allocation next id and ll_sync thread can be cause
741 of set wrong osc next id and can be kill valid ost objects.
745 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
749 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
753 Description: Error handling in osc_statfs_interpret() has been improved.
754 Details : Check in osc_statfs_interpret() for EBADR.
758 Description: Do not send statfs() requests to OSTs disabled by administrator.
759 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
761 Severity : enhancement
763 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
765 Severity : enhancement
767 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
769 Severity : enhancement
771 Description: Enhancement for lfs(1) command to use numeric uid/gid.
775 Description: Improved cleanup procedure for conf-sanity test_46a.
779 Description: Do not start bulk transfer, if client tried to reconnect.
780 Details : Do not start bulk transfer, if client tried to reconnect,
781 but failed due to active rpcs. Check for exp_abort_active_req
782 has been implemented.
786 Description: Prevent inconsistences between linux and lustre mount structures.
787 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
788 Make the sleep interruptible.
790 Severity : enhancement
792 Description: provide server to client comms path
793 Details : server to client communications path via new LDLM_SET_INFO rpc
797 Description: truncate starts GFP_FS allocation under transaction causing deadlock
798 Details : ldiskfs_truncate calls grab_cache_page which may start page
799 allocation under an open transaction. This may lead to
800 calling prune_icache with consequent lustre reentrance.
804 Description: send correctly lsm on open replay
805 Details : MDS is trust to LSM size on replay open, but client can set wrong size
810 Description: enable client interface failover
811 Details : When a child reconnects from another NID, properly update export
812 nid hash position and ldlm reverse import.
814 Severity : enhancement
816 Description: Communicate OST degraded/readonly state via statfs to MDS
817 Details : Flags in the statfs returned from OSTs indicate whether the
818 OST is in a degraded RAID state, or if the filesystem has
819 turned read-only after a filesystem error is detected.
824 Description: don't panic if EPROTO was hit when reading symlink
825 Details : correctly handling request reference in error cases.
827 Severity : enhancement
829 Description: Add LustreNetLink, a kernel-userspace communcation path.
831 Severity : enhancement
833 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
838 Description: access to llog context before init.
839 Details : move handling CATALOGS file at osc layer and forbid access to llog
843 Frequency : rare, only when client is writing to OST on local node
845 Description: Avoid deadlock for local client writes
846 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
847 memory freeing context. This allows OST threads to set the
848 PF_MEMALLOC flag on task structures in order to allocate memory
849 from reserved pools and complete IO.
850 Use GFP_HIGHUSER for OST allocations for non-local client writes,
851 so that the OST threads generate memory pressure and allow
852 inactive pages to be reclaimed.
854 Severity : enhancement
856 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
861 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
862 Details : move ldlm namespace creation in setup phase to avoid grab
863 _lprocfs_lock with cli_sem held.
867 Description: Temporarily disable grant shrink.
868 Details : Disable the feature for debugging.
871 Frequency : only when formatting test filesystems
873 Description: Unable to run several mkfs.lustre on loop devices at the same time
874 Details : mkfs.lustre returns error 256 on the concurrent loop devices
875 formatting. The solution is to proper handle the error.
877 Severity : enhancement
879 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
881 Severity : enhancement
883 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
888 Description: resolve race between obd_disconnect and class_disconnect_exports
889 Details : if obd_disconnect will be called to already disconnected export he
890 forget release one reference and osc module can't unloaded.
892 Severity : enhancement
894 Description: Allow tuning service thread via /proc
895 Details : For each service a new
896 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
897 created that can be used to set min/max thread counts, and get the
898 current number of running threads.
903 Descriptoin: don't return error if have partially created objects for file.
904 Details : lov_update_create_set uses set->set_success as index for created
905 objects, so if some requests will be failed, they will have hole
906 at end of array and qos_shrink_lsm for allocate correct lsm.
908 Severity : enhancement
910 Description: Update OFED support to 1.4.1
915 Description: Reduce small size read RPC
916 Details : Set read-ahead limit for every file and only do read-ahead when
917 available read-ahead pages are bigger than 1M to avoid small size
920 Severity : enhancement
922 Description: move AT tunable parameters for more consistent usage
923 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
925 Severity : enhancement
927 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
928 Details : allow skip disconnected ost for send statfs request and hide error
934 Description: fix racy locking of mballoc block bitmaps causing BUG
935 Details : The locking of the mballoc buddy bitmap and the in-memory
936 block bitmap was using two different spin locks in some
937 cases. This made it possible to incorrectly access the
938 mballoc bitmap while another process was modifying it,
939 causing a sanity assertion to fail. While no on-disk corruption
940 was reported, there was some risk of this happening.
943 Frequency : rare, on llog test 6
945 Descriptoin: don't allow connect to already connected import
946 Details : allowing connect to already connected import is hide connecting problem.
951 Descriptoin: allow kill process which wait statahead result
952 Details : for some reasons 'ls' can stick in waiting result from statahead,
953 in this case need way for kill this process.
955 Severity : enhancement
957 Description: Add state history info file, enhance import info file
958 Details : Track import connection state changes in a new osc/mdc proc file;
959 add overview-type data to the osc/mdc import proc file.
961 Severity : enhancement
963 Description: MDS create should not wait for statfs RPC while holding DLM lock.
965 Severity : enhancement
967 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
972 Descriptoin: OST grant too much space to client even there are not enough space.
973 Details : Client will shrink its grant cache to OST if there are no write
974 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
975 this grant cache if there are already not enough avaible space
976 (left_space < total_clients * 32M).
979 Frequency : start MDS on uncleanly shutdowned MDS device
981 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
982 Details : stay in waiting mds<>ost recovery finished produce random bugs
983 due race between two ll_sync thread for one lov target. send
984 ACTIVATE event only if connect realy finished and import have
988 Frequency : rare, connect and disconnect target at same time
990 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
991 Details : don't call obd_disconnect under lov_lock. this long time
992 operation and can block ptlrpcd which answer to connect request.
997 Descriptoin: don't lose wakeup for imp_recovery_waitq
998 Details : recover_import_no_retry or invalidate_import and import_close can
999 both sleep on imp_recovery_waitq, but we was send only one wakeup
1003 Frequency : always with long access acl
1005 Descriptoin: mds can't pack reply with long acl.
1006 Details : mds don't control size of acl but they limited by reint/getattr
1009 Severity : enhancement
1011 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1013 Severity : enhancement
1015 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1018 Frequency : start MDS on uncleanly shutdowned MDS device
1020 Descriptoin: aborting recovery hang on MDS
1021 Details : don't throttle destroy RPCs for the MDT.
1024 Frequency : on remount
1026 Description: external journal device not working after the remount
1027 Details : clear dev_rdonly flag for external journal devices in
1033 Description: shutdown vs evict race
1034 Details : client_disconnect_export vs connect request race.
1035 if client will evicted at this time - we start invalidate
1036 thread without referece to import and import can be freed
1040 Frequency : rare, need acl's on inode.
1042 Description: client can't handle ost additional correctly
1043 Details : if ost was added after client connected to mds client can have
1044 hit lnet_try_match_md ... to big messages to wide striped files.
1045 in this case need teach client to handle config events about add
1046 lov target and update client max ea size at that event.
1048 Severity : enhancement
1050 Description: Changelogs
1051 Details : Changelogs are a lightweight mechanism to track filesystem
1052 metadata and namespace changes. The changelog is recorded
1053 permanently on the MDTs, and is periodically "consumed" / purged
1054 when records are no longer needed.
1056 Severity : enhancement
1058 Description: compact fld format with extents
1059 Details : Store range of seq rather than every seq in FLD. Seq
1060 controller update FLD rather than clients. In Case of CMD, mdt0
1061 has FLD, all other metadata server act as non persistent proxy
1062 for FLD queries and cache fld entries in fld cache.
1067 Description: don't skip ost target if they assigned to file
1068 Details : Drop slow OSCs if we can, but not for requested start idx.
1069 This means "if OSC is slow and it is not the requested
1070 start OST, then it can be skipped, otherwise skip it only
1071 if it is inactive/recovering/out-of-space.
1075 Description: more cleanup in mds_lov
1076 Details : not send LOV EA under replay, we can't know about they size at this
1077 time. Don't allow client connect to mds before any ost connected,
1078 for avoid problems with LOV EA size and returning EIO to client.
1080 Severity : enhancement
1082 Description: Interoperability at server side (Disk interoperability)
1084 Severity : enhancement
1086 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1088 Severity : enhancement
1090 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1092 Severity : enhancement
1094 Description: New client IO stack (CLIO).
1096 Severity : enhancement
1098 Description: Commit on sharing. Eliminate inter-client dependencies between
1099 uncommitted transactions by doing transaction commits.
1100 Thereby clients may recovery independently.
1103 Frequency : Create a symlink file with a very long name
1105 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1106 Details : If there is no extra space in the request for early cancels,
1107 ldlm_req_handles_avail() returns 0 instead of a negative value.
1109 Severity : enhancement
1111 Description: Add /proc entry for import status
1112 Details : The mdc, osc, and mgc import directories now have
1113 an import directory that contains useful import data for debugging
1114 connection problems.
1116 Severity : enhancement
1118 Description: Re-disable certain /proc logging
1119 Details : Enable and disable client's offset_stats, extents_stats and
1120 extents_stats_per_process stats logging on the fly.
1123 Frequency : Only on FC kernels 2.6.22+
1125 Description: oops in statahead
1126 Details : Do not drop reference count for the dentry from VFS when lookup,
1127 VFS will do that by itself.
1129 Severity : enhancement
1131 Description: Generic /proc file permissions
1132 Details : Set /Proc file permissions in a more generic way to enable non-
1133 root users operate on some /proc files.
1137 Description: Hitting mdc_commit_close() ASSERTION
1138 Details : Properly handle request reference release in
1139 ll_release_openhandle().
1143 Description: quota recovery deadlock during mds failover
1144 Details : This patch includes att18982, att18236, att18237 in bz14840.
1146 1. fix osts hang when mds does failover with quotaon
1147 2. prevent watchdog storm when osts threads wait for the
1152 Frequency : only patchless client
1153 Description: add workaround for race between add/remove dentry from hash
1155 Severity : enhancement
1157 Description: Allow OST glimpses to return PW locks
1161 Description: LBUG when llog conf file is full
1162 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1166 Description: Prevent import from entering FULL state when server in recovery
1170 Description: service mount cannot take device name with ":"
1171 Details : Only when device name contains ":/" will mount treat it as
1177 Description: replace ptlrpcd with the statahead thread to interpret the async
1178 statahead RPC callback
1182 Frequency : on recovery
1183 Description: I/O failures after umount during fail back
1184 Details : if client reconnected to restarted server we need join to recovery
1185 instead of find server handler is changed and process self
1186 eviction with cancel all locks.
1188 Severity : enhancement
1190 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1192 Severity : enhancement
1194 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1196 Severity : enhancement
1198 Description: Upcall on Lustre log has been dumped
1199 Details : Allow for a user mode script to be called once a Lustre log has
1200 been dumped. It passes the filename of the dumped log to the
1201 script, the location of the script can be specified via
1202 /proc/sys/lnet/debug_log_upcall.
1207 Description: avoid idr_remove called for id which is not allocated.
1208 Details : Move assigment s_dev for clustered nfs to end of initialization,
1209 to avoid problem with error handling.
1214 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1215 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1216 into CONN_USED_HASH and this prodice warning when put connection
1217 again in unused hash.
1220 Severity : enhancement
1222 Description: File striping can now be set to use an arbitrary pool of OSTs.
1224 Severity : enhancement
1226 Description: Export bytes_read/bytes_write count on OSC/OST.
1230 Description: Early reply size mismatch, MGC loses connection
1231 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1232 the connect flags are properly negotiated.
1237 Description: quotas are not honored with O_DIRECT
1238 Details : all writes with the flag O_DIRECT will use grants which leads to
1239 this problem. Now using OBD_BRW_SYNC to guard this.
1243 Description: add quota statistics
1244 Details : 1. sort out quota proc entries and proc code.
1245 2. add quota statistics
1247 Severity : enhancement
1249 Description: enable quota support for HEAD.
1253 Description: Properly propagate oinfo flags from lov to osc for statfs
1254 Details : restore missing copy oi_flags to lov requests.
1256 Severity : enhancement
1258 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1259 : l_getgroups(8), lst(8), routerstat(8)
1260 Details : included man pages for llobdstat(8), llstat(8),
1261 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1263 Severity : enhancement
1265 Description: configure's --enable-quota should check the
1266 : kernel .config for CONFIG_QUOTA
1267 Details : configure is terminated if --enable-quota is passed but
1268 : no quota support is in kernel
1272 Description: Remove portals compatibility
1273 Details : Remove portals compatibility, not interoperable with releases
1278 Description: Resolve device initialization race
1279 Details : Prevent proc handler from accessing devices added to the
1280 obd_devs array but yet be intialized.
1282 Severity : enhancement
1284 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1286 Severity : enhancement
1288 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1293 Description: Using wrong pointer in osc_brw_prep_request
1294 Details : Access to array[-1] can produce panic if kernel compiled with
1295 CONFIG_PAGE_ALLOC enabled
1299 Description: Client runs out of low memory
1300 Details : Consider only lowmem when counting initial number of llap pages
1304 Description: *optional* service tags registration
1305 Details : if the "service tags" package is installed on a Lustre node
1306 When the filesystem is mounted, a local-node service tag will
1307 be created. See http://inventory.sun.com/ for more information
1308 about the Service Tags asset management system.
1312 Description: Kernel BUG tries to release flock
1313 Details : Lustre does not destroy flock lock before last reference goes
1314 away. So always drop flock locks when client is evicted and
1315 perform unlock regardless of successfulness of speaking to MDS.
1319 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1323 Description: sanity test 65a fails if stripecount of -1 is set
1324 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1329 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1330 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1333 Severity : enhancement
1335 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1339 Description: do not process already freed flock
1340 Details : flock can possibly be freed by another thread before it reaches
1341 to ldlm_flock_completion_ast.
1345 Description: LBUG during stress test
1346 Details : Need properly lock accesses the flock deadlock detection list.
1350 Description: oops in page fault handler
1351 Details : kernel page fault handler can return two special 'pages' in error
1352 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1356 Description: timeout with invalidate import.
1357 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1358 should be handled by ptlrpcd. This produce long age waiting and
1359 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1361 Severity : enhancement
1363 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1367 Description: enable MGS and MDT services start separately
1368 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1369 a co-located MGS without starting the MGS, which is a complement
1370 to 'nosvc' mount option.
1374 Frequency : after recovery
1375 Description: precreate to many object's after del orphan.
1376 Details : del orphan st in oscc last_id == next_id and this triger growing
1377 count of precreated objects. Set flag LOW to skip increase count
1378 of precreated objects.
1382 Frequency : rare, on clear nid stats
1383 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1384 Details : when clean nid stats sometimes try destroy live entry,
1385 and this produce panic in free.
1389 Description: Stack overflow during MDS log replay
1390 ease stack pressure by using a thread dealing llog_process.
1394 Description: wait until IO finished before start new when do lock cancel.
1395 Details : VM protocol want old IO finished before start new, in this case
1396 need wait until PG_writeback is cleared until check dirty flag and
1397 call writepages in lock cancel callback.
1399 Severity : enhancement
1401 Description: using special macro for print time and cleanup in includes.
1405 Description: mds_mfd_close() ASSERTION(rc == 0)
1406 Details : In mds_mfd_close(), we need protect inode's writecount change
1407 within its orphan write semaphore to prevent possible races.
1411 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1415 Frequency : rare, on shutdown ost
1416 Description: don't hit live lock with umount ost.
1417 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1418 use shrink_dcache_sb instead.
1422 Description: don't panic with use echo client
1423 Details : echo client pass NULL as client nid pointer and this produce null
1424 pointer dereference.
1428 Description: fix build on ppc32
1429 Details : compile code with -m64 flag produce wrong object file for ppc32.
1433 Description: add message levels for liblustreapi
1437 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1438 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1439 will result in an -EAGAIN and in the exit call path we call
1440 alloc_rr() which will with increasing aggressiveness attempt to
1441 aquire precreated objects on the minimum number of required OSCs.
1445 Description: Use old size assignment to avoid deadlock
1446 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1447 the scheduling while holding a spinlock. We do not need locking
1448 for size in ll_update_inode() because size is only updated from
1449 the MDS for directories or files without objects, so there is no
1450 other place to do the update, and concurrent access to such inodes
1451 are protected by the inode lock.
1455 Description: resolve "_IOWR redefined" build error on SLES10
1459 Description: dump the memory debugging after all modules are unloaded to
1460 suppress false negative in conf_sanity test 39
1462 Severity : enhancement
1464 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1469 Description: fix SLES kernel versioning
1470 Details : the kernel version for our SLES 10 kernel did not include a "-"
1471 before the "smp" at the end. while this was not a problem in
1472 general, it did mean that software trying to use the kernel
1473 version to try to detect a vendor specific kernel would fail.
1474 this was most evident by the OFED build scripts.
1478 Description: Don't update lov_desc members until making sure they are valid
1479 Details : When updating lov_desc members via proc fs, need fix their
1480 validities before doing the real update.
1484 Description: don't put request into delay list while invalidate in flight.
1485 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1486 import in flight. this produce timeout for invalidate and sometimes
1487 can cause stale data.
1491 Frequency : on ppc only
1492 Description: not convert ost objects for directory because it's not exist.
1493 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1497 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1499 Severity : enhancement
1501 Description: Update to support 2.6.22.14 vanilla kernel.
1505 Frequency : rare, on recovery
1506 Description: read procfs can produce deadlock in some situation
1507 Details : Holding lprocfs lock which send rpc can produce block for destroy
1508 obd objects and this also block reconnect with -EALREADY. This
1509 isn't fix all lprocfs bugs - but make it rare.
1511 Severity : enhancement
1513 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1516 Frequency : frequent on X2 node
1518 Description: mdc_set_open_replay_data LBUG
1519 Details : Set replay data for requests that are eligible for replay.
1523 Description: lustre_mgs: operation 101 on unconnected MGS
1524 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1525 MGC, and late on MGC cannot successfully connect to MGS and a lot
1526 of the error messages complaining that MGS is not connected.
1530 Frequency : on network error
1531 Description: panic with double free request if network error
1532 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1533 but it's true only for synchronus enqueue, for async enqueue
1534 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1537 Severity : enhancement
1539 Description: client-side metadata stat-ahead during readdir(directory readahead)
1540 Details : perform client-side metadata stat-ahead when the client detects
1541 readdir and sequential stat of dir entries therein
1544 Frequency : on start mds
1546 Description: Implement get_info(last_id) in obdfilter.
1549 Frequency : occasional
1551 Description: Correctly check stale fid, not start epoch if ost not support SOM
1552 Details : open with flag O_CREATE need set old fid in op_fid3 because
1553 op_fid2 was overwritten with new generated fid, but mds can answer
1554 with one of these two fids and both is not stale. Setattr
1555 incorrectly started an epoch and assume will be called
1556 done_writing, but without SOM done_writing ever being called.
1559 Frequency : rare, depends on device drivers and load
1561 Description: MDS or OSS nodes crash due to stack overflow
1562 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1563 In some cases, in conjunction with device drivers that use a lot
1564 of stack the MDS (or possibly OSS) service threads could overflow
1565 the stack. One change which was identified to consume additional
1566 stack has been reworked to avoid the extra stack usage.
1569 Frequency : occasional
1571 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1572 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1573 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1574 oscc_has_objects and see OSC has no objests and call
1575 oscc_internal_create to resend create request.
1577 Severity : enhancement
1579 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1581 Severity : enhancement
1583 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1586 Frequency : very rare
1588 Description: Fix replay if there is an un-replied request and open
1589 Details : In some cases, older replay request will revert the
1590 mcd->mcd_last_xid on MDS which is used to record the client's
1591 latest sent request.
1593 Severity : enhancement
1595 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1597 Severity : enhancement
1599 Description: Add rhel5 support to HEAD.
1601 Serverity : enhancement
1603 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1608 Description: Don't allow skipping OSTs if index has been specified.
1609 Details : Don't allow skipping OSTs if index has been specified, make
1610 locking in internal create lots better.
1614 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1615 Details : access to bitfield in structure is always rounded to long
1616 and this produce problem with not atomic change any bit.
1620 Description: Lustre make rpms failed.
1621 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1622 builds Lustre rpms from tarball.
1625 Frequency : rare on shutdown ost
1627 Description: If llog cancel was not send before clean_exports phase, this can
1628 produce deadlock in llog code.
1629 Details : If llog thread has last reference to obd and call class_import_put
1630 this produce deadlock because llog_cleanup_commit_master wait when
1631 last llog_commit_thread exited, but this never success because was
1632 called from llog_commit_thread.
1636 Description: allow userland application know is lost one of stripes.
1637 Details : fill lvb_blocks with error code on ost and return it to
1638 application if error flag found.
1642 Description: NULL lov_tgts causing MDS oops
1643 Details : more safe checks for NULL lov_tgts for avoid oops.
1645 Severity : enhancement
1647 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1651 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1652 Details : If an OST is down the MDS will hang indefinitely in
1653 obd_statfs_rqset() waiting for the statfs data. While for
1654 MDS QOS usage of statfs, it should not stuck in waiting.
1656 Severity : enhancement
1658 Description: remote_acl support
1659 Details : Support ACL-based permission check for remote user.
1660 Support setfacl/getfacl for remote user with the utils
1661 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1662 the system "{s,g}etfacl" utils.
1664 Severity : enhancement
1666 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1668 Severity : enhancement
1670 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1674 Description: make mgs_setparam() handle fsname containing dash
1675 Details : fsname containing a dash does not work with lctl conf_param
1677 Severity : enhancement
1679 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1681 Severity : enhancement
1683 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1687 Frequency : rare, at shutdown
1688 Description: access already free / zero obd_namespace.
1689 Details : if client_disconnect_export was called without force flag set,
1690 and exist connect request in flight, this can produce access to
1691 NULL pointer (or already free pointer) when connect_interpret
1692 store ocd flags in obd_namespace.
1696 Frequency : only at startup
1697 Description: not alloc memory with spinlock held.
1698 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1699 if any spinlock held.
1701 Severity : enhancement
1703 Description: make lustre randomly fail allocating memory
1704 Details : Make lustre randomly failed allocating memory for testing purpose.
1706 Severity : enhancement
1708 Description: lost problems with lov objid file
1709 Details : Fixes some scability and access to not inited memory problems
1710 in work with lov objdid file.
1715 Description: lfs find does not continue on file error
1716 Details : Continue other files processing when a file/dir is absent.
1720 Description: Inconsistent usage of lustre_pack_reply()
1721 Details : Standardize the usage of lustre_pack_reply() such that it
1722 always generate a CERROR on failure.
1725 Frequency : occasional
1727 Description: MDS hang or stay in waiting lock
1728 Details : If client receive lock with CBPENDING flag ldlm need send lock
1729 cancel as separate rpc, to avoid situation when cancel request
1730 can't processed due all i/o threads stay in wait lock.
1734 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1738 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1742 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1746 Description: Update to RHEL4 latest kernel.
1748 Severity : enhancement
1750 Description: Build SLES10 patchless client fails
1751 Details : The configure was broken by run ./configure with
1752 --with-linux-obj=.... argument for patchless client. When the
1753 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1754 can't search header adequately. Use absolute path such as
1755 -I($LINUX)/include instead.
1759 Description: interrupt oig_wait produce painc on resend.
1760 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1761 private set, and this produce situation when rq_ptlrpcd_data not
1762 copyed to new allocated request and triggered LBUG on assert
1763 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1764 ptlrpcd set if request is changed and can be safety changed to use
1769 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1770 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1771 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1772 these macros and replace them with their difinitions instead.
1776 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1777 Details : Update which_patch & target file for SLES10 latest kernel.
1780 Description: Remove client patches from SLES 10 kernel.
1781 Details : This causes SLES 10 clients to behave as patchless clients
1782 even on a Lustre-patched (server) kernel.
1784 Severity : enhancement
1786 Description: self-adjustable client's lru lists
1787 Details : use adaptive algorithm for managing client cached locks lru
1788 lists according to current server load, other client's work
1789 pattern, memory activities, etc. Both, server and client
1790 side namespaces provide number of proc tunables for controlling
1793 Severity : enhancement
1795 Description: light-weight GSS support
1796 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1797 and reduce performance overhead.
1799 Severity : enhancement
1801 Description: Linux keyring support
1802 Details : Support using service of Linux keyring for Lustre GSS internal
1803 context refresh/cache mechanism.
1807 Description: Fix errors in lfs documentation
1808 Details : Fixes man pages
1812 Description: don't use GFP_* in generic Lustre code.
1813 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1817 Description: obdclass is limited by single OBD_ALLOC(idarray)
1818 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1822 Description: warnings with build patchless client with vanila 2.6.19 and up
1823 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1824 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1828 Frequency : Only for SLES
1829 Description: sanity_quota fail test_1
1830 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1831 file, which makes TSTID[2] inunique.
1835 Frequency : Only for SLES9
1836 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1837 Details : Update target/ChangeLog/which_patch .
1841 Frequency : Only for test_10 in sanity-quota.sh
1842 Description: a bug in quota test code
1843 Details : checking a test flag in a wrong place
1847 Frequency : Only for test_13 in sanity-quota.sh
1848 Description: a bug in quota test script
1849 Details : shouldn't check overfull allocated quota
1853 Description: jbd statistics
1854 Details : Port older jbd statistics patch for sles10
1856 Severity : enhancement
1858 Description: Add printing inode info into message about error in writepage.
1863 Description: reply_lock_interpret crash due to race with it and lock cancel.
1864 Details : Do not replay locks that are being cancelled. Do not reference
1865 locks by their address during replay, just by their handle.
1869 Frequency : When flocks are used.
1870 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1871 Details : Flock locks might destroy just granted lock if it could be merged
1872 with another existing flock, this is done in completion handler,
1873 so teach ldlm_cli_enquque_fini that this is a valid case for
1879 Description: Oops in read and write path when failing to allocate lock.
1880 Details : Check if lock allocation failed and return error back.
1884 Description: lstripe command fails for valid OST index
1885 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1886 instead of lov->desc.ld_active_tgt_count.
1890 Description: sanity.sh failed test 103
1891 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1896 Description: df doesn't work properly if diskfs blocksize != 4K
1897 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1901 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1902 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1904 Severity : enhancement
1906 Description: Adaptive timeouts
1907 Details : RPC timeouts adapt to changing server load and network
1908 conditions to reduce resend attempts and improve recovery time.
1912 Description: llapi_file_create() does not allow some changes
1913 Details : add llapi_file_open() that allows specifying the mode and
1914 open flags, and also returns an open file handle.
1918 Description: merge and cleanup kernel patches.
1919 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1923 Description: Add journal checksum support.(Kernel part)
1924 Details : The journal checksum feature adds two new flags i.e
1925 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1926 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1927 indicates that the commit block contains the checksum for
1928 the blocks described by the descriptor blocks. Now commit
1929 record can be sent to disk without waiting for descriptor
1930 blocks to be written to disk. This behavior is controlled
1931 using JBD2_FEATURE_ASYNC_COMMIT flag.
1935 Description: OSS needs mutliple precreate threads
1936 Details : Add ability to start more than one create thread per OSS.
1940 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1941 Details : execute lfs setstripe on client
1945 Description: mds_obd_create error creating tmp object
1946 Details : When the user sets quota on root, llog will be affected and can't
1947 create files and write files.
1950 Frequency : Always on ia64 patchless client, and possibly others.
1952 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1953 Details : This allows the patchless client to be loaded on architectures
1954 without this export.
1958 Description: RedHat Update kernel for RHEL5
1959 Details : Add the kernel config file for RHEL5.
1963 Description: RedHat Update kernel for RHEL5
1964 Details : Modify the kernel config file more closer RHEL5.
1968 Description: Build failure against Centos5 (RHEL5)
1969 Details : Define PAGE_SIZE when it isn't present.
1973 Description: test_7 on sanity-quota.sh can't be used on separate machines
1974 Details : add facet to handle it
1978 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1979 Details : with NFS, the anon dentry's parent was set to itself in
1980 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1981 corresponding dentry other than use rec->ur_name.
1983 Severity : enhancement
1985 Description: lfs setstripe enhancement
1986 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1990 Description: enable data checksumming by default
1991 Details : enable checksum by default, allow --disable-checksum
1992 configure option and "-o nochecksum" mount option. Checksums
1993 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1997 Description: lustre support for RHEL5
1998 Details : Add support for RHEL5.
2002 Description: Client eviction due to failover config
2003 Details : after a connection loss, the lustre client should attempt to
2004 reconnect to the last active server first before trying the
2005 other potential connections.
2009 Description: when mds and osts use different quota unit(32bit and 64bit),
2010 quota will be released repeatly.
2011 Details : void sending multiple quota reqs to mds, which will keep the
2012 status between the reqs.
2016 Description: osts not allocated evenly to files
2017 Details : change the condition to increase offset_idx
2020 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2022 Description: Data corruption for OSTs that are formatted larger than 2TB
2024 Details : When generating the bio request for lustre file writes the
2025 sector number would overflow a temporary variable before being
2026 used for the IO. The data reads correctly from Lustre (which
2027 will overflow in a similar manner) but other file data or
2028 filesystem metadata may be corrupted in some cases.
2032 Description: Tune the kernel for good SCSI performance.
2033 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2034 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2039 Description: rewrite ext2-derived code in obdclass/uuid.c
2040 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2041 from scratch preserving functionality.
2045 Description: "lfs find -obd UUID" prints directories
2046 Details : "lfs find -obd UUID" will return all directory names instead
2047 of just file names. It is incorrect because the directories
2048 do not reside on the OSTs.
2052 Description: use i_size_read and i_size_write in 2.6 port
2053 Details : replace inode->i_size access with i_size_read/write()
2056 Frequency : when removing large files
2058 Description: scheduling issue during removal of large Lustre files
2059 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2060 It causes scheduling issues when removing large files (17TB in the
2064 Frequency : only with liblustre clients on XT3
2066 Description: evictions taking too long
2067 Details : allow llrd to evict clients directly on OSTs
2070 Frequency : only on ppc
2072 Description: /proc/fs/lustre/devices broken on ppc
2073 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2074 We should make sure the type of 'index' is loff_t and then cast
2075 explicitly as needed below. Do not assign an explicitly cast
2079 Frequency : only for rhel5
2081 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2082 Details : Modify the target file & which_kernel.
2084 Severity : enhancement
2086 Description: omit set fsid for export NFS
2087 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2088 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2092 Frequency : Always, for kernels after 2.6.16
2093 Description: Fix warning idr_remove called for id=.. which is not allocated.
2094 Details : Last kernels save old s_dev before kill super and not allow
2095 to restore from callback - restore it before call kill_anon_super.
2099 Description: buffer overruns could theoretically occur
2100 Details : llapi_semantic_traverse() modifies the "path" argument by
2101 appending values to the end of the origin string, and a
2102 overrun may occur. Adding buffer overrun check in liblustreapi.
2106 Description: Fix error on 'ls .' at the top of the Lustre mount.
2107 Details : Don't revalidate dentry if it is a root dentry.
2111 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2112 Details : Modify vm-tunables-rhel4.patch.
2116 Description: Kernel config for 2.6.18-vanilla.
2117 Details : Modify targets/2.6-vanilla.target.in.
2118 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2119 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2120 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2121 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2125 Description: improve handling recoverable errors
2126 Details : if request processig with error which can be recoverable on server
2127 request should be resend, otherwise page released from cache and
2132 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2133 Details : Modify target file & which_patch.
2134 A flaw was found in the IA32 system call emulation provided
2135 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2136 value could be stored in the %RAX register, which could trigger an
2137 out-of-bounds system call table access. An untrusted local user
2138 could exploit this flaw to run code in the kernel
2139 (ie a root privilege escalation). (CVE-2007-4573).
2143 Description: O_DIRECT bypasses client statistics.
2144 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2145 read_ahead_stats were updated. copied the stats section from
2146 osc_send_oap_rpc() into async_internal().
2150 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2154 Description: change order of libsysio includes
2155 Details : '#include sysio.h' should always come before '#include xtio.h'
2159 Description: handle "serious error: objid * already exists" more gracefully
2160 Details : If LAST_ID value on disk is smaller than the objects existing in
2161 the O/0/d* directories, it indicates disk corruption and causes an
2162 LBUG(). If the object is 0-length, then we should use the existing
2163 object. This will help to avoid a full fsck in most cases.
2165 Severity : enhancement
2167 Description: adapt the lustre_config script to support the upgrade case
2168 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2169 server targets to 1.6 in parallel.
2173 Description: To avoid grant space > avaible space when the disk is almost
2174 full. Without this patch you might see the error "grant XXXX >
2175 available" or some LBUG about grant, when the disk is almost
2177 Details : In filter_check_grant, for non_grant cache write, we should
2178 check the left space by if (*left > ungranted + bytes), instead
2179 of (*left > ungranted), because only we are sure the left space
2180 is enough for another "bytes", then the ungrant space should be
2181 increase. In client, we should update cl_avail_grant only there
2182 is OBD_MD_FLGRANT in the reply.
2186 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2187 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2188 described in CVE-2007-4573. This problem could allow untrusted
2189 local users to gain root access.
2192 Frequency : when using O_DIRECT and quotas
2194 Description: Incorrect file ownership on O_DIRECT output files
2195 Details : block usage reported by 'lfs quota' does not take into account
2196 files that have been written with O_DIRECT.
2201 Description: touch file failed when fs is not full
2202 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2203 otherwise we can get ENOSP while fs is not full.
2207 Description: parallel lock callbacks
2208 Details : Instead of sending blocking and completion callbacks as separated
2209 requests, adding them to a set and sending in parallel.
2212 Frequency : only for Cray XT3
2213 Bugzilla : 12829/13455
2214 Description: Changing primary group doesn't change the group lustre assigns to
2216 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2217 overridden with the primary group provided by the group upcall,
2218 whereas the supplied fsgid can be trusted if it is in the list of
2219 supplementary groups returned by the group upcall.
2221 Severity : enhancement
2223 Description: Allow masking D_WARNING, D_ERROR messages from console
2224 Details : Console messages can now be disabled via lnet.printk.
2228 Description: User code with malformed file open parameter crashes client node
2229 Details : Before packing join_file req, all the related reference should be
2230 checked carefully in case some malformed flags cause fake
2231 join_file req on client.
2235 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2236 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2237 mode is equal to the granted mode. However, it should also check
2238 that LDLM_FL_CP_REQD is not set.
2242 Description: Heavy nfs access might result in deadlocks
2243 Details : After ELC code landed, it is now improper to enqueue any mds
2244 locks under och_sem, because enqueue might want to decide to
2245 cancel open locks for same inode we are holding och_sem for.
2249 Description: Client eviction while running blogbench
2250 Details : A lot of unlink operations with concurrent I/O can lead to a
2251 deadlock causing evictions. To address the problem, the number of
2252 oustanding OST_DESTROY requests is now throttled to
2253 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2254 callbacks are processed in priority.
2258 Description: enable ACLs on MDS by default
2259 Details : ACLs must be enabled on MDS by default.
2262 Frequency : PPC/PPC64 only
2264 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2265 Details : fix duplicated definitions between asm-ppc64/types.h and
2266 lustre_types.h on PPC.
2269 Frequency : PPC/PPC64 only
2271 Description: asm-ppc/segment.h does not exist
2272 Details : fix compile issue on PPC.
2276 Description: better handle error messages in extents code
2279 Frequency : RHEL4 only
2281 Description: mkfs is very slow on IA64/RHEL4
2282 Details : A performance regression has been discovered in the MPT Fusion
2283 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2284 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2285 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2287 Severity : enhancement
2289 Description: SNMP support enhancement
2290 Details : Adding total number of sampled request for an MDS node in snmp
2293 Severity : enhancement
2295 Description: Optimize ldlm waiting list processing for PR extent locks
2296 Details : When processing waiting list for read extent lock and meeting read
2297 lock that is same or wider to it that is not contended, skip
2298 processing rest of the list and immediatelly return current
2299 status of conflictness, since we are guaranteed there are no
2300 conflicting locks in the rest of the list.
2304 Description: Time out and refuse to reconnect
2305 Details : When the failover node is the primary node, it is possible
2306 to have two identical connections in imp_conn_list. We must
2307 compare not conn's pointers but NIDs, otherwise we can defeat
2308 connection throttling.
2312 Description: port llog fixes from b1_6 into HEAD
2313 Details : Port llog reference couting and some llog cleanups from b1_6
2314 (bug 10800) into HEAD, for protect from panic and access to already
2315 free llog structures.
2319 Description: Detect stride IO mode in read-ahead
2320 Details : When a client does stride read, read-ahead should detect that and
2321 read-ahead pages according to the detected stride pattern.
2325 Description: data checksumming impacts single node performance
2326 Details : add support for several checksum algorithm. Currently, only CRC32
2327 and Adler-32 are supported. The checksum type can be changed on
2328 the fly via /proc/fs/lustre/osc/*/checksum_type.
2332 Description: use adler32 for page checksums
2333 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2338 Description: build for x2 fails
2339 Details : fix compile issue on Cray systems.
2343 Description: Properly match for duplicate locks
2344 Details : Due to different lock order from skiplists code, we need to
2345 traverse entire list for now
2348 Frequency : only on PPC/SLES10
2350 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2351 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2352 be defined. Add a hack in mkfs_lustre.c to work around this compile
2357 Description: LASSERT on MDS when client holding flock lock dies
2358 Details : ldlm pool logic depends on number of granted locks equal to
2359 number of released locks which is not true for flock locks, so
2360 just exclude such locks from consideration.
2364 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2365 Details : Use fsfilt_sync() for both the whole filesystem sync and
2366 individual file sync to eliminate dangerous inode locking
2367 with I_LOCK that can lead to a deadlock.
2371 Description: performance in 1.6.3
2372 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2373 sure that 1MB requests are not fragmented by the block layer.
2375 Severity : enhancement
2377 Description: organize the server-side client stats on per-nid basis
2378 Details : Change the structure of stats under obdfilter and mds to
2386 The "uuid"s file would list the uuids of _active_ exports.
2387 And the clear entry is to clear all stats and stale nids.
2389 Severity : enhancement
2391 Description: eliminate client locks in face of contention
2392 Details : file contention detection and lockless i/o implementation
2393 for contended files.
2397 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2400 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2402 Description: MDS may lose file striping (and hence file data) in some cases
2403 Details : If there are additional extended attributes stored on the MDS,
2404 in particular ACLs, SELinux, or user attributes (if user_xattr
2405 is specified for the client mount options) then there is a risk
2406 of attribute loss. Additionally, the Lustre file striping
2407 needs to be larger than default (e.g. striped over all OSTs),
2408 and an additional attribute must be stored initially in the
2409 inode and then increase in size enough to be moved to the
2410 external attribute block (e.g. ACL growing in size) for file
2415 Description: skiplist implementation simplification
2416 Details : skiplists are used to group compatible locks on granted list
2417 that was implemented as tracking first and last lock of each
2418 lock group the patch changes that to using doubly linked lists
2422 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2423 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2424 because LASSERT on both the data supplied by a client, and the
2425 data on disk is dangerous and incorrect.
2427 Severity : enhancement
2429 Description: Slow truncate/writes to huge files at high offsets.
2430 Details : Directly associate cached pages to lock that protect those pages,
2431 this allows us to quickly find what pages to write and remove
2432 once lock callback is received.
2436 Description: more ldlm soft lockups
2437 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2438 starve other threads from the resource lock for a long time in
2439 case of long waiting queue, so change the debug level from
2440 D_OTHER to the less frequently used D_INFO.
2442 Severity : enhancement
2444 Description: add -gid, -group, -uid, -user options to lfs find
2448 Description: Hung threads in invalidate_inode_pages2_range
2449 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2450 once one is completed. As a result, some RPCs are stuck in the
2451 queue and are never sent.
2455 Description: filter threads hungs on waiting journal commit
2456 Details : Cleanup filter group llog code, then only filter group llog will
2457 be only created in the MDS/OST syncing process.
2461 Description: Procfs and llog threads access destoryed import sometimes.
2462 Details : Sync the import destoryed process with procfs and llog threads by
2463 the import refcount and semaphore.
2465 Severity : enhancement
2467 Description: openlock cache of b1_6 port to HEAD
2472 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2473 Details : If insertion of an extent fails, then discard the inode
2474 preallocation and free data blocks else it can lead to duplicate
2479 Description: don't always update ctime in ext3_xattr_set_handle()
2480 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2481 In some cases the ctime should not be updated, for example for
2482 2.0->1.8 compatibility it is necessary to delete an xattr and it
2483 should not update the ctime.
2487 Bugzilla : 15713/16362
2488 Description: Assertion in iopen_connect_dentry in 1.6.3
2489 Details : looking up an inode via iopen with the wrong generation number can
2490 populate the dcache with a disconneced dentry while the inode
2491 number is in the process of being reallocated. This causes an
2492 assertion failure in iopen since the inode's dentry list contains
2493 both a connected and disconnected dentry.
2497 Description: assertion failure in ldlm_handle2lock()
2498 Details : fix a race between class_handle_unhash() and class_handle2object()
2499 introduced in lustre 1.6.5 by bug 13622.
2504 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2505 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2510 Description: oops in sd_iostats_seq_show()
2511 Details : unloading/reloading the scsi low level driver triggers a kernel
2512 bug when trying to access the sd iostat file.
2517 Description: Kernel panics during QLogic driver reload
2518 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2519 patch, causing memory corruption.
2524 Description: journal_dev option does not work in b1_6
2525 Details : pass mount option during pre-mount.
2527 Severity : enhancement
2529 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2530 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2531 extent information of a file. It can be used to map logical blocks
2532 in a file to physical blocks in the block device.
2536 Description: LDLM soft lockups - improvement
2537 Details : It is be possible to send the lock handle along with each read
2538 or write request because the client is already doing a lock match
2539 itself so there isn't any reason the OST should have to re-do that
2543 Frequency : only on Cray X2
2545 Description: X2 build failures
2546 Details : fix build failures on Cray X2.
2550 Description: xid & resent requests
2551 Details : Initialize RPC XID from clock at startup (randomly if clock is
2554 Severity : enhancement
2556 Description: Add lustre_start utility to start or stop multiple Lustre servers
2561 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2562 Details : In case of memory pressure, list_del() can be called twice on
2563 req->rq_history_list, causing a kernel oops.
2567 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2568 Details : fix stack overflow in the distributed lock manager by defering
2569 export eviction after a failed AST to the elt thread instead of
2570 handling it in the dlm interpret routine.
2574 Description: Convert some comments to new format.
2575 Details : Update documenting comments to match doxygen conventions.
2579 Description: Grammar fixes.
2580 Details : A couple of trivial spelling fixes.
2584 Description: OSD_COUNTERS-mandatory
2585 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2586 to simplify the code.
2590 Description: simplify cmm_device freeing logic.
2591 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2595 Description: Add lockdep support to dt_object_operations locking interface.
2596 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2597 parameter indicating lock ordering. Update mdd code to use new
2602 Description: Introduce failloc constants for lockless IO tests.
2603 Details : Add two new failloc constants to test lockless IO. Only one of
2604 them in implemented---another is checked in yet to be landed
2609 Description: Add lockdep support for inode mutex.
2610 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2611 in the situations where Lustre has to lock more than one inode
2616 Description: Add optional invariants checking support.
2617 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2618 configure switch. This macro is to be used for consistency and
2619 sanity checks that are too expensive to be left in `production'
2624 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2625 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2630 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2631 Details : Use spin_lock_nested() in (the only) situation where more than
2632 one ldlm_lock is locked simultaneously. Also, fix possible
2633 dead-lock in ldlm_lock_change_resource() by enforcing particular
2638 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2639 Details : Instead of specifying each ldlm_lock call-back through separate
2640 parameter, wrap them into struct ldlm_callback_suite.
2644 Description: Kill join_lru obd method and its callers.
2645 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2646 in memory---a requirement implemented through ->o_join_lru() obd
2647 method in HEAD. Remove this method and its users.
2651 Description: Add asynchronous ldlm ENQUEUE completion handler.
2652 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2653 case that stock ldlm_completion_ast() cannot handle as it waits
2654 until lock is granted. Introduce new ldlm_completion_ast_async()
2655 for this. Also comment ldlm_completion_ast().
2659 Description: ldlm_error <-> errno conversion.
2660 Details : Add functions to map (rather arbitrary) between LDLM error codes
2661 and standard errno values. CLIO needs this to prevent LDLM specific
2662 constants from escaping ldlm and osc.
2666 Description: Kill unused ldlm_handle2lock_ns() function.
2667 Details : Kill unused ldlm_handle2lock_ns() function.
2671 Description: Add lu_ref support to ldlm_lock
2672 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2673 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2674 and ldlm_resource. LDLM interface has to be changed, because code that
2675 releases a reference on a lock, has to "know" what reference this is.
2676 In the most frequent case
2678 lock = ldlm_handle2lock(handle);
2680 LDLM_LOCK_PUT(lock);
2682 no changes are required. When any other reference (received _not_ from
2683 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2684 instead of LDLM_LOCK_PUT().
2686 Arguably, changes are pervasive, and interface requires some discipline
2687 for proper use. On the other hand, it was very instrumental in finding
2688 a few leaked lock references.
2692 Description: Add ldlm_lock_addref_try().
2693 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2694 attempts to addref a lock that might be being canceled
2699 Description: Add ldlm_weigh_callback().
2700 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2701 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2702 hard-coded `number of pages' logic.
2706 Description: Add lockdep annotations to llog code.
2707 Details : Use appropriately tagged _nested() locking calls in the places
2708 where llog takes more than one ->lgh_lock lock.
2712 Description: Add loi_kms_set().
2713 Details : Wrap kms updates into a helper function.
2717 Description: Constify instances of struct lsm_operations.
2718 Details : Constify instances of struct lsm_operations.
2722 Description: lu_conf support.
2723 Details : On a server, a file system object is uniquely identified
2724 by a fid, which is sufficient to locate and load all object
2725 state (inode). On a client, on the other hand, more data are
2726 necessary instantiate an object. Change lu_object_find() and
2727 friends to take additional `lu_conf' argument describing object.
2728 Typically this includes layout information.
2732 Description: lu_context fixes.
2733 Details : Introduce new lu_context functions that are needed on the client
2734 side, where some system threads (ptlrpcd) are shared by multiple
2735 modules, and so cannot be stopped during module shutdown.
2739 Description: Add start and stop methods to lu_device_type_operations.
2740 Details : Introduce two new methods in lu_device_type_operations, that are
2741 invoked when first instance of a given type is created and last one
2742 is destroyed respectively. This is need by CLIO.
2746 Description: Add lu_ref support to struct lu_device.
2747 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2748 track leaked references.
2752 Description: Introduce lu_kmem_descr.
2753 Details : lu_kmem_descr and its companion interface allow to create
2754 and destroy a number of kmem caches at once.
2758 Description: Fix lu_object finalization race.
2759 Details : Fix a race between lu_object_find() finding an object and its
2760 concurrent finalization. This race is (most likely) not possible
2761 on the server, but might happen on the client.
2765 Description: Introduce lu_ref interface.
2766 Details : lu_ref is a debugging module allowing to track references to
2767 a given object. It is quite cpu expensive, and has to be
2768 explicitly enabled with --enable-lu_ref. See usage description
2773 Description: Factor lu_site procfs stats into a separate function.
2774 Details : Separate lu_site stats printing code into a separate function
2775 that can be reused on a client.
2779 Description: Constify instances of struct {lu,dt,md}_device_operations.
2780 Details : Constify instances of struct {lu,dt,md}_device_operations.
2784 Description: Introduce struct md_site and move meta-data specific parts of
2785 struct lu_site here.
2786 Details : Move md-specific fields out of struct lu_site into special struct
2787 md_site, so that lu_site can be used on a client.
2791 Description: Kill mdd_lov_destroy().
2792 Details : Remove unused mdd code.
2796 Description: Add st_block checking to multistat.c.
2797 Details : Add st_block checking to multistat.c.
2801 Description: Add lu_ref support to struct obd_device.
2802 Details : Add lu_ref tracking to obd_device.
2806 Description: Kill obd_set_fail_loc().
2807 Details : Remove unused code.
2811 Description: Add special type for ptlrpc_request interpret functions.
2812 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2813 there. Actual usage will be in CLIO.
2817 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2818 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2819 doesn't work with lockdep.
2823 Description: Add rwv.c test program.
2824 Details : New testing program exercising readv(2) and writev(2) (Qian).
2828 Description: Add sendfile.c test program.
2829 Details : New testing program exercising sendfile(2) (Jay).
2833 Description: Ratelimit a message that can be very frequent.
2834 Details : Ratelimit a memory allocation failure message that can
2839 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2840 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2844 Description: Liblustre build fixes.
2845 Details : Liblustre build fixes.
2849 Description: libcfs: add cfs_{need,cond}_resched() interface.
2850 Details : libcfs: add cfs_{need,cond}_resched() definition and
2851 implementations for Linux, NT, and liblustre.
2853 Severity : enhancement
2855 Description: More exported tunables for mballoc
2856 Details : Add support for tunable preallocation window and new tunables for
2857 large/small requests
2861 Description: Detect corruption of block bitmap and checking for preallocations
2862 Details : Checks validity of on-disk block bitmap. Also it does better
2863 checking of number of applied preallocations. When corruption is
2864 found, it turns filesystem readonly to prevent further corruptions.
2868 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2869 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2870 window to 0. If it does hit cache miss, and read-pattern is still
2871 stride-io mode, does not reset the stride window, but also does not
2872 increase the stride window length in this case.
2876 Frequency : only for big-endian servers
2877 Description: Check if system is big-endian while mounting fs with extents feature
2878 Details : Mounting a filesystem with extents feature will fail on big-endian
2879 systems since ext3-based ldiskfs is not supported on big-endian
2880 systems. This can be over-riden with "bigendian_extents" mount option.
2882 Severity : enhancement
2884 Description: The root squash functionality
2885 Details : A security feature, which is to prevent users from being able
2886 to mount lustre on their desktop, run as root, and delete
2887 all of the files in the filesystem. The goal is accomplished by
2888 remapping user id (UID) and group id (GID) of the root user to
2889 a UID and GID specified by the system administartor via Lustre
2890 configuration management server (MGS). The functionality also
2891 allows to specify sets of clients for which the remapping does
2896 Description: Excessive recovery window
2897 Details : With AT enabled, the recovery window can be excessively long (6000+
2898 seconds). To address this problem, we no longer use
2899 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2900 timeout no longer depends on the service time, it is set to
2901 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2902 time via pb_service_time.
2906 Description: Watchdog triggered on MDS failover
2907 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2908 the OSTs know that the MDS "UUID" can be reused for the same export
2909 from a different NID, so we do not need to wait for the export to be
2913 Frequency : rare, only if using MMP with Linux RAID
2915 Description: MMP doesn't work with Linux RAID
2916 Details : While using HA for Lustre servers with Linux RAID, it is possible
2917 that MMP will not detect multiple mounts. To make this work we
2918 need to unplug the device queue in RAID when the MMP block is being
2919 written. Also while reading the MMP block, we should read it from
2920 disk and not the cached one.
2922 Severity : enhancement
2924 Description: open file using fid
2925 Details : A file can be opened using just its fid, like
2926 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2929 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2932 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2933 Details : add proper locking for creation of jbd_4k slab cache
2937 Description: MMP check in ext3_remount() fails without displaying any error
2938 Details : When multiple mount protection fails during remount, proper error
2941 Severity : enhancement
2943 Description: Allow stripe size to be up to 4G-64k
2944 Details : Fix math logic to allow large stripe sizes.
2948 Description: add check for >8TB ldiskfs filesystems
2949 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2950 Don't allow >8TB ldiskfs filesystems to be mounted without
2951 force_over_8tb mount option
2956 Descriptoin: OSS complains "too many blocks" and crashes
2957 Details : In filter_commitrw_write():
2958 1. lquota_chkquota();
2959 2. obdo_from_inode();
2960 3. lquota_pending_commit();
2961 "2." has changed uid and gid in this case, then caused
2962 lquota_chkquota() and lquota_pending_commit() to use
2963 different uid and gid. That is the root of the bug.
2967 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2968 Details : Detect and print out the 'stripe count = -1' case properly.
2972 Description: Out or order replies might be lost on replay
2973 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2974 request with tid smaller than one currently being added, add it
2975 to the start, not end of the list.
2977 --------------------------------------------------------------------------------
2979 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2981 * Support for kernels:
2982 2.6.5-7.283 (SLES 9),
2983 2.6.9-55.EL (RHEL 4),
2984 2.6.16.46-0.14 (SLES 10),
2985 2.6.18.8 vanilla (kernel.org)
2986 * Client support for unpatched kernels:
2987 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2988 2.6.16 - 2.6.22 vanilla (kernel.org)
2989 * Due to recently discovered recovery problems, we do not recommend
2990 using patchless RHEL 4 clients with this or any earlier release.
2991 * Recommended e2fsprogs version: 1.39.cfs8
2992 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2993 * Starting with this release, the ldiskfs backing filesystem required
2994 by Lustre is now in its own package, lustre-ldiskfs. This package
2995 should be installed. It is versioned separately from Lustre and
2996 may be released separately in future.
3000 Description: block reactivating mgc import until all deactivates complete
3001 Details : Fix race when failing back MDT/MGS to itself (testing)
3004 Frequency : at statup only
3006 Description: mds_lov_synchronize race leads to various problems
3007 Details : simultaneous MDT->OST connections at startup can cause the
3008 sync to abort, leaving the OSC in a bad state.
3010 Severity : enhancement
3012 Description: add optional extra BUILD_VERSION info
3013 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3014 to override the lustre version.
3017 Frequency : 2.6.18 servers only
3019 Description: ll_kern_mount() doesn't release the module reference
3020 Details : The ldiskfs module reference count never drops down to 0
3021 because ll_kern_mount() doesn't release the module reference.
3026 Description: server LBUG when using old ost_num_threads parameter
3027 Details : Accept the old ost_num_threads parameter but warn that it
3028 is deprecated, and fix an off-by-one error that caused an LBUG.
3033 Description: Transient SCSI error results in persistent IO issue
3034 Details : iobuf->dr_error is not reinitialized to 0 between two
3038 Frequency : sometimes when underlying device returns I/O errors
3040 Description: OSTs not going read-only during write failures
3041 Details : OSTs are not remounted read-only when the journal commit threads
3042 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3043 instead of the ext3 wrappers.
3047 Description: poor connect scaling with increasing client count
3048 Details : Don't run filter_grant_sanity_check for more than 100 exports
3049 to improve scaling for large numbers of clients.
3052 Frequency : SLES10 only
3054 Description: sanity-quota.sh quotacheck failed: rc = -22
3055 Details : Quotas cannot be enabled on SLES10.
3058 Frequency : liblustre clients only
3060 Description: getdirentries does not give error when run on compute nodes
3061 Details : getdirentries does not fail when the size specified as an argument
3062 is too small to contain at least one entry
3064 Severity : enhancement
3066 Description: Add LNET router traceability for debug purposes
3067 Details : If a checksum failure occurs with a router as part of the
3068 IO path, the NID of the last router that forwarded the bulk data
3069 is printed so it can be identified.
3074 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3075 Details : Due to a race condition, liblustre clients were occasionally
3076 evicted incorrectly.
3078 Severity : enhancement
3080 Description: lfs setstripe use optional parameters instead of postional
3083 Severity : enhancement
3085 Description: Nanosecond timestamp support for ldiskfs
3086 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3087 resolution timestamps. There is not yet support for this at
3088 the Lustre filesystem level.
3091 Frequency : during server recovery
3093 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3094 Details : request with rq_no_resend flag not awake l_wait_event if they get
3098 Frequency : nfs export on patchless client
3100 Description: connectathon hang when test nfs export over patchless client
3101 Details : Disconnected dentry cannot be found with lookup, so we do not need
3102 to unhash it or make it invalid
3105 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3106 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3107 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3112 Description: open req refcounting wrong on reconnect
3113 Details : If reconnect happened between getting open reply from server and
3114 call to mdc_set_replay_data in ll_file_open, we will schedule
3115 replay for unreferenced request that we are about to free.
3116 Subsequent close will crash in variety of ways.
3117 Check that request is still eligible for replay in
3118 mdc_set_replay_data().
3123 Description: disable writes to filesystem when reading health_check file
3124 Details : the default for reading the health_check proc file has changed
3125 to NOT do a journal transaction and write to disk, because this
3126 can cause reads of the /proc file to hang and block HA state
3127 checking on a healthy but otherwise heavily loaded system. It
3128 is possible to return to the previous behaviour during configure
3129 with --enable-health-write.
3131 Severity : enhancement
3133 Description: 64-bit inode version
3134 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3135 to the inode. This will be required for version-based recovery.
3140 Description: MDS fails to start if a duplicate client export is detected
3141 Details : in some rare cases it was possible for a client to connect to
3142 an MDS multiple times. Upon recovery the MDS would detect this
3143 and fail during startup. Handle this more gracefully.
3145 Severity : enhancement
3147 Description: Add -o localflock option to simulate old noflock
3149 Details : This will achieve local-only flock/fcntl locks
3155 Description: log_commit_thread vs filter_destroy race leads to crash
3156 Details : Take import reference before releasing llog record semaphore
3161 Description: Wrong request locking in request set processing
3162 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3163 imp_delayed_list, in this place should be used imp_lock.
3166 Frequency : when reconnection
3168 Description: Grant Leak when osc reconnect to OST
3169 Details : When osc reconnect ost, OST(filter) should check whether it
3170 should grant more space to client by comparing fed_grant and
3171 cl_avail_grant, and return the granted space to client instead
3172 of "new granted" space, because client will call osc_init_grant
3173 to update the client grant space info.
3176 Frequency : when client reconnect to OST
3178 Description: Grant Leak when osc do resend and replay bulk write
3179 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3180 bulk write request, because the grant info while be sync between
3181 OSC and OST when reconnect, and we should ignore the grant info
3182 these of resend/replay write req.
3187 Description: Grant space more than available left space sometimes.
3188 Details : When then OST is about to be full, if two bulk writing from
3189 different clients came to OST. Accord the avaliable space of the
3190 OST, the first req should be permitted, and the second one
3191 should be denied by ENOSPC. But if the seconde arrived before
3192 the first one is committed. The OST might wrongly permit second
3193 writing, which will cause grant space > available space.
3196 Frequency : when client is evicted
3198 Description: Grant might be wrongly erased when osc is evicted by OST
3199 Details : when the import is evicted by server, it will fork another
3200 thread ptlrpc_invalidate_import_thread to invalidate the
3201 import, where the grant will be set to 0. While the original
3202 thread will update the grant it got when connecting. So if
3203 the former happened latter, the grant will be wrongly errased
3204 because of this race.
3209 Description: Checking Stale with correct fid
3210 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3211 to check whether it is stale, because sometimes, we want the
3212 enqueue happened anyway, and op_data.fid2 will not be initialized.
3214 Severity : enhancement
3216 Description: update patchless client
3217 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3220 Frequency : only with 2.4 kernel
3222 Description: random memory corruption
3223 Details : size of struct ll_inode_info is to big for union inode.u and this
3224 can be cause of random memory corruption.
3229 Description: Memory leak in recovery
3230 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3231 It should also check obd_fail before fsfilt_start, otherwise if
3232 fsfilt_start return -EROFS,(failover mds during mds recovery).
3233 then the req will return with repmsg->transno = 0 and rc = EROFS.
3234 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3235 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3236 be freed no matter whether fsfilt_commit success or not.
3239 Frequency : only with huge count clients
3241 Description: Prevents from taking the superblock lock in llap_from_page for
3243 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3244 for llap_from_page call in ll_removepage prevents from taking the
3245 superblock lock for a soon died page.
3250 Description: Not check open intent error before release open handle
3251 Details : in some rare cases, the open intent error is not checked before
3252 release open handle, which may cause
3253 ASSERTION(open_req->rq_transno != 0), because it tries to release
3254 the failed open handle.
3259 Description: Set cat log bitmap only after create log success.
3260 Details : in some rare cases, the cat log bitmap is set too early. and it
3261 should be set only after create log success.
3265 Description: Accessing a block bevice can re-enable I/O when Lustre is
3266 tearing down a device.
3267 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3272 Description: service threads may hog cpus when there are a lot of requests
3273 Details : Insert cond_resched to give other threads a chance to use some CPU
3278 Description: the cat log was not initialized in recovery
3279 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3280 unlink log on mds will not be initialized until mds post
3281 recovery. And also in mds post recovery, the unlink log will
3282 initialization will be done asynchronausly, so there will be race
3283 between add unlink log and unlink log initialization.
3287 Description: brw_stats were being printed incorrectly
3288 Details : brw_stats were being printed as log2 but all of them were not
3289 recorded as log2. Also remove some code duplication arising from
3290 filter_tally_{read,write}.
3294 Frequency : rare, only in recovery.
3295 Description: ASSERTION(req->rq_type != LI_POISON) failed
3296 Details : imp_lock should be held while iterating over imp_sending_list for
3297 prevent destroy request after get timeout in ptlrpc_queue_wait.
3301 Description: replay-single.sh test 52 fails
3302 Details : A lock's skiplist need to be cleanup when it being unlinked
3303 from its resource list.
3307 Description: Short directio read returns full requested size rather than
3309 Details : Direct I/O operations should return actual amount of bytes
3310 transferred rather than requested size.
3312 Severity : enhancement
3314 Description: metadata RPC reduction (e.g. for rm performance)
3315 Details : decrease the amount of synchronous RPC between clients and servers
3316 by canceling conflicing lock before the operation on the client
3317 and packing thier handles into the main operation RPC to server.
3319 Severity : enhancement
3321 Description: Async OSC create to avoid the blocking unnecessarily.
3322 Details : If an OST has no remaining object, system will block on the
3323 creation when it needs to create a new object on this OST. Now,
3324 ways use pre-created objects when available, instead of blocking on
3325 an empty osc while others are not empty. If we must block, we
3326 block for the shortest possible period of time.
3330 Description: improve handling recoverable errors
3331 Details : if request processig with error which can be recoverable on server
3332 request should be resend, otherwise page released from cache and
3335 Severity : enhancement
3337 Description: refine locking for avoid write wrong info into lov_objid file
3338 Details : fix possible races with add new target and write/update data in
3341 --------------------------------------------------------------------------------
3343 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3348 Frequency : on some architectures
3350 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3351 Details : Uninitialized flags sometimes cause configuration commands to
3355 Frequency : patchless clients only
3357 Description: missing __iget() symbol export
3358 Details : The __iget() symbol export is missing. To avoid the need for
3359 this on patchless clients the deathrow inode reaper is turned
3360 off, and we depend on the VM to clean up old inodes. This
3361 dependency was during via the fix for bug 12181.
3363 --------------------------------------------------------------------------------
3365 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3367 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3368 INTEROPERATE with older versions automatically. In many cases a
3369 special upgrade step is needed. Please read the
3370 user documentation before upgrading any part of a live system.
3371 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3372 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3373 * WARNING: Lustre configuration and startup changes are required with
3374 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3378 Severity : enhancement
3380 Description: MountConf
3381 Details : Lustre configuration is now managed via mkfs and mount
3382 commands instead of lmc and lconf. New obd types (MGS, MGC)
3383 are added for dynamic configuration management. See
3384 https://mail.clusterfs.com/wikis/lustre/MountConf for
3387 Severity : enhancement
3389 Description: dynamic OST addition
3390 Details : OSTs can now be added to a live filesystem
3392 Severity : enhancement
3394 Description: startup order invariance
3395 Details : MDTs and OSTs can be started in any order. Clients only
3396 require the MDT to complete startup.
3398 Severity : enhancement
3400 Description: parallel, asynchronous orphan cleanup
3401 Details : orphan cleanup is now performed in separate threads for each
3402 OST, allowing parallel non-blocking operation.
3404 Severity : enhancement
3406 Description: optimized stripe assignment
3407 Details : stripe assignments are now made based on ost space available,
3408 ost previous usage, and OSS previous usage, in order to try
3409 to optimize storage space and networking resources.
3411 Severity : enhancement
3413 Description: Permanently set tunables
3414 Details : All writable /proc/fs/lustre tunables can now be permanently
3415 set on a per-server basis, at mkfs time or on a live system.
3417 Severity : enhancement
3419 Description: Lustre message v2
3420 Details : Add lustre message format v2.
3422 Severity : enhancement
3424 Description: client OST exclusion list
3425 Details : Clients can be started with a list of OSTs that should be
3426 declared "inactive" for known non-responsive OSTs.
3430 Description: SPEC SFS validation failure on NFS v2 over lustre.
3431 Details : Changes the blocksize for regular files to be 2x RPC size,
3432 and not depend on stripe size.
3434 Severity : enhancement
3436 Description: Multiple MD RPCs in flight.
3437 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3438 To avoid overly-overloading MDS, introduce a limit on number of
3439 MDS RPCs in flight for a single client and add /proc controls
3440 to adjust this limit.
3442 Severity : enhancement
3444 Description: client read/write statistics
3445 Details : Add client read/write call usage stats for performance
3446 analysis of user processes.
3447 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3448 file access. extents_stats shows chunk size distribution.
3449 extents_stats_per_process show chunk size distribution per
3452 Severity : enhancement
3454 Description: mds statistics
3455 Details : Add detailed mds operations statistics in
3456 /proc/fs/lustre/mds/*/stats.
3460 Description: Failure of copying files with lustre special EAs.
3461 Details : Client side always return success for setxattr call for lustre
3462 special xattr (currently only "trusted.lov").
3467 Description: Refcount LNET uuids
3468 Details : The global LNET uuid list grew linearly with every startup;
3469 refcount repeated list entries instead of always adding to
3472 Severity : enhancement
3474 Description: Dynamic service threads
3475 Details : Within a small range, start extra service threads
3476 automatically when the request queue builds up.
3479 Frequency : mixed-endian client/server environments
3481 Description: mixed-endian crashes
3482 Details : The new msg_v2 system had some failures in mixed-endian
3485 Severity : enhancement
3487 Description: Easy OST removal
3488 Details : OSTs can be permanently deactivated with e.g. 'lctl
3489 conf_param lustre-OST0001.osc.active=0'
3491 Severity : enhancement
3493 Description: MGS proc entries
3494 Details : Added basic proc entries for the MGS showing what filesystems
3497 Severity : enhancement
3499 Description: provide MGS failover
3500 Details : Added config lock reacquisition after MGS server failover.
3502 Severity : enhancement
3504 Description: add Linux 2.4 support
3505 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3509 Description: a large application tries to do I/O to the same resource and dies
3510 in the middle of it.
3511 Details : Check the req->rq_arrival time after the call to
3512 ost_brw_lock_get(), but before we do anything about
3513 processing it & sending the BULK transfer request. This
3514 should help move old stale pending locks off the queue as
3515 quickly as obd_timeout.
3518 Frequency : when an incorrect nid is specified during startup
3520 Description: ptlrpc connect to non-existant node causes kernel crash
3521 Details : LNET can't be re-entered from an event callback, which
3522 happened when we expire a message after the export has been
3523 cleaned up. Instead, hand the zombie cleanup off to another
3526 Severity : enhancement
3528 Description: plain/inodebits lock performance improvement
3529 Details : Grouping plain/inodebits in granted list by their request modes
3530 and bits policy, thus improving the performance of search through
3534 Frequency : only if OST filesystem is corrupted
3536 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3537 Details : for a short time RPCs with bulk IO are in the replay list,
3538 but replay of bulk IOs is unimplemented. If the OST filesystem
3539 is corrupted due to disk cache incoherency and then replay is
3540 started it is possible to trip an assertion. Avoid putting
3541 committed RPCs into the replay list at all to avoid this issue.
3544 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3546 Description: System hang on startup
3547 Details : This bug allowed the liblustre (e.g. catamount) client to
3548 return to the app before handling all startup RPCs. This
3549 could leave the node unresponsive to lustre network traffic
3550 and manifested as a server ptllnd timeout.
3552 Severity : enhancement
3554 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3555 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3556 boolean to control whether to print peer debug info when a
3557 client's RPC times out.
3560 Frequency : only for kernels with patches from Lustre below 1.4.3
3562 Description: Remove old rdonly API
3563 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3566 Frequency : only for devices with external journals
3568 Description: Set external device read-only also
3569 Details : During a commanded failover stop, we set the disk device
3570 read-only while the server shuts down. We now also set any
3571 external journal device read-only at the same time.
3574 Frequency : when upgrading from 1.4 while trying to change parameters
3576 Description: The wrong (new) MDC name was used when setting parameters for
3577 upgraded MDT's. Also allows changing of OSC (and MDC)
3578 parameters if --writeconf is specified at tunefs upgrade time.
3581 Frequency : when setting specific ost indicies
3583 Description: QOS code breaks on skipped indicies
3584 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3585 created with --index need not be sequential.
3587 Severity : enhancement
3589 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3590 Details : The uninit_groups feature works in conjunction with the kernel
3591 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3592 up the pass1 processing of e2fsck. This is a read-only feature
3593 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3594 mount filesystems that have had this feature enabled.
3596 Severity : enhancement
3598 Description: Improve multi-block allocation algorithm to avoid fragmentation
3599 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3600 allocation locality and avoid filesystem fragmentation.
3602 ------------------------------------------------------------------------------
3604 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3606 * Support for kernels:
3607 2.4.21-47.0.1.EL (RHEL 3)
3608 2.6.5-7.283 (SLES 9)
3609 2.6.9-42.0.10.EL (RHEL 4)
3610 2.6.12.6 vanilla (kernel.org)
3611 2.6.16.27-0.9 (SLES 10)
3612 * Recommended e2fsprogs version: 1.39.cfs5
3614 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3618 Frequency : occasional, depends on client load and configuration
3619 Bugzilla : 12181, 12203
3620 Description: data loss for recently-modified files
3622 Details : In some cases it is possible that recently written or created
3623 files may not be written to disk in a timely manner (this should
3624 normally be within 30s unless client IO load is very high).
3625 The problem appears as zero-length files or files that are a
3626 multiple of 1MB in size after a client crash or client eviction
3627 that are missing data at the end of the file.
3629 This problem is more likely to be hit on clients where files are
3630 repeatedly created and unlinked in the same directory, clients
3631 have a large amount of RAM, have many CPUs, the filesystem has
3632 many OSTs, the clients are rebooted frequently, and/or the files
3633 are not accessed by other nodes after being written.
3635 The presence of the problem can be detected by looking at
3636 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3637 smaller than the second (nr_unused) then dirty files will not
3638 be flushed automatically to disk. "sync; sleep 10" should be
3639 run several times on the node before unmounting it to update
3640 Lustre (this is also safe to run on nodes without this problem).
3642 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3643 that can cause this same problem, so customers using that kernel
3644 also need to update the kernel in addition to Lustre. In order
3645 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3647 It is normal that files written just before a client crash (less
3648 than 30s) may not yet have been flushed to disk, even for local
3652 Frequency : frequent on thin XT3 nodes
3654 Description: UUID collision on thin XT3 Linux nodes
3655 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3656 randomly, since we relied on an insufficiently-seeded PRNG.
3661 Description: OSS hangs after "All ost request buffers busy"
3662 Details : A deadlock between quota and journal operations caused OSS
3663 hangs after printing "All ost request buffers busy."
3666 Frequency : always on liblustre builds
3668 Description: Cleanup compiler warnings on liblustre
3671 Frequency : always on liblustre builds on XT3
3673 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3677 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3678 > 32000 subdirectories
3679 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3680 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3681 subdirectory count crosses 32000. This will aid e2fsck to
3682 correctly handle more than 32000 subdirectories.
3685 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3687 Description: System hang on startup
3688 Details : This bug allowed the liblustre (e.g. catamount) client to
3689 return to the app before handling all startup RPCs. This
3690 could leave the node unresponsive to lustre network traffic
3691 and manifested as a server ptllnd timeout.
3693 Severity : enhancement
3695 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3696 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3697 boolean to control whether to print peer debug info when a
3698 client's RPC times out.
3703 Description: make O_SYNC working on 2.6 kernels
3704 Details : 2.6 kernels use different method for mark pages for write,
3705 so need add a code to lustre for O_SYNC work.
3710 Description: Failure to close file and release space on NFS
3711 Details : Put inode details into lock acquired in ll_intent_file_open.
3712 Use mdc_intent_lock in ll_intent_open to properly
3713 detect all kind of errors unhandled by mdc_enqueue.
3718 Description: proc file read during shutdown sometimes raced obd removal,
3720 Details : Add lock to prevent obd access after proc file removal.
3723 Frequency : Only for files larger than 4GB on 32-bit clients.
3725 Description: improperly doing page alignment of locks
3726 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3727 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3732 Description: RPCs being resent when they shouldn't be
3733 Details : Some RPCs that should not be resent are being resent. This
3734 can cause inconsistencies in the RPC state machine. Do not
3735 resend such requests.
3738 Frequency : rare, only with NFS export
3740 Description: Crash on NFS re-export node
3741 Details : under very unusual load conditions an assertion is hit in
3742 ll_intent_file_open()
3745 Frequency : only if OST filesystem is corrupted
3747 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3748 Details : for a short time RPCs with bulk IO are in the replay list,
3749 but replay of bulk IOs is unimplemented. If the OST filesystem
3750 is corrupted due to disk cache incoherency and then replay is
3751 started it is possible to trip an assertion. Avoid putting
3752 committed RPCs into the replay list at all to avoid this issue.
3757 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3758 Details : Large single O_DIRECT read and write calls can fail to allocate
3759 a sufficiently large buffer to process the request. In case of
3760 allocation failure the allocation is retried with a smaller
3761 buffer and broken into smaller requests.
3763 Severity : enhancement
3765 Description: Add -o localflock option to simulate old noflock behaviour.
3766 Details : This will achieve local-only flock/fcntl locks coherentness.
3771 Description: versioning check is incomplete
3772 Details : Checking the version difference of client vs. server, report
3773 error if the gap is too big.
3778 Description: add support PG_writeback bit
3779 Details : add support for PG_writeback bit for Lustre, for more carefull
3780 work with page cache in 2.6 kernel. This also fix some deadlocks
3781 and remove hack for work O_SYNC with 2.6 kernel.
3783 Severity : enhancement
3785 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3786 Details : The uninit_groups feature works in conjunction with the kernel
3787 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3788 up the pass1 processing of e2fsck. This is a read-only feature
3789 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3790 mount filesystems that have had this feature enabled.
3792 Severity : enhancement
3794 Description: Improve multi-block allocation algorithm to avoid fragmentation
3795 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3796 allocation locality and avoid filesystem fragmentation.
3798 ------------------------------------------------------------------------------
3800 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3802 * Support for kernels:
3803 2.6.9-42.0.3.EL (RHEL 4)
3804 2.6.5-7.276 (SLES 9)
3805 2.4.21-47.0.1.EL (RHEL 3)
3806 2.6.12.6 vanilla (kernel.org)
3807 2.6.16.21-0.8 (SLES10)
3808 * Recommended e2fsprogs version: 1.39.cfs2-0
3810 * The backwards-compatible /proc/sys/portals symlink has been removed
3811 in this release. Before upgrading, please ensure that you change
3812 any configuration scripts or /etc/sysctl.conf files that access
3813 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3814 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3815 in advance of the upgrade on any system running Lustre 1.4.6 or
3816 newer, since /proc/sys/lnet was added in that version.
3817 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3821 Frequency : only when quota is used
3823 Description: avoid scanning export list for quota master
3824 Details : Change the algorithms to avoid scanning export list in order
3825 to improve the efficiency.
3828 Frequency : MDS failover only, very rarely
3830 Description: "went back in time" messages on mds failover
3831 Details : The greatest transno may be lost when the current operation
3832 finishes with an error (transno==0) and the client's last_rcvd
3833 record is over-written. Save the greatest transno in the
3834 mds_last_transno for this case.
3837 Frequency : always for specific kernels and striping counts
3839 Description: client may get "Matching packet too big" without ACL support
3840 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3841 when trying to access files in certain configurations. The
3842 clients should in fact be denied when mounting because they do
3843 not understand ACLs.
3846 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3848 Description: many clients connecting with IO in progress causes connect timeouts
3849 Details : Avoid synchronous journal commits to avoid delays caused by many
3850 clients connecting/disconnecting when bulk IO is in progress.
3851 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3852 OST_IO_PORTAL to avoid delays behind potentially many pending
3856 Frequency : occasionally with multiple writers to a single file
3858 Description: shared writes to file may result in wrong size reported by stat()
3859 Details : Allow growing of kms when extent lock is cancelled
3862 Frequency : always with random mmap IO to multi-striped file
3864 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3865 Details : Only if the hole is at the end of OST object so that kms is too
3866 small. Fix is to increase kms accordingly in ll_nopage.
3869 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3871 Description: writes to a missing object would leak memory on the OST
3872 Details : If there is an inconsistency between the MDS and OST filesystems,
3873 such that the MDS references an object that doesn't exist, writes
3874 to that object will leak memory due to incorrect cleanup in the
3875 error handling path, eventually running out of memory on the OST.
3880 Description: Creating too long symlink causes lustre errors
3881 Details : Check symlink and name lengths before sending requests to MDS.
3884 Frequency : only if flock is enabled (not on by default)
3886 Description: posix locks not released on fd closure on 2.6.9+
3887 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3888 kernels, this caused such locks not to be released on fd close and
3889 then assertions on fs unmount about still used locks.
3892 Frequency : MDS failover only, very rarely
3894 Description: clients may get ASSERTION(granted_lock != NULL)
3895 Details : When request was taking a long time, and a client was resending
3896 a getattr by name lock request. The were multiple lock requests
3897 with the same client lock handle and
3898 mds_getattr_name->fixup_handle_for_resent_request found one of the
3899 lock handles but later failed with ASSERTION(granted_lock != NULL).
3904 Description: handle->h_buffer_credits > 0, assertion failure
3905 Details : h_buffer_credits is zero after truncate, causing assertion
3906 failure. This patch extends the transaction or creates a new
3910 Frequency : NFS re-export or patchless client
3911 Bugzilla : 11179, 10796
3912 Description: Crash on NFS re-export node (__d_move)
3913 Details : We do not want to hash the dentry if we don't have a lock.
3914 But if this dentry is later used in d_move, we'd hit uninitialised
3915 list head d_hash, so we just do this to init d_hash field but
3916 leave dentry unhashed.
3919 Frequency : NFS re-export or patchless client
3921 Description: NFS exports has problem with symbolic link
3922 Details : lustre client didn't properly install dentry when re-exported
3923 to NFS or running patchless client.
3926 Frequency : NFS re-export or patchless client
3928 Description: Various nfs/patchless fixes.
3929 Details : fixes reuse disconected alias for lookup process - this fixes
3930 warning "find_exported_dentry: npd != pd",
3931 fix permission error with open files at nfs.
3932 fix apply umask when do revalidate.
3935 Frequency : occasional
3937 Description: Crash on NFS re-export node
3938 Details : calling clear_page() on the wrong pointer triggered oops in
3939 generic_mapping_read().
3942 Frequency : rarely, using O_DIRECT IO
3944 Description: unaligned directio crashes client with LASSERT
3945 Details : check for unaligned buffers before trying any requests.
3948 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3950 Description: stale data returned from RAID cache
3951 Details : If only a small amount of IO is done to the RAID device before
3952 reading it again it is possible to get stale data from the RAID
3953 cache instead of reading it from disk.
3956 Frequency : always for sles10 kernel
3958 Description: sles10 support
3959 Details : ll_follow_link: compile fixes and using of nd_set_link
3960 under newer kernels.
3963 Frequency : depends on arch, kernel and compiler version, always on sles10
3966 Description: recursive or deep enough symlinks cause stack overflow
3967 Details : getting rid of large stack-allocated variable in
3971 Frequency : depends on hardware
3973 Description: lustre write performance loss in the SLES10 kernel
3974 Details : the performance loss is caused by using of write barriers in the
3975 ext3 code. The SLES10 kernel turns barrier support on by
3976 default. The fix is to undo that change for ldiskfs.
3978 ------------------------------------------------------------------------------
3980 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3982 * Support for kernels:
3983 2.6.9-42.0.3EL (RHEL 4)
3984 2.6.5-7.276 (SLES 9)
3985 2.4.21-47.0.1.EL (RHEL 3)
3986 2.6.12.6 vanilla (kernel.org)
3990 Frequency : quota enabled and large files being deleted
3992 Description: releasing more than 4GB of quota at once hangs OST
3993 Details : If a user deletes more than 4GB of files on a single OST it
3994 will cause the OST to spin in an infinite loop. Release
3995 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4000 Description: statfs data retrieved from /proc may be stale or zero
4001 Details : When reading per-device statfs data from /proc, in the
4002 {kbytes,files}_{total,free,avail} files, it may appear
4003 as zero or be out of date.
4006 Frequency : systems with MD RAID1 external journal devices
4008 Description: lconf's call to blkid is confused by RAID1 journal devices
4009 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4010 of returning all block devices that match the journal UUID.
4013 Frequency : always, for aggregate stripe size over 4GB
4015 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4016 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4017 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4020 Frequency : always if "lfs find" used on a local file/directory
4022 Description: "lfs find" segfaults if used on a local file/directory
4023 Details : The case where a directory component was not specified wasn't
4024 handled correctly. Handle this properly.
4027 Frequency : always on ppc64
4029 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4030 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4031 perform endianness conversion.
4034 Frequency : rarely (truncate to non-zero file size after write under load)
4035 Bugzilla : 10730, 10687
4036 Description: Files padded with zeros to next 4K multiple
4037 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4038 it is possible that files that are truncated to a non-zero size
4039 immediately after being written are filled with zero bytes beyond
4040 the truncated size. No file data is lost.
4042 Severity : enhancement
4044 Description: Allow recovery/failover for liblustre clients.
4045 Details : liblustre clients were unaware of failover configurations until
4048 Severity : enhancement
4050 Description: user file locks should fail when not mounting with flock option
4051 Details : Set up an error-returning stub in ll_file_operations.lock field
4052 to prevent incorrect behaviour when client is mounted without
4053 flock option. Also, set up properly f_op->flock field for
4057 Frequency : always on ia64
4059 Description: "lfs df" loops on printing out MDS statfs information
4060 Details : The obd_ioctl_data was not initialized and in some systems
4061 this caused a failure during the ioctl that did not return
4062 an error. Initialize the struct and return an error on failure.
4065 Frequency : SLES 9 only
4067 Description: Error of copying files with lustre special EAs as root
4068 Details : Client side always return success for setxattr call for lustre
4069 special xattr (currently only "trusted.lov").
4072 Frequency : rarely on clusters with both ia64+i386 clients
4074 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4075 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4076 clients are doing concurrent writes to the same file it is
4077 possible that the ia64 clients may LASSERT because the OST
4078 extent locks are not PAGE_SIZE aligned. Ensure that grown
4079 locks are always aligned on the request boundary.
4082 Frequency : specific use, occasional
4084 Description: Overwriting in use executable truncates on-disk binary image
4085 Details : If one node attempts to overwrite an executable in use by
4086 another node, we now correctly return ETXTBSY instead of
4087 truncating the file.
4089 Severity : enhancement
4091 Description: Async OSC create to avoid the blocking unnecessarily.
4092 Details : If an OST has no remaining object, system will block on the creating
4093 when need to create a new object on this OST. Now, ways use
4094 pre-created objects when available, instead of blocking on an
4095 empty osc while others are not empty. If we must block, we block
4096 for the shortest possible period of time.
4101 Description: chmod on Lustre root is propagated to other clients
4102 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4103 invalid by the follow_mount time.
4106 Frequency : liblustre clients only
4108 Description: Race in 'instant cancel' lock handling could lead to such locks
4109 never to be granted in case of SMP MDS
4110 Details : Do not destroy not yet granted but cbpending locks in
4114 Frequency : replay/resend of open
4116 Description: non null lock assetion failure in mds_intent_policy
4117 Details : Trying to replay/resend lockless open requests resulted in
4118 mds_open() returning 0 with no lock. Now it sets a flag if
4119 a lock is going to be returned.
4121 Severity : enhancement
4123 Description: Checksum enhancements
4124 Details : New checksum enhancements allow for resending RPCs that failed
4127 Severity : enhancement
4129 Description: Tunables on number of dirty pages in cacche
4130 Details : Allow to set limit on number of dirty pages cached.
4135 Description: client crash on unmount - lock still has references
4136 Details : In some error handling cases it was possible to leak a lock
4137 reference on a client while accessing a file. This was not
4138 harmful to the client during operation, but would cause the
4139 client to crash when the filesystem is unmounted.
4142 Frequency : specific case, rare
4144 Description: ETXTBSY on mds though file not in use by client
4145 Details : ETXTBSY is no longer incorrectly returned when attempting to
4146 chmod or chown a directory that the user previously tried to
4147 execute or a currently-executing binary.
4150 Frequency : extremely rare except on liblustre-based clients
4152 Description: Lustre space not freed when files are deleted
4153 Details : Clean up open-unlinked files after client eviction. Previously
4154 the unlink was skipped and the files remained as orphans.
4159 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4160 Details : In some cases it was possible to send a blocking callback to
4161 a client doing a glimpse, even though that client didn't get
4162 a lock granted. When the glimpse lock is cancelled on the OST
4163 the freed lock is left on the waiting list and corrupted the list.
4166 Frequency : all core dumps
4168 Description: Broke core dumps to lustre
4169 Details : Negative dentry may be unhashed if parent does not have UPDATE
4170 lock, but some callers, e.g. do_coredump, expect dentry to be
4171 hashed after successful create, hash it in ll_create_it.
4173 ------------------------------------------------------------------------------
4175 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4177 * Support for kernels:
4178 2.6.9-42.0.2.EL (RHEL 4)
4179 2.6.5-7.276 (SLES 9)
4180 2.4.21-40.EL (RHEL 3)
4181 2.6.12.6 vanilla (kernel.org)
4185 Frequency : always on RHEL 3
4187 Description: Number of open files grows over time
4188 Details : The number of open files grows over time, whether or not
4189 Lustre is started. This was due to a filp leak introduced
4190 by one of our kernel patches.
4192 ------------------------------------------------------------------------------
4194 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4196 * Support for kernels:
4197 2.6.9-42.EL (RHEL 4)
4198 2.6.5-7.276 (SLES 9)
4199 2.4.21-40.EL (RHEL 3)
4200 2.6.12.6 vanilla (kernel.org)
4205 Bugzilla : 5719, 9635, 9792, 9684
4206 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4207 Details : If a server is under heavy load and cannot reply to new
4208 connection requests before the client resends the (re)connect,
4209 the connection handling code can behave badly if two service
4210 threads are concurrently handing separate (re)connections from
4211 the same client. Add better locking to the connection handling
4212 code, and ensure that only a single connection will be processed
4213 for a given client UUID, even if the lock is dropped.
4215 Severity : enhancement
4217 Description: add TCP zero-copy support to kernel
4218 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4219 sends if the hardware supports scatter-gather and checksumming.
4220 This allows socklnd to do client-write and server-read more
4221 efficiently and reduce CPU utilization from skbuf copying.
4224 Frequency : only if NFS exporting from client
4226 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4227 Details : If setting attributes on a file created under NFS that had
4228 never been opened it would be possible to oops the client
4229 if the file had no objects.
4232 Frequency : always for liblustre
4234 Description: liblustre client does MDS+OSTs setattr RPC for each write
4235 Details : When doing a write from a liblustre client, the client
4236 incorrectly issued an RPC to the MDS and each OST the file was
4237 striped over in order to update the timestamps. When writing
4238 with small chunks and many clients this could overwhelm the MDS
4239 with RPCs. In all cases it would slow down the write because
4240 these RPCs are unnecessary.
4242 Severity : enhancement
4244 Description: allow number of MDS service threads to be changed at module load
4245 Details : It is now possible to change the number of MDS service threads
4246 running. Adding "options mds mds_num_threads={N}" to the MDS's
4247 /etc/modprobe.conf will set the number of threads for the next
4248 time Lustre is restarted (assuming the "mds" module is also
4249 reloaded at that time). The default number of threads will
4250 stay the same, 32 for most systems.
4255 Description: OST crash if filesystem is unformatted or corrupt
4256 Details : If an OST is started on a device that has never been formatted
4257 or if the filesystem is corrupt and cannot even mount then the
4258 error handling cleanup routines would dereference a NULL pointer.
4263 Description: NULL pointer deref in llap_from_page.
4264 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4265 incorrect) mapping if the page was truncated/reclaimed while it was
4266 searched for. Check for this condition and skip such pages when
4267 doing readahead. Introduce extra check to llap_from_page() to
4268 verify page->mapping->host is non-NULL (so page is not anonymous).
4271 Frequency : Sometimes when using sys_sendfile
4273 Description: "page not covered by a lock" warnings from ll_readpage
4274 Details : sendfile called ll_readpage without right page locks present.
4275 Now we introduced ll_file_sendfile that does necessary locking
4276 around call to generic_file_sendfile() much like we do in
4280 Frequency : with certain MDS communication failures at client mount time
4282 Description: NULL pointer deref after failed client mount
4283 Details : a client connection request may delayed by the network layer
4284 and not be sent until after the PTLRPC layer has timed out the
4285 request. If the client fails the mount immediately it will try
4286 to clean up before the network times out the request. Add a
4287 reference from the request import to the obd device and delay
4288 the cleanup until the network drops the request.
4291 Frequency : occasionally during client (re)connect
4293 Description: assertion failure during client (re)connect
4294 Details : processing a client connection request may be delayed by the
4295 client or server longer than the client connect timeout. This
4296 causes the client to resend the connection request. If the
4297 original connection request is replied in this interval, the
4298 client may trip an assertion failure in ptlrpc_connect_interpret()
4299 which thought it would be the only running connect process.
4302 Frequency : only with obd_echo servers and clients that are rebooted
4304 Description: kernel BUG accessing uninitialized data structure
4305 Details : When running an obd_echo server it did not start the ping_evictor
4306 thread, and when a client was evicted an uninitialized data
4307 structure was accessed. Start the ping_evictor in the RPC
4308 service startup instead of the OBD startup.
4310 Severity : enhancement
4311 Bugzilla : 10193 (patchless)
4312 Description: Remove dependency on various unexported kernel interfaces.
4313 Details : No longer need reparent_to_init, exit_mm, exit_files,
4314 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4317 Frequency : rare (only users of deprecated and unsupported LDAP config)
4319 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4320 Details : LDAP apparently contains 'lustreName' attributes instead of
4321 'name'. A simple remapping of the name is sufficient.
4324 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4326 Description: waiting_locks_callback trips kernel BUG if client is evicted
4327 Details : Running with the dump_on_timeout debug flag turned on makes
4328 it possible that the waiting_locks_callback() can try to dump
4329 the Lustre kernel debug logs from an interrupt handler. Defer
4330 this log dumping to the expired_lock_main() thread.
4332 Severity : enhancement
4334 Description: Support NFS exporting on 2.6 kernels.
4335 Details : Implement non-rawops metadata methods for NFS server to use without
4336 changing NFS server code.
4339 Frequency : very rare (synthetic metadata workload only)
4341 Description: two racing renames might cause an MDS thread to deadlock
4342 Details : Running the "racer" program may cause one MDS thread to rename
4343 a file from being the source of a rename to being the target of
4344 a rename at exactly the same time that another thread is doing
4345 so, and the second thread has already enqueued these locks after
4346 doing a lookup of the target and is trying to relock them in
4347 order. Ensure that we don't try to re-lock the same resource.
4350 Frequency : only very large systems with liblustre clients
4352 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4353 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4354 all OSTs in parallel. This allows the eviction of stale liblustre
4355 clients to proceed much faster than if they were done in series,
4356 and also offers similar improvements for other set_info RPCs.
4361 Description: excessive CPU usage during initial read phase on client
4362 Details : During the initial read phase on a client, it would agressively
4363 retry readahead on the file, consuming too much CPU and impacting
4364 performance (since 1.4.5.8). Improve the readahead algorithm
4365 to avoid this, and also improve some other common cases (read
4366 of small files in particular, where "small" is files smaller than
4367 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4372 Description: MDS crash when receiving packet with unknown intent.
4373 Details : Do not LBUG in unknown intent case, just return -EFAULT
4375 Severity : enhancement
4376 Bugzilla : 9293, 9385
4377 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4378 Details : Do not serialize getattr (non-intent version) and statfs.
4381 Frequency : occasional, when OST network is overloaded/intermittent
4383 Description: client evicted by OST after bulk IO timeout
4384 Details : If a client sends a bulk IO request (read or write) the OST
4385 may evict the client if it is unresposive to its data GET/PUT
4386 request. This is incorrect if the network is overloaded (takes
4387 too long to transfer the RPC data) or dropped the OST GET/PUT
4388 request. There is no need to evict the client at all, since
4389 the pinger and/or lock callbacks will handle this, and the
4390 client can restart the bulk request.
4393 Frequency : Always when mmapping file with no objects
4395 Description: client crashes when mmapping file with no objects
4396 Details : Check that we actually have objects in a file before doing any
4397 operations on objects in ll_vm_open, ll_vm_close and
4403 Description: Request leak when working with deleted CWD
4404 Details : Introduce advanced request refcount tracking for requests
4405 referenced from lustre intent.
4407 Severity : Enhancement
4409 Description: Cache open file handles on client.
4410 Details : MDS now will return special lock along with openhandle, if
4411 requested and client is allowed to hold openhandle, even if unused,
4412 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4413 closing files for every read/write openration.
4415 Severity : Enhancement
4417 Description: Cache open negative dentries on client when possible.
4418 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4419 negative dentries on lock revocation.
4424 Description: Remounting a client read-only wasn't possible with a zconf mount
4425 Details : It wasn't possible to remount a client read-only with llmount.
4427 Severity : enhancement
4428 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4429 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4430 MPICH is included to add Lustre-specific ADIO interfaces.
4431 This is based closely on the UFS ADIO layer and only differs
4432 in file creation, in order to allow the OST striping to be set.
4433 This is user-contributed code and not supported by CFS.
4438 Description: extended inode attributes (immutable, append-only) work improperly
4439 when 2.4 and 2.6 kernels are used on client/server or vice versa
4440 Details : Introduce kernel-independent values for these flags.
4442 Severity : enhancement
4445 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4446 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4447 granularity. Now, Fractional MB granularity is be supported,
4448 this is very useful for low memory system.
4450 Severity : enhancement
4452 Description: Getattr by fid
4453 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4454 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4455 have locks and avoids getattr rpc requests for it on every stat.
4458 Frequency : Always, for filesystems larger than 2TB
4460 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4461 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4462 filesystems larger than 2TB to be mounted with mballoc enabled.
4465 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4467 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4468 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4469 filesystems larger than 2TB is mounted then the kernel will
4470 silently corrupt the start of the filesystem. CONFIG_LBD is
4471 enabled for all CFS-supported kernels, but the possibility of
4472 this happening with a modified kernel config exists.
4474 Severity : enhancement
4476 Description: add client O_DIRECT support for 2.6 kernels
4477 Details : It is now possible to do O_DIRECT reads and writes to files
4478 in the Lustre client mountpoint on 2.6 kernel clients.
4480 Severity : enhancement
4482 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4483 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4484 parallel, not waiting for response from every OST before sending
4485 a rpc to the next OST.
4490 Description: setattr vs write race when updating file timestamps
4491 Details : Client processes that update a file timestamp into the past
4492 right after writing to the file (e.g. tar) it is possible that
4493 the updated file modification time can be reset to the current
4494 time due to a race between processing the setattr and write RPC.
4496 Severity : enhancement
4498 Description: Bring 'lfs find' closer in line with regular Linux find.
4499 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4500 -print0 options and obtains all the needed info through the lustre
4503 Severity : enhancement
4505 Description: support up to 1024 configured devices on one node
4506 Details : change obd_dev array from statically allocated to dynamically
4507 allocated structs as they are first used to reduce memory usage
4512 Description: Flush dirty partially truncated pages during truncate
4513 Details : Immediatelly flush partially truncated pages in filter_setattr,
4514 this way we completely avoid having any pages in page cache on OST
4515 and can retire ugly workarounds during writes to flush such pages.
4520 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4521 Details : protect inode from truncation within vfs_unlink() context
4522 just take a reference before calling vfs_unlink() and release it
4523 when parent's i_sem is free.
4528 Description: last_id value checked outside lock on OST caused LASSERT failure
4529 Details : If there were multiple MDS->OST object precreate requests in
4530 flight, it was possible that the OST's last object id was checked
4531 outside a lock and incorrectly tripped an assertion. Move checks
4532 inside locks, and discard old precreate requests.
4535 Frequency : always, if extents are used on OSTs
4537 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4538 Details : OSTs using the extents format would not zero the high 16 bits of
4539 the index physical block number. This is not a problem for any
4540 OST filesystems smaller than 16TB, and no kernels support ext3
4541 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4542 new/modified files) and can be fixed for existing filesystems
4543 with e2fsprogs-1.39-cfs1.
4548 Description: import connection selection may be incorrect if timer wraps
4549 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4550 import connections to be ignored if the 32-bit jiffies counter
4551 wraps. Use a 64-bit jiffies counter.
4554 Frequency : very large clusters immediately after boot
4556 Description: LNET request buffers exhausted under heavy short-term load
4557 Details : If a large number of client requests are generated on a service
4558 that has previously never seen so many requests it is possible
4559 that the request buffer growth cannot keep up with the spike in
4560 demand. Instead of dropping incoming requests, they are held in
4561 the LND until the RPC service can accept more requests.
4564 Frequency : Sometimes during replay
4566 Description: Assertion failure in ll_local_open after replay.
4567 Details : If replay happened on an open request reply before we were able
4568 to set replay handler, reply will become not swabbed tripping the
4569 assertion in ll_local_open. Now we set the handler right after
4570 recognising of open request
4573 Frequency : very rare
4575 Description: kernel reports "badness in vsnprintf"
4576 Details : Reading from the "recovery_status" /proc file in small chunks
4577 may cause a negative length in lprocfs_obd_rd_recovery_status()
4578 call to vsnprintf() (which is otherwise harmless). Exit early
4579 if there is no more space in the output buffer.
4581 Severity : enhancement
4583 Description: clear OBD RPC statistics by writing to them
4584 Details : It is now possible to clear the OBD RPC statistics by writing
4585 to the "stats" file.
4590 Description: Client mtime is not the same on different clients after utimes
4591 Details : In some cases, the client was using the utimes() syscall on
4592 a file cached on another node. The clients now validate the
4593 ctime from the MDS + OSTs to determine which one is right.
4598 Description: Inability to activate failout mode
4599 Details : lconf script incorrectly assumed that in pythong string's numeric
4600 value is used in comparisons.
4603 Frequency : always with multiple stripes per file
4605 Description: Inefficient object allocation for mutli-stripe files
4606 Details : When selecting which OSTs to stripe files over, for files with
4607 a stripe count that divides evenly into the number of OSTs,
4608 the MDS is always picking the same starting OST for each file.
4609 Return the OST selection heuristic to the original design.
4614 Description: mount failures may take full timeout to return an error
4615 Details : Under some heavy load conditions it is possible that a
4616 failed mount can wait for the full obd_timeout interval,
4617 possibly several minutes, before reporting an error.
4618 Instead return an error as soon as the status is known.
4620 Frequency : quota enabled and large files being deleted
4622 Description: releasing more than 4GB of quota at once hangs OST
4623 Details : If a user deletes more than 4GB of files on a single OST it
4624 will cause the OST to spin in an infinite loop. Release
4625 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4630 Description: statfs data retrieved from /proc may be stale or zero
4631 Details : When reading per-device statfs data from /proc, in the
4632 {kbytes,files}_{total,free,avail} files, it may appear
4633 as zero or be out of date.
4636 Frequency : systems with MD RAID1 external journal devices
4638 Description: lconf's call to blkid is confused by RAID1 journal devices
4639 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4640 of returning all block devices that match the journal UUID.
4643 Frequency : always, for aggregate stripe size over 4GB
4645 Description: assertion fails when trying to use 4GB stripe size
4646 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4647 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4650 Frequency : always on ppc64
4652 Description: the first write on an ext3 filesystem with mballoc got stuck
4653 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4654 perform endianness conversion.
4656 ------------------------------------------------------------------------------
4658 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4660 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4661 INTEROPERATE with older versions automatically. Please read the
4662 user documentation before upgrading any part of a live system.
4663 * WARNING: Lustre networking configuration changes are required with
4664 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4667 * Support for kernels:
4668 2.6.9-22.0.2.EL (RHEL 4)
4669 2.6.5-7.244 (SLES 9)
4670 2.6.12.6 vanilla (kernel.org)
4673 Severity : enhancement
4674 Bugzilla : 7981/8208
4675 Description: Introduced Lustre Networking (LNET)
4676 Details : LNET is new networking infrastructure for Lustre, it includes
4677 a reorganized network configuration mode (see the user
4678 documentation for full details) as well as support for routing
4679 between different network fabrics. Lustre Networking Devices
4680 (LNDs) for the supported network fabrics have also been
4681 created for this new infrastructure.
4683 Severity : enhancement
4684 Description: Introduced Access control lists
4685 Details : clients can set ACLs on files and directories in order to have
4686 more fine-grained permissions than the standard Unix UGO+RWX.
4687 The MDS must be started with the "-o acl" mount option.
4689 Severity : enhancement
4690 Description: Introduced filesystem quotas
4691 Details : Administrators may now establish per-user quotas on the
4694 Severity : enhancement
4696 Description: Configuration change for the XT3
4697 The PTLLND is now used to run Lustre over Portals on the XT3
4698 The configure option(s) --with-cray-portals are no longer used.
4699 Rather --with-portals=<path-to-portals-includes> is used to
4700 enable building on the XT3. In addition to enable XT3 specific
4701 features the option --enable-cray-xt3 must be used.
4706 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4707 Details : A race between allocating a new llcd and re-getting the llcd_lock
4708 allowed another thread to grab newly-allocated llcd.
4710 Severity : enhancement
4712 Description: 2.6 OST async journal commit and locking fix to improve performance
4713 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4714 2.6 kernels are now async as they already were in 2.4 kernels so
4715 that they can commit concurrently with the network bulk transfer.
4716 For block-allocated files the filter allocation semaphore is held
4717 to avoid filesystem fragmentation during allocation. BKL lock
4718 removed for 2.6 xattr operations where it is no longer needed.
4723 Description: lconf incorrectly determined whether two IP networks could talk
4724 Details : In some more complicated routing and multiple-network
4725 configurations, lconf will avoid trying to make a network
4726 connection to a disjoint part of the IP space. It was doing the
4727 math incorrectly for one set of cases.
4732 Description: Fix for potential infinite loop processing records in an llog.
4733 Details : If an llog record is corrupted/zeroed, it is possible to loop
4734 forever in llog_process(). Validate the llog record length
4735 and skip the remainder of the block on error.
4738 Frequency : occasional (liblustre only)
4740 Description: liblustre could not open files whose last component is a symlink
4741 Details : sysio_path_walk() would incorrectly pass the open intent to
4742 intermediate path components.
4745 Frequency : rare (liblustre only with non-standard tuning)
4746 Bugzilla : 7201 (7350)
4747 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4748 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4749 after it has been released, only do this for referenced locks
4751 Severity : enhancement
4753 Description: specifying an (invalid) directory default stripe_size of -1
4754 would reset the directory default striping
4755 Details : stripe_size -1 was used internally to signal directory stripe
4756 removal, now use "all default" to signal dir stripe removal
4757 as a directory striping of "all default" is not useful
4760 Frequency : common for large clusters running liblustre clients
4762 Description: doing an ls when liblustre clients are running is slow
4763 Details : sending a glimpse AST to a liblustre client waits for every AST
4764 to time out, as liblustre clients will not respond. Since they
4765 cannot cache data we refresh the OST lock LVB from disk instead.
4767 Severity : enhancement
4769 Description: doing an ls at the same time as file IO can be slow
4770 Details : enqueue and other "small" requests can be blocked behind many
4771 large IO requests. Create a new OST IO portal for non-IO
4772 requests so they can be processed faster.
4775 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4777 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4778 created file with mode 000
4779 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4780 allow an ftruncate() as a non-root user to a file with mode 000.
4781 Lustre now respects this flag to disable mode checks when
4782 truncating a file owned by the user
4785 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4788 Description: Revoking locks from clients that went dead or catatonic might take
4790 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4791 cancellation of such locks instant on servers without waiting for
4792 any reply from clients. Clients drops these locks when cancel
4793 notification from server is received without replying.
4796 Frequency : liblustre-only, when liblustre client dies or becomes busy
4798 Description: Doing ls on Linux clients can take a long time with active
4800 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4801 granting such locks to it in the first place if possible. Locks
4802 are taken by proxy on the OST during the read or write and
4803 dropped immediately afterward. Add connect flags handling, do
4804 not grant locks to liblustre clients for glimpse ASTs.
4806 Severity : enhancement
4808 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4809 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4810 sequential reads and behaves badly with applications doing random
4811 reads. Improve it by reading ahead at least read region, and
4812 avoiding excessive large RPC for small reads.
4814 Severity : enhancement
4816 Description: Creating more than 1000 files for a single job may cause a load
4817 imbalance on the OSTs if there are also a large number of OSTs.
4818 Details : qos_prep_create() uses an OST index reseed value that is an
4819 even multiple of the number of available OSTs so that if the
4820 reseed happens in the middle of the object allocation it will
4821 still utilize the OSTs as uniformly as possible.
4826 Description: OST or MDS may oops in ping_evictor_main()
4827 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4828 but doesn't restart at beginning of obd_exports_timed list
4831 Severity : enhancement
4833 Description: improve by-nid export eviction on the MDS and OST
4834 Details : allow multiple exports with the same NID to be evicted at one
4835 time without re-searching the exports list.
4838 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4840 Description: MDS may oops in groups_free()
4841 Details : in rare race conditions a newly allocated group_info struct is
4842 freed again, and this can be NULL. The 2.4 compatibility code
4843 for groups_free() checked for a NULL pointer, but 2.6 did not.
4846 Frequency : common for liblustre clients doing little filesystem IO
4847 Bugzilla : 9352, 7313
4848 Description: server may evict liblustre clients accessing contended locks
4849 Details : if a client is granted a lock or receives a completion AST
4850 with a blocking AST already set it would not reply to the AST
4851 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4854 Frequency : lfs setstripe, only systems with more than 160 OSTs
4856 Description: unable to set striping with a starting offset beyond OST 160
4857 Details : llapi_create_file() incorrectly limited the starting stripe
4858 index to the maximum single-file stripe count.
4861 Frequency : LDAP users only
4863 Description: lconf did not handle in-kernel recovery with LDAP properly
4864 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4866 Severity : enhancement
4868 Description: bind OST threads to NUMA nodes to improve performance
4869 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4870 node and do their allocations there to localize memory access
4872 Severity : enhancement
4874 Description: llmount can determine client NID directly from Myrinet (GM)
4875 Details : the client NID code from gmnalnid was moved directly into
4876 llmount, removing the need to use this or specifying the
4877 client NID explicitly when mounting GM clients with zeroconf
4880 Frequency : if client is started with down MDS
4882 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4883 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4884 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4885 ptlrpc_import_delayed_req() because we didn't check if the
4886 request was interrupted, and we also didn't break out of the
4887 event loop if there was no timeout
4892 Description: data loss during non-page-aligned writes to a single file from
4893 both multiple nodes and multiple threads on one node at same time
4894 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4895 inconsistency led to false short-reads, that were cached and later
4896 used by ->prepare_write() to fill in partially written page,
4897 leading to data loss.
4900 Frequency : always, if lconf --abort_recovery used
4902 Description: lconf --abort_recovery fails with 'Operation not supported'
4903 Details : lconf was attempting to abort recovery on the MDT device and not
4906 Severity : enhancement
4908 Description: remove cleanup logs
4909 Details : replace lconf-generated cleanup logs with lustre internal
4910 cleanup routines. Eliminates the need for client-cleanup and
4913 Severity : enhancement
4915 Description: add support for EAs (user and system) on lustre filesystems
4916 Details : it is now possible to store extended attributes in the Lustre
4917 client filesystem, and with the user_xattr mount option it
4918 is possible to allow users to store EAs on their files also
4920 Severity : enhancement
4922 Description: Add possibility (config option) to show minimal available OST free
4924 Details : When compiled with --enable-mindf configure option, statfs(2)
4925 (and so, df) will return least minimal free space available from
4926 all OSTs as amount of free space on FS, instead of summary of
4927 free spaces of all OSTs.
4929 Severity : enhancement
4931 Description: do not expand extent locks acquired on OST-side
4932 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4933 by server: they are not cached anyway.
4936 Frequency : when mmap is used/binaries executed from Lustre
4938 Description: Unmmap pages before throwing them away from read cache.
4939 Details : llap_shrink cache now attempts to unmap pages before discarding
4940 them (if unmapping failed - do not discard). SLES9 kernel has
4941 extra checks that trigger if this unmapping is not done first.
4946 Description: lconf didn't resolve symlinks before checking to see whether a
4947 given mountpoint was already in use
4950 Frequency : when migrating failover services
4951 Bugzilla : 6395, 9514
4952 Description: When migrating a subset of services from a node (e.g. failback
4953 from a failover service node) the remaining services would
4954 time out and evict clients.
4955 Details : lconf --force (implied by --failover) sets the global obd_timeout
4956 to 5 seconds in order to quickly disconnect, but this caused
4957 other RPCs to time out too quickly. Do not change the global
4958 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4960 Severity : enhancement
4961 Frequency : if MDS is started with down OST
4962 Bugzilla : 9439,5706
4963 Description: Allow startup/shutdown of an MDS without depending on the
4964 availability of the OSTs.
4965 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4966 Add appropriate locking and lov-osc refcounts for safe
4967 cleaning. Add osc abort_inflight calls in case the
4968 synchronize never started.
4971 Frequency : occasional (Cray XT3 only)
4973 Description: root not authorized to access files in CRAY_PORTALS environment
4974 Details : The client process capabilities were not honoured on the MDS in
4975 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4976 been accessed by an authorized user then root was able to access
4977 the file on the local client also. The root user capabilities
4978 are now allowed on the MDS, as this environment has secure UID.
4981 Frequency : occasional
4983 Description: ldiskfs "too long searching" message happens too often
4984 Details : A debugging message (otherwise harmless) prints too often on
4985 the OST console. This has been reduced to only happen when
4986 there are fragmentation problems on the filesystem.
4991 Description: Division by zero in statfs when all OSCs are inactive
4992 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4993 lov_statfs divides by value returned by lov_get_stripecnt().
4997 Bugzilla : 9489, 3273
4998 Description: First write from each client to each OST was only 4kB in size,
4999 to initialize client writeback cache, which caused sub-optimal
5000 RPCs and poor layout on disk for the first written file.
5001 Details : Clients now request an initial cache grant at (re)connect time
5002 and so that they can start streaming writes to the cache right
5003 away and always do full-sized RPCs if there is enough data.
5004 If the OST is rebooted the client also re-establishes its grant
5005 so that client cached writes will be honoured under the grant.
5010 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5011 Details : Now I/O RPCs go to different portal number and (presumably) fast
5012 lock requests (and glimses) and other RPCs get their own service
5013 threads pool that should be able to service those RPCs
5016 Severity : enhancement
5018 Description: Ability to exchange lustre version between client and servers and
5019 issue warnings at client side if client is too old. Also for
5020 liblustre clients there is ability to refuse connection of too old
5022 Details : New 'version' field is added to connect data structure that is
5023 filled with version info. That info is later checked by server and
5027 Frequency : rare, liblustre only.
5028 Bugzilla : 9296, 9581
5029 Description: Two simultaneous writes from liblustre at offset within same page
5030 might proceed at the same time overwriting eachother with stale
5032 Details : I/O lock withing llu_file_prwv was released too early, before data
5033 actually was hitting the wire. Extended lock-holding time until
5034 server acknowledges receiving data.
5037 Frequency : extremely rare. Never observed in practice.
5039 Description: avoid generating lustre_handle cookie of 0.
5040 Details : class_handle_hash() generates handle cookies by incrementing
5041 global counter, and can hit 0 occasionaly (this is unlikely, but
5042 not impossible, because initial value of cookie counter is
5043 selected randonly). Value of 0 is used as a sentinel meaning
5044 "unassigned handle" --- avoid it. Also coalesce two critical
5045 sections in this function into one.
5047 Severity : enhancement
5049 Description: allow liblustre clients to delegate truncate locking to OST
5050 Details : To avoid overhead of locking, liblustre client instructs OST to
5051 take extent lock in ost_punch() on client's behalf. New connection
5052 flag is added to handle backward compatibility.
5054 Severity : enhancement
5055 Bugzilla : 4928, 7341, 9758
5056 Description: allow number of OST service threads to be specified
5057 Details : a module parameter allows the number of OST service threads
5058 to be specified via "options ost ost_num_threads={N}" in the
5059 OSS's /etc/modules.conf or /etc/modprobe.conf.
5063 Bugzilla : 6146, 9635, 9895
5064 Description: servers crash with bad pointer in target_handle_connect()
5065 Details : In rare cases when a client is reconnecting it was possible that
5066 the connection request was the last reference for that export.
5067 We would temporarily drop the export reference and get a new
5068 one, but this may have been the last reference and the export
5069 was just destroyed. Get new reference before dropping old one.
5071 Severity : enhancement
5072 Frequency : if client is started with failover MDS
5074 Description: Allow multiple MDS hostnames in the mount command
5075 Details : Try to read the configuration from all specified MDS
5076 hostnames during a client mount in case the "primary"
5079 Severity : enhancement
5081 Description: Stop sending data to evicted clients as soon as possible.
5082 Details : Check if the client we are about to send or are sending data to
5083 was evicted already. (Check is done every second of waiting,
5084 for which l_wait_event interface was extended to allow checking
5085 of exit condition at specified intervals).
5088 Frequency : rare, normally only when NFS exporting is done from client
5090 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5092 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5095 Frequency : always, if srand() is called before liblustre initialization
5097 Description: Liblustre uses system PRNG disturbing its usage by user application
5098 Details : Introduce internal to lustre fast and high-quality PRNG for
5099 lustre usage and make liblustre and some other places in generic
5100 lustre code to use it.
5102 Severity : enhancement
5103 Bugzilla : 9477, 9557, 9870
5104 Description: Verify that the MDS configuration logs are updated when xml is
5105 Details : Check if the .xml configuration logs are newer than the config
5106 logs stored on the MDS and report an error if this is the case.
5107 Request --write-conf, or allow starting with --old_conf.
5109 Severity : enhancement
5111 Description: Handle symlinks in the path when checking if Lustre is mounted.
5112 Details : Resolve intermediate symlinks when checking if a client has
5113 mounted a filesystem to avoid duplicate client mounts.
5118 Description: lconf can hit an error exception but still return success.
5119 Details : The lconf command catches the Command error exception at the top
5120 level script context and will exit with the associated exit
5121 status, but doesn't ensure that this exit status is non-zero.
5126 Description: failure of ptlrpc thread startup can cause oops
5127 Details : Starting a ptlrpc service thread can fail if there are a large
5128 number of threads or the server memory is very fragmented.
5129 Handle this without oopsing.
5132 Frequency : always, only if liblustre and non-default acceptor port was used
5134 Description: liblustre cannot connect to servers with non-default acceptor port
5135 Details : tcpnal_set_default_params() was not called and was therefore
5136 ignoring the environment varaible TCPNAL_PORT, as well as other
5137 TCPNAL_ environment variables
5142 Description: two objects could be created on the same OST for a single file
5143 Details : If an OST is down, in some cases it was possible to create two
5144 objects on a single OST for a single file. No problems other
5145 than potential performance impact and spurious error messages.
5149 Bugzilla : 5681, 9562
5150 Description: Client may oops in ll_unhash_aliases
5151 Details : Client dcache may become inconsistent in race condition.
5152 In some cases "getcwd" can fail if the current directory is
5158 Description: Inode refcounting problems in NFS export code
5159 Details : link_raw functions used to call d_instantiate without obtaining
5160 extra inode reference first.
5164 Bugzilla : 9942, 9903
5165 Description: Referencing freed requests leading to crash, memleaks with NFS.
5166 Details : We used to require that call to ll_revalidate_it was always
5167 followed by ll_lookup_it. Also with revalidate_special() it is
5168 possible to call ll_revalidate_it() twice for the same dentry
5169 even if first occurence returned success. This fix changes semantic
5170 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5171 reference on a request referred from the intent.
5172 ll_intent_release() then releases such a request.
5175 Frequency : rare, normally benchmark loads only
5177 Description: unlinked inodes were kept in memory on the client
5178 Details : If a client is repeatedly creating and unlinking files it
5179 can accumulate a lot of stale inodes in the inode slab cache.
5180 If there is no other client load running this can cause the
5181 client node to run out of memory. Instead flush old inodes
5182 from client cache that have the same inode number as a new inode.
5185 Frequency : SLES9 2.6.5 kernel and long filenames only
5186 Bugzilla : 9969, 10379
5187 Description: utime reports stale NFS file handle
5188 Details : SLES9 uses out-of-dentry names in some cases, which confused
5189 the lustre dentry revalidation. Change it to always use the
5193 Frequency : rare, unless heavy write-truncate concurrency is continuous
5194 Bugzilla : 4180, 6984, 7171, 9963, 9331
5195 Description: OST becomes very slow and/or deadlocked during object unlink
5196 Details : filter_destroy() was holding onto the parent directory lock
5197 while truncating+unlinking objects. For very large objects this
5198 may block other threads for a long time and slow overall OST
5199 responsiveness. It may also be possible to get a lock ordering
5200 deadlock in this case, or run out of journal credits because of
5201 the combined truncate+unlink. Solution is to do object truncate
5202 first in one transaction without parent lock, and then do the
5203 final unlink in a new transaction with the parent lock. This
5204 reduces the lock hold time dramatically.
5207 Frequency : rare, 2.4 kernels only
5209 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5210 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5211 before unmounting their filesystem in order to avoid deadlock.
5212 The kernel_locked() function in 2.4 kernels only checks whether
5213 the kernel lock is held, not whether it is this process that is
5214 holding it as 2.6 kernels do.
5219 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5220 Details : The client ptlrpc code may be trying to reconnect to a down
5221 server before a previous connection attempt has timed out.
5222 Increase the reconnect interval to be longer than the connection
5223 timeout interval to avoid sending duplicate connections to
5227 Frequency : echo_client brw_test command
5229 Description: fix echo_client to work with OST preallocated code
5230 Details : OST preallocation code (5137) didn't take echo_client IO path
5231 into account: echo_client calls filter methods outside of any
5232 OST thread and, hence, there is no per-thread preallocated
5233 pages and buffers to use. Solution: hijack pga pages for IO. As
5234 a byproduct, this avoids unnecessary data copying.
5238 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5239 Description: Client can oops in mdc_commit_close() after open replay
5240 Details : It was possible for the MDS to return an open request with no
5241 transaction number in mds_finish_transno() if the client was
5242 evicted, but without actually returning an error. Clients
5243 would later try to replay that open and may trip an assertion
5244 Simplify the client close codepath, and always return an error
5245 from the MDS in case the open is not successful.
5248 Frequency : rare, 2.6 OSTs only
5250 Description: OST may deadlock under high load on fragmented files
5251 Details : If there was a heavy load and highly-fragmented OST filesystems
5252 it was possible to have all the OST threads deadlock waiting on
5253 allocation of biovecs, because the biovecs were not released
5254 until the entire RPC IO was completed. Instead, release biovecs
5255 as soon as they are complete to ensure forward IO progress.
5257 Severity : enhancement
5259 Description: Support for specifying external journal device at mount
5260 Details : If an OST or MDS device is formatted with an external journal
5261 device, this device major/minor is stored in the ext3 superblock
5262 and may not be valid for failover. Allow detecting and
5263 specifying the external journal at mount time.
5268 Description: Mounting an MDS with pending unlinked files may cause oops
5269 Details : target_finish_recovery() calls mds_postrecov() which returned
5270 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5271 considers this an error and immediately begins cleaning up the
5272 lov, just after starting the mds_lov process
5274 Severity : enhancement
5276 Description: Implement 'lfs df' to report actual free space on per-OST basis
5277 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5278 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5279 usage of objects; '-h' to report in human readable format.
5281 ------------------------------------------------------------------------------
5283 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5290 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5291 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5292 at startup during orphan unlinking. Instead, skip buddy update
5293 before setup, as it will be regenerated after recovery is complete.
5298 Description: If an OST is inactive, its locks might reference stale inodes.
5299 Details : lov_change_cbdata() must iterate over all namespaces, even if
5300 they are inactive to clear inode references from the lock.
5302 Severity : enhancement
5303 Frequency : occasional, if non-standard max_dirty_mb used
5305 Description: Client will block write RPCs if not enough grant
5306 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5307 then the client will block writes while waiting for another RPC
5308 to complete instead of consuming its dirty limit. With change
5309 we get improved performance when max_dirty_mb is small.
5311 Severity : enhancement
5312 Bugzilla : 3389, 6253
5313 Description: Add support for supplementary groups on the MDS.
5314 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5315 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5316 lookups for user supplementary groups into a cache.
5320 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5321 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5322 for liblustre O_EXCL is enforced by MDS.
5326 Description: readdir never returns NULL in liblustre.
5327 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5328 dentry in struct dirent.
5332 Frequency : liblustre only, depends on application IO pattern
5333 Description: liblustre clients evicted if not contacting servers
5334 Details : Don't put liblustre clients into the ping_evictor list, so
5335 they will not be evicted by the pinger ever.
5337 Severity : enhancement
5339 Description: Add ability to evict clients by NID from MDS.
5340 Details : By echoing "nid:$NID" string into
5341 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5342 $NID would be instantly evicted from this MDS and from all active
5343 OSTs connected to it.
5347 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5348 Details : lookup_it_finish() used to query file size from OSTs that was not
5353 Description: service threads change working directory to that of init
5354 Details : Starting lustre service threads may pin the working directory
5355 of the parent thread, making that filesystem busy. Threads
5356 now change to the working directory of init to avoid this.
5360 Frequency : during shutdown only
5361 Description: shutdown with a failed MDS or OST can cause unmount to hang
5362 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5367 Frequency : 2.6 only
5368 Description: chmod/chown may include an extra supplementary group
5369 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5370 supplementary group and if none is specified this is used.
5373 Bugzilla : 5479 (6816)
5375 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5376 Details : If lookup is in progress on a file that is unlinked we might try
5377 to revalidate the inode and fail in revalidate after lookup is
5378 complete and ll_file_open() enqueues the open again but
5379 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5382 Frequency : always, if lconf --abort_recovery used
5384 Description: lconf --abort_recovery fails with 'Operation not supported'
5385 Details : lconf was attempting to abort recovery on the MDT device and not
5388 ------------------------------------------------------------------------------
5390 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5395 Frequency : rare (only unsupported configurations with a node running as an
5397 Bugzilla : 6514, 5137
5398 Description: Mounting a Lustre file system on a node running as an OST could
5400 Details : OSTs now preallocates memory needed to write out data at
5401 startup, instead of when needed, to avoid having to
5402 allocate memory in possibly low memory situations.
5403 Specifically, if the file system is mounted on on OST,
5404 memory pressure could force it to try to write out data,
5405 which it needed to allocate memory to do. Due to the low
5406 memory, it would be unable to do so and the node would
5407 become unresponsive.
5409 Severity : enhancement
5411 Description: Addition of lconf --service command line option
5412 Details : lconf now accepts a '--service <arg>' option, which is
5413 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5415 Severity : enhancement
5417 Description: Failover mode is now the default for OSTs.
5418 Details : By default, OSTs will now run in failover mode. To return to
5419 the old behaviour, add '--failout' to the lmc line for OSTs.
5421 Severity : enhancement
5423 Description: Health checks are now provided for MDS and OSTs
5424 Details : Additional detailed health check information on MSD and OSTs
5425 is now provided through the procfs health_check value.
5428 Frequency : occasional, depends on IO load
5430 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5431 after numerous create/delete cycles
5432 Details : The ext3 inode allocation policy would not allocate new inodes
5433 very well on the OSTs because there are no new directories
5434 being created. Instead we look for groups with free space if
5435 the parent directories are nearly full.
5440 Description: Network or server problems during mount may cause partially
5441 mounted clients instead of returning an error.
5442 Details : The config llog parsing code may overwrite the error return
5443 code during mount error handling, returning success instead
5449 Description: MDS can fail to allocate large reply buffers
5450 Details : After long uptimes the MDS can fail to allocate large reply
5451 buffers (e.g. zconf client mount config records) due to memory
5452 fragmentation or consumption by the buffer cache. Preallocate
5453 some large reply buffers so that these replies can be sent even
5454 under memory pressure.
5458 Frequency : rare (liblustre)
5459 Description: fsx running with liblustre complained that using truncate() to
5460 extend the file doesn't work. This patch corrects that issue.
5461 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5462 ATTR_SIZE and lsm use in llu_setattr_raw.
5466 Frequency : rare, only 2.6 kernels
5467 Description: Unusual file access patterns on the MDS may result in inode
5468 data being lost in very rare circumstances.
5469 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5470 code in the 2.6 kernel caused the inode and/or EA data not to
5471 be read from disk, causing single-file corruption.
5475 Frequency : rare, only 2.6 filesystems using extents
5476 Description: Heavy concurrent write and delete load may cause data corruption.
5477 Details : It was possible under high-load situations to have an extent
5478 metadata block in the block device cache from a just-unlinked
5479 file overwrite a newly-allocated data block. We now unmap any
5480 metadata buffers that alias just-allocated data blocks.
5484 Frequency : filesystems with default stripe_count larger than 77
5485 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5486 Details : lconf specifies an inode size of 4096 bytes when the default
5487 stripe_count is larger than 77. This conflicts with the default
5488 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5489 this case to avoid pinning too much memory for large EAs.
5491 ------------------------------------------------------------------------------
5493 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5498 Frequency : rare (extremely heavy IO load with hundreds of clients)
5500 Description: Client is evicted, gets IO error writing to file
5501 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5502 caused clients to be evicted for AST timeouts. The fixes in
5503 bug 5192 mean we no longer need to have such short AST timeouts
5504 so ldlm_timeout has been increased.
5507 Frequency : occasional during --force or --failover shutdown under load
5508 Bugzilla : 5949, 4834
5509 Description: Server oops/LBUG if stopped with --force or --failover under load
5510 Details : a collection of import/export refcount and cleanup ordering
5511 issues fixed for safer force cleanup
5514 Frequency : only filesystems larger than 120 OSTs
5515 Bugzilla : 5990, 6223
5516 Description: lfs getstripe would oops on a very large filesystem
5517 Details : lov_getconfig used kfree on vmalloc'd memory
5520 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5521 Bugzilla : 6242, 6243
5522 Description: reading from files that had been truncated to a non-zero size
5523 but never opened returned no data
5524 Details : ll_file_read() reads zeros from no-object files to EOF
5529 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5530 Details : There's little harm in aborting MDS/OSS recovery and letting it
5531 try again, so I removed the LASSERT and return an error instead.
5533 Severity : enhancement
5535 Description: New debugging infrastructure for tracking down data corruption
5536 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5537 (b) cover more of the client-side code path, and (c) try to narrow
5538 down where problems occurred
5542 Bugzilla : 3819, 4364, 4397, 6313
5543 Description: Racing close and eviction MDS could cause assertion in mds_close
5544 Details : It was possible to get multiple mfd references during close and
5545 client eviction, leading to one thread referencing a freed mfd.
5547 Severity: : enhancement
5548 Bugzilla : 3262, 6359
5549 Description: Attempts to reconnect to servers are now more aggressive.
5550 Details : This builds on the enhanced upcall-less recovery that was added
5551 in 1.4.2. When trying to reconnect to servers, clients will
5552 now try each server in the failover group every 10 seconds. By
5553 default, clients would previously try one server every 25 seconds.
5558 Description: After recovery, certain operations trigger a failed
5559 assertion on a client.
5560 Details : Failing over an mds, using lconf -d --failover, while a
5561 client was doing a readdir() call would cause the client to
5562 LBUG after recovery completed and the readdir() was resent.
5564 Severity : enhancement
5566 Description: Default groups are now added by lconf
5567 Details : You can now run lconf --group <servicename> without having to
5568 manually add groups with lmc.
5571 Frequency : occasional
5573 Description: Nodes with an elan id of 0 trigger a failed assertion
5576 Frequency : always when accessing e.g. tty/console device nodes
5578 Description: tty and some other devices nodes cannot be used on lustre
5579 Details : file's private_data field is used by device data and lustre
5580 values in there got lost. New field was added to struct file to
5581 store fs-specific private data.
5584 Frequency : when exporting Lustre via NFS
5586 Description: NFSD failed occasionally when looking up a path component
5587 Details : NFSD is looking up ".." which was broken in ext3 directories
5588 that had grown large enough to become hashed.
5591 Frequency : Clusters with multiple interfaces not on the same subnet
5593 Description: Nodes will repeatedly try to reconnect to an interface which it
5594 cannot reach and report an error to the log.
5595 Details : Extra peer list entries will be created by lconf with some peers
5596 unreachable. lconf now validates the peer before adding it.
5599 Frequency : Only if a default stripe is set on the filesystem root.
5601 Description: Setting a default stripe on the filesystem root prevented the
5602 filesystem from being remounted.
5603 Details : The client was sending extra request flags in the root getattr
5604 request and did not allocate a reply buffer for the dir EA.
5607 Frequency : occasional, higher if lots of files are accessed by one client
5608 Bugzilla : 6159, 6097
5609 Description: Client trips assertion regarding lsm mismatch/magic
5610 Details : While revalidating inodes the VFS looks up inodes with ifind()
5611 and in rare cases can find an inode that is being freed.
5612 The ll_test_inode() code will free the lsm during ifind()
5613 when it finds an existing inode and then the VFS later attaches
5614 this free lsm to a new inode.
5618 Bugzilla : 6422, 7030
5619 Description: MDS deadlock between mkdir and client eviction
5620 Details : Creating a new file via mkdir or mknod (starting a transaction
5621 and getting the ns lock) can deadlock with client eviction
5622 (gets ns lock and trying to finish a synchronous transaction).
5625 Frequency : occasional
5626 Description: While starting a server, the fsfilt_ext3 module could not be
5628 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5629 kernels. Previously, lconf would still use the ext3 name
5630 when trying to load modules. Now, it will correctly use
5631 ext3 on 2.4 and ldiskfs on 2.6.
5633 Severity : enhancement
5634 Description: The default stripe count has been changed to 1
5635 Details : The interpretation of the default stripe count (0, to lfs
5636 or lmc) has been changed to mean striping across a single
5637 OST, rather than all available. For general usage we have
5638 found a stripe count of 1 or 2 works best.
5640 Severity : enhancement
5641 Description: Add support for compiling against Cray portals.
5642 Details : Conditional compiling for some areas that are different
5646 Frequency : occasional
5647 Bugzilla : 6409, 6834
5648 Description: Creating files with an explicit stripe count may lead to
5649 a failed assertion on the MDS
5650 Details : If some OSTs are full or unavailable, creating files may
5651 trigger a failed assertion on the MDS. Now, Lustre will
5652 try to use other servers or return an error to the
5656 Frequency : occasional
5658 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5659 caused lock timeout during readahead (since 1.4.2).
5660 Details : Processes doing readahead might match a lock that hasn't been
5661 granted yet if there are overlapping and conflicting lock
5662 requests. The readahead process waits on ungranted lock
5663 (original lock is CBPENDING), while OST waits for that process
5664 to cancel CBPENDING read lock and eventually evicts client.
5666 Severity : enhancement
5668 Description: Initial enabling of flock support for clients
5669 Details : Implements fcntl advisory locking and file status functions.
5670 This feature is provided as an optional mount flag (default
5671 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5672 locking are implemented yet, and those that are are not guaranteed
5673 to be completely correct in production environments.
5674 mount -t lustre -o [flock|noflock] ...
5677 Frequency : occasional
5679 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5680 assertion in the ext3 JBD (journaling) layer.
5681 Details : The b_committed_data struct is protected by the big kernel lock
5682 in 2.4 kernels, serializing journal_commit_transaction() and
5683 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5684 access to this struct. In 2.6 kernels there is finer grained
5685 locking to improve SMP performance of the JBD layer.
5689 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5694 Description: merge and cleanup kernel patches.
5697 -----------------------------------------------------------------------------
5699 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5701 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5702 versions of Lustre. Please update all servers and clients to
5703 version 1.4.2 or later at the same time. You must also run
5704 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5705 to update the configuration logs.
5707 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5708 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5709 - swab llog records of type '0' so we get proper header size/idx (5861)
5710 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5711 - fix rename of one directory over another leaking an inode (5953)
5712 - avoid SetPageDirty on 2.6 (5981)
5713 - don't re-add just-being-destroyed locks to the waiting list (5653)
5714 - when creating new directories, inherit the parent's custom
5715 striping settings if present parent (3048)
5716 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5717 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5718 - don't hold client locks on temporary worklist from l_lru (5666)
5719 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5720 - automatically evict dead clients (5921)
5721 - Update file size properly in create+truncate+fstat case (6196)
5722 - Do not unhash mountpoint dentries, do not allow removal of
5724 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5725 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5726 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5727 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5728 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5729 - Correctly handle reads of files with no objects (6243)
5730 - lctl recover will also mark a device active if deactivate used (5933)
5732 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5733 - allow --write-conf on an MDS with different nettype than client (5619)
5734 - don't write config llogs to MDS for mounts not from that MDS (5617)
5735 - lconf should create multiple TCP connections from a client (5201)
5736 - init scripts are now turned off by default; run chkconfig --on
5737 lustre and chkconfig --on lustrefs to use them
5738 - upcalls are no longer needed for clients to recover to failover
5740 - add --abort-recovery option to lconf to abort recovery on device
5742 - add support for an arbitrary number of OSTs (3026)
5743 - Quota support protocol changes.
5744 - forward compatibility changes to wire structs (6007)
5745 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5746 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5747 - improved "lustre status" script
5748 - initialize blocksize for non-regular files (6062)
5749 - added --disable-server and --disable-client configure options (5782)
5750 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5751 - Vanilla 2.4.29 support
5752 - increase maximum number of obd devices to 520 (6242)
5753 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5754 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5755 SLES 9 2.6.5 kernels
5756 - limit stripes per file to 160 (the maximum EA size) (6093)
5758 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5761 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5762 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5763 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5764 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5765 - small ext3 extents cleanups and fixes (5733)
5766 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5767 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5768 - avoid needless client->OST connect, fix handle mismatch (5317)
5769 - fix DLM error path that led to out-of-sync client, long delays (5779)
5770 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5771 - fix several locking issues related to i_size (5492,5624,5654,5672)
5772 - don't move pending lock onto export if it is already evicted (5683)
5773 - fix kernel oops when creating .foo in unlinked directory (5548)
5774 - fix deadlock in obdfilter statistics vs. object create (5811)
5775 - use time_{before,after} to avoid timer jiffies wrap (5882)
5776 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5777 - Do not leak request if server was not able to process it (5154)
5778 - If mds_open unable to find parent dir, make that negative lookup(5154)
5779 - don't create new directories with extent-mapping (5909, 5936)
5781 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5782 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5783 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5784 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5785 - support for RHEL4 kernel on i686 (5773)
5786 - provide error messages when incompatible logs are encountered (5898)
5788 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5789 * version 1.4.0.10 (1.4.1 release candidate 1)
5791 - don't keep a lock reference when lock is not granted (4238)
5792 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5793 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5794 - drop import inflight refcount on signal_completed_replay error (5255)
5795 - unlock page after async write error during send (3677)
5796 - handle missing objects in filter_preprw_read properly (5265)
5797 - no transno return for symlink open, don't save no-trasno open (3440)
5798 - don't try to complete elan receive that already failed (4012)
5799 - free RPC server reply state on error (5406)
5800 - clean up thread from ptlrpc_start_thread() on error (5160)
5801 - readahead could read extra page into cache that wasn't ejected (5388)
5802 - prevent races in class_attach/setup/cleanup/detach (5260)
5803 - don't dereference de->d_inode after l_dput of de (5458)
5804 - use "int" for stripe value returned from lock_to_stripe (5544)
5805 - mballoc allocation and error-checking fixes in 2.6 (5504)
5806 - block device patches to fix I/O request sizes in 2.6 (5482)
5807 - look up hostnames for IB nals (5602)
5808 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5809 - don't start multiple acceptors for the same port (5277)
5810 - fix incorrect LASSERT in mds_getattr_name (5635)
5811 - export a proc file for general "ping" checking (5628)
5812 - fix "lfs check" to not block when the MDS is down (5628)
5814 - service request history (4965)
5815 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5816 - create an "evict_client" /proc entry on OSTs, like the MDS has
5817 - fix mount usage message, return errors per mount(8) (5168)
5818 - change grep [] to grep "[]" in tests so they work in more UMLs
5819 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5820 - remove ancient LOV_MAGIC_V0 EA support (5047)
5821 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5822 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5823 - allow readahead allocations to fail when low on memory (5383)
5824 - mmap locking landed again, after considerable improvement (2828)
5825 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5827 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5830 - send OST transaction number in read/write reply to free req (4966)
5831 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5832 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5833 - the watchdog thread now runs as interruptible (5246)
5834 - flock/lockf fixes (but it's still disabled, pending 5135)
5835 - don't use EXT3 constants in llite code (5094)
5836 - memory shortage at startup could cause assertion (5176)
5838 - reorganization of lov code
5839 - single portals codebase
5841 - add extents/mballoc support (5025)
5842 - direct I/O reads in the obdfilter (4048)
5843 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5845 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5848 - send OST transaction number in read/write reply to free req (4966)
5849 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5850 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5851 - the watchdog thread now runs as interruptible (5246)
5852 - handle missing objects in filter_preprw_read properly (5265)
5853 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5854 - drop import inflight refcount on signal_completed_replay error (5255)
5855 - unlock page after async write error during send (3677)
5856 - return original error code on reconstructed replies (3761)
5857 - no transno return for symlink open, don't save no-trasno open (3440)
5859 - add pid to ldlm debugging output (4922)
5860 - bump the watchdog timeouts -- we can't handle 30sec yet
5861 - extra debugging for orphan dentry/inode bug (5259)
5863 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5866 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5867 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5868 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5869 - fix NULL dereference in /proc/sys/portals/routes (4827)
5870 - allow failed mdc_close() operations to be interrupted (4561)
5871 - stop precreate on OST before MDS would time out on it (4778)
5872 - don't send partial-page writes before EOF from client (4410)
5873 - discard client grant for sub-page writes on large-page clients (4520)
5874 - don't free dentries not owned by NFS code, check generation (4806)
5875 - fix lsm leak if mds_create_objects() fails (4801)
5876 - limit debug_daemon file size, always print CERROR messages (4789)
5877 - use transno after validating reply (3892)
5878 - process timed out requests if import state changes (3754)
5879 - update mtime on OST during writes, return in glimpse (4829)
5880 - add mkfsoptions to LDAP (4679)
5881 - use ->max_readahead method instead of zapping global ra (5039)
5882 - don't interrupt __l_wait_event() during strace
5884 - add software watchdogs to catch hung threads quickly (4941)
5885 - make lustrefs init script start after nfs is mounted
5886 - fix CWARN/ERROR duplication (4930)
5887 - return async write errors to application if possible (2248)
5888 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5889 - print NAL number in %x format (4645)
5890 - update barely-supported suse-2.4.21-171 series (4842)
5891 - support for sles 9 %post scripts
5892 - support for building 2.6 kernel-source packages
5893 - support for sles km_* packages
5895 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5898 - ignore -ENOENT errors in osc_destroy (3639)
5899 - notify osc create thread that OSC is being cleaned up (4600)
5900 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5901 - reconstruct ost_handle() like mds_handle() (4657)
5902 - create a new thread to do import eviction to avoid deadlock (3969)
5903 - let lconf resolve symlinked-to devices (4629)
5904 - don't unlink "objects" from directory with default EA (4554)
5905 - hold socknal file ref over connect in case target is down (4394)
5906 - allow more than 32000 subdirectories in a single directory (3244)
5907 - fix blocks count for O_DIRECT writes (3751)
5908 - OST returns ENOSPC from object create when no space left (4539)
5909 - don't send truncate RPC if file size isn't changing (4410)
5910 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5911 - bind to privileged port in socknal and tcpnal (3689)
5913 - rate limit CERROR/CWARN console message to avoid overload (4519)
5914 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5915 - basic mmap support (3918)
5916 - kernel patch series update from b1_4 (4711)
5918 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5921 - avoid crash during MDS cleanup with OST shut down (2775)
5922 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5923 - don't use bad inodes on the MDS (3744)
5924 - dynamic object preallocation to improve recovery speed (4236)
5925 - don't hold spinlock over lock dumping or change debug flags (4401)
5926 - don't zero obd_dev when it is force cleaned (3651)
5927 - print grants to console if they go negative (4431)
5928 - "lctl deactivate" will stop automatic recovery attempts (3406)
5929 - look for existing locks in ldlm_handle_enqueue() (3764)
5930 - don't resolve lock handle twice in recovery avoiding race (4401)
5931 - revalidate should check working dir is a directory (4134)
5933 - don't always mark "slow" obdfilter messages as errors (4418)
5935 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5938 - don't close LustreDB during write_conf until it is done (3860)
5939 - fix typo in lconf for_each_profile (3821)
5940 - allow dumping logs from multiple threads at one time (3820)
5941 - don't allow multiple threads in OSC recovery (3812)
5942 - fix debug_size parameters (3864)
5943 - fix mds_postrecov to initialize import for llog ctxt (3121)
5944 - replace config semaphore with spinlock (3306)
5945 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5946 - don't allow enqueue to complete on a destroyed export (3822)
5947 - down write_lock before checking llog header bitmap (3825)
5948 - recover from lock replay timeout (3764)
5949 - up llog sem before sending rpc (3652)
5950 - reduce ns lock hold times when setting kms (3267)
5951 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5952 - fix NULL deref and obd_dev leak on setup error (3312)
5953 - replace some LBUG about llog ops with error handling (3841)
5954 - don't match INVALID dentries from d_lookup and spin (3784)
5955 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5956 - fix invalid assertion in ptlrpc_set_wait (3880)
5958 - add libwrap support for the TCP acceptor (3996)
5959 - add /proc/sys/portals/routes for non-root route listing (3994)
5960 - allow setting MDS UUID in .xml (2580)
5961 - print the stack of a process that LBUGs (4228)
5963 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5966 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5967 - make sure to unset replay flag from failed open requests (3440)
5968 - if default stripe count is 0, use OST count for inode size (3636)
5969 - update parent mtime/ctime on client for create/unlink (2611)
5970 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5971 - free recovery state on server during a forced cleanup (3571)
5972 - unregister_reply for resent reqs (3063)
5973 - loop back devices mounting and status check on 2.6 (3563)
5974 - fix resource-creation race that can provoke i_size == 0 (3513)
5975 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5976 - more debugging for page-accounting assertion (3746)
5977 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5978 - avoid deadlock after precreation failure (3758)
5979 - fix race and lock order deadlock in orphan handling (3450, 3750)
5980 - add validity checks when grabbing inodes from l_ast_data (3599)
5982 - add /proc/.../recovery_status to obdfilter (3428)
5983 - lightweight CDEBUG infrastructure, debug daemon (3668)
5984 - change default OSC RPC parameters to be better on small clusters
5985 - turn off OST read cache for files smaller than 32MB
5986 - install man pages and include them in rpms (3100)
5987 - add new init script for (un)mounting lustre filesystems (2593)
5988 - run chkconfig in %post for init scripts (3701)
5989 - drop scimac NAL (unmaintained)
5991 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5994 - clean kiobufs before and after use (3485)
5995 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5996 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5997 - change init script to fail more clearly if not run as root (1528)
5998 - allow clients to reconnect during replay (1742)
5999 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6000 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6001 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6002 - fix cond_resched (3554)
6003 - only evict unfinished clients after recovery (3515)
6004 - allow bulk resend, prevent data loss (3570)
6005 - dynamic ptlrpc request buffer allocation (2102)
6006 - don't allow unlinking open directory if it isn't empty (2904)
6007 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6008 - remove extraneous obd dereference causing LASSERT failure (3334)
6009 - don't use get_cycles() when creating temp. files on the mds (3156)
6010 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6011 - handle EEXIST for set-stripe, set proper directory name (3336)
6013 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6014 - fix ksocknal_fmb_callback() error messages (2918)
6016 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6019 - don't copy lvb into (possibly NULL) reply on error (2983)
6020 - don't deref dentry after dput, don't free lvb on error (2922)
6021 - use the kms to determine writeback rpc length (2947)
6022 - increment oti_logcookies when osc is inactive (2948)
6023 - update client's i_blocks count via lvb messages (2543)
6024 - handle intent open/close of special files properly (1557)
6025 - mount MDS with errors=remount-ro, like obdfilter (2009)
6026 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6027 - don't use cancelling-locks' kms values (2947)
6028 - use highest lock extent for kms, not last one (2925)
6029 - don't dereference ERR_PTR() dentry in error handling path (3107)
6030 - fix thread race in portals_debug_dumplog() (3122)
6031 - create lprocfs device entries at setup instead of at attach (1519)
6032 - common AST error handler, don't evict client on completion race (3145)
6033 - zero nameidata in detach_mnt in 2.6 (3118)
6034 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6035 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6036 - initialize RPC timeout timer earlier for 2.6 (3219)
6037 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6038 - print nal/nid for unknown nid (3258)
6039 - additional checks for oscc recovery before doing precreate (3284)
6040 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6041 - don't crash in mdc_close for bad permissions on open (3285)
6042 - zero i_rdev for non-device files (3147)
6043 - clear page->private before handing to FS, better assertion (3119)
6044 - tune the read pipeline (3236)
6045 - fix incorrect decref of invalidated dentry (2350)
6046 - provide read-ahead stats and refine rpc in flight stats (3328)
6047 - don't hold journal transaction open across create RPC (3313)
6048 - update atime on MDS at close time (3265)
6049 - close LDAP connection when recovering to avoid server load (3315)
6050 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6051 - don't leak open file on MDS after open resend (3325)
6052 - serialize filter_precreate and filter_destroy_precreated (3329)
6053 - loop device shouldn't call sync_dev() for nul device (3092)
6054 - clear page cache after eviction (2766)
6055 - resynchronize MDS->OST in background (2824)
6056 - refuse to mount the same filesystem twice on same mountpoint (3394)
6057 - allow llmount to create routes for mounting behind routers (3320)
6058 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6059 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6060 - fix rare problem with rename on htree directories (3417)
6062 - allow default OST striping configuration per directory (1414)
6063 - fix compilation for qswnal for 2.6 kernels (3125)
6064 - increase maximum number of MDS request buffers for large systems
6065 - change liblustreapi to be useful for external progs like lfsck (3098)
6066 - increase local configuration timeout for slow disks (3353)
6067 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6069 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6072 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6073 - don't overwrite extent policy data in reply if lock was blocked (2901)
6074 - drop filter export grants atomically with removal from device (2663)
6075 - del obd_self_export from work_list in class_disconnect_exports (2908)
6076 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6077 - swab reply message in mdc_close, other PPC fixes (2464)
6078 - fix destroying of named logs (2325)
6079 - overwrite old logs when running lconf --write_conf (2264)
6080 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6081 - fix race in target_handle_connect (2898)
6082 - mds_reint_create() should take same inode create lock (2926)
6083 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6084 - don't close files for self_export to avoid uninitialized obd (2936)
6085 - allow MDS with the same name as client node (2939)
6086 - hold dentry reference for closed log files for unlink (2325)
6087 - reserve space for all logs during transactions (2059)
6088 - don't evict page beyond end of stripe extent (2925)
6089 - don't oops on a deleted current working directory (2399)
6090 - handle hard links to targets without a parent properly (2517)
6091 - don't dereference NULL lock when racing during eviction (2867)
6092 - don't grow lock extents when lots of conflicting locks (2919)
6094 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6097 - account for cache space usage on clients to avoid data loss (974)
6098 - lfsck support in lustre kernel code (2349)
6099 - reduce journal credits needed for BRW writes (2370)
6100 - orphan handling to avoid losing space on client/server crashes
6101 - ptlrpcd can be blocked, stopping ALL progress (2477)
6102 - use lock value blocks to assist in proper KMS, faster stat (1021)
6103 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6104 - recovery for initial connections (2355)
6105 - fixes for mds_cleanup_orphans (1934)
6106 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6107 - block all file creations until orphan recovery completes (1901)
6108 - client remove rq_connection from request struct (2423)
6109 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6110 - recovery timer race (2670)
6111 - mdc_close recovey bug (2532)
6112 - ptlrpc cleanup bug (2710)
6113 - mds timeout on local locks (2588)
6114 - namespace lock held during RPCs (2431)
6115 - handle interrupted sync write properly (2503)
6116 - don't try to handle a message that hasn't been replied to (2699)
6117 - client assert failure during cleanup after abort recovery (2701)
6118 - leak mdc device after failed mount (2712)
6119 - ptlrpc_check_set allows timedout requests to complete (2714)
6120 - wait for inflight reqs when ptlrpcd finishes (2710)
6121 - make sure unregistered services are removed from the srv_list
6122 - reset bulk XID's when resending them (caught by 1138 test)
6123 - unregister_bulk after timeout
6124 - fix lconf error (2694)
6125 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6126 - readahead locks pages, leaves pending causing memory pressure (2673)
6127 - increase OST request buffers to 4096 on large machines (2729)
6128 - fix up permission of existing directories in simple_mkdir (2661)
6129 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6130 - don't assign transno to errored transactions (2742)
6131 - don't delete objects on OST if given a bogus objid from MDS (2751)
6132 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6133 - if rq_no_resend, then timeout request after recovery (2432)
6134 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6135 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6136 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6137 - MDS recovery shouldn't delete objects that it has given out (2730)
6138 - if enqueue arrives after completion, don't clobber LVB (2819)
6139 - don't unlock pages twice when trigger_group_io returns error (2814)
6140 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6141 - don't write pages to disk if there was an error (1450)
6142 - don't ping imports that have recovery disabled (2676)
6143 - take buffered bytes into account when balancing socknal conn (2817)
6144 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6145 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6146 - remove little-endian swabbing of llog records (1987)
6147 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6148 - retry reposting request buffers if they fail (1191)
6149 - grow extent at grant time to avoid granting a revoked lock (2809)
6150 - lock revoke doesn't evict page if covered by a second lock (2765)
6151 - disable VM readahead to avoid reading outside lock extents (2805)
6153 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6154 - updated kernel patches for hp-2.4.20 kernel (2681)
6156 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6159 - fix truncated write corruption (2366)
6160 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6162 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6163 - ensure len(uuid) < 37 in lmc (1171)
6164 - fix ia64 OOPS in llog_test (2255)
6165 - zero end of page at obdfilter for partial page writes (2648)
6166 - don't leave stale dentries around after renames (bug 2428)
6167 - fix timeouts when evicting a client with a single lock held (2642)
6168 - set deadline for the initial HELLO message to drain (2634)
6169 - print out dotted-quad IP addresses in the socknal (2302)
6171 - additional debugging for MDS client eviction problem (2443)
6172 - fix mkfsoptions support for osts (2603, 2604)
6174 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6177 - add series for the vanilla 2.6.0 kernel
6178 - add series for the vanilla 2.4.24 kernel
6179 - add series for a cray x86/64 UL kernel drop
6180 - fix xattr patches for the vanilla 2.4.19 series
6182 - generate true UUIDs in lmc (1171)
6183 - have portals stack dumping break in UML (2466)
6184 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6185 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6186 - wake callback waiting threads on client eviction (2460)
6187 - Add --ptldebug and --subsystem to lmc (1719)
6188 - update assertion to allow safe interrupt allocation
6189 - set rq_no_resend for cancel requests (2432)
6190 - recalculate ptlrpcd timeout after resend (2494)
6191 - call vfs_rmdir when removing pending directories (2368)
6192 - fix renaming a file to itself (2429)
6193 - lmc creates a default one-stripe lov (2454)
6194 - expand procfs space to handle large clusters (2326)
6195 - increase UML stack to avoid overflow
6196 - update lconf's list of debug and subsystem masks
6197 - fix lfs find --obd (2510)
6198 - /proc tunable for disabling filter read caching (2591)
6199 - stop rpm packages from altering slapd.conf (2301)
6200 - disable nagle in the socknal under 0conf (2578)
6201 - choose mds inode size based on stripe count (2572)
6202 - fix kernel-source rpm problems (2516)
6204 - add --disable-doc to avoid pdf generation (2421)
6205 - update documentation, tests, type-os, comments
6206 - avoid format warnings on ia64
6207 - remove the TOE NAL
6208 - tiny code cleanups by removing unused fields
6210 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6213 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6214 - remove the most blatant lies from BUILDING file (2371)
6215 - change default debug level to reasonable production setting
6216 - reduce client side cache size to reduce cache flush time
6217 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6218 - make TCP zerocopy and pinger support enabled by default (2476)
6219 - sync writes completed after process exits caused crashes (2319)
6220 - maintain correct mount count on the MDS (2356)
6221 - backout 1557, because 2316 wasn't really fixed
6222 - better file I/O statistics gathering in /proc
6223 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6224 - another kernel patch to fix zero-copy TCP function export
6225 - don't take duplicate lock when processing re-sent getattr (2420)
6226 - lctl uses obd_self_export instead of creating new conn (2353)
6227 - MDS/OST recovery case which requires object creation asserted (2425)
6228 - move lfs from /usr/sbin to /usr/bin in packages
6229 - fix race between mds_client_add and mds_client_free (2417)
6230 - use kmalloc instead of slabs in portals (2430)
6231 - don't create duplicate records when a failover MDS is present (2442)
6232 - remove unnecessary mount age check (2332)
6233 - don't remove directory inodes from locks prematurely (2451)
6234 - don't break if MDS service name is the same as hostname (2103)
6235 - fix races in client write RPC generation when cache full (2482)
6237 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6240 - remove now-unused request->rq_obd (278)
6241 - if an allocation fails, print out how much memory we've used (1933)
6242 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6243 - add the "configurable stack size" patch to most series files (1256)
6244 - ability to write large log records, for 100+ OST configs (2306)
6245 - fix NULL deref when filter_prep fails (2314)
6246 - fix operator precedence error in filter_sync
6247 - dynamic allocation of socknal TX descriptors (2315)
6248 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6249 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6250 - ensure that CWARN messages in Portals always get to the syslog
6251 - __init/__exit are not for prototype decls (ldlm_init/exit)
6252 - x86-64 compile warning fixes
6253 - fix gateway LMC keyword conflict (2318)
6254 - fix MDS lock inversions in getattr/reint paths (1844)
6255 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6256 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6257 - DLM race condition prevented some lock evictions (2328)
6258 - ENOMEM detection and retry on socknal sends (2230)
6259 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6260 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6261 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6262 - better error messages when a client is rejected during recovery (1505)
6263 - avoid cancelling locks which were never granted, after failure (2330)
6264 - fix i_sem/journal inversion in mds_client_add (2333)
6265 - fix truncate/getattr lock cycle deadlock (2334)
6266 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6267 - fix two rare exit paths which could leak an l_lock() ref (2321)
6268 - fencepost error in MDS/OST orphan recovery (2226)
6269 - make log record alignment 8 bytes (1988)
6270 - lstripe now fails when requested offset > ost_count (2237)
6271 - ensure that all kernel series have a complete list.h (1607)
6272 - fix crashes in special-file operations (2316)
6273 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6275 - allow configurable automake binary, for testing new versions
6276 - small update to the lfs documentation
6278 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6280 * fix negative export reference count in fsfilt_sync (2312)
6282 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6283 * release candidate 0.9.1
6285 - orphans are moved into the PENDING directory for possible recovery
6286 - replayed opens now open by fid for orphan/rename safety (1042)
6287 - last close of an orphan inode generates a transno (683)
6288 - chdir() and mount() now pin the directory entry (1020)
6289 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6290 - discard very old requests without processing them (1502)
6291 - remove some common, well-understood CERRORs (1505)
6292 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6293 - clear "grant" flags in OST replies until OST grant code lands (1644)
6294 - fix read performance by not clobbering i_blksize on client (1598)
6295 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6296 - make LRU size a /proc tunable, clears locks when reduced (707)
6297 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6298 - server locks take references on exports now (1558)
6299 - build fixes for 2.4.20-rh trees (1663)
6300 - return an error from lov_create if all OSCs are inactive (1751)
6301 - fix import levels when a reconnect happens without a timeout (1597)
6302 - exit early from mds_open if we get a lookup error (1749)
6303 - partial page read at EOF wouldn't wait for disk before sending (1642)
6304 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6305 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6306 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6307 - write/truncate lock inversion (1639)
6308 - Don't auto-load obdclass, portals modules during cleanup (1495)
6309 - fix timestamps from jumping to "now" (1763)
6310 - extra journal assertions (1648)
6311 - add an extra multiunlink test (1771)
6312 - fix read_record/write_record API (1776)
6313 - fix leak of offset_extent, possible incorrect i_size later (1772)
6314 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6315 - Debugging for the kqswnal_get_idle_tx problems (1820)
6316 - Allow recovery to be attempted multiple times (1536)
6317 - Write out MDS last_rcvd file after it is first created (1600)
6318 - Fix tx_descriptor leak in failed transmit situations (1827)
6319 - ext3 journaling fixes for assertion failure after IO error (1871)
6320 - class_export_put() on freed export after completion AST error (1896)
6321 - Fix revalidate looping in VFS (1322)
6322 - Don't access a freed export during MDS_REINT timeout (1521)
6323 - Add open-unlink recovery support on the MDS (1673,1764)
6324 - Return an error if no MDS data was read from last_rcvd (1946)
6325 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6326 - Don't setup a disk device that doesn't match exported UUID (317)
6327 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6328 - avoid committing NULL handle in force close
6329 - local.sh is now a one-stripe LOV configuration
6330 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6331 - fix ext3 htree duplicate directory entry corruption (1516)
6332 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6333 - update to new LOV EA format (2097)
6334 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6336 2003-06-15 Phil Schwan <phil@clusterfs.com>
6339 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6340 - per-import/export recovery handling (958, 931, 959)
6341 - multiple last-rcvd slots, for serving multiple FSes (949)
6342 - connections are again shared between multiple imp/exports (963, 964)
6343 - "umount -f" would hang if any requests needed to be sent (393, 978)
6344 - avoid pinning large req buffer by copying for queued messages (989)
6345 - add "uuid" to "lctl device" command to help upcalls (991)
6346 - "open" RPCs with transnos would confuse recovery counters (1037)
6347 - do proper endian conversion of all wire messages (288, 340, 891)
6348 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6349 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6350 - fix ptlrpc_connection leak in target_handle_connect (1174)
6351 - fix import refcounting bug in OST and MDS cleanup (1134)
6352 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6353 - fix very unlikely obd_types race condition (501)
6354 - remove osc_open hack for echo_client (1187)
6355 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6356 - a failure in read_inode2 leads to deadlock (1139)
6357 - cancel ack-locks as soon as transaction is committed (1072)
6358 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6359 - make sure to commitrw after any preprw to avoid deadlock (1162)
6360 - failing to execute a file in a lustre FS would lock inode (1203)
6361 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6362 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6363 - obd_uuid2tgt would walk off the end of the list (1255)
6364 - on IA64 the osc would give portals incorrect bulk size (1258)
6365 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6366 - fix lock inversion caused by new llite matching code (1282)
6367 - limit the number of dirty pages on a client to 10MB (1286)
6368 - timed out locks were not being corrected cancelled (1289)
6369 - fix O_DIRECT above 4GB on IA-32 (1292)
6370 * major user-visible changes
6371 - fail out/fail over policy now controlled by the upcall (993)
6373 - add OBD_PING to check server availability and failure (954)
6374 - lustre messages are now sent in sending host order (288, 340, 891)
6375 - add eadatalen to MDS getattr reply (340)
6376 - OST read replies may contain second buffer, with per-page status (593)
6378 2003-03-11 Phil Schwan <phil@clusterfs.com>
6381 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6382 - failed open()s could cause deadlock; fixed (867, 869)
6383 - stop cancelling OST locks when files are closed (481)
6384 - overlapping XID spaces caused network corruption (851, 853)
6385 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6386 - setattr_raw added, to do single-RPC, server-side setattrs
6387 - lmc/lconf syntax change for OST UUIDs
6388 - fix crashy race condition between ptlrpc_free_req and osc_close
6389 - don't use request in mdc_enqueue if we hit a timeout (889)
6390 - don't set the inode i_size for regular files from the MDS (896)
6391 - handle out of order completion AST (842)
6392 - don't LBUG if a lock request times out after receiving AST (913)
6393 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6394 - if a bad lock AST arrives, send an error instead of dropping entirely
6395 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6396 - fix leak in bulk IO when only partially completed (899, 900, 926)
6397 - fix O_DIRECT for ia64 (55)
6398 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6399 - C-z after timeout could hang a process forever; fixed (977)
6401 - client-side I/O cache (678, 924, 929, 941, 970)
6403 - READPAGE and SETATTRs which don't take server-side locks get
6406 2003-02-11 Phil Schwan <phil@clusterfs.com>
6409 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6410 - Fix fsfilt_extN_readpage() to read a full page of directory
6411 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6412 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6413 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6414 - Most intent operations were removed, in favour of a new RPC mode
6415 that does a single RPC to the server and bypasses most of the VFS
6416 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6417 - Aggressively cancel local locks on DLM servers
6418 - mds_reint_unlink sends EA to the client if it's the last nlink.
6419 client uses that EA to unlink OST objects.
6420 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6421 - recursive symlinks were fixed (439)
6422 - fixed NULL deref in DEBUG_REQ
6423 - filter_update_lastobjid no longer calls sync, which annoyed extN
6424 - fixed multi-client small-writes to a single file problem (445)
6425 - fixed mtime updates during file writes (607)
6426 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6427 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6428 - fixed Linux OST size reporting problem (444, 656)
6429 - OST now updates object mtime with writes or setattr (607, 619)
6430 - client verifies file size before zeroing page past EOF (445)
6431 - OST now writes last allocated objid to disk with allocation (108)
6432 - LOV on echo now works (409)
6434 - mds_reint_unlink sends a new buffer, with the EA included. this
6435 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6436 set if a regular file was being unlinked, and it was the last link
6437 - use PtlGet from the target for bulk writes (315)
6438 - OST now updates object mtime with writes or setattr (607, 619)
6439 - LDLM now has a grant-time callback to revalidate locked items, if
6441 - Many MDS operations were reorganized to combat race conditions
6443 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6444 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6446 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6449 - Fully reactivate OST imports after reconnection (512, others)
6450 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6451 - More graceful error handling for truncating on dead OST (515)
6452 - Don't error out unless we're actually accessing dead stripes (474)
6453 - Fix garbage sizes when stripes are missing (410)
6454 - LRU counters were broken, causing constant lock purge (433, 432)
6455 - garbage on read from stripes with failed OSTs (441)
6456 - mark OSCs as active before reconnecting during recovery (438)
6457 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6458 - lfind did not preserve OST order in output (443)
6459 - symlinks cause hung clients, incorrect data (439)
6460 - stop dereferencing request after dropping refcount (457)
6461 - don't LASSERT(spin_is_locked) on non-SMP (455)
6462 - fixes for many rename() bugs
6463 - fstat didn't correctly synchronize attributes (399)
6464 - server must handle lock cancellation during blocking AST prep (487)
6465 - bulk descriptors were free()d too soon (511)
6466 - fix paths in lconf, which would load incorrect modules (451, 507)
6467 - fix confusing lconf 'host not found' error message (386)
6468 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6469 - fix race condition in mdc_blocking_ast() for inode access (526)
6470 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6471 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6472 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6473 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6474 - fix lstripe in several ways (kernel side) (527)
6475 - fix request leak in ldlm_cli_enqueue (262)
6476 - incorrect OSC was marked inactive after OST failure
6477 - call mds_fs_cleanup before unmounting filesystem (524)
6478 - fix races between taking ns_lock and ldlm_lock_change_resource
6479 - fix races updating LOV export open file list
6480 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6481 - fix recovery NULL deref in ldlm_cli_cancel_unused
6482 - fix some DLM races by using new hash table for lock handles (419)
6483 - permit the client to specify desired inodes, at replay
6484 - duplicate requests when we queue them for replay reintegration
6485 - fix last_rcvd offset calculation
6486 - sync after each recovered transaction, so we always make progress
6487 - never, not always, ERESTART requests without transnos
6488 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6490 - skip replay if the MDS says that the client is already connected
6491 - don't check for a recovery-enabled export to match lctl's UUID
6492 - don't INC_USE_COUNT for phantom exports
6493 - don't crash when cleaning up phantom exports (567)
6494 - don't double-finish or set replay data for errored mdc_open requests
6495 - abort requests when they time out, so we don't get old replies
6496 - send/receive replies for AST messages again
6497 - if the client says that it doesn't have the lock, cancel it on the
6499 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6500 mark it as destroyed, it will all work out when decref is called
6501 - fix module use counts (22, 581)
6503 - ASTs now expect a reply (server cancels lock on error reply)
6505 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6508 - fix many simultaneous client startup (392)
6509 - fix dentry->d_it clobbering
6510 - credentials weren't being shipped for readdir/getattr operations
6511 - remove invalid assertions triggered during some concurrent MD
6513 - proper Lustre versions added (336, 389)
6514 - fix memory leak for create error case (398)
6515 - fix LOV locking bug that would get cli/srv out of sync
6516 - fix echo client over LOV (409)
6517 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6518 - fix double-O_EXCL intent crash (424)
6519 - avoid sending multiple lock CANCELs (352)
6521 - MDS can do multi-client recovery (modulo bugs in new code)
6523 - many updates, edits, cleanups
6525 2002-11-18 Phil Schwan <phil@clusterfs.com>
6528 - fix null d_it dereference (346)
6529 - fix full OST/dbench hang (333)
6530 - fix permission problem with file removal (286)
6531 - fix removal of OSCs from LOV when they fail
6532 - fix NULL deref during bulk timeout (214)
6533 - fix problems related to multiple filesystems on one MDS (241)
6534 - fixed serious subtle metadata locking bugs
6535 - free locks on clients when inodes are removed due to memory
6537 - fix inode pointer in lock data (285)
6538 - partial support for multiple MDS on a single host (241)
6539 - data locks weren't cancelled at clear_inode time (290, 311)
6540 - intent locks could lead to unbounded lock growth (205)
6541 - added a maximum lock count, an LRU list, and a flusher
6542 - fix multiple rename (365)
6543 - properly abstracted the echo client
6544 - OSC locked 1 byte too many; fixed
6545 - rewrote brw callback code:
6546 - fixed recovery bugs related to LOVs (306)
6547 - fixed too-many-pages-in-one-write crash (191)
6548 - fixed (again) crash in sync_io_timeout (214)
6549 - probably fixed callback-related race (385)
6551 - Add capability to MDS protocol
6552 - LDLM cancellations and callbacks on different portals
6554 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6557 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6558 - defer open object destruction to close time (601981, 138)
6559 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6560 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6561 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6562 - don't call obd_setattr in ll_file_release for destroyed objects
6563 * protocol change to lustre_msg: move |version| and add |flags|
6564 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6565 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6566 * added replay of create, unlink, link and rename operations during
6567 MDS failover; recovery should be much more robust now
6568 * remove failed OSCs from LOVs (only lov_create uses this so far)
6569 * the lustre-HOWTO was brought (more) up to date (582544)
6571 2002-10-23 Phil Schwan <phil@clusterfs.com>
6574 - in-use dentries weren't being reused properly (617851)
6575 - prevent multiple LDLM setup (599178)
6576 - fix LOV size calculations for truncate (617853)
6577 - fix client handling of MDS intent errors (POSIX)
6578 - fix permission bug in lovstripe.c test (624321)
6579 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6580 - truncate past end of file could corrupt data
6581 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6582 - a race in recovery could return ETIMEDOUT to apps (623947)
6583 - building outside the source directory was fixed
6584 * the lustre-HOWTO was brought (more) up to date (582544)
6585 * major progress was made on recovery functionality
6587 2002-10-10 Phil Schwan <phil@clusterfs.com>
6590 - recovery deadlock fix
6591 - rm -rf causes LBUG fix (617817)
6592 - file open by multiple tasks fix (618962)
6593 - directory permissions bugs (602707 and 620007)
6594 - journal_stop fixed with locking (611313)
6595 - O_APPEND failures resolved (618273, perhaps 614459)
6596 - lconf PATH fix (619770)
6597 - IA64 build fix (621450)
6598 - RPC buffer sizes scale with amount of memory
6600 2002-10-01 Phil Schwan <phil@clusterfs.com>
6603 - locks would be cancelled without throwing away data pages,
6604 resulting in inconsistent data (605627)
6605 - inode attributes were not always being refreshed (605627, 612449)
6606 - lconf now continues to cleanup after lctl reports an error
6607 - MDS now enforces user permissions (602707)
6608 - lprocfs cleanup fixed, but not yet enabled (614157)
6609 - fixed infinite server hang, should a client not respond to an AST
6610 - avoid going into recovery if user calls readlink() with a buffer
6611 that's too small (613941)
6612 - AST RPCs no longer require replies (614867) -- this may be changed
6613 - don't crash server if client sends an IOV that's too big (611336)
6614 - fixed lock conversion deadlock (611892)
6615 - fixed the following of symlinks (614622)
6616 * recovery: the server can remove locks from a client that dies, other
6617 clients can make progress
6618 * more extN patch fixes
6619 * compile-time configurable ptlrpc buffer allocations
6621 - collaborative read cache document
6622 - Lustre Lite Performance CDR document-in-progress
6624 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6627 - fix typo in patch-2.4.18
6629 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6632 - clear ptlrpc request each time in handle_incoming_request()
6633 - unlink of files now destroys the object on the OST
6635 2002-09-19 Peter Braam <braam@clusterfs.com>
6637 * add hard link support
6638 * change obdfile creation method
6639 * kernel patch changed
6641 2002-09-19 Peter Braam <braam@clusterfs.com>
6644 - stack overflow bug in extN fixed
6646 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6648 * documentation updates
6649 - add man pages for config tools
6650 - update tests/README to describe testing with new config tools
6651 - finish metadata API descriptions
6652 * bug fixes and cleanups
6653 - statfs workaround for 16TB limit
6654 - LOV stripe allocation improved, can stripe on subset of OSTs
6655 - LOV file size/IO offset was wrong for files > 4GB in size
6656 - object EA data was being dropped, caused files to be unreadable
6657 - memory overflow with non-LOV OST caused memory corruption
6658 - fixed regression tests to work with new config tools, obdfilter
6659 - fixed bug when directory size became larger than 1 block
6660 - fixed bug (for single client case) when PWD was deleted
6661 - invalidate local directory pages when doing intent-based ops
6662 - avoid LDLM oops when lock callback contained bad data
6664 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6666 * documentation updates
6667 * bug fixes and cleanups
6668 - configuration tools
6671 - 64-bit compile warnings
6672 - 64-bit internal statfs data
6674 * test_brw on persistent OST devices
6676 * lprocfs (disabled)
6678 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6680 * documentation updates
6681 * bug fixes and cleanups
6682 * configuration tools
6684 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6687 * many small fixes to 0_5_4
6688 * io/network handling
6689 * thinkos in MDS operations
6691 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6694 * crucial basic fixes to 0.5.3
6695 * IOR, Iozone work over Elan
6698 2002-08-07 Phil Schwan <phil@clusterfs.com>
6699 * version 0_5_3, our first alpha
6700 * we use the new Portals iovs
6701 * documentation updates
6702 * bug fixes and cleanups
6703 * small changes in the DLM wire protocol
6705 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6706 * version 0_5_1 with some initial stability,
6707 * locking on MD and file I/O.
6708 * documentation updates
6709 * several bug fixes since 0.5.0
6710 * small changes in wire protocol
6712 2002-07-18 Phil Schwan <phil@clusterfs.com>
6714 * delivered as Lustre Light Alpha
6715 * fixed a crash after handling invalid MDS requests
6716 * fixed directory pages for architectures with non-4k pages sizes
6718 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6719 * release version v0_4_4
6720 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6722 * Fixes a number of interruption problems with OST operations.
6723 * Update documentation for portals header changes
6724 * Move all wire protocol structs/defines to lustre_idl.h
6725 * Fixes symlink length bug.
6726 * Add tcpdump to repository.
6728 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6729 * release version v0_4_3
6730 * Fixes statfs for inodes on extN.
6731 * Fixes bug in runtests which would delete /etc/hosts.
6732 * Use 64-bit object IDs wherever possible (not into VFS though)
6733 Remove ost_get_info, which is unused by lustre, and out of date.
6735 2002-07-03 Peter Braam <braam@clusterfs.com>
6736 * release version v0_4_2 Fixes a lookup error (type not passed)
6737 * move forward to head of Portals
6738 * move forward to latest Lustre kernel
6740 2002-06-25 Peter Braam <braam@clusterfs.com>
6741 * release version v0_4_1. Hopefully stable on single node use.