3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.1.1.el7 (RHEL7.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 4.18.0-80.7.1.el8 (RHEL8.0)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.100 (SLES12 SP3)
13 4.4.0-131 (Ubuntu 16.04)
14 4.15.0-32 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-1062.1.1.el7 (RHEL7.7)
19 4.12.14-95.29 (SLES12 SP4)
20 4.15.0-48 (Ubuntu 18.04)
21 * Other clients known to build on unpatched kernels at some point (others may also work):
22 3.10.0-862.14.4.el7 (RHEL7.5)
23 3.10.0-957.27.2.el7 (RHEL7.6)
24 4.14.0-49.13.1.el7a (RHEL7.5)
25 4.14.0-115.2.2.el7a (RHEL7.6)
26 4.18.0-80.7.1.el8 (RHEL8.0)
27 4.4.120-92.70 (SLES12 SP2)
28 4.4.180-94.100 (SLES12 SP3)
29 4.4.0-131 (Ubuntu 16.04)
30 vanilla linux 4.15.0 (ZFS only)
31 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
32 * Recommended ZFS / SPL version: 0.7.13
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
38 * Note for Ubuntu 14.04.5:
39 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
40 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
42 --------------------------------------------------------------------------------
46 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
47 for currently supported client and server kernel versions.
48 * Server known to build on the following kernels (others may also work):
49 2.6.32-696.18.7.el6 (RHEL6.9)
50 3.10.0-862.14.4.el7 (RHEL7.5)
51 3.10.0-957.el7 (RHEL7.6)
52 4.4.120-92.70 (SLES12 SP2)
53 4.4.162-94.69 (SLES12 SP3)
54 4.4.0-131 (Ubuntu 16.04)
55 4.15.0-32 (Ubuntu 18.04)
56 vanilla linux 4.6.7 (ZFS only)
57 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
58 * Client known to build on unpatched kernels (others may also work):
59 2.6.32-696.18.7.el6 (RHEL6.9)
60 2.6.32-754.6.3.el6 (RHEL6.10)
61 3.10.0-862.14.4.el7 (RHEL7.5)
62 3.10.0-957.el7 (RHEL7.6)
63 4.14.0-49.13.1.el7a (RHEL7.5)
64 4.14.0-115.2.2.el7a (RHEL7.6)
65 4.4.120-92.70 (SLES12 SP2)
66 4.4.162-94.69 (SLES12 SP3)
67 4.4.0-131 (Ubuntu 16.04)
68 4.15.0-32 (Ubuntu 18.04)
69 vanilla linux 4.15.0 (ZFS only)
70 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
71 * Recommended ZFS / SPL version: 0.7.9
72 * Tested with ZFS / SPL version: 0.7.9
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
78 * Note for Ubuntu 14.04.5:
79 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
80 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
82 --------------------------------------------------------------------------------
84 2018-04-03 Intel Corporation
86 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
87 for currently supported client and server kernel versions.
88 * Server known to build on patched kernels:
89 2.6.32-431.29.2.el6 (RHEL6.5)
90 2.6.32-504.30.3.el6 (RHEL6.6)
91 2.6.32-573.26.1.el6 (RHEL6.7)
92 2.6.32-642.15.1.el6 (RHEL6.8)
93 2.6.32-696.18.7.el6 (RHEL6.9)
94 3.10.0-693.21.1.el7 (RHEL7.4)
95 3.0.101-0.47.71 (SLES11 SP3)
96 3.0.101-107 (SLES11 SP4)
97 3.12.74-60.64.40 (SLES12 SP1)
98 4.4.114-92.64 (SLES12 SP2)
99 4.4.114-94.11 (SLES12 SP3)
100 3.13.0-101 (Ubuntu 14.04, ZFS only)
101 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
102 4.4.0-116 (Ubuntu 16.04)
103 vanilla linux 4.6.7 (ZFS only)
104 * Client known to build on unpatched kernels:
105 2.6.32-431.29.2.el6 (RHEL6.5)
106 2.6.32-504.30.3.el6 (RHEL6.6)
107 2.6.32-573.26.1.el6 (RHEL6.7)
108 2.6.32-642.15.1.el6 (RHEL6.8)
109 2.6.32-696.18.7.el6 (RHEL6.9)
110 3.10.0-693.21.1.el7 (RHEL7.4)
111 3.0.101-0.47.71 (SLES11 SP3)
112 3.0.101-107 (SLES11 SP4)
113 3.12.74-60.64.40 (SLES12 SP1)
114 4.4.114-92.64 (SLES12 SP2)
115 4.4.114-94.11 (SLES12 SP3)
116 3.13.0-101 (Ubuntu 14.04)
117 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
118 4.4.0-116 (Ubuntu 16.04)
120 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
121 * Recommended ZFS / SPL version: 0.7.6
122 * Tested with ZFS / SPL version: 0.7.6
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
128 * Note for Ubuntu 14.04.5:
129 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
130 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
132 --------------------------------------------------------------------------------
134 2017-07-05 Intel Corporation
136 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
137 for currently supported client and server kernel versions.
138 * Server known to build on patched kernels:
139 2.6.32-431.29.2.el6 (RHEL6.5)
140 2.6.32-504.30.3.el6 (RHEL6.6)
141 2.6.32-573.26.1.el6 (RHEL6.7)
142 2.6.32-642.15.1.el6 (RHEL6.8)
143 2.6.32-696.3.1.el6 (RHEL6.9)
144 3.10.0-514.21.1.el7 (RHEL7.3)
145 3.0.101-0.47.71 (SLES11 SP3)
146 3.0.101-100 (SLES11 SP4)
147 3.12.74-60.64.40 (SLES12 SP1)
148 4.4.59-92.17 (SLES12 SP2)
149 vanilla linux 4.6.7 (ZFS only)
150 * Client known to build on unpatched kernels:
151 2.6.32-431.29.2.el6 (RHEL6.5)
152 2.6.32-504.30.3.el6 (RHEL6.6)
153 2.6.32-573.26.1.el6 (RHEL6.7)
154 2.6.32-642.15.1.el6 (RHEL6.8)
155 2.6.32-696.3.1.el6 (RHEL6.9)
156 3.10.0-514.21.1.el7 (RHEL7.3)
157 3.0.101-0.47.71 (SLES11 SP3)
158 3.0.101-100 (SLES11 SP4)
159 3.12.74-60.64.40 (SLES12 SP1)
160 4.4.59-92.17 (SLES12 SP2)
162 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
163 * Recommended ZFS / SPL version: 0.6.5.9
164 * Tested with ZFS / SPL version: 0.6.5.9
165 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
166 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
167 stack overflow. For more information, please refer to bugzilla 17630.
168 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
169 the re-exporting nodes to be mounted with "32bitapi" mount option
171 --------------------------------------------------------------------------------
173 12-07-2016 Intel Corporation
175 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
176 for currently supported client and server kernel versions.
177 * Server known to build on patched kernels:
178 2.6.32-431.29.2.el6 (RHEL6.5)
179 2.6.32-504.30.3.el6 (RHEL6.6)
180 2.6.32-573.26.1.el6 (RHEL6.7)
181 2.6.32-642.11.1.el6 (RHEL6.8)
182 3.10.0-514.el7 (RHEL7.3)
183 3.0.101-0.47.71 (SLES11 SP3)
184 3.0.101-80 (SLES11 SP4)
185 3.12.62-60.64.8 (SLES12 SP1)
186 vanilla linux 4.5.7 (ZFS only)
187 * Client known to build on unpatched kernels:
188 2.6.32-431.29.2.el6 (RHEL6.5)
189 2.6.32-504.30.3.el6 (RHEL6.6)
190 2.6.32-573.26.1.el6 (RHEL6.7)
191 2.6.32-642.11.1.el6 (RHEL6.8)
192 3.10.0-514.el7 (RHEL7.3)
193 3.0.101-0.47.71 (SLES11 SP3)
194 3.0.101-80 (SLES11 SP4)
195 3.12.62-60.64.8 (SLES12 SP1)
197 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
198 * Recommended ZFS / SPL version: 0.6.5.8
199 * Tested with ZFS / SPL version: 0.6.5.8
200 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
201 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
202 stack overflow. For more information, please refer to bugzilla 17630.
203 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
204 the re-exporting nodes to be mounted with "32bitapi" mount option
206 --------------------------------------------------------------------------------
208 02-29-2016 Intel Corporation
210 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
211 for currently supported client and server kernel versions.
212 * Server known to build on patched kernels:
213 2.6.32-431.29.2.el6 (RHEL6.5)
214 2.6.32-504.30.3.el6 (RHEL6.6)
215 2.6.32-573.12.1.el6 (RHEL6.7)
216 3.10.0-327.3.1.el7 (RHEL7.2)
217 3.0.101-0.47.71 (SLES11 SP3)
218 3.0.101-68 (SLES11 SP4)
219 vanilla linux 4.2.1 (ZFS only)
220 * Client known to build on unpatched kernels:
221 2.6.32-431.29.2.el6 (RHEL6.5)
222 2.6.32-504.30.3.el6 (RHEL6.6)
223 2.6.32-573.12.1.el6 (RHEL6.7)
224 3.10.0-327.3.1.el7 (RHEL7.2)
225 3.0.101-0.47.71 (SLES11 SP3)
226 3.0.101-68 (SLES11 SP4)
229 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
230 * Recommended ZFS / SPL version: 0.6.4.2
231 * Tested with ZFS / SPL version: 0.6.4.2
232 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
233 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
234 stack overflow. For more information, please refer to bugzilla 17630.
235 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
236 the re-exporting nodes to be mounted with "32bitapi" mount option
238 --------------------------------------------------------------------------------
240 03-10-2015 Intel Corporation
242 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
243 for currently supported client and server kernel versions.
244 * Server known to build on patched kernels:
245 2.6.32-431.29.2.el6 (RHEL6.5)
246 2.6.32-504.8.1.el6 (RHEL6.6)
247 3.0.101-0.46 (SLES11 SP3)
248 * Client known to build on unpatched kernels:
249 2.6.32-431.29.2.el6 (RHEL6.5)
250 2.6.32-504.8.1.el6 (RHEL6.6)
251 3.10.0-123.20.1.el7 (RHEL7)
252 3.0.101-0.46 (SLES11 SP3)
253 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
254 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
255 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
256 stack overflow. For more information, please refer to bugzilla 17630.
257 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
258 the re-exporting nodes to be mounted with "32bitapi" mount option
260 Severity : enhancement
262 Description: control OST-index in IDIF via ROCOMPAT flag.
263 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
264 last_rcvd file. For new formatted OST device, it will be auto set;
265 for the case of upgrading from old OST device, you can enable it
266 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
267 enabled, for new created OST-object, its IDIF-in-LMA will contain
268 the OST-index; for the existing OST-object, the OSD will convert
269 old format IDIF as new format IDIF with OST-index stored in the
270 LMA EA when accessing such OST-object or via OI scrub. Once such
271 flag is enabled, it cannot be reverted back, so the system cannot
272 be downgraded to the orignal incompatible version.
274 --------------------------------------------------------------------------------
276 07-30-2014 Intel Corporation
278 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
279 for currently supported client and server kernel versions.
280 * Server known to build on patched kernels:
281 2.6.32-431.20.3.el6 (RHEL6.5)
282 3.0.101-0.31 (SLES11 SP3)
283 * Client known to build on unpatched kernels:
284 2.6.32-431.20.3.el6 (RHEL6.5)
285 3.0.101-0.31 (SLES11 SP3)
288 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
289 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
290 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
291 stack overflow. For more information, please refer to bugzilla 17630.
292 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
293 the re-exporting nodes to be mounted with "32bitapi" mount option
295 --------------------------------------------------------------------------------
297 10-31-2013 Intel Corporation
299 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
300 * Server builds for kernels:
301 2.6.32-358.18.1.el6 (RHEL6.4)
302 * Client builds for unpatched kernels:
303 2.6.32-358.18.1.el6 (RHEL6.4)
304 3.0.92-0.8 (SLES11 SP3)
306 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
307 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
308 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
309 stack overflow. For more information, please refer to bugzilla 17630.
310 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
311 the re-exporting nodes to be mounted with "32bitapi" mount option
313 --------------------------------------------------------------------------------
315 05-30-2013 Intel Corporation
317 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
318 * Server builds for kernels:
319 2.6.32-358.6.2.el6 (RHEL6.4)
320 * Client builds for unpatched kernels:
321 2.6.32-358.6.2.el6 (RHEL6.4)
322 3.0.74-0.6.6 (SLES11 SP2)
324 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
325 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
326 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
327 stack overflow. For more information, please refer to bugzilla 17630.
328 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
329 the re-exporting nodes to be mounted with "32bitapi" mount option
331 --------------------------------------------------------------------------------
333 9-30-2012 Whamcloud, Inc.
335 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
336 * Server support for kernels:
337 2.6.32-279.5.1.el6 (RHEL6)
338 * Client support for unpatched kernels:
339 2.6.32-279.5.1.el6 (RHEL6)
341 2.6.18-238.19.1.el5 (RHEL5)
342 2.6.16.60-0.69.1 (SLES10)
343 * Recommended e2fsprogs version: 1.41.90.wc4
344 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
345 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
346 stack overflow. For more information, please refer to bugzilla 17630.
347 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
348 the re-exporting nodes to be mounted with "32bitapi" mount option
350 Severity : enhancement
352 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
354 Severity : enhancement
356 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
358 Severity : enhancement
360 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
362 Severity : enhancement
364 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
366 Severity : enhancement
368 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
370 --------------------------------------------------------------------------------
372 04-30-2012 Whamcloud, Inc.
374 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
375 * Server support for kernels:
376 2.6.32-220.4.2.el6 (RHEL6)
377 * Client support for unpatched kernels:
378 2.6.32-220.4.2.el6 (RHEL6)
380 2.6.18-238.19.1.el5 (RHEL5)
381 2.6.16.60-0.69.1 (SLES10)
382 * Recommended e2fsprogs version: 1.41.90.wc4
383 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
384 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
385 stack overflow. For more information, please refer to bugzilla 17630.
386 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
387 the re-exporting nodes to be mounted with "32bitapi" mount option
389 Severity : enhancement
391 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
393 Severity : enhancement
395 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
397 Severity : enhancement
399 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
401 Severity : enhancement
403 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
405 --------------------------------------------------------------------------------
407 09-30-2011 Whamcloud, Inc.
409 * Server support for kernels:
410 2.6.18-238.19.1.el5 (RHEL5)
411 2.6.32-131.6.1.el6 (RHEL6)
412 * Client support for unpatched kernels:
413 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
414 2.6.18-238.19.1.el5 (RHEL5)
415 2.6.32-131.6.1.el6 (RHEL6)
416 2.6.29.4-167.fc11 (FC11)
417 2.6.16.60-0.69.1 (SLES10)
419 * Recommended e2fsprogs version: 1.41.90.wc3
420 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
421 Lustre filesystem with 4K stack may cause a stack overflow. For more
422 information, please refer to bugzilla 17630.
423 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
424 mounted with "32bitapi" mount option
426 Severity : enhancement
427 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
429 Severity : enhancement
430 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
432 Severity : enhancement
433 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
436 Frequency : only with SLES10
438 Description: Use OFED "KMP" provided by Novell
439 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
440 Lustre is now built against this package. That means you need to
441 install the ofed-kmp package from Novell for the patchless client.
442 Note that the ofed-kmp that Novell ships may not exactly match the
443 kernel version but should still be compatible.
445 Severity : enhancement
447 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
448 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
450 --------------------------------------------------------------------------------
452 2010-07-15 Oracle, Inc.
454 * Server support for kernels:
455 2.6.18-164.11.1.el5 (RHEL 5)
456 2.6.18-164.11.1.0.1.el5 (OEL 5)
457 * Client support for unpatched kernels:
458 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
459 2.6.18-164.11.1.el5 (RHEL 5),
460 2.6.18-164.11.1.0.1.el5 (OEL 5)
461 2.6.16.60-0.42.8 (SLES 10),
463 2.6.29.4-167.fc11 (FC11)
464 * Recommended e2fsprogs version: 1.41.10-sun2
466 Severity : enhancement
468 Description: Distribute lfs_migrate script with lustre
469 Details : The lfs_migrate mentioned in the manual is now included in the
470 lustre tree in lustre/scripts and will distributed in lustre RPMs.
472 Severity : enhancement
474 Description: Failover nid list added to import proc file.
476 Severity : enhancement
478 Description: "som_preview" mount option required to use SOM
479 Details : The clients will need to use som_preview mount option
483 Description: Lustre does not do 1MB IOs to HW RAID
484 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
486 Severity : enhancement
488 Description: lfs quota output is not very convenient for awk/sed-parsing
489 Details : Some positions in lfs quota output table could be empty or
490 non-empty which made it hard to parse it with scripts, now
491 a dash is put instead of space where there is not supposed
496 Description: Update to OFED 1.5.1
497 Details : OFED 1.5.1 will be used in environments that don't use
498 vendor supplied OFED.
502 Description: Remove lnet. prefix from lctl params display
503 Details : Also renamed lnet.memused to lnet_memused to
508 Description: Use conf_param -d to remove permanant settings.
512 Description: Add list_param -R
513 Details : lctl list_param now supports recursive mode.
516 Bugzilla : 21829, 22312
517 Description: Fix llobdstat.
518 Details : 1. Now updating the write_bytes counter
519 2. Port script fixes from 1.8.
523 Description: decrease the usage of memory on clients.
524 Details : 1. On clients, recycle dentries and inodes unused.
525 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
530 Description: llobdstat fix and enhancement
531 Details : add a counter to set a limit to how many samples will be returned
532 fix a wildcard in the path to limit to obdfilter stats only
536 Description: lfs check only allowed for root
537 Details : Code cleanups and added a regresion test of non-root use of
542 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
543 Details : Fix hr->hr_index overload.
547 Description: Fix FIEMAP functionality.
551 Description: FC11 patchless client support added
555 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
556 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
560 Description: New acc-sm script and various test-framework enhancments.
564 Description: Previously landed patch was reverted.
568 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
572 Description: Add ldiskfs maxdirsize mount option.
576 Description: Fix mkfs.lustre support for 16TB LUNs.
580 Description: Print mmp_check_interval
581 Details : print mmp_check_interval and make it possible to abort
582 mount operation in case it takes too long.
586 Description: Add cascading_rw to lustre mpi tests.
590 Description: Store SOM data in EA.
591 Details : Also several SOM related cleanups.
595 Description: Add refcount to md_open_data
599 Description: Use refcount on lov_request_set to prevent memory corruption.
603 Description: Statahead: use parent dir's i_mutex for synchronization.
607 Description: loadgen improvements.
611 Description: Enable peer health detection in ksocklnd and ko2iblnd.
615 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
619 Description: Fix race on obd_devcie:md_stats
623 Description: Fix SLES 11 client builds.
624 Details : Resolved uninitalized pointer gcc warnings.
628 Description: Allow non-root access for "lfs check".
629 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
633 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
635 Severity : enhancement
637 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
638 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
642 Description: Sanity test_27p fails.
643 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
644 Sanity tests 27 were cleaned up and improved.
648 Description: Unify req format on client/servers
649 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
650 buffers and swabbers approach..
655 Description: drop join file stuff.
657 Severity : enhancement
659 Description: Add 2.6.27 and SLES11 (patchless client) support
663 Description: Fix in ptlrpc_expire_one_request() to print the signed time
666 Severity : enhancement
668 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
671 Severity : enhancement
673 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
675 Severity : enhancement
677 Description: correctly handle big reply message.
678 Details : send LNet event if reply is bigger than buffer and adjust this buffer
683 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
685 Severity : enhancement
687 Description: Add support for OEL5.
691 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
693 Severity : enhancement
695 Description: "weak-modules" support
696 Details : Implement "weak-modules" support which enables kernel modules
697 to be used with any kernel that implements the same kABI. In
698 order to achieve this modules are now installed in
699 /lib/modules/$(uname -r)/updates/kernel on all distributions.
703 Description: client could not reconnect to OST because of an active request.
704 Details : The client now retries to reconnect to the same server, if a
705 connect request failed with EBUSY or -EAGAIN. Test case 35b
706 conf-sanity.sh has been created.
710 Description: Don't allow make backward step on assignin osc next id.
711 Details : race between allocation next id and ll_sync thread can be cause
712 of set wrong osc next id and can be kill valid ost objects.
716 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
720 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
724 Description: Error handling in osc_statfs_interpret() has been improved.
725 Details : Check in osc_statfs_interpret() for EBADR.
729 Description: Do not send statfs() requests to OSTs disabled by administrator.
730 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
732 Severity : enhancement
734 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
736 Severity : enhancement
738 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
740 Severity : enhancement
742 Description: Enhancement for lfs(1) command to use numeric uid/gid.
746 Description: Improved cleanup procedure for conf-sanity test_46a.
750 Description: Do not start bulk transfer, if client tried to reconnect.
751 Details : Do not start bulk transfer, if client tried to reconnect,
752 but failed due to active rpcs. Check for exp_abort_active_req
753 has been implemented.
757 Description: Prevent inconsistences between linux and lustre mount structures.
758 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
759 Make the sleep interruptible.
761 Severity : enhancement
763 Description: provide server to client comms path
764 Details : server to client communications path via new LDLM_SET_INFO rpc
768 Description: truncate starts GFP_FS allocation under transaction causing deadlock
769 Details : ldiskfs_truncate calls grab_cache_page which may start page
770 allocation under an open transaction. This may lead to
771 calling prune_icache with consequent lustre reentrance.
775 Description: send correctly lsm on open replay
776 Details : MDS is trust to LSM size on replay open, but client can set wrong size
781 Description: enable client interface failover
782 Details : When a child reconnects from another NID, properly update export
783 nid hash position and ldlm reverse import.
785 Severity : enhancement
787 Description: Communicate OST degraded/readonly state via statfs to MDS
788 Details : Flags in the statfs returned from OSTs indicate whether the
789 OST is in a degraded RAID state, or if the filesystem has
790 turned read-only after a filesystem error is detected.
795 Description: don't panic if EPROTO was hit when reading symlink
796 Details : correctly handling request reference in error cases.
798 Severity : enhancement
800 Description: Add LustreNetLink, a kernel-userspace communcation path.
802 Severity : enhancement
804 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
809 Description: access to llog context before init.
810 Details : move handling CATALOGS file at osc layer and forbid access to llog
814 Frequency : rare, only when client is writing to OST on local node
816 Description: Avoid deadlock for local client writes
817 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
818 memory freeing context. This allows OST threads to set the
819 PF_MEMALLOC flag on task structures in order to allocate memory
820 from reserved pools and complete IO.
821 Use GFP_HIGHUSER for OST allocations for non-local client writes,
822 so that the OST threads generate memory pressure and allow
823 inactive pages to be reclaimed.
825 Severity : enhancement
827 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
832 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
833 Details : move ldlm namespace creation in setup phase to avoid grab
834 _lprocfs_lock with cli_sem held.
838 Description: Temporarily disable grant shrink.
839 Details : Disable the feature for debugging.
842 Frequency : only when formatting test filesystems
844 Description: Unable to run several mkfs.lustre on loop devices at the same time
845 Details : mkfs.lustre returns error 256 on the concurrent loop devices
846 formatting. The solution is to proper handle the error.
848 Severity : enhancement
850 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
852 Severity : enhancement
854 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
859 Description: resolve race between obd_disconnect and class_disconnect_exports
860 Details : if obd_disconnect will be called to already disconnected export he
861 forget release one reference and osc module can't unloaded.
863 Severity : enhancement
865 Description: Allow tuning service thread via /proc
866 Details : For each service a new
867 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
868 created that can be used to set min/max thread counts, and get the
869 current number of running threads.
874 Descriptoin: don't return error if have partially created objects for file.
875 Details : lov_update_create_set uses set->set_success as index for created
876 objects, so if some requests will be failed, they will have hole
877 at end of array and qos_shrink_lsm for allocate correct lsm.
879 Severity : enhancement
881 Description: Update OFED support to 1.4.1
886 Description: Reduce small size read RPC
887 Details : Set read-ahead limit for every file and only do read-ahead when
888 available read-ahead pages are bigger than 1M to avoid small size
891 Severity : enhancement
893 Description: move AT tunable parameters for more consistent usage
894 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
896 Severity : enhancement
898 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
899 Details : allow skip disconnected ost for send statfs request and hide error
905 Description: fix racy locking of mballoc block bitmaps causing BUG
906 Details : The locking of the mballoc buddy bitmap and the in-memory
907 block bitmap was using two different spin locks in some
908 cases. This made it possible to incorrectly access the
909 mballoc bitmap while another process was modifying it,
910 causing a sanity assertion to fail. While no on-disk corruption
911 was reported, there was some risk of this happening.
914 Frequency : rare, on llog test 6
916 Descriptoin: don't allow connect to already connected import
917 Details : allowing connect to already connected import is hide connecting problem.
922 Descriptoin: allow kill process which wait statahead result
923 Details : for some reasons 'ls' can stick in waiting result from statahead,
924 in this case need way for kill this process.
926 Severity : enhancement
928 Description: Add state history info file, enhance import info file
929 Details : Track import connection state changes in a new osc/mdc proc file;
930 add overview-type data to the osc/mdc import proc file.
932 Severity : enhancement
934 Description: MDS create should not wait for statfs RPC while holding DLM lock.
936 Severity : enhancement
938 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
943 Descriptoin: OST grant too much space to client even there are not enough space.
944 Details : Client will shrink its grant cache to OST if there are no write
945 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
946 this grant cache if there are already not enough avaible space
947 (left_space < total_clients * 32M).
950 Frequency : start MDS on uncleanly shutdowned MDS device
952 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
953 Details : stay in waiting mds<>ost recovery finished produce random bugs
954 due race between two ll_sync thread for one lov target. send
955 ACTIVATE event only if connect realy finished and import have
959 Frequency : rare, connect and disconnect target at same time
961 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
962 Details : don't call obd_disconnect under lov_lock. this long time
963 operation and can block ptlrpcd which answer to connect request.
968 Descriptoin: don't lose wakeup for imp_recovery_waitq
969 Details : recover_import_no_retry or invalidate_import and import_close can
970 both sleep on imp_recovery_waitq, but we was send only one wakeup
974 Frequency : always with long access acl
976 Descriptoin: mds can't pack reply with long acl.
977 Details : mds don't control size of acl but they limited by reint/getattr
980 Severity : enhancement
982 Description: Update to SLES10 kernel-2.6.16.60-0.33.
984 Severity : enhancement
986 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
989 Frequency : start MDS on uncleanly shutdowned MDS device
991 Descriptoin: aborting recovery hang on MDS
992 Details : don't throttle destroy RPCs for the MDT.
995 Frequency : on remount
997 Description: external journal device not working after the remount
998 Details : clear dev_rdonly flag for external journal devices in
1004 Description: shutdown vs evict race
1005 Details : client_disconnect_export vs connect request race.
1006 if client will evicted at this time - we start invalidate
1007 thread without referece to import and import can be freed
1011 Frequency : rare, need acl's on inode.
1013 Description: client can't handle ost additional correctly
1014 Details : if ost was added after client connected to mds client can have
1015 hit lnet_try_match_md ... to big messages to wide striped files.
1016 in this case need teach client to handle config events about add
1017 lov target and update client max ea size at that event.
1019 Severity : enhancement
1021 Description: Changelogs
1022 Details : Changelogs are a lightweight mechanism to track filesystem
1023 metadata and namespace changes. The changelog is recorded
1024 permanently on the MDTs, and is periodically "consumed" / purged
1025 when records are no longer needed.
1027 Severity : enhancement
1029 Description: compact fld format with extents
1030 Details : Store range of seq rather than every seq in FLD. Seq
1031 controller update FLD rather than clients. In Case of CMD, mdt0
1032 has FLD, all other metadata server act as non persistent proxy
1033 for FLD queries and cache fld entries in fld cache.
1038 Description: don't skip ost target if they assigned to file
1039 Details : Drop slow OSCs if we can, but not for requested start idx.
1040 This means "if OSC is slow and it is not the requested
1041 start OST, then it can be skipped, otherwise skip it only
1042 if it is inactive/recovering/out-of-space.
1046 Description: more cleanup in mds_lov
1047 Details : not send LOV EA under replay, we can't know about they size at this
1048 time. Don't allow client connect to mds before any ost connected,
1049 for avoid problems with LOV EA size and returning EIO to client.
1051 Severity : enhancement
1053 Description: Interoperability at server side (Disk interoperability)
1055 Severity : enhancement
1057 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1059 Severity : enhancement
1061 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1063 Severity : enhancement
1065 Description: New client IO stack (CLIO).
1067 Severity : enhancement
1069 Description: Commit on sharing. Eliminate inter-client dependencies between
1070 uncommitted transactions by doing transaction commits.
1071 Thereby clients may recovery independently.
1074 Frequency : Create a symlink file with a very long name
1076 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1077 Details : If there is no extra space in the request for early cancels,
1078 ldlm_req_handles_avail() returns 0 instead of a negative value.
1080 Severity : enhancement
1082 Description: Add /proc entry for import status
1083 Details : The mdc, osc, and mgc import directories now have
1084 an import directory that contains useful import data for debugging
1085 connection problems.
1087 Severity : enhancement
1089 Description: Re-disable certain /proc logging
1090 Details : Enable and disable client's offset_stats, extents_stats and
1091 extents_stats_per_process stats logging on the fly.
1094 Frequency : Only on FC kernels 2.6.22+
1096 Description: oops in statahead
1097 Details : Do not drop reference count for the dentry from VFS when lookup,
1098 VFS will do that by itself.
1100 Severity : enhancement
1102 Description: Generic /proc file permissions
1103 Details : Set /Proc file permissions in a more generic way to enable non-
1104 root users operate on some /proc files.
1108 Description: Hitting mdc_commit_close() ASSERTION
1109 Details : Properly handle request reference release in
1110 ll_release_openhandle().
1114 Description: quota recovery deadlock during mds failover
1115 Details : This patch includes att18982, att18236, att18237 in bz14840.
1117 1. fix osts hang when mds does failover with quotaon
1118 2. prevent watchdog storm when osts threads wait for the
1123 Frequency : only patchless client
1124 Description: add workaround for race between add/remove dentry from hash
1126 Severity : enhancement
1128 Description: Allow OST glimpses to return PW locks
1132 Description: LBUG when llog conf file is full
1133 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1137 Description: Prevent import from entering FULL state when server in recovery
1141 Description: service mount cannot take device name with ":"
1142 Details : Only when device name contains ":/" will mount treat it as
1148 Description: replace ptlrpcd with the statahead thread to interpret the async
1149 statahead RPC callback
1153 Frequency : on recovery
1154 Description: I/O failures after umount during fail back
1155 Details : if client reconnected to restarted server we need join to recovery
1156 instead of find server handler is changed and process self
1157 eviction with cancel all locks.
1159 Severity : enhancement
1161 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1163 Severity : enhancement
1165 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1167 Severity : enhancement
1169 Description: Upcall on Lustre log has been dumped
1170 Details : Allow for a user mode script to be called once a Lustre log has
1171 been dumped. It passes the filename of the dumped log to the
1172 script, the location of the script can be specified via
1173 /proc/sys/lnet/debug_log_upcall.
1178 Description: avoid idr_remove called for id which is not allocated.
1179 Details : Move assigment s_dev for clustered nfs to end of initialization,
1180 to avoid problem with error handling.
1185 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1186 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1187 into CONN_USED_HASH and this prodice warning when put connection
1188 again in unused hash.
1191 Severity : enhancement
1193 Description: File striping can now be set to use an arbitrary pool of OSTs.
1195 Severity : enhancement
1197 Description: Export bytes_read/bytes_write count on OSC/OST.
1201 Description: Early reply size mismatch, MGC loses connection
1202 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1203 the connect flags are properly negotiated.
1208 Description: quotas are not honored with O_DIRECT
1209 Details : all writes with the flag O_DIRECT will use grants which leads to
1210 this problem. Now using OBD_BRW_SYNC to guard this.
1214 Description: add quota statistics
1215 Details : 1. sort out quota proc entries and proc code.
1216 2. add quota statistics
1218 Severity : enhancement
1220 Description: enable quota support for HEAD.
1224 Description: Properly propagate oinfo flags from lov to osc for statfs
1225 Details : restore missing copy oi_flags to lov requests.
1227 Severity : enhancement
1229 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1230 : l_getgroups(8), lst(8), routerstat(8)
1231 Details : included man pages for llobdstat(8), llstat(8),
1232 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1234 Severity : enhancement
1236 Description: configure's --enable-quota should check the
1237 : kernel .config for CONFIG_QUOTA
1238 Details : configure is terminated if --enable-quota is passed but
1239 : no quota support is in kernel
1243 Description: Remove portals compatibility
1244 Details : Remove portals compatibility, not interoperable with releases
1249 Description: Resolve device initialization race
1250 Details : Prevent proc handler from accessing devices added to the
1251 obd_devs array but yet be intialized.
1253 Severity : enhancement
1255 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1257 Severity : enhancement
1259 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1264 Description: Using wrong pointer in osc_brw_prep_request
1265 Details : Access to array[-1] can produce panic if kernel compiled with
1266 CONFIG_PAGE_ALLOC enabled
1270 Description: Client runs out of low memory
1271 Details : Consider only lowmem when counting initial number of llap pages
1275 Description: *optional* service tags registration
1276 Details : if the "service tags" package is installed on a Lustre node
1277 When the filesystem is mounted, a local-node service tag will
1278 be created. See http://inventory.sun.com/ for more information
1279 about the Service Tags asset management system.
1283 Description: Kernel BUG tries to release flock
1284 Details : Lustre does not destroy flock lock before last reference goes
1285 away. So always drop flock locks when client is evicted and
1286 perform unlock regardless of successfulness of speaking to MDS.
1290 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1294 Description: sanity test 65a fails if stripecount of -1 is set
1295 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1300 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1301 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1304 Severity : enhancement
1306 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1310 Description: do not process already freed flock
1311 Details : flock can possibly be freed by another thread before it reaches
1312 to ldlm_flock_completion_ast.
1316 Description: LBUG during stress test
1317 Details : Need properly lock accesses the flock deadlock detection list.
1321 Description: oops in page fault handler
1322 Details : kernel page fault handler can return two special 'pages' in error
1323 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1327 Description: timeout with invalidate import.
1328 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1329 should be handled by ptlrpcd. This produce long age waiting and
1330 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1332 Severity : enhancement
1334 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1338 Description: enable MGS and MDT services start separately
1339 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1340 a co-located MGS without starting the MGS, which is a complement
1341 to 'nosvc' mount option.
1345 Frequency : after recovery
1346 Description: precreate to many object's after del orphan.
1347 Details : del orphan st in oscc last_id == next_id and this triger growing
1348 count of precreated objects. Set flag LOW to skip increase count
1349 of precreated objects.
1353 Frequency : rare, on clear nid stats
1354 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1355 Details : when clean nid stats sometimes try destroy live entry,
1356 and this produce panic in free.
1360 Description: Stack overflow during MDS log replay
1361 ease stack pressure by using a thread dealing llog_process.
1365 Description: wait until IO finished before start new when do lock cancel.
1366 Details : VM protocol want old IO finished before start new, in this case
1367 need wait until PG_writeback is cleared until check dirty flag and
1368 call writepages in lock cancel callback.
1370 Severity : enhancement
1372 Description: using special macro for print time and cleanup in includes.
1376 Description: mds_mfd_close() ASSERTION(rc == 0)
1377 Details : In mds_mfd_close(), we need protect inode's writecount change
1378 within its orphan write semaphore to prevent possible races.
1382 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1386 Frequency : rare, on shutdown ost
1387 Description: don't hit live lock with umount ost.
1388 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1389 use shrink_dcache_sb instead.
1393 Description: don't panic with use echo client
1394 Details : echo client pass NULL as client nid pointer and this produce null
1395 pointer dereference.
1399 Description: fix build on ppc32
1400 Details : compile code with -m64 flag produce wrong object file for ppc32.
1404 Description: add message levels for liblustreapi
1408 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1409 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1410 will result in an -EAGAIN and in the exit call path we call
1411 alloc_rr() which will with increasing aggressiveness attempt to
1412 aquire precreated objects on the minimum number of required OSCs.
1416 Description: Use old size assignment to avoid deadlock
1417 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1418 the scheduling while holding a spinlock. We do not need locking
1419 for size in ll_update_inode() because size is only updated from
1420 the MDS for directories or files without objects, so there is no
1421 other place to do the update, and concurrent access to such inodes
1422 are protected by the inode lock.
1426 Description: resolve "_IOWR redefined" build error on SLES10
1430 Description: dump the memory debugging after all modules are unloaded to
1431 suppress false negative in conf_sanity test 39
1433 Severity : enhancement
1435 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1440 Description: fix SLES kernel versioning
1441 Details : the kernel version for our SLES 10 kernel did not include a "-"
1442 before the "smp" at the end. while this was not a problem in
1443 general, it did mean that software trying to use the kernel
1444 version to try to detect a vendor specific kernel would fail.
1445 this was most evident by the OFED build scripts.
1449 Description: Don't update lov_desc members until making sure they are valid
1450 Details : When updating lov_desc members via proc fs, need fix their
1451 validities before doing the real update.
1455 Description: don't put request into delay list while invalidate in flight.
1456 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1457 import in flight. this produce timeout for invalidate and sometimes
1458 can cause stale data.
1462 Frequency : on ppc only
1463 Description: not convert ost objects for directory because it's not exist.
1464 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1468 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1470 Severity : enhancement
1472 Description: Update to support 2.6.22.14 vanilla kernel.
1476 Frequency : rare, on recovery
1477 Description: read procfs can produce deadlock in some situation
1478 Details : Holding lprocfs lock which send rpc can produce block for destroy
1479 obd objects and this also block reconnect with -EALREADY. This
1480 isn't fix all lprocfs bugs - but make it rare.
1482 Severity : enhancement
1484 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1487 Frequency : frequent on X2 node
1489 Description: mdc_set_open_replay_data LBUG
1490 Details : Set replay data for requests that are eligible for replay.
1494 Description: lustre_mgs: operation 101 on unconnected MGS
1495 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1496 MGC, and late on MGC cannot successfully connect to MGS and a lot
1497 of the error messages complaining that MGS is not connected.
1501 Frequency : on network error
1502 Description: panic with double free request if network error
1503 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1504 but it's true only for synchronus enqueue, for async enqueue
1505 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1508 Severity : enhancement
1510 Description: client-side metadata stat-ahead during readdir(directory readahead)
1511 Details : perform client-side metadata stat-ahead when the client detects
1512 readdir and sequential stat of dir entries therein
1515 Frequency : on start mds
1517 Description: Implement get_info(last_id) in obdfilter.
1520 Frequency : occasional
1522 Description: Correctly check stale fid, not start epoch if ost not support SOM
1523 Details : open with flag O_CREATE need set old fid in op_fid3 because
1524 op_fid2 was overwritten with new generated fid, but mds can answer
1525 with one of these two fids and both is not stale. Setattr
1526 incorrectly started an epoch and assume will be called
1527 done_writing, but without SOM done_writing ever being called.
1530 Frequency : rare, depends on device drivers and load
1532 Description: MDS or OSS nodes crash due to stack overflow
1533 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1534 In some cases, in conjunction with device drivers that use a lot
1535 of stack the MDS (or possibly OSS) service threads could overflow
1536 the stack. One change which was identified to consume additional
1537 stack has been reworked to avoid the extra stack usage.
1540 Frequency : occasional
1542 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1543 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1544 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1545 oscc_has_objects and see OSC has no objests and call
1546 oscc_internal_create to resend create request.
1548 Severity : enhancement
1550 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1552 Severity : enhancement
1554 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1557 Frequency : very rare
1559 Description: Fix replay if there is an un-replied request and open
1560 Details : In some cases, older replay request will revert the
1561 mcd->mcd_last_xid on MDS which is used to record the client's
1562 latest sent request.
1564 Severity : enhancement
1566 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1568 Severity : enhancement
1570 Description: Add rhel5 support to HEAD.
1572 Serverity : enhancement
1574 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1579 Description: Don't allow skipping OSTs if index has been specified.
1580 Details : Don't allow skipping OSTs if index has been specified, make
1581 locking in internal create lots better.
1585 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1586 Details : access to bitfield in structure is always rounded to long
1587 and this produce problem with not atomic change any bit.
1591 Description: Lustre make rpms failed.
1592 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1593 builds Lustre rpms from tarball.
1596 Frequency : rare on shutdown ost
1598 Description: If llog cancel was not send before clean_exports phase, this can
1599 produce deadlock in llog code.
1600 Details : If llog thread has last reference to obd and call class_import_put
1601 this produce deadlock because llog_cleanup_commit_master wait when
1602 last llog_commit_thread exited, but this never success because was
1603 called from llog_commit_thread.
1607 Description: allow userland application know is lost one of stripes.
1608 Details : fill lvb_blocks with error code on ost and return it to
1609 application if error flag found.
1613 Description: NULL lov_tgts causing MDS oops
1614 Details : more safe checks for NULL lov_tgts for avoid oops.
1616 Severity : enhancement
1618 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1622 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1623 Details : If an OST is down the MDS will hang indefinitely in
1624 obd_statfs_rqset() waiting for the statfs data. While for
1625 MDS QOS usage of statfs, it should not stuck in waiting.
1627 Severity : enhancement
1629 Description: remote_acl support
1630 Details : Support ACL-based permission check for remote user.
1631 Support setfacl/getfacl for remote user with the utils
1632 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1633 the system "{s,g}etfacl" utils.
1635 Severity : enhancement
1637 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1639 Severity : enhancement
1641 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1645 Description: make mgs_setparam() handle fsname containing dash
1646 Details : fsname containing a dash does not work with lctl conf_param
1648 Severity : enhancement
1650 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1652 Severity : enhancement
1654 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1658 Frequency : rare, at shutdown
1659 Description: access already free / zero obd_namespace.
1660 Details : if client_disconnect_export was called without force flag set,
1661 and exist connect request in flight, this can produce access to
1662 NULL pointer (or already free pointer) when connect_interpret
1663 store ocd flags in obd_namespace.
1667 Frequency : only at startup
1668 Description: not alloc memory with spinlock held.
1669 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1670 if any spinlock held.
1672 Severity : enhancement
1674 Description: make lustre randomly fail allocating memory
1675 Details : Make lustre randomly failed allocating memory for testing purpose.
1677 Severity : enhancement
1679 Description: lost problems with lov objid file
1680 Details : Fixes some scability and access to not inited memory problems
1681 in work with lov objdid file.
1686 Description: lfs find does not continue on file error
1687 Details : Continue other files processing when a file/dir is absent.
1691 Description: Inconsistent usage of lustre_pack_reply()
1692 Details : Standardize the usage of lustre_pack_reply() such that it
1693 always generate a CERROR on failure.
1696 Frequency : occasional
1698 Description: MDS hang or stay in waiting lock
1699 Details : If client receive lock with CBPENDING flag ldlm need send lock
1700 cancel as separate rpc, to avoid situation when cancel request
1701 can't processed due all i/o threads stay in wait lock.
1705 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1709 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1713 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1717 Description: Update to RHEL4 latest kernel.
1719 Severity : enhancement
1721 Description: Build SLES10 patchless client fails
1722 Details : The configure was broken by run ./configure with
1723 --with-linux-obj=.... argument for patchless client. When the
1724 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1725 can't search header adequately. Use absolute path such as
1726 -I($LINUX)/include instead.
1730 Description: interrupt oig_wait produce painc on resend.
1731 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1732 private set, and this produce situation when rq_ptlrpcd_data not
1733 copyed to new allocated request and triggered LBUG on assert
1734 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1735 ptlrpcd set if request is changed and can be safety changed to use
1740 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1741 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1742 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1743 these macros and replace them with their difinitions instead.
1747 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1748 Details : Update which_patch & target file for SLES10 latest kernel.
1751 Description: Remove client patches from SLES 10 kernel.
1752 Details : This causes SLES 10 clients to behave as patchless clients
1753 even on a Lustre-patched (server) kernel.
1755 Severity : enhancement
1757 Description: self-adjustable client's lru lists
1758 Details : use adaptive algorithm for managing client cached locks lru
1759 lists according to current server load, other client's work
1760 pattern, memory activities, etc. Both, server and client
1761 side namespaces provide number of proc tunables for controlling
1764 Severity : enhancement
1766 Description: light-weight GSS support
1767 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1768 and reduce performance overhead.
1770 Severity : enhancement
1772 Description: Linux keyring support
1773 Details : Support using service of Linux keyring for Lustre GSS internal
1774 context refresh/cache mechanism.
1778 Description: Fix errors in lfs documentation
1779 Details : Fixes man pages
1783 Description: don't use GFP_* in generic Lustre code.
1784 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1788 Description: obdclass is limited by single OBD_ALLOC(idarray)
1789 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1793 Description: warnings with build patchless client with vanila 2.6.19 and up
1794 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1795 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1799 Frequency : Only for SLES
1800 Description: sanity_quota fail test_1
1801 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1802 file, which makes TSTID[2] inunique.
1806 Frequency : Only for SLES9
1807 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1808 Details : Update target/ChangeLog/which_patch .
1812 Frequency : Only for test_10 in sanity-quota.sh
1813 Description: a bug in quota test code
1814 Details : checking a test flag in a wrong place
1818 Frequency : Only for test_13 in sanity-quota.sh
1819 Description: a bug in quota test script
1820 Details : shouldn't check overfull allocated quota
1824 Description: jbd statistics
1825 Details : Port older jbd statistics patch for sles10
1827 Severity : enhancement
1829 Description: Add printing inode info into message about error in writepage.
1834 Description: reply_lock_interpret crash due to race with it and lock cancel.
1835 Details : Do not replay locks that are being cancelled. Do not reference
1836 locks by their address during replay, just by their handle.
1840 Frequency : When flocks are used.
1841 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1842 Details : Flock locks might destroy just granted lock if it could be merged
1843 with another existing flock, this is done in completion handler,
1844 so teach ldlm_cli_enquque_fini that this is a valid case for
1850 Description: Oops in read and write path when failing to allocate lock.
1851 Details : Check if lock allocation failed and return error back.
1855 Description: lstripe command fails for valid OST index
1856 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1857 instead of lov->desc.ld_active_tgt_count.
1861 Description: sanity.sh failed test 103
1862 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1867 Description: df doesn't work properly if diskfs blocksize != 4K
1868 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1872 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1873 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1875 Severity : enhancement
1877 Description: Adaptive timeouts
1878 Details : RPC timeouts adapt to changing server load and network
1879 conditions to reduce resend attempts and improve recovery time.
1883 Description: llapi_file_create() does not allow some changes
1884 Details : add llapi_file_open() that allows specifying the mode and
1885 open flags, and also returns an open file handle.
1889 Description: merge and cleanup kernel patches.
1890 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1894 Description: Add journal checksum support.(Kernel part)
1895 Details : The journal checksum feature adds two new flags i.e
1896 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1897 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1898 indicates that the commit block contains the checksum for
1899 the blocks described by the descriptor blocks. Now commit
1900 record can be sent to disk without waiting for descriptor
1901 blocks to be written to disk. This behavior is controlled
1902 using JBD2_FEATURE_ASYNC_COMMIT flag.
1906 Description: OSS needs mutliple precreate threads
1907 Details : Add ability to start more than one create thread per OSS.
1911 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1912 Details : execute lfs setstripe on client
1916 Description: mds_obd_create error creating tmp object
1917 Details : When the user sets quota on root, llog will be affected and can't
1918 create files and write files.
1921 Frequency : Always on ia64 patchless client, and possibly others.
1923 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1924 Details : This allows the patchless client to be loaded on architectures
1925 without this export.
1929 Description: RedHat Update kernel for RHEL5
1930 Details : Add the kernel config file for RHEL5.
1934 Description: RedHat Update kernel for RHEL5
1935 Details : Modify the kernel config file more closer RHEL5.
1939 Description: Build failure against Centos5 (RHEL5)
1940 Details : Define PAGE_SIZE when it isn't present.
1944 Description: test_7 on sanity-quota.sh can't be used on separate machines
1945 Details : add facet to handle it
1949 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1950 Details : with NFS, the anon dentry's parent was set to itself in
1951 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1952 corresponding dentry other than use rec->ur_name.
1954 Severity : enhancement
1956 Description: lfs setstripe enhancement
1957 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1961 Description: enable data checksumming by default
1962 Details : enable checksum by default, allow --disable-checksum
1963 configure option and "-o nochecksum" mount option. Checksums
1964 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1968 Description: lustre support for RHEL5
1969 Details : Add support for RHEL5.
1973 Description: Client eviction due to failover config
1974 Details : after a connection loss, the lustre client should attempt to
1975 reconnect to the last active server first before trying the
1976 other potential connections.
1980 Description: when mds and osts use different quota unit(32bit and 64bit),
1981 quota will be released repeatly.
1982 Details : void sending multiple quota reqs to mds, which will keep the
1983 status between the reqs.
1987 Description: osts not allocated evenly to files
1988 Details : change the condition to increase offset_idx
1991 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1993 Description: Data corruption for OSTs that are formatted larger than 2TB
1995 Details : When generating the bio request for lustre file writes the
1996 sector number would overflow a temporary variable before being
1997 used for the IO. The data reads correctly from Lustre (which
1998 will overflow in a similar manner) but other file data or
1999 filesystem metadata may be corrupted in some cases.
2003 Description: Tune the kernel for good SCSI performance.
2004 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2005 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2010 Description: rewrite ext2-derived code in obdclass/uuid.c
2011 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2012 from scratch preserving functionality.
2016 Description: "lfs find -obd UUID" prints directories
2017 Details : "lfs find -obd UUID" will return all directory names instead
2018 of just file names. It is incorrect because the directories
2019 do not reside on the OSTs.
2023 Description: use i_size_read and i_size_write in 2.6 port
2024 Details : replace inode->i_size access with i_size_read/write()
2027 Frequency : when removing large files
2029 Description: scheduling issue during removal of large Lustre files
2030 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2031 It causes scheduling issues when removing large files (17TB in the
2035 Frequency : only with liblustre clients on XT3
2037 Description: evictions taking too long
2038 Details : allow llrd to evict clients directly on OSTs
2041 Frequency : only on ppc
2043 Description: /proc/fs/lustre/devices broken on ppc
2044 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2045 We should make sure the type of 'index' is loff_t and then cast
2046 explicitly as needed below. Do not assign an explicitly cast
2050 Frequency : only for rhel5
2052 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2053 Details : Modify the target file & which_kernel.
2055 Severity : enhancement
2057 Description: omit set fsid for export NFS
2058 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2059 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2063 Frequency : Always, for kernels after 2.6.16
2064 Description: Fix warning idr_remove called for id=.. which is not allocated.
2065 Details : Last kernels save old s_dev before kill super and not allow
2066 to restore from callback - restore it before call kill_anon_super.
2070 Description: buffer overruns could theoretically occur
2071 Details : llapi_semantic_traverse() modifies the "path" argument by
2072 appending values to the end of the origin string, and a
2073 overrun may occur. Adding buffer overrun check in liblustreapi.
2077 Description: Fix error on 'ls .' at the top of the Lustre mount.
2078 Details : Don't revalidate dentry if it is a root dentry.
2082 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2083 Details : Modify vm-tunables-rhel4.patch.
2087 Description: Kernel config for 2.6.18-vanilla.
2088 Details : Modify targets/2.6-vanilla.target.in.
2089 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2090 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2091 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2092 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2096 Description: improve handling recoverable errors
2097 Details : if request processig with error which can be recoverable on server
2098 request should be resend, otherwise page released from cache and
2103 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2104 Details : Modify target file & which_patch.
2105 A flaw was found in the IA32 system call emulation provided
2106 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2107 value could be stored in the %RAX register, which could trigger an
2108 out-of-bounds system call table access. An untrusted local user
2109 could exploit this flaw to run code in the kernel
2110 (ie a root privilege escalation). (CVE-2007-4573).
2114 Description: O_DIRECT bypasses client statistics.
2115 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2116 read_ahead_stats were updated. copied the stats section from
2117 osc_send_oap_rpc() into async_internal().
2121 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2125 Description: change order of libsysio includes
2126 Details : '#include sysio.h' should always come before '#include xtio.h'
2130 Description: handle "serious error: objid * already exists" more gracefully
2131 Details : If LAST_ID value on disk is smaller than the objects existing in
2132 the O/0/d* directories, it indicates disk corruption and causes an
2133 LBUG(). If the object is 0-length, then we should use the existing
2134 object. This will help to avoid a full fsck in most cases.
2136 Severity : enhancement
2138 Description: adapt the lustre_config script to support the upgrade case
2139 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2140 server targets to 1.6 in parallel.
2144 Description: To avoid grant space > avaible space when the disk is almost
2145 full. Without this patch you might see the error "grant XXXX >
2146 available" or some LBUG about grant, when the disk is almost
2148 Details : In filter_check_grant, for non_grant cache write, we should
2149 check the left space by if (*left > ungranted + bytes), instead
2150 of (*left > ungranted), because only we are sure the left space
2151 is enough for another "bytes", then the ungrant space should be
2152 increase. In client, we should update cl_avail_grant only there
2153 is OBD_MD_FLGRANT in the reply.
2157 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2158 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2159 described in CVE-2007-4573. This problem could allow untrusted
2160 local users to gain root access.
2163 Frequency : when using O_DIRECT and quotas
2165 Description: Incorrect file ownership on O_DIRECT output files
2166 Details : block usage reported by 'lfs quota' does not take into account
2167 files that have been written with O_DIRECT.
2172 Description: touch file failed when fs is not full
2173 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2174 otherwise we can get ENOSP while fs is not full.
2178 Description: parallel lock callbacks
2179 Details : Instead of sending blocking and completion callbacks as separated
2180 requests, adding them to a set and sending in parallel.
2183 Frequency : only for Cray XT3
2184 Bugzilla : 12829/13455
2185 Description: Changing primary group doesn't change the group lustre assigns to
2187 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2188 overridden with the primary group provided by the group upcall,
2189 whereas the supplied fsgid can be trusted if it is in the list of
2190 supplementary groups returned by the group upcall.
2192 Severity : enhancement
2194 Description: Allow masking D_WARNING, D_ERROR messages from console
2195 Details : Console messages can now be disabled via lnet.printk.
2199 Description: User code with malformed file open parameter crashes client node
2200 Details : Before packing join_file req, all the related reference should be
2201 checked carefully in case some malformed flags cause fake
2202 join_file req on client.
2206 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2207 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2208 mode is equal to the granted mode. However, it should also check
2209 that LDLM_FL_CP_REQD is not set.
2213 Description: Heavy nfs access might result in deadlocks
2214 Details : After ELC code landed, it is now improper to enqueue any mds
2215 locks under och_sem, because enqueue might want to decide to
2216 cancel open locks for same inode we are holding och_sem for.
2220 Description: Client eviction while running blogbench
2221 Details : A lot of unlink operations with concurrent I/O can lead to a
2222 deadlock causing evictions. To address the problem, the number of
2223 oustanding OST_DESTROY requests is now throttled to
2224 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2225 callbacks are processed in priority.
2229 Description: enable ACLs on MDS by default
2230 Details : ACLs must be enabled on MDS by default.
2233 Frequency : PPC/PPC64 only
2235 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2236 Details : fix duplicated definitions between asm-ppc64/types.h and
2237 lustre_types.h on PPC.
2240 Frequency : PPC/PPC64 only
2242 Description: asm-ppc/segment.h does not exist
2243 Details : fix compile issue on PPC.
2247 Description: better handle error messages in extents code
2250 Frequency : RHEL4 only
2252 Description: mkfs is very slow on IA64/RHEL4
2253 Details : A performance regression has been discovered in the MPT Fusion
2254 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2255 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2256 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2258 Severity : enhancement
2260 Description: SNMP support enhancement
2261 Details : Adding total number of sampled request for an MDS node in snmp
2264 Severity : enhancement
2266 Description: Optimize ldlm waiting list processing for PR extent locks
2267 Details : When processing waiting list for read extent lock and meeting read
2268 lock that is same or wider to it that is not contended, skip
2269 processing rest of the list and immediatelly return current
2270 status of conflictness, since we are guaranteed there are no
2271 conflicting locks in the rest of the list.
2275 Description: Time out and refuse to reconnect
2276 Details : When the failover node is the primary node, it is possible
2277 to have two identical connections in imp_conn_list. We must
2278 compare not conn's pointers but NIDs, otherwise we can defeat
2279 connection throttling.
2283 Description: port llog fixes from b1_6 into HEAD
2284 Details : Port llog reference couting and some llog cleanups from b1_6
2285 (bug 10800) into HEAD, for protect from panic and access to already
2286 free llog structures.
2290 Description: Detect stride IO mode in read-ahead
2291 Details : When a client does stride read, read-ahead should detect that and
2292 read-ahead pages according to the detected stride pattern.
2296 Description: data checksumming impacts single node performance
2297 Details : add support for several checksum algorithm. Currently, only CRC32
2298 and Adler-32 are supported. The checksum type can be changed on
2299 the fly via /proc/fs/lustre/osc/*/checksum_type.
2303 Description: use adler32 for page checksums
2304 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2309 Description: build for x2 fails
2310 Details : fix compile issue on Cray systems.
2314 Description: Properly match for duplicate locks
2315 Details : Due to different lock order from skiplists code, we need to
2316 traverse entire list for now
2319 Frequency : only on PPC/SLES10
2321 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2322 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2323 be defined. Add a hack in mkfs_lustre.c to work around this compile
2328 Description: LASSERT on MDS when client holding flock lock dies
2329 Details : ldlm pool logic depends on number of granted locks equal to
2330 number of released locks which is not true for flock locks, so
2331 just exclude such locks from consideration.
2335 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2336 Details : Use fsfilt_sync() for both the whole filesystem sync and
2337 individual file sync to eliminate dangerous inode locking
2338 with I_LOCK that can lead to a deadlock.
2342 Description: performance in 1.6.3
2343 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2344 sure that 1MB requests are not fragmented by the block layer.
2346 Severity : enhancement
2348 Description: organize the server-side client stats on per-nid basis
2349 Details : Change the structure of stats under obdfilter and mds to
2357 The "uuid"s file would list the uuids of _active_ exports.
2358 And the clear entry is to clear all stats and stale nids.
2360 Severity : enhancement
2362 Description: eliminate client locks in face of contention
2363 Details : file contention detection and lockless i/o implementation
2364 for contended files.
2368 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2371 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2373 Description: MDS may lose file striping (and hence file data) in some cases
2374 Details : If there are additional extended attributes stored on the MDS,
2375 in particular ACLs, SELinux, or user attributes (if user_xattr
2376 is specified for the client mount options) then there is a risk
2377 of attribute loss. Additionally, the Lustre file striping
2378 needs to be larger than default (e.g. striped over all OSTs),
2379 and an additional attribute must be stored initially in the
2380 inode and then increase in size enough to be moved to the
2381 external attribute block (e.g. ACL growing in size) for file
2386 Description: skiplist implementation simplification
2387 Details : skiplists are used to group compatible locks on granted list
2388 that was implemented as tracking first and last lock of each
2389 lock group the patch changes that to using doubly linked lists
2393 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2394 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2395 because LASSERT on both the data supplied by a client, and the
2396 data on disk is dangerous and incorrect.
2398 Severity : enhancement
2400 Description: Slow truncate/writes to huge files at high offsets.
2401 Details : Directly associate cached pages to lock that protect those pages,
2402 this allows us to quickly find what pages to write and remove
2403 once lock callback is received.
2407 Description: more ldlm soft lockups
2408 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2409 starve other threads from the resource lock for a long time in
2410 case of long waiting queue, so change the debug level from
2411 D_OTHER to the less frequently used D_INFO.
2413 Severity : enhancement
2415 Description: add -gid, -group, -uid, -user options to lfs find
2419 Description: Hung threads in invalidate_inode_pages2_range
2420 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2421 once one is completed. As a result, some RPCs are stuck in the
2422 queue and are never sent.
2426 Description: filter threads hungs on waiting journal commit
2427 Details : Cleanup filter group llog code, then only filter group llog will
2428 be only created in the MDS/OST syncing process.
2432 Description: Procfs and llog threads access destoryed import sometimes.
2433 Details : Sync the import destoryed process with procfs and llog threads by
2434 the import refcount and semaphore.
2436 Severity : enhancement
2438 Description: openlock cache of b1_6 port to HEAD
2443 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2444 Details : If insertion of an extent fails, then discard the inode
2445 preallocation and free data blocks else it can lead to duplicate
2450 Description: don't always update ctime in ext3_xattr_set_handle()
2451 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2452 In some cases the ctime should not be updated, for example for
2453 2.0->1.8 compatibility it is necessary to delete an xattr and it
2454 should not update the ctime.
2458 Bugzilla : 15713/16362
2459 Description: Assertion in iopen_connect_dentry in 1.6.3
2460 Details : looking up an inode via iopen with the wrong generation number can
2461 populate the dcache with a disconneced dentry while the inode
2462 number is in the process of being reallocated. This causes an
2463 assertion failure in iopen since the inode's dentry list contains
2464 both a connected and disconnected dentry.
2468 Description: assertion failure in ldlm_handle2lock()
2469 Details : fix a race between class_handle_unhash() and class_handle2object()
2470 introduced in lustre 1.6.5 by bug 13622.
2475 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2476 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2481 Description: oops in sd_iostats_seq_show()
2482 Details : unloading/reloading the scsi low level driver triggers a kernel
2483 bug when trying to access the sd iostat file.
2488 Description: Kernel panics during QLogic driver reload
2489 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2490 patch, causing memory corruption.
2495 Description: journal_dev option does not work in b1_6
2496 Details : pass mount option during pre-mount.
2498 Severity : enhancement
2500 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2501 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2502 extent information of a file. It can be used to map logical blocks
2503 in a file to physical blocks in the block device.
2507 Description: LDLM soft lockups - improvement
2508 Details : It is be possible to send the lock handle along with each read
2509 or write request because the client is already doing a lock match
2510 itself so there isn't any reason the OST should have to re-do that
2514 Frequency : only on Cray X2
2516 Description: X2 build failures
2517 Details : fix build failures on Cray X2.
2521 Description: xid & resent requests
2522 Details : Initialize RPC XID from clock at startup (randomly if clock is
2525 Severity : enhancement
2527 Description: Add lustre_start utility to start or stop multiple Lustre servers
2532 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2533 Details : In case of memory pressure, list_del() can be called twice on
2534 req->rq_history_list, causing a kernel oops.
2538 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2539 Details : fix stack overflow in the distributed lock manager by defering
2540 export eviction after a failed AST to the elt thread instead of
2541 handling it in the dlm interpret routine.
2545 Description: Convert some comments to new format.
2546 Details : Update documenting comments to match doxygen conventions.
2550 Description: Grammar fixes.
2551 Details : A couple of trivial spelling fixes.
2555 Description: OSD_COUNTERS-mandatory
2556 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2557 to simplify the code.
2561 Description: simplify cmm_device freeing logic.
2562 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2566 Description: Add lockdep support to dt_object_operations locking interface.
2567 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2568 parameter indicating lock ordering. Update mdd code to use new
2573 Description: Introduce failloc constants for lockless IO tests.
2574 Details : Add two new failloc constants to test lockless IO. Only one of
2575 them in implemented---another is checked in yet to be landed
2580 Description: Add lockdep support for inode mutex.
2581 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2582 in the situations where Lustre has to lock more than one inode
2587 Description: Add optional invariants checking support.
2588 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2589 configure switch. This macro is to be used for consistency and
2590 sanity checks that are too expensive to be left in `production'
2595 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2596 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2601 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2602 Details : Use spin_lock_nested() in (the only) situation where more than
2603 one ldlm_lock is locked simultaneously. Also, fix possible
2604 dead-lock in ldlm_lock_change_resource() by enforcing particular
2609 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2610 Details : Instead of specifying each ldlm_lock call-back through separate
2611 parameter, wrap them into struct ldlm_callback_suite.
2615 Description: Kill join_lru obd method and its callers.
2616 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2617 in memory---a requirement implemented through ->o_join_lru() obd
2618 method in HEAD. Remove this method and its users.
2622 Description: Add asynchronous ldlm ENQUEUE completion handler.
2623 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2624 case that stock ldlm_completion_ast() cannot handle as it waits
2625 until lock is granted. Introduce new ldlm_completion_ast_async()
2626 for this. Also comment ldlm_completion_ast().
2630 Description: ldlm_error <-> errno conversion.
2631 Details : Add functions to map (rather arbitrary) between LDLM error codes
2632 and standard errno values. CLIO needs this to prevent LDLM specific
2633 constants from escaping ldlm and osc.
2637 Description: Kill unused ldlm_handle2lock_ns() function.
2638 Details : Kill unused ldlm_handle2lock_ns() function.
2642 Description: Add lu_ref support to ldlm_lock
2643 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2644 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2645 and ldlm_resource. LDLM interface has to be changed, because code that
2646 releases a reference on a lock, has to "know" what reference this is.
2647 In the most frequent case
2649 lock = ldlm_handle2lock(handle);
2651 LDLM_LOCK_PUT(lock);
2653 no changes are required. When any other reference (received _not_ from
2654 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2655 instead of LDLM_LOCK_PUT().
2657 Arguably, changes are pervasive, and interface requires some discipline
2658 for proper use. On the other hand, it was very instrumental in finding
2659 a few leaked lock references.
2663 Description: Add ldlm_lock_addref_try().
2664 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2665 attempts to addref a lock that might be being canceled
2670 Description: Add ldlm_weigh_callback().
2671 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2672 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2673 hard-coded `number of pages' logic.
2677 Description: Add lockdep annotations to llog code.
2678 Details : Use appropriately tagged _nested() locking calls in the places
2679 where llog takes more than one ->lgh_lock lock.
2683 Description: Add loi_kms_set().
2684 Details : Wrap kms updates into a helper function.
2688 Description: Constify instances of struct lsm_operations.
2689 Details : Constify instances of struct lsm_operations.
2693 Description: lu_conf support.
2694 Details : On a server, a file system object is uniquely identified
2695 by a fid, which is sufficient to locate and load all object
2696 state (inode). On a client, on the other hand, more data are
2697 necessary instantiate an object. Change lu_object_find() and
2698 friends to take additional `lu_conf' argument describing object.
2699 Typically this includes layout information.
2703 Description: lu_context fixes.
2704 Details : Introduce new lu_context functions that are needed on the client
2705 side, where some system threads (ptlrpcd) are shared by multiple
2706 modules, and so cannot be stopped during module shutdown.
2710 Description: Add start and stop methods to lu_device_type_operations.
2711 Details : Introduce two new methods in lu_device_type_operations, that are
2712 invoked when first instance of a given type is created and last one
2713 is destroyed respectively. This is need by CLIO.
2717 Description: Add lu_ref support to struct lu_device.
2718 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2719 track leaked references.
2723 Description: Introduce lu_kmem_descr.
2724 Details : lu_kmem_descr and its companion interface allow to create
2725 and destroy a number of kmem caches at once.
2729 Description: Fix lu_object finalization race.
2730 Details : Fix a race between lu_object_find() finding an object and its
2731 concurrent finalization. This race is (most likely) not possible
2732 on the server, but might happen on the client.
2736 Description: Introduce lu_ref interface.
2737 Details : lu_ref is a debugging module allowing to track references to
2738 a given object. It is quite cpu expensive, and has to be
2739 explicitly enabled with --enable-lu_ref. See usage description
2744 Description: Factor lu_site procfs stats into a separate function.
2745 Details : Separate lu_site stats printing code into a separate function
2746 that can be reused on a client.
2750 Description: Constify instances of struct {lu,dt,md}_device_operations.
2751 Details : Constify instances of struct {lu,dt,md}_device_operations.
2755 Description: Introduce struct md_site and move meta-data specific parts of
2756 struct lu_site here.
2757 Details : Move md-specific fields out of struct lu_site into special struct
2758 md_site, so that lu_site can be used on a client.
2762 Description: Kill mdd_lov_destroy().
2763 Details : Remove unused mdd code.
2767 Description: Add st_block checking to multistat.c.
2768 Details : Add st_block checking to multistat.c.
2772 Description: Add lu_ref support to struct obd_device.
2773 Details : Add lu_ref tracking to obd_device.
2777 Description: Kill obd_set_fail_loc().
2778 Details : Remove unused code.
2782 Description: Add special type for ptlrpc_request interpret functions.
2783 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2784 there. Actual usage will be in CLIO.
2788 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2789 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2790 doesn't work with lockdep.
2794 Description: Add rwv.c test program.
2795 Details : New testing program exercising readv(2) and writev(2) (Qian).
2799 Description: Add sendfile.c test program.
2800 Details : New testing program exercising sendfile(2) (Jay).
2804 Description: Ratelimit a message that can be very frequent.
2805 Details : Ratelimit a memory allocation failure message that can
2810 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2811 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2815 Description: Liblustre build fixes.
2816 Details : Liblustre build fixes.
2820 Description: libcfs: add cfs_{need,cond}_resched() interface.
2821 Details : libcfs: add cfs_{need,cond}_resched() definition and
2822 implementations for Linux, NT, and liblustre.
2824 Severity : enhancement
2826 Description: More exported tunables for mballoc
2827 Details : Add support for tunable preallocation window and new tunables for
2828 large/small requests
2832 Description: Detect corruption of block bitmap and checking for preallocations
2833 Details : Checks validity of on-disk block bitmap. Also it does better
2834 checking of number of applied preallocations. When corruption is
2835 found, it turns filesystem readonly to prevent further corruptions.
2839 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2840 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2841 window to 0. If it does hit cache miss, and read-pattern is still
2842 stride-io mode, does not reset the stride window, but also does not
2843 increase the stride window length in this case.
2847 Frequency : only for big-endian servers
2848 Description: Check if system is big-endian while mounting fs with extents feature
2849 Details : Mounting a filesystem with extents feature will fail on big-endian
2850 systems since ext3-based ldiskfs is not supported on big-endian
2851 systems. This can be over-riden with "bigendian_extents" mount option.
2853 Severity : enhancement
2855 Description: The root squash functionality
2856 Details : A security feature, which is to prevent users from being able
2857 to mount lustre on their desktop, run as root, and delete
2858 all of the files in the filesystem. The goal is accomplished by
2859 remapping user id (UID) and group id (GID) of the root user to
2860 a UID and GID specified by the system administartor via Lustre
2861 configuration management server (MGS). The functionality also
2862 allows to specify sets of clients for which the remapping does
2867 Description: Excessive recovery window
2868 Details : With AT enabled, the recovery window can be excessively long (6000+
2869 seconds). To address this problem, we no longer use
2870 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2871 timeout no longer depends on the service time, it is set to
2872 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2873 time via pb_service_time.
2877 Description: Watchdog triggered on MDS failover
2878 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2879 the OSTs know that the MDS "UUID" can be reused for the same export
2880 from a different NID, so we do not need to wait for the export to be
2884 Frequency : rare, only if using MMP with Linux RAID
2886 Description: MMP doesn't work with Linux RAID
2887 Details : While using HA for Lustre servers with Linux RAID, it is possible
2888 that MMP will not detect multiple mounts. To make this work we
2889 need to unplug the device queue in RAID when the MMP block is being
2890 written. Also while reading the MMP block, we should read it from
2891 disk and not the cached one.
2893 Severity : enhancement
2895 Description: open file using fid
2896 Details : A file can be opened using just its fid, like
2897 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2900 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2903 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2904 Details : add proper locking for creation of jbd_4k slab cache
2908 Description: MMP check in ext3_remount() fails without displaying any error
2909 Details : When multiple mount protection fails during remount, proper error
2912 Severity : enhancement
2914 Description: Allow stripe size to be up to 4G-64k
2915 Details : Fix math logic to allow large stripe sizes.
2919 Description: add check for >8TB ldiskfs filesystems
2920 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2921 Don't allow >8TB ldiskfs filesystems to be mounted without
2922 force_over_8tb mount option
2927 Descriptoin: OSS complains "too many blocks" and crashes
2928 Details : In filter_commitrw_write():
2929 1. lquota_chkquota();
2930 2. obdo_from_inode();
2931 3. lquota_pending_commit();
2932 "2." has changed uid and gid in this case, then caused
2933 lquota_chkquota() and lquota_pending_commit() to use
2934 different uid and gid. That is the root of the bug.
2938 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2939 Details : Detect and print out the 'stripe count = -1' case properly.
2943 Description: Out or order replies might be lost on replay
2944 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2945 request with tid smaller than one currently being added, add it
2946 to the start, not end of the list.
2948 --------------------------------------------------------------------------------
2950 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2952 * Support for kernels:
2953 2.6.5-7.283 (SLES 9),
2954 2.6.9-55.EL (RHEL 4),
2955 2.6.16.46-0.14 (SLES 10),
2956 2.6.18.8 vanilla (kernel.org)
2957 * Client support for unpatched kernels:
2958 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2959 2.6.16 - 2.6.22 vanilla (kernel.org)
2960 * Due to recently discovered recovery problems, we do not recommend
2961 using patchless RHEL 4 clients with this or any earlier release.
2962 * Recommended e2fsprogs version: 1.39.cfs8
2963 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2964 * Starting with this release, the ldiskfs backing filesystem required
2965 by Lustre is now in its own package, lustre-ldiskfs. This package
2966 should be installed. It is versioned separately from Lustre and
2967 may be released separately in future.
2971 Description: block reactivating mgc import until all deactivates complete
2972 Details : Fix race when failing back MDT/MGS to itself (testing)
2975 Frequency : at statup only
2977 Description: mds_lov_synchronize race leads to various problems
2978 Details : simultaneous MDT->OST connections at startup can cause the
2979 sync to abort, leaving the OSC in a bad state.
2981 Severity : enhancement
2983 Description: add optional extra BUILD_VERSION info
2984 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2985 to override the lustre version.
2988 Frequency : 2.6.18 servers only
2990 Description: ll_kern_mount() doesn't release the module reference
2991 Details : The ldiskfs module reference count never drops down to 0
2992 because ll_kern_mount() doesn't release the module reference.
2997 Description: server LBUG when using old ost_num_threads parameter
2998 Details : Accept the old ost_num_threads parameter but warn that it
2999 is deprecated, and fix an off-by-one error that caused an LBUG.
3004 Description: Transient SCSI error results in persistent IO issue
3005 Details : iobuf->dr_error is not reinitialized to 0 between two
3009 Frequency : sometimes when underlying device returns I/O errors
3011 Description: OSTs not going read-only during write failures
3012 Details : OSTs are not remounted read-only when the journal commit threads
3013 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3014 instead of the ext3 wrappers.
3018 Description: poor connect scaling with increasing client count
3019 Details : Don't run filter_grant_sanity_check for more than 100 exports
3020 to improve scaling for large numbers of clients.
3023 Frequency : SLES10 only
3025 Description: sanity-quota.sh quotacheck failed: rc = -22
3026 Details : Quotas cannot be enabled on SLES10.
3029 Frequency : liblustre clients only
3031 Description: getdirentries does not give error when run on compute nodes
3032 Details : getdirentries does not fail when the size specified as an argument
3033 is too small to contain at least one entry
3035 Severity : enhancement
3037 Description: Add LNET router traceability for debug purposes
3038 Details : If a checksum failure occurs with a router as part of the
3039 IO path, the NID of the last router that forwarded the bulk data
3040 is printed so it can be identified.
3045 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3046 Details : Due to a race condition, liblustre clients were occasionally
3047 evicted incorrectly.
3049 Severity : enhancement
3051 Description: lfs setstripe use optional parameters instead of postional
3054 Severity : enhancement
3056 Description: Nanosecond timestamp support for ldiskfs
3057 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3058 resolution timestamps. There is not yet support for this at
3059 the Lustre filesystem level.
3062 Frequency : during server recovery
3064 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3065 Details : request with rq_no_resend flag not awake l_wait_event if they get
3069 Frequency : nfs export on patchless client
3071 Description: connectathon hang when test nfs export over patchless client
3072 Details : Disconnected dentry cannot be found with lookup, so we do not need
3073 to unhash it or make it invalid
3076 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3077 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3078 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3083 Description: open req refcounting wrong on reconnect
3084 Details : If reconnect happened between getting open reply from server and
3085 call to mdc_set_replay_data in ll_file_open, we will schedule
3086 replay for unreferenced request that we are about to free.
3087 Subsequent close will crash in variety of ways.
3088 Check that request is still eligible for replay in
3089 mdc_set_replay_data().
3094 Description: disable writes to filesystem when reading health_check file
3095 Details : the default for reading the health_check proc file has changed
3096 to NOT do a journal transaction and write to disk, because this
3097 can cause reads of the /proc file to hang and block HA state
3098 checking on a healthy but otherwise heavily loaded system. It
3099 is possible to return to the previous behaviour during configure
3100 with --enable-health-write.
3102 Severity : enhancement
3104 Description: 64-bit inode version
3105 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3106 to the inode. This will be required for version-based recovery.
3111 Description: MDS fails to start if a duplicate client export is detected
3112 Details : in some rare cases it was possible for a client to connect to
3113 an MDS multiple times. Upon recovery the MDS would detect this
3114 and fail during startup. Handle this more gracefully.
3116 Severity : enhancement
3118 Description: Add -o localflock option to simulate old noflock
3120 Details : This will achieve local-only flock/fcntl locks
3126 Description: log_commit_thread vs filter_destroy race leads to crash
3127 Details : Take import reference before releasing llog record semaphore
3132 Description: Wrong request locking in request set processing
3133 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3134 imp_delayed_list, in this place should be used imp_lock.
3137 Frequency : when reconnection
3139 Description: Grant Leak when osc reconnect to OST
3140 Details : When osc reconnect ost, OST(filter) should check whether it
3141 should grant more space to client by comparing fed_grant and
3142 cl_avail_grant, and return the granted space to client instead
3143 of "new granted" space, because client will call osc_init_grant
3144 to update the client grant space info.
3147 Frequency : when client reconnect to OST
3149 Description: Grant Leak when osc do resend and replay bulk write
3150 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3151 bulk write request, because the grant info while be sync between
3152 OSC and OST when reconnect, and we should ignore the grant info
3153 these of resend/replay write req.
3158 Description: Grant space more than available left space sometimes.
3159 Details : When then OST is about to be full, if two bulk writing from
3160 different clients came to OST. Accord the avaliable space of the
3161 OST, the first req should be permitted, and the second one
3162 should be denied by ENOSPC. But if the seconde arrived before
3163 the first one is committed. The OST might wrongly permit second
3164 writing, which will cause grant space > available space.
3167 Frequency : when client is evicted
3169 Description: Grant might be wrongly erased when osc is evicted by OST
3170 Details : when the import is evicted by server, it will fork another
3171 thread ptlrpc_invalidate_import_thread to invalidate the
3172 import, where the grant will be set to 0. While the original
3173 thread will update the grant it got when connecting. So if
3174 the former happened latter, the grant will be wrongly errased
3175 because of this race.
3180 Description: Checking Stale with correct fid
3181 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3182 to check whether it is stale, because sometimes, we want the
3183 enqueue happened anyway, and op_data.fid2 will not be initialized.
3185 Severity : enhancement
3187 Description: update patchless client
3188 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3191 Frequency : only with 2.4 kernel
3193 Description: random memory corruption
3194 Details : size of struct ll_inode_info is to big for union inode.u and this
3195 can be cause of random memory corruption.
3200 Description: Memory leak in recovery
3201 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3202 It should also check obd_fail before fsfilt_start, otherwise if
3203 fsfilt_start return -EROFS,(failover mds during mds recovery).
3204 then the req will return with repmsg->transno = 0 and rc = EROFS.
3205 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3206 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3207 be freed no matter whether fsfilt_commit success or not.
3210 Frequency : only with huge count clients
3212 Description: Prevents from taking the superblock lock in llap_from_page for
3214 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3215 for llap_from_page call in ll_removepage prevents from taking the
3216 superblock lock for a soon died page.
3221 Description: Not check open intent error before release open handle
3222 Details : in some rare cases, the open intent error is not checked before
3223 release open handle, which may cause
3224 ASSERTION(open_req->rq_transno != 0), because it tries to release
3225 the failed open handle.
3230 Description: Set cat log bitmap only after create log success.
3231 Details : in some rare cases, the cat log bitmap is set too early. and it
3232 should be set only after create log success.
3236 Description: Accessing a block bevice can re-enable I/O when Lustre is
3237 tearing down a device.
3238 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3243 Description: service threads may hog cpus when there are a lot of requests
3244 Details : Insert cond_resched to give other threads a chance to use some CPU
3249 Description: the cat log was not initialized in recovery
3250 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3251 unlink log on mds will not be initialized until mds post
3252 recovery. And also in mds post recovery, the unlink log will
3253 initialization will be done asynchronausly, so there will be race
3254 between add unlink log and unlink log initialization.
3258 Description: brw_stats were being printed incorrectly
3259 Details : brw_stats were being printed as log2 but all of them were not
3260 recorded as log2. Also remove some code duplication arising from
3261 filter_tally_{read,write}.
3265 Frequency : rare, only in recovery.
3266 Description: ASSERTION(req->rq_type != LI_POISON) failed
3267 Details : imp_lock should be held while iterating over imp_sending_list for
3268 prevent destroy request after get timeout in ptlrpc_queue_wait.
3272 Description: replay-single.sh test 52 fails
3273 Details : A lock's skiplist need to be cleanup when it being unlinked
3274 from its resource list.
3278 Description: Short directio read returns full requested size rather than
3280 Details : Direct I/O operations should return actual amount of bytes
3281 transferred rather than requested size.
3283 Severity : enhancement
3285 Description: metadata RPC reduction (e.g. for rm performance)
3286 Details : decrease the amount of synchronous RPC between clients and servers
3287 by canceling conflicing lock before the operation on the client
3288 and packing thier handles into the main operation RPC to server.
3290 Severity : enhancement
3292 Description: Async OSC create to avoid the blocking unnecessarily.
3293 Details : If an OST has no remaining object, system will block on the
3294 creation when it needs to create a new object on this OST. Now,
3295 ways use pre-created objects when available, instead of blocking on
3296 an empty osc while others are not empty. If we must block, we
3297 block for the shortest possible period of time.
3301 Description: improve handling recoverable errors
3302 Details : if request processig with error which can be recoverable on server
3303 request should be resend, otherwise page released from cache and
3306 Severity : enhancement
3308 Description: refine locking for avoid write wrong info into lov_objid file
3309 Details : fix possible races with add new target and write/update data in
3312 --------------------------------------------------------------------------------
3314 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3319 Frequency : on some architectures
3321 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3322 Details : Uninitialized flags sometimes cause configuration commands to
3326 Frequency : patchless clients only
3328 Description: missing __iget() symbol export
3329 Details : The __iget() symbol export is missing. To avoid the need for
3330 this on patchless clients the deathrow inode reaper is turned
3331 off, and we depend on the VM to clean up old inodes. This
3332 dependency was during via the fix for bug 12181.
3334 --------------------------------------------------------------------------------
3336 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3338 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3339 INTEROPERATE with older versions automatically. In many cases a
3340 special upgrade step is needed. Please read the
3341 user documentation before upgrading any part of a live system.
3342 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3343 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3344 * WARNING: Lustre configuration and startup changes are required with
3345 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3349 Severity : enhancement
3351 Description: MountConf
3352 Details : Lustre configuration is now managed via mkfs and mount
3353 commands instead of lmc and lconf. New obd types (MGS, MGC)
3354 are added for dynamic configuration management. See
3355 https://mail.clusterfs.com/wikis/lustre/MountConf for
3358 Severity : enhancement
3360 Description: dynamic OST addition
3361 Details : OSTs can now be added to a live filesystem
3363 Severity : enhancement
3365 Description: startup order invariance
3366 Details : MDTs and OSTs can be started in any order. Clients only
3367 require the MDT to complete startup.
3369 Severity : enhancement
3371 Description: parallel, asynchronous orphan cleanup
3372 Details : orphan cleanup is now performed in separate threads for each
3373 OST, allowing parallel non-blocking operation.
3375 Severity : enhancement
3377 Description: optimized stripe assignment
3378 Details : stripe assignments are now made based on ost space available,
3379 ost previous usage, and OSS previous usage, in order to try
3380 to optimize storage space and networking resources.
3382 Severity : enhancement
3384 Description: Permanently set tunables
3385 Details : All writable /proc/fs/lustre tunables can now be permanently
3386 set on a per-server basis, at mkfs time or on a live system.
3388 Severity : enhancement
3390 Description: Lustre message v2
3391 Details : Add lustre message format v2.
3393 Severity : enhancement
3395 Description: client OST exclusion list
3396 Details : Clients can be started with a list of OSTs that should be
3397 declared "inactive" for known non-responsive OSTs.
3401 Description: SPEC SFS validation failure on NFS v2 over lustre.
3402 Details : Changes the blocksize for regular files to be 2x RPC size,
3403 and not depend on stripe size.
3405 Severity : enhancement
3407 Description: Multiple MD RPCs in flight.
3408 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3409 To avoid overly-overloading MDS, introduce a limit on number of
3410 MDS RPCs in flight for a single client and add /proc controls
3411 to adjust this limit.
3413 Severity : enhancement
3415 Description: client read/write statistics
3416 Details : Add client read/write call usage stats for performance
3417 analysis of user processes.
3418 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3419 file access. extents_stats shows chunk size distribution.
3420 extents_stats_per_process show chunk size distribution per
3423 Severity : enhancement
3425 Description: mds statistics
3426 Details : Add detailed mds operations statistics in
3427 /proc/fs/lustre/mds/*/stats.
3431 Description: Failure of copying files with lustre special EAs.
3432 Details : Client side always return success for setxattr call for lustre
3433 special xattr (currently only "trusted.lov").
3438 Description: Refcount LNET uuids
3439 Details : The global LNET uuid list grew linearly with every startup;
3440 refcount repeated list entries instead of always adding to
3443 Severity : enhancement
3445 Description: Dynamic service threads
3446 Details : Within a small range, start extra service threads
3447 automatically when the request queue builds up.
3450 Frequency : mixed-endian client/server environments
3452 Description: mixed-endian crashes
3453 Details : The new msg_v2 system had some failures in mixed-endian
3456 Severity : enhancement
3458 Description: Easy OST removal
3459 Details : OSTs can be permanently deactivated with e.g. 'lctl
3460 conf_param lustre-OST0001.osc.active=0'
3462 Severity : enhancement
3464 Description: MGS proc entries
3465 Details : Added basic proc entries for the MGS showing what filesystems
3468 Severity : enhancement
3470 Description: provide MGS failover
3471 Details : Added config lock reacquisition after MGS server failover.
3473 Severity : enhancement
3475 Description: add Linux 2.4 support
3476 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3480 Description: a large application tries to do I/O to the same resource and dies
3481 in the middle of it.
3482 Details : Check the req->rq_arrival time after the call to
3483 ost_brw_lock_get(), but before we do anything about
3484 processing it & sending the BULK transfer request. This
3485 should help move old stale pending locks off the queue as
3486 quickly as obd_timeout.
3489 Frequency : when an incorrect nid is specified during startup
3491 Description: ptlrpc connect to non-existant node causes kernel crash
3492 Details : LNET can't be re-entered from an event callback, which
3493 happened when we expire a message after the export has been
3494 cleaned up. Instead, hand the zombie cleanup off to another
3497 Severity : enhancement
3499 Description: plain/inodebits lock performance improvement
3500 Details : Grouping plain/inodebits in granted list by their request modes
3501 and bits policy, thus improving the performance of search through
3505 Frequency : only if OST filesystem is corrupted
3507 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3508 Details : for a short time RPCs with bulk IO are in the replay list,
3509 but replay of bulk IOs is unimplemented. If the OST filesystem
3510 is corrupted due to disk cache incoherency and then replay is
3511 started it is possible to trip an assertion. Avoid putting
3512 committed RPCs into the replay list at all to avoid this issue.
3515 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3517 Description: System hang on startup
3518 Details : This bug allowed the liblustre (e.g. catamount) client to
3519 return to the app before handling all startup RPCs. This
3520 could leave the node unresponsive to lustre network traffic
3521 and manifested as a server ptllnd timeout.
3523 Severity : enhancement
3525 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3526 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3527 boolean to control whether to print peer debug info when a
3528 client's RPC times out.
3531 Frequency : only for kernels with patches from Lustre below 1.4.3
3533 Description: Remove old rdonly API
3534 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3537 Frequency : only for devices with external journals
3539 Description: Set external device read-only also
3540 Details : During a commanded failover stop, we set the disk device
3541 read-only while the server shuts down. We now also set any
3542 external journal device read-only at the same time.
3545 Frequency : when upgrading from 1.4 while trying to change parameters
3547 Description: The wrong (new) MDC name was used when setting parameters for
3548 upgraded MDT's. Also allows changing of OSC (and MDC)
3549 parameters if --writeconf is specified at tunefs upgrade time.
3552 Frequency : when setting specific ost indicies
3554 Description: QOS code breaks on skipped indicies
3555 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3556 created with --index need not be sequential.
3558 Severity : enhancement
3560 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3561 Details : The uninit_groups feature works in conjunction with the kernel
3562 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3563 up the pass1 processing of e2fsck. This is a read-only feature
3564 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3565 mount filesystems that have had this feature enabled.
3567 Severity : enhancement
3569 Description: Improve multi-block allocation algorithm to avoid fragmentation
3570 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3571 allocation locality and avoid filesystem fragmentation.
3573 ------------------------------------------------------------------------------
3575 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3577 * Support for kernels:
3578 2.4.21-47.0.1.EL (RHEL 3)
3579 2.6.5-7.283 (SLES 9)
3580 2.6.9-42.0.10.EL (RHEL 4)
3581 2.6.12.6 vanilla (kernel.org)
3582 2.6.16.27-0.9 (SLES 10)
3583 * Recommended e2fsprogs version: 1.39.cfs5
3585 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3589 Frequency : occasional, depends on client load and configuration
3590 Bugzilla : 12181, 12203
3591 Description: data loss for recently-modified files
3593 Details : In some cases it is possible that recently written or created
3594 files may not be written to disk in a timely manner (this should
3595 normally be within 30s unless client IO load is very high).
3596 The problem appears as zero-length files or files that are a
3597 multiple of 1MB in size after a client crash or client eviction
3598 that are missing data at the end of the file.
3600 This problem is more likely to be hit on clients where files are
3601 repeatedly created and unlinked in the same directory, clients
3602 have a large amount of RAM, have many CPUs, the filesystem has
3603 many OSTs, the clients are rebooted frequently, and/or the files
3604 are not accessed by other nodes after being written.
3606 The presence of the problem can be detected by looking at
3607 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3608 smaller than the second (nr_unused) then dirty files will not
3609 be flushed automatically to disk. "sync; sleep 10" should be
3610 run several times on the node before unmounting it to update
3611 Lustre (this is also safe to run on nodes without this problem).
3613 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3614 that can cause this same problem, so customers using that kernel
3615 also need to update the kernel in addition to Lustre. In order
3616 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3618 It is normal that files written just before a client crash (less
3619 than 30s) may not yet have been flushed to disk, even for local
3623 Frequency : frequent on thin XT3 nodes
3625 Description: UUID collision on thin XT3 Linux nodes
3626 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3627 randomly, since we relied on an insufficiently-seeded PRNG.
3632 Description: OSS hangs after "All ost request buffers busy"
3633 Details : A deadlock between quota and journal operations caused OSS
3634 hangs after printing "All ost request buffers busy."
3637 Frequency : always on liblustre builds
3639 Description: Cleanup compiler warnings on liblustre
3642 Frequency : always on liblustre builds on XT3
3644 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3648 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3649 > 32000 subdirectories
3650 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3651 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3652 subdirectory count crosses 32000. This will aid e2fsck to
3653 correctly handle more than 32000 subdirectories.
3656 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3658 Description: System hang on startup
3659 Details : This bug allowed the liblustre (e.g. catamount) client to
3660 return to the app before handling all startup RPCs. This
3661 could leave the node unresponsive to lustre network traffic
3662 and manifested as a server ptllnd timeout.
3664 Severity : enhancement
3666 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3667 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3668 boolean to control whether to print peer debug info when a
3669 client's RPC times out.
3674 Description: make O_SYNC working on 2.6 kernels
3675 Details : 2.6 kernels use different method for mark pages for write,
3676 so need add a code to lustre for O_SYNC work.
3681 Description: Failure to close file and release space on NFS
3682 Details : Put inode details into lock acquired in ll_intent_file_open.
3683 Use mdc_intent_lock in ll_intent_open to properly
3684 detect all kind of errors unhandled by mdc_enqueue.
3689 Description: proc file read during shutdown sometimes raced obd removal,
3691 Details : Add lock to prevent obd access after proc file removal.
3694 Frequency : Only for files larger than 4GB on 32-bit clients.
3696 Description: improperly doing page alignment of locks
3697 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3698 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3703 Description: RPCs being resent when they shouldn't be
3704 Details : Some RPCs that should not be resent are being resent. This
3705 can cause inconsistencies in the RPC state machine. Do not
3706 resend such requests.
3709 Frequency : rare, only with NFS export
3711 Description: Crash on NFS re-export node
3712 Details : under very unusual load conditions an assertion is hit in
3713 ll_intent_file_open()
3716 Frequency : only if OST filesystem is corrupted
3718 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3719 Details : for a short time RPCs with bulk IO are in the replay list,
3720 but replay of bulk IOs is unimplemented. If the OST filesystem
3721 is corrupted due to disk cache incoherency and then replay is
3722 started it is possible to trip an assertion. Avoid putting
3723 committed RPCs into the replay list at all to avoid this issue.
3728 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3729 Details : Large single O_DIRECT read and write calls can fail to allocate
3730 a sufficiently large buffer to process the request. In case of
3731 allocation failure the allocation is retried with a smaller
3732 buffer and broken into smaller requests.
3734 Severity : enhancement
3736 Description: Add -o localflock option to simulate old noflock behaviour.
3737 Details : This will achieve local-only flock/fcntl locks coherentness.
3742 Description: versioning check is incomplete
3743 Details : Checking the version difference of client vs. server, report
3744 error if the gap is too big.
3749 Description: add support PG_writeback bit
3750 Details : add support for PG_writeback bit for Lustre, for more carefull
3751 work with page cache in 2.6 kernel. This also fix some deadlocks
3752 and remove hack for work O_SYNC with 2.6 kernel.
3754 Severity : enhancement
3756 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3757 Details : The uninit_groups feature works in conjunction with the kernel
3758 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3759 up the pass1 processing of e2fsck. This is a read-only feature
3760 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3761 mount filesystems that have had this feature enabled.
3763 Severity : enhancement
3765 Description: Improve multi-block allocation algorithm to avoid fragmentation
3766 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3767 allocation locality and avoid filesystem fragmentation.
3769 ------------------------------------------------------------------------------
3771 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3773 * Support for kernels:
3774 2.6.9-42.0.3.EL (RHEL 4)
3775 2.6.5-7.276 (SLES 9)
3776 2.4.21-47.0.1.EL (RHEL 3)
3777 2.6.12.6 vanilla (kernel.org)
3778 2.6.16.21-0.8 (SLES10)
3779 * Recommended e2fsprogs version: 1.39.cfs2-0
3781 * The backwards-compatible /proc/sys/portals symlink has been removed
3782 in this release. Before upgrading, please ensure that you change
3783 any configuration scripts or /etc/sysctl.conf files that access
3784 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3785 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3786 in advance of the upgrade on any system running Lustre 1.4.6 or
3787 newer, since /proc/sys/lnet was added in that version.
3788 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3792 Frequency : only when quota is used
3794 Description: avoid scanning export list for quota master
3795 Details : Change the algorithms to avoid scanning export list in order
3796 to improve the efficiency.
3799 Frequency : MDS failover only, very rarely
3801 Description: "went back in time" messages on mds failover
3802 Details : The greatest transno may be lost when the current operation
3803 finishes with an error (transno==0) and the client's last_rcvd
3804 record is over-written. Save the greatest transno in the
3805 mds_last_transno for this case.
3808 Frequency : always for specific kernels and striping counts
3810 Description: client may get "Matching packet too big" without ACL support
3811 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3812 when trying to access files in certain configurations. The
3813 clients should in fact be denied when mounting because they do
3814 not understand ACLs.
3817 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3819 Description: many clients connecting with IO in progress causes connect timeouts
3820 Details : Avoid synchronous journal commits to avoid delays caused by many
3821 clients connecting/disconnecting when bulk IO is in progress.
3822 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3823 OST_IO_PORTAL to avoid delays behind potentially many pending
3827 Frequency : occasionally with multiple writers to a single file
3829 Description: shared writes to file may result in wrong size reported by stat()
3830 Details : Allow growing of kms when extent lock is cancelled
3833 Frequency : always with random mmap IO to multi-striped file
3835 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3836 Details : Only if the hole is at the end of OST object so that kms is too
3837 small. Fix is to increase kms accordingly in ll_nopage.
3840 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3842 Description: writes to a missing object would leak memory on the OST
3843 Details : If there is an inconsistency between the MDS and OST filesystems,
3844 such that the MDS references an object that doesn't exist, writes
3845 to that object will leak memory due to incorrect cleanup in the
3846 error handling path, eventually running out of memory on the OST.
3851 Description: Creating too long symlink causes lustre errors
3852 Details : Check symlink and name lengths before sending requests to MDS.
3855 Frequency : only if flock is enabled (not on by default)
3857 Description: posix locks not released on fd closure on 2.6.9+
3858 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3859 kernels, this caused such locks not to be released on fd close and
3860 then assertions on fs unmount about still used locks.
3863 Frequency : MDS failover only, very rarely
3865 Description: clients may get ASSERTION(granted_lock != NULL)
3866 Details : When request was taking a long time, and a client was resending
3867 a getattr by name lock request. The were multiple lock requests
3868 with the same client lock handle and
3869 mds_getattr_name->fixup_handle_for_resent_request found one of the
3870 lock handles but later failed with ASSERTION(granted_lock != NULL).
3875 Description: handle->h_buffer_credits > 0, assertion failure
3876 Details : h_buffer_credits is zero after truncate, causing assertion
3877 failure. This patch extends the transaction or creates a new
3881 Frequency : NFS re-export or patchless client
3882 Bugzilla : 11179, 10796
3883 Description: Crash on NFS re-export node (__d_move)
3884 Details : We do not want to hash the dentry if we don't have a lock.
3885 But if this dentry is later used in d_move, we'd hit uninitialised
3886 list head d_hash, so we just do this to init d_hash field but
3887 leave dentry unhashed.
3890 Frequency : NFS re-export or patchless client
3892 Description: NFS exports has problem with symbolic link
3893 Details : lustre client didn't properly install dentry when re-exported
3894 to NFS or running patchless client.
3897 Frequency : NFS re-export or patchless client
3899 Description: Various nfs/patchless fixes.
3900 Details : fixes reuse disconected alias for lookup process - this fixes
3901 warning "find_exported_dentry: npd != pd",
3902 fix permission error with open files at nfs.
3903 fix apply umask when do revalidate.
3906 Frequency : occasional
3908 Description: Crash on NFS re-export node
3909 Details : calling clear_page() on the wrong pointer triggered oops in
3910 generic_mapping_read().
3913 Frequency : rarely, using O_DIRECT IO
3915 Description: unaligned directio crashes client with LASSERT
3916 Details : check for unaligned buffers before trying any requests.
3919 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3921 Description: stale data returned from RAID cache
3922 Details : If only a small amount of IO is done to the RAID device before
3923 reading it again it is possible to get stale data from the RAID
3924 cache instead of reading it from disk.
3927 Frequency : always for sles10 kernel
3929 Description: sles10 support
3930 Details : ll_follow_link: compile fixes and using of nd_set_link
3931 under newer kernels.
3934 Frequency : depends on arch, kernel and compiler version, always on sles10
3937 Description: recursive or deep enough symlinks cause stack overflow
3938 Details : getting rid of large stack-allocated variable in
3942 Frequency : depends on hardware
3944 Description: lustre write performance loss in the SLES10 kernel
3945 Details : the performance loss is caused by using of write barriers in the
3946 ext3 code. The SLES10 kernel turns barrier support on by
3947 default. The fix is to undo that change for ldiskfs.
3949 ------------------------------------------------------------------------------
3951 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3953 * Support for kernels:
3954 2.6.9-42.0.3EL (RHEL 4)
3955 2.6.5-7.276 (SLES 9)
3956 2.4.21-47.0.1.EL (RHEL 3)
3957 2.6.12.6 vanilla (kernel.org)
3961 Frequency : quota enabled and large files being deleted
3963 Description: releasing more than 4GB of quota at once hangs OST
3964 Details : If a user deletes more than 4GB of files on a single OST it
3965 will cause the OST to spin in an infinite loop. Release
3966 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3971 Description: statfs data retrieved from /proc may be stale or zero
3972 Details : When reading per-device statfs data from /proc, in the
3973 {kbytes,files}_{total,free,avail} files, it may appear
3974 as zero or be out of date.
3977 Frequency : systems with MD RAID1 external journal devices
3979 Description: lconf's call to blkid is confused by RAID1 journal devices
3980 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3981 of returning all block devices that match the journal UUID.
3984 Frequency : always, for aggregate stripe size over 4GB
3986 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3987 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3988 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3991 Frequency : always if "lfs find" used on a local file/directory
3993 Description: "lfs find" segfaults if used on a local file/directory
3994 Details : The case where a directory component was not specified wasn't
3995 handled correctly. Handle this properly.
3998 Frequency : always on ppc64
4000 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4001 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4002 perform endianness conversion.
4005 Frequency : rarely (truncate to non-zero file size after write under load)
4006 Bugzilla : 10730, 10687
4007 Description: Files padded with zeros to next 4K multiple
4008 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4009 it is possible that files that are truncated to a non-zero size
4010 immediately after being written are filled with zero bytes beyond
4011 the truncated size. No file data is lost.
4013 Severity : enhancement
4015 Description: Allow recovery/failover for liblustre clients.
4016 Details : liblustre clients were unaware of failover configurations until
4019 Severity : enhancement
4021 Description: user file locks should fail when not mounting with flock option
4022 Details : Set up an error-returning stub in ll_file_operations.lock field
4023 to prevent incorrect behaviour when client is mounted without
4024 flock option. Also, set up properly f_op->flock field for
4028 Frequency : always on ia64
4030 Description: "lfs df" loops on printing out MDS statfs information
4031 Details : The obd_ioctl_data was not initialized and in some systems
4032 this caused a failure during the ioctl that did not return
4033 an error. Initialize the struct and return an error on failure.
4036 Frequency : SLES 9 only
4038 Description: Error of copying files with lustre special EAs as root
4039 Details : Client side always return success for setxattr call for lustre
4040 special xattr (currently only "trusted.lov").
4043 Frequency : rarely on clusters with both ia64+i386 clients
4045 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4046 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4047 clients are doing concurrent writes to the same file it is
4048 possible that the ia64 clients may LASSERT because the OST
4049 extent locks are not PAGE_SIZE aligned. Ensure that grown
4050 locks are always aligned on the request boundary.
4053 Frequency : specific use, occasional
4055 Description: Overwriting in use executable truncates on-disk binary image
4056 Details : If one node attempts to overwrite an executable in use by
4057 another node, we now correctly return ETXTBSY instead of
4058 truncating the file.
4060 Severity : enhancement
4062 Description: Async OSC create to avoid the blocking unnecessarily.
4063 Details : If an OST has no remaining object, system will block on the creating
4064 when need to create a new object on this OST. Now, ways use
4065 pre-created objects when available, instead of blocking on an
4066 empty osc while others are not empty. If we must block, we block
4067 for the shortest possible period of time.
4072 Description: chmod on Lustre root is propagated to other clients
4073 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4074 invalid by the follow_mount time.
4077 Frequency : liblustre clients only
4079 Description: Race in 'instant cancel' lock handling could lead to such locks
4080 never to be granted in case of SMP MDS
4081 Details : Do not destroy not yet granted but cbpending locks in
4085 Frequency : replay/resend of open
4087 Description: non null lock assetion failure in mds_intent_policy
4088 Details : Trying to replay/resend lockless open requests resulted in
4089 mds_open() returning 0 with no lock. Now it sets a flag if
4090 a lock is going to be returned.
4092 Severity : enhancement
4094 Description: Checksum enhancements
4095 Details : New checksum enhancements allow for resending RPCs that failed
4098 Severity : enhancement
4100 Description: Tunables on number of dirty pages in cacche
4101 Details : Allow to set limit on number of dirty pages cached.
4106 Description: client crash on unmount - lock still has references
4107 Details : In some error handling cases it was possible to leak a lock
4108 reference on a client while accessing a file. This was not
4109 harmful to the client during operation, but would cause the
4110 client to crash when the filesystem is unmounted.
4113 Frequency : specific case, rare
4115 Description: ETXTBSY on mds though file not in use by client
4116 Details : ETXTBSY is no longer incorrectly returned when attempting to
4117 chmod or chown a directory that the user previously tried to
4118 execute or a currently-executing binary.
4121 Frequency : extremely rare except on liblustre-based clients
4123 Description: Lustre space not freed when files are deleted
4124 Details : Clean up open-unlinked files after client eviction. Previously
4125 the unlink was skipped and the files remained as orphans.
4130 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4131 Details : In some cases it was possible to send a blocking callback to
4132 a client doing a glimpse, even though that client didn't get
4133 a lock granted. When the glimpse lock is cancelled on the OST
4134 the freed lock is left on the waiting list and corrupted the list.
4137 Frequency : all core dumps
4139 Description: Broke core dumps to lustre
4140 Details : Negative dentry may be unhashed if parent does not have UPDATE
4141 lock, but some callers, e.g. do_coredump, expect dentry to be
4142 hashed after successful create, hash it in ll_create_it.
4144 ------------------------------------------------------------------------------
4146 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4148 * Support for kernels:
4149 2.6.9-42.0.2.EL (RHEL 4)
4150 2.6.5-7.276 (SLES 9)
4151 2.4.21-40.EL (RHEL 3)
4152 2.6.12.6 vanilla (kernel.org)
4156 Frequency : always on RHEL 3
4158 Description: Number of open files grows over time
4159 Details : The number of open files grows over time, whether or not
4160 Lustre is started. This was due to a filp leak introduced
4161 by one of our kernel patches.
4163 ------------------------------------------------------------------------------
4165 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4167 * Support for kernels:
4168 2.6.9-42.EL (RHEL 4)
4169 2.6.5-7.276 (SLES 9)
4170 2.4.21-40.EL (RHEL 3)
4171 2.6.12.6 vanilla (kernel.org)
4176 Bugzilla : 5719, 9635, 9792, 9684
4177 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4178 Details : If a server is under heavy load and cannot reply to new
4179 connection requests before the client resends the (re)connect,
4180 the connection handling code can behave badly if two service
4181 threads are concurrently handing separate (re)connections from
4182 the same client. Add better locking to the connection handling
4183 code, and ensure that only a single connection will be processed
4184 for a given client UUID, even if the lock is dropped.
4186 Severity : enhancement
4188 Description: add TCP zero-copy support to kernel
4189 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4190 sends if the hardware supports scatter-gather and checksumming.
4191 This allows socklnd to do client-write and server-read more
4192 efficiently and reduce CPU utilization from skbuf copying.
4195 Frequency : only if NFS exporting from client
4197 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4198 Details : If setting attributes on a file created under NFS that had
4199 never been opened it would be possible to oops the client
4200 if the file had no objects.
4203 Frequency : always for liblustre
4205 Description: liblustre client does MDS+OSTs setattr RPC for each write
4206 Details : When doing a write from a liblustre client, the client
4207 incorrectly issued an RPC to the MDS and each OST the file was
4208 striped over in order to update the timestamps. When writing
4209 with small chunks and many clients this could overwhelm the MDS
4210 with RPCs. In all cases it would slow down the write because
4211 these RPCs are unnecessary.
4213 Severity : enhancement
4215 Description: allow number of MDS service threads to be changed at module load
4216 Details : It is now possible to change the number of MDS service threads
4217 running. Adding "options mds mds_num_threads={N}" to the MDS's
4218 /etc/modprobe.conf will set the number of threads for the next
4219 time Lustre is restarted (assuming the "mds" module is also
4220 reloaded at that time). The default number of threads will
4221 stay the same, 32 for most systems.
4226 Description: OST crash if filesystem is unformatted or corrupt
4227 Details : If an OST is started on a device that has never been formatted
4228 or if the filesystem is corrupt and cannot even mount then the
4229 error handling cleanup routines would dereference a NULL pointer.
4234 Description: NULL pointer deref in llap_from_page.
4235 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4236 incorrect) mapping if the page was truncated/reclaimed while it was
4237 searched for. Check for this condition and skip such pages when
4238 doing readahead. Introduce extra check to llap_from_page() to
4239 verify page->mapping->host is non-NULL (so page is not anonymous).
4242 Frequency : Sometimes when using sys_sendfile
4244 Description: "page not covered by a lock" warnings from ll_readpage
4245 Details : sendfile called ll_readpage without right page locks present.
4246 Now we introduced ll_file_sendfile that does necessary locking
4247 around call to generic_file_sendfile() much like we do in
4251 Frequency : with certain MDS communication failures at client mount time
4253 Description: NULL pointer deref after failed client mount
4254 Details : a client connection request may delayed by the network layer
4255 and not be sent until after the PTLRPC layer has timed out the
4256 request. If the client fails the mount immediately it will try
4257 to clean up before the network times out the request. Add a
4258 reference from the request import to the obd device and delay
4259 the cleanup until the network drops the request.
4262 Frequency : occasionally during client (re)connect
4264 Description: assertion failure during client (re)connect
4265 Details : processing a client connection request may be delayed by the
4266 client or server longer than the client connect timeout. This
4267 causes the client to resend the connection request. If the
4268 original connection request is replied in this interval, the
4269 client may trip an assertion failure in ptlrpc_connect_interpret()
4270 which thought it would be the only running connect process.
4273 Frequency : only with obd_echo servers and clients that are rebooted
4275 Description: kernel BUG accessing uninitialized data structure
4276 Details : When running an obd_echo server it did not start the ping_evictor
4277 thread, and when a client was evicted an uninitialized data
4278 structure was accessed. Start the ping_evictor in the RPC
4279 service startup instead of the OBD startup.
4281 Severity : enhancement
4282 Bugzilla : 10193 (patchless)
4283 Description: Remove dependency on various unexported kernel interfaces.
4284 Details : No longer need reparent_to_init, exit_mm, exit_files,
4285 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4288 Frequency : rare (only users of deprecated and unsupported LDAP config)
4290 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4291 Details : LDAP apparently contains 'lustreName' attributes instead of
4292 'name'. A simple remapping of the name is sufficient.
4295 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4297 Description: waiting_locks_callback trips kernel BUG if client is evicted
4298 Details : Running with the dump_on_timeout debug flag turned on makes
4299 it possible that the waiting_locks_callback() can try to dump
4300 the Lustre kernel debug logs from an interrupt handler. Defer
4301 this log dumping to the expired_lock_main() thread.
4303 Severity : enhancement
4305 Description: Support NFS exporting on 2.6 kernels.
4306 Details : Implement non-rawops metadata methods for NFS server to use without
4307 changing NFS server code.
4310 Frequency : very rare (synthetic metadata workload only)
4312 Description: two racing renames might cause an MDS thread to deadlock
4313 Details : Running the "racer" program may cause one MDS thread to rename
4314 a file from being the source of a rename to being the target of
4315 a rename at exactly the same time that another thread is doing
4316 so, and the second thread has already enqueued these locks after
4317 doing a lookup of the target and is trying to relock them in
4318 order. Ensure that we don't try to re-lock the same resource.
4321 Frequency : only very large systems with liblustre clients
4323 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4324 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4325 all OSTs in parallel. This allows the eviction of stale liblustre
4326 clients to proceed much faster than if they were done in series,
4327 and also offers similar improvements for other set_info RPCs.
4332 Description: excessive CPU usage during initial read phase on client
4333 Details : During the initial read phase on a client, it would agressively
4334 retry readahead on the file, consuming too much CPU and impacting
4335 performance (since 1.4.5.8). Improve the readahead algorithm
4336 to avoid this, and also improve some other common cases (read
4337 of small files in particular, where "small" is files smaller than
4338 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4343 Description: MDS crash when receiving packet with unknown intent.
4344 Details : Do not LBUG in unknown intent case, just return -EFAULT
4346 Severity : enhancement
4347 Bugzilla : 9293, 9385
4348 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4349 Details : Do not serialize getattr (non-intent version) and statfs.
4352 Frequency : occasional, when OST network is overloaded/intermittent
4354 Description: client evicted by OST after bulk IO timeout
4355 Details : If a client sends a bulk IO request (read or write) the OST
4356 may evict the client if it is unresposive to its data GET/PUT
4357 request. This is incorrect if the network is overloaded (takes
4358 too long to transfer the RPC data) or dropped the OST GET/PUT
4359 request. There is no need to evict the client at all, since
4360 the pinger and/or lock callbacks will handle this, and the
4361 client can restart the bulk request.
4364 Frequency : Always when mmapping file with no objects
4366 Description: client crashes when mmapping file with no objects
4367 Details : Check that we actually have objects in a file before doing any
4368 operations on objects in ll_vm_open, ll_vm_close and
4374 Description: Request leak when working with deleted CWD
4375 Details : Introduce advanced request refcount tracking for requests
4376 referenced from lustre intent.
4378 Severity : Enhancement
4380 Description: Cache open file handles on client.
4381 Details : MDS now will return special lock along with openhandle, if
4382 requested and client is allowed to hold openhandle, even if unused,
4383 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4384 closing files for every read/write openration.
4386 Severity : Enhancement
4388 Description: Cache open negative dentries on client when possible.
4389 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4390 negative dentries on lock revocation.
4395 Description: Remounting a client read-only wasn't possible with a zconf mount
4396 Details : It wasn't possible to remount a client read-only with llmount.
4398 Severity : enhancement
4399 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4400 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4401 MPICH is included to add Lustre-specific ADIO interfaces.
4402 This is based closely on the UFS ADIO layer and only differs
4403 in file creation, in order to allow the OST striping to be set.
4404 This is user-contributed code and not supported by CFS.
4409 Description: extended inode attributes (immutable, append-only) work improperly
4410 when 2.4 and 2.6 kernels are used on client/server or vice versa
4411 Details : Introduce kernel-independent values for these flags.
4413 Severity : enhancement
4416 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4417 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4418 granularity. Now, Fractional MB granularity is be supported,
4419 this is very useful for low memory system.
4421 Severity : enhancement
4423 Description: Getattr by fid
4424 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4425 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4426 have locks and avoids getattr rpc requests for it on every stat.
4429 Frequency : Always, for filesystems larger than 2TB
4431 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4432 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4433 filesystems larger than 2TB to be mounted with mballoc enabled.
4436 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4438 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4439 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4440 filesystems larger than 2TB is mounted then the kernel will
4441 silently corrupt the start of the filesystem. CONFIG_LBD is
4442 enabled for all CFS-supported kernels, but the possibility of
4443 this happening with a modified kernel config exists.
4445 Severity : enhancement
4447 Description: add client O_DIRECT support for 2.6 kernels
4448 Details : It is now possible to do O_DIRECT reads and writes to files
4449 in the Lustre client mountpoint on 2.6 kernel clients.
4451 Severity : enhancement
4453 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4454 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4455 parallel, not waiting for response from every OST before sending
4456 a rpc to the next OST.
4461 Description: setattr vs write race when updating file timestamps
4462 Details : Client processes that update a file timestamp into the past
4463 right after writing to the file (e.g. tar) it is possible that
4464 the updated file modification time can be reset to the current
4465 time due to a race between processing the setattr and write RPC.
4467 Severity : enhancement
4469 Description: Bring 'lfs find' closer in line with regular Linux find.
4470 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4471 -print0 options and obtains all the needed info through the lustre
4474 Severity : enhancement
4476 Description: support up to 1024 configured devices on one node
4477 Details : change obd_dev array from statically allocated to dynamically
4478 allocated structs as they are first used to reduce memory usage
4483 Description: Flush dirty partially truncated pages during truncate
4484 Details : Immediatelly flush partially truncated pages in filter_setattr,
4485 this way we completely avoid having any pages in page cache on OST
4486 and can retire ugly workarounds during writes to flush such pages.
4491 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4492 Details : protect inode from truncation within vfs_unlink() context
4493 just take a reference before calling vfs_unlink() and release it
4494 when parent's i_sem is free.
4499 Description: last_id value checked outside lock on OST caused LASSERT failure
4500 Details : If there were multiple MDS->OST object precreate requests in
4501 flight, it was possible that the OST's last object id was checked
4502 outside a lock and incorrectly tripped an assertion. Move checks
4503 inside locks, and discard old precreate requests.
4506 Frequency : always, if extents are used on OSTs
4508 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4509 Details : OSTs using the extents format would not zero the high 16 bits of
4510 the index physical block number. This is not a problem for any
4511 OST filesystems smaller than 16TB, and no kernels support ext3
4512 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4513 new/modified files) and can be fixed for existing filesystems
4514 with e2fsprogs-1.39-cfs1.
4519 Description: import connection selection may be incorrect if timer wraps
4520 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4521 import connections to be ignored if the 32-bit jiffies counter
4522 wraps. Use a 64-bit jiffies counter.
4525 Frequency : very large clusters immediately after boot
4527 Description: LNET request buffers exhausted under heavy short-term load
4528 Details : If a large number of client requests are generated on a service
4529 that has previously never seen so many requests it is possible
4530 that the request buffer growth cannot keep up with the spike in
4531 demand. Instead of dropping incoming requests, they are held in
4532 the LND until the RPC service can accept more requests.
4535 Frequency : Sometimes during replay
4537 Description: Assertion failure in ll_local_open after replay.
4538 Details : If replay happened on an open request reply before we were able
4539 to set replay handler, reply will become not swabbed tripping the
4540 assertion in ll_local_open. Now we set the handler right after
4541 recognising of open request
4544 Frequency : very rare
4546 Description: kernel reports "badness in vsnprintf"
4547 Details : Reading from the "recovery_status" /proc file in small chunks
4548 may cause a negative length in lprocfs_obd_rd_recovery_status()
4549 call to vsnprintf() (which is otherwise harmless). Exit early
4550 if there is no more space in the output buffer.
4552 Severity : enhancement
4554 Description: clear OBD RPC statistics by writing to them
4555 Details : It is now possible to clear the OBD RPC statistics by writing
4556 to the "stats" file.
4561 Description: Client mtime is not the same on different clients after utimes
4562 Details : In some cases, the client was using the utimes() syscall on
4563 a file cached on another node. The clients now validate the
4564 ctime from the MDS + OSTs to determine which one is right.
4569 Description: Inability to activate failout mode
4570 Details : lconf script incorrectly assumed that in pythong string's numeric
4571 value is used in comparisons.
4574 Frequency : always with multiple stripes per file
4576 Description: Inefficient object allocation for mutli-stripe files
4577 Details : When selecting which OSTs to stripe files over, for files with
4578 a stripe count that divides evenly into the number of OSTs,
4579 the MDS is always picking the same starting OST for each file.
4580 Return the OST selection heuristic to the original design.
4585 Description: mount failures may take full timeout to return an error
4586 Details : Under some heavy load conditions it is possible that a
4587 failed mount can wait for the full obd_timeout interval,
4588 possibly several minutes, before reporting an error.
4589 Instead return an error as soon as the status is known.
4591 Frequency : quota enabled and large files being deleted
4593 Description: releasing more than 4GB of quota at once hangs OST
4594 Details : If a user deletes more than 4GB of files on a single OST it
4595 will cause the OST to spin in an infinite loop. Release
4596 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4601 Description: statfs data retrieved from /proc may be stale or zero
4602 Details : When reading per-device statfs data from /proc, in the
4603 {kbytes,files}_{total,free,avail} files, it may appear
4604 as zero or be out of date.
4607 Frequency : systems with MD RAID1 external journal devices
4609 Description: lconf's call to blkid is confused by RAID1 journal devices
4610 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4611 of returning all block devices that match the journal UUID.
4614 Frequency : always, for aggregate stripe size over 4GB
4616 Description: assertion fails when trying to use 4GB stripe size
4617 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4618 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4621 Frequency : always on ppc64
4623 Description: the first write on an ext3 filesystem with mballoc got stuck
4624 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4625 perform endianness conversion.
4627 ------------------------------------------------------------------------------
4629 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4631 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4632 INTEROPERATE with older versions automatically. Please read the
4633 user documentation before upgrading any part of a live system.
4634 * WARNING: Lustre networking configuration changes are required with
4635 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4638 * Support for kernels:
4639 2.6.9-22.0.2.EL (RHEL 4)
4640 2.6.5-7.244 (SLES 9)
4641 2.6.12.6 vanilla (kernel.org)
4644 Severity : enhancement
4645 Bugzilla : 7981/8208
4646 Description: Introduced Lustre Networking (LNET)
4647 Details : LNET is new networking infrastructure for Lustre, it includes
4648 a reorganized network configuration mode (see the user
4649 documentation for full details) as well as support for routing
4650 between different network fabrics. Lustre Networking Devices
4651 (LNDs) for the supported network fabrics have also been
4652 created for this new infrastructure.
4654 Severity : enhancement
4655 Description: Introduced Access control lists
4656 Details : clients can set ACLs on files and directories in order to have
4657 more fine-grained permissions than the standard Unix UGO+RWX.
4658 The MDS must be started with the "-o acl" mount option.
4660 Severity : enhancement
4661 Description: Introduced filesystem quotas
4662 Details : Administrators may now establish per-user quotas on the
4665 Severity : enhancement
4667 Description: Configuration change for the XT3
4668 The PTLLND is now used to run Lustre over Portals on the XT3
4669 The configure option(s) --with-cray-portals are no longer used.
4670 Rather --with-portals=<path-to-portals-includes> is used to
4671 enable building on the XT3. In addition to enable XT3 specific
4672 features the option --enable-cray-xt3 must be used.
4677 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4678 Details : A race between allocating a new llcd and re-getting the llcd_lock
4679 allowed another thread to grab newly-allocated llcd.
4681 Severity : enhancement
4683 Description: 2.6 OST async journal commit and locking fix to improve performance
4684 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4685 2.6 kernels are now async as they already were in 2.4 kernels so
4686 that they can commit concurrently with the network bulk transfer.
4687 For block-allocated files the filter allocation semaphore is held
4688 to avoid filesystem fragmentation during allocation. BKL lock
4689 removed for 2.6 xattr operations where it is no longer needed.
4694 Description: lconf incorrectly determined whether two IP networks could talk
4695 Details : In some more complicated routing and multiple-network
4696 configurations, lconf will avoid trying to make a network
4697 connection to a disjoint part of the IP space. It was doing the
4698 math incorrectly for one set of cases.
4703 Description: Fix for potential infinite loop processing records in an llog.
4704 Details : If an llog record is corrupted/zeroed, it is possible to loop
4705 forever in llog_process(). Validate the llog record length
4706 and skip the remainder of the block on error.
4709 Frequency : occasional (liblustre only)
4711 Description: liblustre could not open files whose last component is a symlink
4712 Details : sysio_path_walk() would incorrectly pass the open intent to
4713 intermediate path components.
4716 Frequency : rare (liblustre only with non-standard tuning)
4717 Bugzilla : 7201 (7350)
4718 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4719 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4720 after it has been released, only do this for referenced locks
4722 Severity : enhancement
4724 Description: specifying an (invalid) directory default stripe_size of -1
4725 would reset the directory default striping
4726 Details : stripe_size -1 was used internally to signal directory stripe
4727 removal, now use "all default" to signal dir stripe removal
4728 as a directory striping of "all default" is not useful
4731 Frequency : common for large clusters running liblustre clients
4733 Description: doing an ls when liblustre clients are running is slow
4734 Details : sending a glimpse AST to a liblustre client waits for every AST
4735 to time out, as liblustre clients will not respond. Since they
4736 cannot cache data we refresh the OST lock LVB from disk instead.
4738 Severity : enhancement
4740 Description: doing an ls at the same time as file IO can be slow
4741 Details : enqueue and other "small" requests can be blocked behind many
4742 large IO requests. Create a new OST IO portal for non-IO
4743 requests so they can be processed faster.
4746 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4748 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4749 created file with mode 000
4750 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4751 allow an ftruncate() as a non-root user to a file with mode 000.
4752 Lustre now respects this flag to disable mode checks when
4753 truncating a file owned by the user
4756 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4759 Description: Revoking locks from clients that went dead or catatonic might take
4761 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4762 cancellation of such locks instant on servers without waiting for
4763 any reply from clients. Clients drops these locks when cancel
4764 notification from server is received without replying.
4767 Frequency : liblustre-only, when liblustre client dies or becomes busy
4769 Description: Doing ls on Linux clients can take a long time with active
4771 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4772 granting such locks to it in the first place if possible. Locks
4773 are taken by proxy on the OST during the read or write and
4774 dropped immediately afterward. Add connect flags handling, do
4775 not grant locks to liblustre clients for glimpse ASTs.
4777 Severity : enhancement
4779 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4780 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4781 sequential reads and behaves badly with applications doing random
4782 reads. Improve it by reading ahead at least read region, and
4783 avoiding excessive large RPC for small reads.
4785 Severity : enhancement
4787 Description: Creating more than 1000 files for a single job may cause a load
4788 imbalance on the OSTs if there are also a large number of OSTs.
4789 Details : qos_prep_create() uses an OST index reseed value that is an
4790 even multiple of the number of available OSTs so that if the
4791 reseed happens in the middle of the object allocation it will
4792 still utilize the OSTs as uniformly as possible.
4797 Description: OST or MDS may oops in ping_evictor_main()
4798 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4799 but doesn't restart at beginning of obd_exports_timed list
4802 Severity : enhancement
4804 Description: improve by-nid export eviction on the MDS and OST
4805 Details : allow multiple exports with the same NID to be evicted at one
4806 time without re-searching the exports list.
4809 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4811 Description: MDS may oops in groups_free()
4812 Details : in rare race conditions a newly allocated group_info struct is
4813 freed again, and this can be NULL. The 2.4 compatibility code
4814 for groups_free() checked for a NULL pointer, but 2.6 did not.
4817 Frequency : common for liblustre clients doing little filesystem IO
4818 Bugzilla : 9352, 7313
4819 Description: server may evict liblustre clients accessing contended locks
4820 Details : if a client is granted a lock or receives a completion AST
4821 with a blocking AST already set it would not reply to the AST
4822 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4825 Frequency : lfs setstripe, only systems with more than 160 OSTs
4827 Description: unable to set striping with a starting offset beyond OST 160
4828 Details : llapi_create_file() incorrectly limited the starting stripe
4829 index to the maximum single-file stripe count.
4832 Frequency : LDAP users only
4834 Description: lconf did not handle in-kernel recovery with LDAP properly
4835 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4837 Severity : enhancement
4839 Description: bind OST threads to NUMA nodes to improve performance
4840 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4841 node and do their allocations there to localize memory access
4843 Severity : enhancement
4845 Description: llmount can determine client NID directly from Myrinet (GM)
4846 Details : the client NID code from gmnalnid was moved directly into
4847 llmount, removing the need to use this or specifying the
4848 client NID explicitly when mounting GM clients with zeroconf
4851 Frequency : if client is started with down MDS
4853 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4854 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4855 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4856 ptlrpc_import_delayed_req() because we didn't check if the
4857 request was interrupted, and we also didn't break out of the
4858 event loop if there was no timeout
4863 Description: data loss during non-page-aligned writes to a single file from
4864 both multiple nodes and multiple threads on one node at same time
4865 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4866 inconsistency led to false short-reads, that were cached and later
4867 used by ->prepare_write() to fill in partially written page,
4868 leading to data loss.
4871 Frequency : always, if lconf --abort_recovery used
4873 Description: lconf --abort_recovery fails with 'Operation not supported'
4874 Details : lconf was attempting to abort recovery on the MDT device and not
4877 Severity : enhancement
4879 Description: remove cleanup logs
4880 Details : replace lconf-generated cleanup logs with lustre internal
4881 cleanup routines. Eliminates the need for client-cleanup and
4884 Severity : enhancement
4886 Description: add support for EAs (user and system) on lustre filesystems
4887 Details : it is now possible to store extended attributes in the Lustre
4888 client filesystem, and with the user_xattr mount option it
4889 is possible to allow users to store EAs on their files also
4891 Severity : enhancement
4893 Description: Add possibility (config option) to show minimal available OST free
4895 Details : When compiled with --enable-mindf configure option, statfs(2)
4896 (and so, df) will return least minimal free space available from
4897 all OSTs as amount of free space on FS, instead of summary of
4898 free spaces of all OSTs.
4900 Severity : enhancement
4902 Description: do not expand extent locks acquired on OST-side
4903 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4904 by server: they are not cached anyway.
4907 Frequency : when mmap is used/binaries executed from Lustre
4909 Description: Unmmap pages before throwing them away from read cache.
4910 Details : llap_shrink cache now attempts to unmap pages before discarding
4911 them (if unmapping failed - do not discard). SLES9 kernel has
4912 extra checks that trigger if this unmapping is not done first.
4917 Description: lconf didn't resolve symlinks before checking to see whether a
4918 given mountpoint was already in use
4921 Frequency : when migrating failover services
4922 Bugzilla : 6395, 9514
4923 Description: When migrating a subset of services from a node (e.g. failback
4924 from a failover service node) the remaining services would
4925 time out and evict clients.
4926 Details : lconf --force (implied by --failover) sets the global obd_timeout
4927 to 5 seconds in order to quickly disconnect, but this caused
4928 other RPCs to time out too quickly. Do not change the global
4929 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4931 Severity : enhancement
4932 Frequency : if MDS is started with down OST
4933 Bugzilla : 9439,5706
4934 Description: Allow startup/shutdown of an MDS without depending on the
4935 availability of the OSTs.
4936 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4937 Add appropriate locking and lov-osc refcounts for safe
4938 cleaning. Add osc abort_inflight calls in case the
4939 synchronize never started.
4942 Frequency : occasional (Cray XT3 only)
4944 Description: root not authorized to access files in CRAY_PORTALS environment
4945 Details : The client process capabilities were not honoured on the MDS in
4946 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4947 been accessed by an authorized user then root was able to access
4948 the file on the local client also. The root user capabilities
4949 are now allowed on the MDS, as this environment has secure UID.
4952 Frequency : occasional
4954 Description: ldiskfs "too long searching" message happens too often
4955 Details : A debugging message (otherwise harmless) prints too often on
4956 the OST console. This has been reduced to only happen when
4957 there are fragmentation problems on the filesystem.
4962 Description: Division by zero in statfs when all OSCs are inactive
4963 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4964 lov_statfs divides by value returned by lov_get_stripecnt().
4968 Bugzilla : 9489, 3273
4969 Description: First write from each client to each OST was only 4kB in size,
4970 to initialize client writeback cache, which caused sub-optimal
4971 RPCs and poor layout on disk for the first written file.
4972 Details : Clients now request an initial cache grant at (re)connect time
4973 and so that they can start streaming writes to the cache right
4974 away and always do full-sized RPCs if there is enough data.
4975 If the OST is rebooted the client also re-establishes its grant
4976 so that client cached writes will be honoured under the grant.
4981 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4982 Details : Now I/O RPCs go to different portal number and (presumably) fast
4983 lock requests (and glimses) and other RPCs get their own service
4984 threads pool that should be able to service those RPCs
4987 Severity : enhancement
4989 Description: Ability to exchange lustre version between client and servers and
4990 issue warnings at client side if client is too old. Also for
4991 liblustre clients there is ability to refuse connection of too old
4993 Details : New 'version' field is added to connect data structure that is
4994 filled with version info. That info is later checked by server and
4998 Frequency : rare, liblustre only.
4999 Bugzilla : 9296, 9581
5000 Description: Two simultaneous writes from liblustre at offset within same page
5001 might proceed at the same time overwriting eachother with stale
5003 Details : I/O lock withing llu_file_prwv was released too early, before data
5004 actually was hitting the wire. Extended lock-holding time until
5005 server acknowledges receiving data.
5008 Frequency : extremely rare. Never observed in practice.
5010 Description: avoid generating lustre_handle cookie of 0.
5011 Details : class_handle_hash() generates handle cookies by incrementing
5012 global counter, and can hit 0 occasionaly (this is unlikely, but
5013 not impossible, because initial value of cookie counter is
5014 selected randonly). Value of 0 is used as a sentinel meaning
5015 "unassigned handle" --- avoid it. Also coalesce two critical
5016 sections in this function into one.
5018 Severity : enhancement
5020 Description: allow liblustre clients to delegate truncate locking to OST
5021 Details : To avoid overhead of locking, liblustre client instructs OST to
5022 take extent lock in ost_punch() on client's behalf. New connection
5023 flag is added to handle backward compatibility.
5025 Severity : enhancement
5026 Bugzilla : 4928, 7341, 9758
5027 Description: allow number of OST service threads to be specified
5028 Details : a module parameter allows the number of OST service threads
5029 to be specified via "options ost ost_num_threads={N}" in the
5030 OSS's /etc/modules.conf or /etc/modprobe.conf.
5034 Bugzilla : 6146, 9635, 9895
5035 Description: servers crash with bad pointer in target_handle_connect()
5036 Details : In rare cases when a client is reconnecting it was possible that
5037 the connection request was the last reference for that export.
5038 We would temporarily drop the export reference and get a new
5039 one, but this may have been the last reference and the export
5040 was just destroyed. Get new reference before dropping old one.
5042 Severity : enhancement
5043 Frequency : if client is started with failover MDS
5045 Description: Allow multiple MDS hostnames in the mount command
5046 Details : Try to read the configuration from all specified MDS
5047 hostnames during a client mount in case the "primary"
5050 Severity : enhancement
5052 Description: Stop sending data to evicted clients as soon as possible.
5053 Details : Check if the client we are about to send or are sending data to
5054 was evicted already. (Check is done every second of waiting,
5055 for which l_wait_event interface was extended to allow checking
5056 of exit condition at specified intervals).
5059 Frequency : rare, normally only when NFS exporting is done from client
5061 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5063 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5066 Frequency : always, if srand() is called before liblustre initialization
5068 Description: Liblustre uses system PRNG disturbing its usage by user application
5069 Details : Introduce internal to lustre fast and high-quality PRNG for
5070 lustre usage and make liblustre and some other places in generic
5071 lustre code to use it.
5073 Severity : enhancement
5074 Bugzilla : 9477, 9557, 9870
5075 Description: Verify that the MDS configuration logs are updated when xml is
5076 Details : Check if the .xml configuration logs are newer than the config
5077 logs stored on the MDS and report an error if this is the case.
5078 Request --write-conf, or allow starting with --old_conf.
5080 Severity : enhancement
5082 Description: Handle symlinks in the path when checking if Lustre is mounted.
5083 Details : Resolve intermediate symlinks when checking if a client has
5084 mounted a filesystem to avoid duplicate client mounts.
5089 Description: lconf can hit an error exception but still return success.
5090 Details : The lconf command catches the Command error exception at the top
5091 level script context and will exit with the associated exit
5092 status, but doesn't ensure that this exit status is non-zero.
5097 Description: failure of ptlrpc thread startup can cause oops
5098 Details : Starting a ptlrpc service thread can fail if there are a large
5099 number of threads or the server memory is very fragmented.
5100 Handle this without oopsing.
5103 Frequency : always, only if liblustre and non-default acceptor port was used
5105 Description: liblustre cannot connect to servers with non-default acceptor port
5106 Details : tcpnal_set_default_params() was not called and was therefore
5107 ignoring the environment varaible TCPNAL_PORT, as well as other
5108 TCPNAL_ environment variables
5113 Description: two objects could be created on the same OST for a single file
5114 Details : If an OST is down, in some cases it was possible to create two
5115 objects on a single OST for a single file. No problems other
5116 than potential performance impact and spurious error messages.
5120 Bugzilla : 5681, 9562
5121 Description: Client may oops in ll_unhash_aliases
5122 Details : Client dcache may become inconsistent in race condition.
5123 In some cases "getcwd" can fail if the current directory is
5129 Description: Inode refcounting problems in NFS export code
5130 Details : link_raw functions used to call d_instantiate without obtaining
5131 extra inode reference first.
5135 Bugzilla : 9942, 9903
5136 Description: Referencing freed requests leading to crash, memleaks with NFS.
5137 Details : We used to require that call to ll_revalidate_it was always
5138 followed by ll_lookup_it. Also with revalidate_special() it is
5139 possible to call ll_revalidate_it() twice for the same dentry
5140 even if first occurence returned success. This fix changes semantic
5141 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5142 reference on a request referred from the intent.
5143 ll_intent_release() then releases such a request.
5146 Frequency : rare, normally benchmark loads only
5148 Description: unlinked inodes were kept in memory on the client
5149 Details : If a client is repeatedly creating and unlinking files it
5150 can accumulate a lot of stale inodes in the inode slab cache.
5151 If there is no other client load running this can cause the
5152 client node to run out of memory. Instead flush old inodes
5153 from client cache that have the same inode number as a new inode.
5156 Frequency : SLES9 2.6.5 kernel and long filenames only
5157 Bugzilla : 9969, 10379
5158 Description: utime reports stale NFS file handle
5159 Details : SLES9 uses out-of-dentry names in some cases, which confused
5160 the lustre dentry revalidation. Change it to always use the
5164 Frequency : rare, unless heavy write-truncate concurrency is continuous
5165 Bugzilla : 4180, 6984, 7171, 9963, 9331
5166 Description: OST becomes very slow and/or deadlocked during object unlink
5167 Details : filter_destroy() was holding onto the parent directory lock
5168 while truncating+unlinking objects. For very large objects this
5169 may block other threads for a long time and slow overall OST
5170 responsiveness. It may also be possible to get a lock ordering
5171 deadlock in this case, or run out of journal credits because of
5172 the combined truncate+unlink. Solution is to do object truncate
5173 first in one transaction without parent lock, and then do the
5174 final unlink in a new transaction with the parent lock. This
5175 reduces the lock hold time dramatically.
5178 Frequency : rare, 2.4 kernels only
5180 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5181 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5182 before unmounting their filesystem in order to avoid deadlock.
5183 The kernel_locked() function in 2.4 kernels only checks whether
5184 the kernel lock is held, not whether it is this process that is
5185 holding it as 2.6 kernels do.
5190 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5191 Details : The client ptlrpc code may be trying to reconnect to a down
5192 server before a previous connection attempt has timed out.
5193 Increase the reconnect interval to be longer than the connection
5194 timeout interval to avoid sending duplicate connections to
5198 Frequency : echo_client brw_test command
5200 Description: fix echo_client to work with OST preallocated code
5201 Details : OST preallocation code (5137) didn't take echo_client IO path
5202 into account: echo_client calls filter methods outside of any
5203 OST thread and, hence, there is no per-thread preallocated
5204 pages and buffers to use. Solution: hijack pga pages for IO. As
5205 a byproduct, this avoids unnecessary data copying.
5209 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5210 Description: Client can oops in mdc_commit_close() after open replay
5211 Details : It was possible for the MDS to return an open request with no
5212 transaction number in mds_finish_transno() if the client was
5213 evicted, but without actually returning an error. Clients
5214 would later try to replay that open and may trip an assertion
5215 Simplify the client close codepath, and always return an error
5216 from the MDS in case the open is not successful.
5219 Frequency : rare, 2.6 OSTs only
5221 Description: OST may deadlock under high load on fragmented files
5222 Details : If there was a heavy load and highly-fragmented OST filesystems
5223 it was possible to have all the OST threads deadlock waiting on
5224 allocation of biovecs, because the biovecs were not released
5225 until the entire RPC IO was completed. Instead, release biovecs
5226 as soon as they are complete to ensure forward IO progress.
5228 Severity : enhancement
5230 Description: Support for specifying external journal device at mount
5231 Details : If an OST or MDS device is formatted with an external journal
5232 device, this device major/minor is stored in the ext3 superblock
5233 and may not be valid for failover. Allow detecting and
5234 specifying the external journal at mount time.
5239 Description: Mounting an MDS with pending unlinked files may cause oops
5240 Details : target_finish_recovery() calls mds_postrecov() which returned
5241 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5242 considers this an error and immediately begins cleaning up the
5243 lov, just after starting the mds_lov process
5245 Severity : enhancement
5247 Description: Implement 'lfs df' to report actual free space on per-OST basis
5248 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5249 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5250 usage of objects; '-h' to report in human readable format.
5252 ------------------------------------------------------------------------------
5254 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5261 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5262 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5263 at startup during orphan unlinking. Instead, skip buddy update
5264 before setup, as it will be regenerated after recovery is complete.
5269 Description: If an OST is inactive, its locks might reference stale inodes.
5270 Details : lov_change_cbdata() must iterate over all namespaces, even if
5271 they are inactive to clear inode references from the lock.
5273 Severity : enhancement
5274 Frequency : occasional, if non-standard max_dirty_mb used
5276 Description: Client will block write RPCs if not enough grant
5277 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5278 then the client will block writes while waiting for another RPC
5279 to complete instead of consuming its dirty limit. With change
5280 we get improved performance when max_dirty_mb is small.
5282 Severity : enhancement
5283 Bugzilla : 3389, 6253
5284 Description: Add support for supplementary groups on the MDS.
5285 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5286 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5287 lookups for user supplementary groups into a cache.
5291 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5292 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5293 for liblustre O_EXCL is enforced by MDS.
5297 Description: readdir never returns NULL in liblustre.
5298 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5299 dentry in struct dirent.
5303 Frequency : liblustre only, depends on application IO pattern
5304 Description: liblustre clients evicted if not contacting servers
5305 Details : Don't put liblustre clients into the ping_evictor list, so
5306 they will not be evicted by the pinger ever.
5308 Severity : enhancement
5310 Description: Add ability to evict clients by NID from MDS.
5311 Details : By echoing "nid:$NID" string into
5312 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5313 $NID would be instantly evicted from this MDS and from all active
5314 OSTs connected to it.
5318 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5319 Details : lookup_it_finish() used to query file size from OSTs that was not
5324 Description: service threads change working directory to that of init
5325 Details : Starting lustre service threads may pin the working directory
5326 of the parent thread, making that filesystem busy. Threads
5327 now change to the working directory of init to avoid this.
5331 Frequency : during shutdown only
5332 Description: shutdown with a failed MDS or OST can cause unmount to hang
5333 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5338 Frequency : 2.6 only
5339 Description: chmod/chown may include an extra supplementary group
5340 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5341 supplementary group and if none is specified this is used.
5344 Bugzilla : 5479 (6816)
5346 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5347 Details : If lookup is in progress on a file that is unlinked we might try
5348 to revalidate the inode and fail in revalidate after lookup is
5349 complete and ll_file_open() enqueues the open again but
5350 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5353 Frequency : always, if lconf --abort_recovery used
5355 Description: lconf --abort_recovery fails with 'Operation not supported'
5356 Details : lconf was attempting to abort recovery on the MDT device and not
5359 ------------------------------------------------------------------------------
5361 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5366 Frequency : rare (only unsupported configurations with a node running as an
5368 Bugzilla : 6514, 5137
5369 Description: Mounting a Lustre file system on a node running as an OST could
5371 Details : OSTs now preallocates memory needed to write out data at
5372 startup, instead of when needed, to avoid having to
5373 allocate memory in possibly low memory situations.
5374 Specifically, if the file system is mounted on on OST,
5375 memory pressure could force it to try to write out data,
5376 which it needed to allocate memory to do. Due to the low
5377 memory, it would be unable to do so and the node would
5378 become unresponsive.
5380 Severity : enhancement
5382 Description: Addition of lconf --service command line option
5383 Details : lconf now accepts a '--service <arg>' option, which is
5384 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5386 Severity : enhancement
5388 Description: Failover mode is now the default for OSTs.
5389 Details : By default, OSTs will now run in failover mode. To return to
5390 the old behaviour, add '--failout' to the lmc line for OSTs.
5392 Severity : enhancement
5394 Description: Health checks are now provided for MDS and OSTs
5395 Details : Additional detailed health check information on MSD and OSTs
5396 is now provided through the procfs health_check value.
5399 Frequency : occasional, depends on IO load
5401 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5402 after numerous create/delete cycles
5403 Details : The ext3 inode allocation policy would not allocate new inodes
5404 very well on the OSTs because there are no new directories
5405 being created. Instead we look for groups with free space if
5406 the parent directories are nearly full.
5411 Description: Network or server problems during mount may cause partially
5412 mounted clients instead of returning an error.
5413 Details : The config llog parsing code may overwrite the error return
5414 code during mount error handling, returning success instead
5420 Description: MDS can fail to allocate large reply buffers
5421 Details : After long uptimes the MDS can fail to allocate large reply
5422 buffers (e.g. zconf client mount config records) due to memory
5423 fragmentation or consumption by the buffer cache. Preallocate
5424 some large reply buffers so that these replies can be sent even
5425 under memory pressure.
5429 Frequency : rare (liblustre)
5430 Description: fsx running with liblustre complained that using truncate() to
5431 extend the file doesn't work. This patch corrects that issue.
5432 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5433 ATTR_SIZE and lsm use in llu_setattr_raw.
5437 Frequency : rare, only 2.6 kernels
5438 Description: Unusual file access patterns on the MDS may result in inode
5439 data being lost in very rare circumstances.
5440 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5441 code in the 2.6 kernel caused the inode and/or EA data not to
5442 be read from disk, causing single-file corruption.
5446 Frequency : rare, only 2.6 filesystems using extents
5447 Description: Heavy concurrent write and delete load may cause data corruption.
5448 Details : It was possible under high-load situations to have an extent
5449 metadata block in the block device cache from a just-unlinked
5450 file overwrite a newly-allocated data block. We now unmap any
5451 metadata buffers that alias just-allocated data blocks.
5455 Frequency : filesystems with default stripe_count larger than 77
5456 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5457 Details : lconf specifies an inode size of 4096 bytes when the default
5458 stripe_count is larger than 77. This conflicts with the default
5459 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5460 this case to avoid pinning too much memory for large EAs.
5462 ------------------------------------------------------------------------------
5464 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5469 Frequency : rare (extremely heavy IO load with hundreds of clients)
5471 Description: Client is evicted, gets IO error writing to file
5472 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5473 caused clients to be evicted for AST timeouts. The fixes in
5474 bug 5192 mean we no longer need to have such short AST timeouts
5475 so ldlm_timeout has been increased.
5478 Frequency : occasional during --force or --failover shutdown under load
5479 Bugzilla : 5949, 4834
5480 Description: Server oops/LBUG if stopped with --force or --failover under load
5481 Details : a collection of import/export refcount and cleanup ordering
5482 issues fixed for safer force cleanup
5485 Frequency : only filesystems larger than 120 OSTs
5486 Bugzilla : 5990, 6223
5487 Description: lfs getstripe would oops on a very large filesystem
5488 Details : lov_getconfig used kfree on vmalloc'd memory
5491 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5492 Bugzilla : 6242, 6243
5493 Description: reading from files that had been truncated to a non-zero size
5494 but never opened returned no data
5495 Details : ll_file_read() reads zeros from no-object files to EOF
5500 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5501 Details : There's little harm in aborting MDS/OSS recovery and letting it
5502 try again, so I removed the LASSERT and return an error instead.
5504 Severity : enhancement
5506 Description: New debugging infrastructure for tracking down data corruption
5507 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5508 (b) cover more of the client-side code path, and (c) try to narrow
5509 down where problems occurred
5513 Bugzilla : 3819, 4364, 4397, 6313
5514 Description: Racing close and eviction MDS could cause assertion in mds_close
5515 Details : It was possible to get multiple mfd references during close and
5516 client eviction, leading to one thread referencing a freed mfd.
5518 Severity: : enhancement
5519 Bugzilla : 3262, 6359
5520 Description: Attempts to reconnect to servers are now more aggressive.
5521 Details : This builds on the enhanced upcall-less recovery that was added
5522 in 1.4.2. When trying to reconnect to servers, clients will
5523 now try each server in the failover group every 10 seconds. By
5524 default, clients would previously try one server every 25 seconds.
5529 Description: After recovery, certain operations trigger a failed
5530 assertion on a client.
5531 Details : Failing over an mds, using lconf -d --failover, while a
5532 client was doing a readdir() call would cause the client to
5533 LBUG after recovery completed and the readdir() was resent.
5535 Severity : enhancement
5537 Description: Default groups are now added by lconf
5538 Details : You can now run lconf --group <servicename> without having to
5539 manually add groups with lmc.
5542 Frequency : occasional
5544 Description: Nodes with an elan id of 0 trigger a failed assertion
5547 Frequency : always when accessing e.g. tty/console device nodes
5549 Description: tty and some other devices nodes cannot be used on lustre
5550 Details : file's private_data field is used by device data and lustre
5551 values in there got lost. New field was added to struct file to
5552 store fs-specific private data.
5555 Frequency : when exporting Lustre via NFS
5557 Description: NFSD failed occasionally when looking up a path component
5558 Details : NFSD is looking up ".." which was broken in ext3 directories
5559 that had grown large enough to become hashed.
5562 Frequency : Clusters with multiple interfaces not on the same subnet
5564 Description: Nodes will repeatedly try to reconnect to an interface which it
5565 cannot reach and report an error to the log.
5566 Details : Extra peer list entries will be created by lconf with some peers
5567 unreachable. lconf now validates the peer before adding it.
5570 Frequency : Only if a default stripe is set on the filesystem root.
5572 Description: Setting a default stripe on the filesystem root prevented the
5573 filesystem from being remounted.
5574 Details : The client was sending extra request flags in the root getattr
5575 request and did not allocate a reply buffer for the dir EA.
5578 Frequency : occasional, higher if lots of files are accessed by one client
5579 Bugzilla : 6159, 6097
5580 Description: Client trips assertion regarding lsm mismatch/magic
5581 Details : While revalidating inodes the VFS looks up inodes with ifind()
5582 and in rare cases can find an inode that is being freed.
5583 The ll_test_inode() code will free the lsm during ifind()
5584 when it finds an existing inode and then the VFS later attaches
5585 this free lsm to a new inode.
5589 Bugzilla : 6422, 7030
5590 Description: MDS deadlock between mkdir and client eviction
5591 Details : Creating a new file via mkdir or mknod (starting a transaction
5592 and getting the ns lock) can deadlock with client eviction
5593 (gets ns lock and trying to finish a synchronous transaction).
5596 Frequency : occasional
5597 Description: While starting a server, the fsfilt_ext3 module could not be
5599 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5600 kernels. Previously, lconf would still use the ext3 name
5601 when trying to load modules. Now, it will correctly use
5602 ext3 on 2.4 and ldiskfs on 2.6.
5604 Severity : enhancement
5605 Description: The default stripe count has been changed to 1
5606 Details : The interpretation of the default stripe count (0, to lfs
5607 or lmc) has been changed to mean striping across a single
5608 OST, rather than all available. For general usage we have
5609 found a stripe count of 1 or 2 works best.
5611 Severity : enhancement
5612 Description: Add support for compiling against Cray portals.
5613 Details : Conditional compiling for some areas that are different
5617 Frequency : occasional
5618 Bugzilla : 6409, 6834
5619 Description: Creating files with an explicit stripe count may lead to
5620 a failed assertion on the MDS
5621 Details : If some OSTs are full or unavailable, creating files may
5622 trigger a failed assertion on the MDS. Now, Lustre will
5623 try to use other servers or return an error to the
5627 Frequency : occasional
5629 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5630 caused lock timeout during readahead (since 1.4.2).
5631 Details : Processes doing readahead might match a lock that hasn't been
5632 granted yet if there are overlapping and conflicting lock
5633 requests. The readahead process waits on ungranted lock
5634 (original lock is CBPENDING), while OST waits for that process
5635 to cancel CBPENDING read lock and eventually evicts client.
5637 Severity : enhancement
5639 Description: Initial enabling of flock support for clients
5640 Details : Implements fcntl advisory locking and file status functions.
5641 This feature is provided as an optional mount flag (default
5642 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5643 locking are implemented yet, and those that are are not guaranteed
5644 to be completely correct in production environments.
5645 mount -t lustre -o [flock|noflock] ...
5648 Frequency : occasional
5650 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5651 assertion in the ext3 JBD (journaling) layer.
5652 Details : The b_committed_data struct is protected by the big kernel lock
5653 in 2.4 kernels, serializing journal_commit_transaction() and
5654 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5655 access to this struct. In 2.6 kernels there is finer grained
5656 locking to improve SMP performance of the JBD layer.
5660 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5665 Description: merge and cleanup kernel patches.
5668 -----------------------------------------------------------------------------
5670 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5672 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5673 versions of Lustre. Please update all servers and clients to
5674 version 1.4.2 or later at the same time. You must also run
5675 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5676 to update the configuration logs.
5678 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5679 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5680 - swab llog records of type '0' so we get proper header size/idx (5861)
5681 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5682 - fix rename of one directory over another leaking an inode (5953)
5683 - avoid SetPageDirty on 2.6 (5981)
5684 - don't re-add just-being-destroyed locks to the waiting list (5653)
5685 - when creating new directories, inherit the parent's custom
5686 striping settings if present parent (3048)
5687 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5688 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5689 - don't hold client locks on temporary worklist from l_lru (5666)
5690 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5691 - automatically evict dead clients (5921)
5692 - Update file size properly in create+truncate+fstat case (6196)
5693 - Do not unhash mountpoint dentries, do not allow removal of
5695 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5696 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5697 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5698 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5699 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5700 - Correctly handle reads of files with no objects (6243)
5701 - lctl recover will also mark a device active if deactivate used (5933)
5703 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5704 - allow --write-conf on an MDS with different nettype than client (5619)
5705 - don't write config llogs to MDS for mounts not from that MDS (5617)
5706 - lconf should create multiple TCP connections from a client (5201)
5707 - init scripts are now turned off by default; run chkconfig --on
5708 lustre and chkconfig --on lustrefs to use them
5709 - upcalls are no longer needed for clients to recover to failover
5711 - add --abort-recovery option to lconf to abort recovery on device
5713 - add support for an arbitrary number of OSTs (3026)
5714 - Quota support protocol changes.
5715 - forward compatibility changes to wire structs (6007)
5716 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5717 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5718 - improved "lustre status" script
5719 - initialize blocksize for non-regular files (6062)
5720 - added --disable-server and --disable-client configure options (5782)
5721 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5722 - Vanilla 2.4.29 support
5723 - increase maximum number of obd devices to 520 (6242)
5724 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5725 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5726 SLES 9 2.6.5 kernels
5727 - limit stripes per file to 160 (the maximum EA size) (6093)
5729 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5732 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5733 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5734 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5735 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5736 - small ext3 extents cleanups and fixes (5733)
5737 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5738 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5739 - avoid needless client->OST connect, fix handle mismatch (5317)
5740 - fix DLM error path that led to out-of-sync client, long delays (5779)
5741 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5742 - fix several locking issues related to i_size (5492,5624,5654,5672)
5743 - don't move pending lock onto export if it is already evicted (5683)
5744 - fix kernel oops when creating .foo in unlinked directory (5548)
5745 - fix deadlock in obdfilter statistics vs. object create (5811)
5746 - use time_{before,after} to avoid timer jiffies wrap (5882)
5747 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5748 - Do not leak request if server was not able to process it (5154)
5749 - If mds_open unable to find parent dir, make that negative lookup(5154)
5750 - don't create new directories with extent-mapping (5909, 5936)
5752 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5753 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5754 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5755 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5756 - support for RHEL4 kernel on i686 (5773)
5757 - provide error messages when incompatible logs are encountered (5898)
5759 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5760 * version 1.4.0.10 (1.4.1 release candidate 1)
5762 - don't keep a lock reference when lock is not granted (4238)
5763 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5764 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5765 - drop import inflight refcount on signal_completed_replay error (5255)
5766 - unlock page after async write error during send (3677)
5767 - handle missing objects in filter_preprw_read properly (5265)
5768 - no transno return for symlink open, don't save no-trasno open (3440)
5769 - don't try to complete elan receive that already failed (4012)
5770 - free RPC server reply state on error (5406)
5771 - clean up thread from ptlrpc_start_thread() on error (5160)
5772 - readahead could read extra page into cache that wasn't ejected (5388)
5773 - prevent races in class_attach/setup/cleanup/detach (5260)
5774 - don't dereference de->d_inode after l_dput of de (5458)
5775 - use "int" for stripe value returned from lock_to_stripe (5544)
5776 - mballoc allocation and error-checking fixes in 2.6 (5504)
5777 - block device patches to fix I/O request sizes in 2.6 (5482)
5778 - look up hostnames for IB nals (5602)
5779 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5780 - don't start multiple acceptors for the same port (5277)
5781 - fix incorrect LASSERT in mds_getattr_name (5635)
5782 - export a proc file for general "ping" checking (5628)
5783 - fix "lfs check" to not block when the MDS is down (5628)
5785 - service request history (4965)
5786 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5787 - create an "evict_client" /proc entry on OSTs, like the MDS has
5788 - fix mount usage message, return errors per mount(8) (5168)
5789 - change grep [] to grep "[]" in tests so they work in more UMLs
5790 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5791 - remove ancient LOV_MAGIC_V0 EA support (5047)
5792 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5793 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5794 - allow readahead allocations to fail when low on memory (5383)
5795 - mmap locking landed again, after considerable improvement (2828)
5796 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5798 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5801 - send OST transaction number in read/write reply to free req (4966)
5802 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5803 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5804 - the watchdog thread now runs as interruptible (5246)
5805 - flock/lockf fixes (but it's still disabled, pending 5135)
5806 - don't use EXT3 constants in llite code (5094)
5807 - memory shortage at startup could cause assertion (5176)
5809 - reorganization of lov code
5810 - single portals codebase
5812 - add extents/mballoc support (5025)
5813 - direct I/O reads in the obdfilter (4048)
5814 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5816 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5819 - send OST transaction number in read/write reply to free req (4966)
5820 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5821 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5822 - the watchdog thread now runs as interruptible (5246)
5823 - handle missing objects in filter_preprw_read properly (5265)
5824 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5825 - drop import inflight refcount on signal_completed_replay error (5255)
5826 - unlock page after async write error during send (3677)
5827 - return original error code on reconstructed replies (3761)
5828 - no transno return for symlink open, don't save no-trasno open (3440)
5830 - add pid to ldlm debugging output (4922)
5831 - bump the watchdog timeouts -- we can't handle 30sec yet
5832 - extra debugging for orphan dentry/inode bug (5259)
5834 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5837 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5838 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5839 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5840 - fix NULL dereference in /proc/sys/portals/routes (4827)
5841 - allow failed mdc_close() operations to be interrupted (4561)
5842 - stop precreate on OST before MDS would time out on it (4778)
5843 - don't send partial-page writes before EOF from client (4410)
5844 - discard client grant for sub-page writes on large-page clients (4520)
5845 - don't free dentries not owned by NFS code, check generation (4806)
5846 - fix lsm leak if mds_create_objects() fails (4801)
5847 - limit debug_daemon file size, always print CERROR messages (4789)
5848 - use transno after validating reply (3892)
5849 - process timed out requests if import state changes (3754)
5850 - update mtime on OST during writes, return in glimpse (4829)
5851 - add mkfsoptions to LDAP (4679)
5852 - use ->max_readahead method instead of zapping global ra (5039)
5853 - don't interrupt __l_wait_event() during strace
5855 - add software watchdogs to catch hung threads quickly (4941)
5856 - make lustrefs init script start after nfs is mounted
5857 - fix CWARN/ERROR duplication (4930)
5858 - return async write errors to application if possible (2248)
5859 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5860 - print NAL number in %x format (4645)
5861 - update barely-supported suse-2.4.21-171 series (4842)
5862 - support for sles 9 %post scripts
5863 - support for building 2.6 kernel-source packages
5864 - support for sles km_* packages
5866 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5869 - ignore -ENOENT errors in osc_destroy (3639)
5870 - notify osc create thread that OSC is being cleaned up (4600)
5871 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5872 - reconstruct ost_handle() like mds_handle() (4657)
5873 - create a new thread to do import eviction to avoid deadlock (3969)
5874 - let lconf resolve symlinked-to devices (4629)
5875 - don't unlink "objects" from directory with default EA (4554)
5876 - hold socknal file ref over connect in case target is down (4394)
5877 - allow more than 32000 subdirectories in a single directory (3244)
5878 - fix blocks count for O_DIRECT writes (3751)
5879 - OST returns ENOSPC from object create when no space left (4539)
5880 - don't send truncate RPC if file size isn't changing (4410)
5881 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5882 - bind to privileged port in socknal and tcpnal (3689)
5884 - rate limit CERROR/CWARN console message to avoid overload (4519)
5885 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5886 - basic mmap support (3918)
5887 - kernel patch series update from b1_4 (4711)
5889 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5892 - avoid crash during MDS cleanup with OST shut down (2775)
5893 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5894 - don't use bad inodes on the MDS (3744)
5895 - dynamic object preallocation to improve recovery speed (4236)
5896 - don't hold spinlock over lock dumping or change debug flags (4401)
5897 - don't zero obd_dev when it is force cleaned (3651)
5898 - print grants to console if they go negative (4431)
5899 - "lctl deactivate" will stop automatic recovery attempts (3406)
5900 - look for existing locks in ldlm_handle_enqueue() (3764)
5901 - don't resolve lock handle twice in recovery avoiding race (4401)
5902 - revalidate should check working dir is a directory (4134)
5904 - don't always mark "slow" obdfilter messages as errors (4418)
5906 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5909 - don't close LustreDB during write_conf until it is done (3860)
5910 - fix typo in lconf for_each_profile (3821)
5911 - allow dumping logs from multiple threads at one time (3820)
5912 - don't allow multiple threads in OSC recovery (3812)
5913 - fix debug_size parameters (3864)
5914 - fix mds_postrecov to initialize import for llog ctxt (3121)
5915 - replace config semaphore with spinlock (3306)
5916 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5917 - don't allow enqueue to complete on a destroyed export (3822)
5918 - down write_lock before checking llog header bitmap (3825)
5919 - recover from lock replay timeout (3764)
5920 - up llog sem before sending rpc (3652)
5921 - reduce ns lock hold times when setting kms (3267)
5922 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5923 - fix NULL deref and obd_dev leak on setup error (3312)
5924 - replace some LBUG about llog ops with error handling (3841)
5925 - don't match INVALID dentries from d_lookup and spin (3784)
5926 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5927 - fix invalid assertion in ptlrpc_set_wait (3880)
5929 - add libwrap support for the TCP acceptor (3996)
5930 - add /proc/sys/portals/routes for non-root route listing (3994)
5931 - allow setting MDS UUID in .xml (2580)
5932 - print the stack of a process that LBUGs (4228)
5934 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5938 - make sure to unset replay flag from failed open requests (3440)
5939 - if default stripe count is 0, use OST count for inode size (3636)
5940 - update parent mtime/ctime on client for create/unlink (2611)
5941 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5942 - free recovery state on server during a forced cleanup (3571)
5943 - unregister_reply for resent reqs (3063)
5944 - loop back devices mounting and status check on 2.6 (3563)
5945 - fix resource-creation race that can provoke i_size == 0 (3513)
5946 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5947 - more debugging for page-accounting assertion (3746)
5948 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5949 - avoid deadlock after precreation failure (3758)
5950 - fix race and lock order deadlock in orphan handling (3450, 3750)
5951 - add validity checks when grabbing inodes from l_ast_data (3599)
5953 - add /proc/.../recovery_status to obdfilter (3428)
5954 - lightweight CDEBUG infrastructure, debug daemon (3668)
5955 - change default OSC RPC parameters to be better on small clusters
5956 - turn off OST read cache for files smaller than 32MB
5957 - install man pages and include them in rpms (3100)
5958 - add new init script for (un)mounting lustre filesystems (2593)
5959 - run chkconfig in %post for init scripts (3701)
5960 - drop scimac NAL (unmaintained)
5962 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5965 - clean kiobufs before and after use (3485)
5966 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5967 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5968 - change init script to fail more clearly if not run as root (1528)
5969 - allow clients to reconnect during replay (1742)
5970 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5971 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5972 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5973 - fix cond_resched (3554)
5974 - only evict unfinished clients after recovery (3515)
5975 - allow bulk resend, prevent data loss (3570)
5976 - dynamic ptlrpc request buffer allocation (2102)
5977 - don't allow unlinking open directory if it isn't empty (2904)
5978 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5979 - remove extraneous obd dereference causing LASSERT failure (3334)
5980 - don't use get_cycles() when creating temp. files on the mds (3156)
5981 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5982 - handle EEXIST for set-stripe, set proper directory name (3336)
5984 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5985 - fix ksocknal_fmb_callback() error messages (2918)
5987 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5990 - don't copy lvb into (possibly NULL) reply on error (2983)
5991 - don't deref dentry after dput, don't free lvb on error (2922)
5992 - use the kms to determine writeback rpc length (2947)
5993 - increment oti_logcookies when osc is inactive (2948)
5994 - update client's i_blocks count via lvb messages (2543)
5995 - handle intent open/close of special files properly (1557)
5996 - mount MDS with errors=remount-ro, like obdfilter (2009)
5997 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5998 - don't use cancelling-locks' kms values (2947)
5999 - use highest lock extent for kms, not last one (2925)
6000 - don't dereference ERR_PTR() dentry in error handling path (3107)
6001 - fix thread race in portals_debug_dumplog() (3122)
6002 - create lprocfs device entries at setup instead of at attach (1519)
6003 - common AST error handler, don't evict client on completion race (3145)
6004 - zero nameidata in detach_mnt in 2.6 (3118)
6005 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6006 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6007 - initialize RPC timeout timer earlier for 2.6 (3219)
6008 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6009 - print nal/nid for unknown nid (3258)
6010 - additional checks for oscc recovery before doing precreate (3284)
6011 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6012 - don't crash in mdc_close for bad permissions on open (3285)
6013 - zero i_rdev for non-device files (3147)
6014 - clear page->private before handing to FS, better assertion (3119)
6015 - tune the read pipeline (3236)
6016 - fix incorrect decref of invalidated dentry (2350)
6017 - provide read-ahead stats and refine rpc in flight stats (3328)
6018 - don't hold journal transaction open across create RPC (3313)
6019 - update atime on MDS at close time (3265)
6020 - close LDAP connection when recovering to avoid server load (3315)
6021 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6022 - don't leak open file on MDS after open resend (3325)
6023 - serialize filter_precreate and filter_destroy_precreated (3329)
6024 - loop device shouldn't call sync_dev() for nul device (3092)
6025 - clear page cache after eviction (2766)
6026 - resynchronize MDS->OST in background (2824)
6027 - refuse to mount the same filesystem twice on same mountpoint (3394)
6028 - allow llmount to create routes for mounting behind routers (3320)
6029 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6030 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6031 - fix rare problem with rename on htree directories (3417)
6033 - allow default OST striping configuration per directory (1414)
6034 - fix compilation for qswnal for 2.6 kernels (3125)
6035 - increase maximum number of MDS request buffers for large systems
6036 - change liblustreapi to be useful for external progs like lfsck (3098)
6037 - increase local configuration timeout for slow disks (3353)
6038 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6040 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6043 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6044 - don't overwrite extent policy data in reply if lock was blocked (2901)
6045 - drop filter export grants atomically with removal from device (2663)
6046 - del obd_self_export from work_list in class_disconnect_exports (2908)
6047 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6048 - swab reply message in mdc_close, other PPC fixes (2464)
6049 - fix destroying of named logs (2325)
6050 - overwrite old logs when running lconf --write_conf (2264)
6051 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6052 - fix race in target_handle_connect (2898)
6053 - mds_reint_create() should take same inode create lock (2926)
6054 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6055 - don't close files for self_export to avoid uninitialized obd (2936)
6056 - allow MDS with the same name as client node (2939)
6057 - hold dentry reference for closed log files for unlink (2325)
6058 - reserve space for all logs during transactions (2059)
6059 - don't evict page beyond end of stripe extent (2925)
6060 - don't oops on a deleted current working directory (2399)
6061 - handle hard links to targets without a parent properly (2517)
6062 - don't dereference NULL lock when racing during eviction (2867)
6063 - don't grow lock extents when lots of conflicting locks (2919)
6065 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6068 - account for cache space usage on clients to avoid data loss (974)
6069 - lfsck support in lustre kernel code (2349)
6070 - reduce journal credits needed for BRW writes (2370)
6071 - orphan handling to avoid losing space on client/server crashes
6072 - ptlrpcd can be blocked, stopping ALL progress (2477)
6073 - use lock value blocks to assist in proper KMS, faster stat (1021)
6074 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6075 - recovery for initial connections (2355)
6076 - fixes for mds_cleanup_orphans (1934)
6077 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6078 - block all file creations until orphan recovery completes (1901)
6079 - client remove rq_connection from request struct (2423)
6080 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6081 - recovery timer race (2670)
6082 - mdc_close recovey bug (2532)
6083 - ptlrpc cleanup bug (2710)
6084 - mds timeout on local locks (2588)
6085 - namespace lock held during RPCs (2431)
6086 - handle interrupted sync write properly (2503)
6087 - don't try to handle a message that hasn't been replied to (2699)
6088 - client assert failure during cleanup after abort recovery (2701)
6089 - leak mdc device after failed mount (2712)
6090 - ptlrpc_check_set allows timedout requests to complete (2714)
6091 - wait for inflight reqs when ptlrpcd finishes (2710)
6092 - make sure unregistered services are removed from the srv_list
6093 - reset bulk XID's when resending them (caught by 1138 test)
6094 - unregister_bulk after timeout
6095 - fix lconf error (2694)
6096 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6097 - readahead locks pages, leaves pending causing memory pressure (2673)
6098 - increase OST request buffers to 4096 on large machines (2729)
6099 - fix up permission of existing directories in simple_mkdir (2661)
6100 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6101 - don't assign transno to errored transactions (2742)
6102 - don't delete objects on OST if given a bogus objid from MDS (2751)
6103 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6104 - if rq_no_resend, then timeout request after recovery (2432)
6105 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6106 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6107 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6108 - MDS recovery shouldn't delete objects that it has given out (2730)
6109 - if enqueue arrives after completion, don't clobber LVB (2819)
6110 - don't unlock pages twice when trigger_group_io returns error (2814)
6111 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6112 - don't write pages to disk if there was an error (1450)
6113 - don't ping imports that have recovery disabled (2676)
6114 - take buffered bytes into account when balancing socknal conn (2817)
6115 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6116 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6117 - remove little-endian swabbing of llog records (1987)
6118 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6119 - retry reposting request buffers if they fail (1191)
6120 - grow extent at grant time to avoid granting a revoked lock (2809)
6121 - lock revoke doesn't evict page if covered by a second lock (2765)
6122 - disable VM readahead to avoid reading outside lock extents (2805)
6124 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6125 - updated kernel patches for hp-2.4.20 kernel (2681)
6127 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6130 - fix truncated write corruption (2366)
6131 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6133 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6134 - ensure len(uuid) < 37 in lmc (1171)
6135 - fix ia64 OOPS in llog_test (2255)
6136 - zero end of page at obdfilter for partial page writes (2648)
6137 - don't leave stale dentries around after renames (bug 2428)
6138 - fix timeouts when evicting a client with a single lock held (2642)
6139 - set deadline for the initial HELLO message to drain (2634)
6140 - print out dotted-quad IP addresses in the socknal (2302)
6142 - additional debugging for MDS client eviction problem (2443)
6143 - fix mkfsoptions support for osts (2603, 2604)
6145 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6148 - add series for the vanilla 2.6.0 kernel
6149 - add series for the vanilla 2.4.24 kernel
6150 - add series for a cray x86/64 UL kernel drop
6151 - fix xattr patches for the vanilla 2.4.19 series
6153 - generate true UUIDs in lmc (1171)
6154 - have portals stack dumping break in UML (2466)
6155 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6156 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6157 - wake callback waiting threads on client eviction (2460)
6158 - Add --ptldebug and --subsystem to lmc (1719)
6159 - update assertion to allow safe interrupt allocation
6160 - set rq_no_resend for cancel requests (2432)
6161 - recalculate ptlrpcd timeout after resend (2494)
6162 - call vfs_rmdir when removing pending directories (2368)
6163 - fix renaming a file to itself (2429)
6164 - lmc creates a default one-stripe lov (2454)
6165 - expand procfs space to handle large clusters (2326)
6166 - increase UML stack to avoid overflow
6167 - update lconf's list of debug and subsystem masks
6168 - fix lfs find --obd (2510)
6169 - /proc tunable for disabling filter read caching (2591)
6170 - stop rpm packages from altering slapd.conf (2301)
6171 - disable nagle in the socknal under 0conf (2578)
6172 - choose mds inode size based on stripe count (2572)
6173 - fix kernel-source rpm problems (2516)
6175 - add --disable-doc to avoid pdf generation (2421)
6176 - update documentation, tests, type-os, comments
6177 - avoid format warnings on ia64
6178 - remove the TOE NAL
6179 - tiny code cleanups by removing unused fields
6181 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6185 - remove the most blatant lies from BUILDING file (2371)
6186 - change default debug level to reasonable production setting
6187 - reduce client side cache size to reduce cache flush time
6188 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6189 - make TCP zerocopy and pinger support enabled by default (2476)
6190 - sync writes completed after process exits caused crashes (2319)
6191 - maintain correct mount count on the MDS (2356)
6192 - backout 1557, because 2316 wasn't really fixed
6193 - better file I/O statistics gathering in /proc
6194 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6195 - another kernel patch to fix zero-copy TCP function export
6196 - don't take duplicate lock when processing re-sent getattr (2420)
6197 - lctl uses obd_self_export instead of creating new conn (2353)
6198 - MDS/OST recovery case which requires object creation asserted (2425)
6199 - move lfs from /usr/sbin to /usr/bin in packages
6200 - fix race between mds_client_add and mds_client_free (2417)
6201 - use kmalloc instead of slabs in portals (2430)
6202 - don't create duplicate records when a failover MDS is present (2442)
6203 - remove unnecessary mount age check (2332)
6204 - don't remove directory inodes from locks prematurely (2451)
6205 - don't break if MDS service name is the same as hostname (2103)
6206 - fix races in client write RPC generation when cache full (2482)
6208 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6211 - remove now-unused request->rq_obd (278)
6212 - if an allocation fails, print out how much memory we've used (1933)
6213 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6214 - add the "configurable stack size" patch to most series files (1256)
6215 - ability to write large log records, for 100+ OST configs (2306)
6216 - fix NULL deref when filter_prep fails (2314)
6217 - fix operator precedence error in filter_sync
6218 - dynamic allocation of socknal TX descriptors (2315)
6219 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6220 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6221 - ensure that CWARN messages in Portals always get to the syslog
6222 - __init/__exit are not for prototype decls (ldlm_init/exit)
6223 - x86-64 compile warning fixes
6224 - fix gateway LMC keyword conflict (2318)
6225 - fix MDS lock inversions in getattr/reint paths (1844)
6226 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6227 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6228 - DLM race condition prevented some lock evictions (2328)
6229 - ENOMEM detection and retry on socknal sends (2230)
6230 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6231 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6232 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6233 - better error messages when a client is rejected during recovery (1505)
6234 - avoid cancelling locks which were never granted, after failure (2330)
6235 - fix i_sem/journal inversion in mds_client_add (2333)
6236 - fix truncate/getattr lock cycle deadlock (2334)
6237 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6238 - fix two rare exit paths which could leak an l_lock() ref (2321)
6239 - fencepost error in MDS/OST orphan recovery (2226)
6240 - make log record alignment 8 bytes (1988)
6241 - lstripe now fails when requested offset > ost_count (2237)
6242 - ensure that all kernel series have a complete list.h (1607)
6243 - fix crashes in special-file operations (2316)
6244 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6246 - allow configurable automake binary, for testing new versions
6247 - small update to the lfs documentation
6249 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6251 * fix negative export reference count in fsfilt_sync (2312)
6253 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6254 * release candidate 0.9.1
6256 - orphans are moved into the PENDING directory for possible recovery
6257 - replayed opens now open by fid for orphan/rename safety (1042)
6258 - last close of an orphan inode generates a transno (683)
6259 - chdir() and mount() now pin the directory entry (1020)
6260 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6261 - discard very old requests without processing them (1502)
6262 - remove some common, well-understood CERRORs (1505)
6263 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6264 - clear "grant" flags in OST replies until OST grant code lands (1644)
6265 - fix read performance by not clobbering i_blksize on client (1598)
6266 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6267 - make LRU size a /proc tunable, clears locks when reduced (707)
6268 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6269 - server locks take references on exports now (1558)
6270 - build fixes for 2.4.20-rh trees (1663)
6271 - return an error from lov_create if all OSCs are inactive (1751)
6272 - fix import levels when a reconnect happens without a timeout (1597)
6273 - exit early from mds_open if we get a lookup error (1749)
6274 - partial page read at EOF wouldn't wait for disk before sending (1642)
6275 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6276 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6277 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6278 - write/truncate lock inversion (1639)
6279 - Don't auto-load obdclass, portals modules during cleanup (1495)
6280 - fix timestamps from jumping to "now" (1763)
6281 - extra journal assertions (1648)
6282 - add an extra multiunlink test (1771)
6283 - fix read_record/write_record API (1776)
6284 - fix leak of offset_extent, possible incorrect i_size later (1772)
6285 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6286 - Debugging for the kqswnal_get_idle_tx problems (1820)
6287 - Allow recovery to be attempted multiple times (1536)
6288 - Write out MDS last_rcvd file after it is first created (1600)
6289 - Fix tx_descriptor leak in failed transmit situations (1827)
6290 - ext3 journaling fixes for assertion failure after IO error (1871)
6291 - class_export_put() on freed export after completion AST error (1896)
6292 - Fix revalidate looping in VFS (1322)
6293 - Don't access a freed export during MDS_REINT timeout (1521)
6294 - Add open-unlink recovery support on the MDS (1673,1764)
6295 - Return an error if no MDS data was read from last_rcvd (1946)
6296 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6297 - Don't setup a disk device that doesn't match exported UUID (317)
6298 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6299 - avoid committing NULL handle in force close
6300 - local.sh is now a one-stripe LOV configuration
6301 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6302 - fix ext3 htree duplicate directory entry corruption (1516)
6303 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6304 - update to new LOV EA format (2097)
6305 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6307 2003-06-15 Phil Schwan <phil@clusterfs.com>
6310 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6311 - per-import/export recovery handling (958, 931, 959)
6312 - multiple last-rcvd slots, for serving multiple FSes (949)
6313 - connections are again shared between multiple imp/exports (963, 964)
6314 - "umount -f" would hang if any requests needed to be sent (393, 978)
6315 - avoid pinning large req buffer by copying for queued messages (989)
6316 - add "uuid" to "lctl device" command to help upcalls (991)
6317 - "open" RPCs with transnos would confuse recovery counters (1037)
6318 - do proper endian conversion of all wire messages (288, 340, 891)
6319 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6320 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6321 - fix ptlrpc_connection leak in target_handle_connect (1174)
6322 - fix import refcounting bug in OST and MDS cleanup (1134)
6323 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6324 - fix very unlikely obd_types race condition (501)
6325 - remove osc_open hack for echo_client (1187)
6326 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6327 - a failure in read_inode2 leads to deadlock (1139)
6328 - cancel ack-locks as soon as transaction is committed (1072)
6329 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6330 - make sure to commitrw after any preprw to avoid deadlock (1162)
6331 - failing to execute a file in a lustre FS would lock inode (1203)
6332 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6333 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6334 - obd_uuid2tgt would walk off the end of the list (1255)
6335 - on IA64 the osc would give portals incorrect bulk size (1258)
6336 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6337 - fix lock inversion caused by new llite matching code (1282)
6338 - limit the number of dirty pages on a client to 10MB (1286)
6339 - timed out locks were not being corrected cancelled (1289)
6340 - fix O_DIRECT above 4GB on IA-32 (1292)
6341 * major user-visible changes
6342 - fail out/fail over policy now controlled by the upcall (993)
6344 - add OBD_PING to check server availability and failure (954)
6345 - lustre messages are now sent in sending host order (288, 340, 891)
6346 - add eadatalen to MDS getattr reply (340)
6347 - OST read replies may contain second buffer, with per-page status (593)
6349 2003-03-11 Phil Schwan <phil@clusterfs.com>
6352 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6353 - failed open()s could cause deadlock; fixed (867, 869)
6354 - stop cancelling OST locks when files are closed (481)
6355 - overlapping XID spaces caused network corruption (851, 853)
6356 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6357 - setattr_raw added, to do single-RPC, server-side setattrs
6358 - lmc/lconf syntax change for OST UUIDs
6359 - fix crashy race condition between ptlrpc_free_req and osc_close
6360 - don't use request in mdc_enqueue if we hit a timeout (889)
6361 - don't set the inode i_size for regular files from the MDS (896)
6362 - handle out of order completion AST (842)
6363 - don't LBUG if a lock request times out after receiving AST (913)
6364 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6365 - if a bad lock AST arrives, send an error instead of dropping entirely
6366 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6367 - fix leak in bulk IO when only partially completed (899, 900, 926)
6368 - fix O_DIRECT for ia64 (55)
6369 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6370 - C-z after timeout could hang a process forever; fixed (977)
6372 - client-side I/O cache (678, 924, 929, 941, 970)
6374 - READPAGE and SETATTRs which don't take server-side locks get
6377 2003-02-11 Phil Schwan <phil@clusterfs.com>
6380 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6381 - Fix fsfilt_extN_readpage() to read a full page of directory
6382 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6383 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6384 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6385 - Most intent operations were removed, in favour of a new RPC mode
6386 that does a single RPC to the server and bypasses most of the VFS
6387 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6388 - Aggressively cancel local locks on DLM servers
6389 - mds_reint_unlink sends EA to the client if it's the last nlink.
6390 client uses that EA to unlink OST objects.
6391 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6392 - recursive symlinks were fixed (439)
6393 - fixed NULL deref in DEBUG_REQ
6394 - filter_update_lastobjid no longer calls sync, which annoyed extN
6395 - fixed multi-client small-writes to a single file problem (445)
6396 - fixed mtime updates during file writes (607)
6397 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6398 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6399 - fixed Linux OST size reporting problem (444, 656)
6400 - OST now updates object mtime with writes or setattr (607, 619)
6401 - client verifies file size before zeroing page past EOF (445)
6402 - OST now writes last allocated objid to disk with allocation (108)
6403 - LOV on echo now works (409)
6405 - mds_reint_unlink sends a new buffer, with the EA included. this
6406 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6407 set if a regular file was being unlinked, and it was the last link
6408 - use PtlGet from the target for bulk writes (315)
6409 - OST now updates object mtime with writes or setattr (607, 619)
6410 - LDLM now has a grant-time callback to revalidate locked items, if
6412 - Many MDS operations were reorganized to combat race conditions
6414 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6415 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6417 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6420 - Fully reactivate OST imports after reconnection (512, others)
6421 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6422 - More graceful error handling for truncating on dead OST (515)
6423 - Don't error out unless we're actually accessing dead stripes (474)
6424 - Fix garbage sizes when stripes are missing (410)
6425 - LRU counters were broken, causing constant lock purge (433, 432)
6426 - garbage on read from stripes with failed OSTs (441)
6427 - mark OSCs as active before reconnecting during recovery (438)
6428 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6429 - lfind did not preserve OST order in output (443)
6430 - symlinks cause hung clients, incorrect data (439)
6431 - stop dereferencing request after dropping refcount (457)
6432 - don't LASSERT(spin_is_locked) on non-SMP (455)
6433 - fixes for many rename() bugs
6434 - fstat didn't correctly synchronize attributes (399)
6435 - server must handle lock cancellation during blocking AST prep (487)
6436 - bulk descriptors were free()d too soon (511)
6437 - fix paths in lconf, which would load incorrect modules (451, 507)
6438 - fix confusing lconf 'host not found' error message (386)
6439 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6440 - fix race condition in mdc_blocking_ast() for inode access (526)
6441 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6442 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6443 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6444 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6445 - fix lstripe in several ways (kernel side) (527)
6446 - fix request leak in ldlm_cli_enqueue (262)
6447 - incorrect OSC was marked inactive after OST failure
6448 - call mds_fs_cleanup before unmounting filesystem (524)
6449 - fix races between taking ns_lock and ldlm_lock_change_resource
6450 - fix races updating LOV export open file list
6451 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6452 - fix recovery NULL deref in ldlm_cli_cancel_unused
6453 - fix some DLM races by using new hash table for lock handles (419)
6454 - permit the client to specify desired inodes, at replay
6455 - duplicate requests when we queue them for replay reintegration
6456 - fix last_rcvd offset calculation
6457 - sync after each recovered transaction, so we always make progress
6458 - never, not always, ERESTART requests without transnos
6459 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6461 - skip replay if the MDS says that the client is already connected
6462 - don't check for a recovery-enabled export to match lctl's UUID
6463 - don't INC_USE_COUNT for phantom exports
6464 - don't crash when cleaning up phantom exports (567)
6465 - don't double-finish or set replay data for errored mdc_open requests
6466 - abort requests when they time out, so we don't get old replies
6467 - send/receive replies for AST messages again
6468 - if the client says that it doesn't have the lock, cancel it on the
6470 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6471 mark it as destroyed, it will all work out when decref is called
6472 - fix module use counts (22, 581)
6474 - ASTs now expect a reply (server cancels lock on error reply)
6476 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6479 - fix many simultaneous client startup (392)
6480 - fix dentry->d_it clobbering
6481 - credentials weren't being shipped for readdir/getattr operations
6482 - remove invalid assertions triggered during some concurrent MD
6484 - proper Lustre versions added (336, 389)
6485 - fix memory leak for create error case (398)
6486 - fix LOV locking bug that would get cli/srv out of sync
6487 - fix echo client over LOV (409)
6488 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6489 - fix double-O_EXCL intent crash (424)
6490 - avoid sending multiple lock CANCELs (352)
6492 - MDS can do multi-client recovery (modulo bugs in new code)
6494 - many updates, edits, cleanups
6496 2002-11-18 Phil Schwan <phil@clusterfs.com>
6499 - fix null d_it dereference (346)
6500 - fix full OST/dbench hang (333)
6501 - fix permission problem with file removal (286)
6502 - fix removal of OSCs from LOV when they fail
6503 - fix NULL deref during bulk timeout (214)
6504 - fix problems related to multiple filesystems on one MDS (241)
6505 - fixed serious subtle metadata locking bugs
6506 - free locks on clients when inodes are removed due to memory
6508 - fix inode pointer in lock data (285)
6509 - partial support for multiple MDS on a single host (241)
6510 - data locks weren't cancelled at clear_inode time (290, 311)
6511 - intent locks could lead to unbounded lock growth (205)
6512 - added a maximum lock count, an LRU list, and a flusher
6513 - fix multiple rename (365)
6514 - properly abstracted the echo client
6515 - OSC locked 1 byte too many; fixed
6516 - rewrote brw callback code:
6517 - fixed recovery bugs related to LOVs (306)
6518 - fixed too-many-pages-in-one-write crash (191)
6519 - fixed (again) crash in sync_io_timeout (214)
6520 - probably fixed callback-related race (385)
6522 - Add capability to MDS protocol
6523 - LDLM cancellations and callbacks on different portals
6525 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6528 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6529 - defer open object destruction to close time (601981, 138)
6530 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6531 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6532 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6533 - don't call obd_setattr in ll_file_release for destroyed objects
6534 * protocol change to lustre_msg: move |version| and add |flags|
6535 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6536 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6537 * added replay of create, unlink, link and rename operations during
6538 MDS failover; recovery should be much more robust now
6539 * remove failed OSCs from LOVs (only lov_create uses this so far)
6540 * the lustre-HOWTO was brought (more) up to date (582544)
6542 2002-10-23 Phil Schwan <phil@clusterfs.com>
6545 - in-use dentries weren't being reused properly (617851)
6546 - prevent multiple LDLM setup (599178)
6547 - fix LOV size calculations for truncate (617853)
6548 - fix client handling of MDS intent errors (POSIX)
6549 - fix permission bug in lovstripe.c test (624321)
6550 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6551 - truncate past end of file could corrupt data
6552 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6553 - a race in recovery could return ETIMEDOUT to apps (623947)
6554 - building outside the source directory was fixed
6555 * the lustre-HOWTO was brought (more) up to date (582544)
6556 * major progress was made on recovery functionality
6558 2002-10-10 Phil Schwan <phil@clusterfs.com>
6561 - recovery deadlock fix
6562 - rm -rf causes LBUG fix (617817)
6563 - file open by multiple tasks fix (618962)
6564 - directory permissions bugs (602707 and 620007)
6565 - journal_stop fixed with locking (611313)
6566 - O_APPEND failures resolved (618273, perhaps 614459)
6567 - lconf PATH fix (619770)
6568 - IA64 build fix (621450)
6569 - RPC buffer sizes scale with amount of memory
6571 2002-10-01 Phil Schwan <phil@clusterfs.com>
6574 - locks would be cancelled without throwing away data pages,
6575 resulting in inconsistent data (605627)
6576 - inode attributes were not always being refreshed (605627, 612449)
6577 - lconf now continues to cleanup after lctl reports an error
6578 - MDS now enforces user permissions (602707)
6579 - lprocfs cleanup fixed, but not yet enabled (614157)
6580 - fixed infinite server hang, should a client not respond to an AST
6581 - avoid going into recovery if user calls readlink() with a buffer
6582 that's too small (613941)
6583 - AST RPCs no longer require replies (614867) -- this may be changed
6584 - don't crash server if client sends an IOV that's too big (611336)
6585 - fixed lock conversion deadlock (611892)
6586 - fixed the following of symlinks (614622)
6587 * recovery: the server can remove locks from a client that dies, other
6588 clients can make progress
6589 * more extN patch fixes
6590 * compile-time configurable ptlrpc buffer allocations
6592 - collaborative read cache document
6593 - Lustre Lite Performance CDR document-in-progress
6595 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6598 - fix typo in patch-2.4.18
6600 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6603 - clear ptlrpc request each time in handle_incoming_request()
6604 - unlink of files now destroys the object on the OST
6606 2002-09-19 Peter Braam <braam@clusterfs.com>
6608 * add hard link support
6609 * change obdfile creation method
6610 * kernel patch changed
6612 2002-09-19 Peter Braam <braam@clusterfs.com>
6615 - stack overflow bug in extN fixed
6617 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6619 * documentation updates
6620 - add man pages for config tools
6621 - update tests/README to describe testing with new config tools
6622 - finish metadata API descriptions
6623 * bug fixes and cleanups
6624 - statfs workaround for 16TB limit
6625 - LOV stripe allocation improved, can stripe on subset of OSTs
6626 - LOV file size/IO offset was wrong for files > 4GB in size
6627 - object EA data was being dropped, caused files to be unreadable
6628 - memory overflow with non-LOV OST caused memory corruption
6629 - fixed regression tests to work with new config tools, obdfilter
6630 - fixed bug when directory size became larger than 1 block
6631 - fixed bug (for single client case) when PWD was deleted
6632 - invalidate local directory pages when doing intent-based ops
6633 - avoid LDLM oops when lock callback contained bad data
6635 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6637 * documentation updates
6638 * bug fixes and cleanups
6639 - configuration tools
6642 - 64-bit compile warnings
6643 - 64-bit internal statfs data
6645 * test_brw on persistent OST devices
6647 * lprocfs (disabled)
6649 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6651 * documentation updates
6652 * bug fixes and cleanups
6653 * configuration tools
6655 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6658 * many small fixes to 0_5_4
6659 * io/network handling
6660 * thinkos in MDS operations
6662 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6665 * crucial basic fixes to 0.5.3
6666 * IOR, Iozone work over Elan
6669 2002-08-07 Phil Schwan <phil@clusterfs.com>
6670 * version 0_5_3, our first alpha
6671 * we use the new Portals iovs
6672 * documentation updates
6673 * bug fixes and cleanups
6674 * small changes in the DLM wire protocol
6676 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6677 * version 0_5_1 with some initial stability,
6678 * locking on MD and file I/O.
6679 * documentation updates
6680 * several bug fixes since 0.5.0
6681 * small changes in wire protocol
6683 2002-07-18 Phil Schwan <phil@clusterfs.com>
6685 * delivered as Lustre Light Alpha
6686 * fixed a crash after handling invalid MDS requests
6687 * fixed directory pages for architectures with non-4k pages sizes
6689 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6690 * release version v0_4_4
6691 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6693 * Fixes a number of interruption problems with OST operations.
6694 * Update documentation for portals header changes
6695 * Move all wire protocol structs/defines to lustre_idl.h
6696 * Fixes symlink length bug.
6697 * Add tcpdump to repository.
6699 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6700 * release version v0_4_3
6701 * Fixes statfs for inodes on extN.
6702 * Fixes bug in runtests which would delete /etc/hosts.
6703 * Use 64-bit object IDs wherever possible (not into VFS though)
6704 Remove ost_get_info, which is unused by lustre, and out of date.
6706 2002-07-03 Peter Braam <braam@clusterfs.com>
6707 * release version v0_4_2 Fixes a lookup error (type not passed)
6708 * move forward to head of Portals
6709 * move forward to latest Lustre kernel
6711 2002-06-25 Peter Braam <braam@clusterfs.com>
6712 * release version v0_4_1. Hopefully stable on single node use.