3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.10.1.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.162-94.69 (SLES12 SP3)
11 4.4.0-142 (Ubuntu 16.04)
12 4.15.0-45 (Ubuntu 18.04)
13 vanilla linux 4.6.7 (ZFS only)
14 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
15 * Client known to build on unpatched kernels (others may also work):
16 2.6.32-696.18.7.el6 (RHEL6.9)
17 2.6.32-754.6.3.el6 (RHEL6.10)
18 3.10.0-862.14.4.el7 (RHEL7.5)
19 3.10.0-957.10.1.el7 (RHEL7.6)
20 4.14.0-49.13.1.el7a (RHEL7.5)
21 4.14.0-115.2.2.el7a (RHEL7.6)
22 4.4.120-92.70 (SLES12 SP2)
23 4.4.162-94.69 (SLES12 SP3)
24 4.12.14-95.13 (SLES12 SP4)
25 4.4.0-142 (Ubuntu 16.04)
26 4.15.0-45 (Ubuntu 18.04)
27 vanilla linux 4.15.0 (ZFS only)
28 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
29 * Recommended ZFS / SPL version: 0.7.13
30 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
31 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
32 stack overflow. For more information, please refer to bugzilla 17630.
33 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
34 the re-exporting nodes to be mounted with "32bitapi" mount option
35 * Note for Ubuntu 14.04.5:
36 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
37 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
41 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
42 for currently supported client and server kernel versions.
43 * Server known to build on the following kernels (others may also work):
44 2.6.32-696.18.7.el6 (RHEL6.9)
45 3.10.0-862.14.4.el7 (RHEL7.5)
46 3.10.0-957.el7 (RHEL7.6)
47 4.4.120-92.70 (SLES12 SP2)
48 4.4.162-94.69 (SLES12 SP3)
49 4.4.0-131 (Ubuntu 16.04)
50 4.15.0-32 (Ubuntu 18.04)
51 vanilla linux 4.6.7 (ZFS only)
52 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
53 * Client known to build on unpatched kernels (others may also work):
54 2.6.32-696.18.7.el6 (RHEL6.9)
55 2.6.32-754.6.3.el6 (RHEL6.10)
56 3.10.0-862.14.4.el7 (RHEL7.5)
57 3.10.0-957.el7 (RHEL7.6)
58 4.14.0-49.13.1.el7a (RHEL7.5)
59 4.14.0-115.2.2.el7a (RHEL7.6)
60 4.4.120-92.70 (SLES12 SP2)
61 4.4.162-94.69 (SLES12 SP3)
62 4.4.0-131 (Ubuntu 16.04)
63 4.15.0-32 (Ubuntu 18.04)
64 vanilla linux 4.15.0 (ZFS only)
65 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
66 * Recommended ZFS / SPL version: 0.7.9
67 * Tested with ZFS / SPL version: 0.7.9
68 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
69 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
70 stack overflow. For more information, please refer to bugzilla 17630.
71 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
72 the re-exporting nodes to be mounted with "32bitapi" mount option
73 * Note for Ubuntu 14.04.5:
74 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
75 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
77 --------------------------------------------------------------------------------
80 2018-04-03 Intel Corporation
82 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
83 for currently supported client and server kernel versions.
84 * Server known to build on patched kernels:
85 2.6.32-431.29.2.el6 (RHEL6.5)
86 2.6.32-504.30.3.el6 (RHEL6.6)
87 2.6.32-573.26.1.el6 (RHEL6.7)
88 2.6.32-642.15.1.el6 (RHEL6.8)
89 2.6.32-696.18.7.el6 (RHEL6.9)
90 3.10.0-693.21.1.el7 (RHEL7.4)
91 3.0.101-0.47.71 (SLES11 SP3)
92 3.0.101-107 (SLES11 SP4)
93 3.12.74-60.64.40 (SLES12 SP1)
94 4.4.114-92.64 (SLES12 SP2)
95 4.4.114-94.11 (SLES12 SP3)
96 3.13.0-101 (Ubuntu 14.04, ZFS only)
97 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
98 4.4.0-116 (Ubuntu 16.04)
99 vanilla linux 4.6.7 (ZFS only)
100 * Client known to build on unpatched kernels:
101 2.6.32-431.29.2.el6 (RHEL6.5)
102 2.6.32-504.30.3.el6 (RHEL6.6)
103 2.6.32-573.26.1.el6 (RHEL6.7)
104 2.6.32-642.15.1.el6 (RHEL6.8)
105 2.6.32-696.18.7.el6 (RHEL6.9)
106 3.10.0-693.21.1.el7 (RHEL7.4)
107 3.0.101-0.47.71 (SLES11 SP3)
108 3.0.101-107 (SLES11 SP4)
109 3.12.74-60.64.40 (SLES12 SP1)
110 4.4.114-92.64 (SLES12 SP2)
111 4.4.114-94.11 (SLES12 SP3)
112 3.13.0-101 (Ubuntu 14.04)
113 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
114 4.4.0-116 (Ubuntu 16.04)
116 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
117 * Recommended ZFS / SPL version: 0.7.6
118 * Tested with ZFS / SPL version: 0.7.6
119 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
120 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
121 stack overflow. For more information, please refer to bugzilla 17630.
122 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
123 the re-exporting nodes to be mounted with "32bitapi" mount option
124 * Note for Ubuntu 14.04.5:
125 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
126 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
128 --------------------------------------------------------------------------------
130 2017-07-05 Intel Corporation
132 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
133 for currently supported client and server kernel versions.
134 * Server known to build on patched kernels:
135 2.6.32-431.29.2.el6 (RHEL6.5)
136 2.6.32-504.30.3.el6 (RHEL6.6)
137 2.6.32-573.26.1.el6 (RHEL6.7)
138 2.6.32-642.15.1.el6 (RHEL6.8)
139 2.6.32-696.3.1.el6 (RHEL6.9)
140 3.10.0-514.21.1.el7 (RHEL7.3)
141 3.0.101-0.47.71 (SLES11 SP3)
142 3.0.101-100 (SLES11 SP4)
143 3.12.74-60.64.40 (SLES12 SP1)
144 4.4.59-92.17 (SLES12 SP2)
145 vanilla linux 4.6.7 (ZFS only)
146 * Client known to build on unpatched kernels:
147 2.6.32-431.29.2.el6 (RHEL6.5)
148 2.6.32-504.30.3.el6 (RHEL6.6)
149 2.6.32-573.26.1.el6 (RHEL6.7)
150 2.6.32-642.15.1.el6 (RHEL6.8)
151 2.6.32-696.3.1.el6 (RHEL6.9)
152 3.10.0-514.21.1.el7 (RHEL7.3)
153 3.0.101-0.47.71 (SLES11 SP3)
154 3.0.101-100 (SLES11 SP4)
155 3.12.74-60.64.40 (SLES12 SP1)
156 4.4.59-92.17 (SLES12 SP2)
158 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
159 * Recommended ZFS / SPL version: 0.6.5.9
160 * Tested with ZFS / SPL version: 0.6.5.9
161 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
162 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
163 stack overflow. For more information, please refer to bugzilla 17630.
164 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
165 the re-exporting nodes to be mounted with "32bitapi" mount option
167 --------------------------------------------------------------------------------
169 12-07-2016 Intel Corporation
171 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
172 for currently supported client and server kernel versions.
173 * Server known to build on patched kernels:
174 2.6.32-431.29.2.el6 (RHEL6.5)
175 2.6.32-504.30.3.el6 (RHEL6.6)
176 2.6.32-573.26.1.el6 (RHEL6.7)
177 2.6.32-642.11.1.el6 (RHEL6.8)
178 3.10.0-514.el7 (RHEL7.3)
179 3.0.101-0.47.71 (SLES11 SP3)
180 3.0.101-80 (SLES11 SP4)
181 3.12.62-60.64.8 (SLES12 SP1)
182 vanilla linux 4.5.7 (ZFS only)
183 * Client known to build on unpatched kernels:
184 2.6.32-431.29.2.el6 (RHEL6.5)
185 2.6.32-504.30.3.el6 (RHEL6.6)
186 2.6.32-573.26.1.el6 (RHEL6.7)
187 2.6.32-642.11.1.el6 (RHEL6.8)
188 3.10.0-514.el7 (RHEL7.3)
189 3.0.101-0.47.71 (SLES11 SP3)
190 3.0.101-80 (SLES11 SP4)
191 3.12.62-60.64.8 (SLES12 SP1)
193 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
194 * Recommended ZFS / SPL version: 0.6.5.8
195 * Tested with ZFS / SPL version: 0.6.5.8
196 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
197 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
198 stack overflow. For more information, please refer to bugzilla 17630.
199 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
200 the re-exporting nodes to be mounted with "32bitapi" mount option
202 --------------------------------------------------------------------------------
204 02-29-2016 Intel Corporation
206 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
207 for currently supported client and server kernel versions.
208 * Server known to build on patched kernels:
209 2.6.32-431.29.2.el6 (RHEL6.5)
210 2.6.32-504.30.3.el6 (RHEL6.6)
211 2.6.32-573.12.1.el6 (RHEL6.7)
212 3.10.0-327.3.1.el7 (RHEL7.2)
213 3.0.101-0.47.71 (SLES11 SP3)
214 3.0.101-68 (SLES11 SP4)
215 vanilla linux 4.2.1 (ZFS only)
216 * Client known to build on unpatched kernels:
217 2.6.32-431.29.2.el6 (RHEL6.5)
218 2.6.32-504.30.3.el6 (RHEL6.6)
219 2.6.32-573.12.1.el6 (RHEL6.7)
220 3.10.0-327.3.1.el7 (RHEL7.2)
221 3.0.101-0.47.71 (SLES11 SP3)
222 3.0.101-68 (SLES11 SP4)
225 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
226 * Recommended ZFS / SPL version: 0.6.4.2
227 * Tested with ZFS / SPL version: 0.6.4.2
228 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
229 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
230 stack overflow. For more information, please refer to bugzilla 17630.
231 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
232 the re-exporting nodes to be mounted with "32bitapi" mount option
234 --------------------------------------------------------------------------------
236 03-10-2015 Intel Corporation
238 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
239 for currently supported client and server kernel versions.
240 * Server known to build on patched kernels:
241 2.6.32-431.29.2.el6 (RHEL6.5)
242 2.6.32-504.8.1.el6 (RHEL6.6)
243 3.0.101-0.46 (SLES11 SP3)
244 * Client known to build on unpatched kernels:
245 2.6.32-431.29.2.el6 (RHEL6.5)
246 2.6.32-504.8.1.el6 (RHEL6.6)
247 3.10.0-123.20.1.el7 (RHEL7)
248 3.0.101-0.46 (SLES11 SP3)
249 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
250 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
251 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
252 stack overflow. For more information, please refer to bugzilla 17630.
253 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
254 the re-exporting nodes to be mounted with "32bitapi" mount option
256 Severity : enhancement
258 Description: control OST-index in IDIF via ROCOMPAT flag.
259 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
260 last_rcvd file. For new formatted OST device, it will be auto set;
261 for the case of upgrading from old OST device, you can enable it
262 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
263 enabled, for new created OST-object, its IDIF-in-LMA will contain
264 the OST-index; for the existing OST-object, the OSD will convert
265 old format IDIF as new format IDIF with OST-index stored in the
266 LMA EA when accessing such OST-object or via OI scrub. Once such
267 flag is enabled, it cannot be reverted back, so the system cannot
268 be downgraded to the orignal incompatible version.
270 --------------------------------------------------------------------------------
272 07-30-2014 Intel Corporation
274 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
275 for currently supported client and server kernel versions.
276 * Server known to build on patched kernels:
277 2.6.32-431.20.3.el6 (RHEL6.5)
278 3.0.101-0.31 (SLES11 SP3)
279 * Client known to build on unpatched kernels:
280 2.6.32-431.20.3.el6 (RHEL6.5)
281 3.0.101-0.31 (SLES11 SP3)
284 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
285 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
286 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
287 stack overflow. For more information, please refer to bugzilla 17630.
288 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
289 the re-exporting nodes to be mounted with "32bitapi" mount option
291 --------------------------------------------------------------------------------
293 10-31-2013 Intel Corporation
295 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
296 * Server builds for kernels:
297 2.6.32-358.18.1.el6 (RHEL6.4)
298 * Client builds for unpatched kernels:
299 2.6.32-358.18.1.el6 (RHEL6.4)
300 3.0.92-0.8 (SLES11 SP3)
302 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
303 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
304 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
305 stack overflow. For more information, please refer to bugzilla 17630.
306 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
307 the re-exporting nodes to be mounted with "32bitapi" mount option
309 --------------------------------------------------------------------------------
311 05-30-2013 Intel Corporation
313 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
314 * Server builds for kernels:
315 2.6.32-358.6.2.el6 (RHEL6.4)
316 * Client builds for unpatched kernels:
317 2.6.32-358.6.2.el6 (RHEL6.4)
318 3.0.74-0.6.6 (SLES11 SP2)
320 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
321 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
322 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
323 stack overflow. For more information, please refer to bugzilla 17630.
324 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
325 the re-exporting nodes to be mounted with "32bitapi" mount option
327 --------------------------------------------------------------------------------
329 9-30-2012 Whamcloud, Inc.
331 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
332 * Server support for kernels:
333 2.6.32-279.5.1.el6 (RHEL6)
334 * Client support for unpatched kernels:
335 2.6.32-279.5.1.el6 (RHEL6)
337 2.6.18-238.19.1.el5 (RHEL5)
338 2.6.16.60-0.69.1 (SLES10)
339 * Recommended e2fsprogs version: 1.41.90.wc4
340 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
341 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
342 stack overflow. For more information, please refer to bugzilla 17630.
343 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
344 the re-exporting nodes to be mounted with "32bitapi" mount option
346 Severity : enhancement
348 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
350 Severity : enhancement
352 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
354 Severity : enhancement
356 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
358 Severity : enhancement
360 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
362 Severity : enhancement
364 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
366 --------------------------------------------------------------------------------
368 04-30-2012 Whamcloud, Inc.
370 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
371 * Server support for kernels:
372 2.6.32-220.4.2.el6 (RHEL6)
373 * Client support for unpatched kernels:
374 2.6.32-220.4.2.el6 (RHEL6)
376 2.6.18-238.19.1.el5 (RHEL5)
377 2.6.16.60-0.69.1 (SLES10)
378 * Recommended e2fsprogs version: 1.41.90.wc4
379 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
380 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
381 stack overflow. For more information, please refer to bugzilla 17630.
382 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
383 the re-exporting nodes to be mounted with "32bitapi" mount option
385 Severity : enhancement
387 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
389 Severity : enhancement
391 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
393 Severity : enhancement
395 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
397 Severity : enhancement
399 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
401 --------------------------------------------------------------------------------
403 09-30-2011 Whamcloud, Inc.
405 * Server support for kernels:
406 2.6.18-238.19.1.el5 (RHEL5)
407 2.6.32-131.6.1.el6 (RHEL6)
408 * Client support for unpatched kernels:
409 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
410 2.6.18-238.19.1.el5 (RHEL5)
411 2.6.32-131.6.1.el6 (RHEL6)
412 2.6.29.4-167.fc11 (FC11)
413 2.6.16.60-0.69.1 (SLES10)
415 * Recommended e2fsprogs version: 1.41.90.wc3
416 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
417 Lustre filesystem with 4K stack may cause a stack overflow. For more
418 information, please refer to bugzilla 17630.
419 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
420 mounted with "32bitapi" mount option
422 Severity : enhancement
423 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
425 Severity : enhancement
426 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
428 Severity : enhancement
429 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
432 Frequency : only with SLES10
434 Description: Use OFED "KMP" provided by Novell
435 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
436 Lustre is now built against this package. That means you need to
437 install the ofed-kmp package from Novell for the patchless client.
438 Note that the ofed-kmp that Novell ships may not exactly match the
439 kernel version but should still be compatible.
441 Severity : enhancement
443 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
444 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
446 --------------------------------------------------------------------------------
448 2010-07-15 Oracle, Inc.
450 * Server support for kernels:
451 2.6.18-164.11.1.el5 (RHEL 5)
452 2.6.18-164.11.1.0.1.el5 (OEL 5)
453 * Client support for unpatched kernels:
454 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
455 2.6.18-164.11.1.el5 (RHEL 5),
456 2.6.18-164.11.1.0.1.el5 (OEL 5)
457 2.6.16.60-0.42.8 (SLES 10),
459 2.6.29.4-167.fc11 (FC11)
460 * Recommended e2fsprogs version: 1.41.10-sun2
462 Severity : enhancement
464 Description: Distribute lfs_migrate script with lustre
465 Details : The lfs_migrate mentioned in the manual is now included in the
466 lustre tree in lustre/scripts and will distributed in lustre RPMs.
468 Severity : enhancement
470 Description: Failover nid list added to import proc file.
472 Severity : enhancement
474 Description: "som_preview" mount option required to use SOM
475 Details : The clients will need to use som_preview mount option
479 Description: Lustre does not do 1MB IOs to HW RAID
480 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
482 Severity : enhancement
484 Description: lfs quota output is not very convenient for awk/sed-parsing
485 Details : Some positions in lfs quota output table could be empty or
486 non-empty which made it hard to parse it with scripts, now
487 a dash is put instead of space where there is not supposed
492 Description: Update to OFED 1.5.1
493 Details : OFED 1.5.1 will be used in environments that don't use
494 vendor supplied OFED.
498 Description: Remove lnet. prefix from lctl params display
499 Details : Also renamed lnet.memused to lnet_memused to
504 Description: Use conf_param -d to remove permanant settings.
508 Description: Add list_param -R
509 Details : lctl list_param now supports recursive mode.
512 Bugzilla : 21829, 22312
513 Description: Fix llobdstat.
514 Details : 1. Now updating the write_bytes counter
515 2. Port script fixes from 1.8.
519 Description: decrease the usage of memory on clients.
520 Details : 1. On clients, recycle dentries and inodes unused.
521 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
526 Description: llobdstat fix and enhancement
527 Details : add a counter to set a limit to how many samples will be returned
528 fix a wildcard in the path to limit to obdfilter stats only
532 Description: lfs check only allowed for root
533 Details : Code cleanups and added a regresion test of non-root use of
538 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
539 Details : Fix hr->hr_index overload.
543 Description: Fix FIEMAP functionality.
547 Description: FC11 patchless client support added
551 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
552 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
556 Description: New acc-sm script and various test-framework enhancments.
560 Description: Previously landed patch was reverted.
564 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
568 Description: Add ldiskfs maxdirsize mount option.
572 Description: Fix mkfs.lustre support for 16TB LUNs.
576 Description: Print mmp_check_interval
577 Details : print mmp_check_interval and make it possible to abort
578 mount operation in case it takes too long.
582 Description: Add cascading_rw to lustre mpi tests.
586 Description: Store SOM data in EA.
587 Details : Also several SOM related cleanups.
591 Description: Add refcount to md_open_data
595 Description: Use refcount on lov_request_set to prevent memory corruption.
599 Description: Statahead: use parent dir's i_mutex for synchronization.
603 Description: loadgen improvements.
607 Description: Enable peer health detection in ksocklnd and ko2iblnd.
611 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
615 Description: Fix race on obd_devcie:md_stats
619 Description: Fix SLES 11 client builds.
620 Details : Resolved uninitalized pointer gcc warnings.
624 Description: Allow non-root access for "lfs check".
625 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
629 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
631 Severity : enhancement
633 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
634 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
638 Description: Sanity test_27p fails.
639 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
640 Sanity tests 27 were cleaned up and improved.
644 Description: Unify req format on client/servers
645 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
646 buffers and swabbers approach..
651 Description: drop join file stuff.
653 Severity : enhancement
655 Description: Add 2.6.27 and SLES11 (patchless client) support
659 Description: Fix in ptlrpc_expire_one_request() to print the signed time
662 Severity : enhancement
664 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
667 Severity : enhancement
669 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
671 Severity : enhancement
673 Description: correctly handle big reply message.
674 Details : send LNet event if reply is bigger than buffer and adjust this buffer
679 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
681 Severity : enhancement
683 Description: Add support for OEL5.
687 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
689 Severity : enhancement
691 Description: "weak-modules" support
692 Details : Implement "weak-modules" support which enables kernel modules
693 to be used with any kernel that implements the same kABI. In
694 order to achieve this modules are now installed in
695 /lib/modules/$(uname -r)/updates/kernel on all distributions.
699 Description: client could not reconnect to OST because of an active request.
700 Details : The client now retries to reconnect to the same server, if a
701 connect request failed with EBUSY or -EAGAIN. Test case 35b
702 conf-sanity.sh has been created.
706 Description: Don't allow make backward step on assignin osc next id.
707 Details : race between allocation next id and ll_sync thread can be cause
708 of set wrong osc next id and can be kill valid ost objects.
712 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
716 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
720 Description: Error handling in osc_statfs_interpret() has been improved.
721 Details : Check in osc_statfs_interpret() for EBADR.
725 Description: Do not send statfs() requests to OSTs disabled by administrator.
726 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
728 Severity : enhancement
730 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
732 Severity : enhancement
734 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
736 Severity : enhancement
738 Description: Enhancement for lfs(1) command to use numeric uid/gid.
742 Description: Improved cleanup procedure for conf-sanity test_46a.
746 Description: Do not start bulk transfer, if client tried to reconnect.
747 Details : Do not start bulk transfer, if client tried to reconnect,
748 but failed due to active rpcs. Check for exp_abort_active_req
749 has been implemented.
753 Description: Prevent inconsistences between linux and lustre mount structures.
754 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
755 Make the sleep interruptible.
757 Severity : enhancement
759 Description: provide server to client comms path
760 Details : server to client communications path via new LDLM_SET_INFO rpc
764 Description: truncate starts GFP_FS allocation under transaction causing deadlock
765 Details : ldiskfs_truncate calls grab_cache_page which may start page
766 allocation under an open transaction. This may lead to
767 calling prune_icache with consequent lustre reentrance.
771 Description: send correctly lsm on open replay
772 Details : MDS is trust to LSM size on replay open, but client can set wrong size
777 Description: enable client interface failover
778 Details : When a child reconnects from another NID, properly update export
779 nid hash position and ldlm reverse import.
781 Severity : enhancement
783 Description: Communicate OST degraded/readonly state via statfs to MDS
784 Details : Flags in the statfs returned from OSTs indicate whether the
785 OST is in a degraded RAID state, or if the filesystem has
786 turned read-only after a filesystem error is detected.
791 Description: don't panic if EPROTO was hit when reading symlink
792 Details : correctly handling request reference in error cases.
794 Severity : enhancement
796 Description: Add LustreNetLink, a kernel-userspace communcation path.
798 Severity : enhancement
800 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
805 Description: access to llog context before init.
806 Details : move handling CATALOGS file at osc layer and forbid access to llog
810 Frequency : rare, only when client is writing to OST on local node
812 Description: Avoid deadlock for local client writes
813 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
814 memory freeing context. This allows OST threads to set the
815 PF_MEMALLOC flag on task structures in order to allocate memory
816 from reserved pools and complete IO.
817 Use GFP_HIGHUSER for OST allocations for non-local client writes,
818 so that the OST threads generate memory pressure and allow
819 inactive pages to be reclaimed.
821 Severity : enhancement
823 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
828 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
829 Details : move ldlm namespace creation in setup phase to avoid grab
830 _lprocfs_lock with cli_sem held.
834 Description: Temporarily disable grant shrink.
835 Details : Disable the feature for debugging.
838 Frequency : only when formatting test filesystems
840 Description: Unable to run several mkfs.lustre on loop devices at the same time
841 Details : mkfs.lustre returns error 256 on the concurrent loop devices
842 formatting. The solution is to proper handle the error.
844 Severity : enhancement
846 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
848 Severity : enhancement
850 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
855 Description: resolve race between obd_disconnect and class_disconnect_exports
856 Details : if obd_disconnect will be called to already disconnected export he
857 forget release one reference and osc module can't unloaded.
859 Severity : enhancement
861 Description: Allow tuning service thread via /proc
862 Details : For each service a new
863 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
864 created that can be used to set min/max thread counts, and get the
865 current number of running threads.
870 Descriptoin: don't return error if have partially created objects for file.
871 Details : lov_update_create_set uses set->set_success as index for created
872 objects, so if some requests will be failed, they will have hole
873 at end of array and qos_shrink_lsm for allocate correct lsm.
875 Severity : enhancement
877 Description: Update OFED support to 1.4.1
882 Description: Reduce small size read RPC
883 Details : Set read-ahead limit for every file and only do read-ahead when
884 available read-ahead pages are bigger than 1M to avoid small size
887 Severity : enhancement
889 Description: move AT tunable parameters for more consistent usage
890 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
892 Severity : enhancement
894 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
895 Details : allow skip disconnected ost for send statfs request and hide error
901 Description: fix racy locking of mballoc block bitmaps causing BUG
902 Details : The locking of the mballoc buddy bitmap and the in-memory
903 block bitmap was using two different spin locks in some
904 cases. This made it possible to incorrectly access the
905 mballoc bitmap while another process was modifying it,
906 causing a sanity assertion to fail. While no on-disk corruption
907 was reported, there was some risk of this happening.
910 Frequency : rare, on llog test 6
912 Descriptoin: don't allow connect to already connected import
913 Details : allowing connect to already connected import is hide connecting problem.
918 Descriptoin: allow kill process which wait statahead result
919 Details : for some reasons 'ls' can stick in waiting result from statahead,
920 in this case need way for kill this process.
922 Severity : enhancement
924 Description: Add state history info file, enhance import info file
925 Details : Track import connection state changes in a new osc/mdc proc file;
926 add overview-type data to the osc/mdc import proc file.
928 Severity : enhancement
930 Description: MDS create should not wait for statfs RPC while holding DLM lock.
932 Severity : enhancement
934 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
939 Descriptoin: OST grant too much space to client even there are not enough space.
940 Details : Client will shrink its grant cache to OST if there are no write
941 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
942 this grant cache if there are already not enough avaible space
943 (left_space < total_clients * 32M).
946 Frequency : start MDS on uncleanly shutdowned MDS device
948 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
949 Details : stay in waiting mds<>ost recovery finished produce random bugs
950 due race between two ll_sync thread for one lov target. send
951 ACTIVATE event only if connect realy finished and import have
955 Frequency : rare, connect and disconnect target at same time
957 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
958 Details : don't call obd_disconnect under lov_lock. this long time
959 operation and can block ptlrpcd which answer to connect request.
964 Descriptoin: don't lose wakeup for imp_recovery_waitq
965 Details : recover_import_no_retry or invalidate_import and import_close can
966 both sleep on imp_recovery_waitq, but we was send only one wakeup
970 Frequency : always with long access acl
972 Descriptoin: mds can't pack reply with long acl.
973 Details : mds don't control size of acl but they limited by reint/getattr
976 Severity : enhancement
978 Description: Update to SLES10 kernel-2.6.16.60-0.33.
980 Severity : enhancement
982 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
985 Frequency : start MDS on uncleanly shutdowned MDS device
987 Descriptoin: aborting recovery hang on MDS
988 Details : don't throttle destroy RPCs for the MDT.
991 Frequency : on remount
993 Description: external journal device not working after the remount
994 Details : clear dev_rdonly flag for external journal devices in
1000 Description: shutdown vs evict race
1001 Details : client_disconnect_export vs connect request race.
1002 if client will evicted at this time - we start invalidate
1003 thread without referece to import and import can be freed
1007 Frequency : rare, need acl's on inode.
1009 Description: client can't handle ost additional correctly
1010 Details : if ost was added after client connected to mds client can have
1011 hit lnet_try_match_md ... to big messages to wide striped files.
1012 in this case need teach client to handle config events about add
1013 lov target and update client max ea size at that event.
1015 Severity : enhancement
1017 Description: Changelogs
1018 Details : Changelogs are a lightweight mechanism to track filesystem
1019 metadata and namespace changes. The changelog is recorded
1020 permanently on the MDTs, and is periodically "consumed" / purged
1021 when records are no longer needed.
1023 Severity : enhancement
1025 Description: compact fld format with extents
1026 Details : Store range of seq rather than every seq in FLD. Seq
1027 controller update FLD rather than clients. In Case of CMD, mdt0
1028 has FLD, all other metadata server act as non persistent proxy
1029 for FLD queries and cache fld entries in fld cache.
1034 Description: don't skip ost target if they assigned to file
1035 Details : Drop slow OSCs if we can, but not for requested start idx.
1036 This means "if OSC is slow and it is not the requested
1037 start OST, then it can be skipped, otherwise skip it only
1038 if it is inactive/recovering/out-of-space.
1042 Description: more cleanup in mds_lov
1043 Details : not send LOV EA under replay, we can't know about they size at this
1044 time. Don't allow client connect to mds before any ost connected,
1045 for avoid problems with LOV EA size and returning EIO to client.
1047 Severity : enhancement
1049 Description: Interoperability at server side (Disk interoperability)
1051 Severity : enhancement
1053 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1055 Severity : enhancement
1057 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1059 Severity : enhancement
1061 Description: New client IO stack (CLIO).
1063 Severity : enhancement
1065 Description: Commit on sharing. Eliminate inter-client dependencies between
1066 uncommitted transactions by doing transaction commits.
1067 Thereby clients may recovery independently.
1070 Frequency : Create a symlink file with a very long name
1072 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1073 Details : If there is no extra space in the request for early cancels,
1074 ldlm_req_handles_avail() returns 0 instead of a negative value.
1076 Severity : enhancement
1078 Description: Add /proc entry for import status
1079 Details : The mdc, osc, and mgc import directories now have
1080 an import directory that contains useful import data for debugging
1081 connection problems.
1083 Severity : enhancement
1085 Description: Re-disable certain /proc logging
1086 Details : Enable and disable client's offset_stats, extents_stats and
1087 extents_stats_per_process stats logging on the fly.
1090 Frequency : Only on FC kernels 2.6.22+
1092 Description: oops in statahead
1093 Details : Do not drop reference count for the dentry from VFS when lookup,
1094 VFS will do that by itself.
1096 Severity : enhancement
1098 Description: Generic /proc file permissions
1099 Details : Set /Proc file permissions in a more generic way to enable non-
1100 root users operate on some /proc files.
1104 Description: Hitting mdc_commit_close() ASSERTION
1105 Details : Properly handle request reference release in
1106 ll_release_openhandle().
1110 Description: quota recovery deadlock during mds failover
1111 Details : This patch includes att18982, att18236, att18237 in bz14840.
1113 1. fix osts hang when mds does failover with quotaon
1114 2. prevent watchdog storm when osts threads wait for the
1119 Frequency : only patchless client
1120 Description: add workaround for race between add/remove dentry from hash
1122 Severity : enhancement
1124 Description: Allow OST glimpses to return PW locks
1128 Description: LBUG when llog conf file is full
1129 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1133 Description: Prevent import from entering FULL state when server in recovery
1137 Description: service mount cannot take device name with ":"
1138 Details : Only when device name contains ":/" will mount treat it as
1144 Description: replace ptlrpcd with the statahead thread to interpret the async
1145 statahead RPC callback
1149 Frequency : on recovery
1150 Description: I/O failures after umount during fail back
1151 Details : if client reconnected to restarted server we need join to recovery
1152 instead of find server handler is changed and process self
1153 eviction with cancel all locks.
1155 Severity : enhancement
1157 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1159 Severity : enhancement
1161 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1163 Severity : enhancement
1165 Description: Upcall on Lustre log has been dumped
1166 Details : Allow for a user mode script to be called once a Lustre log has
1167 been dumped. It passes the filename of the dumped log to the
1168 script, the location of the script can be specified via
1169 /proc/sys/lnet/debug_log_upcall.
1174 Description: avoid idr_remove called for id which is not allocated.
1175 Details : Move assigment s_dev for clustered nfs to end of initialization,
1176 to avoid problem with error handling.
1181 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1182 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1183 into CONN_USED_HASH and this prodice warning when put connection
1184 again in unused hash.
1187 Severity : enhancement
1189 Description: File striping can now be set to use an arbitrary pool of OSTs.
1191 Severity : enhancement
1193 Description: Export bytes_read/bytes_write count on OSC/OST.
1197 Description: Early reply size mismatch, MGC loses connection
1198 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1199 the connect flags are properly negotiated.
1204 Description: quotas are not honored with O_DIRECT
1205 Details : all writes with the flag O_DIRECT will use grants which leads to
1206 this problem. Now using OBD_BRW_SYNC to guard this.
1210 Description: add quota statistics
1211 Details : 1. sort out quota proc entries and proc code.
1212 2. add quota statistics
1214 Severity : enhancement
1216 Description: enable quota support for HEAD.
1220 Description: Properly propagate oinfo flags from lov to osc for statfs
1221 Details : restore missing copy oi_flags to lov requests.
1223 Severity : enhancement
1225 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1226 : l_getgroups(8), lst(8), routerstat(8)
1227 Details : included man pages for llobdstat(8), llstat(8),
1228 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1230 Severity : enhancement
1232 Description: configure's --enable-quota should check the
1233 : kernel .config for CONFIG_QUOTA
1234 Details : configure is terminated if --enable-quota is passed but
1235 : no quota support is in kernel
1239 Description: Remove portals compatibility
1240 Details : Remove portals compatibility, not interoperable with releases
1245 Description: Resolve device initialization race
1246 Details : Prevent proc handler from accessing devices added to the
1247 obd_devs array but yet be intialized.
1249 Severity : enhancement
1251 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1253 Severity : enhancement
1255 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1260 Description: Using wrong pointer in osc_brw_prep_request
1261 Details : Access to array[-1] can produce panic if kernel compiled with
1262 CONFIG_PAGE_ALLOC enabled
1266 Description: Client runs out of low memory
1267 Details : Consider only lowmem when counting initial number of llap pages
1271 Description: *optional* service tags registration
1272 Details : if the "service tags" package is installed on a Lustre node
1273 When the filesystem is mounted, a local-node service tag will
1274 be created. See http://inventory.sun.com/ for more information
1275 about the Service Tags asset management system.
1279 Description: Kernel BUG tries to release flock
1280 Details : Lustre does not destroy flock lock before last reference goes
1281 away. So always drop flock locks when client is evicted and
1282 perform unlock regardless of successfulness of speaking to MDS.
1286 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1290 Description: sanity test 65a fails if stripecount of -1 is set
1291 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1296 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1297 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1300 Severity : enhancement
1302 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1306 Description: do not process already freed flock
1307 Details : flock can possibly be freed by another thread before it reaches
1308 to ldlm_flock_completion_ast.
1312 Description: LBUG during stress test
1313 Details : Need properly lock accesses the flock deadlock detection list.
1317 Description: oops in page fault handler
1318 Details : kernel page fault handler can return two special 'pages' in error
1319 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1323 Description: timeout with invalidate import.
1324 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1325 should be handled by ptlrpcd. This produce long age waiting and
1326 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1328 Severity : enhancement
1330 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1334 Description: enable MGS and MDT services start separately
1335 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1336 a co-located MGS without starting the MGS, which is a complement
1337 to 'nosvc' mount option.
1341 Frequency : after recovery
1342 Description: precreate to many object's after del orphan.
1343 Details : del orphan st in oscc last_id == next_id and this triger growing
1344 count of precreated objects. Set flag LOW to skip increase count
1345 of precreated objects.
1349 Frequency : rare, on clear nid stats
1350 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1351 Details : when clean nid stats sometimes try destroy live entry,
1352 and this produce panic in free.
1356 Description: Stack overflow during MDS log replay
1357 ease stack pressure by using a thread dealing llog_process.
1361 Description: wait until IO finished before start new when do lock cancel.
1362 Details : VM protocol want old IO finished before start new, in this case
1363 need wait until PG_writeback is cleared until check dirty flag and
1364 call writepages in lock cancel callback.
1366 Severity : enhancement
1368 Description: using special macro for print time and cleanup in includes.
1372 Description: mds_mfd_close() ASSERTION(rc == 0)
1373 Details : In mds_mfd_close(), we need protect inode's writecount change
1374 within its orphan write semaphore to prevent possible races.
1378 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1382 Frequency : rare, on shutdown ost
1383 Description: don't hit live lock with umount ost.
1384 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1385 use shrink_dcache_sb instead.
1389 Description: don't panic with use echo client
1390 Details : echo client pass NULL as client nid pointer and this produce null
1391 pointer dereference.
1395 Description: fix build on ppc32
1396 Details : compile code with -m64 flag produce wrong object file for ppc32.
1400 Description: add message levels for liblustreapi
1404 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1405 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1406 will result in an -EAGAIN and in the exit call path we call
1407 alloc_rr() which will with increasing aggressiveness attempt to
1408 aquire precreated objects on the minimum number of required OSCs.
1412 Description: Use old size assignment to avoid deadlock
1413 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1414 the scheduling while holding a spinlock. We do not need locking
1415 for size in ll_update_inode() because size is only updated from
1416 the MDS for directories or files without objects, so there is no
1417 other place to do the update, and concurrent access to such inodes
1418 are protected by the inode lock.
1422 Description: resolve "_IOWR redefined" build error on SLES10
1426 Description: dump the memory debugging after all modules are unloaded to
1427 suppress false negative in conf_sanity test 39
1429 Severity : enhancement
1431 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1436 Description: fix SLES kernel versioning
1437 Details : the kernel version for our SLES 10 kernel did not include a "-"
1438 before the "smp" at the end. while this was not a problem in
1439 general, it did mean that software trying to use the kernel
1440 version to try to detect a vendor specific kernel would fail.
1441 this was most evident by the OFED build scripts.
1445 Description: Don't update lov_desc members until making sure they are valid
1446 Details : When updating lov_desc members via proc fs, need fix their
1447 validities before doing the real update.
1451 Description: don't put request into delay list while invalidate in flight.
1452 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1453 import in flight. this produce timeout for invalidate and sometimes
1454 can cause stale data.
1458 Frequency : on ppc only
1459 Description: not convert ost objects for directory because it's not exist.
1460 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1464 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1466 Severity : enhancement
1468 Description: Update to support 2.6.22.14 vanilla kernel.
1472 Frequency : rare, on recovery
1473 Description: read procfs can produce deadlock in some situation
1474 Details : Holding lprocfs lock which send rpc can produce block for destroy
1475 obd objects and this also block reconnect with -EALREADY. This
1476 isn't fix all lprocfs bugs - but make it rare.
1478 Severity : enhancement
1480 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1483 Frequency : frequent on X2 node
1485 Description: mdc_set_open_replay_data LBUG
1486 Details : Set replay data for requests that are eligible for replay.
1490 Description: lustre_mgs: operation 101 on unconnected MGS
1491 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1492 MGC, and late on MGC cannot successfully connect to MGS and a lot
1493 of the error messages complaining that MGS is not connected.
1497 Frequency : on network error
1498 Description: panic with double free request if network error
1499 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1500 but it's true only for synchronus enqueue, for async enqueue
1501 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1504 Severity : enhancement
1506 Description: client-side metadata stat-ahead during readdir(directory readahead)
1507 Details : perform client-side metadata stat-ahead when the client detects
1508 readdir and sequential stat of dir entries therein
1511 Frequency : on start mds
1513 Description: Implement get_info(last_id) in obdfilter.
1516 Frequency : occasional
1518 Description: Correctly check stale fid, not start epoch if ost not support SOM
1519 Details : open with flag O_CREATE need set old fid in op_fid3 because
1520 op_fid2 was overwritten with new generated fid, but mds can answer
1521 with one of these two fids and both is not stale. Setattr
1522 incorrectly started an epoch and assume will be called
1523 done_writing, but without SOM done_writing ever being called.
1526 Frequency : rare, depends on device drivers and load
1528 Description: MDS or OSS nodes crash due to stack overflow
1529 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1530 In some cases, in conjunction with device drivers that use a lot
1531 of stack the MDS (or possibly OSS) service threads could overflow
1532 the stack. One change which was identified to consume additional
1533 stack has been reworked to avoid the extra stack usage.
1536 Frequency : occasional
1538 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1539 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1540 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1541 oscc_has_objects and see OSC has no objests and call
1542 oscc_internal_create to resend create request.
1544 Severity : enhancement
1546 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1548 Severity : enhancement
1550 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1553 Frequency : very rare
1555 Description: Fix replay if there is an un-replied request and open
1556 Details : In some cases, older replay request will revert the
1557 mcd->mcd_last_xid on MDS which is used to record the client's
1558 latest sent request.
1560 Severity : enhancement
1562 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1564 Severity : enhancement
1566 Description: Add rhel5 support to HEAD.
1568 Serverity : enhancement
1570 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1575 Description: Don't allow skipping OSTs if index has been specified.
1576 Details : Don't allow skipping OSTs if index has been specified, make
1577 locking in internal create lots better.
1581 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1582 Details : access to bitfield in structure is always rounded to long
1583 and this produce problem with not atomic change any bit.
1587 Description: Lustre make rpms failed.
1588 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1589 builds Lustre rpms from tarball.
1592 Frequency : rare on shutdown ost
1594 Description: If llog cancel was not send before clean_exports phase, this can
1595 produce deadlock in llog code.
1596 Details : If llog thread has last reference to obd and call class_import_put
1597 this produce deadlock because llog_cleanup_commit_master wait when
1598 last llog_commit_thread exited, but this never success because was
1599 called from llog_commit_thread.
1603 Description: allow userland application know is lost one of stripes.
1604 Details : fill lvb_blocks with error code on ost and return it to
1605 application if error flag found.
1609 Description: NULL lov_tgts causing MDS oops
1610 Details : more safe checks for NULL lov_tgts for avoid oops.
1612 Severity : enhancement
1614 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1618 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1619 Details : If an OST is down the MDS will hang indefinitely in
1620 obd_statfs_rqset() waiting for the statfs data. While for
1621 MDS QOS usage of statfs, it should not stuck in waiting.
1623 Severity : enhancement
1625 Description: remote_acl support
1626 Details : Support ACL-based permission check for remote user.
1627 Support setfacl/getfacl for remote user with the utils
1628 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1629 the system "{s,g}etfacl" utils.
1631 Severity : enhancement
1633 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1635 Severity : enhancement
1637 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1641 Description: make mgs_setparam() handle fsname containing dash
1642 Details : fsname containing a dash does not work with lctl conf_param
1644 Severity : enhancement
1646 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1648 Severity : enhancement
1650 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1654 Frequency : rare, at shutdown
1655 Description: access already free / zero obd_namespace.
1656 Details : if client_disconnect_export was called without force flag set,
1657 and exist connect request in flight, this can produce access to
1658 NULL pointer (or already free pointer) when connect_interpret
1659 store ocd flags in obd_namespace.
1663 Frequency : only at startup
1664 Description: not alloc memory with spinlock held.
1665 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1666 if any spinlock held.
1668 Severity : enhancement
1670 Description: make lustre randomly fail allocating memory
1671 Details : Make lustre randomly failed allocating memory for testing purpose.
1673 Severity : enhancement
1675 Description: lost problems with lov objid file
1676 Details : Fixes some scability and access to not inited memory problems
1677 in work with lov objdid file.
1682 Description: lfs find does not continue on file error
1683 Details : Continue other files processing when a file/dir is absent.
1687 Description: Inconsistent usage of lustre_pack_reply()
1688 Details : Standardize the usage of lustre_pack_reply() such that it
1689 always generate a CERROR on failure.
1692 Frequency : occasional
1694 Description: MDS hang or stay in waiting lock
1695 Details : If client receive lock with CBPENDING flag ldlm need send lock
1696 cancel as separate rpc, to avoid situation when cancel request
1697 can't processed due all i/o threads stay in wait lock.
1701 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1705 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1709 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1713 Description: Update to RHEL4 latest kernel.
1715 Severity : enhancement
1717 Description: Build SLES10 patchless client fails
1718 Details : The configure was broken by run ./configure with
1719 --with-linux-obj=.... argument for patchless client. When the
1720 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1721 can't search header adequately. Use absolute path such as
1722 -I($LINUX)/include instead.
1726 Description: interrupt oig_wait produce painc on resend.
1727 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1728 private set, and this produce situation when rq_ptlrpcd_data not
1729 copyed to new allocated request and triggered LBUG on assert
1730 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1731 ptlrpcd set if request is changed and can be safety changed to use
1736 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1737 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1738 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1739 these macros and replace them with their difinitions instead.
1743 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1744 Details : Update which_patch & target file for SLES10 latest kernel.
1747 Description: Remove client patches from SLES 10 kernel.
1748 Details : This causes SLES 10 clients to behave as patchless clients
1749 even on a Lustre-patched (server) kernel.
1751 Severity : enhancement
1753 Description: self-adjustable client's lru lists
1754 Details : use adaptive algorithm for managing client cached locks lru
1755 lists according to current server load, other client's work
1756 pattern, memory activities, etc. Both, server and client
1757 side namespaces provide number of proc tunables for controlling
1760 Severity : enhancement
1762 Description: light-weight GSS support
1763 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1764 and reduce performance overhead.
1766 Severity : enhancement
1768 Description: Linux keyring support
1769 Details : Support using service of Linux keyring for Lustre GSS internal
1770 context refresh/cache mechanism.
1774 Description: Fix errors in lfs documentation
1775 Details : Fixes man pages
1779 Description: don't use GFP_* in generic Lustre code.
1780 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1784 Description: obdclass is limited by single OBD_ALLOC(idarray)
1785 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1789 Description: warnings with build patchless client with vanila 2.6.19 and up
1790 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1791 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1795 Frequency : Only for SLES
1796 Description: sanity_quota fail test_1
1797 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1798 file, which makes TSTID[2] inunique.
1802 Frequency : Only for SLES9
1803 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1804 Details : Update target/ChangeLog/which_patch .
1808 Frequency : Only for test_10 in sanity-quota.sh
1809 Description: a bug in quota test code
1810 Details : checking a test flag in a wrong place
1814 Frequency : Only for test_13 in sanity-quota.sh
1815 Description: a bug in quota test script
1816 Details : shouldn't check overfull allocated quota
1820 Description: jbd statistics
1821 Details : Port older jbd statistics patch for sles10
1823 Severity : enhancement
1825 Description: Add printing inode info into message about error in writepage.
1830 Description: reply_lock_interpret crash due to race with it and lock cancel.
1831 Details : Do not replay locks that are being cancelled. Do not reference
1832 locks by their address during replay, just by their handle.
1836 Frequency : When flocks are used.
1837 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1838 Details : Flock locks might destroy just granted lock if it could be merged
1839 with another existing flock, this is done in completion handler,
1840 so teach ldlm_cli_enquque_fini that this is a valid case for
1846 Description: Oops in read and write path when failing to allocate lock.
1847 Details : Check if lock allocation failed and return error back.
1851 Description: lstripe command fails for valid OST index
1852 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1853 instead of lov->desc.ld_active_tgt_count.
1857 Description: sanity.sh failed test 103
1858 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1863 Description: df doesn't work properly if diskfs blocksize != 4K
1864 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1868 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1869 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1871 Severity : enhancement
1873 Description: Adaptive timeouts
1874 Details : RPC timeouts adapt to changing server load and network
1875 conditions to reduce resend attempts and improve recovery time.
1879 Description: llapi_file_create() does not allow some changes
1880 Details : add llapi_file_open() that allows specifying the mode and
1881 open flags, and also returns an open file handle.
1885 Description: merge and cleanup kernel patches.
1886 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1890 Description: Add journal checksum support.(Kernel part)
1891 Details : The journal checksum feature adds two new flags i.e
1892 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1893 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1894 indicates that the commit block contains the checksum for
1895 the blocks described by the descriptor blocks. Now commit
1896 record can be sent to disk without waiting for descriptor
1897 blocks to be written to disk. This behavior is controlled
1898 using JBD2_FEATURE_ASYNC_COMMIT flag.
1902 Description: OSS needs mutliple precreate threads
1903 Details : Add ability to start more than one create thread per OSS.
1907 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1908 Details : execute lfs setstripe on client
1912 Description: mds_obd_create error creating tmp object
1913 Details : When the user sets quota on root, llog will be affected and can't
1914 create files and write files.
1917 Frequency : Always on ia64 patchless client, and possibly others.
1919 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1920 Details : This allows the patchless client to be loaded on architectures
1921 without this export.
1925 Description: RedHat Update kernel for RHEL5
1926 Details : Add the kernel config file for RHEL5.
1930 Description: RedHat Update kernel for RHEL5
1931 Details : Modify the kernel config file more closer RHEL5.
1935 Description: Build failure against Centos5 (RHEL5)
1936 Details : Define PAGE_SIZE when it isn't present.
1940 Description: test_7 on sanity-quota.sh can't be used on separate machines
1941 Details : add facet to handle it
1945 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1946 Details : with NFS, the anon dentry's parent was set to itself in
1947 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1948 corresponding dentry other than use rec->ur_name.
1950 Severity : enhancement
1952 Description: lfs setstripe enhancement
1953 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1957 Description: enable data checksumming by default
1958 Details : enable checksum by default, allow --disable-checksum
1959 configure option and "-o nochecksum" mount option. Checksums
1960 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1964 Description: lustre support for RHEL5
1965 Details : Add support for RHEL5.
1969 Description: Client eviction due to failover config
1970 Details : after a connection loss, the lustre client should attempt to
1971 reconnect to the last active server first before trying the
1972 other potential connections.
1976 Description: when mds and osts use different quota unit(32bit and 64bit),
1977 quota will be released repeatly.
1978 Details : void sending multiple quota reqs to mds, which will keep the
1979 status between the reqs.
1983 Description: osts not allocated evenly to files
1984 Details : change the condition to increase offset_idx
1987 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1989 Description: Data corruption for OSTs that are formatted larger than 2TB
1991 Details : When generating the bio request for lustre file writes the
1992 sector number would overflow a temporary variable before being
1993 used for the IO. The data reads correctly from Lustre (which
1994 will overflow in a similar manner) but other file data or
1995 filesystem metadata may be corrupted in some cases.
1999 Description: Tune the kernel for good SCSI performance.
2000 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2001 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2006 Description: rewrite ext2-derived code in obdclass/uuid.c
2007 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2008 from scratch preserving functionality.
2012 Description: "lfs find -obd UUID" prints directories
2013 Details : "lfs find -obd UUID" will return all directory names instead
2014 of just file names. It is incorrect because the directories
2015 do not reside on the OSTs.
2019 Description: use i_size_read and i_size_write in 2.6 port
2020 Details : replace inode->i_size access with i_size_read/write()
2023 Frequency : when removing large files
2025 Description: scheduling issue during removal of large Lustre files
2026 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2027 It causes scheduling issues when removing large files (17TB in the
2031 Frequency : only with liblustre clients on XT3
2033 Description: evictions taking too long
2034 Details : allow llrd to evict clients directly on OSTs
2037 Frequency : only on ppc
2039 Description: /proc/fs/lustre/devices broken on ppc
2040 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2041 We should make sure the type of 'index' is loff_t and then cast
2042 explicitly as needed below. Do not assign an explicitly cast
2046 Frequency : only for rhel5
2048 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2049 Details : Modify the target file & which_kernel.
2051 Severity : enhancement
2053 Description: omit set fsid for export NFS
2054 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2055 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2059 Frequency : Always, for kernels after 2.6.16
2060 Description: Fix warning idr_remove called for id=.. which is not allocated.
2061 Details : Last kernels save old s_dev before kill super and not allow
2062 to restore from callback - restore it before call kill_anon_super.
2066 Description: buffer overruns could theoretically occur
2067 Details : llapi_semantic_traverse() modifies the "path" argument by
2068 appending values to the end of the origin string, and a
2069 overrun may occur. Adding buffer overrun check in liblustreapi.
2073 Description: Fix error on 'ls .' at the top of the Lustre mount.
2074 Details : Don't revalidate dentry if it is a root dentry.
2078 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2079 Details : Modify vm-tunables-rhel4.patch.
2083 Description: Kernel config for 2.6.18-vanilla.
2084 Details : Modify targets/2.6-vanilla.target.in.
2085 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2086 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2087 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2088 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2092 Description: improve handling recoverable errors
2093 Details : if request processig with error which can be recoverable on server
2094 request should be resend, otherwise page released from cache and
2099 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2100 Details : Modify target file & which_patch.
2101 A flaw was found in the IA32 system call emulation provided
2102 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2103 value could be stored in the %RAX register, which could trigger an
2104 out-of-bounds system call table access. An untrusted local user
2105 could exploit this flaw to run code in the kernel
2106 (ie a root privilege escalation). (CVE-2007-4573).
2110 Description: O_DIRECT bypasses client statistics.
2111 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2112 read_ahead_stats were updated. copied the stats section from
2113 osc_send_oap_rpc() into async_internal().
2117 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2121 Description: change order of libsysio includes
2122 Details : '#include sysio.h' should always come before '#include xtio.h'
2126 Description: handle "serious error: objid * already exists" more gracefully
2127 Details : If LAST_ID value on disk is smaller than the objects existing in
2128 the O/0/d* directories, it indicates disk corruption and causes an
2129 LBUG(). If the object is 0-length, then we should use the existing
2130 object. This will help to avoid a full fsck in most cases.
2132 Severity : enhancement
2134 Description: adapt the lustre_config script to support the upgrade case
2135 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2136 server targets to 1.6 in parallel.
2140 Description: To avoid grant space > avaible space when the disk is almost
2141 full. Without this patch you might see the error "grant XXXX >
2142 available" or some LBUG about grant, when the disk is almost
2144 Details : In filter_check_grant, for non_grant cache write, we should
2145 check the left space by if (*left > ungranted + bytes), instead
2146 of (*left > ungranted), because only we are sure the left space
2147 is enough for another "bytes", then the ungrant space should be
2148 increase. In client, we should update cl_avail_grant only there
2149 is OBD_MD_FLGRANT in the reply.
2153 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2154 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2155 described in CVE-2007-4573. This problem could allow untrusted
2156 local users to gain root access.
2159 Frequency : when using O_DIRECT and quotas
2161 Description: Incorrect file ownership on O_DIRECT output files
2162 Details : block usage reported by 'lfs quota' does not take into account
2163 files that have been written with O_DIRECT.
2168 Description: touch file failed when fs is not full
2169 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2170 otherwise we can get ENOSP while fs is not full.
2174 Description: parallel lock callbacks
2175 Details : Instead of sending blocking and completion callbacks as separated
2176 requests, adding them to a set and sending in parallel.
2179 Frequency : only for Cray XT3
2180 Bugzilla : 12829/13455
2181 Description: Changing primary group doesn't change the group lustre assigns to
2183 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2184 overridden with the primary group provided by the group upcall,
2185 whereas the supplied fsgid can be trusted if it is in the list of
2186 supplementary groups returned by the group upcall.
2188 Severity : enhancement
2190 Description: Allow masking D_WARNING, D_ERROR messages from console
2191 Details : Console messages can now be disabled via lnet.printk.
2195 Description: User code with malformed file open parameter crashes client node
2196 Details : Before packing join_file req, all the related reference should be
2197 checked carefully in case some malformed flags cause fake
2198 join_file req on client.
2202 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2203 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2204 mode is equal to the granted mode. However, it should also check
2205 that LDLM_FL_CP_REQD is not set.
2209 Description: Heavy nfs access might result in deadlocks
2210 Details : After ELC code landed, it is now improper to enqueue any mds
2211 locks under och_sem, because enqueue might want to decide to
2212 cancel open locks for same inode we are holding och_sem for.
2216 Description: Client eviction while running blogbench
2217 Details : A lot of unlink operations with concurrent I/O can lead to a
2218 deadlock causing evictions. To address the problem, the number of
2219 oustanding OST_DESTROY requests is now throttled to
2220 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2221 callbacks are processed in priority.
2225 Description: enable ACLs on MDS by default
2226 Details : ACLs must be enabled on MDS by default.
2229 Frequency : PPC/PPC64 only
2231 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2232 Details : fix duplicated definitions between asm-ppc64/types.h and
2233 lustre_types.h on PPC.
2236 Frequency : PPC/PPC64 only
2238 Description: asm-ppc/segment.h does not exist
2239 Details : fix compile issue on PPC.
2243 Description: better handle error messages in extents code
2246 Frequency : RHEL4 only
2248 Description: mkfs is very slow on IA64/RHEL4
2249 Details : A performance regression has been discovered in the MPT Fusion
2250 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2251 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2252 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2254 Severity : enhancement
2256 Description: SNMP support enhancement
2257 Details : Adding total number of sampled request for an MDS node in snmp
2260 Severity : enhancement
2262 Description: Optimize ldlm waiting list processing for PR extent locks
2263 Details : When processing waiting list for read extent lock and meeting read
2264 lock that is same or wider to it that is not contended, skip
2265 processing rest of the list and immediatelly return current
2266 status of conflictness, since we are guaranteed there are no
2267 conflicting locks in the rest of the list.
2271 Description: Time out and refuse to reconnect
2272 Details : When the failover node is the primary node, it is possible
2273 to have two identical connections in imp_conn_list. We must
2274 compare not conn's pointers but NIDs, otherwise we can defeat
2275 connection throttling.
2279 Description: port llog fixes from b1_6 into HEAD
2280 Details : Port llog reference couting and some llog cleanups from b1_6
2281 (bug 10800) into HEAD, for protect from panic and access to already
2282 free llog structures.
2286 Description: Detect stride IO mode in read-ahead
2287 Details : When a client does stride read, read-ahead should detect that and
2288 read-ahead pages according to the detected stride pattern.
2292 Description: data checksumming impacts single node performance
2293 Details : add support for several checksum algorithm. Currently, only CRC32
2294 and Adler-32 are supported. The checksum type can be changed on
2295 the fly via /proc/fs/lustre/osc/*/checksum_type.
2299 Description: use adler32 for page checksums
2300 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2305 Description: build for x2 fails
2306 Details : fix compile issue on Cray systems.
2310 Description: Properly match for duplicate locks
2311 Details : Due to different lock order from skiplists code, we need to
2312 traverse entire list for now
2315 Frequency : only on PPC/SLES10
2317 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2318 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2319 be defined. Add a hack in mkfs_lustre.c to work around this compile
2324 Description: LASSERT on MDS when client holding flock lock dies
2325 Details : ldlm pool logic depends on number of granted locks equal to
2326 number of released locks which is not true for flock locks, so
2327 just exclude such locks from consideration.
2331 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2332 Details : Use fsfilt_sync() for both the whole filesystem sync and
2333 individual file sync to eliminate dangerous inode locking
2334 with I_LOCK that can lead to a deadlock.
2338 Description: performance in 1.6.3
2339 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2340 sure that 1MB requests are not fragmented by the block layer.
2342 Severity : enhancement
2344 Description: organize the server-side client stats on per-nid basis
2345 Details : Change the structure of stats under obdfilter and mds to
2353 The "uuid"s file would list the uuids of _active_ exports.
2354 And the clear entry is to clear all stats and stale nids.
2356 Severity : enhancement
2358 Description: eliminate client locks in face of contention
2359 Details : file contention detection and lockless i/o implementation
2360 for contended files.
2364 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2367 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2369 Description: MDS may lose file striping (and hence file data) in some cases
2370 Details : If there are additional extended attributes stored on the MDS,
2371 in particular ACLs, SELinux, or user attributes (if user_xattr
2372 is specified for the client mount options) then there is a risk
2373 of attribute loss. Additionally, the Lustre file striping
2374 needs to be larger than default (e.g. striped over all OSTs),
2375 and an additional attribute must be stored initially in the
2376 inode and then increase in size enough to be moved to the
2377 external attribute block (e.g. ACL growing in size) for file
2382 Description: skiplist implementation simplification
2383 Details : skiplists are used to group compatible locks on granted list
2384 that was implemented as tracking first and last lock of each
2385 lock group the patch changes that to using doubly linked lists
2389 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2390 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2391 because LASSERT on both the data supplied by a client, and the
2392 data on disk is dangerous and incorrect.
2394 Severity : enhancement
2396 Description: Slow truncate/writes to huge files at high offsets.
2397 Details : Directly associate cached pages to lock that protect those pages,
2398 this allows us to quickly find what pages to write and remove
2399 once lock callback is received.
2403 Description: more ldlm soft lockups
2404 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2405 starve other threads from the resource lock for a long time in
2406 case of long waiting queue, so change the debug level from
2407 D_OTHER to the less frequently used D_INFO.
2409 Severity : enhancement
2411 Description: add -gid, -group, -uid, -user options to lfs find
2415 Description: Hung threads in invalidate_inode_pages2_range
2416 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2417 once one is completed. As a result, some RPCs are stuck in the
2418 queue and are never sent.
2422 Description: filter threads hungs on waiting journal commit
2423 Details : Cleanup filter group llog code, then only filter group llog will
2424 be only created in the MDS/OST syncing process.
2428 Description: Procfs and llog threads access destoryed import sometimes.
2429 Details : Sync the import destoryed process with procfs and llog threads by
2430 the import refcount and semaphore.
2432 Severity : enhancement
2434 Description: openlock cache of b1_6 port to HEAD
2439 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2440 Details : If insertion of an extent fails, then discard the inode
2441 preallocation and free data blocks else it can lead to duplicate
2446 Description: don't always update ctime in ext3_xattr_set_handle()
2447 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2448 In some cases the ctime should not be updated, for example for
2449 2.0->1.8 compatibility it is necessary to delete an xattr and it
2450 should not update the ctime.
2454 Bugzilla : 15713/16362
2455 Description: Assertion in iopen_connect_dentry in 1.6.3
2456 Details : looking up an inode via iopen with the wrong generation number can
2457 populate the dcache with a disconneced dentry while the inode
2458 number is in the process of being reallocated. This causes an
2459 assertion failure in iopen since the inode's dentry list contains
2460 both a connected and disconnected dentry.
2464 Description: assertion failure in ldlm_handle2lock()
2465 Details : fix a race between class_handle_unhash() and class_handle2object()
2466 introduced in lustre 1.6.5 by bug 13622.
2471 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2472 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2477 Description: oops in sd_iostats_seq_show()
2478 Details : unloading/reloading the scsi low level driver triggers a kernel
2479 bug when trying to access the sd iostat file.
2484 Description: Kernel panics during QLogic driver reload
2485 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2486 patch, causing memory corruption.
2491 Description: journal_dev option does not work in b1_6
2492 Details : pass mount option during pre-mount.
2494 Severity : enhancement
2496 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2497 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2498 extent information of a file. It can be used to map logical blocks
2499 in a file to physical blocks in the block device.
2503 Description: LDLM soft lockups - improvement
2504 Details : It is be possible to send the lock handle along with each read
2505 or write request because the client is already doing a lock match
2506 itself so there isn't any reason the OST should have to re-do that
2510 Frequency : only on Cray X2
2512 Description: X2 build failures
2513 Details : fix build failures on Cray X2.
2517 Description: xid & resent requests
2518 Details : Initialize RPC XID from clock at startup (randomly if clock is
2521 Severity : enhancement
2523 Description: Add lustre_start utility to start or stop multiple Lustre servers
2528 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2529 Details : In case of memory pressure, list_del() can be called twice on
2530 req->rq_history_list, causing a kernel oops.
2534 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2535 Details : fix stack overflow in the distributed lock manager by defering
2536 export eviction after a failed AST to the elt thread instead of
2537 handling it in the dlm interpret routine.
2541 Description: Convert some comments to new format.
2542 Details : Update documenting comments to match doxygen conventions.
2546 Description: Grammar fixes.
2547 Details : A couple of trivial spelling fixes.
2551 Description: OSD_COUNTERS-mandatory
2552 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2553 to simplify the code.
2557 Description: simplify cmm_device freeing logic.
2558 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2562 Description: Add lockdep support to dt_object_operations locking interface.
2563 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2564 parameter indicating lock ordering. Update mdd code to use new
2569 Description: Introduce failloc constants for lockless IO tests.
2570 Details : Add two new failloc constants to test lockless IO. Only one of
2571 them in implemented---another is checked in yet to be landed
2576 Description: Add lockdep support for inode mutex.
2577 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2578 in the situations where Lustre has to lock more than one inode
2583 Description: Add optional invariants checking support.
2584 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2585 configure switch. This macro is to be used for consistency and
2586 sanity checks that are too expensive to be left in `production'
2591 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2592 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2597 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2598 Details : Use spin_lock_nested() in (the only) situation where more than
2599 one ldlm_lock is locked simultaneously. Also, fix possible
2600 dead-lock in ldlm_lock_change_resource() by enforcing particular
2605 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2606 Details : Instead of specifying each ldlm_lock call-back through separate
2607 parameter, wrap them into struct ldlm_callback_suite.
2611 Description: Kill join_lru obd method and its callers.
2612 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2613 in memory---a requirement implemented through ->o_join_lru() obd
2614 method in HEAD. Remove this method and its users.
2618 Description: Add asynchronous ldlm ENQUEUE completion handler.
2619 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2620 case that stock ldlm_completion_ast() cannot handle as it waits
2621 until lock is granted. Introduce new ldlm_completion_ast_async()
2622 for this. Also comment ldlm_completion_ast().
2626 Description: ldlm_error <-> errno conversion.
2627 Details : Add functions to map (rather arbitrary) between LDLM error codes
2628 and standard errno values. CLIO needs this to prevent LDLM specific
2629 constants from escaping ldlm and osc.
2633 Description: Kill unused ldlm_handle2lock_ns() function.
2634 Details : Kill unused ldlm_handle2lock_ns() function.
2638 Description: Add lu_ref support to ldlm_lock
2639 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2640 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2641 and ldlm_resource. LDLM interface has to be changed, because code that
2642 releases a reference on a lock, has to "know" what reference this is.
2643 In the most frequent case
2645 lock = ldlm_handle2lock(handle);
2647 LDLM_LOCK_PUT(lock);
2649 no changes are required. When any other reference (received _not_ from
2650 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2651 instead of LDLM_LOCK_PUT().
2653 Arguably, changes are pervasive, and interface requires some discipline
2654 for proper use. On the other hand, it was very instrumental in finding
2655 a few leaked lock references.
2659 Description: Add ldlm_lock_addref_try().
2660 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2661 attempts to addref a lock that might be being canceled
2666 Description: Add ldlm_weigh_callback().
2667 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2668 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2669 hard-coded `number of pages' logic.
2673 Description: Add lockdep annotations to llog code.
2674 Details : Use appropriately tagged _nested() locking calls in the places
2675 where llog takes more than one ->lgh_lock lock.
2679 Description: Add loi_kms_set().
2680 Details : Wrap kms updates into a helper function.
2684 Description: Constify instances of struct lsm_operations.
2685 Details : Constify instances of struct lsm_operations.
2689 Description: lu_conf support.
2690 Details : On a server, a file system object is uniquely identified
2691 by a fid, which is sufficient to locate and load all object
2692 state (inode). On a client, on the other hand, more data are
2693 necessary instantiate an object. Change lu_object_find() and
2694 friends to take additional `lu_conf' argument describing object.
2695 Typically this includes layout information.
2699 Description: lu_context fixes.
2700 Details : Introduce new lu_context functions that are needed on the client
2701 side, where some system threads (ptlrpcd) are shared by multiple
2702 modules, and so cannot be stopped during module shutdown.
2706 Description: Add start and stop methods to lu_device_type_operations.
2707 Details : Introduce two new methods in lu_device_type_operations, that are
2708 invoked when first instance of a given type is created and last one
2709 is destroyed respectively. This is need by CLIO.
2713 Description: Add lu_ref support to struct lu_device.
2714 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2715 track leaked references.
2719 Description: Introduce lu_kmem_descr.
2720 Details : lu_kmem_descr and its companion interface allow to create
2721 and destroy a number of kmem caches at once.
2725 Description: Fix lu_object finalization race.
2726 Details : Fix a race between lu_object_find() finding an object and its
2727 concurrent finalization. This race is (most likely) not possible
2728 on the server, but might happen on the client.
2732 Description: Introduce lu_ref interface.
2733 Details : lu_ref is a debugging module allowing to track references to
2734 a given object. It is quite cpu expensive, and has to be
2735 explicitly enabled with --enable-lu_ref. See usage description
2740 Description: Factor lu_site procfs stats into a separate function.
2741 Details : Separate lu_site stats printing code into a separate function
2742 that can be reused on a client.
2746 Description: Constify instances of struct {lu,dt,md}_device_operations.
2747 Details : Constify instances of struct {lu,dt,md}_device_operations.
2751 Description: Introduce struct md_site and move meta-data specific parts of
2752 struct lu_site here.
2753 Details : Move md-specific fields out of struct lu_site into special struct
2754 md_site, so that lu_site can be used on a client.
2758 Description: Kill mdd_lov_destroy().
2759 Details : Remove unused mdd code.
2763 Description: Add st_block checking to multistat.c.
2764 Details : Add st_block checking to multistat.c.
2768 Description: Add lu_ref support to struct obd_device.
2769 Details : Add lu_ref tracking to obd_device.
2773 Description: Kill obd_set_fail_loc().
2774 Details : Remove unused code.
2778 Description: Add special type for ptlrpc_request interpret functions.
2779 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2780 there. Actual usage will be in CLIO.
2784 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2785 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2786 doesn't work with lockdep.
2790 Description: Add rwv.c test program.
2791 Details : New testing program exercising readv(2) and writev(2) (Qian).
2795 Description: Add sendfile.c test program.
2796 Details : New testing program exercising sendfile(2) (Jay).
2800 Description: Ratelimit a message that can be very frequent.
2801 Details : Ratelimit a memory allocation failure message that can
2806 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2807 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2811 Description: Liblustre build fixes.
2812 Details : Liblustre build fixes.
2816 Description: libcfs: add cfs_{need,cond}_resched() interface.
2817 Details : libcfs: add cfs_{need,cond}_resched() definition and
2818 implementations for Linux, NT, and liblustre.
2820 Severity : enhancement
2822 Description: More exported tunables for mballoc
2823 Details : Add support for tunable preallocation window and new tunables for
2824 large/small requests
2828 Description: Detect corruption of block bitmap and checking for preallocations
2829 Details : Checks validity of on-disk block bitmap. Also it does better
2830 checking of number of applied preallocations. When corruption is
2831 found, it turns filesystem readonly to prevent further corruptions.
2835 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2836 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2837 window to 0. If it does hit cache miss, and read-pattern is still
2838 stride-io mode, does not reset the stride window, but also does not
2839 increase the stride window length in this case.
2843 Frequency : only for big-endian servers
2844 Description: Check if system is big-endian while mounting fs with extents feature
2845 Details : Mounting a filesystem with extents feature will fail on big-endian
2846 systems since ext3-based ldiskfs is not supported on big-endian
2847 systems. This can be over-riden with "bigendian_extents" mount option.
2849 Severity : enhancement
2851 Description: The root squash functionality
2852 Details : A security feature, which is to prevent users from being able
2853 to mount lustre on their desktop, run as root, and delete
2854 all of the files in the filesystem. The goal is accomplished by
2855 remapping user id (UID) and group id (GID) of the root user to
2856 a UID and GID specified by the system administartor via Lustre
2857 configuration management server (MGS). The functionality also
2858 allows to specify sets of clients for which the remapping does
2863 Description: Excessive recovery window
2864 Details : With AT enabled, the recovery window can be excessively long (6000+
2865 seconds). To address this problem, we no longer use
2866 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2867 timeout no longer depends on the service time, it is set to
2868 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2869 time via pb_service_time.
2873 Description: Watchdog triggered on MDS failover
2874 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2875 the OSTs know that the MDS "UUID" can be reused for the same export
2876 from a different NID, so we do not need to wait for the export to be
2880 Frequency : rare, only if using MMP with Linux RAID
2882 Description: MMP doesn't work with Linux RAID
2883 Details : While using HA for Lustre servers with Linux RAID, it is possible
2884 that MMP will not detect multiple mounts. To make this work we
2885 need to unplug the device queue in RAID when the MMP block is being
2886 written. Also while reading the MMP block, we should read it from
2887 disk and not the cached one.
2889 Severity : enhancement
2891 Description: open file using fid
2892 Details : A file can be opened using just its fid, like
2893 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2896 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2899 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2900 Details : add proper locking for creation of jbd_4k slab cache
2904 Description: MMP check in ext3_remount() fails without displaying any error
2905 Details : When multiple mount protection fails during remount, proper error
2908 Severity : enhancement
2910 Description: Allow stripe size to be up to 4G-64k
2911 Details : Fix math logic to allow large stripe sizes.
2915 Description: add check for >8TB ldiskfs filesystems
2916 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2917 Don't allow >8TB ldiskfs filesystems to be mounted without
2918 force_over_8tb mount option
2923 Descriptoin: OSS complains "too many blocks" and crashes
2924 Details : In filter_commitrw_write():
2925 1. lquota_chkquota();
2926 2. obdo_from_inode();
2927 3. lquota_pending_commit();
2928 "2." has changed uid and gid in this case, then caused
2929 lquota_chkquota() and lquota_pending_commit() to use
2930 different uid and gid. That is the root of the bug.
2934 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2935 Details : Detect and print out the 'stripe count = -1' case properly.
2939 Description: Out or order replies might be lost on replay
2940 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2941 request with tid smaller than one currently being added, add it
2942 to the start, not end of the list.
2944 --------------------------------------------------------------------------------
2946 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2948 * Support for kernels:
2949 2.6.5-7.283 (SLES 9),
2950 2.6.9-55.EL (RHEL 4),
2951 2.6.16.46-0.14 (SLES 10),
2952 2.6.18.8 vanilla (kernel.org)
2953 * Client support for unpatched kernels:
2954 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2955 2.6.16 - 2.6.22 vanilla (kernel.org)
2956 * Due to recently discovered recovery problems, we do not recommend
2957 using patchless RHEL 4 clients with this or any earlier release.
2958 * Recommended e2fsprogs version: 1.39.cfs8
2959 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2960 * Starting with this release, the ldiskfs backing filesystem required
2961 by Lustre is now in its own package, lustre-ldiskfs. This package
2962 should be installed. It is versioned separately from Lustre and
2963 may be released separately in future.
2967 Description: block reactivating mgc import until all deactivates complete
2968 Details : Fix race when failing back MDT/MGS to itself (testing)
2971 Frequency : at statup only
2973 Description: mds_lov_synchronize race leads to various problems
2974 Details : simultaneous MDT->OST connections at startup can cause the
2975 sync to abort, leaving the OSC in a bad state.
2977 Severity : enhancement
2979 Description: add optional extra BUILD_VERSION info
2980 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2981 to override the lustre version.
2984 Frequency : 2.6.18 servers only
2986 Description: ll_kern_mount() doesn't release the module reference
2987 Details : The ldiskfs module reference count never drops down to 0
2988 because ll_kern_mount() doesn't release the module reference.
2993 Description: server LBUG when using old ost_num_threads parameter
2994 Details : Accept the old ost_num_threads parameter but warn that it
2995 is deprecated, and fix an off-by-one error that caused an LBUG.
3000 Description: Transient SCSI error results in persistent IO issue
3001 Details : iobuf->dr_error is not reinitialized to 0 between two
3005 Frequency : sometimes when underlying device returns I/O errors
3007 Description: OSTs not going read-only during write failures
3008 Details : OSTs are not remounted read-only when the journal commit threads
3009 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3010 instead of the ext3 wrappers.
3014 Description: poor connect scaling with increasing client count
3015 Details : Don't run filter_grant_sanity_check for more than 100 exports
3016 to improve scaling for large numbers of clients.
3019 Frequency : SLES10 only
3021 Description: sanity-quota.sh quotacheck failed: rc = -22
3022 Details : Quotas cannot be enabled on SLES10.
3025 Frequency : liblustre clients only
3027 Description: getdirentries does not give error when run on compute nodes
3028 Details : getdirentries does not fail when the size specified as an argument
3029 is too small to contain at least one entry
3031 Severity : enhancement
3033 Description: Add LNET router traceability for debug purposes
3034 Details : If a checksum failure occurs with a router as part of the
3035 IO path, the NID of the last router that forwarded the bulk data
3036 is printed so it can be identified.
3041 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3042 Details : Due to a race condition, liblustre clients were occasionally
3043 evicted incorrectly.
3045 Severity : enhancement
3047 Description: lfs setstripe use optional parameters instead of postional
3050 Severity : enhancement
3052 Description: Nanosecond timestamp support for ldiskfs
3053 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3054 resolution timestamps. There is not yet support for this at
3055 the Lustre filesystem level.
3058 Frequency : during server recovery
3060 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3061 Details : request with rq_no_resend flag not awake l_wait_event if they get
3065 Frequency : nfs export on patchless client
3067 Description: connectathon hang when test nfs export over patchless client
3068 Details : Disconnected dentry cannot be found with lookup, so we do not need
3069 to unhash it or make it invalid
3072 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3073 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3074 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3079 Description: open req refcounting wrong on reconnect
3080 Details : If reconnect happened between getting open reply from server and
3081 call to mdc_set_replay_data in ll_file_open, we will schedule
3082 replay for unreferenced request that we are about to free.
3083 Subsequent close will crash in variety of ways.
3084 Check that request is still eligible for replay in
3085 mdc_set_replay_data().
3090 Description: disable writes to filesystem when reading health_check file
3091 Details : the default for reading the health_check proc file has changed
3092 to NOT do a journal transaction and write to disk, because this
3093 can cause reads of the /proc file to hang and block HA state
3094 checking on a healthy but otherwise heavily loaded system. It
3095 is possible to return to the previous behaviour during configure
3096 with --enable-health-write.
3098 Severity : enhancement
3100 Description: 64-bit inode version
3101 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3102 to the inode. This will be required for version-based recovery.
3107 Description: MDS fails to start if a duplicate client export is detected
3108 Details : in some rare cases it was possible for a client to connect to
3109 an MDS multiple times. Upon recovery the MDS would detect this
3110 and fail during startup. Handle this more gracefully.
3112 Severity : enhancement
3114 Description: Add -o localflock option to simulate old noflock
3116 Details : This will achieve local-only flock/fcntl locks
3122 Description: log_commit_thread vs filter_destroy race leads to crash
3123 Details : Take import reference before releasing llog record semaphore
3128 Description: Wrong request locking in request set processing
3129 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3130 imp_delayed_list, in this place should be used imp_lock.
3133 Frequency : when reconnection
3135 Description: Grant Leak when osc reconnect to OST
3136 Details : When osc reconnect ost, OST(filter) should check whether it
3137 should grant more space to client by comparing fed_grant and
3138 cl_avail_grant, and return the granted space to client instead
3139 of "new granted" space, because client will call osc_init_grant
3140 to update the client grant space info.
3143 Frequency : when client reconnect to OST
3145 Description: Grant Leak when osc do resend and replay bulk write
3146 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3147 bulk write request, because the grant info while be sync between
3148 OSC and OST when reconnect, and we should ignore the grant info
3149 these of resend/replay write req.
3154 Description: Grant space more than available left space sometimes.
3155 Details : When then OST is about to be full, if two bulk writing from
3156 different clients came to OST. Accord the avaliable space of the
3157 OST, the first req should be permitted, and the second one
3158 should be denied by ENOSPC. But if the seconde arrived before
3159 the first one is committed. The OST might wrongly permit second
3160 writing, which will cause grant space > available space.
3163 Frequency : when client is evicted
3165 Description: Grant might be wrongly erased when osc is evicted by OST
3166 Details : when the import is evicted by server, it will fork another
3167 thread ptlrpc_invalidate_import_thread to invalidate the
3168 import, where the grant will be set to 0. While the original
3169 thread will update the grant it got when connecting. So if
3170 the former happened latter, the grant will be wrongly errased
3171 because of this race.
3176 Description: Checking Stale with correct fid
3177 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3178 to check whether it is stale, because sometimes, we want the
3179 enqueue happened anyway, and op_data.fid2 will not be initialized.
3181 Severity : enhancement
3183 Description: update patchless client
3184 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3187 Frequency : only with 2.4 kernel
3189 Description: random memory corruption
3190 Details : size of struct ll_inode_info is to big for union inode.u and this
3191 can be cause of random memory corruption.
3196 Description: Memory leak in recovery
3197 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3198 It should also check obd_fail before fsfilt_start, otherwise if
3199 fsfilt_start return -EROFS,(failover mds during mds recovery).
3200 then the req will return with repmsg->transno = 0 and rc = EROFS.
3201 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3202 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3203 be freed no matter whether fsfilt_commit success or not.
3206 Frequency : only with huge count clients
3208 Description: Prevents from taking the superblock lock in llap_from_page for
3210 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3211 for llap_from_page call in ll_removepage prevents from taking the
3212 superblock lock for a soon died page.
3217 Description: Not check open intent error before release open handle
3218 Details : in some rare cases, the open intent error is not checked before
3219 release open handle, which may cause
3220 ASSERTION(open_req->rq_transno != 0), because it tries to release
3221 the failed open handle.
3226 Description: Set cat log bitmap only after create log success.
3227 Details : in some rare cases, the cat log bitmap is set too early. and it
3228 should be set only after create log success.
3232 Description: Accessing a block bevice can re-enable I/O when Lustre is
3233 tearing down a device.
3234 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3239 Description: service threads may hog cpus when there are a lot of requests
3240 Details : Insert cond_resched to give other threads a chance to use some CPU
3245 Description: the cat log was not initialized in recovery
3246 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3247 unlink log on mds will not be initialized until mds post
3248 recovery. And also in mds post recovery, the unlink log will
3249 initialization will be done asynchronausly, so there will be race
3250 between add unlink log and unlink log initialization.
3254 Description: brw_stats were being printed incorrectly
3255 Details : brw_stats were being printed as log2 but all of them were not
3256 recorded as log2. Also remove some code duplication arising from
3257 filter_tally_{read,write}.
3261 Frequency : rare, only in recovery.
3262 Description: ASSERTION(req->rq_type != LI_POISON) failed
3263 Details : imp_lock should be held while iterating over imp_sending_list for
3264 prevent destroy request after get timeout in ptlrpc_queue_wait.
3268 Description: replay-single.sh test 52 fails
3269 Details : A lock's skiplist need to be cleanup when it being unlinked
3270 from its resource list.
3274 Description: Short directio read returns full requested size rather than
3276 Details : Direct I/O operations should return actual amount of bytes
3277 transferred rather than requested size.
3279 Severity : enhancement
3281 Description: metadata RPC reduction (e.g. for rm performance)
3282 Details : decrease the amount of synchronous RPC between clients and servers
3283 by canceling conflicing lock before the operation on the client
3284 and packing thier handles into the main operation RPC to server.
3286 Severity : enhancement
3288 Description: Async OSC create to avoid the blocking unnecessarily.
3289 Details : If an OST has no remaining object, system will block on the
3290 creation when it needs to create a new object on this OST. Now,
3291 ways use pre-created objects when available, instead of blocking on
3292 an empty osc while others are not empty. If we must block, we
3293 block for the shortest possible period of time.
3297 Description: improve handling recoverable errors
3298 Details : if request processig with error which can be recoverable on server
3299 request should be resend, otherwise page released from cache and
3302 Severity : enhancement
3304 Description: refine locking for avoid write wrong info into lov_objid file
3305 Details : fix possible races with add new target and write/update data in
3308 --------------------------------------------------------------------------------
3310 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3315 Frequency : on some architectures
3317 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3318 Details : Uninitialized flags sometimes cause configuration commands to
3322 Frequency : patchless clients only
3324 Description: missing __iget() symbol export
3325 Details : The __iget() symbol export is missing. To avoid the need for
3326 this on patchless clients the deathrow inode reaper is turned
3327 off, and we depend on the VM to clean up old inodes. This
3328 dependency was during via the fix for bug 12181.
3330 --------------------------------------------------------------------------------
3332 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3334 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3335 INTEROPERATE with older versions automatically. In many cases a
3336 special upgrade step is needed. Please read the
3337 user documentation before upgrading any part of a live system.
3338 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3339 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3340 * WARNING: Lustre configuration and startup changes are required with
3341 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3345 Severity : enhancement
3347 Description: MountConf
3348 Details : Lustre configuration is now managed via mkfs and mount
3349 commands instead of lmc and lconf. New obd types (MGS, MGC)
3350 are added for dynamic configuration management. See
3351 https://mail.clusterfs.com/wikis/lustre/MountConf for
3354 Severity : enhancement
3356 Description: dynamic OST addition
3357 Details : OSTs can now be added to a live filesystem
3359 Severity : enhancement
3361 Description: startup order invariance
3362 Details : MDTs and OSTs can be started in any order. Clients only
3363 require the MDT to complete startup.
3365 Severity : enhancement
3367 Description: parallel, asynchronous orphan cleanup
3368 Details : orphan cleanup is now performed in separate threads for each
3369 OST, allowing parallel non-blocking operation.
3371 Severity : enhancement
3373 Description: optimized stripe assignment
3374 Details : stripe assignments are now made based on ost space available,
3375 ost previous usage, and OSS previous usage, in order to try
3376 to optimize storage space and networking resources.
3378 Severity : enhancement
3380 Description: Permanently set tunables
3381 Details : All writable /proc/fs/lustre tunables can now be permanently
3382 set on a per-server basis, at mkfs time or on a live system.
3384 Severity : enhancement
3386 Description: Lustre message v2
3387 Details : Add lustre message format v2.
3389 Severity : enhancement
3391 Description: client OST exclusion list
3392 Details : Clients can be started with a list of OSTs that should be
3393 declared "inactive" for known non-responsive OSTs.
3397 Description: SPEC SFS validation failure on NFS v2 over lustre.
3398 Details : Changes the blocksize for regular files to be 2x RPC size,
3399 and not depend on stripe size.
3401 Severity : enhancement
3403 Description: Multiple MD RPCs in flight.
3404 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3405 To avoid overly-overloading MDS, introduce a limit on number of
3406 MDS RPCs in flight for a single client and add /proc controls
3407 to adjust this limit.
3409 Severity : enhancement
3411 Description: client read/write statistics
3412 Details : Add client read/write call usage stats for performance
3413 analysis of user processes.
3414 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3415 file access. extents_stats shows chunk size distribution.
3416 extents_stats_per_process show chunk size distribution per
3419 Severity : enhancement
3421 Description: mds statistics
3422 Details : Add detailed mds operations statistics in
3423 /proc/fs/lustre/mds/*/stats.
3427 Description: Failure of copying files with lustre special EAs.
3428 Details : Client side always return success for setxattr call for lustre
3429 special xattr (currently only "trusted.lov").
3434 Description: Refcount LNET uuids
3435 Details : The global LNET uuid list grew linearly with every startup;
3436 refcount repeated list entries instead of always adding to
3439 Severity : enhancement
3441 Description: Dynamic service threads
3442 Details : Within a small range, start extra service threads
3443 automatically when the request queue builds up.
3446 Frequency : mixed-endian client/server environments
3448 Description: mixed-endian crashes
3449 Details : The new msg_v2 system had some failures in mixed-endian
3452 Severity : enhancement
3454 Description: Easy OST removal
3455 Details : OSTs can be permanently deactivated with e.g. 'lctl
3456 conf_param lustre-OST0001.osc.active=0'
3458 Severity : enhancement
3460 Description: MGS proc entries
3461 Details : Added basic proc entries for the MGS showing what filesystems
3464 Severity : enhancement
3466 Description: provide MGS failover
3467 Details : Added config lock reacquisition after MGS server failover.
3469 Severity : enhancement
3471 Description: add Linux 2.4 support
3472 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3476 Description: a large application tries to do I/O to the same resource and dies
3477 in the middle of it.
3478 Details : Check the req->rq_arrival time after the call to
3479 ost_brw_lock_get(), but before we do anything about
3480 processing it & sending the BULK transfer request. This
3481 should help move old stale pending locks off the queue as
3482 quickly as obd_timeout.
3485 Frequency : when an incorrect nid is specified during startup
3487 Description: ptlrpc connect to non-existant node causes kernel crash
3488 Details : LNET can't be re-entered from an event callback, which
3489 happened when we expire a message after the export has been
3490 cleaned up. Instead, hand the zombie cleanup off to another
3493 Severity : enhancement
3495 Description: plain/inodebits lock performance improvement
3496 Details : Grouping plain/inodebits in granted list by their request modes
3497 and bits policy, thus improving the performance of search through
3501 Frequency : only if OST filesystem is corrupted
3503 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3504 Details : for a short time RPCs with bulk IO are in the replay list,
3505 but replay of bulk IOs is unimplemented. If the OST filesystem
3506 is corrupted due to disk cache incoherency and then replay is
3507 started it is possible to trip an assertion. Avoid putting
3508 committed RPCs into the replay list at all to avoid this issue.
3511 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3513 Description: System hang on startup
3514 Details : This bug allowed the liblustre (e.g. catamount) client to
3515 return to the app before handling all startup RPCs. This
3516 could leave the node unresponsive to lustre network traffic
3517 and manifested as a server ptllnd timeout.
3519 Severity : enhancement
3521 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3522 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3523 boolean to control whether to print peer debug info when a
3524 client's RPC times out.
3527 Frequency : only for kernels with patches from Lustre below 1.4.3
3529 Description: Remove old rdonly API
3530 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3533 Frequency : only for devices with external journals
3535 Description: Set external device read-only also
3536 Details : During a commanded failover stop, we set the disk device
3537 read-only while the server shuts down. We now also set any
3538 external journal device read-only at the same time.
3541 Frequency : when upgrading from 1.4 while trying to change parameters
3543 Description: The wrong (new) MDC name was used when setting parameters for
3544 upgraded MDT's. Also allows changing of OSC (and MDC)
3545 parameters if --writeconf is specified at tunefs upgrade time.
3548 Frequency : when setting specific ost indicies
3550 Description: QOS code breaks on skipped indicies
3551 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3552 created with --index need not be sequential.
3554 Severity : enhancement
3556 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3557 Details : The uninit_groups feature works in conjunction with the kernel
3558 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3559 up the pass1 processing of e2fsck. This is a read-only feature
3560 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3561 mount filesystems that have had this feature enabled.
3563 Severity : enhancement
3565 Description: Improve multi-block allocation algorithm to avoid fragmentation
3566 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3567 allocation locality and avoid filesystem fragmentation.
3569 ------------------------------------------------------------------------------
3571 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3573 * Support for kernels:
3574 2.4.21-47.0.1.EL (RHEL 3)
3575 2.6.5-7.283 (SLES 9)
3576 2.6.9-42.0.10.EL (RHEL 4)
3577 2.6.12.6 vanilla (kernel.org)
3578 2.6.16.27-0.9 (SLES 10)
3579 * Recommended e2fsprogs version: 1.39.cfs5
3581 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3585 Frequency : occasional, depends on client load and configuration
3586 Bugzilla : 12181, 12203
3587 Description: data loss for recently-modified files
3589 Details : In some cases it is possible that recently written or created
3590 files may not be written to disk in a timely manner (this should
3591 normally be within 30s unless client IO load is very high).
3592 The problem appears as zero-length files or files that are a
3593 multiple of 1MB in size after a client crash or client eviction
3594 that are missing data at the end of the file.
3596 This problem is more likely to be hit on clients where files are
3597 repeatedly created and unlinked in the same directory, clients
3598 have a large amount of RAM, have many CPUs, the filesystem has
3599 many OSTs, the clients are rebooted frequently, and/or the files
3600 are not accessed by other nodes after being written.
3602 The presence of the problem can be detected by looking at
3603 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3604 smaller than the second (nr_unused) then dirty files will not
3605 be flushed automatically to disk. "sync; sleep 10" should be
3606 run several times on the node before unmounting it to update
3607 Lustre (this is also safe to run on nodes without this problem).
3609 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3610 that can cause this same problem, so customers using that kernel
3611 also need to update the kernel in addition to Lustre. In order
3612 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3614 It is normal that files written just before a client crash (less
3615 than 30s) may not yet have been flushed to disk, even for local
3619 Frequency : frequent on thin XT3 nodes
3621 Description: UUID collision on thin XT3 Linux nodes
3622 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3623 randomly, since we relied on an insufficiently-seeded PRNG.
3628 Description: OSS hangs after "All ost request buffers busy"
3629 Details : A deadlock between quota and journal operations caused OSS
3630 hangs after printing "All ost request buffers busy."
3633 Frequency : always on liblustre builds
3635 Description: Cleanup compiler warnings on liblustre
3638 Frequency : always on liblustre builds on XT3
3640 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3644 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3645 > 32000 subdirectories
3646 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3647 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3648 subdirectory count crosses 32000. This will aid e2fsck to
3649 correctly handle more than 32000 subdirectories.
3652 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3654 Description: System hang on startup
3655 Details : This bug allowed the liblustre (e.g. catamount) client to
3656 return to the app before handling all startup RPCs. This
3657 could leave the node unresponsive to lustre network traffic
3658 and manifested as a server ptllnd timeout.
3660 Severity : enhancement
3662 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3663 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3664 boolean to control whether to print peer debug info when a
3665 client's RPC times out.
3670 Description: make O_SYNC working on 2.6 kernels
3671 Details : 2.6 kernels use different method for mark pages for write,
3672 so need add a code to lustre for O_SYNC work.
3677 Description: Failure to close file and release space on NFS
3678 Details : Put inode details into lock acquired in ll_intent_file_open.
3679 Use mdc_intent_lock in ll_intent_open to properly
3680 detect all kind of errors unhandled by mdc_enqueue.
3685 Description: proc file read during shutdown sometimes raced obd removal,
3687 Details : Add lock to prevent obd access after proc file removal.
3690 Frequency : Only for files larger than 4GB on 32-bit clients.
3692 Description: improperly doing page alignment of locks
3693 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3694 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3699 Description: RPCs being resent when they shouldn't be
3700 Details : Some RPCs that should not be resent are being resent. This
3701 can cause inconsistencies in the RPC state machine. Do not
3702 resend such requests.
3705 Frequency : rare, only with NFS export
3707 Description: Crash on NFS re-export node
3708 Details : under very unusual load conditions an assertion is hit in
3709 ll_intent_file_open()
3712 Frequency : only if OST filesystem is corrupted
3714 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3715 Details : for a short time RPCs with bulk IO are in the replay list,
3716 but replay of bulk IOs is unimplemented. If the OST filesystem
3717 is corrupted due to disk cache incoherency and then replay is
3718 started it is possible to trip an assertion. Avoid putting
3719 committed RPCs into the replay list at all to avoid this issue.
3724 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3725 Details : Large single O_DIRECT read and write calls can fail to allocate
3726 a sufficiently large buffer to process the request. In case of
3727 allocation failure the allocation is retried with a smaller
3728 buffer and broken into smaller requests.
3730 Severity : enhancement
3732 Description: Add -o localflock option to simulate old noflock behaviour.
3733 Details : This will achieve local-only flock/fcntl locks coherentness.
3738 Description: versioning check is incomplete
3739 Details : Checking the version difference of client vs. server, report
3740 error if the gap is too big.
3745 Description: add support PG_writeback bit
3746 Details : add support for PG_writeback bit for Lustre, for more carefull
3747 work with page cache in 2.6 kernel. This also fix some deadlocks
3748 and remove hack for work O_SYNC with 2.6 kernel.
3750 Severity : enhancement
3752 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3753 Details : The uninit_groups feature works in conjunction with the kernel
3754 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3755 up the pass1 processing of e2fsck. This is a read-only feature
3756 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3757 mount filesystems that have had this feature enabled.
3759 Severity : enhancement
3761 Description: Improve multi-block allocation algorithm to avoid fragmentation
3762 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3763 allocation locality and avoid filesystem fragmentation.
3765 ------------------------------------------------------------------------------
3767 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3769 * Support for kernels:
3770 2.6.9-42.0.3.EL (RHEL 4)
3771 2.6.5-7.276 (SLES 9)
3772 2.4.21-47.0.1.EL (RHEL 3)
3773 2.6.12.6 vanilla (kernel.org)
3774 2.6.16.21-0.8 (SLES10)
3775 * Recommended e2fsprogs version: 1.39.cfs2-0
3777 * The backwards-compatible /proc/sys/portals symlink has been removed
3778 in this release. Before upgrading, please ensure that you change
3779 any configuration scripts or /etc/sysctl.conf files that access
3780 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3781 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3782 in advance of the upgrade on any system running Lustre 1.4.6 or
3783 newer, since /proc/sys/lnet was added in that version.
3784 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3788 Frequency : only when quota is used
3790 Description: avoid scanning export list for quota master
3791 Details : Change the algorithms to avoid scanning export list in order
3792 to improve the efficiency.
3795 Frequency : MDS failover only, very rarely
3797 Description: "went back in time" messages on mds failover
3798 Details : The greatest transno may be lost when the current operation
3799 finishes with an error (transno==0) and the client's last_rcvd
3800 record is over-written. Save the greatest transno in the
3801 mds_last_transno for this case.
3804 Frequency : always for specific kernels and striping counts
3806 Description: client may get "Matching packet too big" without ACL support
3807 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3808 when trying to access files in certain configurations. The
3809 clients should in fact be denied when mounting because they do
3810 not understand ACLs.
3813 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3815 Description: many clients connecting with IO in progress causes connect timeouts
3816 Details : Avoid synchronous journal commits to avoid delays caused by many
3817 clients connecting/disconnecting when bulk IO is in progress.
3818 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3819 OST_IO_PORTAL to avoid delays behind potentially many pending
3823 Frequency : occasionally with multiple writers to a single file
3825 Description: shared writes to file may result in wrong size reported by stat()
3826 Details : Allow growing of kms when extent lock is cancelled
3829 Frequency : always with random mmap IO to multi-striped file
3831 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3832 Details : Only if the hole is at the end of OST object so that kms is too
3833 small. Fix is to increase kms accordingly in ll_nopage.
3836 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3838 Description: writes to a missing object would leak memory on the OST
3839 Details : If there is an inconsistency between the MDS and OST filesystems,
3840 such that the MDS references an object that doesn't exist, writes
3841 to that object will leak memory due to incorrect cleanup in the
3842 error handling path, eventually running out of memory on the OST.
3847 Description: Creating too long symlink causes lustre errors
3848 Details : Check symlink and name lengths before sending requests to MDS.
3851 Frequency : only if flock is enabled (not on by default)
3853 Description: posix locks not released on fd closure on 2.6.9+
3854 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3855 kernels, this caused such locks not to be released on fd close and
3856 then assertions on fs unmount about still used locks.
3859 Frequency : MDS failover only, very rarely
3861 Description: clients may get ASSERTION(granted_lock != NULL)
3862 Details : When request was taking a long time, and a client was resending
3863 a getattr by name lock request. The were multiple lock requests
3864 with the same client lock handle and
3865 mds_getattr_name->fixup_handle_for_resent_request found one of the
3866 lock handles but later failed with ASSERTION(granted_lock != NULL).
3871 Description: handle->h_buffer_credits > 0, assertion failure
3872 Details : h_buffer_credits is zero after truncate, causing assertion
3873 failure. This patch extends the transaction or creates a new
3877 Frequency : NFS re-export or patchless client
3878 Bugzilla : 11179, 10796
3879 Description: Crash on NFS re-export node (__d_move)
3880 Details : We do not want to hash the dentry if we don't have a lock.
3881 But if this dentry is later used in d_move, we'd hit uninitialised
3882 list head d_hash, so we just do this to init d_hash field but
3883 leave dentry unhashed.
3886 Frequency : NFS re-export or patchless client
3888 Description: NFS exports has problem with symbolic link
3889 Details : lustre client didn't properly install dentry when re-exported
3890 to NFS or running patchless client.
3893 Frequency : NFS re-export or patchless client
3895 Description: Various nfs/patchless fixes.
3896 Details : fixes reuse disconected alias for lookup process - this fixes
3897 warning "find_exported_dentry: npd != pd",
3898 fix permission error with open files at nfs.
3899 fix apply umask when do revalidate.
3902 Frequency : occasional
3904 Description: Crash on NFS re-export node
3905 Details : calling clear_page() on the wrong pointer triggered oops in
3906 generic_mapping_read().
3909 Frequency : rarely, using O_DIRECT IO
3911 Description: unaligned directio crashes client with LASSERT
3912 Details : check for unaligned buffers before trying any requests.
3915 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3917 Description: stale data returned from RAID cache
3918 Details : If only a small amount of IO is done to the RAID device before
3919 reading it again it is possible to get stale data from the RAID
3920 cache instead of reading it from disk.
3923 Frequency : always for sles10 kernel
3925 Description: sles10 support
3926 Details : ll_follow_link: compile fixes and using of nd_set_link
3927 under newer kernels.
3930 Frequency : depends on arch, kernel and compiler version, always on sles10
3933 Description: recursive or deep enough symlinks cause stack overflow
3934 Details : getting rid of large stack-allocated variable in
3938 Frequency : depends on hardware
3940 Description: lustre write performance loss in the SLES10 kernel
3941 Details : the performance loss is caused by using of write barriers in the
3942 ext3 code. The SLES10 kernel turns barrier support on by
3943 default. The fix is to undo that change for ldiskfs.
3945 ------------------------------------------------------------------------------
3947 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3949 * Support for kernels:
3950 2.6.9-42.0.3EL (RHEL 4)
3951 2.6.5-7.276 (SLES 9)
3952 2.4.21-47.0.1.EL (RHEL 3)
3953 2.6.12.6 vanilla (kernel.org)
3957 Frequency : quota enabled and large files being deleted
3959 Description: releasing more than 4GB of quota at once hangs OST
3960 Details : If a user deletes more than 4GB of files on a single OST it
3961 will cause the OST to spin in an infinite loop. Release
3962 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3967 Description: statfs data retrieved from /proc may be stale or zero
3968 Details : When reading per-device statfs data from /proc, in the
3969 {kbytes,files}_{total,free,avail} files, it may appear
3970 as zero or be out of date.
3973 Frequency : systems with MD RAID1 external journal devices
3975 Description: lconf's call to blkid is confused by RAID1 journal devices
3976 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3977 of returning all block devices that match the journal UUID.
3980 Frequency : always, for aggregate stripe size over 4GB
3982 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3983 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3984 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3987 Frequency : always if "lfs find" used on a local file/directory
3989 Description: "lfs find" segfaults if used on a local file/directory
3990 Details : The case where a directory component was not specified wasn't
3991 handled correctly. Handle this properly.
3994 Frequency : always on ppc64
3996 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3997 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3998 perform endianness conversion.
4001 Frequency : rarely (truncate to non-zero file size after write under load)
4002 Bugzilla : 10730, 10687
4003 Description: Files padded with zeros to next 4K multiple
4004 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4005 it is possible that files that are truncated to a non-zero size
4006 immediately after being written are filled with zero bytes beyond
4007 the truncated size. No file data is lost.
4009 Severity : enhancement
4011 Description: Allow recovery/failover for liblustre clients.
4012 Details : liblustre clients were unaware of failover configurations until
4015 Severity : enhancement
4017 Description: user file locks should fail when not mounting with flock option
4018 Details : Set up an error-returning stub in ll_file_operations.lock field
4019 to prevent incorrect behaviour when client is mounted without
4020 flock option. Also, set up properly f_op->flock field for
4024 Frequency : always on ia64
4026 Description: "lfs df" loops on printing out MDS statfs information
4027 Details : The obd_ioctl_data was not initialized and in some systems
4028 this caused a failure during the ioctl that did not return
4029 an error. Initialize the struct and return an error on failure.
4032 Frequency : SLES 9 only
4034 Description: Error of copying files with lustre special EAs as root
4035 Details : Client side always return success for setxattr call for lustre
4036 special xattr (currently only "trusted.lov").
4039 Frequency : rarely on clusters with both ia64+i386 clients
4041 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4042 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4043 clients are doing concurrent writes to the same file it is
4044 possible that the ia64 clients may LASSERT because the OST
4045 extent locks are not PAGE_SIZE aligned. Ensure that grown
4046 locks are always aligned on the request boundary.
4049 Frequency : specific use, occasional
4051 Description: Overwriting in use executable truncates on-disk binary image
4052 Details : If one node attempts to overwrite an executable in use by
4053 another node, we now correctly return ETXTBSY instead of
4054 truncating the file.
4056 Severity : enhancement
4058 Description: Async OSC create to avoid the blocking unnecessarily.
4059 Details : If an OST has no remaining object, system will block on the creating
4060 when need to create a new object on this OST. Now, ways use
4061 pre-created objects when available, instead of blocking on an
4062 empty osc while others are not empty. If we must block, we block
4063 for the shortest possible period of time.
4068 Description: chmod on Lustre root is propagated to other clients
4069 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4070 invalid by the follow_mount time.
4073 Frequency : liblustre clients only
4075 Description: Race in 'instant cancel' lock handling could lead to such locks
4076 never to be granted in case of SMP MDS
4077 Details : Do not destroy not yet granted but cbpending locks in
4081 Frequency : replay/resend of open
4083 Description: non null lock assetion failure in mds_intent_policy
4084 Details : Trying to replay/resend lockless open requests resulted in
4085 mds_open() returning 0 with no lock. Now it sets a flag if
4086 a lock is going to be returned.
4088 Severity : enhancement
4090 Description: Checksum enhancements
4091 Details : New checksum enhancements allow for resending RPCs that failed
4094 Severity : enhancement
4096 Description: Tunables on number of dirty pages in cacche
4097 Details : Allow to set limit on number of dirty pages cached.
4102 Description: client crash on unmount - lock still has references
4103 Details : In some error handling cases it was possible to leak a lock
4104 reference on a client while accessing a file. This was not
4105 harmful to the client during operation, but would cause the
4106 client to crash when the filesystem is unmounted.
4109 Frequency : specific case, rare
4111 Description: ETXTBSY on mds though file not in use by client
4112 Details : ETXTBSY is no longer incorrectly returned when attempting to
4113 chmod or chown a directory that the user previously tried to
4114 execute or a currently-executing binary.
4117 Frequency : extremely rare except on liblustre-based clients
4119 Description: Lustre space not freed when files are deleted
4120 Details : Clean up open-unlinked files after client eviction. Previously
4121 the unlink was skipped and the files remained as orphans.
4126 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4127 Details : In some cases it was possible to send a blocking callback to
4128 a client doing a glimpse, even though that client didn't get
4129 a lock granted. When the glimpse lock is cancelled on the OST
4130 the freed lock is left on the waiting list and corrupted the list.
4133 Frequency : all core dumps
4135 Description: Broke core dumps to lustre
4136 Details : Negative dentry may be unhashed if parent does not have UPDATE
4137 lock, but some callers, e.g. do_coredump, expect dentry to be
4138 hashed after successful create, hash it in ll_create_it.
4140 ------------------------------------------------------------------------------
4142 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4144 * Support for kernels:
4145 2.6.9-42.0.2.EL (RHEL 4)
4146 2.6.5-7.276 (SLES 9)
4147 2.4.21-40.EL (RHEL 3)
4148 2.6.12.6 vanilla (kernel.org)
4152 Frequency : always on RHEL 3
4154 Description: Number of open files grows over time
4155 Details : The number of open files grows over time, whether or not
4156 Lustre is started. This was due to a filp leak introduced
4157 by one of our kernel patches.
4159 ------------------------------------------------------------------------------
4161 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4163 * Support for kernels:
4164 2.6.9-42.EL (RHEL 4)
4165 2.6.5-7.276 (SLES 9)
4166 2.4.21-40.EL (RHEL 3)
4167 2.6.12.6 vanilla (kernel.org)
4172 Bugzilla : 5719, 9635, 9792, 9684
4173 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4174 Details : If a server is under heavy load and cannot reply to new
4175 connection requests before the client resends the (re)connect,
4176 the connection handling code can behave badly if two service
4177 threads are concurrently handing separate (re)connections from
4178 the same client. Add better locking to the connection handling
4179 code, and ensure that only a single connection will be processed
4180 for a given client UUID, even if the lock is dropped.
4182 Severity : enhancement
4184 Description: add TCP zero-copy support to kernel
4185 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4186 sends if the hardware supports scatter-gather and checksumming.
4187 This allows socklnd to do client-write and server-read more
4188 efficiently and reduce CPU utilization from skbuf copying.
4191 Frequency : only if NFS exporting from client
4193 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4194 Details : If setting attributes on a file created under NFS that had
4195 never been opened it would be possible to oops the client
4196 if the file had no objects.
4199 Frequency : always for liblustre
4201 Description: liblustre client does MDS+OSTs setattr RPC for each write
4202 Details : When doing a write from a liblustre client, the client
4203 incorrectly issued an RPC to the MDS and each OST the file was
4204 striped over in order to update the timestamps. When writing
4205 with small chunks and many clients this could overwhelm the MDS
4206 with RPCs. In all cases it would slow down the write because
4207 these RPCs are unnecessary.
4209 Severity : enhancement
4211 Description: allow number of MDS service threads to be changed at module load
4212 Details : It is now possible to change the number of MDS service threads
4213 running. Adding "options mds mds_num_threads={N}" to the MDS's
4214 /etc/modprobe.conf will set the number of threads for the next
4215 time Lustre is restarted (assuming the "mds" module is also
4216 reloaded at that time). The default number of threads will
4217 stay the same, 32 for most systems.
4222 Description: OST crash if filesystem is unformatted or corrupt
4223 Details : If an OST is started on a device that has never been formatted
4224 or if the filesystem is corrupt and cannot even mount then the
4225 error handling cleanup routines would dereference a NULL pointer.
4230 Description: NULL pointer deref in llap_from_page.
4231 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4232 incorrect) mapping if the page was truncated/reclaimed while it was
4233 searched for. Check for this condition and skip such pages when
4234 doing readahead. Introduce extra check to llap_from_page() to
4235 verify page->mapping->host is non-NULL (so page is not anonymous).
4238 Frequency : Sometimes when using sys_sendfile
4240 Description: "page not covered by a lock" warnings from ll_readpage
4241 Details : sendfile called ll_readpage without right page locks present.
4242 Now we introduced ll_file_sendfile that does necessary locking
4243 around call to generic_file_sendfile() much like we do in
4247 Frequency : with certain MDS communication failures at client mount time
4249 Description: NULL pointer deref after failed client mount
4250 Details : a client connection request may delayed by the network layer
4251 and not be sent until after the PTLRPC layer has timed out the
4252 request. If the client fails the mount immediately it will try
4253 to clean up before the network times out the request. Add a
4254 reference from the request import to the obd device and delay
4255 the cleanup until the network drops the request.
4258 Frequency : occasionally during client (re)connect
4260 Description: assertion failure during client (re)connect
4261 Details : processing a client connection request may be delayed by the
4262 client or server longer than the client connect timeout. This
4263 causes the client to resend the connection request. If the
4264 original connection request is replied in this interval, the
4265 client may trip an assertion failure in ptlrpc_connect_interpret()
4266 which thought it would be the only running connect process.
4269 Frequency : only with obd_echo servers and clients that are rebooted
4271 Description: kernel BUG accessing uninitialized data structure
4272 Details : When running an obd_echo server it did not start the ping_evictor
4273 thread, and when a client was evicted an uninitialized data
4274 structure was accessed. Start the ping_evictor in the RPC
4275 service startup instead of the OBD startup.
4277 Severity : enhancement
4278 Bugzilla : 10193 (patchless)
4279 Description: Remove dependency on various unexported kernel interfaces.
4280 Details : No longer need reparent_to_init, exit_mm, exit_files,
4281 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4284 Frequency : rare (only users of deprecated and unsupported LDAP config)
4286 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4287 Details : LDAP apparently contains 'lustreName' attributes instead of
4288 'name'. A simple remapping of the name is sufficient.
4291 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4293 Description: waiting_locks_callback trips kernel BUG if client is evicted
4294 Details : Running with the dump_on_timeout debug flag turned on makes
4295 it possible that the waiting_locks_callback() can try to dump
4296 the Lustre kernel debug logs from an interrupt handler. Defer
4297 this log dumping to the expired_lock_main() thread.
4299 Severity : enhancement
4301 Description: Support NFS exporting on 2.6 kernels.
4302 Details : Implement non-rawops metadata methods for NFS server to use without
4303 changing NFS server code.
4306 Frequency : very rare (synthetic metadata workload only)
4308 Description: two racing renames might cause an MDS thread to deadlock
4309 Details : Running the "racer" program may cause one MDS thread to rename
4310 a file from being the source of a rename to being the target of
4311 a rename at exactly the same time that another thread is doing
4312 so, and the second thread has already enqueued these locks after
4313 doing a lookup of the target and is trying to relock them in
4314 order. Ensure that we don't try to re-lock the same resource.
4317 Frequency : only very large systems with liblustre clients
4319 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4320 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4321 all OSTs in parallel. This allows the eviction of stale liblustre
4322 clients to proceed much faster than if they were done in series,
4323 and also offers similar improvements for other set_info RPCs.
4328 Description: excessive CPU usage during initial read phase on client
4329 Details : During the initial read phase on a client, it would agressively
4330 retry readahead on the file, consuming too much CPU and impacting
4331 performance (since 1.4.5.8). Improve the readahead algorithm
4332 to avoid this, and also improve some other common cases (read
4333 of small files in particular, where "small" is files smaller than
4334 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4339 Description: MDS crash when receiving packet with unknown intent.
4340 Details : Do not LBUG in unknown intent case, just return -EFAULT
4342 Severity : enhancement
4343 Bugzilla : 9293, 9385
4344 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4345 Details : Do not serialize getattr (non-intent version) and statfs.
4348 Frequency : occasional, when OST network is overloaded/intermittent
4350 Description: client evicted by OST after bulk IO timeout
4351 Details : If a client sends a bulk IO request (read or write) the OST
4352 may evict the client if it is unresposive to its data GET/PUT
4353 request. This is incorrect if the network is overloaded (takes
4354 too long to transfer the RPC data) or dropped the OST GET/PUT
4355 request. There is no need to evict the client at all, since
4356 the pinger and/or lock callbacks will handle this, and the
4357 client can restart the bulk request.
4360 Frequency : Always when mmapping file with no objects
4362 Description: client crashes when mmapping file with no objects
4363 Details : Check that we actually have objects in a file before doing any
4364 operations on objects in ll_vm_open, ll_vm_close and
4370 Description: Request leak when working with deleted CWD
4371 Details : Introduce advanced request refcount tracking for requests
4372 referenced from lustre intent.
4374 Severity : Enhancement
4376 Description: Cache open file handles on client.
4377 Details : MDS now will return special lock along with openhandle, if
4378 requested and client is allowed to hold openhandle, even if unused,
4379 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4380 closing files for every read/write openration.
4382 Severity : Enhancement
4384 Description: Cache open negative dentries on client when possible.
4385 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4386 negative dentries on lock revocation.
4391 Description: Remounting a client read-only wasn't possible with a zconf mount
4392 Details : It wasn't possible to remount a client read-only with llmount.
4394 Severity : enhancement
4395 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4396 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4397 MPICH is included to add Lustre-specific ADIO interfaces.
4398 This is based closely on the UFS ADIO layer and only differs
4399 in file creation, in order to allow the OST striping to be set.
4400 This is user-contributed code and not supported by CFS.
4405 Description: extended inode attributes (immutable, append-only) work improperly
4406 when 2.4 and 2.6 kernels are used on client/server or vice versa
4407 Details : Introduce kernel-independent values for these flags.
4409 Severity : enhancement
4412 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4413 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4414 granularity. Now, Fractional MB granularity is be supported,
4415 this is very useful for low memory system.
4417 Severity : enhancement
4419 Description: Getattr by fid
4420 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4421 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4422 have locks and avoids getattr rpc requests for it on every stat.
4425 Frequency : Always, for filesystems larger than 2TB
4427 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4428 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4429 filesystems larger than 2TB to be mounted with mballoc enabled.
4432 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4434 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4435 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4436 filesystems larger than 2TB is mounted then the kernel will
4437 silently corrupt the start of the filesystem. CONFIG_LBD is
4438 enabled for all CFS-supported kernels, but the possibility of
4439 this happening with a modified kernel config exists.
4441 Severity : enhancement
4443 Description: add client O_DIRECT support for 2.6 kernels
4444 Details : It is now possible to do O_DIRECT reads and writes to files
4445 in the Lustre client mountpoint on 2.6 kernel clients.
4447 Severity : enhancement
4449 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4450 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4451 parallel, not waiting for response from every OST before sending
4452 a rpc to the next OST.
4457 Description: setattr vs write race when updating file timestamps
4458 Details : Client processes that update a file timestamp into the past
4459 right after writing to the file (e.g. tar) it is possible that
4460 the updated file modification time can be reset to the current
4461 time due to a race between processing the setattr and write RPC.
4463 Severity : enhancement
4465 Description: Bring 'lfs find' closer in line with regular Linux find.
4466 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4467 -print0 options and obtains all the needed info through the lustre
4470 Severity : enhancement
4472 Description: support up to 1024 configured devices on one node
4473 Details : change obd_dev array from statically allocated to dynamically
4474 allocated structs as they are first used to reduce memory usage
4479 Description: Flush dirty partially truncated pages during truncate
4480 Details : Immediatelly flush partially truncated pages in filter_setattr,
4481 this way we completely avoid having any pages in page cache on OST
4482 and can retire ugly workarounds during writes to flush such pages.
4487 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4488 Details : protect inode from truncation within vfs_unlink() context
4489 just take a reference before calling vfs_unlink() and release it
4490 when parent's i_sem is free.
4495 Description: last_id value checked outside lock on OST caused LASSERT failure
4496 Details : If there were multiple MDS->OST object precreate requests in
4497 flight, it was possible that the OST's last object id was checked
4498 outside a lock and incorrectly tripped an assertion. Move checks
4499 inside locks, and discard old precreate requests.
4502 Frequency : always, if extents are used on OSTs
4504 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4505 Details : OSTs using the extents format would not zero the high 16 bits of
4506 the index physical block number. This is not a problem for any
4507 OST filesystems smaller than 16TB, and no kernels support ext3
4508 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4509 new/modified files) and can be fixed for existing filesystems
4510 with e2fsprogs-1.39-cfs1.
4515 Description: import connection selection may be incorrect if timer wraps
4516 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4517 import connections to be ignored if the 32-bit jiffies counter
4518 wraps. Use a 64-bit jiffies counter.
4521 Frequency : very large clusters immediately after boot
4523 Description: LNET request buffers exhausted under heavy short-term load
4524 Details : If a large number of client requests are generated on a service
4525 that has previously never seen so many requests it is possible
4526 that the request buffer growth cannot keep up with the spike in
4527 demand. Instead of dropping incoming requests, they are held in
4528 the LND until the RPC service can accept more requests.
4531 Frequency : Sometimes during replay
4533 Description: Assertion failure in ll_local_open after replay.
4534 Details : If replay happened on an open request reply before we were able
4535 to set replay handler, reply will become not swabbed tripping the
4536 assertion in ll_local_open. Now we set the handler right after
4537 recognising of open request
4540 Frequency : very rare
4542 Description: kernel reports "badness in vsnprintf"
4543 Details : Reading from the "recovery_status" /proc file in small chunks
4544 may cause a negative length in lprocfs_obd_rd_recovery_status()
4545 call to vsnprintf() (which is otherwise harmless). Exit early
4546 if there is no more space in the output buffer.
4548 Severity : enhancement
4550 Description: clear OBD RPC statistics by writing to them
4551 Details : It is now possible to clear the OBD RPC statistics by writing
4552 to the "stats" file.
4557 Description: Client mtime is not the same on different clients after utimes
4558 Details : In some cases, the client was using the utimes() syscall on
4559 a file cached on another node. The clients now validate the
4560 ctime from the MDS + OSTs to determine which one is right.
4565 Description: Inability to activate failout mode
4566 Details : lconf script incorrectly assumed that in pythong string's numeric
4567 value is used in comparisons.
4570 Frequency : always with multiple stripes per file
4572 Description: Inefficient object allocation for mutli-stripe files
4573 Details : When selecting which OSTs to stripe files over, for files with
4574 a stripe count that divides evenly into the number of OSTs,
4575 the MDS is always picking the same starting OST for each file.
4576 Return the OST selection heuristic to the original design.
4581 Description: mount failures may take full timeout to return an error
4582 Details : Under some heavy load conditions it is possible that a
4583 failed mount can wait for the full obd_timeout interval,
4584 possibly several minutes, before reporting an error.
4585 Instead return an error as soon as the status is known.
4587 Frequency : quota enabled and large files being deleted
4589 Description: releasing more than 4GB of quota at once hangs OST
4590 Details : If a user deletes more than 4GB of files on a single OST it
4591 will cause the OST to spin in an infinite loop. Release
4592 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4597 Description: statfs data retrieved from /proc may be stale or zero
4598 Details : When reading per-device statfs data from /proc, in the
4599 {kbytes,files}_{total,free,avail} files, it may appear
4600 as zero or be out of date.
4603 Frequency : systems with MD RAID1 external journal devices
4605 Description: lconf's call to blkid is confused by RAID1 journal devices
4606 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4607 of returning all block devices that match the journal UUID.
4610 Frequency : always, for aggregate stripe size over 4GB
4612 Description: assertion fails when trying to use 4GB stripe size
4613 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4614 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4617 Frequency : always on ppc64
4619 Description: the first write on an ext3 filesystem with mballoc got stuck
4620 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4621 perform endianness conversion.
4623 ------------------------------------------------------------------------------
4625 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4627 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4628 INTEROPERATE with older versions automatically. Please read the
4629 user documentation before upgrading any part of a live system.
4630 * WARNING: Lustre networking configuration changes are required with
4631 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4634 * Support for kernels:
4635 2.6.9-22.0.2.EL (RHEL 4)
4636 2.6.5-7.244 (SLES 9)
4637 2.6.12.6 vanilla (kernel.org)
4640 Severity : enhancement
4641 Bugzilla : 7981/8208
4642 Description: Introduced Lustre Networking (LNET)
4643 Details : LNET is new networking infrastructure for Lustre, it includes
4644 a reorganized network configuration mode (see the user
4645 documentation for full details) as well as support for routing
4646 between different network fabrics. Lustre Networking Devices
4647 (LNDs) for the supported network fabrics have also been
4648 created for this new infrastructure.
4650 Severity : enhancement
4651 Description: Introduced Access control lists
4652 Details : clients can set ACLs on files and directories in order to have
4653 more fine-grained permissions than the standard Unix UGO+RWX.
4654 The MDS must be started with the "-o acl" mount option.
4656 Severity : enhancement
4657 Description: Introduced filesystem quotas
4658 Details : Administrators may now establish per-user quotas on the
4661 Severity : enhancement
4663 Description: Configuration change for the XT3
4664 The PTLLND is now used to run Lustre over Portals on the XT3
4665 The configure option(s) --with-cray-portals are no longer used.
4666 Rather --with-portals=<path-to-portals-includes> is used to
4667 enable building on the XT3. In addition to enable XT3 specific
4668 features the option --enable-cray-xt3 must be used.
4673 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4674 Details : A race between allocating a new llcd and re-getting the llcd_lock
4675 allowed another thread to grab newly-allocated llcd.
4677 Severity : enhancement
4679 Description: 2.6 OST async journal commit and locking fix to improve performance
4680 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4681 2.6 kernels are now async as they already were in 2.4 kernels so
4682 that they can commit concurrently with the network bulk transfer.
4683 For block-allocated files the filter allocation semaphore is held
4684 to avoid filesystem fragmentation during allocation. BKL lock
4685 removed for 2.6 xattr operations where it is no longer needed.
4690 Description: lconf incorrectly determined whether two IP networks could talk
4691 Details : In some more complicated routing and multiple-network
4692 configurations, lconf will avoid trying to make a network
4693 connection to a disjoint part of the IP space. It was doing the
4694 math incorrectly for one set of cases.
4699 Description: Fix for potential infinite loop processing records in an llog.
4700 Details : If an llog record is corrupted/zeroed, it is possible to loop
4701 forever in llog_process(). Validate the llog record length
4702 and skip the remainder of the block on error.
4705 Frequency : occasional (liblustre only)
4707 Description: liblustre could not open files whose last component is a symlink
4708 Details : sysio_path_walk() would incorrectly pass the open intent to
4709 intermediate path components.
4712 Frequency : rare (liblustre only with non-standard tuning)
4713 Bugzilla : 7201 (7350)
4714 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4715 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4716 after it has been released, only do this for referenced locks
4718 Severity : enhancement
4720 Description: specifying an (invalid) directory default stripe_size of -1
4721 would reset the directory default striping
4722 Details : stripe_size -1 was used internally to signal directory stripe
4723 removal, now use "all default" to signal dir stripe removal
4724 as a directory striping of "all default" is not useful
4727 Frequency : common for large clusters running liblustre clients
4729 Description: doing an ls when liblustre clients are running is slow
4730 Details : sending a glimpse AST to a liblustre client waits for every AST
4731 to time out, as liblustre clients will not respond. Since they
4732 cannot cache data we refresh the OST lock LVB from disk instead.
4734 Severity : enhancement
4736 Description: doing an ls at the same time as file IO can be slow
4737 Details : enqueue and other "small" requests can be blocked behind many
4738 large IO requests. Create a new OST IO portal for non-IO
4739 requests so they can be processed faster.
4742 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4744 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4745 created file with mode 000
4746 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4747 allow an ftruncate() as a non-root user to a file with mode 000.
4748 Lustre now respects this flag to disable mode checks when
4749 truncating a file owned by the user
4752 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4755 Description: Revoking locks from clients that went dead or catatonic might take
4757 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4758 cancellation of such locks instant on servers without waiting for
4759 any reply from clients. Clients drops these locks when cancel
4760 notification from server is received without replying.
4763 Frequency : liblustre-only, when liblustre client dies or becomes busy
4765 Description: Doing ls on Linux clients can take a long time with active
4767 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4768 granting such locks to it in the first place if possible. Locks
4769 are taken by proxy on the OST during the read or write and
4770 dropped immediately afterward. Add connect flags handling, do
4771 not grant locks to liblustre clients for glimpse ASTs.
4773 Severity : enhancement
4775 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4776 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4777 sequential reads and behaves badly with applications doing random
4778 reads. Improve it by reading ahead at least read region, and
4779 avoiding excessive large RPC for small reads.
4781 Severity : enhancement
4783 Description: Creating more than 1000 files for a single job may cause a load
4784 imbalance on the OSTs if there are also a large number of OSTs.
4785 Details : qos_prep_create() uses an OST index reseed value that is an
4786 even multiple of the number of available OSTs so that if the
4787 reseed happens in the middle of the object allocation it will
4788 still utilize the OSTs as uniformly as possible.
4793 Description: OST or MDS may oops in ping_evictor_main()
4794 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4795 but doesn't restart at beginning of obd_exports_timed list
4798 Severity : enhancement
4800 Description: improve by-nid export eviction on the MDS and OST
4801 Details : allow multiple exports with the same NID to be evicted at one
4802 time without re-searching the exports list.
4805 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4807 Description: MDS may oops in groups_free()
4808 Details : in rare race conditions a newly allocated group_info struct is
4809 freed again, and this can be NULL. The 2.4 compatibility code
4810 for groups_free() checked for a NULL pointer, but 2.6 did not.
4813 Frequency : common for liblustre clients doing little filesystem IO
4814 Bugzilla : 9352, 7313
4815 Description: server may evict liblustre clients accessing contended locks
4816 Details : if a client is granted a lock or receives a completion AST
4817 with a blocking AST already set it would not reply to the AST
4818 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4821 Frequency : lfs setstripe, only systems with more than 160 OSTs
4823 Description: unable to set striping with a starting offset beyond OST 160
4824 Details : llapi_create_file() incorrectly limited the starting stripe
4825 index to the maximum single-file stripe count.
4828 Frequency : LDAP users only
4830 Description: lconf did not handle in-kernel recovery with LDAP properly
4831 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4833 Severity : enhancement
4835 Description: bind OST threads to NUMA nodes to improve performance
4836 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4837 node and do their allocations there to localize memory access
4839 Severity : enhancement
4841 Description: llmount can determine client NID directly from Myrinet (GM)
4842 Details : the client NID code from gmnalnid was moved directly into
4843 llmount, removing the need to use this or specifying the
4844 client NID explicitly when mounting GM clients with zeroconf
4847 Frequency : if client is started with down MDS
4849 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4850 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4851 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4852 ptlrpc_import_delayed_req() because we didn't check if the
4853 request was interrupted, and we also didn't break out of the
4854 event loop if there was no timeout
4859 Description: data loss during non-page-aligned writes to a single file from
4860 both multiple nodes and multiple threads on one node at same time
4861 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4862 inconsistency led to false short-reads, that were cached and later
4863 used by ->prepare_write() to fill in partially written page,
4864 leading to data loss.
4867 Frequency : always, if lconf --abort_recovery used
4869 Description: lconf --abort_recovery fails with 'Operation not supported'
4870 Details : lconf was attempting to abort recovery on the MDT device and not
4873 Severity : enhancement
4875 Description: remove cleanup logs
4876 Details : replace lconf-generated cleanup logs with lustre internal
4877 cleanup routines. Eliminates the need for client-cleanup and
4880 Severity : enhancement
4882 Description: add support for EAs (user and system) on lustre filesystems
4883 Details : it is now possible to store extended attributes in the Lustre
4884 client filesystem, and with the user_xattr mount option it
4885 is possible to allow users to store EAs on their files also
4887 Severity : enhancement
4889 Description: Add possibility (config option) to show minimal available OST free
4891 Details : When compiled with --enable-mindf configure option, statfs(2)
4892 (and so, df) will return least minimal free space available from
4893 all OSTs as amount of free space on FS, instead of summary of
4894 free spaces of all OSTs.
4896 Severity : enhancement
4898 Description: do not expand extent locks acquired on OST-side
4899 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4900 by server: they are not cached anyway.
4903 Frequency : when mmap is used/binaries executed from Lustre
4905 Description: Unmmap pages before throwing them away from read cache.
4906 Details : llap_shrink cache now attempts to unmap pages before discarding
4907 them (if unmapping failed - do not discard). SLES9 kernel has
4908 extra checks that trigger if this unmapping is not done first.
4913 Description: lconf didn't resolve symlinks before checking to see whether a
4914 given mountpoint was already in use
4917 Frequency : when migrating failover services
4918 Bugzilla : 6395, 9514
4919 Description: When migrating a subset of services from a node (e.g. failback
4920 from a failover service node) the remaining services would
4921 time out and evict clients.
4922 Details : lconf --force (implied by --failover) sets the global obd_timeout
4923 to 5 seconds in order to quickly disconnect, but this caused
4924 other RPCs to time out too quickly. Do not change the global
4925 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4927 Severity : enhancement
4928 Frequency : if MDS is started with down OST
4929 Bugzilla : 9439,5706
4930 Description: Allow startup/shutdown of an MDS without depending on the
4931 availability of the OSTs.
4932 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4933 Add appropriate locking and lov-osc refcounts for safe
4934 cleaning. Add osc abort_inflight calls in case the
4935 synchronize never started.
4938 Frequency : occasional (Cray XT3 only)
4940 Description: root not authorized to access files in CRAY_PORTALS environment
4941 Details : The client process capabilities were not honoured on the MDS in
4942 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4943 been accessed by an authorized user then root was able to access
4944 the file on the local client also. The root user capabilities
4945 are now allowed on the MDS, as this environment has secure UID.
4948 Frequency : occasional
4950 Description: ldiskfs "too long searching" message happens too often
4951 Details : A debugging message (otherwise harmless) prints too often on
4952 the OST console. This has been reduced to only happen when
4953 there are fragmentation problems on the filesystem.
4958 Description: Division by zero in statfs when all OSCs are inactive
4959 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4960 lov_statfs divides by value returned by lov_get_stripecnt().
4964 Bugzilla : 9489, 3273
4965 Description: First write from each client to each OST was only 4kB in size,
4966 to initialize client writeback cache, which caused sub-optimal
4967 RPCs and poor layout on disk for the first written file.
4968 Details : Clients now request an initial cache grant at (re)connect time
4969 and so that they can start streaming writes to the cache right
4970 away and always do full-sized RPCs if there is enough data.
4971 If the OST is rebooted the client also re-establishes its grant
4972 so that client cached writes will be honoured under the grant.
4977 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4978 Details : Now I/O RPCs go to different portal number and (presumably) fast
4979 lock requests (and glimses) and other RPCs get their own service
4980 threads pool that should be able to service those RPCs
4983 Severity : enhancement
4985 Description: Ability to exchange lustre version between client and servers and
4986 issue warnings at client side if client is too old. Also for
4987 liblustre clients there is ability to refuse connection of too old
4989 Details : New 'version' field is added to connect data structure that is
4990 filled with version info. That info is later checked by server and
4994 Frequency : rare, liblustre only.
4995 Bugzilla : 9296, 9581
4996 Description: Two simultaneous writes from liblustre at offset within same page
4997 might proceed at the same time overwriting eachother with stale
4999 Details : I/O lock withing llu_file_prwv was released too early, before data
5000 actually was hitting the wire. Extended lock-holding time until
5001 server acknowledges receiving data.
5004 Frequency : extremely rare. Never observed in practice.
5006 Description: avoid generating lustre_handle cookie of 0.
5007 Details : class_handle_hash() generates handle cookies by incrementing
5008 global counter, and can hit 0 occasionaly (this is unlikely, but
5009 not impossible, because initial value of cookie counter is
5010 selected randonly). Value of 0 is used as a sentinel meaning
5011 "unassigned handle" --- avoid it. Also coalesce two critical
5012 sections in this function into one.
5014 Severity : enhancement
5016 Description: allow liblustre clients to delegate truncate locking to OST
5017 Details : To avoid overhead of locking, liblustre client instructs OST to
5018 take extent lock in ost_punch() on client's behalf. New connection
5019 flag is added to handle backward compatibility.
5021 Severity : enhancement
5022 Bugzilla : 4928, 7341, 9758
5023 Description: allow number of OST service threads to be specified
5024 Details : a module parameter allows the number of OST service threads
5025 to be specified via "options ost ost_num_threads={N}" in the
5026 OSS's /etc/modules.conf or /etc/modprobe.conf.
5030 Bugzilla : 6146, 9635, 9895
5031 Description: servers crash with bad pointer in target_handle_connect()
5032 Details : In rare cases when a client is reconnecting it was possible that
5033 the connection request was the last reference for that export.
5034 We would temporarily drop the export reference and get a new
5035 one, but this may have been the last reference and the export
5036 was just destroyed. Get new reference before dropping old one.
5038 Severity : enhancement
5039 Frequency : if client is started with failover MDS
5041 Description: Allow multiple MDS hostnames in the mount command
5042 Details : Try to read the configuration from all specified MDS
5043 hostnames during a client mount in case the "primary"
5046 Severity : enhancement
5048 Description: Stop sending data to evicted clients as soon as possible.
5049 Details : Check if the client we are about to send or are sending data to
5050 was evicted already. (Check is done every second of waiting,
5051 for which l_wait_event interface was extended to allow checking
5052 of exit condition at specified intervals).
5055 Frequency : rare, normally only when NFS exporting is done from client
5057 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5059 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5062 Frequency : always, if srand() is called before liblustre initialization
5064 Description: Liblustre uses system PRNG disturbing its usage by user application
5065 Details : Introduce internal to lustre fast and high-quality PRNG for
5066 lustre usage and make liblustre and some other places in generic
5067 lustre code to use it.
5069 Severity : enhancement
5070 Bugzilla : 9477, 9557, 9870
5071 Description: Verify that the MDS configuration logs are updated when xml is
5072 Details : Check if the .xml configuration logs are newer than the config
5073 logs stored on the MDS and report an error if this is the case.
5074 Request --write-conf, or allow starting with --old_conf.
5076 Severity : enhancement
5078 Description: Handle symlinks in the path when checking if Lustre is mounted.
5079 Details : Resolve intermediate symlinks when checking if a client has
5080 mounted a filesystem to avoid duplicate client mounts.
5085 Description: lconf can hit an error exception but still return success.
5086 Details : The lconf command catches the Command error exception at the top
5087 level script context and will exit with the associated exit
5088 status, but doesn't ensure that this exit status is non-zero.
5093 Description: failure of ptlrpc thread startup can cause oops
5094 Details : Starting a ptlrpc service thread can fail if there are a large
5095 number of threads or the server memory is very fragmented.
5096 Handle this without oopsing.
5099 Frequency : always, only if liblustre and non-default acceptor port was used
5101 Description: liblustre cannot connect to servers with non-default acceptor port
5102 Details : tcpnal_set_default_params() was not called and was therefore
5103 ignoring the environment varaible TCPNAL_PORT, as well as other
5104 TCPNAL_ environment variables
5109 Description: two objects could be created on the same OST for a single file
5110 Details : If an OST is down, in some cases it was possible to create two
5111 objects on a single OST for a single file. No problems other
5112 than potential performance impact and spurious error messages.
5116 Bugzilla : 5681, 9562
5117 Description: Client may oops in ll_unhash_aliases
5118 Details : Client dcache may become inconsistent in race condition.
5119 In some cases "getcwd" can fail if the current directory is
5125 Description: Inode refcounting problems in NFS export code
5126 Details : link_raw functions used to call d_instantiate without obtaining
5127 extra inode reference first.
5131 Bugzilla : 9942, 9903
5132 Description: Referencing freed requests leading to crash, memleaks with NFS.
5133 Details : We used to require that call to ll_revalidate_it was always
5134 followed by ll_lookup_it. Also with revalidate_special() it is
5135 possible to call ll_revalidate_it() twice for the same dentry
5136 even if first occurence returned success. This fix changes semantic
5137 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5138 reference on a request referred from the intent.
5139 ll_intent_release() then releases such a request.
5142 Frequency : rare, normally benchmark loads only
5144 Description: unlinked inodes were kept in memory on the client
5145 Details : If a client is repeatedly creating and unlinking files it
5146 can accumulate a lot of stale inodes in the inode slab cache.
5147 If there is no other client load running this can cause the
5148 client node to run out of memory. Instead flush old inodes
5149 from client cache that have the same inode number as a new inode.
5152 Frequency : SLES9 2.6.5 kernel and long filenames only
5153 Bugzilla : 9969, 10379
5154 Description: utime reports stale NFS file handle
5155 Details : SLES9 uses out-of-dentry names in some cases, which confused
5156 the lustre dentry revalidation. Change it to always use the
5160 Frequency : rare, unless heavy write-truncate concurrency is continuous
5161 Bugzilla : 4180, 6984, 7171, 9963, 9331
5162 Description: OST becomes very slow and/or deadlocked during object unlink
5163 Details : filter_destroy() was holding onto the parent directory lock
5164 while truncating+unlinking objects. For very large objects this
5165 may block other threads for a long time and slow overall OST
5166 responsiveness. It may also be possible to get a lock ordering
5167 deadlock in this case, or run out of journal credits because of
5168 the combined truncate+unlink. Solution is to do object truncate
5169 first in one transaction without parent lock, and then do the
5170 final unlink in a new transaction with the parent lock. This
5171 reduces the lock hold time dramatically.
5174 Frequency : rare, 2.4 kernels only
5176 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5177 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5178 before unmounting their filesystem in order to avoid deadlock.
5179 The kernel_locked() function in 2.4 kernels only checks whether
5180 the kernel lock is held, not whether it is this process that is
5181 holding it as 2.6 kernels do.
5186 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5187 Details : The client ptlrpc code may be trying to reconnect to a down
5188 server before a previous connection attempt has timed out.
5189 Increase the reconnect interval to be longer than the connection
5190 timeout interval to avoid sending duplicate connections to
5194 Frequency : echo_client brw_test command
5196 Description: fix echo_client to work with OST preallocated code
5197 Details : OST preallocation code (5137) didn't take echo_client IO path
5198 into account: echo_client calls filter methods outside of any
5199 OST thread and, hence, there is no per-thread preallocated
5200 pages and buffers to use. Solution: hijack pga pages for IO. As
5201 a byproduct, this avoids unnecessary data copying.
5205 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5206 Description: Client can oops in mdc_commit_close() after open replay
5207 Details : It was possible for the MDS to return an open request with no
5208 transaction number in mds_finish_transno() if the client was
5209 evicted, but without actually returning an error. Clients
5210 would later try to replay that open and may trip an assertion
5211 Simplify the client close codepath, and always return an error
5212 from the MDS in case the open is not successful.
5215 Frequency : rare, 2.6 OSTs only
5217 Description: OST may deadlock under high load on fragmented files
5218 Details : If there was a heavy load and highly-fragmented OST filesystems
5219 it was possible to have all the OST threads deadlock waiting on
5220 allocation of biovecs, because the biovecs were not released
5221 until the entire RPC IO was completed. Instead, release biovecs
5222 as soon as they are complete to ensure forward IO progress.
5224 Severity : enhancement
5226 Description: Support for specifying external journal device at mount
5227 Details : If an OST or MDS device is formatted with an external journal
5228 device, this device major/minor is stored in the ext3 superblock
5229 and may not be valid for failover. Allow detecting and
5230 specifying the external journal at mount time.
5235 Description: Mounting an MDS with pending unlinked files may cause oops
5236 Details : target_finish_recovery() calls mds_postrecov() which returned
5237 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5238 considers this an error and immediately begins cleaning up the
5239 lov, just after starting the mds_lov process
5241 Severity : enhancement
5243 Description: Implement 'lfs df' to report actual free space on per-OST basis
5244 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5245 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5246 usage of objects; '-h' to report in human readable format.
5248 ------------------------------------------------------------------------------
5250 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5257 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5258 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5259 at startup during orphan unlinking. Instead, skip buddy update
5260 before setup, as it will be regenerated after recovery is complete.
5265 Description: If an OST is inactive, its locks might reference stale inodes.
5266 Details : lov_change_cbdata() must iterate over all namespaces, even if
5267 they are inactive to clear inode references from the lock.
5269 Severity : enhancement
5270 Frequency : occasional, if non-standard max_dirty_mb used
5272 Description: Client will block write RPCs if not enough grant
5273 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5274 then the client will block writes while waiting for another RPC
5275 to complete instead of consuming its dirty limit. With change
5276 we get improved performance when max_dirty_mb is small.
5278 Severity : enhancement
5279 Bugzilla : 3389, 6253
5280 Description: Add support for supplementary groups on the MDS.
5281 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5282 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5283 lookups for user supplementary groups into a cache.
5287 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5288 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5289 for liblustre O_EXCL is enforced by MDS.
5293 Description: readdir never returns NULL in liblustre.
5294 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5295 dentry in struct dirent.
5299 Frequency : liblustre only, depends on application IO pattern
5300 Description: liblustre clients evicted if not contacting servers
5301 Details : Don't put liblustre clients into the ping_evictor list, so
5302 they will not be evicted by the pinger ever.
5304 Severity : enhancement
5306 Description: Add ability to evict clients by NID from MDS.
5307 Details : By echoing "nid:$NID" string into
5308 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5309 $NID would be instantly evicted from this MDS and from all active
5310 OSTs connected to it.
5314 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5315 Details : lookup_it_finish() used to query file size from OSTs that was not
5320 Description: service threads change working directory to that of init
5321 Details : Starting lustre service threads may pin the working directory
5322 of the parent thread, making that filesystem busy. Threads
5323 now change to the working directory of init to avoid this.
5327 Frequency : during shutdown only
5328 Description: shutdown with a failed MDS or OST can cause unmount to hang
5329 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5334 Frequency : 2.6 only
5335 Description: chmod/chown may include an extra supplementary group
5336 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5337 supplementary group and if none is specified this is used.
5340 Bugzilla : 5479 (6816)
5342 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5343 Details : If lookup is in progress on a file that is unlinked we might try
5344 to revalidate the inode and fail in revalidate after lookup is
5345 complete and ll_file_open() enqueues the open again but
5346 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5349 Frequency : always, if lconf --abort_recovery used
5351 Description: lconf --abort_recovery fails with 'Operation not supported'
5352 Details : lconf was attempting to abort recovery on the MDT device and not
5355 ------------------------------------------------------------------------------
5357 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5362 Frequency : rare (only unsupported configurations with a node running as an
5364 Bugzilla : 6514, 5137
5365 Description: Mounting a Lustre file system on a node running as an OST could
5367 Details : OSTs now preallocates memory needed to write out data at
5368 startup, instead of when needed, to avoid having to
5369 allocate memory in possibly low memory situations.
5370 Specifically, if the file system is mounted on on OST,
5371 memory pressure could force it to try to write out data,
5372 which it needed to allocate memory to do. Due to the low
5373 memory, it would be unable to do so and the node would
5374 become unresponsive.
5376 Severity : enhancement
5378 Description: Addition of lconf --service command line option
5379 Details : lconf now accepts a '--service <arg>' option, which is
5380 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5382 Severity : enhancement
5384 Description: Failover mode is now the default for OSTs.
5385 Details : By default, OSTs will now run in failover mode. To return to
5386 the old behaviour, add '--failout' to the lmc line for OSTs.
5388 Severity : enhancement
5390 Description: Health checks are now provided for MDS and OSTs
5391 Details : Additional detailed health check information on MSD and OSTs
5392 is now provided through the procfs health_check value.
5395 Frequency : occasional, depends on IO load
5397 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5398 after numerous create/delete cycles
5399 Details : The ext3 inode allocation policy would not allocate new inodes
5400 very well on the OSTs because there are no new directories
5401 being created. Instead we look for groups with free space if
5402 the parent directories are nearly full.
5407 Description: Network or server problems during mount may cause partially
5408 mounted clients instead of returning an error.
5409 Details : The config llog parsing code may overwrite the error return
5410 code during mount error handling, returning success instead
5416 Description: MDS can fail to allocate large reply buffers
5417 Details : After long uptimes the MDS can fail to allocate large reply
5418 buffers (e.g. zconf client mount config records) due to memory
5419 fragmentation or consumption by the buffer cache. Preallocate
5420 some large reply buffers so that these replies can be sent even
5421 under memory pressure.
5425 Frequency : rare (liblustre)
5426 Description: fsx running with liblustre complained that using truncate() to
5427 extend the file doesn't work. This patch corrects that issue.
5428 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5429 ATTR_SIZE and lsm use in llu_setattr_raw.
5433 Frequency : rare, only 2.6 kernels
5434 Description: Unusual file access patterns on the MDS may result in inode
5435 data being lost in very rare circumstances.
5436 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5437 code in the 2.6 kernel caused the inode and/or EA data not to
5438 be read from disk, causing single-file corruption.
5442 Frequency : rare, only 2.6 filesystems using extents
5443 Description: Heavy concurrent write and delete load may cause data corruption.
5444 Details : It was possible under high-load situations to have an extent
5445 metadata block in the block device cache from a just-unlinked
5446 file overwrite a newly-allocated data block. We now unmap any
5447 metadata buffers that alias just-allocated data blocks.
5451 Frequency : filesystems with default stripe_count larger than 77
5452 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5453 Details : lconf specifies an inode size of 4096 bytes when the default
5454 stripe_count is larger than 77. This conflicts with the default
5455 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5456 this case to avoid pinning too much memory for large EAs.
5458 ------------------------------------------------------------------------------
5460 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5465 Frequency : rare (extremely heavy IO load with hundreds of clients)
5467 Description: Client is evicted, gets IO error writing to file
5468 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5469 caused clients to be evicted for AST timeouts. The fixes in
5470 bug 5192 mean we no longer need to have such short AST timeouts
5471 so ldlm_timeout has been increased.
5474 Frequency : occasional during --force or --failover shutdown under load
5475 Bugzilla : 5949, 4834
5476 Description: Server oops/LBUG if stopped with --force or --failover under load
5477 Details : a collection of import/export refcount and cleanup ordering
5478 issues fixed for safer force cleanup
5481 Frequency : only filesystems larger than 120 OSTs
5482 Bugzilla : 5990, 6223
5483 Description: lfs getstripe would oops on a very large filesystem
5484 Details : lov_getconfig used kfree on vmalloc'd memory
5487 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5488 Bugzilla : 6242, 6243
5489 Description: reading from files that had been truncated to a non-zero size
5490 but never opened returned no data
5491 Details : ll_file_read() reads zeros from no-object files to EOF
5496 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5497 Details : There's little harm in aborting MDS/OSS recovery and letting it
5498 try again, so I removed the LASSERT and return an error instead.
5500 Severity : enhancement
5502 Description: New debugging infrastructure for tracking down data corruption
5503 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5504 (b) cover more of the client-side code path, and (c) try to narrow
5505 down where problems occurred
5509 Bugzilla : 3819, 4364, 4397, 6313
5510 Description: Racing close and eviction MDS could cause assertion in mds_close
5511 Details : It was possible to get multiple mfd references during close and
5512 client eviction, leading to one thread referencing a freed mfd.
5514 Severity: : enhancement
5515 Bugzilla : 3262, 6359
5516 Description: Attempts to reconnect to servers are now more aggressive.
5517 Details : This builds on the enhanced upcall-less recovery that was added
5518 in 1.4.2. When trying to reconnect to servers, clients will
5519 now try each server in the failover group every 10 seconds. By
5520 default, clients would previously try one server every 25 seconds.
5525 Description: After recovery, certain operations trigger a failed
5526 assertion on a client.
5527 Details : Failing over an mds, using lconf -d --failover, while a
5528 client was doing a readdir() call would cause the client to
5529 LBUG after recovery completed and the readdir() was resent.
5531 Severity : enhancement
5533 Description: Default groups are now added by lconf
5534 Details : You can now run lconf --group <servicename> without having to
5535 manually add groups with lmc.
5538 Frequency : occasional
5540 Description: Nodes with an elan id of 0 trigger a failed assertion
5543 Frequency : always when accessing e.g. tty/console device nodes
5545 Description: tty and some other devices nodes cannot be used on lustre
5546 Details : file's private_data field is used by device data and lustre
5547 values in there got lost. New field was added to struct file to
5548 store fs-specific private data.
5551 Frequency : when exporting Lustre via NFS
5553 Description: NFSD failed occasionally when looking up a path component
5554 Details : NFSD is looking up ".." which was broken in ext3 directories
5555 that had grown large enough to become hashed.
5558 Frequency : Clusters with multiple interfaces not on the same subnet
5560 Description: Nodes will repeatedly try to reconnect to an interface which it
5561 cannot reach and report an error to the log.
5562 Details : Extra peer list entries will be created by lconf with some peers
5563 unreachable. lconf now validates the peer before adding it.
5566 Frequency : Only if a default stripe is set on the filesystem root.
5568 Description: Setting a default stripe on the filesystem root prevented the
5569 filesystem from being remounted.
5570 Details : The client was sending extra request flags in the root getattr
5571 request and did not allocate a reply buffer for the dir EA.
5574 Frequency : occasional, higher if lots of files are accessed by one client
5575 Bugzilla : 6159, 6097
5576 Description: Client trips assertion regarding lsm mismatch/magic
5577 Details : While revalidating inodes the VFS looks up inodes with ifind()
5578 and in rare cases can find an inode that is being freed.
5579 The ll_test_inode() code will free the lsm during ifind()
5580 when it finds an existing inode and then the VFS later attaches
5581 this free lsm to a new inode.
5585 Bugzilla : 6422, 7030
5586 Description: MDS deadlock between mkdir and client eviction
5587 Details : Creating a new file via mkdir or mknod (starting a transaction
5588 and getting the ns lock) can deadlock with client eviction
5589 (gets ns lock and trying to finish a synchronous transaction).
5592 Frequency : occasional
5593 Description: While starting a server, the fsfilt_ext3 module could not be
5595 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5596 kernels. Previously, lconf would still use the ext3 name
5597 when trying to load modules. Now, it will correctly use
5598 ext3 on 2.4 and ldiskfs on 2.6.
5600 Severity : enhancement
5601 Description: The default stripe count has been changed to 1
5602 Details : The interpretation of the default stripe count (0, to lfs
5603 or lmc) has been changed to mean striping across a single
5604 OST, rather than all available. For general usage we have
5605 found a stripe count of 1 or 2 works best.
5607 Severity : enhancement
5608 Description: Add support for compiling against Cray portals.
5609 Details : Conditional compiling for some areas that are different
5613 Frequency : occasional
5614 Bugzilla : 6409, 6834
5615 Description: Creating files with an explicit stripe count may lead to
5616 a failed assertion on the MDS
5617 Details : If some OSTs are full or unavailable, creating files may
5618 trigger a failed assertion on the MDS. Now, Lustre will
5619 try to use other servers or return an error to the
5623 Frequency : occasional
5625 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5626 caused lock timeout during readahead (since 1.4.2).
5627 Details : Processes doing readahead might match a lock that hasn't been
5628 granted yet if there are overlapping and conflicting lock
5629 requests. The readahead process waits on ungranted lock
5630 (original lock is CBPENDING), while OST waits for that process
5631 to cancel CBPENDING read lock and eventually evicts client.
5633 Severity : enhancement
5635 Description: Initial enabling of flock support for clients
5636 Details : Implements fcntl advisory locking and file status functions.
5637 This feature is provided as an optional mount flag (default
5638 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5639 locking are implemented yet, and those that are are not guaranteed
5640 to be completely correct in production environments.
5641 mount -t lustre -o [flock|noflock] ...
5644 Frequency : occasional
5646 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5647 assertion in the ext3 JBD (journaling) layer.
5648 Details : The b_committed_data struct is protected by the big kernel lock
5649 in 2.4 kernels, serializing journal_commit_transaction() and
5650 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5651 access to this struct. In 2.6 kernels there is finer grained
5652 locking to improve SMP performance of the JBD layer.
5656 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5661 Description: merge and cleanup kernel patches.
5664 -----------------------------------------------------------------------------
5666 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5668 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5669 versions of Lustre. Please update all servers and clients to
5670 version 1.4.2 or later at the same time. You must also run
5671 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5672 to update the configuration logs.
5674 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5675 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5676 - swab llog records of type '0' so we get proper header size/idx (5861)
5677 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5678 - fix rename of one directory over another leaking an inode (5953)
5679 - avoid SetPageDirty on 2.6 (5981)
5680 - don't re-add just-being-destroyed locks to the waiting list (5653)
5681 - when creating new directories, inherit the parent's custom
5682 striping settings if present parent (3048)
5683 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5684 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5685 - don't hold client locks on temporary worklist from l_lru (5666)
5686 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5687 - automatically evict dead clients (5921)
5688 - Update file size properly in create+truncate+fstat case (6196)
5689 - Do not unhash mountpoint dentries, do not allow removal of
5691 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5692 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5693 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5694 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5695 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5696 - Correctly handle reads of files with no objects (6243)
5697 - lctl recover will also mark a device active if deactivate used (5933)
5699 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5700 - allow --write-conf on an MDS with different nettype than client (5619)
5701 - don't write config llogs to MDS for mounts not from that MDS (5617)
5702 - lconf should create multiple TCP connections from a client (5201)
5703 - init scripts are now turned off by default; run chkconfig --on
5704 lustre and chkconfig --on lustrefs to use them
5705 - upcalls are no longer needed for clients to recover to failover
5707 - add --abort-recovery option to lconf to abort recovery on device
5709 - add support for an arbitrary number of OSTs (3026)
5710 - Quota support protocol changes.
5711 - forward compatibility changes to wire structs (6007)
5712 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5713 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5714 - improved "lustre status" script
5715 - initialize blocksize for non-regular files (6062)
5716 - added --disable-server and --disable-client configure options (5782)
5717 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5718 - Vanilla 2.4.29 support
5719 - increase maximum number of obd devices to 520 (6242)
5720 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5721 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5722 SLES 9 2.6.5 kernels
5723 - limit stripes per file to 160 (the maximum EA size) (6093)
5725 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5728 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5729 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5730 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5731 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5732 - small ext3 extents cleanups and fixes (5733)
5733 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5734 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5735 - avoid needless client->OST connect, fix handle mismatch (5317)
5736 - fix DLM error path that led to out-of-sync client, long delays (5779)
5737 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5738 - fix several locking issues related to i_size (5492,5624,5654,5672)
5739 - don't move pending lock onto export if it is already evicted (5683)
5740 - fix kernel oops when creating .foo in unlinked directory (5548)
5741 - fix deadlock in obdfilter statistics vs. object create (5811)
5742 - use time_{before,after} to avoid timer jiffies wrap (5882)
5743 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5744 - Do not leak request if server was not able to process it (5154)
5745 - If mds_open unable to find parent dir, make that negative lookup(5154)
5746 - don't create new directories with extent-mapping (5909, 5936)
5748 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5749 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5750 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5751 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5752 - support for RHEL4 kernel on i686 (5773)
5753 - provide error messages when incompatible logs are encountered (5898)
5755 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5756 * version 1.4.0.10 (1.4.1 release candidate 1)
5758 - don't keep a lock reference when lock is not granted (4238)
5759 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5760 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5761 - drop import inflight refcount on signal_completed_replay error (5255)
5762 - unlock page after async write error during send (3677)
5763 - handle missing objects in filter_preprw_read properly (5265)
5764 - no transno return for symlink open, don't save no-trasno open (3440)
5765 - don't try to complete elan receive that already failed (4012)
5766 - free RPC server reply state on error (5406)
5767 - clean up thread from ptlrpc_start_thread() on error (5160)
5768 - readahead could read extra page into cache that wasn't ejected (5388)
5769 - prevent races in class_attach/setup/cleanup/detach (5260)
5770 - don't dereference de->d_inode after l_dput of de (5458)
5771 - use "int" for stripe value returned from lock_to_stripe (5544)
5772 - mballoc allocation and error-checking fixes in 2.6 (5504)
5773 - block device patches to fix I/O request sizes in 2.6 (5482)
5774 - look up hostnames for IB nals (5602)
5775 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5776 - don't start multiple acceptors for the same port (5277)
5777 - fix incorrect LASSERT in mds_getattr_name (5635)
5778 - export a proc file for general "ping" checking (5628)
5779 - fix "lfs check" to not block when the MDS is down (5628)
5781 - service request history (4965)
5782 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5783 - create an "evict_client" /proc entry on OSTs, like the MDS has
5784 - fix mount usage message, return errors per mount(8) (5168)
5785 - change grep [] to grep "[]" in tests so they work in more UMLs
5786 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5787 - remove ancient LOV_MAGIC_V0 EA support (5047)
5788 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5789 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5790 - allow readahead allocations to fail when low on memory (5383)
5791 - mmap locking landed again, after considerable improvement (2828)
5792 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5794 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5797 - send OST transaction number in read/write reply to free req (4966)
5798 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5799 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5800 - the watchdog thread now runs as interruptible (5246)
5801 - flock/lockf fixes (but it's still disabled, pending 5135)
5802 - don't use EXT3 constants in llite code (5094)
5803 - memory shortage at startup could cause assertion (5176)
5805 - reorganization of lov code
5806 - single portals codebase
5808 - add extents/mballoc support (5025)
5809 - direct I/O reads in the obdfilter (4048)
5810 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5812 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5815 - send OST transaction number in read/write reply to free req (4966)
5816 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5817 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5818 - the watchdog thread now runs as interruptible (5246)
5819 - handle missing objects in filter_preprw_read properly (5265)
5820 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5821 - drop import inflight refcount on signal_completed_replay error (5255)
5822 - unlock page after async write error during send (3677)
5823 - return original error code on reconstructed replies (3761)
5824 - no transno return for symlink open, don't save no-trasno open (3440)
5826 - add pid to ldlm debugging output (4922)
5827 - bump the watchdog timeouts -- we can't handle 30sec yet
5828 - extra debugging for orphan dentry/inode bug (5259)
5830 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5833 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5834 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5835 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5836 - fix NULL dereference in /proc/sys/portals/routes (4827)
5837 - allow failed mdc_close() operations to be interrupted (4561)
5838 - stop precreate on OST before MDS would time out on it (4778)
5839 - don't send partial-page writes before EOF from client (4410)
5840 - discard client grant for sub-page writes on large-page clients (4520)
5841 - don't free dentries not owned by NFS code, check generation (4806)
5842 - fix lsm leak if mds_create_objects() fails (4801)
5843 - limit debug_daemon file size, always print CERROR messages (4789)
5844 - use transno after validating reply (3892)
5845 - process timed out requests if import state changes (3754)
5846 - update mtime on OST during writes, return in glimpse (4829)
5847 - add mkfsoptions to LDAP (4679)
5848 - use ->max_readahead method instead of zapping global ra (5039)
5849 - don't interrupt __l_wait_event() during strace
5851 - add software watchdogs to catch hung threads quickly (4941)
5852 - make lustrefs init script start after nfs is mounted
5853 - fix CWARN/ERROR duplication (4930)
5854 - return async write errors to application if possible (2248)
5855 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5856 - print NAL number in %x format (4645)
5857 - update barely-supported suse-2.4.21-171 series (4842)
5858 - support for sles 9 %post scripts
5859 - support for building 2.6 kernel-source packages
5860 - support for sles km_* packages
5862 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5865 - ignore -ENOENT errors in osc_destroy (3639)
5866 - notify osc create thread that OSC is being cleaned up (4600)
5867 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5868 - reconstruct ost_handle() like mds_handle() (4657)
5869 - create a new thread to do import eviction to avoid deadlock (3969)
5870 - let lconf resolve symlinked-to devices (4629)
5871 - don't unlink "objects" from directory with default EA (4554)
5872 - hold socknal file ref over connect in case target is down (4394)
5873 - allow more than 32000 subdirectories in a single directory (3244)
5874 - fix blocks count for O_DIRECT writes (3751)
5875 - OST returns ENOSPC from object create when no space left (4539)
5876 - don't send truncate RPC if file size isn't changing (4410)
5877 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5878 - bind to privileged port in socknal and tcpnal (3689)
5880 - rate limit CERROR/CWARN console message to avoid overload (4519)
5881 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5882 - basic mmap support (3918)
5883 - kernel patch series update from b1_4 (4711)
5885 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5888 - avoid crash during MDS cleanup with OST shut down (2775)
5889 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5890 - don't use bad inodes on the MDS (3744)
5891 - dynamic object preallocation to improve recovery speed (4236)
5892 - don't hold spinlock over lock dumping or change debug flags (4401)
5893 - don't zero obd_dev when it is force cleaned (3651)
5894 - print grants to console if they go negative (4431)
5895 - "lctl deactivate" will stop automatic recovery attempts (3406)
5896 - look for existing locks in ldlm_handle_enqueue() (3764)
5897 - don't resolve lock handle twice in recovery avoiding race (4401)
5898 - revalidate should check working dir is a directory (4134)
5900 - don't always mark "slow" obdfilter messages as errors (4418)
5902 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5905 - don't close LustreDB during write_conf until it is done (3860)
5906 - fix typo in lconf for_each_profile (3821)
5907 - allow dumping logs from multiple threads at one time (3820)
5908 - don't allow multiple threads in OSC recovery (3812)
5909 - fix debug_size parameters (3864)
5910 - fix mds_postrecov to initialize import for llog ctxt (3121)
5911 - replace config semaphore with spinlock (3306)
5912 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5913 - don't allow enqueue to complete on a destroyed export (3822)
5914 - down write_lock before checking llog header bitmap (3825)
5915 - recover from lock replay timeout (3764)
5916 - up llog sem before sending rpc (3652)
5917 - reduce ns lock hold times when setting kms (3267)
5918 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5919 - fix NULL deref and obd_dev leak on setup error (3312)
5920 - replace some LBUG about llog ops with error handling (3841)
5921 - don't match INVALID dentries from d_lookup and spin (3784)
5922 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5923 - fix invalid assertion in ptlrpc_set_wait (3880)
5925 - add libwrap support for the TCP acceptor (3996)
5926 - add /proc/sys/portals/routes for non-root route listing (3994)
5927 - allow setting MDS UUID in .xml (2580)
5928 - print the stack of a process that LBUGs (4228)
5930 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5933 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5934 - make sure to unset replay flag from failed open requests (3440)
5935 - if default stripe count is 0, use OST count for inode size (3636)
5936 - update parent mtime/ctime on client for create/unlink (2611)
5937 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5938 - free recovery state on server during a forced cleanup (3571)
5939 - unregister_reply for resent reqs (3063)
5940 - loop back devices mounting and status check on 2.6 (3563)
5941 - fix resource-creation race that can provoke i_size == 0 (3513)
5942 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5943 - more debugging for page-accounting assertion (3746)
5944 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5945 - avoid deadlock after precreation failure (3758)
5946 - fix race and lock order deadlock in orphan handling (3450, 3750)
5947 - add validity checks when grabbing inodes from l_ast_data (3599)
5949 - add /proc/.../recovery_status to obdfilter (3428)
5950 - lightweight CDEBUG infrastructure, debug daemon (3668)
5951 - change default OSC RPC parameters to be better on small clusters
5952 - turn off OST read cache for files smaller than 32MB
5953 - install man pages and include them in rpms (3100)
5954 - add new init script for (un)mounting lustre filesystems (2593)
5955 - run chkconfig in %post for init scripts (3701)
5956 - drop scimac NAL (unmaintained)
5958 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5961 - clean kiobufs before and after use (3485)
5962 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5963 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5964 - change init script to fail more clearly if not run as root (1528)
5965 - allow clients to reconnect during replay (1742)
5966 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5967 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5968 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5969 - fix cond_resched (3554)
5970 - only evict unfinished clients after recovery (3515)
5971 - allow bulk resend, prevent data loss (3570)
5972 - dynamic ptlrpc request buffer allocation (2102)
5973 - don't allow unlinking open directory if it isn't empty (2904)
5974 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5975 - remove extraneous obd dereference causing LASSERT failure (3334)
5976 - don't use get_cycles() when creating temp. files on the mds (3156)
5977 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5978 - handle EEXIST for set-stripe, set proper directory name (3336)
5980 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5981 - fix ksocknal_fmb_callback() error messages (2918)
5983 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5986 - don't copy lvb into (possibly NULL) reply on error (2983)
5987 - don't deref dentry after dput, don't free lvb on error (2922)
5988 - use the kms to determine writeback rpc length (2947)
5989 - increment oti_logcookies when osc is inactive (2948)
5990 - update client's i_blocks count via lvb messages (2543)
5991 - handle intent open/close of special files properly (1557)
5992 - mount MDS with errors=remount-ro, like obdfilter (2009)
5993 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5994 - don't use cancelling-locks' kms values (2947)
5995 - use highest lock extent for kms, not last one (2925)
5996 - don't dereference ERR_PTR() dentry in error handling path (3107)
5997 - fix thread race in portals_debug_dumplog() (3122)
5998 - create lprocfs device entries at setup instead of at attach (1519)
5999 - common AST error handler, don't evict client on completion race (3145)
6000 - zero nameidata in detach_mnt in 2.6 (3118)
6001 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6002 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6003 - initialize RPC timeout timer earlier for 2.6 (3219)
6004 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6005 - print nal/nid for unknown nid (3258)
6006 - additional checks for oscc recovery before doing precreate (3284)
6007 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6008 - don't crash in mdc_close for bad permissions on open (3285)
6009 - zero i_rdev for non-device files (3147)
6010 - clear page->private before handing to FS, better assertion (3119)
6011 - tune the read pipeline (3236)
6012 - fix incorrect decref of invalidated dentry (2350)
6013 - provide read-ahead stats and refine rpc in flight stats (3328)
6014 - don't hold journal transaction open across create RPC (3313)
6015 - update atime on MDS at close time (3265)
6016 - close LDAP connection when recovering to avoid server load (3315)
6017 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6018 - don't leak open file on MDS after open resend (3325)
6019 - serialize filter_precreate and filter_destroy_precreated (3329)
6020 - loop device shouldn't call sync_dev() for nul device (3092)
6021 - clear page cache after eviction (2766)
6022 - resynchronize MDS->OST in background (2824)
6023 - refuse to mount the same filesystem twice on same mountpoint (3394)
6024 - allow llmount to create routes for mounting behind routers (3320)
6025 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6026 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6027 - fix rare problem with rename on htree directories (3417)
6029 - allow default OST striping configuration per directory (1414)
6030 - fix compilation for qswnal for 2.6 kernels (3125)
6031 - increase maximum number of MDS request buffers for large systems
6032 - change liblustreapi to be useful for external progs like lfsck (3098)
6033 - increase local configuration timeout for slow disks (3353)
6034 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6036 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6039 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6040 - don't overwrite extent policy data in reply if lock was blocked (2901)
6041 - drop filter export grants atomically with removal from device (2663)
6042 - del obd_self_export from work_list in class_disconnect_exports (2908)
6043 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6044 - swab reply message in mdc_close, other PPC fixes (2464)
6045 - fix destroying of named logs (2325)
6046 - overwrite old logs when running lconf --write_conf (2264)
6047 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6048 - fix race in target_handle_connect (2898)
6049 - mds_reint_create() should take same inode create lock (2926)
6050 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6051 - don't close files for self_export to avoid uninitialized obd (2936)
6052 - allow MDS with the same name as client node (2939)
6053 - hold dentry reference for closed log files for unlink (2325)
6054 - reserve space for all logs during transactions (2059)
6055 - don't evict page beyond end of stripe extent (2925)
6056 - don't oops on a deleted current working directory (2399)
6057 - handle hard links to targets without a parent properly (2517)
6058 - don't dereference NULL lock when racing during eviction (2867)
6059 - don't grow lock extents when lots of conflicting locks (2919)
6061 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6064 - account for cache space usage on clients to avoid data loss (974)
6065 - lfsck support in lustre kernel code (2349)
6066 - reduce journal credits needed for BRW writes (2370)
6067 - orphan handling to avoid losing space on client/server crashes
6068 - ptlrpcd can be blocked, stopping ALL progress (2477)
6069 - use lock value blocks to assist in proper KMS, faster stat (1021)
6070 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6071 - recovery for initial connections (2355)
6072 - fixes for mds_cleanup_orphans (1934)
6073 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6074 - block all file creations until orphan recovery completes (1901)
6075 - client remove rq_connection from request struct (2423)
6076 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6077 - recovery timer race (2670)
6078 - mdc_close recovey bug (2532)
6079 - ptlrpc cleanup bug (2710)
6080 - mds timeout on local locks (2588)
6081 - namespace lock held during RPCs (2431)
6082 - handle interrupted sync write properly (2503)
6083 - don't try to handle a message that hasn't been replied to (2699)
6084 - client assert failure during cleanup after abort recovery (2701)
6085 - leak mdc device after failed mount (2712)
6086 - ptlrpc_check_set allows timedout requests to complete (2714)
6087 - wait for inflight reqs when ptlrpcd finishes (2710)
6088 - make sure unregistered services are removed from the srv_list
6089 - reset bulk XID's when resending them (caught by 1138 test)
6090 - unregister_bulk after timeout
6091 - fix lconf error (2694)
6092 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6093 - readahead locks pages, leaves pending causing memory pressure (2673)
6094 - increase OST request buffers to 4096 on large machines (2729)
6095 - fix up permission of existing directories in simple_mkdir (2661)
6096 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6097 - don't assign transno to errored transactions (2742)
6098 - don't delete objects on OST if given a bogus objid from MDS (2751)
6099 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6100 - if rq_no_resend, then timeout request after recovery (2432)
6101 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6102 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6103 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6104 - MDS recovery shouldn't delete objects that it has given out (2730)
6105 - if enqueue arrives after completion, don't clobber LVB (2819)
6106 - don't unlock pages twice when trigger_group_io returns error (2814)
6107 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6108 - don't write pages to disk if there was an error (1450)
6109 - don't ping imports that have recovery disabled (2676)
6110 - take buffered bytes into account when balancing socknal conn (2817)
6111 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6112 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6113 - remove little-endian swabbing of llog records (1987)
6114 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6115 - retry reposting request buffers if they fail (1191)
6116 - grow extent at grant time to avoid granting a revoked lock (2809)
6117 - lock revoke doesn't evict page if covered by a second lock (2765)
6118 - disable VM readahead to avoid reading outside lock extents (2805)
6120 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6121 - updated kernel patches for hp-2.4.20 kernel (2681)
6123 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6126 - fix truncated write corruption (2366)
6127 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6129 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6130 - ensure len(uuid) < 37 in lmc (1171)
6131 - fix ia64 OOPS in llog_test (2255)
6132 - zero end of page at obdfilter for partial page writes (2648)
6133 - don't leave stale dentries around after renames (bug 2428)
6134 - fix timeouts when evicting a client with a single lock held (2642)
6135 - set deadline for the initial HELLO message to drain (2634)
6136 - print out dotted-quad IP addresses in the socknal (2302)
6138 - additional debugging for MDS client eviction problem (2443)
6139 - fix mkfsoptions support for osts (2603, 2604)
6141 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6144 - add series for the vanilla 2.6.0 kernel
6145 - add series for the vanilla 2.4.24 kernel
6146 - add series for a cray x86/64 UL kernel drop
6147 - fix xattr patches for the vanilla 2.4.19 series
6149 - generate true UUIDs in lmc (1171)
6150 - have portals stack dumping break in UML (2466)
6151 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6152 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6153 - wake callback waiting threads on client eviction (2460)
6154 - Add --ptldebug and --subsystem to lmc (1719)
6155 - update assertion to allow safe interrupt allocation
6156 - set rq_no_resend for cancel requests (2432)
6157 - recalculate ptlrpcd timeout after resend (2494)
6158 - call vfs_rmdir when removing pending directories (2368)
6159 - fix renaming a file to itself (2429)
6160 - lmc creates a default one-stripe lov (2454)
6161 - expand procfs space to handle large clusters (2326)
6162 - increase UML stack to avoid overflow
6163 - update lconf's list of debug and subsystem masks
6164 - fix lfs find --obd (2510)
6165 - /proc tunable for disabling filter read caching (2591)
6166 - stop rpm packages from altering slapd.conf (2301)
6167 - disable nagle in the socknal under 0conf (2578)
6168 - choose mds inode size based on stripe count (2572)
6169 - fix kernel-source rpm problems (2516)
6171 - add --disable-doc to avoid pdf generation (2421)
6172 - update documentation, tests, type-os, comments
6173 - avoid format warnings on ia64
6174 - remove the TOE NAL
6175 - tiny code cleanups by removing unused fields
6177 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6180 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6181 - remove the most blatant lies from BUILDING file (2371)
6182 - change default debug level to reasonable production setting
6183 - reduce client side cache size to reduce cache flush time
6184 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6185 - make TCP zerocopy and pinger support enabled by default (2476)
6186 - sync writes completed after process exits caused crashes (2319)
6187 - maintain correct mount count on the MDS (2356)
6188 - backout 1557, because 2316 wasn't really fixed
6189 - better file I/O statistics gathering in /proc
6190 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6191 - another kernel patch to fix zero-copy TCP function export
6192 - don't take duplicate lock when processing re-sent getattr (2420)
6193 - lctl uses obd_self_export instead of creating new conn (2353)
6194 - MDS/OST recovery case which requires object creation asserted (2425)
6195 - move lfs from /usr/sbin to /usr/bin in packages
6196 - fix race between mds_client_add and mds_client_free (2417)
6197 - use kmalloc instead of slabs in portals (2430)
6198 - don't create duplicate records when a failover MDS is present (2442)
6199 - remove unnecessary mount age check (2332)
6200 - don't remove directory inodes from locks prematurely (2451)
6201 - don't break if MDS service name is the same as hostname (2103)
6202 - fix races in client write RPC generation when cache full (2482)
6204 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6207 - remove now-unused request->rq_obd (278)
6208 - if an allocation fails, print out how much memory we've used (1933)
6209 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6210 - add the "configurable stack size" patch to most series files (1256)
6211 - ability to write large log records, for 100+ OST configs (2306)
6212 - fix NULL deref when filter_prep fails (2314)
6213 - fix operator precedence error in filter_sync
6214 - dynamic allocation of socknal TX descriptors (2315)
6215 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6216 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6217 - ensure that CWARN messages in Portals always get to the syslog
6218 - __init/__exit are not for prototype decls (ldlm_init/exit)
6219 - x86-64 compile warning fixes
6220 - fix gateway LMC keyword conflict (2318)
6221 - fix MDS lock inversions in getattr/reint paths (1844)
6222 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6223 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6224 - DLM race condition prevented some lock evictions (2328)
6225 - ENOMEM detection and retry on socknal sends (2230)
6226 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6227 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6228 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6229 - better error messages when a client is rejected during recovery (1505)
6230 - avoid cancelling locks which were never granted, after failure (2330)
6231 - fix i_sem/journal inversion in mds_client_add (2333)
6232 - fix truncate/getattr lock cycle deadlock (2334)
6233 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6234 - fix two rare exit paths which could leak an l_lock() ref (2321)
6235 - fencepost error in MDS/OST orphan recovery (2226)
6236 - make log record alignment 8 bytes (1988)
6237 - lstripe now fails when requested offset > ost_count (2237)
6238 - ensure that all kernel series have a complete list.h (1607)
6239 - fix crashes in special-file operations (2316)
6240 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6242 - allow configurable automake binary, for testing new versions
6243 - small update to the lfs documentation
6245 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6247 * fix negative export reference count in fsfilt_sync (2312)
6249 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6250 * release candidate 0.9.1
6252 - orphans are moved into the PENDING directory for possible recovery
6253 - replayed opens now open by fid for orphan/rename safety (1042)
6254 - last close of an orphan inode generates a transno (683)
6255 - chdir() and mount() now pin the directory entry (1020)
6256 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6257 - discard very old requests without processing them (1502)
6258 - remove some common, well-understood CERRORs (1505)
6259 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6260 - clear "grant" flags in OST replies until OST grant code lands (1644)
6261 - fix read performance by not clobbering i_blksize on client (1598)
6262 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6263 - make LRU size a /proc tunable, clears locks when reduced (707)
6264 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6265 - server locks take references on exports now (1558)
6266 - build fixes for 2.4.20-rh trees (1663)
6267 - return an error from lov_create if all OSCs are inactive (1751)
6268 - fix import levels when a reconnect happens without a timeout (1597)
6269 - exit early from mds_open if we get a lookup error (1749)
6270 - partial page read at EOF wouldn't wait for disk before sending (1642)
6271 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6272 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6273 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6274 - write/truncate lock inversion (1639)
6275 - Don't auto-load obdclass, portals modules during cleanup (1495)
6276 - fix timestamps from jumping to "now" (1763)
6277 - extra journal assertions (1648)
6278 - add an extra multiunlink test (1771)
6279 - fix read_record/write_record API (1776)
6280 - fix leak of offset_extent, possible incorrect i_size later (1772)
6281 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6282 - Debugging for the kqswnal_get_idle_tx problems (1820)
6283 - Allow recovery to be attempted multiple times (1536)
6284 - Write out MDS last_rcvd file after it is first created (1600)
6285 - Fix tx_descriptor leak in failed transmit situations (1827)
6286 - ext3 journaling fixes for assertion failure after IO error (1871)
6287 - class_export_put() on freed export after completion AST error (1896)
6288 - Fix revalidate looping in VFS (1322)
6289 - Don't access a freed export during MDS_REINT timeout (1521)
6290 - Add open-unlink recovery support on the MDS (1673,1764)
6291 - Return an error if no MDS data was read from last_rcvd (1946)
6292 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6293 - Don't setup a disk device that doesn't match exported UUID (317)
6294 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6295 - avoid committing NULL handle in force close
6296 - local.sh is now a one-stripe LOV configuration
6297 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6298 - fix ext3 htree duplicate directory entry corruption (1516)
6299 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6300 - update to new LOV EA format (2097)
6301 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6303 2003-06-15 Phil Schwan <phil@clusterfs.com>
6306 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6307 - per-import/export recovery handling (958, 931, 959)
6308 - multiple last-rcvd slots, for serving multiple FSes (949)
6309 - connections are again shared between multiple imp/exports (963, 964)
6310 - "umount -f" would hang if any requests needed to be sent (393, 978)
6311 - avoid pinning large req buffer by copying for queued messages (989)
6312 - add "uuid" to "lctl device" command to help upcalls (991)
6313 - "open" RPCs with transnos would confuse recovery counters (1037)
6314 - do proper endian conversion of all wire messages (288, 340, 891)
6315 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6316 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6317 - fix ptlrpc_connection leak in target_handle_connect (1174)
6318 - fix import refcounting bug in OST and MDS cleanup (1134)
6319 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6320 - fix very unlikely obd_types race condition (501)
6321 - remove osc_open hack for echo_client (1187)
6322 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6323 - a failure in read_inode2 leads to deadlock (1139)
6324 - cancel ack-locks as soon as transaction is committed (1072)
6325 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6326 - make sure to commitrw after any preprw to avoid deadlock (1162)
6327 - failing to execute a file in a lustre FS would lock inode (1203)
6328 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6329 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6330 - obd_uuid2tgt would walk off the end of the list (1255)
6331 - on IA64 the osc would give portals incorrect bulk size (1258)
6332 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6333 - fix lock inversion caused by new llite matching code (1282)
6334 - limit the number of dirty pages on a client to 10MB (1286)
6335 - timed out locks were not being corrected cancelled (1289)
6336 - fix O_DIRECT above 4GB on IA-32 (1292)
6337 * major user-visible changes
6338 - fail out/fail over policy now controlled by the upcall (993)
6340 - add OBD_PING to check server availability and failure (954)
6341 - lustre messages are now sent in sending host order (288, 340, 891)
6342 - add eadatalen to MDS getattr reply (340)
6343 - OST read replies may contain second buffer, with per-page status (593)
6345 2003-03-11 Phil Schwan <phil@clusterfs.com>
6348 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6349 - failed open()s could cause deadlock; fixed (867, 869)
6350 - stop cancelling OST locks when files are closed (481)
6351 - overlapping XID spaces caused network corruption (851, 853)
6352 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6353 - setattr_raw added, to do single-RPC, server-side setattrs
6354 - lmc/lconf syntax change for OST UUIDs
6355 - fix crashy race condition between ptlrpc_free_req and osc_close
6356 - don't use request in mdc_enqueue if we hit a timeout (889)
6357 - don't set the inode i_size for regular files from the MDS (896)
6358 - handle out of order completion AST (842)
6359 - don't LBUG if a lock request times out after receiving AST (913)
6360 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6361 - if a bad lock AST arrives, send an error instead of dropping entirely
6362 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6363 - fix leak in bulk IO when only partially completed (899, 900, 926)
6364 - fix O_DIRECT for ia64 (55)
6365 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6366 - C-z after timeout could hang a process forever; fixed (977)
6368 - client-side I/O cache (678, 924, 929, 941, 970)
6370 - READPAGE and SETATTRs which don't take server-side locks get
6373 2003-02-11 Phil Schwan <phil@clusterfs.com>
6376 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6377 - Fix fsfilt_extN_readpage() to read a full page of directory
6378 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6379 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6380 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6381 - Most intent operations were removed, in favour of a new RPC mode
6382 that does a single RPC to the server and bypasses most of the VFS
6383 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6384 - Aggressively cancel local locks on DLM servers
6385 - mds_reint_unlink sends EA to the client if it's the last nlink.
6386 client uses that EA to unlink OST objects.
6387 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6388 - recursive symlinks were fixed (439)
6389 - fixed NULL deref in DEBUG_REQ
6390 - filter_update_lastobjid no longer calls sync, which annoyed extN
6391 - fixed multi-client small-writes to a single file problem (445)
6392 - fixed mtime updates during file writes (607)
6393 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6394 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6395 - fixed Linux OST size reporting problem (444, 656)
6396 - OST now updates object mtime with writes or setattr (607, 619)
6397 - client verifies file size before zeroing page past EOF (445)
6398 - OST now writes last allocated objid to disk with allocation (108)
6399 - LOV on echo now works (409)
6401 - mds_reint_unlink sends a new buffer, with the EA included. this
6402 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6403 set if a regular file was being unlinked, and it was the last link
6404 - use PtlGet from the target for bulk writes (315)
6405 - OST now updates object mtime with writes or setattr (607, 619)
6406 - LDLM now has a grant-time callback to revalidate locked items, if
6408 - Many MDS operations were reorganized to combat race conditions
6410 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6411 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6413 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6416 - Fully reactivate OST imports after reconnection (512, others)
6417 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6418 - More graceful error handling for truncating on dead OST (515)
6419 - Don't error out unless we're actually accessing dead stripes (474)
6420 - Fix garbage sizes when stripes are missing (410)
6421 - LRU counters were broken, causing constant lock purge (433, 432)
6422 - garbage on read from stripes with failed OSTs (441)
6423 - mark OSCs as active before reconnecting during recovery (438)
6424 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6425 - lfind did not preserve OST order in output (443)
6426 - symlinks cause hung clients, incorrect data (439)
6427 - stop dereferencing request after dropping refcount (457)
6428 - don't LASSERT(spin_is_locked) on non-SMP (455)
6429 - fixes for many rename() bugs
6430 - fstat didn't correctly synchronize attributes (399)
6431 - server must handle lock cancellation during blocking AST prep (487)
6432 - bulk descriptors were free()d too soon (511)
6433 - fix paths in lconf, which would load incorrect modules (451, 507)
6434 - fix confusing lconf 'host not found' error message (386)
6435 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6436 - fix race condition in mdc_blocking_ast() for inode access (526)
6437 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6438 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6439 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6440 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6441 - fix lstripe in several ways (kernel side) (527)
6442 - fix request leak in ldlm_cli_enqueue (262)
6443 - incorrect OSC was marked inactive after OST failure
6444 - call mds_fs_cleanup before unmounting filesystem (524)
6445 - fix races between taking ns_lock and ldlm_lock_change_resource
6446 - fix races updating LOV export open file list
6447 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6448 - fix recovery NULL deref in ldlm_cli_cancel_unused
6449 - fix some DLM races by using new hash table for lock handles (419)
6450 - permit the client to specify desired inodes, at replay
6451 - duplicate requests when we queue them for replay reintegration
6452 - fix last_rcvd offset calculation
6453 - sync after each recovered transaction, so we always make progress
6454 - never, not always, ERESTART requests without transnos
6455 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6457 - skip replay if the MDS says that the client is already connected
6458 - don't check for a recovery-enabled export to match lctl's UUID
6459 - don't INC_USE_COUNT for phantom exports
6460 - don't crash when cleaning up phantom exports (567)
6461 - don't double-finish or set replay data for errored mdc_open requests
6462 - abort requests when they time out, so we don't get old replies
6463 - send/receive replies for AST messages again
6464 - if the client says that it doesn't have the lock, cancel it on the
6466 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6467 mark it as destroyed, it will all work out when decref is called
6468 - fix module use counts (22, 581)
6470 - ASTs now expect a reply (server cancels lock on error reply)
6472 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6475 - fix many simultaneous client startup (392)
6476 - fix dentry->d_it clobbering
6477 - credentials weren't being shipped for readdir/getattr operations
6478 - remove invalid assertions triggered during some concurrent MD
6480 - proper Lustre versions added (336, 389)
6481 - fix memory leak for create error case (398)
6482 - fix LOV locking bug that would get cli/srv out of sync
6483 - fix echo client over LOV (409)
6484 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6485 - fix double-O_EXCL intent crash (424)
6486 - avoid sending multiple lock CANCELs (352)
6488 - MDS can do multi-client recovery (modulo bugs in new code)
6490 - many updates, edits, cleanups
6492 2002-11-18 Phil Schwan <phil@clusterfs.com>
6495 - fix null d_it dereference (346)
6496 - fix full OST/dbench hang (333)
6497 - fix permission problem with file removal (286)
6498 - fix removal of OSCs from LOV when they fail
6499 - fix NULL deref during bulk timeout (214)
6500 - fix problems related to multiple filesystems on one MDS (241)
6501 - fixed serious subtle metadata locking bugs
6502 - free locks on clients when inodes are removed due to memory
6504 - fix inode pointer in lock data (285)
6505 - partial support for multiple MDS on a single host (241)
6506 - data locks weren't cancelled at clear_inode time (290, 311)
6507 - intent locks could lead to unbounded lock growth (205)
6508 - added a maximum lock count, an LRU list, and a flusher
6509 - fix multiple rename (365)
6510 - properly abstracted the echo client
6511 - OSC locked 1 byte too many; fixed
6512 - rewrote brw callback code:
6513 - fixed recovery bugs related to LOVs (306)
6514 - fixed too-many-pages-in-one-write crash (191)
6515 - fixed (again) crash in sync_io_timeout (214)
6516 - probably fixed callback-related race (385)
6518 - Add capability to MDS protocol
6519 - LDLM cancellations and callbacks on different portals
6521 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6524 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6525 - defer open object destruction to close time (601981, 138)
6526 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6527 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6528 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6529 - don't call obd_setattr in ll_file_release for destroyed objects
6530 * protocol change to lustre_msg: move |version| and add |flags|
6531 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6532 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6533 * added replay of create, unlink, link and rename operations during
6534 MDS failover; recovery should be much more robust now
6535 * remove failed OSCs from LOVs (only lov_create uses this so far)
6536 * the lustre-HOWTO was brought (more) up to date (582544)
6538 2002-10-23 Phil Schwan <phil@clusterfs.com>
6541 - in-use dentries weren't being reused properly (617851)
6542 - prevent multiple LDLM setup (599178)
6543 - fix LOV size calculations for truncate (617853)
6544 - fix client handling of MDS intent errors (POSIX)
6545 - fix permission bug in lovstripe.c test (624321)
6546 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6547 - truncate past end of file could corrupt data
6548 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6549 - a race in recovery could return ETIMEDOUT to apps (623947)
6550 - building outside the source directory was fixed
6551 * the lustre-HOWTO was brought (more) up to date (582544)
6552 * major progress was made on recovery functionality
6554 2002-10-10 Phil Schwan <phil@clusterfs.com>
6557 - recovery deadlock fix
6558 - rm -rf causes LBUG fix (617817)
6559 - file open by multiple tasks fix (618962)
6560 - directory permissions bugs (602707 and 620007)
6561 - journal_stop fixed with locking (611313)
6562 - O_APPEND failures resolved (618273, perhaps 614459)
6563 - lconf PATH fix (619770)
6564 - IA64 build fix (621450)
6565 - RPC buffer sizes scale with amount of memory
6567 2002-10-01 Phil Schwan <phil@clusterfs.com>
6570 - locks would be cancelled without throwing away data pages,
6571 resulting in inconsistent data (605627)
6572 - inode attributes were not always being refreshed (605627, 612449)
6573 - lconf now continues to cleanup after lctl reports an error
6574 - MDS now enforces user permissions (602707)
6575 - lprocfs cleanup fixed, but not yet enabled (614157)
6576 - fixed infinite server hang, should a client not respond to an AST
6577 - avoid going into recovery if user calls readlink() with a buffer
6578 that's too small (613941)
6579 - AST RPCs no longer require replies (614867) -- this may be changed
6580 - don't crash server if client sends an IOV that's too big (611336)
6581 - fixed lock conversion deadlock (611892)
6582 - fixed the following of symlinks (614622)
6583 * recovery: the server can remove locks from a client that dies, other
6584 clients can make progress
6585 * more extN patch fixes
6586 * compile-time configurable ptlrpc buffer allocations
6588 - collaborative read cache document
6589 - Lustre Lite Performance CDR document-in-progress
6591 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6594 - fix typo in patch-2.4.18
6596 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6599 - clear ptlrpc request each time in handle_incoming_request()
6600 - unlink of files now destroys the object on the OST
6602 2002-09-19 Peter Braam <braam@clusterfs.com>
6604 * add hard link support
6605 * change obdfile creation method
6606 * kernel patch changed
6608 2002-09-19 Peter Braam <braam@clusterfs.com>
6611 - stack overflow bug in extN fixed
6613 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6615 * documentation updates
6616 - add man pages for config tools
6617 - update tests/README to describe testing with new config tools
6618 - finish metadata API descriptions
6619 * bug fixes and cleanups
6620 - statfs workaround for 16TB limit
6621 - LOV stripe allocation improved, can stripe on subset of OSTs
6622 - LOV file size/IO offset was wrong for files > 4GB in size
6623 - object EA data was being dropped, caused files to be unreadable
6624 - memory overflow with non-LOV OST caused memory corruption
6625 - fixed regression tests to work with new config tools, obdfilter
6626 - fixed bug when directory size became larger than 1 block
6627 - fixed bug (for single client case) when PWD was deleted
6628 - invalidate local directory pages when doing intent-based ops
6629 - avoid LDLM oops when lock callback contained bad data
6631 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6633 * documentation updates
6634 * bug fixes and cleanups
6635 - configuration tools
6638 - 64-bit compile warnings
6639 - 64-bit internal statfs data
6641 * test_brw on persistent OST devices
6643 * lprocfs (disabled)
6645 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6647 * documentation updates
6648 * bug fixes and cleanups
6649 * configuration tools
6651 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6654 * many small fixes to 0_5_4
6655 * io/network handling
6656 * thinkos in MDS operations
6658 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6661 * crucial basic fixes to 0.5.3
6662 * IOR, Iozone work over Elan
6665 2002-08-07 Phil Schwan <phil@clusterfs.com>
6666 * version 0_5_3, our first alpha
6667 * we use the new Portals iovs
6668 * documentation updates
6669 * bug fixes and cleanups
6670 * small changes in the DLM wire protocol
6672 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6673 * version 0_5_1 with some initial stability,
6674 * locking on MD and file I/O.
6675 * documentation updates
6676 * several bug fixes since 0.5.0
6677 * small changes in wire protocol
6679 2002-07-18 Phil Schwan <phil@clusterfs.com>
6681 * delivered as Lustre Light Alpha
6682 * fixed a crash after handling invalid MDS requests
6683 * fixed directory pages for architectures with non-4k pages sizes
6685 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6686 * release version v0_4_4
6687 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6689 * Fixes a number of interruption problems with OST operations.
6690 * Update documentation for portals header changes
6691 * Move all wire protocol structs/defines to lustre_idl.h
6692 * Fixes symlink length bug.
6693 * Add tcpdump to repository.
6695 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6696 * release version v0_4_3
6697 * Fixes statfs for inodes on extN.
6698 * Fixes bug in runtests which would delete /etc/hosts.
6699 * Use 64-bit object IDs wherever possible (not into VFS though)
6700 Remove ost_get_info, which is unused by lustre, and out of date.
6702 2002-07-03 Peter Braam <braam@clusterfs.com>
6703 * release version v0_4_2 Fixes a lookup error (type not passed)
6704 * move forward to head of Portals
6705 * move forward to latest Lustre kernel
6707 2002-06-25 Peter Braam <braam@clusterfs.com>
6708 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob