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.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.162-94.69 (SLES12 SP3)
11 4.4.0-131 (Ubuntu 16.04)
12 4.15.0-32 (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.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 4.4.0-131 (Ubuntu 16.04)
25 4.15.0-32 (Ubuntu 18.04)
26 vanilla linux 4.15.0 (ZFS only)
27 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
28 * Recommended ZFS / SPL version: 0.7.9
29 * Tested with ZFS / SPL version: 0.7.9
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
39 --------------------------------------------------------------------------------
43 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
44 for currently supported client and server kernel versions.
45 * Server known to build on the following kernels (others may also work):
46 2.6.32-696.18.7.el6 (RHEL6.9)
47 3.10.0-862.14.4.el7 (RHEL7.5)
48 3.10.0-957.el7 (RHEL7.6)
49 4.4.120-92.70 (SLES12 SP2)
50 4.4.162-94.69 (SLES12 SP3)
51 4.4.0-131 (Ubuntu 16.04)
52 4.15.0-32 (Ubuntu 18.04)
53 vanilla linux 4.6.7 (ZFS only)
54 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
55 * Client known to build on unpatched kernels (others may also work):
56 2.6.32-696.18.7.el6 (RHEL6.9)
57 2.6.32-754.6.3.el6 (RHEL6.10)
58 3.10.0-862.14.4.el7 (RHEL7.5)
59 3.10.0-957.el7 (RHEL7.6)
60 4.14.0-49.13.1.el7a (RHEL7.5)
61 4.14.0-115.2.2.el7a (RHEL7.6)
62 4.4.120-92.70 (SLES12 SP2)
63 4.4.162-94.69 (SLES12 SP3)
64 4.4.0-131 (Ubuntu 16.04)
65 4.15.0-32 (Ubuntu 18.04)
66 vanilla linux 4.15.0 (ZFS only)
67 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
68 * Recommended ZFS / SPL version: 0.7.9
69 * Tested with ZFS / SPL version: 0.7.9
70 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
71 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
72 stack overflow. For more information, please refer to bugzilla 17630.
73 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
74 the re-exporting nodes to be mounted with "32bitapi" mount option
75 * Note for Ubuntu 14.04.5:
76 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
77 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
79 --------------------------------------------------------------------------------
82 2018-04-03 Intel Corporation
84 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
85 for currently supported client and server kernel versions.
86 * Server known to build on patched kernels:
87 2.6.32-431.29.2.el6 (RHEL6.5)
88 2.6.32-504.30.3.el6 (RHEL6.6)
89 2.6.32-573.26.1.el6 (RHEL6.7)
90 2.6.32-642.15.1.el6 (RHEL6.8)
91 2.6.32-696.18.7.el6 (RHEL6.9)
92 3.10.0-693.21.1.el7 (RHEL7.4)
93 3.0.101-0.47.71 (SLES11 SP3)
94 3.0.101-107 (SLES11 SP4)
95 3.12.74-60.64.40 (SLES12 SP1)
96 4.4.114-92.64 (SLES12 SP2)
97 4.4.114-94.11 (SLES12 SP3)
98 3.13.0-101 (Ubuntu 14.04, ZFS only)
99 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
100 4.4.0-116 (Ubuntu 16.04)
101 vanilla linux 4.6.7 (ZFS only)
102 * Client known to build on unpatched kernels:
103 2.6.32-431.29.2.el6 (RHEL6.5)
104 2.6.32-504.30.3.el6 (RHEL6.6)
105 2.6.32-573.26.1.el6 (RHEL6.7)
106 2.6.32-642.15.1.el6 (RHEL6.8)
107 2.6.32-696.18.7.el6 (RHEL6.9)
108 3.10.0-693.21.1.el7 (RHEL7.4)
109 3.0.101-0.47.71 (SLES11 SP3)
110 3.0.101-107 (SLES11 SP4)
111 3.12.74-60.64.40 (SLES12 SP1)
112 4.4.114-92.64 (SLES12 SP2)
113 4.4.114-94.11 (SLES12 SP3)
114 3.13.0-101 (Ubuntu 14.04)
115 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
116 4.4.0-116 (Ubuntu 16.04)
118 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
119 * Recommended ZFS / SPL version: 0.7.6
120 * Tested with ZFS / SPL version: 0.7.6
121 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
122 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
123 stack overflow. For more information, please refer to bugzilla 17630.
124 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
125 the re-exporting nodes to be mounted with "32bitapi" mount option
126 * Note for Ubuntu 14.04.5:
127 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
128 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
130 --------------------------------------------------------------------------------
132 2017-07-05 Intel Corporation
134 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
135 for currently supported client and server kernel versions.
136 * Server known to build on patched kernels:
137 2.6.32-431.29.2.el6 (RHEL6.5)
138 2.6.32-504.30.3.el6 (RHEL6.6)
139 2.6.32-573.26.1.el6 (RHEL6.7)
140 2.6.32-642.15.1.el6 (RHEL6.8)
141 2.6.32-696.3.1.el6 (RHEL6.9)
142 3.10.0-514.21.1.el7 (RHEL7.3)
143 3.0.101-0.47.71 (SLES11 SP3)
144 3.0.101-100 (SLES11 SP4)
145 3.12.74-60.64.40 (SLES12 SP1)
146 4.4.59-92.17 (SLES12 SP2)
147 vanilla linux 4.6.7 (ZFS only)
148 * Client known to build on unpatched kernels:
149 2.6.32-431.29.2.el6 (RHEL6.5)
150 2.6.32-504.30.3.el6 (RHEL6.6)
151 2.6.32-573.26.1.el6 (RHEL6.7)
152 2.6.32-642.15.1.el6 (RHEL6.8)
153 2.6.32-696.3.1.el6 (RHEL6.9)
154 3.10.0-514.21.1.el7 (RHEL7.3)
155 3.0.101-0.47.71 (SLES11 SP3)
156 3.0.101-100 (SLES11 SP4)
157 3.12.74-60.64.40 (SLES12 SP1)
158 4.4.59-92.17 (SLES12 SP2)
160 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
161 * Recommended ZFS / SPL version: 0.6.5.9
162 * Tested with ZFS / SPL version: 0.6.5.9
163 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
164 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
165 stack overflow. For more information, please refer to bugzilla 17630.
166 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
167 the re-exporting nodes to be mounted with "32bitapi" mount option
169 --------------------------------------------------------------------------------
171 12-07-2016 Intel Corporation
173 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
174 for currently supported client and server kernel versions.
175 * Server known to build on patched kernels:
176 2.6.32-431.29.2.el6 (RHEL6.5)
177 2.6.32-504.30.3.el6 (RHEL6.6)
178 2.6.32-573.26.1.el6 (RHEL6.7)
179 2.6.32-642.11.1.el6 (RHEL6.8)
180 3.10.0-514.el7 (RHEL7.3)
181 3.0.101-0.47.71 (SLES11 SP3)
182 3.0.101-80 (SLES11 SP4)
183 3.12.62-60.64.8 (SLES12 SP1)
184 vanilla linux 4.5.7 (ZFS only)
185 * Client known to build on unpatched kernels:
186 2.6.32-431.29.2.el6 (RHEL6.5)
187 2.6.32-504.30.3.el6 (RHEL6.6)
188 2.6.32-573.26.1.el6 (RHEL6.7)
189 2.6.32-642.11.1.el6 (RHEL6.8)
190 3.10.0-514.el7 (RHEL7.3)
191 3.0.101-0.47.71 (SLES11 SP3)
192 3.0.101-80 (SLES11 SP4)
193 3.12.62-60.64.8 (SLES12 SP1)
195 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
196 * Recommended ZFS / SPL version: 0.6.5.8
197 * Tested with ZFS / SPL version: 0.6.5.8
198 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
199 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
200 stack overflow. For more information, please refer to bugzilla 17630.
201 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
202 the re-exporting nodes to be mounted with "32bitapi" mount option
204 --------------------------------------------------------------------------------
206 02-29-2016 Intel Corporation
208 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
209 for currently supported client and server kernel versions.
210 * Server known to build on patched kernels:
211 2.6.32-431.29.2.el6 (RHEL6.5)
212 2.6.32-504.30.3.el6 (RHEL6.6)
213 2.6.32-573.12.1.el6 (RHEL6.7)
214 3.10.0-327.3.1.el7 (RHEL7.2)
215 3.0.101-0.47.71 (SLES11 SP3)
216 3.0.101-68 (SLES11 SP4)
217 vanilla linux 4.2.1 (ZFS only)
218 * Client known to build on unpatched kernels:
219 2.6.32-431.29.2.el6 (RHEL6.5)
220 2.6.32-504.30.3.el6 (RHEL6.6)
221 2.6.32-573.12.1.el6 (RHEL6.7)
222 3.10.0-327.3.1.el7 (RHEL7.2)
223 3.0.101-0.47.71 (SLES11 SP3)
224 3.0.101-68 (SLES11 SP4)
227 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
228 * Recommended ZFS / SPL version: 0.6.4.2
229 * Tested with ZFS / SPL version: 0.6.4.2
230 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
231 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
232 stack overflow. For more information, please refer to bugzilla 17630.
233 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
234 the re-exporting nodes to be mounted with "32bitapi" mount option
236 --------------------------------------------------------------------------------
238 03-10-2015 Intel Corporation
240 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
241 for currently supported client and server kernel versions.
242 * Server known to build on patched kernels:
243 2.6.32-431.29.2.el6 (RHEL6.5)
244 2.6.32-504.8.1.el6 (RHEL6.6)
245 3.0.101-0.46 (SLES11 SP3)
246 * Client known to build on unpatched kernels:
247 2.6.32-431.29.2.el6 (RHEL6.5)
248 2.6.32-504.8.1.el6 (RHEL6.6)
249 3.10.0-123.20.1.el7 (RHEL7)
250 3.0.101-0.46 (SLES11 SP3)
251 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
252 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
253 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
254 stack overflow. For more information, please refer to bugzilla 17630.
255 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
256 the re-exporting nodes to be mounted with "32bitapi" mount option
258 Severity : enhancement
260 Description: control OST-index in IDIF via ROCOMPAT flag.
261 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
262 last_rcvd file. For new formatted OST device, it will be auto set;
263 for the case of upgrading from old OST device, you can enable it
264 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
265 enabled, for new created OST-object, its IDIF-in-LMA will contain
266 the OST-index; for the existing OST-object, the OSD will convert
267 old format IDIF as new format IDIF with OST-index stored in the
268 LMA EA when accessing such OST-object or via OI scrub. Once such
269 flag is enabled, it cannot be reverted back, so the system cannot
270 be downgraded to the orignal incompatible version.
272 --------------------------------------------------------------------------------
274 07-30-2014 Intel Corporation
276 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
277 for currently supported client and server kernel versions.
278 * Server known to build on patched kernels:
279 2.6.32-431.20.3.el6 (RHEL6.5)
280 3.0.101-0.31 (SLES11 SP3)
281 * Client known to build on unpatched kernels:
282 2.6.32-431.20.3.el6 (RHEL6.5)
283 3.0.101-0.31 (SLES11 SP3)
286 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
287 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
288 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
289 stack overflow. For more information, please refer to bugzilla 17630.
290 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
291 the re-exporting nodes to be mounted with "32bitapi" mount option
293 --------------------------------------------------------------------------------
295 10-31-2013 Intel Corporation
297 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
298 * Server builds for kernels:
299 2.6.32-358.18.1.el6 (RHEL6.4)
300 * Client builds for unpatched kernels:
301 2.6.32-358.18.1.el6 (RHEL6.4)
302 3.0.92-0.8 (SLES11 SP3)
304 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
305 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
306 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
307 stack overflow. For more information, please refer to bugzilla 17630.
308 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
309 the re-exporting nodes to be mounted with "32bitapi" mount option
311 --------------------------------------------------------------------------------
313 05-30-2013 Intel Corporation
315 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
316 * Server builds for kernels:
317 2.6.32-358.6.2.el6 (RHEL6.4)
318 * Client builds for unpatched kernels:
319 2.6.32-358.6.2.el6 (RHEL6.4)
320 3.0.74-0.6.6 (SLES11 SP2)
322 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
323 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
324 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
325 stack overflow. For more information, please refer to bugzilla 17630.
326 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
327 the re-exporting nodes to be mounted with "32bitapi" mount option
329 --------------------------------------------------------------------------------
331 9-30-2012 Whamcloud, Inc.
333 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
334 * Server support for kernels:
335 2.6.32-279.5.1.el6 (RHEL6)
336 * Client support for unpatched kernels:
337 2.6.32-279.5.1.el6 (RHEL6)
339 2.6.18-238.19.1.el5 (RHEL5)
340 2.6.16.60-0.69.1 (SLES10)
341 * Recommended e2fsprogs version: 1.41.90.wc4
342 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
343 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
344 stack overflow. For more information, please refer to bugzilla 17630.
345 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
346 the re-exporting nodes to be mounted with "32bitapi" mount option
348 Severity : enhancement
350 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
352 Severity : enhancement
354 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
356 Severity : enhancement
358 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
360 Severity : enhancement
362 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
364 Severity : enhancement
366 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
368 --------------------------------------------------------------------------------
370 04-30-2012 Whamcloud, Inc.
372 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
373 * Server support for kernels:
374 2.6.32-220.4.2.el6 (RHEL6)
375 * Client support for unpatched kernels:
376 2.6.32-220.4.2.el6 (RHEL6)
378 2.6.18-238.19.1.el5 (RHEL5)
379 2.6.16.60-0.69.1 (SLES10)
380 * Recommended e2fsprogs version: 1.41.90.wc4
381 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
382 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
383 stack overflow. For more information, please refer to bugzilla 17630.
384 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
385 the re-exporting nodes to be mounted with "32bitapi" mount option
387 Severity : enhancement
389 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
391 Severity : enhancement
393 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
395 Severity : enhancement
397 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
399 Severity : enhancement
401 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
403 --------------------------------------------------------------------------------
405 09-30-2011 Whamcloud, Inc.
407 * Server support for kernels:
408 2.6.18-238.19.1.el5 (RHEL5)
409 2.6.32-131.6.1.el6 (RHEL6)
410 * Client support for unpatched kernels:
411 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
412 2.6.18-238.19.1.el5 (RHEL5)
413 2.6.32-131.6.1.el6 (RHEL6)
414 2.6.29.4-167.fc11 (FC11)
415 2.6.16.60-0.69.1 (SLES10)
417 * Recommended e2fsprogs version: 1.41.90.wc3
418 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
419 Lustre filesystem with 4K stack may cause a stack overflow. For more
420 information, please refer to bugzilla 17630.
421 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
422 mounted with "32bitapi" mount option
424 Severity : enhancement
425 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
427 Severity : enhancement
428 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
430 Severity : enhancement
431 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
434 Frequency : only with SLES10
436 Description: Use OFED "KMP" provided by Novell
437 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
438 Lustre is now built against this package. That means you need to
439 install the ofed-kmp package from Novell for the patchless client.
440 Note that the ofed-kmp that Novell ships may not exactly match the
441 kernel version but should still be compatible.
443 Severity : enhancement
445 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
446 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
448 --------------------------------------------------------------------------------
450 2010-07-15 Oracle, Inc.
452 * Server support for kernels:
453 2.6.18-164.11.1.el5 (RHEL 5)
454 2.6.18-164.11.1.0.1.el5 (OEL 5)
455 * Client support for unpatched kernels:
456 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
457 2.6.18-164.11.1.el5 (RHEL 5),
458 2.6.18-164.11.1.0.1.el5 (OEL 5)
459 2.6.16.60-0.42.8 (SLES 10),
461 2.6.29.4-167.fc11 (FC11)
462 * Recommended e2fsprogs version: 1.41.10-sun2
464 Severity : enhancement
466 Description: Distribute lfs_migrate script with lustre
467 Details : The lfs_migrate mentioned in the manual is now included in the
468 lustre tree in lustre/scripts and will distributed in lustre RPMs.
470 Severity : enhancement
472 Description: Failover nid list added to import proc file.
474 Severity : enhancement
476 Description: "som_preview" mount option required to use SOM
477 Details : The clients will need to use som_preview mount option
481 Description: Lustre does not do 1MB IOs to HW RAID
482 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
484 Severity : enhancement
486 Description: lfs quota output is not very convenient for awk/sed-parsing
487 Details : Some positions in lfs quota output table could be empty or
488 non-empty which made it hard to parse it with scripts, now
489 a dash is put instead of space where there is not supposed
494 Description: Update to OFED 1.5.1
495 Details : OFED 1.5.1 will be used in environments that don't use
496 vendor supplied OFED.
500 Description: Remove lnet. prefix from lctl params display
501 Details : Also renamed lnet.memused to lnet_memused to
506 Description: Use conf_param -d to remove permanant settings.
510 Description: Add list_param -R
511 Details : lctl list_param now supports recursive mode.
514 Bugzilla : 21829, 22312
515 Description: Fix llobdstat.
516 Details : 1. Now updating the write_bytes counter
517 2. Port script fixes from 1.8.
521 Description: decrease the usage of memory on clients.
522 Details : 1. On clients, recycle dentries and inodes unused.
523 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
528 Description: llobdstat fix and enhancement
529 Details : add a counter to set a limit to how many samples will be returned
530 fix a wildcard in the path to limit to obdfilter stats only
534 Description: lfs check only allowed for root
535 Details : Code cleanups and added a regresion test of non-root use of
540 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
541 Details : Fix hr->hr_index overload.
545 Description: Fix FIEMAP functionality.
549 Description: FC11 patchless client support added
553 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
554 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
558 Description: New acc-sm script and various test-framework enhancments.
562 Description: Previously landed patch was reverted.
566 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
570 Description: Add ldiskfs maxdirsize mount option.
574 Description: Fix mkfs.lustre support for 16TB LUNs.
578 Description: Print mmp_check_interval
579 Details : print mmp_check_interval and make it possible to abort
580 mount operation in case it takes too long.
584 Description: Add cascading_rw to lustre mpi tests.
588 Description: Store SOM data in EA.
589 Details : Also several SOM related cleanups.
593 Description: Add refcount to md_open_data
597 Description: Use refcount on lov_request_set to prevent memory corruption.
601 Description: Statahead: use parent dir's i_mutex for synchronization.
605 Description: loadgen improvements.
609 Description: Enable peer health detection in ksocklnd and ko2iblnd.
613 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
617 Description: Fix race on obd_devcie:md_stats
621 Description: Fix SLES 11 client builds.
622 Details : Resolved uninitalized pointer gcc warnings.
626 Description: Allow non-root access for "lfs check".
627 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
631 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
633 Severity : enhancement
635 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
636 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
640 Description: Sanity test_27p fails.
641 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
642 Sanity tests 27 were cleaned up and improved.
646 Description: Unify req format on client/servers
647 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
648 buffers and swabbers approach..
653 Description: drop join file stuff.
655 Severity : enhancement
657 Description: Add 2.6.27 and SLES11 (patchless client) support
661 Description: Fix in ptlrpc_expire_one_request() to print the signed time
664 Severity : enhancement
666 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
669 Severity : enhancement
671 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
673 Severity : enhancement
675 Description: correctly handle big reply message.
676 Details : send LNet event if reply is bigger than buffer and adjust this buffer
681 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
683 Severity : enhancement
685 Description: Add support for OEL5.
689 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
691 Severity : enhancement
693 Description: "weak-modules" support
694 Details : Implement "weak-modules" support which enables kernel modules
695 to be used with any kernel that implements the same kABI. In
696 order to achieve this modules are now installed in
697 /lib/modules/$(uname -r)/updates/kernel on all distributions.
701 Description: client could not reconnect to OST because of an active request.
702 Details : The client now retries to reconnect to the same server, if a
703 connect request failed with EBUSY or -EAGAIN. Test case 35b
704 conf-sanity.sh has been created.
708 Description: Don't allow make backward step on assignin osc next id.
709 Details : race between allocation next id and ll_sync thread can be cause
710 of set wrong osc next id and can be kill valid ost objects.
714 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
718 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
722 Description: Error handling in osc_statfs_interpret() has been improved.
723 Details : Check in osc_statfs_interpret() for EBADR.
727 Description: Do not send statfs() requests to OSTs disabled by administrator.
728 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
730 Severity : enhancement
732 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
734 Severity : enhancement
736 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
738 Severity : enhancement
740 Description: Enhancement for lfs(1) command to use numeric uid/gid.
744 Description: Improved cleanup procedure for conf-sanity test_46a.
748 Description: Do not start bulk transfer, if client tried to reconnect.
749 Details : Do not start bulk transfer, if client tried to reconnect,
750 but failed due to active rpcs. Check for exp_abort_active_req
751 has been implemented.
755 Description: Prevent inconsistences between linux and lustre mount structures.
756 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
757 Make the sleep interruptible.
759 Severity : enhancement
761 Description: provide server to client comms path
762 Details : server to client communications path via new LDLM_SET_INFO rpc
766 Description: truncate starts GFP_FS allocation under transaction causing deadlock
767 Details : ldiskfs_truncate calls grab_cache_page which may start page
768 allocation under an open transaction. This may lead to
769 calling prune_icache with consequent lustre reentrance.
773 Description: send correctly lsm on open replay
774 Details : MDS is trust to LSM size on replay open, but client can set wrong size
779 Description: enable client interface failover
780 Details : When a child reconnects from another NID, properly update export
781 nid hash position and ldlm reverse import.
783 Severity : enhancement
785 Description: Communicate OST degraded/readonly state via statfs to MDS
786 Details : Flags in the statfs returned from OSTs indicate whether the
787 OST is in a degraded RAID state, or if the filesystem has
788 turned read-only after a filesystem error is detected.
793 Description: don't panic if EPROTO was hit when reading symlink
794 Details : correctly handling request reference in error cases.
796 Severity : enhancement
798 Description: Add LustreNetLink, a kernel-userspace communcation path.
800 Severity : enhancement
802 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
807 Description: access to llog context before init.
808 Details : move handling CATALOGS file at osc layer and forbid access to llog
812 Frequency : rare, only when client is writing to OST on local node
814 Description: Avoid deadlock for local client writes
815 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
816 memory freeing context. This allows OST threads to set the
817 PF_MEMALLOC flag on task structures in order to allocate memory
818 from reserved pools and complete IO.
819 Use GFP_HIGHUSER for OST allocations for non-local client writes,
820 so that the OST threads generate memory pressure and allow
821 inactive pages to be reclaimed.
823 Severity : enhancement
825 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
830 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
831 Details : move ldlm namespace creation in setup phase to avoid grab
832 _lprocfs_lock with cli_sem held.
836 Description: Temporarily disable grant shrink.
837 Details : Disable the feature for debugging.
840 Frequency : only when formatting test filesystems
842 Description: Unable to run several mkfs.lustre on loop devices at the same time
843 Details : mkfs.lustre returns error 256 on the concurrent loop devices
844 formatting. The solution is to proper handle the error.
846 Severity : enhancement
848 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
850 Severity : enhancement
852 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
857 Description: resolve race between obd_disconnect and class_disconnect_exports
858 Details : if obd_disconnect will be called to already disconnected export he
859 forget release one reference and osc module can't unloaded.
861 Severity : enhancement
863 Description: Allow tuning service thread via /proc
864 Details : For each service a new
865 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
866 created that can be used to set min/max thread counts, and get the
867 current number of running threads.
872 Descriptoin: don't return error if have partially created objects for file.
873 Details : lov_update_create_set uses set->set_success as index for created
874 objects, so if some requests will be failed, they will have hole
875 at end of array and qos_shrink_lsm for allocate correct lsm.
877 Severity : enhancement
879 Description: Update OFED support to 1.4.1
884 Description: Reduce small size read RPC
885 Details : Set read-ahead limit for every file and only do read-ahead when
886 available read-ahead pages are bigger than 1M to avoid small size
889 Severity : enhancement
891 Description: move AT tunable parameters for more consistent usage
892 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
894 Severity : enhancement
896 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
897 Details : allow skip disconnected ost for send statfs request and hide error
903 Description: fix racy locking of mballoc block bitmaps causing BUG
904 Details : The locking of the mballoc buddy bitmap and the in-memory
905 block bitmap was using two different spin locks in some
906 cases. This made it possible to incorrectly access the
907 mballoc bitmap while another process was modifying it,
908 causing a sanity assertion to fail. While no on-disk corruption
909 was reported, there was some risk of this happening.
912 Frequency : rare, on llog test 6
914 Descriptoin: don't allow connect to already connected import
915 Details : allowing connect to already connected import is hide connecting problem.
920 Descriptoin: allow kill process which wait statahead result
921 Details : for some reasons 'ls' can stick in waiting result from statahead,
922 in this case need way for kill this process.
924 Severity : enhancement
926 Description: Add state history info file, enhance import info file
927 Details : Track import connection state changes in a new osc/mdc proc file;
928 add overview-type data to the osc/mdc import proc file.
930 Severity : enhancement
932 Description: MDS create should not wait for statfs RPC while holding DLM lock.
934 Severity : enhancement
936 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
941 Descriptoin: OST grant too much space to client even there are not enough space.
942 Details : Client will shrink its grant cache to OST if there are no write
943 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
944 this grant cache if there are already not enough avaible space
945 (left_space < total_clients * 32M).
948 Frequency : start MDS on uncleanly shutdowned MDS device
950 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
951 Details : stay in waiting mds<>ost recovery finished produce random bugs
952 due race between two ll_sync thread for one lov target. send
953 ACTIVATE event only if connect realy finished and import have
957 Frequency : rare, connect and disconnect target at same time
959 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
960 Details : don't call obd_disconnect under lov_lock. this long time
961 operation and can block ptlrpcd which answer to connect request.
966 Descriptoin: don't lose wakeup for imp_recovery_waitq
967 Details : recover_import_no_retry or invalidate_import and import_close can
968 both sleep on imp_recovery_waitq, but we was send only one wakeup
972 Frequency : always with long access acl
974 Descriptoin: mds can't pack reply with long acl.
975 Details : mds don't control size of acl but they limited by reint/getattr
978 Severity : enhancement
980 Description: Update to SLES10 kernel-2.6.16.60-0.33.
982 Severity : enhancement
984 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
987 Frequency : start MDS on uncleanly shutdowned MDS device
989 Descriptoin: aborting recovery hang on MDS
990 Details : don't throttle destroy RPCs for the MDT.
993 Frequency : on remount
995 Description: external journal device not working after the remount
996 Details : clear dev_rdonly flag for external journal devices in
1002 Description: shutdown vs evict race
1003 Details : client_disconnect_export vs connect request race.
1004 if client will evicted at this time - we start invalidate
1005 thread without referece to import and import can be freed
1009 Frequency : rare, need acl's on inode.
1011 Description: client can't handle ost additional correctly
1012 Details : if ost was added after client connected to mds client can have
1013 hit lnet_try_match_md ... to big messages to wide striped files.
1014 in this case need teach client to handle config events about add
1015 lov target and update client max ea size at that event.
1017 Severity : enhancement
1019 Description: Changelogs
1020 Details : Changelogs are a lightweight mechanism to track filesystem
1021 metadata and namespace changes. The changelog is recorded
1022 permanently on the MDTs, and is periodically "consumed" / purged
1023 when records are no longer needed.
1025 Severity : enhancement
1027 Description: compact fld format with extents
1028 Details : Store range of seq rather than every seq in FLD. Seq
1029 controller update FLD rather than clients. In Case of CMD, mdt0
1030 has FLD, all other metadata server act as non persistent proxy
1031 for FLD queries and cache fld entries in fld cache.
1036 Description: don't skip ost target if they assigned to file
1037 Details : Drop slow OSCs if we can, but not for requested start idx.
1038 This means "if OSC is slow and it is not the requested
1039 start OST, then it can be skipped, otherwise skip it only
1040 if it is inactive/recovering/out-of-space.
1044 Description: more cleanup in mds_lov
1045 Details : not send LOV EA under replay, we can't know about they size at this
1046 time. Don't allow client connect to mds before any ost connected,
1047 for avoid problems with LOV EA size and returning EIO to client.
1049 Severity : enhancement
1051 Description: Interoperability at server side (Disk interoperability)
1053 Severity : enhancement
1055 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1057 Severity : enhancement
1059 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1061 Severity : enhancement
1063 Description: New client IO stack (CLIO).
1065 Severity : enhancement
1067 Description: Commit on sharing. Eliminate inter-client dependencies between
1068 uncommitted transactions by doing transaction commits.
1069 Thereby clients may recovery independently.
1072 Frequency : Create a symlink file with a very long name
1074 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1075 Details : If there is no extra space in the request for early cancels,
1076 ldlm_req_handles_avail() returns 0 instead of a negative value.
1078 Severity : enhancement
1080 Description: Add /proc entry for import status
1081 Details : The mdc, osc, and mgc import directories now have
1082 an import directory that contains useful import data for debugging
1083 connection problems.
1085 Severity : enhancement
1087 Description: Re-disable certain /proc logging
1088 Details : Enable and disable client's offset_stats, extents_stats and
1089 extents_stats_per_process stats logging on the fly.
1092 Frequency : Only on FC kernels 2.6.22+
1094 Description: oops in statahead
1095 Details : Do not drop reference count for the dentry from VFS when lookup,
1096 VFS will do that by itself.
1098 Severity : enhancement
1100 Description: Generic /proc file permissions
1101 Details : Set /Proc file permissions in a more generic way to enable non-
1102 root users operate on some /proc files.
1106 Description: Hitting mdc_commit_close() ASSERTION
1107 Details : Properly handle request reference release in
1108 ll_release_openhandle().
1112 Description: quota recovery deadlock during mds failover
1113 Details : This patch includes att18982, att18236, att18237 in bz14840.
1115 1. fix osts hang when mds does failover with quotaon
1116 2. prevent watchdog storm when osts threads wait for the
1121 Frequency : only patchless client
1122 Description: add workaround for race between add/remove dentry from hash
1124 Severity : enhancement
1126 Description: Allow OST glimpses to return PW locks
1130 Description: LBUG when llog conf file is full
1131 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1135 Description: Prevent import from entering FULL state when server in recovery
1139 Description: service mount cannot take device name with ":"
1140 Details : Only when device name contains ":/" will mount treat it as
1146 Description: replace ptlrpcd with the statahead thread to interpret the async
1147 statahead RPC callback
1151 Frequency : on recovery
1152 Description: I/O failures after umount during fail back
1153 Details : if client reconnected to restarted server we need join to recovery
1154 instead of find server handler is changed and process self
1155 eviction with cancel all locks.
1157 Severity : enhancement
1159 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1161 Severity : enhancement
1163 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1165 Severity : enhancement
1167 Description: Upcall on Lustre log has been dumped
1168 Details : Allow for a user mode script to be called once a Lustre log has
1169 been dumped. It passes the filename of the dumped log to the
1170 script, the location of the script can be specified via
1171 /proc/sys/lnet/debug_log_upcall.
1176 Description: avoid idr_remove called for id which is not allocated.
1177 Details : Move assigment s_dev for clustered nfs to end of initialization,
1178 to avoid problem with error handling.
1183 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1184 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1185 into CONN_USED_HASH and this prodice warning when put connection
1186 again in unused hash.
1189 Severity : enhancement
1191 Description: File striping can now be set to use an arbitrary pool of OSTs.
1193 Severity : enhancement
1195 Description: Export bytes_read/bytes_write count on OSC/OST.
1199 Description: Early reply size mismatch, MGC loses connection
1200 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1201 the connect flags are properly negotiated.
1206 Description: quotas are not honored with O_DIRECT
1207 Details : all writes with the flag O_DIRECT will use grants which leads to
1208 this problem. Now using OBD_BRW_SYNC to guard this.
1212 Description: add quota statistics
1213 Details : 1. sort out quota proc entries and proc code.
1214 2. add quota statistics
1216 Severity : enhancement
1218 Description: enable quota support for HEAD.
1222 Description: Properly propagate oinfo flags from lov to osc for statfs
1223 Details : restore missing copy oi_flags to lov requests.
1225 Severity : enhancement
1227 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1228 : l_getgroups(8), lst(8), routerstat(8)
1229 Details : included man pages for llobdstat(8), llstat(8),
1230 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1232 Severity : enhancement
1234 Description: configure's --enable-quota should check the
1235 : kernel .config for CONFIG_QUOTA
1236 Details : configure is terminated if --enable-quota is passed but
1237 : no quota support is in kernel
1241 Description: Remove portals compatibility
1242 Details : Remove portals compatibility, not interoperable with releases
1247 Description: Resolve device initialization race
1248 Details : Prevent proc handler from accessing devices added to the
1249 obd_devs array but yet be intialized.
1251 Severity : enhancement
1253 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1255 Severity : enhancement
1257 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1262 Description: Using wrong pointer in osc_brw_prep_request
1263 Details : Access to array[-1] can produce panic if kernel compiled with
1264 CONFIG_PAGE_ALLOC enabled
1268 Description: Client runs out of low memory
1269 Details : Consider only lowmem when counting initial number of llap pages
1273 Description: *optional* service tags registration
1274 Details : if the "service tags" package is installed on a Lustre node
1275 When the filesystem is mounted, a local-node service tag will
1276 be created. See http://inventory.sun.com/ for more information
1277 about the Service Tags asset management system.
1281 Description: Kernel BUG tries to release flock
1282 Details : Lustre does not destroy flock lock before last reference goes
1283 away. So always drop flock locks when client is evicted and
1284 perform unlock regardless of successfulness of speaking to MDS.
1288 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1292 Description: sanity test 65a fails if stripecount of -1 is set
1293 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1298 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1299 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1302 Severity : enhancement
1304 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1308 Description: do not process already freed flock
1309 Details : flock can possibly be freed by another thread before it reaches
1310 to ldlm_flock_completion_ast.
1314 Description: LBUG during stress test
1315 Details : Need properly lock accesses the flock deadlock detection list.
1319 Description: oops in page fault handler
1320 Details : kernel page fault handler can return two special 'pages' in error
1321 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1325 Description: timeout with invalidate import.
1326 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1327 should be handled by ptlrpcd. This produce long age waiting and
1328 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1330 Severity : enhancement
1332 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1336 Description: enable MGS and MDT services start separately
1337 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1338 a co-located MGS without starting the MGS, which is a complement
1339 to 'nosvc' mount option.
1343 Frequency : after recovery
1344 Description: precreate to many object's after del orphan.
1345 Details : del orphan st in oscc last_id == next_id and this triger growing
1346 count of precreated objects. Set flag LOW to skip increase count
1347 of precreated objects.
1351 Frequency : rare, on clear nid stats
1352 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1353 Details : when clean nid stats sometimes try destroy live entry,
1354 and this produce panic in free.
1358 Description: Stack overflow during MDS log replay
1359 ease stack pressure by using a thread dealing llog_process.
1363 Description: wait until IO finished before start new when do lock cancel.
1364 Details : VM protocol want old IO finished before start new, in this case
1365 need wait until PG_writeback is cleared until check dirty flag and
1366 call writepages in lock cancel callback.
1368 Severity : enhancement
1370 Description: using special macro for print time and cleanup in includes.
1374 Description: mds_mfd_close() ASSERTION(rc == 0)
1375 Details : In mds_mfd_close(), we need protect inode's writecount change
1376 within its orphan write semaphore to prevent possible races.
1380 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1384 Frequency : rare, on shutdown ost
1385 Description: don't hit live lock with umount ost.
1386 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1387 use shrink_dcache_sb instead.
1391 Description: don't panic with use echo client
1392 Details : echo client pass NULL as client nid pointer and this produce null
1393 pointer dereference.
1397 Description: fix build on ppc32
1398 Details : compile code with -m64 flag produce wrong object file for ppc32.
1402 Description: add message levels for liblustreapi
1406 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1407 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1408 will result in an -EAGAIN and in the exit call path we call
1409 alloc_rr() which will with increasing aggressiveness attempt to
1410 aquire precreated objects on the minimum number of required OSCs.
1414 Description: Use old size assignment to avoid deadlock
1415 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1416 the scheduling while holding a spinlock. We do not need locking
1417 for size in ll_update_inode() because size is only updated from
1418 the MDS for directories or files without objects, so there is no
1419 other place to do the update, and concurrent access to such inodes
1420 are protected by the inode lock.
1424 Description: resolve "_IOWR redefined" build error on SLES10
1428 Description: dump the memory debugging after all modules are unloaded to
1429 suppress false negative in conf_sanity test 39
1431 Severity : enhancement
1433 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1438 Description: fix SLES kernel versioning
1439 Details : the kernel version for our SLES 10 kernel did not include a "-"
1440 before the "smp" at the end. while this was not a problem in
1441 general, it did mean that software trying to use the kernel
1442 version to try to detect a vendor specific kernel would fail.
1443 this was most evident by the OFED build scripts.
1447 Description: Don't update lov_desc members until making sure they are valid
1448 Details : When updating lov_desc members via proc fs, need fix their
1449 validities before doing the real update.
1453 Description: don't put request into delay list while invalidate in flight.
1454 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1455 import in flight. this produce timeout for invalidate and sometimes
1456 can cause stale data.
1460 Frequency : on ppc only
1461 Description: not convert ost objects for directory because it's not exist.
1462 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1466 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1468 Severity : enhancement
1470 Description: Update to support 2.6.22.14 vanilla kernel.
1474 Frequency : rare, on recovery
1475 Description: read procfs can produce deadlock in some situation
1476 Details : Holding lprocfs lock which send rpc can produce block for destroy
1477 obd objects and this also block reconnect with -EALREADY. This
1478 isn't fix all lprocfs bugs - but make it rare.
1480 Severity : enhancement
1482 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1485 Frequency : frequent on X2 node
1487 Description: mdc_set_open_replay_data LBUG
1488 Details : Set replay data for requests that are eligible for replay.
1492 Description: lustre_mgs: operation 101 on unconnected MGS
1493 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1494 MGC, and late on MGC cannot successfully connect to MGS and a lot
1495 of the error messages complaining that MGS is not connected.
1499 Frequency : on network error
1500 Description: panic with double free request if network error
1501 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1502 but it's true only for synchronus enqueue, for async enqueue
1503 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1506 Severity : enhancement
1508 Description: client-side metadata stat-ahead during readdir(directory readahead)
1509 Details : perform client-side metadata stat-ahead when the client detects
1510 readdir and sequential stat of dir entries therein
1513 Frequency : on start mds
1515 Description: Implement get_info(last_id) in obdfilter.
1518 Frequency : occasional
1520 Description: Correctly check stale fid, not start epoch if ost not support SOM
1521 Details : open with flag O_CREATE need set old fid in op_fid3 because
1522 op_fid2 was overwritten with new generated fid, but mds can answer
1523 with one of these two fids and both is not stale. Setattr
1524 incorrectly started an epoch and assume will be called
1525 done_writing, but without SOM done_writing ever being called.
1528 Frequency : rare, depends on device drivers and load
1530 Description: MDS or OSS nodes crash due to stack overflow
1531 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1532 In some cases, in conjunction with device drivers that use a lot
1533 of stack the MDS (or possibly OSS) service threads could overflow
1534 the stack. One change which was identified to consume additional
1535 stack has been reworked to avoid the extra stack usage.
1538 Frequency : occasional
1540 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1541 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1542 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1543 oscc_has_objects and see OSC has no objests and call
1544 oscc_internal_create to resend create request.
1546 Severity : enhancement
1548 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1550 Severity : enhancement
1552 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1555 Frequency : very rare
1557 Description: Fix replay if there is an un-replied request and open
1558 Details : In some cases, older replay request will revert the
1559 mcd->mcd_last_xid on MDS which is used to record the client's
1560 latest sent request.
1562 Severity : enhancement
1564 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1566 Severity : enhancement
1568 Description: Add rhel5 support to HEAD.
1570 Serverity : enhancement
1572 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1577 Description: Don't allow skipping OSTs if index has been specified.
1578 Details : Don't allow skipping OSTs if index has been specified, make
1579 locking in internal create lots better.
1583 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1584 Details : access to bitfield in structure is always rounded to long
1585 and this produce problem with not atomic change any bit.
1589 Description: Lustre make rpms failed.
1590 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1591 builds Lustre rpms from tarball.
1594 Frequency : rare on shutdown ost
1596 Description: If llog cancel was not send before clean_exports phase, this can
1597 produce deadlock in llog code.
1598 Details : If llog thread has last reference to obd and call class_import_put
1599 this produce deadlock because llog_cleanup_commit_master wait when
1600 last llog_commit_thread exited, but this never success because was
1601 called from llog_commit_thread.
1605 Description: allow userland application know is lost one of stripes.
1606 Details : fill lvb_blocks with error code on ost and return it to
1607 application if error flag found.
1611 Description: NULL lov_tgts causing MDS oops
1612 Details : more safe checks for NULL lov_tgts for avoid oops.
1614 Severity : enhancement
1616 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1620 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1621 Details : If an OST is down the MDS will hang indefinitely in
1622 obd_statfs_rqset() waiting for the statfs data. While for
1623 MDS QOS usage of statfs, it should not stuck in waiting.
1625 Severity : enhancement
1627 Description: remote_acl support
1628 Details : Support ACL-based permission check for remote user.
1629 Support setfacl/getfacl for remote user with the utils
1630 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1631 the system "{s,g}etfacl" utils.
1633 Severity : enhancement
1635 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1637 Severity : enhancement
1639 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1643 Description: make mgs_setparam() handle fsname containing dash
1644 Details : fsname containing a dash does not work with lctl conf_param
1646 Severity : enhancement
1648 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1650 Severity : enhancement
1652 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1656 Frequency : rare, at shutdown
1657 Description: access already free / zero obd_namespace.
1658 Details : if client_disconnect_export was called without force flag set,
1659 and exist connect request in flight, this can produce access to
1660 NULL pointer (or already free pointer) when connect_interpret
1661 store ocd flags in obd_namespace.
1665 Frequency : only at startup
1666 Description: not alloc memory with spinlock held.
1667 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1668 if any spinlock held.
1670 Severity : enhancement
1672 Description: make lustre randomly fail allocating memory
1673 Details : Make lustre randomly failed allocating memory for testing purpose.
1675 Severity : enhancement
1677 Description: lost problems with lov objid file
1678 Details : Fixes some scability and access to not inited memory problems
1679 in work with lov objdid file.
1684 Description: lfs find does not continue on file error
1685 Details : Continue other files processing when a file/dir is absent.
1689 Description: Inconsistent usage of lustre_pack_reply()
1690 Details : Standardize the usage of lustre_pack_reply() such that it
1691 always generate a CERROR on failure.
1694 Frequency : occasional
1696 Description: MDS hang or stay in waiting lock
1697 Details : If client receive lock with CBPENDING flag ldlm need send lock
1698 cancel as separate rpc, to avoid situation when cancel request
1699 can't processed due all i/o threads stay in wait lock.
1703 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1707 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1711 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1715 Description: Update to RHEL4 latest kernel.
1717 Severity : enhancement
1719 Description: Build SLES10 patchless client fails
1720 Details : The configure was broken by run ./configure with
1721 --with-linux-obj=.... argument for patchless client. When the
1722 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1723 can't search header adequately. Use absolute path such as
1724 -I($LINUX)/include instead.
1728 Description: interrupt oig_wait produce painc on resend.
1729 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1730 private set, and this produce situation when rq_ptlrpcd_data not
1731 copyed to new allocated request and triggered LBUG on assert
1732 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1733 ptlrpcd set if request is changed and can be safety changed to use
1738 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1739 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1740 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1741 these macros and replace them with their difinitions instead.
1745 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1746 Details : Update which_patch & target file for SLES10 latest kernel.
1749 Description: Remove client patches from SLES 10 kernel.
1750 Details : This causes SLES 10 clients to behave as patchless clients
1751 even on a Lustre-patched (server) kernel.
1753 Severity : enhancement
1755 Description: self-adjustable client's lru lists
1756 Details : use adaptive algorithm for managing client cached locks lru
1757 lists according to current server load, other client's work
1758 pattern, memory activities, etc. Both, server and client
1759 side namespaces provide number of proc tunables for controlling
1762 Severity : enhancement
1764 Description: light-weight GSS support
1765 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1766 and reduce performance overhead.
1768 Severity : enhancement
1770 Description: Linux keyring support
1771 Details : Support using service of Linux keyring for Lustre GSS internal
1772 context refresh/cache mechanism.
1776 Description: Fix errors in lfs documentation
1777 Details : Fixes man pages
1781 Description: don't use GFP_* in generic Lustre code.
1782 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1786 Description: obdclass is limited by single OBD_ALLOC(idarray)
1787 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1791 Description: warnings with build patchless client with vanila 2.6.19 and up
1792 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1793 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1797 Frequency : Only for SLES
1798 Description: sanity_quota fail test_1
1799 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1800 file, which makes TSTID[2] inunique.
1804 Frequency : Only for SLES9
1805 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1806 Details : Update target/ChangeLog/which_patch .
1810 Frequency : Only for test_10 in sanity-quota.sh
1811 Description: a bug in quota test code
1812 Details : checking a test flag in a wrong place
1816 Frequency : Only for test_13 in sanity-quota.sh
1817 Description: a bug in quota test script
1818 Details : shouldn't check overfull allocated quota
1822 Description: jbd statistics
1823 Details : Port older jbd statistics patch for sles10
1825 Severity : enhancement
1827 Description: Add printing inode info into message about error in writepage.
1832 Description: reply_lock_interpret crash due to race with it and lock cancel.
1833 Details : Do not replay locks that are being cancelled. Do not reference
1834 locks by their address during replay, just by their handle.
1838 Frequency : When flocks are used.
1839 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1840 Details : Flock locks might destroy just granted lock if it could be merged
1841 with another existing flock, this is done in completion handler,
1842 so teach ldlm_cli_enquque_fini that this is a valid case for
1848 Description: Oops in read and write path when failing to allocate lock.
1849 Details : Check if lock allocation failed and return error back.
1853 Description: lstripe command fails for valid OST index
1854 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1855 instead of lov->desc.ld_active_tgt_count.
1859 Description: sanity.sh failed test 103
1860 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1865 Description: df doesn't work properly if diskfs blocksize != 4K
1866 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1870 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1871 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1873 Severity : enhancement
1875 Description: Adaptive timeouts
1876 Details : RPC timeouts adapt to changing server load and network
1877 conditions to reduce resend attempts and improve recovery time.
1881 Description: llapi_file_create() does not allow some changes
1882 Details : add llapi_file_open() that allows specifying the mode and
1883 open flags, and also returns an open file handle.
1887 Description: merge and cleanup kernel patches.
1888 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1892 Description: Add journal checksum support.(Kernel part)
1893 Details : The journal checksum feature adds two new flags i.e
1894 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1895 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1896 indicates that the commit block contains the checksum for
1897 the blocks described by the descriptor blocks. Now commit
1898 record can be sent to disk without waiting for descriptor
1899 blocks to be written to disk. This behavior is controlled
1900 using JBD2_FEATURE_ASYNC_COMMIT flag.
1904 Description: OSS needs mutliple precreate threads
1905 Details : Add ability to start more than one create thread per OSS.
1909 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1910 Details : execute lfs setstripe on client
1914 Description: mds_obd_create error creating tmp object
1915 Details : When the user sets quota on root, llog will be affected and can't
1916 create files and write files.
1919 Frequency : Always on ia64 patchless client, and possibly others.
1921 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1922 Details : This allows the patchless client to be loaded on architectures
1923 without this export.
1927 Description: RedHat Update kernel for RHEL5
1928 Details : Add the kernel config file for RHEL5.
1932 Description: RedHat Update kernel for RHEL5
1933 Details : Modify the kernel config file more closer RHEL5.
1937 Description: Build failure against Centos5 (RHEL5)
1938 Details : Define PAGE_SIZE when it isn't present.
1942 Description: test_7 on sanity-quota.sh can't be used on separate machines
1943 Details : add facet to handle it
1947 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1948 Details : with NFS, the anon dentry's parent was set to itself in
1949 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1950 corresponding dentry other than use rec->ur_name.
1952 Severity : enhancement
1954 Description: lfs setstripe enhancement
1955 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1959 Description: enable data checksumming by default
1960 Details : enable checksum by default, allow --disable-checksum
1961 configure option and "-o nochecksum" mount option. Checksums
1962 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1966 Description: lustre support for RHEL5
1967 Details : Add support for RHEL5.
1971 Description: Client eviction due to failover config
1972 Details : after a connection loss, the lustre client should attempt to
1973 reconnect to the last active server first before trying the
1974 other potential connections.
1978 Description: when mds and osts use different quota unit(32bit and 64bit),
1979 quota will be released repeatly.
1980 Details : void sending multiple quota reqs to mds, which will keep the
1981 status between the reqs.
1985 Description: osts not allocated evenly to files
1986 Details : change the condition to increase offset_idx
1989 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1991 Description: Data corruption for OSTs that are formatted larger than 2TB
1993 Details : When generating the bio request for lustre file writes the
1994 sector number would overflow a temporary variable before being
1995 used for the IO. The data reads correctly from Lustre (which
1996 will overflow in a similar manner) but other file data or
1997 filesystem metadata may be corrupted in some cases.
2001 Description: Tune the kernel for good SCSI performance.
2002 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2003 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2008 Description: rewrite ext2-derived code in obdclass/uuid.c
2009 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2010 from scratch preserving functionality.
2014 Description: "lfs find -obd UUID" prints directories
2015 Details : "lfs find -obd UUID" will return all directory names instead
2016 of just file names. It is incorrect because the directories
2017 do not reside on the OSTs.
2021 Description: use i_size_read and i_size_write in 2.6 port
2022 Details : replace inode->i_size access with i_size_read/write()
2025 Frequency : when removing large files
2027 Description: scheduling issue during removal of large Lustre files
2028 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2029 It causes scheduling issues when removing large files (17TB in the
2033 Frequency : only with liblustre clients on XT3
2035 Description: evictions taking too long
2036 Details : allow llrd to evict clients directly on OSTs
2039 Frequency : only on ppc
2041 Description: /proc/fs/lustre/devices broken on ppc
2042 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2043 We should make sure the type of 'index' is loff_t and then cast
2044 explicitly as needed below. Do not assign an explicitly cast
2048 Frequency : only for rhel5
2050 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2051 Details : Modify the target file & which_kernel.
2053 Severity : enhancement
2055 Description: omit set fsid for export NFS
2056 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2057 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2061 Frequency : Always, for kernels after 2.6.16
2062 Description: Fix warning idr_remove called for id=.. which is not allocated.
2063 Details : Last kernels save old s_dev before kill super and not allow
2064 to restore from callback - restore it before call kill_anon_super.
2068 Description: buffer overruns could theoretically occur
2069 Details : llapi_semantic_traverse() modifies the "path" argument by
2070 appending values to the end of the origin string, and a
2071 overrun may occur. Adding buffer overrun check in liblustreapi.
2075 Description: Fix error on 'ls .' at the top of the Lustre mount.
2076 Details : Don't revalidate dentry if it is a root dentry.
2080 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2081 Details : Modify vm-tunables-rhel4.patch.
2085 Description: Kernel config for 2.6.18-vanilla.
2086 Details : Modify targets/2.6-vanilla.target.in.
2087 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2088 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2089 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2090 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2094 Description: improve handling recoverable errors
2095 Details : if request processig with error which can be recoverable on server
2096 request should be resend, otherwise page released from cache and
2101 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2102 Details : Modify target file & which_patch.
2103 A flaw was found in the IA32 system call emulation provided
2104 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2105 value could be stored in the %RAX register, which could trigger an
2106 out-of-bounds system call table access. An untrusted local user
2107 could exploit this flaw to run code in the kernel
2108 (ie a root privilege escalation). (CVE-2007-4573).
2112 Description: O_DIRECT bypasses client statistics.
2113 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2114 read_ahead_stats were updated. copied the stats section from
2115 osc_send_oap_rpc() into async_internal().
2119 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2123 Description: change order of libsysio includes
2124 Details : '#include sysio.h' should always come before '#include xtio.h'
2128 Description: handle "serious error: objid * already exists" more gracefully
2129 Details : If LAST_ID value on disk is smaller than the objects existing in
2130 the O/0/d* directories, it indicates disk corruption and causes an
2131 LBUG(). If the object is 0-length, then we should use the existing
2132 object. This will help to avoid a full fsck in most cases.
2134 Severity : enhancement
2136 Description: adapt the lustre_config script to support the upgrade case
2137 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2138 server targets to 1.6 in parallel.
2142 Description: To avoid grant space > avaible space when the disk is almost
2143 full. Without this patch you might see the error "grant XXXX >
2144 available" or some LBUG about grant, when the disk is almost
2146 Details : In filter_check_grant, for non_grant cache write, we should
2147 check the left space by if (*left > ungranted + bytes), instead
2148 of (*left > ungranted), because only we are sure the left space
2149 is enough for another "bytes", then the ungrant space should be
2150 increase. In client, we should update cl_avail_grant only there
2151 is OBD_MD_FLGRANT in the reply.
2155 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2156 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2157 described in CVE-2007-4573. This problem could allow untrusted
2158 local users to gain root access.
2161 Frequency : when using O_DIRECT and quotas
2163 Description: Incorrect file ownership on O_DIRECT output files
2164 Details : block usage reported by 'lfs quota' does not take into account
2165 files that have been written with O_DIRECT.
2170 Description: touch file failed when fs is not full
2171 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2172 otherwise we can get ENOSP while fs is not full.
2176 Description: parallel lock callbacks
2177 Details : Instead of sending blocking and completion callbacks as separated
2178 requests, adding them to a set and sending in parallel.
2181 Frequency : only for Cray XT3
2182 Bugzilla : 12829/13455
2183 Description: Changing primary group doesn't change the group lustre assigns to
2185 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2186 overridden with the primary group provided by the group upcall,
2187 whereas the supplied fsgid can be trusted if it is in the list of
2188 supplementary groups returned by the group upcall.
2190 Severity : enhancement
2192 Description: Allow masking D_WARNING, D_ERROR messages from console
2193 Details : Console messages can now be disabled via lnet.printk.
2197 Description: User code with malformed file open parameter crashes client node
2198 Details : Before packing join_file req, all the related reference should be
2199 checked carefully in case some malformed flags cause fake
2200 join_file req on client.
2204 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2205 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2206 mode is equal to the granted mode. However, it should also check
2207 that LDLM_FL_CP_REQD is not set.
2211 Description: Heavy nfs access might result in deadlocks
2212 Details : After ELC code landed, it is now improper to enqueue any mds
2213 locks under och_sem, because enqueue might want to decide to
2214 cancel open locks for same inode we are holding och_sem for.
2218 Description: Client eviction while running blogbench
2219 Details : A lot of unlink operations with concurrent I/O can lead to a
2220 deadlock causing evictions. To address the problem, the number of
2221 oustanding OST_DESTROY requests is now throttled to
2222 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2223 callbacks are processed in priority.
2227 Description: enable ACLs on MDS by default
2228 Details : ACLs must be enabled on MDS by default.
2231 Frequency : PPC/PPC64 only
2233 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2234 Details : fix duplicated definitions between asm-ppc64/types.h and
2235 lustre_types.h on PPC.
2238 Frequency : PPC/PPC64 only
2240 Description: asm-ppc/segment.h does not exist
2241 Details : fix compile issue on PPC.
2245 Description: better handle error messages in extents code
2248 Frequency : RHEL4 only
2250 Description: mkfs is very slow on IA64/RHEL4
2251 Details : A performance regression has been discovered in the MPT Fusion
2252 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2253 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2254 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2256 Severity : enhancement
2258 Description: SNMP support enhancement
2259 Details : Adding total number of sampled request for an MDS node in snmp
2262 Severity : enhancement
2264 Description: Optimize ldlm waiting list processing for PR extent locks
2265 Details : When processing waiting list for read extent lock and meeting read
2266 lock that is same or wider to it that is not contended, skip
2267 processing rest of the list and immediatelly return current
2268 status of conflictness, since we are guaranteed there are no
2269 conflicting locks in the rest of the list.
2273 Description: Time out and refuse to reconnect
2274 Details : When the failover node is the primary node, it is possible
2275 to have two identical connections in imp_conn_list. We must
2276 compare not conn's pointers but NIDs, otherwise we can defeat
2277 connection throttling.
2281 Description: port llog fixes from b1_6 into HEAD
2282 Details : Port llog reference couting and some llog cleanups from b1_6
2283 (bug 10800) into HEAD, for protect from panic and access to already
2284 free llog structures.
2288 Description: Detect stride IO mode in read-ahead
2289 Details : When a client does stride read, read-ahead should detect that and
2290 read-ahead pages according to the detected stride pattern.
2294 Description: data checksumming impacts single node performance
2295 Details : add support for several checksum algorithm. Currently, only CRC32
2296 and Adler-32 are supported. The checksum type can be changed on
2297 the fly via /proc/fs/lustre/osc/*/checksum_type.
2301 Description: use adler32 for page checksums
2302 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2307 Description: build for x2 fails
2308 Details : fix compile issue on Cray systems.
2312 Description: Properly match for duplicate locks
2313 Details : Due to different lock order from skiplists code, we need to
2314 traverse entire list for now
2317 Frequency : only on PPC/SLES10
2319 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2320 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2321 be defined. Add a hack in mkfs_lustre.c to work around this compile
2326 Description: LASSERT on MDS when client holding flock lock dies
2327 Details : ldlm pool logic depends on number of granted locks equal to
2328 number of released locks which is not true for flock locks, so
2329 just exclude such locks from consideration.
2333 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2334 Details : Use fsfilt_sync() for both the whole filesystem sync and
2335 individual file sync to eliminate dangerous inode locking
2336 with I_LOCK that can lead to a deadlock.
2340 Description: performance in 1.6.3
2341 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2342 sure that 1MB requests are not fragmented by the block layer.
2344 Severity : enhancement
2346 Description: organize the server-side client stats on per-nid basis
2347 Details : Change the structure of stats under obdfilter and mds to
2355 The "uuid"s file would list the uuids of _active_ exports.
2356 And the clear entry is to clear all stats and stale nids.
2358 Severity : enhancement
2360 Description: eliminate client locks in face of contention
2361 Details : file contention detection and lockless i/o implementation
2362 for contended files.
2366 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2369 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2371 Description: MDS may lose file striping (and hence file data) in some cases
2372 Details : If there are additional extended attributes stored on the MDS,
2373 in particular ACLs, SELinux, or user attributes (if user_xattr
2374 is specified for the client mount options) then there is a risk
2375 of attribute loss. Additionally, the Lustre file striping
2376 needs to be larger than default (e.g. striped over all OSTs),
2377 and an additional attribute must be stored initially in the
2378 inode and then increase in size enough to be moved to the
2379 external attribute block (e.g. ACL growing in size) for file
2384 Description: skiplist implementation simplification
2385 Details : skiplists are used to group compatible locks on granted list
2386 that was implemented as tracking first and last lock of each
2387 lock group the patch changes that to using doubly linked lists
2391 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2392 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2393 because LASSERT on both the data supplied by a client, and the
2394 data on disk is dangerous and incorrect.
2396 Severity : enhancement
2398 Description: Slow truncate/writes to huge files at high offsets.
2399 Details : Directly associate cached pages to lock that protect those pages,
2400 this allows us to quickly find what pages to write and remove
2401 once lock callback is received.
2405 Description: more ldlm soft lockups
2406 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2407 starve other threads from the resource lock for a long time in
2408 case of long waiting queue, so change the debug level from
2409 D_OTHER to the less frequently used D_INFO.
2411 Severity : enhancement
2413 Description: add -gid, -group, -uid, -user options to lfs find
2417 Description: Hung threads in invalidate_inode_pages2_range
2418 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2419 once one is completed. As a result, some RPCs are stuck in the
2420 queue and are never sent.
2424 Description: filter threads hungs on waiting journal commit
2425 Details : Cleanup filter group llog code, then only filter group llog will
2426 be only created in the MDS/OST syncing process.
2430 Description: Procfs and llog threads access destoryed import sometimes.
2431 Details : Sync the import destoryed process with procfs and llog threads by
2432 the import refcount and semaphore.
2434 Severity : enhancement
2436 Description: openlock cache of b1_6 port to HEAD
2441 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2442 Details : If insertion of an extent fails, then discard the inode
2443 preallocation and free data blocks else it can lead to duplicate
2448 Description: don't always update ctime in ext3_xattr_set_handle()
2449 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2450 In some cases the ctime should not be updated, for example for
2451 2.0->1.8 compatibility it is necessary to delete an xattr and it
2452 should not update the ctime.
2456 Bugzilla : 15713/16362
2457 Description: Assertion in iopen_connect_dentry in 1.6.3
2458 Details : looking up an inode via iopen with the wrong generation number can
2459 populate the dcache with a disconneced dentry while the inode
2460 number is in the process of being reallocated. This causes an
2461 assertion failure in iopen since the inode's dentry list contains
2462 both a connected and disconnected dentry.
2466 Description: assertion failure in ldlm_handle2lock()
2467 Details : fix a race between class_handle_unhash() and class_handle2object()
2468 introduced in lustre 1.6.5 by bug 13622.
2473 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2474 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2479 Description: oops in sd_iostats_seq_show()
2480 Details : unloading/reloading the scsi low level driver triggers a kernel
2481 bug when trying to access the sd iostat file.
2486 Description: Kernel panics during QLogic driver reload
2487 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2488 patch, causing memory corruption.
2493 Description: journal_dev option does not work in b1_6
2494 Details : pass mount option during pre-mount.
2496 Severity : enhancement
2498 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2499 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2500 extent information of a file. It can be used to map logical blocks
2501 in a file to physical blocks in the block device.
2505 Description: LDLM soft lockups - improvement
2506 Details : It is be possible to send the lock handle along with each read
2507 or write request because the client is already doing a lock match
2508 itself so there isn't any reason the OST should have to re-do that
2512 Frequency : only on Cray X2
2514 Description: X2 build failures
2515 Details : fix build failures on Cray X2.
2519 Description: xid & resent requests
2520 Details : Initialize RPC XID from clock at startup (randomly if clock is
2523 Severity : enhancement
2525 Description: Add lustre_start utility to start or stop multiple Lustre servers
2530 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2531 Details : In case of memory pressure, list_del() can be called twice on
2532 req->rq_history_list, causing a kernel oops.
2536 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2537 Details : fix stack overflow in the distributed lock manager by defering
2538 export eviction after a failed AST to the elt thread instead of
2539 handling it in the dlm interpret routine.
2543 Description: Convert some comments to new format.
2544 Details : Update documenting comments to match doxygen conventions.
2548 Description: Grammar fixes.
2549 Details : A couple of trivial spelling fixes.
2553 Description: OSD_COUNTERS-mandatory
2554 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2555 to simplify the code.
2559 Description: simplify cmm_device freeing logic.
2560 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2564 Description: Add lockdep support to dt_object_operations locking interface.
2565 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2566 parameter indicating lock ordering. Update mdd code to use new
2571 Description: Introduce failloc constants for lockless IO tests.
2572 Details : Add two new failloc constants to test lockless IO. Only one of
2573 them in implemented---another is checked in yet to be landed
2578 Description: Add lockdep support for inode mutex.
2579 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2580 in the situations where Lustre has to lock more than one inode
2585 Description: Add optional invariants checking support.
2586 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2587 configure switch. This macro is to be used for consistency and
2588 sanity checks that are too expensive to be left in `production'
2593 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2594 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2599 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2600 Details : Use spin_lock_nested() in (the only) situation where more than
2601 one ldlm_lock is locked simultaneously. Also, fix possible
2602 dead-lock in ldlm_lock_change_resource() by enforcing particular
2607 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2608 Details : Instead of specifying each ldlm_lock call-back through separate
2609 parameter, wrap them into struct ldlm_callback_suite.
2613 Description: Kill join_lru obd method and its callers.
2614 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2615 in memory---a requirement implemented through ->o_join_lru() obd
2616 method in HEAD. Remove this method and its users.
2620 Description: Add asynchronous ldlm ENQUEUE completion handler.
2621 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2622 case that stock ldlm_completion_ast() cannot handle as it waits
2623 until lock is granted. Introduce new ldlm_completion_ast_async()
2624 for this. Also comment ldlm_completion_ast().
2628 Description: ldlm_error <-> errno conversion.
2629 Details : Add functions to map (rather arbitrary) between LDLM error codes
2630 and standard errno values. CLIO needs this to prevent LDLM specific
2631 constants from escaping ldlm and osc.
2635 Description: Kill unused ldlm_handle2lock_ns() function.
2636 Details : Kill unused ldlm_handle2lock_ns() function.
2640 Description: Add lu_ref support to ldlm_lock
2641 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2642 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2643 and ldlm_resource. LDLM interface has to be changed, because code that
2644 releases a reference on a lock, has to "know" what reference this is.
2645 In the most frequent case
2647 lock = ldlm_handle2lock(handle);
2649 LDLM_LOCK_PUT(lock);
2651 no changes are required. When any other reference (received _not_ from
2652 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2653 instead of LDLM_LOCK_PUT().
2655 Arguably, changes are pervasive, and interface requires some discipline
2656 for proper use. On the other hand, it was very instrumental in finding
2657 a few leaked lock references.
2661 Description: Add ldlm_lock_addref_try().
2662 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2663 attempts to addref a lock that might be being canceled
2668 Description: Add ldlm_weigh_callback().
2669 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2670 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2671 hard-coded `number of pages' logic.
2675 Description: Add lockdep annotations to llog code.
2676 Details : Use appropriately tagged _nested() locking calls in the places
2677 where llog takes more than one ->lgh_lock lock.
2681 Description: Add loi_kms_set().
2682 Details : Wrap kms updates into a helper function.
2686 Description: Constify instances of struct lsm_operations.
2687 Details : Constify instances of struct lsm_operations.
2691 Description: lu_conf support.
2692 Details : On a server, a file system object is uniquely identified
2693 by a fid, which is sufficient to locate and load all object
2694 state (inode). On a client, on the other hand, more data are
2695 necessary instantiate an object. Change lu_object_find() and
2696 friends to take additional `lu_conf' argument describing object.
2697 Typically this includes layout information.
2701 Description: lu_context fixes.
2702 Details : Introduce new lu_context functions that are needed on the client
2703 side, where some system threads (ptlrpcd) are shared by multiple
2704 modules, and so cannot be stopped during module shutdown.
2708 Description: Add start and stop methods to lu_device_type_operations.
2709 Details : Introduce two new methods in lu_device_type_operations, that are
2710 invoked when first instance of a given type is created and last one
2711 is destroyed respectively. This is need by CLIO.
2715 Description: Add lu_ref support to struct lu_device.
2716 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2717 track leaked references.
2721 Description: Introduce lu_kmem_descr.
2722 Details : lu_kmem_descr and its companion interface allow to create
2723 and destroy a number of kmem caches at once.
2727 Description: Fix lu_object finalization race.
2728 Details : Fix a race between lu_object_find() finding an object and its
2729 concurrent finalization. This race is (most likely) not possible
2730 on the server, but might happen on the client.
2734 Description: Introduce lu_ref interface.
2735 Details : lu_ref is a debugging module allowing to track references to
2736 a given object. It is quite cpu expensive, and has to be
2737 explicitly enabled with --enable-lu_ref. See usage description
2742 Description: Factor lu_site procfs stats into a separate function.
2743 Details : Separate lu_site stats printing code into a separate function
2744 that can be reused on a client.
2748 Description: Constify instances of struct {lu,dt,md}_device_operations.
2749 Details : Constify instances of struct {lu,dt,md}_device_operations.
2753 Description: Introduce struct md_site and move meta-data specific parts of
2754 struct lu_site here.
2755 Details : Move md-specific fields out of struct lu_site into special struct
2756 md_site, so that lu_site can be used on a client.
2760 Description: Kill mdd_lov_destroy().
2761 Details : Remove unused mdd code.
2765 Description: Add st_block checking to multistat.c.
2766 Details : Add st_block checking to multistat.c.
2770 Description: Add lu_ref support to struct obd_device.
2771 Details : Add lu_ref tracking to obd_device.
2775 Description: Kill obd_set_fail_loc().
2776 Details : Remove unused code.
2780 Description: Add special type for ptlrpc_request interpret functions.
2781 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2782 there. Actual usage will be in CLIO.
2786 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2787 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2788 doesn't work with lockdep.
2792 Description: Add rwv.c test program.
2793 Details : New testing program exercising readv(2) and writev(2) (Qian).
2797 Description: Add sendfile.c test program.
2798 Details : New testing program exercising sendfile(2) (Jay).
2802 Description: Ratelimit a message that can be very frequent.
2803 Details : Ratelimit a memory allocation failure message that can
2808 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2809 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2813 Description: Liblustre build fixes.
2814 Details : Liblustre build fixes.
2818 Description: libcfs: add cfs_{need,cond}_resched() interface.
2819 Details : libcfs: add cfs_{need,cond}_resched() definition and
2820 implementations for Linux, NT, and liblustre.
2822 Severity : enhancement
2824 Description: More exported tunables for mballoc
2825 Details : Add support for tunable preallocation window and new tunables for
2826 large/small requests
2830 Description: Detect corruption of block bitmap and checking for preallocations
2831 Details : Checks validity of on-disk block bitmap. Also it does better
2832 checking of number of applied preallocations. When corruption is
2833 found, it turns filesystem readonly to prevent further corruptions.
2837 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2838 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2839 window to 0. If it does hit cache miss, and read-pattern is still
2840 stride-io mode, does not reset the stride window, but also does not
2841 increase the stride window length in this case.
2845 Frequency : only for big-endian servers
2846 Description: Check if system is big-endian while mounting fs with extents feature
2847 Details : Mounting a filesystem with extents feature will fail on big-endian
2848 systems since ext3-based ldiskfs is not supported on big-endian
2849 systems. This can be over-riden with "bigendian_extents" mount option.
2851 Severity : enhancement
2853 Description: The root squash functionality
2854 Details : A security feature, which is to prevent users from being able
2855 to mount lustre on their desktop, run as root, and delete
2856 all of the files in the filesystem. The goal is accomplished by
2857 remapping user id (UID) and group id (GID) of the root user to
2858 a UID and GID specified by the system administartor via Lustre
2859 configuration management server (MGS). The functionality also
2860 allows to specify sets of clients for which the remapping does
2865 Description: Excessive recovery window
2866 Details : With AT enabled, the recovery window can be excessively long (6000+
2867 seconds). To address this problem, we no longer use
2868 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2869 timeout no longer depends on the service time, it is set to
2870 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2871 time via pb_service_time.
2875 Description: Watchdog triggered on MDS failover
2876 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2877 the OSTs know that the MDS "UUID" can be reused for the same export
2878 from a different NID, so we do not need to wait for the export to be
2882 Frequency : rare, only if using MMP with Linux RAID
2884 Description: MMP doesn't work with Linux RAID
2885 Details : While using HA for Lustre servers with Linux RAID, it is possible
2886 that MMP will not detect multiple mounts. To make this work we
2887 need to unplug the device queue in RAID when the MMP block is being
2888 written. Also while reading the MMP block, we should read it from
2889 disk and not the cached one.
2891 Severity : enhancement
2893 Description: open file using fid
2894 Details : A file can be opened using just its fid, like
2895 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2898 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2901 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2902 Details : add proper locking for creation of jbd_4k slab cache
2906 Description: MMP check in ext3_remount() fails without displaying any error
2907 Details : When multiple mount protection fails during remount, proper error
2910 Severity : enhancement
2912 Description: Allow stripe size to be up to 4G-64k
2913 Details : Fix math logic to allow large stripe sizes.
2917 Description: add check for >8TB ldiskfs filesystems
2918 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2919 Don't allow >8TB ldiskfs filesystems to be mounted without
2920 force_over_8tb mount option
2925 Descriptoin: OSS complains "too many blocks" and crashes
2926 Details : In filter_commitrw_write():
2927 1. lquota_chkquota();
2928 2. obdo_from_inode();
2929 3. lquota_pending_commit();
2930 "2." has changed uid and gid in this case, then caused
2931 lquota_chkquota() and lquota_pending_commit() to use
2932 different uid and gid. That is the root of the bug.
2936 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2937 Details : Detect and print out the 'stripe count = -1' case properly.
2941 Description: Out or order replies might be lost on replay
2942 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2943 request with tid smaller than one currently being added, add it
2944 to the start, not end of the list.
2946 --------------------------------------------------------------------------------
2948 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2950 * Support for kernels:
2951 2.6.5-7.283 (SLES 9),
2952 2.6.9-55.EL (RHEL 4),
2953 2.6.16.46-0.14 (SLES 10),
2954 2.6.18.8 vanilla (kernel.org)
2955 * Client support for unpatched kernels:
2956 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2957 2.6.16 - 2.6.22 vanilla (kernel.org)
2958 * Due to recently discovered recovery problems, we do not recommend
2959 using patchless RHEL 4 clients with this or any earlier release.
2960 * Recommended e2fsprogs version: 1.39.cfs8
2961 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2962 * Starting with this release, the ldiskfs backing filesystem required
2963 by Lustre is now in its own package, lustre-ldiskfs. This package
2964 should be installed. It is versioned separately from Lustre and
2965 may be released separately in future.
2969 Description: block reactivating mgc import until all deactivates complete
2970 Details : Fix race when failing back MDT/MGS to itself (testing)
2973 Frequency : at statup only
2975 Description: mds_lov_synchronize race leads to various problems
2976 Details : simultaneous MDT->OST connections at startup can cause the
2977 sync to abort, leaving the OSC in a bad state.
2979 Severity : enhancement
2981 Description: add optional extra BUILD_VERSION info
2982 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2983 to override the lustre version.
2986 Frequency : 2.6.18 servers only
2988 Description: ll_kern_mount() doesn't release the module reference
2989 Details : The ldiskfs module reference count never drops down to 0
2990 because ll_kern_mount() doesn't release the module reference.
2995 Description: server LBUG when using old ost_num_threads parameter
2996 Details : Accept the old ost_num_threads parameter but warn that it
2997 is deprecated, and fix an off-by-one error that caused an LBUG.
3002 Description: Transient SCSI error results in persistent IO issue
3003 Details : iobuf->dr_error is not reinitialized to 0 between two
3007 Frequency : sometimes when underlying device returns I/O errors
3009 Description: OSTs not going read-only during write failures
3010 Details : OSTs are not remounted read-only when the journal commit threads
3011 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3012 instead of the ext3 wrappers.
3016 Description: poor connect scaling with increasing client count
3017 Details : Don't run filter_grant_sanity_check for more than 100 exports
3018 to improve scaling for large numbers of clients.
3021 Frequency : SLES10 only
3023 Description: sanity-quota.sh quotacheck failed: rc = -22
3024 Details : Quotas cannot be enabled on SLES10.
3027 Frequency : liblustre clients only
3029 Description: getdirentries does not give error when run on compute nodes
3030 Details : getdirentries does not fail when the size specified as an argument
3031 is too small to contain at least one entry
3033 Severity : enhancement
3035 Description: Add LNET router traceability for debug purposes
3036 Details : If a checksum failure occurs with a router as part of the
3037 IO path, the NID of the last router that forwarded the bulk data
3038 is printed so it can be identified.
3043 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3044 Details : Due to a race condition, liblustre clients were occasionally
3045 evicted incorrectly.
3047 Severity : enhancement
3049 Description: lfs setstripe use optional parameters instead of postional
3052 Severity : enhancement
3054 Description: Nanosecond timestamp support for ldiskfs
3055 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3056 resolution timestamps. There is not yet support for this at
3057 the Lustre filesystem level.
3060 Frequency : during server recovery
3062 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3063 Details : request with rq_no_resend flag not awake l_wait_event if they get
3067 Frequency : nfs export on patchless client
3069 Description: connectathon hang when test nfs export over patchless client
3070 Details : Disconnected dentry cannot be found with lookup, so we do not need
3071 to unhash it or make it invalid
3074 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3075 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3076 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3081 Description: open req refcounting wrong on reconnect
3082 Details : If reconnect happened between getting open reply from server and
3083 call to mdc_set_replay_data in ll_file_open, we will schedule
3084 replay for unreferenced request that we are about to free.
3085 Subsequent close will crash in variety of ways.
3086 Check that request is still eligible for replay in
3087 mdc_set_replay_data().
3092 Description: disable writes to filesystem when reading health_check file
3093 Details : the default for reading the health_check proc file has changed
3094 to NOT do a journal transaction and write to disk, because this
3095 can cause reads of the /proc file to hang and block HA state
3096 checking on a healthy but otherwise heavily loaded system. It
3097 is possible to return to the previous behaviour during configure
3098 with --enable-health-write.
3100 Severity : enhancement
3102 Description: 64-bit inode version
3103 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3104 to the inode. This will be required for version-based recovery.
3109 Description: MDS fails to start if a duplicate client export is detected
3110 Details : in some rare cases it was possible for a client to connect to
3111 an MDS multiple times. Upon recovery the MDS would detect this
3112 and fail during startup. Handle this more gracefully.
3114 Severity : enhancement
3116 Description: Add -o localflock option to simulate old noflock
3118 Details : This will achieve local-only flock/fcntl locks
3124 Description: log_commit_thread vs filter_destroy race leads to crash
3125 Details : Take import reference before releasing llog record semaphore
3130 Description: Wrong request locking in request set processing
3131 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3132 imp_delayed_list, in this place should be used imp_lock.
3135 Frequency : when reconnection
3137 Description: Grant Leak when osc reconnect to OST
3138 Details : When osc reconnect ost, OST(filter) should check whether it
3139 should grant more space to client by comparing fed_grant and
3140 cl_avail_grant, and return the granted space to client instead
3141 of "new granted" space, because client will call osc_init_grant
3142 to update the client grant space info.
3145 Frequency : when client reconnect to OST
3147 Description: Grant Leak when osc do resend and replay bulk write
3148 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3149 bulk write request, because the grant info while be sync between
3150 OSC and OST when reconnect, and we should ignore the grant info
3151 these of resend/replay write req.
3156 Description: Grant space more than available left space sometimes.
3157 Details : When then OST is about to be full, if two bulk writing from
3158 different clients came to OST. Accord the avaliable space of the
3159 OST, the first req should be permitted, and the second one
3160 should be denied by ENOSPC. But if the seconde arrived before
3161 the first one is committed. The OST might wrongly permit second
3162 writing, which will cause grant space > available space.
3165 Frequency : when client is evicted
3167 Description: Grant might be wrongly erased when osc is evicted by OST
3168 Details : when the import is evicted by server, it will fork another
3169 thread ptlrpc_invalidate_import_thread to invalidate the
3170 import, where the grant will be set to 0. While the original
3171 thread will update the grant it got when connecting. So if
3172 the former happened latter, the grant will be wrongly errased
3173 because of this race.
3178 Description: Checking Stale with correct fid
3179 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3180 to check whether it is stale, because sometimes, we want the
3181 enqueue happened anyway, and op_data.fid2 will not be initialized.
3183 Severity : enhancement
3185 Description: update patchless client
3186 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3189 Frequency : only with 2.4 kernel
3191 Description: random memory corruption
3192 Details : size of struct ll_inode_info is to big for union inode.u and this
3193 can be cause of random memory corruption.
3198 Description: Memory leak in recovery
3199 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3200 It should also check obd_fail before fsfilt_start, otherwise if
3201 fsfilt_start return -EROFS,(failover mds during mds recovery).
3202 then the req will return with repmsg->transno = 0 and rc = EROFS.
3203 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3204 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3205 be freed no matter whether fsfilt_commit success or not.
3208 Frequency : only with huge count clients
3210 Description: Prevents from taking the superblock lock in llap_from_page for
3212 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3213 for llap_from_page call in ll_removepage prevents from taking the
3214 superblock lock for a soon died page.
3219 Description: Not check open intent error before release open handle
3220 Details : in some rare cases, the open intent error is not checked before
3221 release open handle, which may cause
3222 ASSERTION(open_req->rq_transno != 0), because it tries to release
3223 the failed open handle.
3228 Description: Set cat log bitmap only after create log success.
3229 Details : in some rare cases, the cat log bitmap is set too early. and it
3230 should be set only after create log success.
3234 Description: Accessing a block bevice can re-enable I/O when Lustre is
3235 tearing down a device.
3236 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3241 Description: service threads may hog cpus when there are a lot of requests
3242 Details : Insert cond_resched to give other threads a chance to use some CPU
3247 Description: the cat log was not initialized in recovery
3248 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3249 unlink log on mds will not be initialized until mds post
3250 recovery. And also in mds post recovery, the unlink log will
3251 initialization will be done asynchronausly, so there will be race
3252 between add unlink log and unlink log initialization.
3256 Description: brw_stats were being printed incorrectly
3257 Details : brw_stats were being printed as log2 but all of them were not
3258 recorded as log2. Also remove some code duplication arising from
3259 filter_tally_{read,write}.
3263 Frequency : rare, only in recovery.
3264 Description: ASSERTION(req->rq_type != LI_POISON) failed
3265 Details : imp_lock should be held while iterating over imp_sending_list for
3266 prevent destroy request after get timeout in ptlrpc_queue_wait.
3270 Description: replay-single.sh test 52 fails
3271 Details : A lock's skiplist need to be cleanup when it being unlinked
3272 from its resource list.
3276 Description: Short directio read returns full requested size rather than
3278 Details : Direct I/O operations should return actual amount of bytes
3279 transferred rather than requested size.
3281 Severity : enhancement
3283 Description: metadata RPC reduction (e.g. for rm performance)
3284 Details : decrease the amount of synchronous RPC between clients and servers
3285 by canceling conflicing lock before the operation on the client
3286 and packing thier handles into the main operation RPC to server.
3288 Severity : enhancement
3290 Description: Async OSC create to avoid the blocking unnecessarily.
3291 Details : If an OST has no remaining object, system will block on the
3292 creation when it needs to create a new object on this OST. Now,
3293 ways use pre-created objects when available, instead of blocking on
3294 an empty osc while others are not empty. If we must block, we
3295 block for the shortest possible period of time.
3299 Description: improve handling recoverable errors
3300 Details : if request processig with error which can be recoverable on server
3301 request should be resend, otherwise page released from cache and
3304 Severity : enhancement
3306 Description: refine locking for avoid write wrong info into lov_objid file
3307 Details : fix possible races with add new target and write/update data in
3310 --------------------------------------------------------------------------------
3312 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3317 Frequency : on some architectures
3319 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3320 Details : Uninitialized flags sometimes cause configuration commands to
3324 Frequency : patchless clients only
3326 Description: missing __iget() symbol export
3327 Details : The __iget() symbol export is missing. To avoid the need for
3328 this on patchless clients the deathrow inode reaper is turned
3329 off, and we depend on the VM to clean up old inodes. This
3330 dependency was during via the fix for bug 12181.
3332 --------------------------------------------------------------------------------
3334 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3336 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3337 INTEROPERATE with older versions automatically. In many cases a
3338 special upgrade step is needed. Please read the
3339 user documentation before upgrading any part of a live system.
3340 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3341 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3342 * WARNING: Lustre configuration and startup changes are required with
3343 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3347 Severity : enhancement
3349 Description: MountConf
3350 Details : Lustre configuration is now managed via mkfs and mount
3351 commands instead of lmc and lconf. New obd types (MGS, MGC)
3352 are added for dynamic configuration management. See
3353 https://mail.clusterfs.com/wikis/lustre/MountConf for
3356 Severity : enhancement
3358 Description: dynamic OST addition
3359 Details : OSTs can now be added to a live filesystem
3361 Severity : enhancement
3363 Description: startup order invariance
3364 Details : MDTs and OSTs can be started in any order. Clients only
3365 require the MDT to complete startup.
3367 Severity : enhancement
3369 Description: parallel, asynchronous orphan cleanup
3370 Details : orphan cleanup is now performed in separate threads for each
3371 OST, allowing parallel non-blocking operation.
3373 Severity : enhancement
3375 Description: optimized stripe assignment
3376 Details : stripe assignments are now made based on ost space available,
3377 ost previous usage, and OSS previous usage, in order to try
3378 to optimize storage space and networking resources.
3380 Severity : enhancement
3382 Description: Permanently set tunables
3383 Details : All writable /proc/fs/lustre tunables can now be permanently
3384 set on a per-server basis, at mkfs time or on a live system.
3386 Severity : enhancement
3388 Description: Lustre message v2
3389 Details : Add lustre message format v2.
3391 Severity : enhancement
3393 Description: client OST exclusion list
3394 Details : Clients can be started with a list of OSTs that should be
3395 declared "inactive" for known non-responsive OSTs.
3399 Description: SPEC SFS validation failure on NFS v2 over lustre.
3400 Details : Changes the blocksize for regular files to be 2x RPC size,
3401 and not depend on stripe size.
3403 Severity : enhancement
3405 Description: Multiple MD RPCs in flight.
3406 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3407 To avoid overly-overloading MDS, introduce a limit on number of
3408 MDS RPCs in flight for a single client and add /proc controls
3409 to adjust this limit.
3411 Severity : enhancement
3413 Description: client read/write statistics
3414 Details : Add client read/write call usage stats for performance
3415 analysis of user processes.
3416 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3417 file access. extents_stats shows chunk size distribution.
3418 extents_stats_per_process show chunk size distribution per
3421 Severity : enhancement
3423 Description: mds statistics
3424 Details : Add detailed mds operations statistics in
3425 /proc/fs/lustre/mds/*/stats.
3429 Description: Failure of copying files with lustre special EAs.
3430 Details : Client side always return success for setxattr call for lustre
3431 special xattr (currently only "trusted.lov").
3436 Description: Refcount LNET uuids
3437 Details : The global LNET uuid list grew linearly with every startup;
3438 refcount repeated list entries instead of always adding to
3441 Severity : enhancement
3443 Description: Dynamic service threads
3444 Details : Within a small range, start extra service threads
3445 automatically when the request queue builds up.
3448 Frequency : mixed-endian client/server environments
3450 Description: mixed-endian crashes
3451 Details : The new msg_v2 system had some failures in mixed-endian
3454 Severity : enhancement
3456 Description: Easy OST removal
3457 Details : OSTs can be permanently deactivated with e.g. 'lctl
3458 conf_param lustre-OST0001.osc.active=0'
3460 Severity : enhancement
3462 Description: MGS proc entries
3463 Details : Added basic proc entries for the MGS showing what filesystems
3466 Severity : enhancement
3468 Description: provide MGS failover
3469 Details : Added config lock reacquisition after MGS server failover.
3471 Severity : enhancement
3473 Description: add Linux 2.4 support
3474 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3478 Description: a large application tries to do I/O to the same resource and dies
3479 in the middle of it.
3480 Details : Check the req->rq_arrival time after the call to
3481 ost_brw_lock_get(), but before we do anything about
3482 processing it & sending the BULK transfer request. This
3483 should help move old stale pending locks off the queue as
3484 quickly as obd_timeout.
3487 Frequency : when an incorrect nid is specified during startup
3489 Description: ptlrpc connect to non-existant node causes kernel crash
3490 Details : LNET can't be re-entered from an event callback, which
3491 happened when we expire a message after the export has been
3492 cleaned up. Instead, hand the zombie cleanup off to another
3495 Severity : enhancement
3497 Description: plain/inodebits lock performance improvement
3498 Details : Grouping plain/inodebits in granted list by their request modes
3499 and bits policy, thus improving the performance of search through
3503 Frequency : only if OST filesystem is corrupted
3505 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3506 Details : for a short time RPCs with bulk IO are in the replay list,
3507 but replay of bulk IOs is unimplemented. If the OST filesystem
3508 is corrupted due to disk cache incoherency and then replay is
3509 started it is possible to trip an assertion. Avoid putting
3510 committed RPCs into the replay list at all to avoid this issue.
3513 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3515 Description: System hang on startup
3516 Details : This bug allowed the liblustre (e.g. catamount) client to
3517 return to the app before handling all startup RPCs. This
3518 could leave the node unresponsive to lustre network traffic
3519 and manifested as a server ptllnd timeout.
3521 Severity : enhancement
3523 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3524 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3525 boolean to control whether to print peer debug info when a
3526 client's RPC times out.
3529 Frequency : only for kernels with patches from Lustre below 1.4.3
3531 Description: Remove old rdonly API
3532 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3535 Frequency : only for devices with external journals
3537 Description: Set external device read-only also
3538 Details : During a commanded failover stop, we set the disk device
3539 read-only while the server shuts down. We now also set any
3540 external journal device read-only at the same time.
3543 Frequency : when upgrading from 1.4 while trying to change parameters
3545 Description: The wrong (new) MDC name was used when setting parameters for
3546 upgraded MDT's. Also allows changing of OSC (and MDC)
3547 parameters if --writeconf is specified at tunefs upgrade time.
3550 Frequency : when setting specific ost indicies
3552 Description: QOS code breaks on skipped indicies
3553 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3554 created with --index need not be sequential.
3556 Severity : enhancement
3558 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3559 Details : The uninit_groups feature works in conjunction with the kernel
3560 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3561 up the pass1 processing of e2fsck. This is a read-only feature
3562 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3563 mount filesystems that have had this feature enabled.
3565 Severity : enhancement
3567 Description: Improve multi-block allocation algorithm to avoid fragmentation
3568 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3569 allocation locality and avoid filesystem fragmentation.
3571 ------------------------------------------------------------------------------
3573 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3575 * Support for kernels:
3576 2.4.21-47.0.1.EL (RHEL 3)
3577 2.6.5-7.283 (SLES 9)
3578 2.6.9-42.0.10.EL (RHEL 4)
3579 2.6.12.6 vanilla (kernel.org)
3580 2.6.16.27-0.9 (SLES 10)
3581 * Recommended e2fsprogs version: 1.39.cfs5
3583 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3587 Frequency : occasional, depends on client load and configuration
3588 Bugzilla : 12181, 12203
3589 Description: data loss for recently-modified files
3591 Details : In some cases it is possible that recently written or created
3592 files may not be written to disk in a timely manner (this should
3593 normally be within 30s unless client IO load is very high).
3594 The problem appears as zero-length files or files that are a
3595 multiple of 1MB in size after a client crash or client eviction
3596 that are missing data at the end of the file.
3598 This problem is more likely to be hit on clients where files are
3599 repeatedly created and unlinked in the same directory, clients
3600 have a large amount of RAM, have many CPUs, the filesystem has
3601 many OSTs, the clients are rebooted frequently, and/or the files
3602 are not accessed by other nodes after being written.
3604 The presence of the problem can be detected by looking at
3605 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3606 smaller than the second (nr_unused) then dirty files will not
3607 be flushed automatically to disk. "sync; sleep 10" should be
3608 run several times on the node before unmounting it to update
3609 Lustre (this is also safe to run on nodes without this problem).
3611 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3612 that can cause this same problem, so customers using that kernel
3613 also need to update the kernel in addition to Lustre. In order
3614 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3616 It is normal that files written just before a client crash (less
3617 than 30s) may not yet have been flushed to disk, even for local
3621 Frequency : frequent on thin XT3 nodes
3623 Description: UUID collision on thin XT3 Linux nodes
3624 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3625 randomly, since we relied on an insufficiently-seeded PRNG.
3630 Description: OSS hangs after "All ost request buffers busy"
3631 Details : A deadlock between quota and journal operations caused OSS
3632 hangs after printing "All ost request buffers busy."
3635 Frequency : always on liblustre builds
3637 Description: Cleanup compiler warnings on liblustre
3640 Frequency : always on liblustre builds on XT3
3642 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3646 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3647 > 32000 subdirectories
3648 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3649 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3650 subdirectory count crosses 32000. This will aid e2fsck to
3651 correctly handle more than 32000 subdirectories.
3654 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3656 Description: System hang on startup
3657 Details : This bug allowed the liblustre (e.g. catamount) client to
3658 return to the app before handling all startup RPCs. This
3659 could leave the node unresponsive to lustre network traffic
3660 and manifested as a server ptllnd timeout.
3662 Severity : enhancement
3664 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3665 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3666 boolean to control whether to print peer debug info when a
3667 client's RPC times out.
3672 Description: make O_SYNC working on 2.6 kernels
3673 Details : 2.6 kernels use different method for mark pages for write,
3674 so need add a code to lustre for O_SYNC work.
3679 Description: Failure to close file and release space on NFS
3680 Details : Put inode details into lock acquired in ll_intent_file_open.
3681 Use mdc_intent_lock in ll_intent_open to properly
3682 detect all kind of errors unhandled by mdc_enqueue.
3687 Description: proc file read during shutdown sometimes raced obd removal,
3689 Details : Add lock to prevent obd access after proc file removal.
3692 Frequency : Only for files larger than 4GB on 32-bit clients.
3694 Description: improperly doing page alignment of locks
3695 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3696 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3701 Description: RPCs being resent when they shouldn't be
3702 Details : Some RPCs that should not be resent are being resent. This
3703 can cause inconsistencies in the RPC state machine. Do not
3704 resend such requests.
3707 Frequency : rare, only with NFS export
3709 Description: Crash on NFS re-export node
3710 Details : under very unusual load conditions an assertion is hit in
3711 ll_intent_file_open()
3714 Frequency : only if OST filesystem is corrupted
3716 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3717 Details : for a short time RPCs with bulk IO are in the replay list,
3718 but replay of bulk IOs is unimplemented. If the OST filesystem
3719 is corrupted due to disk cache incoherency and then replay is
3720 started it is possible to trip an assertion. Avoid putting
3721 committed RPCs into the replay list at all to avoid this issue.
3726 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3727 Details : Large single O_DIRECT read and write calls can fail to allocate
3728 a sufficiently large buffer to process the request. In case of
3729 allocation failure the allocation is retried with a smaller
3730 buffer and broken into smaller requests.
3732 Severity : enhancement
3734 Description: Add -o localflock option to simulate old noflock behaviour.
3735 Details : This will achieve local-only flock/fcntl locks coherentness.
3740 Description: versioning check is incomplete
3741 Details : Checking the version difference of client vs. server, report
3742 error if the gap is too big.
3747 Description: add support PG_writeback bit
3748 Details : add support for PG_writeback bit for Lustre, for more carefull
3749 work with page cache in 2.6 kernel. This also fix some deadlocks
3750 and remove hack for work O_SYNC with 2.6 kernel.
3752 Severity : enhancement
3754 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3755 Details : The uninit_groups feature works in conjunction with the kernel
3756 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3757 up the pass1 processing of e2fsck. This is a read-only feature
3758 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3759 mount filesystems that have had this feature enabled.
3761 Severity : enhancement
3763 Description: Improve multi-block allocation algorithm to avoid fragmentation
3764 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3765 allocation locality and avoid filesystem fragmentation.
3767 ------------------------------------------------------------------------------
3769 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3771 * Support for kernels:
3772 2.6.9-42.0.3.EL (RHEL 4)
3773 2.6.5-7.276 (SLES 9)
3774 2.4.21-47.0.1.EL (RHEL 3)
3775 2.6.12.6 vanilla (kernel.org)
3776 2.6.16.21-0.8 (SLES10)
3777 * Recommended e2fsprogs version: 1.39.cfs2-0
3779 * The backwards-compatible /proc/sys/portals symlink has been removed
3780 in this release. Before upgrading, please ensure that you change
3781 any configuration scripts or /etc/sysctl.conf files that access
3782 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3783 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3784 in advance of the upgrade on any system running Lustre 1.4.6 or
3785 newer, since /proc/sys/lnet was added in that version.
3786 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3790 Frequency : only when quota is used
3792 Description: avoid scanning export list for quota master
3793 Details : Change the algorithms to avoid scanning export list in order
3794 to improve the efficiency.
3797 Frequency : MDS failover only, very rarely
3799 Description: "went back in time" messages on mds failover
3800 Details : The greatest transno may be lost when the current operation
3801 finishes with an error (transno==0) and the client's last_rcvd
3802 record is over-written. Save the greatest transno in the
3803 mds_last_transno for this case.
3806 Frequency : always for specific kernels and striping counts
3808 Description: client may get "Matching packet too big" without ACL support
3809 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3810 when trying to access files in certain configurations. The
3811 clients should in fact be denied when mounting because they do
3812 not understand ACLs.
3815 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3817 Description: many clients connecting with IO in progress causes connect timeouts
3818 Details : Avoid synchronous journal commits to avoid delays caused by many
3819 clients connecting/disconnecting when bulk IO is in progress.
3820 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3821 OST_IO_PORTAL to avoid delays behind potentially many pending
3825 Frequency : occasionally with multiple writers to a single file
3827 Description: shared writes to file may result in wrong size reported by stat()
3828 Details : Allow growing of kms when extent lock is cancelled
3831 Frequency : always with random mmap IO to multi-striped file
3833 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3834 Details : Only if the hole is at the end of OST object so that kms is too
3835 small. Fix is to increase kms accordingly in ll_nopage.
3838 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3840 Description: writes to a missing object would leak memory on the OST
3841 Details : If there is an inconsistency between the MDS and OST filesystems,
3842 such that the MDS references an object that doesn't exist, writes
3843 to that object will leak memory due to incorrect cleanup in the
3844 error handling path, eventually running out of memory on the OST.
3849 Description: Creating too long symlink causes lustre errors
3850 Details : Check symlink and name lengths before sending requests to MDS.
3853 Frequency : only if flock is enabled (not on by default)
3855 Description: posix locks not released on fd closure on 2.6.9+
3856 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3857 kernels, this caused such locks not to be released on fd close and
3858 then assertions on fs unmount about still used locks.
3861 Frequency : MDS failover only, very rarely
3863 Description: clients may get ASSERTION(granted_lock != NULL)
3864 Details : When request was taking a long time, and a client was resending
3865 a getattr by name lock request. The were multiple lock requests
3866 with the same client lock handle and
3867 mds_getattr_name->fixup_handle_for_resent_request found one of the
3868 lock handles but later failed with ASSERTION(granted_lock != NULL).
3873 Description: handle->h_buffer_credits > 0, assertion failure
3874 Details : h_buffer_credits is zero after truncate, causing assertion
3875 failure. This patch extends the transaction or creates a new
3879 Frequency : NFS re-export or patchless client
3880 Bugzilla : 11179, 10796
3881 Description: Crash on NFS re-export node (__d_move)
3882 Details : We do not want to hash the dentry if we don't have a lock.
3883 But if this dentry is later used in d_move, we'd hit uninitialised
3884 list head d_hash, so we just do this to init d_hash field but
3885 leave dentry unhashed.
3888 Frequency : NFS re-export or patchless client
3890 Description: NFS exports has problem with symbolic link
3891 Details : lustre client didn't properly install dentry when re-exported
3892 to NFS or running patchless client.
3895 Frequency : NFS re-export or patchless client
3897 Description: Various nfs/patchless fixes.
3898 Details : fixes reuse disconected alias for lookup process - this fixes
3899 warning "find_exported_dentry: npd != pd",
3900 fix permission error with open files at nfs.
3901 fix apply umask when do revalidate.
3904 Frequency : occasional
3906 Description: Crash on NFS re-export node
3907 Details : calling clear_page() on the wrong pointer triggered oops in
3908 generic_mapping_read().
3911 Frequency : rarely, using O_DIRECT IO
3913 Description: unaligned directio crashes client with LASSERT
3914 Details : check for unaligned buffers before trying any requests.
3917 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3919 Description: stale data returned from RAID cache
3920 Details : If only a small amount of IO is done to the RAID device before
3921 reading it again it is possible to get stale data from the RAID
3922 cache instead of reading it from disk.
3925 Frequency : always for sles10 kernel
3927 Description: sles10 support
3928 Details : ll_follow_link: compile fixes and using of nd_set_link
3929 under newer kernels.
3932 Frequency : depends on arch, kernel and compiler version, always on sles10
3935 Description: recursive or deep enough symlinks cause stack overflow
3936 Details : getting rid of large stack-allocated variable in
3940 Frequency : depends on hardware
3942 Description: lustre write performance loss in the SLES10 kernel
3943 Details : the performance loss is caused by using of write barriers in the
3944 ext3 code. The SLES10 kernel turns barrier support on by
3945 default. The fix is to undo that change for ldiskfs.
3947 ------------------------------------------------------------------------------
3949 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3951 * Support for kernels:
3952 2.6.9-42.0.3EL (RHEL 4)
3953 2.6.5-7.276 (SLES 9)
3954 2.4.21-47.0.1.EL (RHEL 3)
3955 2.6.12.6 vanilla (kernel.org)
3959 Frequency : quota enabled and large files being deleted
3961 Description: releasing more than 4GB of quota at once hangs OST
3962 Details : If a user deletes more than 4GB of files on a single OST it
3963 will cause the OST to spin in an infinite loop. Release
3964 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3969 Description: statfs data retrieved from /proc may be stale or zero
3970 Details : When reading per-device statfs data from /proc, in the
3971 {kbytes,files}_{total,free,avail} files, it may appear
3972 as zero or be out of date.
3975 Frequency : systems with MD RAID1 external journal devices
3977 Description: lconf's call to blkid is confused by RAID1 journal devices
3978 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3979 of returning all block devices that match the journal UUID.
3982 Frequency : always, for aggregate stripe size over 4GB
3984 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3985 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3986 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3989 Frequency : always if "lfs find" used on a local file/directory
3991 Description: "lfs find" segfaults if used on a local file/directory
3992 Details : The case where a directory component was not specified wasn't
3993 handled correctly. Handle this properly.
3996 Frequency : always on ppc64
3998 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3999 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4000 perform endianness conversion.
4003 Frequency : rarely (truncate to non-zero file size after write under load)
4004 Bugzilla : 10730, 10687
4005 Description: Files padded with zeros to next 4K multiple
4006 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4007 it is possible that files that are truncated to a non-zero size
4008 immediately after being written are filled with zero bytes beyond
4009 the truncated size. No file data is lost.
4011 Severity : enhancement
4013 Description: Allow recovery/failover for liblustre clients.
4014 Details : liblustre clients were unaware of failover configurations until
4017 Severity : enhancement
4019 Description: user file locks should fail when not mounting with flock option
4020 Details : Set up an error-returning stub in ll_file_operations.lock field
4021 to prevent incorrect behaviour when client is mounted without
4022 flock option. Also, set up properly f_op->flock field for
4026 Frequency : always on ia64
4028 Description: "lfs df" loops on printing out MDS statfs information
4029 Details : The obd_ioctl_data was not initialized and in some systems
4030 this caused a failure during the ioctl that did not return
4031 an error. Initialize the struct and return an error on failure.
4034 Frequency : SLES 9 only
4036 Description: Error of copying files with lustre special EAs as root
4037 Details : Client side always return success for setxattr call for lustre
4038 special xattr (currently only "trusted.lov").
4041 Frequency : rarely on clusters with both ia64+i386 clients
4043 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4044 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4045 clients are doing concurrent writes to the same file it is
4046 possible that the ia64 clients may LASSERT because the OST
4047 extent locks are not PAGE_SIZE aligned. Ensure that grown
4048 locks are always aligned on the request boundary.
4051 Frequency : specific use, occasional
4053 Description: Overwriting in use executable truncates on-disk binary image
4054 Details : If one node attempts to overwrite an executable in use by
4055 another node, we now correctly return ETXTBSY instead of
4056 truncating the file.
4058 Severity : enhancement
4060 Description: Async OSC create to avoid the blocking unnecessarily.
4061 Details : If an OST has no remaining object, system will block on the creating
4062 when need to create a new object on this OST. Now, ways use
4063 pre-created objects when available, instead of blocking on an
4064 empty osc while others are not empty. If we must block, we block
4065 for the shortest possible period of time.
4070 Description: chmod on Lustre root is propagated to other clients
4071 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4072 invalid by the follow_mount time.
4075 Frequency : liblustre clients only
4077 Description: Race in 'instant cancel' lock handling could lead to such locks
4078 never to be granted in case of SMP MDS
4079 Details : Do not destroy not yet granted but cbpending locks in
4083 Frequency : replay/resend of open
4085 Description: non null lock assetion failure in mds_intent_policy
4086 Details : Trying to replay/resend lockless open requests resulted in
4087 mds_open() returning 0 with no lock. Now it sets a flag if
4088 a lock is going to be returned.
4090 Severity : enhancement
4092 Description: Checksum enhancements
4093 Details : New checksum enhancements allow for resending RPCs that failed
4096 Severity : enhancement
4098 Description: Tunables on number of dirty pages in cacche
4099 Details : Allow to set limit on number of dirty pages cached.
4104 Description: client crash on unmount - lock still has references
4105 Details : In some error handling cases it was possible to leak a lock
4106 reference on a client while accessing a file. This was not
4107 harmful to the client during operation, but would cause the
4108 client to crash when the filesystem is unmounted.
4111 Frequency : specific case, rare
4113 Description: ETXTBSY on mds though file not in use by client
4114 Details : ETXTBSY is no longer incorrectly returned when attempting to
4115 chmod or chown a directory that the user previously tried to
4116 execute or a currently-executing binary.
4119 Frequency : extremely rare except on liblustre-based clients
4121 Description: Lustre space not freed when files are deleted
4122 Details : Clean up open-unlinked files after client eviction. Previously
4123 the unlink was skipped and the files remained as orphans.
4128 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4129 Details : In some cases it was possible to send a blocking callback to
4130 a client doing a glimpse, even though that client didn't get
4131 a lock granted. When the glimpse lock is cancelled on the OST
4132 the freed lock is left on the waiting list and corrupted the list.
4135 Frequency : all core dumps
4137 Description: Broke core dumps to lustre
4138 Details : Negative dentry may be unhashed if parent does not have UPDATE
4139 lock, but some callers, e.g. do_coredump, expect dentry to be
4140 hashed after successful create, hash it in ll_create_it.
4142 ------------------------------------------------------------------------------
4144 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4146 * Support for kernels:
4147 2.6.9-42.0.2.EL (RHEL 4)
4148 2.6.5-7.276 (SLES 9)
4149 2.4.21-40.EL (RHEL 3)
4150 2.6.12.6 vanilla (kernel.org)
4154 Frequency : always on RHEL 3
4156 Description: Number of open files grows over time
4157 Details : The number of open files grows over time, whether or not
4158 Lustre is started. This was due to a filp leak introduced
4159 by one of our kernel patches.
4161 ------------------------------------------------------------------------------
4163 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4165 * Support for kernels:
4166 2.6.9-42.EL (RHEL 4)
4167 2.6.5-7.276 (SLES 9)
4168 2.4.21-40.EL (RHEL 3)
4169 2.6.12.6 vanilla (kernel.org)
4174 Bugzilla : 5719, 9635, 9792, 9684
4175 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4176 Details : If a server is under heavy load and cannot reply to new
4177 connection requests before the client resends the (re)connect,
4178 the connection handling code can behave badly if two service
4179 threads are concurrently handing separate (re)connections from
4180 the same client. Add better locking to the connection handling
4181 code, and ensure that only a single connection will be processed
4182 for a given client UUID, even if the lock is dropped.
4184 Severity : enhancement
4186 Description: add TCP zero-copy support to kernel
4187 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4188 sends if the hardware supports scatter-gather and checksumming.
4189 This allows socklnd to do client-write and server-read more
4190 efficiently and reduce CPU utilization from skbuf copying.
4193 Frequency : only if NFS exporting from client
4195 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4196 Details : If setting attributes on a file created under NFS that had
4197 never been opened it would be possible to oops the client
4198 if the file had no objects.
4201 Frequency : always for liblustre
4203 Description: liblustre client does MDS+OSTs setattr RPC for each write
4204 Details : When doing a write from a liblustre client, the client
4205 incorrectly issued an RPC to the MDS and each OST the file was
4206 striped over in order to update the timestamps. When writing
4207 with small chunks and many clients this could overwhelm the MDS
4208 with RPCs. In all cases it would slow down the write because
4209 these RPCs are unnecessary.
4211 Severity : enhancement
4213 Description: allow number of MDS service threads to be changed at module load
4214 Details : It is now possible to change the number of MDS service threads
4215 running. Adding "options mds mds_num_threads={N}" to the MDS's
4216 /etc/modprobe.conf will set the number of threads for the next
4217 time Lustre is restarted (assuming the "mds" module is also
4218 reloaded at that time). The default number of threads will
4219 stay the same, 32 for most systems.
4224 Description: OST crash if filesystem is unformatted or corrupt
4225 Details : If an OST is started on a device that has never been formatted
4226 or if the filesystem is corrupt and cannot even mount then the
4227 error handling cleanup routines would dereference a NULL pointer.
4232 Description: NULL pointer deref in llap_from_page.
4233 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4234 incorrect) mapping if the page was truncated/reclaimed while it was
4235 searched for. Check for this condition and skip such pages when
4236 doing readahead. Introduce extra check to llap_from_page() to
4237 verify page->mapping->host is non-NULL (so page is not anonymous).
4240 Frequency : Sometimes when using sys_sendfile
4242 Description: "page not covered by a lock" warnings from ll_readpage
4243 Details : sendfile called ll_readpage without right page locks present.
4244 Now we introduced ll_file_sendfile that does necessary locking
4245 around call to generic_file_sendfile() much like we do in
4249 Frequency : with certain MDS communication failures at client mount time
4251 Description: NULL pointer deref after failed client mount
4252 Details : a client connection request may delayed by the network layer
4253 and not be sent until after the PTLRPC layer has timed out the
4254 request. If the client fails the mount immediately it will try
4255 to clean up before the network times out the request. Add a
4256 reference from the request import to the obd device and delay
4257 the cleanup until the network drops the request.
4260 Frequency : occasionally during client (re)connect
4262 Description: assertion failure during client (re)connect
4263 Details : processing a client connection request may be delayed by the
4264 client or server longer than the client connect timeout. This
4265 causes the client to resend the connection request. If the
4266 original connection request is replied in this interval, the
4267 client may trip an assertion failure in ptlrpc_connect_interpret()
4268 which thought it would be the only running connect process.
4271 Frequency : only with obd_echo servers and clients that are rebooted
4273 Description: kernel BUG accessing uninitialized data structure
4274 Details : When running an obd_echo server it did not start the ping_evictor
4275 thread, and when a client was evicted an uninitialized data
4276 structure was accessed. Start the ping_evictor in the RPC
4277 service startup instead of the OBD startup.
4279 Severity : enhancement
4280 Bugzilla : 10193 (patchless)
4281 Description: Remove dependency on various unexported kernel interfaces.
4282 Details : No longer need reparent_to_init, exit_mm, exit_files,
4283 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4286 Frequency : rare (only users of deprecated and unsupported LDAP config)
4288 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4289 Details : LDAP apparently contains 'lustreName' attributes instead of
4290 'name'. A simple remapping of the name is sufficient.
4293 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4295 Description: waiting_locks_callback trips kernel BUG if client is evicted
4296 Details : Running with the dump_on_timeout debug flag turned on makes
4297 it possible that the waiting_locks_callback() can try to dump
4298 the Lustre kernel debug logs from an interrupt handler. Defer
4299 this log dumping to the expired_lock_main() thread.
4301 Severity : enhancement
4303 Description: Support NFS exporting on 2.6 kernels.
4304 Details : Implement non-rawops metadata methods for NFS server to use without
4305 changing NFS server code.
4308 Frequency : very rare (synthetic metadata workload only)
4310 Description: two racing renames might cause an MDS thread to deadlock
4311 Details : Running the "racer" program may cause one MDS thread to rename
4312 a file from being the source of a rename to being the target of
4313 a rename at exactly the same time that another thread is doing
4314 so, and the second thread has already enqueued these locks after
4315 doing a lookup of the target and is trying to relock them in
4316 order. Ensure that we don't try to re-lock the same resource.
4319 Frequency : only very large systems with liblustre clients
4321 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4322 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4323 all OSTs in parallel. This allows the eviction of stale liblustre
4324 clients to proceed much faster than if they were done in series,
4325 and also offers similar improvements for other set_info RPCs.
4330 Description: excessive CPU usage during initial read phase on client
4331 Details : During the initial read phase on a client, it would agressively
4332 retry readahead on the file, consuming too much CPU and impacting
4333 performance (since 1.4.5.8). Improve the readahead algorithm
4334 to avoid this, and also improve some other common cases (read
4335 of small files in particular, where "small" is files smaller than
4336 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4341 Description: MDS crash when receiving packet with unknown intent.
4342 Details : Do not LBUG in unknown intent case, just return -EFAULT
4344 Severity : enhancement
4345 Bugzilla : 9293, 9385
4346 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4347 Details : Do not serialize getattr (non-intent version) and statfs.
4350 Frequency : occasional, when OST network is overloaded/intermittent
4352 Description: client evicted by OST after bulk IO timeout
4353 Details : If a client sends a bulk IO request (read or write) the OST
4354 may evict the client if it is unresposive to its data GET/PUT
4355 request. This is incorrect if the network is overloaded (takes
4356 too long to transfer the RPC data) or dropped the OST GET/PUT
4357 request. There is no need to evict the client at all, since
4358 the pinger and/or lock callbacks will handle this, and the
4359 client can restart the bulk request.
4362 Frequency : Always when mmapping file with no objects
4364 Description: client crashes when mmapping file with no objects
4365 Details : Check that we actually have objects in a file before doing any
4366 operations on objects in ll_vm_open, ll_vm_close and
4372 Description: Request leak when working with deleted CWD
4373 Details : Introduce advanced request refcount tracking for requests
4374 referenced from lustre intent.
4376 Severity : Enhancement
4378 Description: Cache open file handles on client.
4379 Details : MDS now will return special lock along with openhandle, if
4380 requested and client is allowed to hold openhandle, even if unused,
4381 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4382 closing files for every read/write openration.
4384 Severity : Enhancement
4386 Description: Cache open negative dentries on client when possible.
4387 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4388 negative dentries on lock revocation.
4393 Description: Remounting a client read-only wasn't possible with a zconf mount
4394 Details : It wasn't possible to remount a client read-only with llmount.
4396 Severity : enhancement
4397 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4398 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4399 MPICH is included to add Lustre-specific ADIO interfaces.
4400 This is based closely on the UFS ADIO layer and only differs
4401 in file creation, in order to allow the OST striping to be set.
4402 This is user-contributed code and not supported by CFS.
4407 Description: extended inode attributes (immutable, append-only) work improperly
4408 when 2.4 and 2.6 kernels are used on client/server or vice versa
4409 Details : Introduce kernel-independent values for these flags.
4411 Severity : enhancement
4414 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4415 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4416 granularity. Now, Fractional MB granularity is be supported,
4417 this is very useful for low memory system.
4419 Severity : enhancement
4421 Description: Getattr by fid
4422 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4423 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4424 have locks and avoids getattr rpc requests for it on every stat.
4427 Frequency : Always, for filesystems larger than 2TB
4429 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4430 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4431 filesystems larger than 2TB to be mounted with mballoc enabled.
4434 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4436 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4437 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4438 filesystems larger than 2TB is mounted then the kernel will
4439 silently corrupt the start of the filesystem. CONFIG_LBD is
4440 enabled for all CFS-supported kernels, but the possibility of
4441 this happening with a modified kernel config exists.
4443 Severity : enhancement
4445 Description: add client O_DIRECT support for 2.6 kernels
4446 Details : It is now possible to do O_DIRECT reads and writes to files
4447 in the Lustre client mountpoint on 2.6 kernel clients.
4449 Severity : enhancement
4451 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4452 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4453 parallel, not waiting for response from every OST before sending
4454 a rpc to the next OST.
4459 Description: setattr vs write race when updating file timestamps
4460 Details : Client processes that update a file timestamp into the past
4461 right after writing to the file (e.g. tar) it is possible that
4462 the updated file modification time can be reset to the current
4463 time due to a race between processing the setattr and write RPC.
4465 Severity : enhancement
4467 Description: Bring 'lfs find' closer in line with regular Linux find.
4468 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4469 -print0 options and obtains all the needed info through the lustre
4472 Severity : enhancement
4474 Description: support up to 1024 configured devices on one node
4475 Details : change obd_dev array from statically allocated to dynamically
4476 allocated structs as they are first used to reduce memory usage
4481 Description: Flush dirty partially truncated pages during truncate
4482 Details : Immediatelly flush partially truncated pages in filter_setattr,
4483 this way we completely avoid having any pages in page cache on OST
4484 and can retire ugly workarounds during writes to flush such pages.
4489 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4490 Details : protect inode from truncation within vfs_unlink() context
4491 just take a reference before calling vfs_unlink() and release it
4492 when parent's i_sem is free.
4497 Description: last_id value checked outside lock on OST caused LASSERT failure
4498 Details : If there were multiple MDS->OST object precreate requests in
4499 flight, it was possible that the OST's last object id was checked
4500 outside a lock and incorrectly tripped an assertion. Move checks
4501 inside locks, and discard old precreate requests.
4504 Frequency : always, if extents are used on OSTs
4506 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4507 Details : OSTs using the extents format would not zero the high 16 bits of
4508 the index physical block number. This is not a problem for any
4509 OST filesystems smaller than 16TB, and no kernels support ext3
4510 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4511 new/modified files) and can be fixed for existing filesystems
4512 with e2fsprogs-1.39-cfs1.
4517 Description: import connection selection may be incorrect if timer wraps
4518 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4519 import connections to be ignored if the 32-bit jiffies counter
4520 wraps. Use a 64-bit jiffies counter.
4523 Frequency : very large clusters immediately after boot
4525 Description: LNET request buffers exhausted under heavy short-term load
4526 Details : If a large number of client requests are generated on a service
4527 that has previously never seen so many requests it is possible
4528 that the request buffer growth cannot keep up with the spike in
4529 demand. Instead of dropping incoming requests, they are held in
4530 the LND until the RPC service can accept more requests.
4533 Frequency : Sometimes during replay
4535 Description: Assertion failure in ll_local_open after replay.
4536 Details : If replay happened on an open request reply before we were able
4537 to set replay handler, reply will become not swabbed tripping the
4538 assertion in ll_local_open. Now we set the handler right after
4539 recognising of open request
4542 Frequency : very rare
4544 Description: kernel reports "badness in vsnprintf"
4545 Details : Reading from the "recovery_status" /proc file in small chunks
4546 may cause a negative length in lprocfs_obd_rd_recovery_status()
4547 call to vsnprintf() (which is otherwise harmless). Exit early
4548 if there is no more space in the output buffer.
4550 Severity : enhancement
4552 Description: clear OBD RPC statistics by writing to them
4553 Details : It is now possible to clear the OBD RPC statistics by writing
4554 to the "stats" file.
4559 Description: Client mtime is not the same on different clients after utimes
4560 Details : In some cases, the client was using the utimes() syscall on
4561 a file cached on another node. The clients now validate the
4562 ctime from the MDS + OSTs to determine which one is right.
4567 Description: Inability to activate failout mode
4568 Details : lconf script incorrectly assumed that in pythong string's numeric
4569 value is used in comparisons.
4572 Frequency : always with multiple stripes per file
4574 Description: Inefficient object allocation for mutli-stripe files
4575 Details : When selecting which OSTs to stripe files over, for files with
4576 a stripe count that divides evenly into the number of OSTs,
4577 the MDS is always picking the same starting OST for each file.
4578 Return the OST selection heuristic to the original design.
4583 Description: mount failures may take full timeout to return an error
4584 Details : Under some heavy load conditions it is possible that a
4585 failed mount can wait for the full obd_timeout interval,
4586 possibly several minutes, before reporting an error.
4587 Instead return an error as soon as the status is known.
4589 Frequency : quota enabled and large files being deleted
4591 Description: releasing more than 4GB of quota at once hangs OST
4592 Details : If a user deletes more than 4GB of files on a single OST it
4593 will cause the OST to spin in an infinite loop. Release
4594 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4599 Description: statfs data retrieved from /proc may be stale or zero
4600 Details : When reading per-device statfs data from /proc, in the
4601 {kbytes,files}_{total,free,avail} files, it may appear
4602 as zero or be out of date.
4605 Frequency : systems with MD RAID1 external journal devices
4607 Description: lconf's call to blkid is confused by RAID1 journal devices
4608 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4609 of returning all block devices that match the journal UUID.
4612 Frequency : always, for aggregate stripe size over 4GB
4614 Description: assertion fails when trying to use 4GB stripe size
4615 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4616 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4619 Frequency : always on ppc64
4621 Description: the first write on an ext3 filesystem with mballoc got stuck
4622 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4623 perform endianness conversion.
4625 ------------------------------------------------------------------------------
4627 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4629 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4630 INTEROPERATE with older versions automatically. Please read the
4631 user documentation before upgrading any part of a live system.
4632 * WARNING: Lustre networking configuration changes are required with
4633 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4636 * Support for kernels:
4637 2.6.9-22.0.2.EL (RHEL 4)
4638 2.6.5-7.244 (SLES 9)
4639 2.6.12.6 vanilla (kernel.org)
4642 Severity : enhancement
4643 Bugzilla : 7981/8208
4644 Description: Introduced Lustre Networking (LNET)
4645 Details : LNET is new networking infrastructure for Lustre, it includes
4646 a reorganized network configuration mode (see the user
4647 documentation for full details) as well as support for routing
4648 between different network fabrics. Lustre Networking Devices
4649 (LNDs) for the supported network fabrics have also been
4650 created for this new infrastructure.
4652 Severity : enhancement
4653 Description: Introduced Access control lists
4654 Details : clients can set ACLs on files and directories in order to have
4655 more fine-grained permissions than the standard Unix UGO+RWX.
4656 The MDS must be started with the "-o acl" mount option.
4658 Severity : enhancement
4659 Description: Introduced filesystem quotas
4660 Details : Administrators may now establish per-user quotas on the
4663 Severity : enhancement
4665 Description: Configuration change for the XT3
4666 The PTLLND is now used to run Lustre over Portals on the XT3
4667 The configure option(s) --with-cray-portals are no longer used.
4668 Rather --with-portals=<path-to-portals-includes> is used to
4669 enable building on the XT3. In addition to enable XT3 specific
4670 features the option --enable-cray-xt3 must be used.
4675 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4676 Details : A race between allocating a new llcd and re-getting the llcd_lock
4677 allowed another thread to grab newly-allocated llcd.
4679 Severity : enhancement
4681 Description: 2.6 OST async journal commit and locking fix to improve performance
4682 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4683 2.6 kernels are now async as they already were in 2.4 kernels so
4684 that they can commit concurrently with the network bulk transfer.
4685 For block-allocated files the filter allocation semaphore is held
4686 to avoid filesystem fragmentation during allocation. BKL lock
4687 removed for 2.6 xattr operations where it is no longer needed.
4692 Description: lconf incorrectly determined whether two IP networks could talk
4693 Details : In some more complicated routing and multiple-network
4694 configurations, lconf will avoid trying to make a network
4695 connection to a disjoint part of the IP space. It was doing the
4696 math incorrectly for one set of cases.
4701 Description: Fix for potential infinite loop processing records in an llog.
4702 Details : If an llog record is corrupted/zeroed, it is possible to loop
4703 forever in llog_process(). Validate the llog record length
4704 and skip the remainder of the block on error.
4707 Frequency : occasional (liblustre only)
4709 Description: liblustre could not open files whose last component is a symlink
4710 Details : sysio_path_walk() would incorrectly pass the open intent to
4711 intermediate path components.
4714 Frequency : rare (liblustre only with non-standard tuning)
4715 Bugzilla : 7201 (7350)
4716 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4717 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4718 after it has been released, only do this for referenced locks
4720 Severity : enhancement
4722 Description: specifying an (invalid) directory default stripe_size of -1
4723 would reset the directory default striping
4724 Details : stripe_size -1 was used internally to signal directory stripe
4725 removal, now use "all default" to signal dir stripe removal
4726 as a directory striping of "all default" is not useful
4729 Frequency : common for large clusters running liblustre clients
4731 Description: doing an ls when liblustre clients are running is slow
4732 Details : sending a glimpse AST to a liblustre client waits for every AST
4733 to time out, as liblustre clients will not respond. Since they
4734 cannot cache data we refresh the OST lock LVB from disk instead.
4736 Severity : enhancement
4738 Description: doing an ls at the same time as file IO can be slow
4739 Details : enqueue and other "small" requests can be blocked behind many
4740 large IO requests. Create a new OST IO portal for non-IO
4741 requests so they can be processed faster.
4744 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4746 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4747 created file with mode 000
4748 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4749 allow an ftruncate() as a non-root user to a file with mode 000.
4750 Lustre now respects this flag to disable mode checks when
4751 truncating a file owned by the user
4754 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4757 Description: Revoking locks from clients that went dead or catatonic might take
4759 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4760 cancellation of such locks instant on servers without waiting for
4761 any reply from clients. Clients drops these locks when cancel
4762 notification from server is received without replying.
4765 Frequency : liblustre-only, when liblustre client dies or becomes busy
4767 Description: Doing ls on Linux clients can take a long time with active
4769 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4770 granting such locks to it in the first place if possible. Locks
4771 are taken by proxy on the OST during the read or write and
4772 dropped immediately afterward. Add connect flags handling, do
4773 not grant locks to liblustre clients for glimpse ASTs.
4775 Severity : enhancement
4777 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4778 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4779 sequential reads and behaves badly with applications doing random
4780 reads. Improve it by reading ahead at least read region, and
4781 avoiding excessive large RPC for small reads.
4783 Severity : enhancement
4785 Description: Creating more than 1000 files for a single job may cause a load
4786 imbalance on the OSTs if there are also a large number of OSTs.
4787 Details : qos_prep_create() uses an OST index reseed value that is an
4788 even multiple of the number of available OSTs so that if the
4789 reseed happens in the middle of the object allocation it will
4790 still utilize the OSTs as uniformly as possible.
4795 Description: OST or MDS may oops in ping_evictor_main()
4796 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4797 but doesn't restart at beginning of obd_exports_timed list
4800 Severity : enhancement
4802 Description: improve by-nid export eviction on the MDS and OST
4803 Details : allow multiple exports with the same NID to be evicted at one
4804 time without re-searching the exports list.
4807 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4809 Description: MDS may oops in groups_free()
4810 Details : in rare race conditions a newly allocated group_info struct is
4811 freed again, and this can be NULL. The 2.4 compatibility code
4812 for groups_free() checked for a NULL pointer, but 2.6 did not.
4815 Frequency : common for liblustre clients doing little filesystem IO
4816 Bugzilla : 9352, 7313
4817 Description: server may evict liblustre clients accessing contended locks
4818 Details : if a client is granted a lock or receives a completion AST
4819 with a blocking AST already set it would not reply to the AST
4820 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4823 Frequency : lfs setstripe, only systems with more than 160 OSTs
4825 Description: unable to set striping with a starting offset beyond OST 160
4826 Details : llapi_create_file() incorrectly limited the starting stripe
4827 index to the maximum single-file stripe count.
4830 Frequency : LDAP users only
4832 Description: lconf did not handle in-kernel recovery with LDAP properly
4833 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4835 Severity : enhancement
4837 Description: bind OST threads to NUMA nodes to improve performance
4838 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4839 node and do their allocations there to localize memory access
4841 Severity : enhancement
4843 Description: llmount can determine client NID directly from Myrinet (GM)
4844 Details : the client NID code from gmnalnid was moved directly into
4845 llmount, removing the need to use this or specifying the
4846 client NID explicitly when mounting GM clients with zeroconf
4849 Frequency : if client is started with down MDS
4851 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4852 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4853 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4854 ptlrpc_import_delayed_req() because we didn't check if the
4855 request was interrupted, and we also didn't break out of the
4856 event loop if there was no timeout
4861 Description: data loss during non-page-aligned writes to a single file from
4862 both multiple nodes and multiple threads on one node at same time
4863 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4864 inconsistency led to false short-reads, that were cached and later
4865 used by ->prepare_write() to fill in partially written page,
4866 leading to data loss.
4869 Frequency : always, if lconf --abort_recovery used
4871 Description: lconf --abort_recovery fails with 'Operation not supported'
4872 Details : lconf was attempting to abort recovery on the MDT device and not
4875 Severity : enhancement
4877 Description: remove cleanup logs
4878 Details : replace lconf-generated cleanup logs with lustre internal
4879 cleanup routines. Eliminates the need for client-cleanup and
4882 Severity : enhancement
4884 Description: add support for EAs (user and system) on lustre filesystems
4885 Details : it is now possible to store extended attributes in the Lustre
4886 client filesystem, and with the user_xattr mount option it
4887 is possible to allow users to store EAs on their files also
4889 Severity : enhancement
4891 Description: Add possibility (config option) to show minimal available OST free
4893 Details : When compiled with --enable-mindf configure option, statfs(2)
4894 (and so, df) will return least minimal free space available from
4895 all OSTs as amount of free space on FS, instead of summary of
4896 free spaces of all OSTs.
4898 Severity : enhancement
4900 Description: do not expand extent locks acquired on OST-side
4901 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4902 by server: they are not cached anyway.
4905 Frequency : when mmap is used/binaries executed from Lustre
4907 Description: Unmmap pages before throwing them away from read cache.
4908 Details : llap_shrink cache now attempts to unmap pages before discarding
4909 them (if unmapping failed - do not discard). SLES9 kernel has
4910 extra checks that trigger if this unmapping is not done first.
4915 Description: lconf didn't resolve symlinks before checking to see whether a
4916 given mountpoint was already in use
4919 Frequency : when migrating failover services
4920 Bugzilla : 6395, 9514
4921 Description: When migrating a subset of services from a node (e.g. failback
4922 from a failover service node) the remaining services would
4923 time out and evict clients.
4924 Details : lconf --force (implied by --failover) sets the global obd_timeout
4925 to 5 seconds in order to quickly disconnect, but this caused
4926 other RPCs to time out too quickly. Do not change the global
4927 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4929 Severity : enhancement
4930 Frequency : if MDS is started with down OST
4931 Bugzilla : 9439,5706
4932 Description: Allow startup/shutdown of an MDS without depending on the
4933 availability of the OSTs.
4934 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4935 Add appropriate locking and lov-osc refcounts for safe
4936 cleaning. Add osc abort_inflight calls in case the
4937 synchronize never started.
4940 Frequency : occasional (Cray XT3 only)
4942 Description: root not authorized to access files in CRAY_PORTALS environment
4943 Details : The client process capabilities were not honoured on the MDS in
4944 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4945 been accessed by an authorized user then root was able to access
4946 the file on the local client also. The root user capabilities
4947 are now allowed on the MDS, as this environment has secure UID.
4950 Frequency : occasional
4952 Description: ldiskfs "too long searching" message happens too often
4953 Details : A debugging message (otherwise harmless) prints too often on
4954 the OST console. This has been reduced to only happen when
4955 there are fragmentation problems on the filesystem.
4960 Description: Division by zero in statfs when all OSCs are inactive
4961 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4962 lov_statfs divides by value returned by lov_get_stripecnt().
4966 Bugzilla : 9489, 3273
4967 Description: First write from each client to each OST was only 4kB in size,
4968 to initialize client writeback cache, which caused sub-optimal
4969 RPCs and poor layout on disk for the first written file.
4970 Details : Clients now request an initial cache grant at (re)connect time
4971 and so that they can start streaming writes to the cache right
4972 away and always do full-sized RPCs if there is enough data.
4973 If the OST is rebooted the client also re-establishes its grant
4974 so that client cached writes will be honoured under the grant.
4979 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4980 Details : Now I/O RPCs go to different portal number and (presumably) fast
4981 lock requests (and glimses) and other RPCs get their own service
4982 threads pool that should be able to service those RPCs
4985 Severity : enhancement
4987 Description: Ability to exchange lustre version between client and servers and
4988 issue warnings at client side if client is too old. Also for
4989 liblustre clients there is ability to refuse connection of too old
4991 Details : New 'version' field is added to connect data structure that is
4992 filled with version info. That info is later checked by server and
4996 Frequency : rare, liblustre only.
4997 Bugzilla : 9296, 9581
4998 Description: Two simultaneous writes from liblustre at offset within same page
4999 might proceed at the same time overwriting eachother with stale
5001 Details : I/O lock withing llu_file_prwv was released too early, before data
5002 actually was hitting the wire. Extended lock-holding time until
5003 server acknowledges receiving data.
5006 Frequency : extremely rare. Never observed in practice.
5008 Description: avoid generating lustre_handle cookie of 0.
5009 Details : class_handle_hash() generates handle cookies by incrementing
5010 global counter, and can hit 0 occasionaly (this is unlikely, but
5011 not impossible, because initial value of cookie counter is
5012 selected randonly). Value of 0 is used as a sentinel meaning
5013 "unassigned handle" --- avoid it. Also coalesce two critical
5014 sections in this function into one.
5016 Severity : enhancement
5018 Description: allow liblustre clients to delegate truncate locking to OST
5019 Details : To avoid overhead of locking, liblustre client instructs OST to
5020 take extent lock in ost_punch() on client's behalf. New connection
5021 flag is added to handle backward compatibility.
5023 Severity : enhancement
5024 Bugzilla : 4928, 7341, 9758
5025 Description: allow number of OST service threads to be specified
5026 Details : a module parameter allows the number of OST service threads
5027 to be specified via "options ost ost_num_threads={N}" in the
5028 OSS's /etc/modules.conf or /etc/modprobe.conf.
5032 Bugzilla : 6146, 9635, 9895
5033 Description: servers crash with bad pointer in target_handle_connect()
5034 Details : In rare cases when a client is reconnecting it was possible that
5035 the connection request was the last reference for that export.
5036 We would temporarily drop the export reference and get a new
5037 one, but this may have been the last reference and the export
5038 was just destroyed. Get new reference before dropping old one.
5040 Severity : enhancement
5041 Frequency : if client is started with failover MDS
5043 Description: Allow multiple MDS hostnames in the mount command
5044 Details : Try to read the configuration from all specified MDS
5045 hostnames during a client mount in case the "primary"
5048 Severity : enhancement
5050 Description: Stop sending data to evicted clients as soon as possible.
5051 Details : Check if the client we are about to send or are sending data to
5052 was evicted already. (Check is done every second of waiting,
5053 for which l_wait_event interface was extended to allow checking
5054 of exit condition at specified intervals).
5057 Frequency : rare, normally only when NFS exporting is done from client
5059 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5061 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5064 Frequency : always, if srand() is called before liblustre initialization
5066 Description: Liblustre uses system PRNG disturbing its usage by user application
5067 Details : Introduce internal to lustre fast and high-quality PRNG for
5068 lustre usage and make liblustre and some other places in generic
5069 lustre code to use it.
5071 Severity : enhancement
5072 Bugzilla : 9477, 9557, 9870
5073 Description: Verify that the MDS configuration logs are updated when xml is
5074 Details : Check if the .xml configuration logs are newer than the config
5075 logs stored on the MDS and report an error if this is the case.
5076 Request --write-conf, or allow starting with --old_conf.
5078 Severity : enhancement
5080 Description: Handle symlinks in the path when checking if Lustre is mounted.
5081 Details : Resolve intermediate symlinks when checking if a client has
5082 mounted a filesystem to avoid duplicate client mounts.
5087 Description: lconf can hit an error exception but still return success.
5088 Details : The lconf command catches the Command error exception at the top
5089 level script context and will exit with the associated exit
5090 status, but doesn't ensure that this exit status is non-zero.
5095 Description: failure of ptlrpc thread startup can cause oops
5096 Details : Starting a ptlrpc service thread can fail if there are a large
5097 number of threads or the server memory is very fragmented.
5098 Handle this without oopsing.
5101 Frequency : always, only if liblustre and non-default acceptor port was used
5103 Description: liblustre cannot connect to servers with non-default acceptor port
5104 Details : tcpnal_set_default_params() was not called and was therefore
5105 ignoring the environment varaible TCPNAL_PORT, as well as other
5106 TCPNAL_ environment variables
5111 Description: two objects could be created on the same OST for a single file
5112 Details : If an OST is down, in some cases it was possible to create two
5113 objects on a single OST for a single file. No problems other
5114 than potential performance impact and spurious error messages.
5118 Bugzilla : 5681, 9562
5119 Description: Client may oops in ll_unhash_aliases
5120 Details : Client dcache may become inconsistent in race condition.
5121 In some cases "getcwd" can fail if the current directory is
5127 Description: Inode refcounting problems in NFS export code
5128 Details : link_raw functions used to call d_instantiate without obtaining
5129 extra inode reference first.
5133 Bugzilla : 9942, 9903
5134 Description: Referencing freed requests leading to crash, memleaks with NFS.
5135 Details : We used to require that call to ll_revalidate_it was always
5136 followed by ll_lookup_it. Also with revalidate_special() it is
5137 possible to call ll_revalidate_it() twice for the same dentry
5138 even if first occurence returned success. This fix changes semantic
5139 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5140 reference on a request referred from the intent.
5141 ll_intent_release() then releases such a request.
5144 Frequency : rare, normally benchmark loads only
5146 Description: unlinked inodes were kept in memory on the client
5147 Details : If a client is repeatedly creating and unlinking files it
5148 can accumulate a lot of stale inodes in the inode slab cache.
5149 If there is no other client load running this can cause the
5150 client node to run out of memory. Instead flush old inodes
5151 from client cache that have the same inode number as a new inode.
5154 Frequency : SLES9 2.6.5 kernel and long filenames only
5155 Bugzilla : 9969, 10379
5156 Description: utime reports stale NFS file handle
5157 Details : SLES9 uses out-of-dentry names in some cases, which confused
5158 the lustre dentry revalidation. Change it to always use the
5162 Frequency : rare, unless heavy write-truncate concurrency is continuous
5163 Bugzilla : 4180, 6984, 7171, 9963, 9331
5164 Description: OST becomes very slow and/or deadlocked during object unlink
5165 Details : filter_destroy() was holding onto the parent directory lock
5166 while truncating+unlinking objects. For very large objects this
5167 may block other threads for a long time and slow overall OST
5168 responsiveness. It may also be possible to get a lock ordering
5169 deadlock in this case, or run out of journal credits because of
5170 the combined truncate+unlink. Solution is to do object truncate
5171 first in one transaction without parent lock, and then do the
5172 final unlink in a new transaction with the parent lock. This
5173 reduces the lock hold time dramatically.
5176 Frequency : rare, 2.4 kernels only
5178 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5179 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5180 before unmounting their filesystem in order to avoid deadlock.
5181 The kernel_locked() function in 2.4 kernels only checks whether
5182 the kernel lock is held, not whether it is this process that is
5183 holding it as 2.6 kernels do.
5188 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5189 Details : The client ptlrpc code may be trying to reconnect to a down
5190 server before a previous connection attempt has timed out.
5191 Increase the reconnect interval to be longer than the connection
5192 timeout interval to avoid sending duplicate connections to
5196 Frequency : echo_client brw_test command
5198 Description: fix echo_client to work with OST preallocated code
5199 Details : OST preallocation code (5137) didn't take echo_client IO path
5200 into account: echo_client calls filter methods outside of any
5201 OST thread and, hence, there is no per-thread preallocated
5202 pages and buffers to use. Solution: hijack pga pages for IO. As
5203 a byproduct, this avoids unnecessary data copying.
5207 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5208 Description: Client can oops in mdc_commit_close() after open replay
5209 Details : It was possible for the MDS to return an open request with no
5210 transaction number in mds_finish_transno() if the client was
5211 evicted, but without actually returning an error. Clients
5212 would later try to replay that open and may trip an assertion
5213 Simplify the client close codepath, and always return an error
5214 from the MDS in case the open is not successful.
5217 Frequency : rare, 2.6 OSTs only
5219 Description: OST may deadlock under high load on fragmented files
5220 Details : If there was a heavy load and highly-fragmented OST filesystems
5221 it was possible to have all the OST threads deadlock waiting on
5222 allocation of biovecs, because the biovecs were not released
5223 until the entire RPC IO was completed. Instead, release biovecs
5224 as soon as they are complete to ensure forward IO progress.
5226 Severity : enhancement
5228 Description: Support for specifying external journal device at mount
5229 Details : If an OST or MDS device is formatted with an external journal
5230 device, this device major/minor is stored in the ext3 superblock
5231 and may not be valid for failover. Allow detecting and
5232 specifying the external journal at mount time.
5237 Description: Mounting an MDS with pending unlinked files may cause oops
5238 Details : target_finish_recovery() calls mds_postrecov() which returned
5239 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5240 considers this an error and immediately begins cleaning up the
5241 lov, just after starting the mds_lov process
5243 Severity : enhancement
5245 Description: Implement 'lfs df' to report actual free space on per-OST basis
5246 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5247 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5248 usage of objects; '-h' to report in human readable format.
5250 ------------------------------------------------------------------------------
5252 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5259 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5260 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5261 at startup during orphan unlinking. Instead, skip buddy update
5262 before setup, as it will be regenerated after recovery is complete.
5267 Description: If an OST is inactive, its locks might reference stale inodes.
5268 Details : lov_change_cbdata() must iterate over all namespaces, even if
5269 they are inactive to clear inode references from the lock.
5271 Severity : enhancement
5272 Frequency : occasional, if non-standard max_dirty_mb used
5274 Description: Client will block write RPCs if not enough grant
5275 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5276 then the client will block writes while waiting for another RPC
5277 to complete instead of consuming its dirty limit. With change
5278 we get improved performance when max_dirty_mb is small.
5280 Severity : enhancement
5281 Bugzilla : 3389, 6253
5282 Description: Add support for supplementary groups on the MDS.
5283 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5284 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5285 lookups for user supplementary groups into a cache.
5289 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5290 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5291 for liblustre O_EXCL is enforced by MDS.
5295 Description: readdir never returns NULL in liblustre.
5296 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5297 dentry in struct dirent.
5301 Frequency : liblustre only, depends on application IO pattern
5302 Description: liblustre clients evicted if not contacting servers
5303 Details : Don't put liblustre clients into the ping_evictor list, so
5304 they will not be evicted by the pinger ever.
5306 Severity : enhancement
5308 Description: Add ability to evict clients by NID from MDS.
5309 Details : By echoing "nid:$NID" string into
5310 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5311 $NID would be instantly evicted from this MDS and from all active
5312 OSTs connected to it.
5316 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5317 Details : lookup_it_finish() used to query file size from OSTs that was not
5322 Description: service threads change working directory to that of init
5323 Details : Starting lustre service threads may pin the working directory
5324 of the parent thread, making that filesystem busy. Threads
5325 now change to the working directory of init to avoid this.
5329 Frequency : during shutdown only
5330 Description: shutdown with a failed MDS or OST can cause unmount to hang
5331 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5336 Frequency : 2.6 only
5337 Description: chmod/chown may include an extra supplementary group
5338 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5339 supplementary group and if none is specified this is used.
5342 Bugzilla : 5479 (6816)
5344 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5345 Details : If lookup is in progress on a file that is unlinked we might try
5346 to revalidate the inode and fail in revalidate after lookup is
5347 complete and ll_file_open() enqueues the open again but
5348 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5351 Frequency : always, if lconf --abort_recovery used
5353 Description: lconf --abort_recovery fails with 'Operation not supported'
5354 Details : lconf was attempting to abort recovery on the MDT device and not
5357 ------------------------------------------------------------------------------
5359 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5364 Frequency : rare (only unsupported configurations with a node running as an
5366 Bugzilla : 6514, 5137
5367 Description: Mounting a Lustre file system on a node running as an OST could
5369 Details : OSTs now preallocates memory needed to write out data at
5370 startup, instead of when needed, to avoid having to
5371 allocate memory in possibly low memory situations.
5372 Specifically, if the file system is mounted on on OST,
5373 memory pressure could force it to try to write out data,
5374 which it needed to allocate memory to do. Due to the low
5375 memory, it would be unable to do so and the node would
5376 become unresponsive.
5378 Severity : enhancement
5380 Description: Addition of lconf --service command line option
5381 Details : lconf now accepts a '--service <arg>' option, which is
5382 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5384 Severity : enhancement
5386 Description: Failover mode is now the default for OSTs.
5387 Details : By default, OSTs will now run in failover mode. To return to
5388 the old behaviour, add '--failout' to the lmc line for OSTs.
5390 Severity : enhancement
5392 Description: Health checks are now provided for MDS and OSTs
5393 Details : Additional detailed health check information on MSD and OSTs
5394 is now provided through the procfs health_check value.
5397 Frequency : occasional, depends on IO load
5399 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5400 after numerous create/delete cycles
5401 Details : The ext3 inode allocation policy would not allocate new inodes
5402 very well on the OSTs because there are no new directories
5403 being created. Instead we look for groups with free space if
5404 the parent directories are nearly full.
5409 Description: Network or server problems during mount may cause partially
5410 mounted clients instead of returning an error.
5411 Details : The config llog parsing code may overwrite the error return
5412 code during mount error handling, returning success instead
5418 Description: MDS can fail to allocate large reply buffers
5419 Details : After long uptimes the MDS can fail to allocate large reply
5420 buffers (e.g. zconf client mount config records) due to memory
5421 fragmentation or consumption by the buffer cache. Preallocate
5422 some large reply buffers so that these replies can be sent even
5423 under memory pressure.
5427 Frequency : rare (liblustre)
5428 Description: fsx running with liblustre complained that using truncate() to
5429 extend the file doesn't work. This patch corrects that issue.
5430 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5431 ATTR_SIZE and lsm use in llu_setattr_raw.
5435 Frequency : rare, only 2.6 kernels
5436 Description: Unusual file access patterns on the MDS may result in inode
5437 data being lost in very rare circumstances.
5438 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5439 code in the 2.6 kernel caused the inode and/or EA data not to
5440 be read from disk, causing single-file corruption.
5444 Frequency : rare, only 2.6 filesystems using extents
5445 Description: Heavy concurrent write and delete load may cause data corruption.
5446 Details : It was possible under high-load situations to have an extent
5447 metadata block in the block device cache from a just-unlinked
5448 file overwrite a newly-allocated data block. We now unmap any
5449 metadata buffers that alias just-allocated data blocks.
5453 Frequency : filesystems with default stripe_count larger than 77
5454 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5455 Details : lconf specifies an inode size of 4096 bytes when the default
5456 stripe_count is larger than 77. This conflicts with the default
5457 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5458 this case to avoid pinning too much memory for large EAs.
5460 ------------------------------------------------------------------------------
5462 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5467 Frequency : rare (extremely heavy IO load with hundreds of clients)
5469 Description: Client is evicted, gets IO error writing to file
5470 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5471 caused clients to be evicted for AST timeouts. The fixes in
5472 bug 5192 mean we no longer need to have such short AST timeouts
5473 so ldlm_timeout has been increased.
5476 Frequency : occasional during --force or --failover shutdown under load
5477 Bugzilla : 5949, 4834
5478 Description: Server oops/LBUG if stopped with --force or --failover under load
5479 Details : a collection of import/export refcount and cleanup ordering
5480 issues fixed for safer force cleanup
5483 Frequency : only filesystems larger than 120 OSTs
5484 Bugzilla : 5990, 6223
5485 Description: lfs getstripe would oops on a very large filesystem
5486 Details : lov_getconfig used kfree on vmalloc'd memory
5489 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5490 Bugzilla : 6242, 6243
5491 Description: reading from files that had been truncated to a non-zero size
5492 but never opened returned no data
5493 Details : ll_file_read() reads zeros from no-object files to EOF
5498 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5499 Details : There's little harm in aborting MDS/OSS recovery and letting it
5500 try again, so I removed the LASSERT and return an error instead.
5502 Severity : enhancement
5504 Description: New debugging infrastructure for tracking down data corruption
5505 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5506 (b) cover more of the client-side code path, and (c) try to narrow
5507 down where problems occurred
5511 Bugzilla : 3819, 4364, 4397, 6313
5512 Description: Racing close and eviction MDS could cause assertion in mds_close
5513 Details : It was possible to get multiple mfd references during close and
5514 client eviction, leading to one thread referencing a freed mfd.
5516 Severity: : enhancement
5517 Bugzilla : 3262, 6359
5518 Description: Attempts to reconnect to servers are now more aggressive.
5519 Details : This builds on the enhanced upcall-less recovery that was added
5520 in 1.4.2. When trying to reconnect to servers, clients will
5521 now try each server in the failover group every 10 seconds. By
5522 default, clients would previously try one server every 25 seconds.
5527 Description: After recovery, certain operations trigger a failed
5528 assertion on a client.
5529 Details : Failing over an mds, using lconf -d --failover, while a
5530 client was doing a readdir() call would cause the client to
5531 LBUG after recovery completed and the readdir() was resent.
5533 Severity : enhancement
5535 Description: Default groups are now added by lconf
5536 Details : You can now run lconf --group <servicename> without having to
5537 manually add groups with lmc.
5540 Frequency : occasional
5542 Description: Nodes with an elan id of 0 trigger a failed assertion
5545 Frequency : always when accessing e.g. tty/console device nodes
5547 Description: tty and some other devices nodes cannot be used on lustre
5548 Details : file's private_data field is used by device data and lustre
5549 values in there got lost. New field was added to struct file to
5550 store fs-specific private data.
5553 Frequency : when exporting Lustre via NFS
5555 Description: NFSD failed occasionally when looking up a path component
5556 Details : NFSD is looking up ".." which was broken in ext3 directories
5557 that had grown large enough to become hashed.
5560 Frequency : Clusters with multiple interfaces not on the same subnet
5562 Description: Nodes will repeatedly try to reconnect to an interface which it
5563 cannot reach and report an error to the log.
5564 Details : Extra peer list entries will be created by lconf with some peers
5565 unreachable. lconf now validates the peer before adding it.
5568 Frequency : Only if a default stripe is set on the filesystem root.
5570 Description: Setting a default stripe on the filesystem root prevented the
5571 filesystem from being remounted.
5572 Details : The client was sending extra request flags in the root getattr
5573 request and did not allocate a reply buffer for the dir EA.
5576 Frequency : occasional, higher if lots of files are accessed by one client
5577 Bugzilla : 6159, 6097
5578 Description: Client trips assertion regarding lsm mismatch/magic
5579 Details : While revalidating inodes the VFS looks up inodes with ifind()
5580 and in rare cases can find an inode that is being freed.
5581 The ll_test_inode() code will free the lsm during ifind()
5582 when it finds an existing inode and then the VFS later attaches
5583 this free lsm to a new inode.
5587 Bugzilla : 6422, 7030
5588 Description: MDS deadlock between mkdir and client eviction
5589 Details : Creating a new file via mkdir or mknod (starting a transaction
5590 and getting the ns lock) can deadlock with client eviction
5591 (gets ns lock and trying to finish a synchronous transaction).
5594 Frequency : occasional
5595 Description: While starting a server, the fsfilt_ext3 module could not be
5597 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5598 kernels. Previously, lconf would still use the ext3 name
5599 when trying to load modules. Now, it will correctly use
5600 ext3 on 2.4 and ldiskfs on 2.6.
5602 Severity : enhancement
5603 Description: The default stripe count has been changed to 1
5604 Details : The interpretation of the default stripe count (0, to lfs
5605 or lmc) has been changed to mean striping across a single
5606 OST, rather than all available. For general usage we have
5607 found a stripe count of 1 or 2 works best.
5609 Severity : enhancement
5610 Description: Add support for compiling against Cray portals.
5611 Details : Conditional compiling for some areas that are different
5615 Frequency : occasional
5616 Bugzilla : 6409, 6834
5617 Description: Creating files with an explicit stripe count may lead to
5618 a failed assertion on the MDS
5619 Details : If some OSTs are full or unavailable, creating files may
5620 trigger a failed assertion on the MDS. Now, Lustre will
5621 try to use other servers or return an error to the
5625 Frequency : occasional
5627 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5628 caused lock timeout during readahead (since 1.4.2).
5629 Details : Processes doing readahead might match a lock that hasn't been
5630 granted yet if there are overlapping and conflicting lock
5631 requests. The readahead process waits on ungranted lock
5632 (original lock is CBPENDING), while OST waits for that process
5633 to cancel CBPENDING read lock and eventually evicts client.
5635 Severity : enhancement
5637 Description: Initial enabling of flock support for clients
5638 Details : Implements fcntl advisory locking and file status functions.
5639 This feature is provided as an optional mount flag (default
5640 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5641 locking are implemented yet, and those that are are not guaranteed
5642 to be completely correct in production environments.
5643 mount -t lustre -o [flock|noflock] ...
5646 Frequency : occasional
5648 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5649 assertion in the ext3 JBD (journaling) layer.
5650 Details : The b_committed_data struct is protected by the big kernel lock
5651 in 2.4 kernels, serializing journal_commit_transaction() and
5652 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5653 access to this struct. In 2.6 kernels there is finer grained
5654 locking to improve SMP performance of the JBD layer.
5658 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5663 Description: merge and cleanup kernel patches.
5666 -----------------------------------------------------------------------------
5668 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5670 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5671 versions of Lustre. Please update all servers and clients to
5672 version 1.4.2 or later at the same time. You must also run
5673 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5674 to update the configuration logs.
5676 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5677 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5678 - swab llog records of type '0' so we get proper header size/idx (5861)
5679 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5680 - fix rename of one directory over another leaking an inode (5953)
5681 - avoid SetPageDirty on 2.6 (5981)
5682 - don't re-add just-being-destroyed locks to the waiting list (5653)
5683 - when creating new directories, inherit the parent's custom
5684 striping settings if present parent (3048)
5685 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5686 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5687 - don't hold client locks on temporary worklist from l_lru (5666)
5688 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5689 - automatically evict dead clients (5921)
5690 - Update file size properly in create+truncate+fstat case (6196)
5691 - Do not unhash mountpoint dentries, do not allow removal of
5693 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5694 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5695 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5696 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5697 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5698 - Correctly handle reads of files with no objects (6243)
5699 - lctl recover will also mark a device active if deactivate used (5933)
5701 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5702 - allow --write-conf on an MDS with different nettype than client (5619)
5703 - don't write config llogs to MDS for mounts not from that MDS (5617)
5704 - lconf should create multiple TCP connections from a client (5201)
5705 - init scripts are now turned off by default; run chkconfig --on
5706 lustre and chkconfig --on lustrefs to use them
5707 - upcalls are no longer needed for clients to recover to failover
5709 - add --abort-recovery option to lconf to abort recovery on device
5711 - add support for an arbitrary number of OSTs (3026)
5712 - Quota support protocol changes.
5713 - forward compatibility changes to wire structs (6007)
5714 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5715 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5716 - improved "lustre status" script
5717 - initialize blocksize for non-regular files (6062)
5718 - added --disable-server and --disable-client configure options (5782)
5719 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5720 - Vanilla 2.4.29 support
5721 - increase maximum number of obd devices to 520 (6242)
5722 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5723 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5724 SLES 9 2.6.5 kernels
5725 - limit stripes per file to 160 (the maximum EA size) (6093)
5727 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5730 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5731 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5732 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5733 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5734 - small ext3 extents cleanups and fixes (5733)
5735 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5736 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5737 - avoid needless client->OST connect, fix handle mismatch (5317)
5738 - fix DLM error path that led to out-of-sync client, long delays (5779)
5739 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5740 - fix several locking issues related to i_size (5492,5624,5654,5672)
5741 - don't move pending lock onto export if it is already evicted (5683)
5742 - fix kernel oops when creating .foo in unlinked directory (5548)
5743 - fix deadlock in obdfilter statistics vs. object create (5811)
5744 - use time_{before,after} to avoid timer jiffies wrap (5882)
5745 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5746 - Do not leak request if server was not able to process it (5154)
5747 - If mds_open unable to find parent dir, make that negative lookup(5154)
5748 - don't create new directories with extent-mapping (5909, 5936)
5750 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5751 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5752 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5753 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5754 - support for RHEL4 kernel on i686 (5773)
5755 - provide error messages when incompatible logs are encountered (5898)
5757 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5758 * version 1.4.0.10 (1.4.1 release candidate 1)
5760 - don't keep a lock reference when lock is not granted (4238)
5761 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5762 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5763 - drop import inflight refcount on signal_completed_replay error (5255)
5764 - unlock page after async write error during send (3677)
5765 - handle missing objects in filter_preprw_read properly (5265)
5766 - no transno return for symlink open, don't save no-trasno open (3440)
5767 - don't try to complete elan receive that already failed (4012)
5768 - free RPC server reply state on error (5406)
5769 - clean up thread from ptlrpc_start_thread() on error (5160)
5770 - readahead could read extra page into cache that wasn't ejected (5388)
5771 - prevent races in class_attach/setup/cleanup/detach (5260)
5772 - don't dereference de->d_inode after l_dput of de (5458)
5773 - use "int" for stripe value returned from lock_to_stripe (5544)
5774 - mballoc allocation and error-checking fixes in 2.6 (5504)
5775 - block device patches to fix I/O request sizes in 2.6 (5482)
5776 - look up hostnames for IB nals (5602)
5777 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5778 - don't start multiple acceptors for the same port (5277)
5779 - fix incorrect LASSERT in mds_getattr_name (5635)
5780 - export a proc file for general "ping" checking (5628)
5781 - fix "lfs check" to not block when the MDS is down (5628)
5783 - service request history (4965)
5784 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5785 - create an "evict_client" /proc entry on OSTs, like the MDS has
5786 - fix mount usage message, return errors per mount(8) (5168)
5787 - change grep [] to grep "[]" in tests so they work in more UMLs
5788 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5789 - remove ancient LOV_MAGIC_V0 EA support (5047)
5790 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5791 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5792 - allow readahead allocations to fail when low on memory (5383)
5793 - mmap locking landed again, after considerable improvement (2828)
5794 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5796 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5799 - send OST transaction number in read/write reply to free req (4966)
5800 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5801 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5802 - the watchdog thread now runs as interruptible (5246)
5803 - flock/lockf fixes (but it's still disabled, pending 5135)
5804 - don't use EXT3 constants in llite code (5094)
5805 - memory shortage at startup could cause assertion (5176)
5807 - reorganization of lov code
5808 - single portals codebase
5810 - add extents/mballoc support (5025)
5811 - direct I/O reads in the obdfilter (4048)
5812 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5814 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5817 - send OST transaction number in read/write reply to free req (4966)
5818 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5819 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5820 - the watchdog thread now runs as interruptible (5246)
5821 - handle missing objects in filter_preprw_read properly (5265)
5822 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5823 - drop import inflight refcount on signal_completed_replay error (5255)
5824 - unlock page after async write error during send (3677)
5825 - return original error code on reconstructed replies (3761)
5826 - no transno return for symlink open, don't save no-trasno open (3440)
5828 - add pid to ldlm debugging output (4922)
5829 - bump the watchdog timeouts -- we can't handle 30sec yet
5830 - extra debugging for orphan dentry/inode bug (5259)
5832 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5835 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5836 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5837 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5838 - fix NULL dereference in /proc/sys/portals/routes (4827)
5839 - allow failed mdc_close() operations to be interrupted (4561)
5840 - stop precreate on OST before MDS would time out on it (4778)
5841 - don't send partial-page writes before EOF from client (4410)
5842 - discard client grant for sub-page writes on large-page clients (4520)
5843 - don't free dentries not owned by NFS code, check generation (4806)
5844 - fix lsm leak if mds_create_objects() fails (4801)
5845 - limit debug_daemon file size, always print CERROR messages (4789)
5846 - use transno after validating reply (3892)
5847 - process timed out requests if import state changes (3754)
5848 - update mtime on OST during writes, return in glimpse (4829)
5849 - add mkfsoptions to LDAP (4679)
5850 - use ->max_readahead method instead of zapping global ra (5039)
5851 - don't interrupt __l_wait_event() during strace
5853 - add software watchdogs to catch hung threads quickly (4941)
5854 - make lustrefs init script start after nfs is mounted
5855 - fix CWARN/ERROR duplication (4930)
5856 - return async write errors to application if possible (2248)
5857 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5858 - print NAL number in %x format (4645)
5859 - update barely-supported suse-2.4.21-171 series (4842)
5860 - support for sles 9 %post scripts
5861 - support for building 2.6 kernel-source packages
5862 - support for sles km_* packages
5864 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5867 - ignore -ENOENT errors in osc_destroy (3639)
5868 - notify osc create thread that OSC is being cleaned up (4600)
5869 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5870 - reconstruct ost_handle() like mds_handle() (4657)
5871 - create a new thread to do import eviction to avoid deadlock (3969)
5872 - let lconf resolve symlinked-to devices (4629)
5873 - don't unlink "objects" from directory with default EA (4554)
5874 - hold socknal file ref over connect in case target is down (4394)
5875 - allow more than 32000 subdirectories in a single directory (3244)
5876 - fix blocks count for O_DIRECT writes (3751)
5877 - OST returns ENOSPC from object create when no space left (4539)
5878 - don't send truncate RPC if file size isn't changing (4410)
5879 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5880 - bind to privileged port in socknal and tcpnal (3689)
5882 - rate limit CERROR/CWARN console message to avoid overload (4519)
5883 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5884 - basic mmap support (3918)
5885 - kernel patch series update from b1_4 (4711)
5887 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5890 - avoid crash during MDS cleanup with OST shut down (2775)
5891 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5892 - don't use bad inodes on the MDS (3744)
5893 - dynamic object preallocation to improve recovery speed (4236)
5894 - don't hold spinlock over lock dumping or change debug flags (4401)
5895 - don't zero obd_dev when it is force cleaned (3651)
5896 - print grants to console if they go negative (4431)
5897 - "lctl deactivate" will stop automatic recovery attempts (3406)
5898 - look for existing locks in ldlm_handle_enqueue() (3764)
5899 - don't resolve lock handle twice in recovery avoiding race (4401)
5900 - revalidate should check working dir is a directory (4134)
5902 - don't always mark "slow" obdfilter messages as errors (4418)
5904 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5907 - don't close LustreDB during write_conf until it is done (3860)
5908 - fix typo in lconf for_each_profile (3821)
5909 - allow dumping logs from multiple threads at one time (3820)
5910 - don't allow multiple threads in OSC recovery (3812)
5911 - fix debug_size parameters (3864)
5912 - fix mds_postrecov to initialize import for llog ctxt (3121)
5913 - replace config semaphore with spinlock (3306)
5914 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5915 - don't allow enqueue to complete on a destroyed export (3822)
5916 - down write_lock before checking llog header bitmap (3825)
5917 - recover from lock replay timeout (3764)
5918 - up llog sem before sending rpc (3652)
5919 - reduce ns lock hold times when setting kms (3267)
5920 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5921 - fix NULL deref and obd_dev leak on setup error (3312)
5922 - replace some LBUG about llog ops with error handling (3841)
5923 - don't match INVALID dentries from d_lookup and spin (3784)
5924 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5925 - fix invalid assertion in ptlrpc_set_wait (3880)
5927 - add libwrap support for the TCP acceptor (3996)
5928 - add /proc/sys/portals/routes for non-root route listing (3994)
5929 - allow setting MDS UUID in .xml (2580)
5930 - print the stack of a process that LBUGs (4228)
5932 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5935 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5936 - make sure to unset replay flag from failed open requests (3440)
5937 - if default stripe count is 0, use OST count for inode size (3636)
5938 - update parent mtime/ctime on client for create/unlink (2611)
5939 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5940 - free recovery state on server during a forced cleanup (3571)
5941 - unregister_reply for resent reqs (3063)
5942 - loop back devices mounting and status check on 2.6 (3563)
5943 - fix resource-creation race that can provoke i_size == 0 (3513)
5944 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5945 - more debugging for page-accounting assertion (3746)
5946 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5947 - avoid deadlock after precreation failure (3758)
5948 - fix race and lock order deadlock in orphan handling (3450, 3750)
5949 - add validity checks when grabbing inodes from l_ast_data (3599)
5951 - add /proc/.../recovery_status to obdfilter (3428)
5952 - lightweight CDEBUG infrastructure, debug daemon (3668)
5953 - change default OSC RPC parameters to be better on small clusters
5954 - turn off OST read cache for files smaller than 32MB
5955 - install man pages and include them in rpms (3100)
5956 - add new init script for (un)mounting lustre filesystems (2593)
5957 - run chkconfig in %post for init scripts (3701)
5958 - drop scimac NAL (unmaintained)
5960 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5963 - clean kiobufs before and after use (3485)
5964 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5965 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5966 - change init script to fail more clearly if not run as root (1528)
5967 - allow clients to reconnect during replay (1742)
5968 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5969 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5970 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5971 - fix cond_resched (3554)
5972 - only evict unfinished clients after recovery (3515)
5973 - allow bulk resend, prevent data loss (3570)
5974 - dynamic ptlrpc request buffer allocation (2102)
5975 - don't allow unlinking open directory if it isn't empty (2904)
5976 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5977 - remove extraneous obd dereference causing LASSERT failure (3334)
5978 - don't use get_cycles() when creating temp. files on the mds (3156)
5979 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5980 - handle EEXIST for set-stripe, set proper directory name (3336)
5982 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5983 - fix ksocknal_fmb_callback() error messages (2918)
5985 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5988 - don't copy lvb into (possibly NULL) reply on error (2983)
5989 - don't deref dentry after dput, don't free lvb on error (2922)
5990 - use the kms to determine writeback rpc length (2947)
5991 - increment oti_logcookies when osc is inactive (2948)
5992 - update client's i_blocks count via lvb messages (2543)
5993 - handle intent open/close of special files properly (1557)
5994 - mount MDS with errors=remount-ro, like obdfilter (2009)
5995 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5996 - don't use cancelling-locks' kms values (2947)
5997 - use highest lock extent for kms, not last one (2925)
5998 - don't dereference ERR_PTR() dentry in error handling path (3107)
5999 - fix thread race in portals_debug_dumplog() (3122)
6000 - create lprocfs device entries at setup instead of at attach (1519)
6001 - common AST error handler, don't evict client on completion race (3145)
6002 - zero nameidata in detach_mnt in 2.6 (3118)
6003 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6004 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6005 - initialize RPC timeout timer earlier for 2.6 (3219)
6006 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6007 - print nal/nid for unknown nid (3258)
6008 - additional checks for oscc recovery before doing precreate (3284)
6009 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6010 - don't crash in mdc_close for bad permissions on open (3285)
6011 - zero i_rdev for non-device files (3147)
6012 - clear page->private before handing to FS, better assertion (3119)
6013 - tune the read pipeline (3236)
6014 - fix incorrect decref of invalidated dentry (2350)
6015 - provide read-ahead stats and refine rpc in flight stats (3328)
6016 - don't hold journal transaction open across create RPC (3313)
6017 - update atime on MDS at close time (3265)
6018 - close LDAP connection when recovering to avoid server load (3315)
6019 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6020 - don't leak open file on MDS after open resend (3325)
6021 - serialize filter_precreate and filter_destroy_precreated (3329)
6022 - loop device shouldn't call sync_dev() for nul device (3092)
6023 - clear page cache after eviction (2766)
6024 - resynchronize MDS->OST in background (2824)
6025 - refuse to mount the same filesystem twice on same mountpoint (3394)
6026 - allow llmount to create routes for mounting behind routers (3320)
6027 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6028 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6029 - fix rare problem with rename on htree directories (3417)
6031 - allow default OST striping configuration per directory (1414)
6032 - fix compilation for qswnal for 2.6 kernels (3125)
6033 - increase maximum number of MDS request buffers for large systems
6034 - change liblustreapi to be useful for external progs like lfsck (3098)
6035 - increase local configuration timeout for slow disks (3353)
6036 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6038 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6041 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6042 - don't overwrite extent policy data in reply if lock was blocked (2901)
6043 - drop filter export grants atomically with removal from device (2663)
6044 - del obd_self_export from work_list in class_disconnect_exports (2908)
6045 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6046 - swab reply message in mdc_close, other PPC fixes (2464)
6047 - fix destroying of named logs (2325)
6048 - overwrite old logs when running lconf --write_conf (2264)
6049 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6050 - fix race in target_handle_connect (2898)
6051 - mds_reint_create() should take same inode create lock (2926)
6052 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6053 - don't close files for self_export to avoid uninitialized obd (2936)
6054 - allow MDS with the same name as client node (2939)
6055 - hold dentry reference for closed log files for unlink (2325)
6056 - reserve space for all logs during transactions (2059)
6057 - don't evict page beyond end of stripe extent (2925)
6058 - don't oops on a deleted current working directory (2399)
6059 - handle hard links to targets without a parent properly (2517)
6060 - don't dereference NULL lock when racing during eviction (2867)
6061 - don't grow lock extents when lots of conflicting locks (2919)
6063 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6066 - account for cache space usage on clients to avoid data loss (974)
6067 - lfsck support in lustre kernel code (2349)
6068 - reduce journal credits needed for BRW writes (2370)
6069 - orphan handling to avoid losing space on client/server crashes
6070 - ptlrpcd can be blocked, stopping ALL progress (2477)
6071 - use lock value blocks to assist in proper KMS, faster stat (1021)
6072 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6073 - recovery for initial connections (2355)
6074 - fixes for mds_cleanup_orphans (1934)
6075 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6076 - block all file creations until orphan recovery completes (1901)
6077 - client remove rq_connection from request struct (2423)
6078 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6079 - recovery timer race (2670)
6080 - mdc_close recovey bug (2532)
6081 - ptlrpc cleanup bug (2710)
6082 - mds timeout on local locks (2588)
6083 - namespace lock held during RPCs (2431)
6084 - handle interrupted sync write properly (2503)
6085 - don't try to handle a message that hasn't been replied to (2699)
6086 - client assert failure during cleanup after abort recovery (2701)
6087 - leak mdc device after failed mount (2712)
6088 - ptlrpc_check_set allows timedout requests to complete (2714)
6089 - wait for inflight reqs when ptlrpcd finishes (2710)
6090 - make sure unregistered services are removed from the srv_list
6091 - reset bulk XID's when resending them (caught by 1138 test)
6092 - unregister_bulk after timeout
6093 - fix lconf error (2694)
6094 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6095 - readahead locks pages, leaves pending causing memory pressure (2673)
6096 - increase OST request buffers to 4096 on large machines (2729)
6097 - fix up permission of existing directories in simple_mkdir (2661)
6098 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6099 - don't assign transno to errored transactions (2742)
6100 - don't delete objects on OST if given a bogus objid from MDS (2751)
6101 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6102 - if rq_no_resend, then timeout request after recovery (2432)
6103 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6104 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6105 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6106 - MDS recovery shouldn't delete objects that it has given out (2730)
6107 - if enqueue arrives after completion, don't clobber LVB (2819)
6108 - don't unlock pages twice when trigger_group_io returns error (2814)
6109 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6110 - don't write pages to disk if there was an error (1450)
6111 - don't ping imports that have recovery disabled (2676)
6112 - take buffered bytes into account when balancing socknal conn (2817)
6113 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6114 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6115 - remove little-endian swabbing of llog records (1987)
6116 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6117 - retry reposting request buffers if they fail (1191)
6118 - grow extent at grant time to avoid granting a revoked lock (2809)
6119 - lock revoke doesn't evict page if covered by a second lock (2765)
6120 - disable VM readahead to avoid reading outside lock extents (2805)
6122 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6123 - updated kernel patches for hp-2.4.20 kernel (2681)
6125 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6128 - fix truncated write corruption (2366)
6129 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6131 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6132 - ensure len(uuid) < 37 in lmc (1171)
6133 - fix ia64 OOPS in llog_test (2255)
6134 - zero end of page at obdfilter for partial page writes (2648)
6135 - don't leave stale dentries around after renames (bug 2428)
6136 - fix timeouts when evicting a client with a single lock held (2642)
6137 - set deadline for the initial HELLO message to drain (2634)
6138 - print out dotted-quad IP addresses in the socknal (2302)
6140 - additional debugging for MDS client eviction problem (2443)
6141 - fix mkfsoptions support for osts (2603, 2604)
6143 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6146 - add series for the vanilla 2.6.0 kernel
6147 - add series for the vanilla 2.4.24 kernel
6148 - add series for a cray x86/64 UL kernel drop
6149 - fix xattr patches for the vanilla 2.4.19 series
6151 - generate true UUIDs in lmc (1171)
6152 - have portals stack dumping break in UML (2466)
6153 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6154 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6155 - wake callback waiting threads on client eviction (2460)
6156 - Add --ptldebug and --subsystem to lmc (1719)
6157 - update assertion to allow safe interrupt allocation
6158 - set rq_no_resend for cancel requests (2432)
6159 - recalculate ptlrpcd timeout after resend (2494)
6160 - call vfs_rmdir when removing pending directories (2368)
6161 - fix renaming a file to itself (2429)
6162 - lmc creates a default one-stripe lov (2454)
6163 - expand procfs space to handle large clusters (2326)
6164 - increase UML stack to avoid overflow
6165 - update lconf's list of debug and subsystem masks
6166 - fix lfs find --obd (2510)
6167 - /proc tunable for disabling filter read caching (2591)
6168 - stop rpm packages from altering slapd.conf (2301)
6169 - disable nagle in the socknal under 0conf (2578)
6170 - choose mds inode size based on stripe count (2572)
6171 - fix kernel-source rpm problems (2516)
6173 - add --disable-doc to avoid pdf generation (2421)
6174 - update documentation, tests, type-os, comments
6175 - avoid format warnings on ia64
6176 - remove the TOE NAL
6177 - tiny code cleanups by removing unused fields
6179 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6182 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6183 - remove the most blatant lies from BUILDING file (2371)
6184 - change default debug level to reasonable production setting
6185 - reduce client side cache size to reduce cache flush time
6186 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6187 - make TCP zerocopy and pinger support enabled by default (2476)
6188 - sync writes completed after process exits caused crashes (2319)
6189 - maintain correct mount count on the MDS (2356)
6190 - backout 1557, because 2316 wasn't really fixed
6191 - better file I/O statistics gathering in /proc
6192 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6193 - another kernel patch to fix zero-copy TCP function export
6194 - don't take duplicate lock when processing re-sent getattr (2420)
6195 - lctl uses obd_self_export instead of creating new conn (2353)
6196 - MDS/OST recovery case which requires object creation asserted (2425)
6197 - move lfs from /usr/sbin to /usr/bin in packages
6198 - fix race between mds_client_add and mds_client_free (2417)
6199 - use kmalloc instead of slabs in portals (2430)
6200 - don't create duplicate records when a failover MDS is present (2442)
6201 - remove unnecessary mount age check (2332)
6202 - don't remove directory inodes from locks prematurely (2451)
6203 - don't break if MDS service name is the same as hostname (2103)
6204 - fix races in client write RPC generation when cache full (2482)
6206 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6209 - remove now-unused request->rq_obd (278)
6210 - if an allocation fails, print out how much memory we've used (1933)
6211 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6212 - add the "configurable stack size" patch to most series files (1256)
6213 - ability to write large log records, for 100+ OST configs (2306)
6214 - fix NULL deref when filter_prep fails (2314)
6215 - fix operator precedence error in filter_sync
6216 - dynamic allocation of socknal TX descriptors (2315)
6217 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6218 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6219 - ensure that CWARN messages in Portals always get to the syslog
6220 - __init/__exit are not for prototype decls (ldlm_init/exit)
6221 - x86-64 compile warning fixes
6222 - fix gateway LMC keyword conflict (2318)
6223 - fix MDS lock inversions in getattr/reint paths (1844)
6224 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6225 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6226 - DLM race condition prevented some lock evictions (2328)
6227 - ENOMEM detection and retry on socknal sends (2230)
6228 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6229 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6230 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6231 - better error messages when a client is rejected during recovery (1505)
6232 - avoid cancelling locks which were never granted, after failure (2330)
6233 - fix i_sem/journal inversion in mds_client_add (2333)
6234 - fix truncate/getattr lock cycle deadlock (2334)
6235 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6236 - fix two rare exit paths which could leak an l_lock() ref (2321)
6237 - fencepost error in MDS/OST orphan recovery (2226)
6238 - make log record alignment 8 bytes (1988)
6239 - lstripe now fails when requested offset > ost_count (2237)
6240 - ensure that all kernel series have a complete list.h (1607)
6241 - fix crashes in special-file operations (2316)
6242 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6244 - allow configurable automake binary, for testing new versions
6245 - small update to the lfs documentation
6247 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6249 * fix negative export reference count in fsfilt_sync (2312)
6251 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6252 * release candidate 0.9.1
6254 - orphans are moved into the PENDING directory for possible recovery
6255 - replayed opens now open by fid for orphan/rename safety (1042)
6256 - last close of an orphan inode generates a transno (683)
6257 - chdir() and mount() now pin the directory entry (1020)
6258 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6259 - discard very old requests without processing them (1502)
6260 - remove some common, well-understood CERRORs (1505)
6261 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6262 - clear "grant" flags in OST replies until OST grant code lands (1644)
6263 - fix read performance by not clobbering i_blksize on client (1598)
6264 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6265 - make LRU size a /proc tunable, clears locks when reduced (707)
6266 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6267 - server locks take references on exports now (1558)
6268 - build fixes for 2.4.20-rh trees (1663)
6269 - return an error from lov_create if all OSCs are inactive (1751)
6270 - fix import levels when a reconnect happens without a timeout (1597)
6271 - exit early from mds_open if we get a lookup error (1749)
6272 - partial page read at EOF wouldn't wait for disk before sending (1642)
6273 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6274 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6275 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6276 - write/truncate lock inversion (1639)
6277 - Don't auto-load obdclass, portals modules during cleanup (1495)
6278 - fix timestamps from jumping to "now" (1763)
6279 - extra journal assertions (1648)
6280 - add an extra multiunlink test (1771)
6281 - fix read_record/write_record API (1776)
6282 - fix leak of offset_extent, possible incorrect i_size later (1772)
6283 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6284 - Debugging for the kqswnal_get_idle_tx problems (1820)
6285 - Allow recovery to be attempted multiple times (1536)
6286 - Write out MDS last_rcvd file after it is first created (1600)
6287 - Fix tx_descriptor leak in failed transmit situations (1827)
6288 - ext3 journaling fixes for assertion failure after IO error (1871)
6289 - class_export_put() on freed export after completion AST error (1896)
6290 - Fix revalidate looping in VFS (1322)
6291 - Don't access a freed export during MDS_REINT timeout (1521)
6292 - Add open-unlink recovery support on the MDS (1673,1764)
6293 - Return an error if no MDS data was read from last_rcvd (1946)
6294 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6295 - Don't setup a disk device that doesn't match exported UUID (317)
6296 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6297 - avoid committing NULL handle in force close
6298 - local.sh is now a one-stripe LOV configuration
6299 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6300 - fix ext3 htree duplicate directory entry corruption (1516)
6301 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6302 - update to new LOV EA format (2097)
6303 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6305 2003-06-15 Phil Schwan <phil@clusterfs.com>
6308 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6309 - per-import/export recovery handling (958, 931, 959)
6310 - multiple last-rcvd slots, for serving multiple FSes (949)
6311 - connections are again shared between multiple imp/exports (963, 964)
6312 - "umount -f" would hang if any requests needed to be sent (393, 978)
6313 - avoid pinning large req buffer by copying for queued messages (989)
6314 - add "uuid" to "lctl device" command to help upcalls (991)
6315 - "open" RPCs with transnos would confuse recovery counters (1037)
6316 - do proper endian conversion of all wire messages (288, 340, 891)
6317 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6318 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6319 - fix ptlrpc_connection leak in target_handle_connect (1174)
6320 - fix import refcounting bug in OST and MDS cleanup (1134)
6321 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6322 - fix very unlikely obd_types race condition (501)
6323 - remove osc_open hack for echo_client (1187)
6324 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6325 - a failure in read_inode2 leads to deadlock (1139)
6326 - cancel ack-locks as soon as transaction is committed (1072)
6327 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6328 - make sure to commitrw after any preprw to avoid deadlock (1162)
6329 - failing to execute a file in a lustre FS would lock inode (1203)
6330 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6331 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6332 - obd_uuid2tgt would walk off the end of the list (1255)
6333 - on IA64 the osc would give portals incorrect bulk size (1258)
6334 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6335 - fix lock inversion caused by new llite matching code (1282)
6336 - limit the number of dirty pages on a client to 10MB (1286)
6337 - timed out locks were not being corrected cancelled (1289)
6338 - fix O_DIRECT above 4GB on IA-32 (1292)
6339 * major user-visible changes
6340 - fail out/fail over policy now controlled by the upcall (993)
6342 - add OBD_PING to check server availability and failure (954)
6343 - lustre messages are now sent in sending host order (288, 340, 891)
6344 - add eadatalen to MDS getattr reply (340)
6345 - OST read replies may contain second buffer, with per-page status (593)
6347 2003-03-11 Phil Schwan <phil@clusterfs.com>
6350 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6351 - failed open()s could cause deadlock; fixed (867, 869)
6352 - stop cancelling OST locks when files are closed (481)
6353 - overlapping XID spaces caused network corruption (851, 853)
6354 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6355 - setattr_raw added, to do single-RPC, server-side setattrs
6356 - lmc/lconf syntax change for OST UUIDs
6357 - fix crashy race condition between ptlrpc_free_req and osc_close
6358 - don't use request in mdc_enqueue if we hit a timeout (889)
6359 - don't set the inode i_size for regular files from the MDS (896)
6360 - handle out of order completion AST (842)
6361 - don't LBUG if a lock request times out after receiving AST (913)
6362 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6363 - if a bad lock AST arrives, send an error instead of dropping entirely
6364 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6365 - fix leak in bulk IO when only partially completed (899, 900, 926)
6366 - fix O_DIRECT for ia64 (55)
6367 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6368 - C-z after timeout could hang a process forever; fixed (977)
6370 - client-side I/O cache (678, 924, 929, 941, 970)
6372 - READPAGE and SETATTRs which don't take server-side locks get
6375 2003-02-11 Phil Schwan <phil@clusterfs.com>
6378 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6379 - Fix fsfilt_extN_readpage() to read a full page of directory
6380 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6381 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6382 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6383 - Most intent operations were removed, in favour of a new RPC mode
6384 that does a single RPC to the server and bypasses most of the VFS
6385 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6386 - Aggressively cancel local locks on DLM servers
6387 - mds_reint_unlink sends EA to the client if it's the last nlink.
6388 client uses that EA to unlink OST objects.
6389 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6390 - recursive symlinks were fixed (439)
6391 - fixed NULL deref in DEBUG_REQ
6392 - filter_update_lastobjid no longer calls sync, which annoyed extN
6393 - fixed multi-client small-writes to a single file problem (445)
6394 - fixed mtime updates during file writes (607)
6395 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6396 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6397 - fixed Linux OST size reporting problem (444, 656)
6398 - OST now updates object mtime with writes or setattr (607, 619)
6399 - client verifies file size before zeroing page past EOF (445)
6400 - OST now writes last allocated objid to disk with allocation (108)
6401 - LOV on echo now works (409)
6403 - mds_reint_unlink sends a new buffer, with the EA included. this
6404 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6405 set if a regular file was being unlinked, and it was the last link
6406 - use PtlGet from the target for bulk writes (315)
6407 - OST now updates object mtime with writes or setattr (607, 619)
6408 - LDLM now has a grant-time callback to revalidate locked items, if
6410 - Many MDS operations were reorganized to combat race conditions
6412 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6413 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6415 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6418 - Fully reactivate OST imports after reconnection (512, others)
6419 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6420 - More graceful error handling for truncating on dead OST (515)
6421 - Don't error out unless we're actually accessing dead stripes (474)
6422 - Fix garbage sizes when stripes are missing (410)
6423 - LRU counters were broken, causing constant lock purge (433, 432)
6424 - garbage on read from stripes with failed OSTs (441)
6425 - mark OSCs as active before reconnecting during recovery (438)
6426 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6427 - lfind did not preserve OST order in output (443)
6428 - symlinks cause hung clients, incorrect data (439)
6429 - stop dereferencing request after dropping refcount (457)
6430 - don't LASSERT(spin_is_locked) on non-SMP (455)
6431 - fixes for many rename() bugs
6432 - fstat didn't correctly synchronize attributes (399)
6433 - server must handle lock cancellation during blocking AST prep (487)
6434 - bulk descriptors were free()d too soon (511)
6435 - fix paths in lconf, which would load incorrect modules (451, 507)
6436 - fix confusing lconf 'host not found' error message (386)
6437 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6438 - fix race condition in mdc_blocking_ast() for inode access (526)
6439 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6440 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6441 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6442 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6443 - fix lstripe in several ways (kernel side) (527)
6444 - fix request leak in ldlm_cli_enqueue (262)
6445 - incorrect OSC was marked inactive after OST failure
6446 - call mds_fs_cleanup before unmounting filesystem (524)
6447 - fix races between taking ns_lock and ldlm_lock_change_resource
6448 - fix races updating LOV export open file list
6449 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6450 - fix recovery NULL deref in ldlm_cli_cancel_unused
6451 - fix some DLM races by using new hash table for lock handles (419)
6452 - permit the client to specify desired inodes, at replay
6453 - duplicate requests when we queue them for replay reintegration
6454 - fix last_rcvd offset calculation
6455 - sync after each recovered transaction, so we always make progress
6456 - never, not always, ERESTART requests without transnos
6457 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6459 - skip replay if the MDS says that the client is already connected
6460 - don't check for a recovery-enabled export to match lctl's UUID
6461 - don't INC_USE_COUNT for phantom exports
6462 - don't crash when cleaning up phantom exports (567)
6463 - don't double-finish or set replay data for errored mdc_open requests
6464 - abort requests when they time out, so we don't get old replies
6465 - send/receive replies for AST messages again
6466 - if the client says that it doesn't have the lock, cancel it on the
6468 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6469 mark it as destroyed, it will all work out when decref is called
6470 - fix module use counts (22, 581)
6472 - ASTs now expect a reply (server cancels lock on error reply)
6474 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6477 - fix many simultaneous client startup (392)
6478 - fix dentry->d_it clobbering
6479 - credentials weren't being shipped for readdir/getattr operations
6480 - remove invalid assertions triggered during some concurrent MD
6482 - proper Lustre versions added (336, 389)
6483 - fix memory leak for create error case (398)
6484 - fix LOV locking bug that would get cli/srv out of sync
6485 - fix echo client over LOV (409)
6486 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6487 - fix double-O_EXCL intent crash (424)
6488 - avoid sending multiple lock CANCELs (352)
6490 - MDS can do multi-client recovery (modulo bugs in new code)
6492 - many updates, edits, cleanups
6494 2002-11-18 Phil Schwan <phil@clusterfs.com>
6497 - fix null d_it dereference (346)
6498 - fix full OST/dbench hang (333)
6499 - fix permission problem with file removal (286)
6500 - fix removal of OSCs from LOV when they fail
6501 - fix NULL deref during bulk timeout (214)
6502 - fix problems related to multiple filesystems on one MDS (241)
6503 - fixed serious subtle metadata locking bugs
6504 - free locks on clients when inodes are removed due to memory
6506 - fix inode pointer in lock data (285)
6507 - partial support for multiple MDS on a single host (241)
6508 - data locks weren't cancelled at clear_inode time (290, 311)
6509 - intent locks could lead to unbounded lock growth (205)
6510 - added a maximum lock count, an LRU list, and a flusher
6511 - fix multiple rename (365)
6512 - properly abstracted the echo client
6513 - OSC locked 1 byte too many; fixed
6514 - rewrote brw callback code:
6515 - fixed recovery bugs related to LOVs (306)
6516 - fixed too-many-pages-in-one-write crash (191)
6517 - fixed (again) crash in sync_io_timeout (214)
6518 - probably fixed callback-related race (385)
6520 - Add capability to MDS protocol
6521 - LDLM cancellations and callbacks on different portals
6523 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6526 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6527 - defer open object destruction to close time (601981, 138)
6528 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6529 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6530 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6531 - don't call obd_setattr in ll_file_release for destroyed objects
6532 * protocol change to lustre_msg: move |version| and add |flags|
6533 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6534 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6535 * added replay of create, unlink, link and rename operations during
6536 MDS failover; recovery should be much more robust now
6537 * remove failed OSCs from LOVs (only lov_create uses this so far)
6538 * the lustre-HOWTO was brought (more) up to date (582544)
6540 2002-10-23 Phil Schwan <phil@clusterfs.com>
6543 - in-use dentries weren't being reused properly (617851)
6544 - prevent multiple LDLM setup (599178)
6545 - fix LOV size calculations for truncate (617853)
6546 - fix client handling of MDS intent errors (POSIX)
6547 - fix permission bug in lovstripe.c test (624321)
6548 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6549 - truncate past end of file could corrupt data
6550 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6551 - a race in recovery could return ETIMEDOUT to apps (623947)
6552 - building outside the source directory was fixed
6553 * the lustre-HOWTO was brought (more) up to date (582544)
6554 * major progress was made on recovery functionality
6556 2002-10-10 Phil Schwan <phil@clusterfs.com>
6559 - recovery deadlock fix
6560 - rm -rf causes LBUG fix (617817)
6561 - file open by multiple tasks fix (618962)
6562 - directory permissions bugs (602707 and 620007)
6563 - journal_stop fixed with locking (611313)
6564 - O_APPEND failures resolved (618273, perhaps 614459)
6565 - lconf PATH fix (619770)
6566 - IA64 build fix (621450)
6567 - RPC buffer sizes scale with amount of memory
6569 2002-10-01 Phil Schwan <phil@clusterfs.com>
6572 - locks would be cancelled without throwing away data pages,
6573 resulting in inconsistent data (605627)
6574 - inode attributes were not always being refreshed (605627, 612449)
6575 - lconf now continues to cleanup after lctl reports an error
6576 - MDS now enforces user permissions (602707)
6577 - lprocfs cleanup fixed, but not yet enabled (614157)
6578 - fixed infinite server hang, should a client not respond to an AST
6579 - avoid going into recovery if user calls readlink() with a buffer
6580 that's too small (613941)
6581 - AST RPCs no longer require replies (614867) -- this may be changed
6582 - don't crash server if client sends an IOV that's too big (611336)
6583 - fixed lock conversion deadlock (611892)
6584 - fixed the following of symlinks (614622)
6585 * recovery: the server can remove locks from a client that dies, other
6586 clients can make progress
6587 * more extN patch fixes
6588 * compile-time configurable ptlrpc buffer allocations
6590 - collaborative read cache document
6591 - Lustre Lite Performance CDR document-in-progress
6593 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6596 - fix typo in patch-2.4.18
6598 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6601 - clear ptlrpc request each time in handle_incoming_request()
6602 - unlink of files now destroys the object on the OST
6604 2002-09-19 Peter Braam <braam@clusterfs.com>
6606 * add hard link support
6607 * change obdfile creation method
6608 * kernel patch changed
6610 2002-09-19 Peter Braam <braam@clusterfs.com>
6613 - stack overflow bug in extN fixed
6615 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6617 * documentation updates
6618 - add man pages for config tools
6619 - update tests/README to describe testing with new config tools
6620 - finish metadata API descriptions
6621 * bug fixes and cleanups
6622 - statfs workaround for 16TB limit
6623 - LOV stripe allocation improved, can stripe on subset of OSTs
6624 - LOV file size/IO offset was wrong for files > 4GB in size
6625 - object EA data was being dropped, caused files to be unreadable
6626 - memory overflow with non-LOV OST caused memory corruption
6627 - fixed regression tests to work with new config tools, obdfilter
6628 - fixed bug when directory size became larger than 1 block
6629 - fixed bug (for single client case) when PWD was deleted
6630 - invalidate local directory pages when doing intent-based ops
6631 - avoid LDLM oops when lock callback contained bad data
6633 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6635 * documentation updates
6636 * bug fixes and cleanups
6637 - configuration tools
6640 - 64-bit compile warnings
6641 - 64-bit internal statfs data
6643 * test_brw on persistent OST devices
6645 * lprocfs (disabled)
6647 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6649 * documentation updates
6650 * bug fixes and cleanups
6651 * configuration tools
6653 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6656 * many small fixes to 0_5_4
6657 * io/network handling
6658 * thinkos in MDS operations
6660 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6663 * crucial basic fixes to 0.5.3
6664 * IOR, Iozone work over Elan
6667 2002-08-07 Phil Schwan <phil@clusterfs.com>
6668 * version 0_5_3, our first alpha
6669 * we use the new Portals iovs
6670 * documentation updates
6671 * bug fixes and cleanups
6672 * small changes in the DLM wire protocol
6674 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6675 * version 0_5_1 with some initial stability,
6676 * locking on MD and file I/O.
6677 * documentation updates
6678 * several bug fixes since 0.5.0
6679 * small changes in wire protocol
6681 2002-07-18 Phil Schwan <phil@clusterfs.com>
6683 * delivered as Lustre Light Alpha
6684 * fixed a crash after handling invalid MDS requests
6685 * fixed directory pages for architectures with non-4k pages sizes
6687 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6688 * release version v0_4_4
6689 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6691 * Fixes a number of interruption problems with OST operations.
6692 * Update documentation for portals header changes
6693 * Move all wire protocol structs/defines to lustre_idl.h
6694 * Fixes symlink length bug.
6695 * Add tcpdump to repository.
6697 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6698 * release version v0_4_3
6699 * Fixes statfs for inodes on extN.
6700 * Fixes bug in runtests which would delete /etc/hosts.
6701 * Use 64-bit object IDs wherever possible (not into VFS though)
6702 Remove ost_get_info, which is unused by lustre, and out of date.
6704 2002-07-03 Peter Braam <braam@clusterfs.com>
6705 * release version v0_4_2 Fixes a lookup error (type not passed)
6706 * move forward to head of Portals
6707 * move forward to latest Lustre kernel
6709 2002-06-25 Peter Braam <braam@clusterfs.com>
6710 * release version v0_4_1. Hopefully stable on single node use.