1 01-31-2018 Intel Corporation
3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.23.1.el6 (RHEL6.9)
11 3.10.0-693.21.1.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.120-94.17 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.23.1.el6 (RHEL6.9)
24 3.10.0-693.21.1.el7 (RHEL7.4)
25 3.0.101-0.47.71 (SLES11 SP3)
26 3.0.101-107 (SLES11 SP4)
27 3.12.74-60.64.40 (SLES12 SP1)
28 4.4.120-92.70 (SLES12 SP2)
29 4.4.120-94.17 (SLES12 SP3)
30 4.4.0-116 (Ubuntu 16.04)
32 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
33 * Recommended ZFS / SPL version: 0.7.5
34 * Tested with ZFS / SPL version: 0.7.5
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
41 --------------------------------------------------------------------------------
43 12-11-2017 Intel Corporation
45 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
46 for currently supported client and server kernel versions.
47 * Server known to build on patched kernels:
48 2.6.32-431.29.2.el6 (RHEL6.5)
49 2.6.32-504.30.3.el6 (RHEL6.6)
50 2.6.32-573.26.1.el6 (RHEL6.7)
51 2.6.32-642.15.1.el6 (RHEL6.8)
52 2.6.32-696.20.1.el6 (RHEL6.9)
53 3.10.0-693.17.1.el7 (RHEL7.4)
54 3.0.101-0.47.71 (SLES11 SP3)
55 3.0.101-107 (SLES11 SP4)
56 3.12.74-60.64.40 (SLES12 SP1)
57 4.4.103-92.56 (SLES12 SP2)
58 4.4.103-6.38 (SLES12 SP3)
59 vanilla linux 4.6.7 (ZFS only)
60 * Client known to build on unpatched kernels:
61 2.6.32-431.29.2.el6 (RHEL6.5)
62 2.6.32-504.30.3.el6 (RHEL6.6)
63 2.6.32-573.26.1.el6 (RHEL6.7)
64 2.6.32-642.15.1.el6 (RHEL6.8)
65 2.6.32-696.20.1.el6 (RHEL6.9)
66 3.10.0-693.17.1.el7 (RHEL7.4)
67 3.0.101-0.47.71 (SLES11 SP3)
68 3.0.101-107 (SLES11 SP4)
69 3.12.74-60.64.40 (SLES12 SP1)
70 4.4.103-92.56 (SLES12 SP2)
71 4.4.103-6.38 (SLES12 SP3)
73 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
74 * Recommended ZFS / SPL version: 0.7.3
75 * Tested with ZFS / SPL version: 0.7.3
76 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
77 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
78 stack overflow. For more information, please refer to bugzilla 17630.
79 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
80 the re-exporting nodes to be mounted with "32bitapi" mount option
82 --------------------------------------------------------------------------------
84 10-02-2017 Intel Corporation
86 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
87 for currently supported client and server kernel versions.
88 * Server known to build on patched kernels:
89 2.6.32-431.29.2.el6 (RHEL6.5)
90 2.6.32-504.30.3.el6 (RHEL6.6)
91 2.6.32-573.26.1.el6 (RHEL6.7)
92 2.6.32-642.15.1.el6 (RHEL6.8)
93 2.6.32-696.10.3.el6 (RHEL6.9)
94 3.10.0-693.2.2.el7 (RHEL7.4)
95 3.0.101-0.47.71 (SLES11 SP3)
96 3.0.101-107 (SLES11 SP4)
97 3.12.74-60.64.40 (SLES12 SP1)
98 4.4.74-92.35 (SLES12 SP2)
99 vanilla linux 4.6.7 (ZFS only)
100 * Client known to build on unpatched kernels:
101 2.6.32-431.29.2.el6 (RHEL6.5)
102 2.6.32-504.30.3.el6 (RHEL6.6)
103 2.6.32-573.26.1.el6 (RHEL6.7)
104 2.6.32-642.15.1.el6 (RHEL6.8)
105 2.6.32-696.10.3.el6 (RHEL6.9)
106 3.10.0-693.2.2.el7 (RHEL7.4)
107 3.0.101-0.47.71 (SLES11 SP3)
108 3.0.101-107 (SLES11 SP4)
109 3.12.74-60.64.40 (SLES12 SP1)
110 4.4.74-92.35 (SLES12 SP2)
112 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
113 * Recommended ZFS / SPL version: 0.7.1
114 * Tested with ZFS / SPL version: 0.7.1
115 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
116 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
117 stack overflow. For more information, please refer to bugzilla 17630.
118 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
119 the re-exporting nodes to be mounted with "32bitapi" mount option
121 --------------------------------------------------------------------------------
123 2017-07-05 Intel Corporation
125 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
126 for currently supported client and server kernel versions.
127 * Server known to build on patched kernels:
128 2.6.32-431.29.2.el6 (RHEL6.5)
129 2.6.32-504.30.3.el6 (RHEL6.6)
130 2.6.32-573.26.1.el6 (RHEL6.7)
131 2.6.32-642.15.1.el6 (RHEL6.8)
132 2.6.32-696.3.1.el6 (RHEL6.9)
133 3.10.0-514.21.1.el7 (RHEL7.3)
134 3.0.101-0.47.71 (SLES11 SP3)
135 3.0.101-100 (SLES11 SP4)
136 3.12.74-60.64.40 (SLES12 SP1)
137 4.4.59-92.17 (SLES12 SP2)
138 vanilla linux 4.6.7 (ZFS only)
139 * Client known to build on unpatched kernels:
140 2.6.32-431.29.2.el6 (RHEL6.5)
141 2.6.32-504.30.3.el6 (RHEL6.6)
142 2.6.32-573.26.1.el6 (RHEL6.7)
143 2.6.32-642.15.1.el6 (RHEL6.8)
144 2.6.32-696.3.1.el6 (RHEL6.9)
145 3.10.0-514.21.1.el7 (RHEL7.3)
146 3.0.101-0.47.71 (SLES11 SP3)
147 3.0.101-100 (SLES11 SP4)
148 3.12.74-60.64.40 (SLES12 SP1)
149 4.4.59-92.17 (SLES12 SP2)
151 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
152 * Recommended ZFS / SPL version: 0.6.5.9
153 * Tested with ZFS / SPL version: 0.6.5.9
154 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
155 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
156 stack overflow. For more information, please refer to bugzilla 17630.
157 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
158 the re-exporting nodes to be mounted with "32bitapi" mount option
160 --------------------------------------------------------------------------------
162 12-07-2016 Intel Corporation
164 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
165 for currently supported client and server kernel versions.
166 * Server known to build on patched kernels:
167 2.6.32-431.29.2.el6 (RHEL6.5)
168 2.6.32-504.30.3.el6 (RHEL6.6)
169 2.6.32-573.26.1.el6 (RHEL6.7)
170 2.6.32-642.11.1.el6 (RHEL6.8)
171 3.10.0-514.el7 (RHEL7.3)
172 3.0.101-0.47.71 (SLES11 SP3)
173 3.0.101-80 (SLES11 SP4)
174 3.12.62-60.64.8 (SLES12 SP1)
175 vanilla linux 4.5.7 (ZFS only)
176 * Client known to build on unpatched kernels:
177 2.6.32-431.29.2.el6 (RHEL6.5)
178 2.6.32-504.30.3.el6 (RHEL6.6)
179 2.6.32-573.26.1.el6 (RHEL6.7)
180 2.6.32-642.11.1.el6 (RHEL6.8)
181 3.10.0-514.el7 (RHEL7.3)
182 3.0.101-0.47.71 (SLES11 SP3)
183 3.0.101-80 (SLES11 SP4)
184 3.12.62-60.64.8 (SLES12 SP1)
186 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
187 * Recommended ZFS / SPL version: 0.6.5.8
188 * Tested with ZFS / SPL version: 0.6.5.8
189 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
190 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
191 stack overflow. For more information, please refer to bugzilla 17630.
192 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
193 the re-exporting nodes to be mounted with "32bitapi" mount option
195 --------------------------------------------------------------------------------
197 02-29-2016 Intel Corporation
199 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
200 for currently supported client and server kernel versions.
201 * Server known to build on patched kernels:
202 2.6.32-431.29.2.el6 (RHEL6.5)
203 2.6.32-504.30.3.el6 (RHEL6.6)
204 2.6.32-573.12.1.el6 (RHEL6.7)
205 3.10.0-327.3.1.el7 (RHEL7.2)
206 3.0.101-0.47.71 (SLES11 SP3)
207 3.0.101-68 (SLES11 SP4)
208 vanilla linux 4.2.1 (ZFS only)
209 * Client known to build on unpatched kernels:
210 2.6.32-431.29.2.el6 (RHEL6.5)
211 2.6.32-504.30.3.el6 (RHEL6.6)
212 2.6.32-573.12.1.el6 (RHEL6.7)
213 3.10.0-327.3.1.el7 (RHEL7.2)
214 3.0.101-0.47.71 (SLES11 SP3)
215 3.0.101-68 (SLES11 SP4)
218 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
219 * Recommended ZFS / SPL version: 0.6.4.2
220 * Tested with ZFS / SPL version: 0.6.4.2
221 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
222 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
223 stack overflow. For more information, please refer to bugzilla 17630.
224 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
225 the re-exporting nodes to be mounted with "32bitapi" mount option
227 --------------------------------------------------------------------------------
229 03-10-2015 Intel Corporation
231 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
232 for currently supported client and server kernel versions.
233 * Server known to build on patched kernels:
234 2.6.32-431.29.2.el6 (RHEL6.5)
235 2.6.32-504.8.1.el6 (RHEL6.6)
236 3.0.101-0.46 (SLES11 SP3)
237 * Client known to build on unpatched kernels:
238 2.6.32-431.29.2.el6 (RHEL6.5)
239 2.6.32-504.8.1.el6 (RHEL6.6)
240 3.10.0-123.20.1.el7 (RHEL7)
241 3.0.101-0.46 (SLES11 SP3)
242 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
243 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
244 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
245 stack overflow. For more information, please refer to bugzilla 17630.
246 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
247 the re-exporting nodes to be mounted with "32bitapi" mount option
249 Severity : enhancement
251 Description: control OST-index in IDIF via ROCOMPAT flag.
252 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
253 last_rcvd file. For new formatted OST device, it will be auto set;
254 for the case of upgrading from old OST device, you can enable it
255 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
256 enabled, for new created OST-object, its IDIF-in-LMA will contain
257 the OST-index; for the existing OST-object, the OSD will convert
258 old format IDIF as new format IDIF with OST-index stored in the
259 LMA EA when accessing such OST-object or via OI scrub. Once such
260 flag is enabled, it cannot be reverted back, so the system cannot
261 be downgraded to the orignal incompatible version.
263 --------------------------------------------------------------------------------
265 07-30-2014 Intel Corporation
267 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
268 for currently supported client and server kernel versions.
269 * Server known to build on patched kernels:
270 2.6.32-431.20.3.el6 (RHEL6.5)
271 3.0.101-0.31 (SLES11 SP3)
272 * Client known to build on unpatched kernels:
273 2.6.32-431.20.3.el6 (RHEL6.5)
274 3.0.101-0.31 (SLES11 SP3)
277 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
278 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
279 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
280 stack overflow. For more information, please refer to bugzilla 17630.
281 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
282 the re-exporting nodes to be mounted with "32bitapi" mount option
284 --------------------------------------------------------------------------------
286 10-31-2013 Intel Corporation
288 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
289 * Server builds for kernels:
290 2.6.32-358.18.1.el6 (RHEL6.4)
291 * Client builds for unpatched kernels:
292 2.6.32-358.18.1.el6 (RHEL6.4)
293 3.0.92-0.8 (SLES11 SP3)
295 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
296 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
297 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
298 stack overflow. For more information, please refer to bugzilla 17630.
299 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
300 the re-exporting nodes to be mounted with "32bitapi" mount option
302 --------------------------------------------------------------------------------
304 05-30-2013 Intel Corporation
306 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
307 * Server builds for kernels:
308 2.6.32-358.6.2.el6 (RHEL6.4)
309 * Client builds for unpatched kernels:
310 2.6.32-358.6.2.el6 (RHEL6.4)
311 3.0.74-0.6.6 (SLES11 SP2)
313 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
314 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
315 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
316 stack overflow. For more information, please refer to bugzilla 17630.
317 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
318 the re-exporting nodes to be mounted with "32bitapi" mount option
320 --------------------------------------------------------------------------------
322 9-30-2012 Whamcloud, Inc.
324 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
325 * Server support for kernels:
326 2.6.32-279.5.1.el6 (RHEL6)
327 * Client support for unpatched kernels:
328 2.6.32-279.5.1.el6 (RHEL6)
330 2.6.18-238.19.1.el5 (RHEL5)
331 2.6.16.60-0.69.1 (SLES10)
332 * Recommended e2fsprogs version: 1.41.90.wc4
333 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
334 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
335 stack overflow. For more information, please refer to bugzilla 17630.
336 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
337 the re-exporting nodes to be mounted with "32bitapi" mount option
339 Severity : enhancement
341 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
343 Severity : enhancement
345 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
347 Severity : enhancement
349 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
351 Severity : enhancement
353 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
355 Severity : enhancement
357 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
359 --------------------------------------------------------------------------------
361 04-30-2012 Whamcloud, Inc.
363 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
364 * Server support for kernels:
365 2.6.32-220.4.2.el6 (RHEL6)
366 * Client support for unpatched kernels:
367 2.6.32-220.4.2.el6 (RHEL6)
369 2.6.18-238.19.1.el5 (RHEL5)
370 2.6.16.60-0.69.1 (SLES10)
371 * Recommended e2fsprogs version: 1.41.90.wc4
372 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
373 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
374 stack overflow. For more information, please refer to bugzilla 17630.
375 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
376 the re-exporting nodes to be mounted with "32bitapi" mount option
378 Severity : enhancement
380 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
382 Severity : enhancement
384 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
386 Severity : enhancement
388 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
390 Severity : enhancement
392 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
394 --------------------------------------------------------------------------------
396 09-30-2011 Whamcloud, Inc.
398 * Server support for kernels:
399 2.6.18-238.19.1.el5 (RHEL5)
400 2.6.32-131.6.1.el6 (RHEL6)
401 * Client support for unpatched kernels:
402 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
403 2.6.18-238.19.1.el5 (RHEL5)
404 2.6.32-131.6.1.el6 (RHEL6)
405 2.6.29.4-167.fc11 (FC11)
406 2.6.16.60-0.69.1 (SLES10)
408 * Recommended e2fsprogs version: 1.41.90.wc3
409 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
410 Lustre filesystem with 4K stack may cause a stack overflow. For more
411 information, please refer to bugzilla 17630.
412 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
413 mounted with "32bitapi" mount option
415 Severity : enhancement
416 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
418 Severity : enhancement
419 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
421 Severity : enhancement
422 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
425 Frequency : only with SLES10
427 Description: Use OFED "KMP" provided by Novell
428 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
429 Lustre is now built against this package. That means you need to
430 install the ofed-kmp package from Novell for the patchless client.
431 Note that the ofed-kmp that Novell ships may not exactly match the
432 kernel version but should still be compatible.
434 Severity : enhancement
436 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
437 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
439 --------------------------------------------------------------------------------
441 2010-07-15 Oracle, Inc.
443 * Server support for kernels:
444 2.6.18-164.11.1.el5 (RHEL 5)
445 2.6.18-164.11.1.0.1.el5 (OEL 5)
446 * Client support for unpatched kernels:
447 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
448 2.6.18-164.11.1.el5 (RHEL 5),
449 2.6.18-164.11.1.0.1.el5 (OEL 5)
450 2.6.16.60-0.42.8 (SLES 10),
452 2.6.29.4-167.fc11 (FC11)
453 * Recommended e2fsprogs version: 1.41.10-sun2
455 Severity : enhancement
457 Description: Distribute lfs_migrate script with lustre
458 Details : The lfs_migrate mentioned in the manual is now included in the
459 lustre tree in lustre/scripts and will distributed in lustre RPMs.
461 Severity : enhancement
463 Description: Failover nid list added to import proc file.
465 Severity : enhancement
467 Description: "som_preview" mount option required to use SOM
468 Details : The clients will need to use som_preview mount option
472 Description: Lustre does not do 1MB IOs to HW RAID
473 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
475 Severity : enhancement
477 Description: lfs quota output is not very convenient for awk/sed-parsing
478 Details : Some positions in lfs quota output table could be empty or
479 non-empty which made it hard to parse it with scripts, now
480 a dash is put instead of space where there is not supposed
485 Description: Update to OFED 1.5.1
486 Details : OFED 1.5.1 will be used in environments that don't use
487 vendor supplied OFED.
491 Description: Remove lnet. prefix from lctl params display
492 Details : Also renamed lnet.memused to lnet_memused to
497 Description: Use conf_param -d to remove permanant settings.
501 Description: Add list_param -R
502 Details : lctl list_param now supports recursive mode.
505 Bugzilla : 21829, 22312
506 Description: Fix llobdstat.
507 Details : 1. Now updating the write_bytes counter
508 2. Port script fixes from 1.8.
512 Description: decrease the usage of memory on clients.
513 Details : 1. On clients, recycle dentries and inodes unused.
514 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
519 Description: llobdstat fix and enhancement
520 Details : add a counter to set a limit to how many samples will be returned
521 fix a wildcard in the path to limit to obdfilter stats only
525 Description: lfs check only allowed for root
526 Details : Code cleanups and added a regresion test of non-root use of
531 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
532 Details : Fix hr->hr_index overload.
536 Description: Fix FIEMAP functionality.
540 Description: FC11 patchless client support added
544 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
545 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
549 Description: New acc-sm script and various test-framework enhancments.
553 Description: Previously landed patch was reverted.
557 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
561 Description: Add ldiskfs maxdirsize mount option.
565 Description: Fix mkfs.lustre support for 16TB LUNs.
569 Description: Print mmp_check_interval
570 Details : print mmp_check_interval and make it possible to abort
571 mount operation in case it takes too long.
575 Description: Add cascading_rw to lustre mpi tests.
579 Description: Store SOM data in EA.
580 Details : Also several SOM related cleanups.
584 Description: Add refcount to md_open_data
588 Description: Use refcount on lov_request_set to prevent memory corruption.
592 Description: Statahead: use parent dir's i_mutex for synchronization.
596 Description: loadgen improvements.
600 Description: Enable peer health detection in ksocklnd and ko2iblnd.
604 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
608 Description: Fix race on obd_devcie:md_stats
612 Description: Fix SLES 11 client builds.
613 Details : Resolved uninitalized pointer gcc warnings.
617 Description: Allow non-root access for "lfs check".
618 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
622 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
624 Severity : enhancement
626 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
627 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
631 Description: Sanity test_27p fails.
632 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
633 Sanity tests 27 were cleaned up and improved.
637 Description: Unify req format on client/servers
638 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
639 buffers and swabbers approach..
644 Description: drop join file stuff.
646 Severity : enhancement
648 Description: Add 2.6.27 and SLES11 (patchless client) support
652 Description: Fix in ptlrpc_expire_one_request() to print the signed time
655 Severity : enhancement
657 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
660 Severity : enhancement
662 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
664 Severity : enhancement
666 Description: correctly handle big reply message.
667 Details : send LNet event if reply is bigger than buffer and adjust this buffer
672 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
674 Severity : enhancement
676 Description: Add support for OEL5.
680 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
682 Severity : enhancement
684 Description: "weak-modules" support
685 Details : Implement "weak-modules" support which enables kernel modules
686 to be used with any kernel that implements the same kABI. In
687 order to achieve this modules are now installed in
688 /lib/modules/$(uname -r)/updates/kernel on all distributions.
692 Description: client could not reconnect to OST because of an active request.
693 Details : The client now retries to reconnect to the same server, if a
694 connect request failed with EBUSY or -EAGAIN. Test case 35b
695 conf-sanity.sh has been created.
699 Description: Don't allow make backward step on assignin osc next id.
700 Details : race between allocation next id and ll_sync thread can be cause
701 of set wrong osc next id and can be kill valid ost objects.
705 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
709 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
713 Description: Error handling in osc_statfs_interpret() has been improved.
714 Details : Check in osc_statfs_interpret() for EBADR.
718 Description: Do not send statfs() requests to OSTs disabled by administrator.
719 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
721 Severity : enhancement
723 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
725 Severity : enhancement
727 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
729 Severity : enhancement
731 Description: Enhancement for lfs(1) command to use numeric uid/gid.
735 Description: Improved cleanup procedure for conf-sanity test_46a.
739 Description: Do not start bulk transfer, if client tried to reconnect.
740 Details : Do not start bulk transfer, if client tried to reconnect,
741 but failed due to active rpcs. Check for exp_abort_active_req
742 has been implemented.
746 Description: Prevent inconsistences between linux and lustre mount structures.
747 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
748 Make the sleep interruptible.
750 Severity : enhancement
752 Description: provide server to client comms path
753 Details : server to client communications path via new LDLM_SET_INFO rpc
757 Description: truncate starts GFP_FS allocation under transaction causing deadlock
758 Details : ldiskfs_truncate calls grab_cache_page which may start page
759 allocation under an open transaction. This may lead to
760 calling prune_icache with consequent lustre reentrance.
764 Description: send correctly lsm on open replay
765 Details : MDS is trust to LSM size on replay open, but client can set wrong size
770 Description: enable client interface failover
771 Details : When a child reconnects from another NID, properly update export
772 nid hash position and ldlm reverse import.
774 Severity : enhancement
776 Description: Communicate OST degraded/readonly state via statfs to MDS
777 Details : Flags in the statfs returned from OSTs indicate whether the
778 OST is in a degraded RAID state, or if the filesystem has
779 turned read-only after a filesystem error is detected.
784 Description: don't panic if EPROTO was hit when reading symlink
785 Details : correctly handling request reference in error cases.
787 Severity : enhancement
789 Description: Add LustreNetLink, a kernel-userspace communcation path.
791 Severity : enhancement
793 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
798 Description: access to llog context before init.
799 Details : move handling CATALOGS file at osc layer and forbid access to llog
803 Frequency : rare, only when client is writing to OST on local node
805 Description: Avoid deadlock for local client writes
806 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
807 memory freeing context. This allows OST threads to set the
808 PF_MEMALLOC flag on task structures in order to allocate memory
809 from reserved pools and complete IO.
810 Use GFP_HIGHUSER for OST allocations for non-local client writes,
811 so that the OST threads generate memory pressure and allow
812 inactive pages to be reclaimed.
814 Severity : enhancement
816 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
821 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
822 Details : move ldlm namespace creation in setup phase to avoid grab
823 _lprocfs_lock with cli_sem held.
827 Description: Temporarily disable grant shrink.
828 Details : Disable the feature for debugging.
831 Frequency : only when formatting test filesystems
833 Description: Unable to run several mkfs.lustre on loop devices at the same time
834 Details : mkfs.lustre returns error 256 on the concurrent loop devices
835 formatting. The solution is to proper handle the error.
837 Severity : enhancement
839 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
841 Severity : enhancement
843 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
848 Description: resolve race between obd_disconnect and class_disconnect_exports
849 Details : if obd_disconnect will be called to already disconnected export he
850 forget release one reference and osc module can't unloaded.
852 Severity : enhancement
854 Description: Allow tuning service thread via /proc
855 Details : For each service a new
856 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
857 created that can be used to set min/max thread counts, and get the
858 current number of running threads.
863 Descriptoin: don't return error if have partially created objects for file.
864 Details : lov_update_create_set uses set->set_success as index for created
865 objects, so if some requests will be failed, they will have hole
866 at end of array and qos_shrink_lsm for allocate correct lsm.
868 Severity : enhancement
870 Description: Update OFED support to 1.4.1
875 Description: Reduce small size read RPC
876 Details : Set read-ahead limit for every file and only do read-ahead when
877 available read-ahead pages are bigger than 1M to avoid small size
880 Severity : enhancement
882 Description: move AT tunable parameters for more consistent usage
883 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
885 Severity : enhancement
887 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
888 Details : allow skip disconnected ost for send statfs request and hide error
894 Description: fix racy locking of mballoc block bitmaps causing BUG
895 Details : The locking of the mballoc buddy bitmap and the in-memory
896 block bitmap was using two different spin locks in some
897 cases. This made it possible to incorrectly access the
898 mballoc bitmap while another process was modifying it,
899 causing a sanity assertion to fail. While no on-disk corruption
900 was reported, there was some risk of this happening.
903 Frequency : rare, on llog test 6
905 Descriptoin: don't allow connect to already connected import
906 Details : allowing connect to already connected import is hide connecting problem.
911 Descriptoin: allow kill process which wait statahead result
912 Details : for some reasons 'ls' can stick in waiting result from statahead,
913 in this case need way for kill this process.
915 Severity : enhancement
917 Description: Add state history info file, enhance import info file
918 Details : Track import connection state changes in a new osc/mdc proc file;
919 add overview-type data to the osc/mdc import proc file.
921 Severity : enhancement
923 Description: MDS create should not wait for statfs RPC while holding DLM lock.
925 Severity : enhancement
927 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
932 Descriptoin: OST grant too much space to client even there are not enough space.
933 Details : Client will shrink its grant cache to OST if there are no write
934 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
935 this grant cache if there are already not enough avaible space
936 (left_space < total_clients * 32M).
939 Frequency : start MDS on uncleanly shutdowned MDS device
941 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
942 Details : stay in waiting mds<>ost recovery finished produce random bugs
943 due race between two ll_sync thread for one lov target. send
944 ACTIVATE event only if connect realy finished and import have
948 Frequency : rare, connect and disconnect target at same time
950 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
951 Details : don't call obd_disconnect under lov_lock. this long time
952 operation and can block ptlrpcd which answer to connect request.
957 Descriptoin: don't lose wakeup for imp_recovery_waitq
958 Details : recover_import_no_retry or invalidate_import and import_close can
959 both sleep on imp_recovery_waitq, but we was send only one wakeup
963 Frequency : always with long access acl
965 Descriptoin: mds can't pack reply with long acl.
966 Details : mds don't control size of acl but they limited by reint/getattr
969 Severity : enhancement
971 Description: Update to SLES10 kernel-2.6.16.60-0.33.
973 Severity : enhancement
975 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
978 Frequency : start MDS on uncleanly shutdowned MDS device
980 Descriptoin: aborting recovery hang on MDS
981 Details : don't throttle destroy RPCs for the MDT.
984 Frequency : on remount
986 Description: external journal device not working after the remount
987 Details : clear dev_rdonly flag for external journal devices in
993 Description: shutdown vs evict race
994 Details : client_disconnect_export vs connect request race.
995 if client will evicted at this time - we start invalidate
996 thread without referece to import and import can be freed
1000 Frequency : rare, need acl's on inode.
1002 Description: client can't handle ost additional correctly
1003 Details : if ost was added after client connected to mds client can have
1004 hit lnet_try_match_md ... to big messages to wide striped files.
1005 in this case need teach client to handle config events about add
1006 lov target and update client max ea size at that event.
1008 Severity : enhancement
1010 Description: Changelogs
1011 Details : Changelogs are a lightweight mechanism to track filesystem
1012 metadata and namespace changes. The changelog is recorded
1013 permanently on the MDTs, and is periodically "consumed" / purged
1014 when records are no longer needed.
1016 Severity : enhancement
1018 Description: compact fld format with extents
1019 Details : Store range of seq rather than every seq in FLD. Seq
1020 controller update FLD rather than clients. In Case of CMD, mdt0
1021 has FLD, all other metadata server act as non persistent proxy
1022 for FLD queries and cache fld entries in fld cache.
1027 Description: don't skip ost target if they assigned to file
1028 Details : Drop slow OSCs if we can, but not for requested start idx.
1029 This means "if OSC is slow and it is not the requested
1030 start OST, then it can be skipped, otherwise skip it only
1031 if it is inactive/recovering/out-of-space.
1035 Description: more cleanup in mds_lov
1036 Details : not send LOV EA under replay, we can't know about they size at this
1037 time. Don't allow client connect to mds before any ost connected,
1038 for avoid problems with LOV EA size and returning EIO to client.
1040 Severity : enhancement
1042 Description: Interoperability at server side (Disk interoperability)
1044 Severity : enhancement
1046 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1048 Severity : enhancement
1050 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1052 Severity : enhancement
1054 Description: New client IO stack (CLIO).
1056 Severity : enhancement
1058 Description: Commit on sharing. Eliminate inter-client dependencies between
1059 uncommitted transactions by doing transaction commits.
1060 Thereby clients may recovery independently.
1063 Frequency : Create a symlink file with a very long name
1065 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1066 Details : If there is no extra space in the request for early cancels,
1067 ldlm_req_handles_avail() returns 0 instead of a negative value.
1069 Severity : enhancement
1071 Description: Add /proc entry for import status
1072 Details : The mdc, osc, and mgc import directories now have
1073 an import directory that contains useful import data for debugging
1074 connection problems.
1076 Severity : enhancement
1078 Description: Re-disable certain /proc logging
1079 Details : Enable and disable client's offset_stats, extents_stats and
1080 extents_stats_per_process stats logging on the fly.
1083 Frequency : Only on FC kernels 2.6.22+
1085 Description: oops in statahead
1086 Details : Do not drop reference count for the dentry from VFS when lookup,
1087 VFS will do that by itself.
1089 Severity : enhancement
1091 Description: Generic /proc file permissions
1092 Details : Set /Proc file permissions in a more generic way to enable non-
1093 root users operate on some /proc files.
1097 Description: Hitting mdc_commit_close() ASSERTION
1098 Details : Properly handle request reference release in
1099 ll_release_openhandle().
1103 Description: quota recovery deadlock during mds failover
1104 Details : This patch includes att18982, att18236, att18237 in bz14840.
1106 1. fix osts hang when mds does failover with quotaon
1107 2. prevent watchdog storm when osts threads wait for the
1112 Frequency : only patchless client
1113 Description: add workaround for race between add/remove dentry from hash
1115 Severity : enhancement
1117 Description: Allow OST glimpses to return PW locks
1121 Description: LBUG when llog conf file is full
1122 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1126 Description: Prevent import from entering FULL state when server in recovery
1130 Description: service mount cannot take device name with ":"
1131 Details : Only when device name contains ":/" will mount treat it as
1137 Description: replace ptlrpcd with the statahead thread to interpret the async
1138 statahead RPC callback
1142 Frequency : on recovery
1143 Description: I/O failures after umount during fail back
1144 Details : if client reconnected to restarted server we need join to recovery
1145 instead of find server handler is changed and process self
1146 eviction with cancel all locks.
1148 Severity : enhancement
1150 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1152 Severity : enhancement
1154 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1156 Severity : enhancement
1158 Description: Upcall on Lustre log has been dumped
1159 Details : Allow for a user mode script to be called once a Lustre log has
1160 been dumped. It passes the filename of the dumped log to the
1161 script, the location of the script can be specified via
1162 /proc/sys/lnet/debug_log_upcall.
1167 Description: avoid idr_remove called for id which is not allocated.
1168 Details : Move assigment s_dev for clustered nfs to end of initialization,
1169 to avoid problem with error handling.
1174 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1175 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1176 into CONN_USED_HASH and this prodice warning when put connection
1177 again in unused hash.
1180 Severity : enhancement
1182 Description: File striping can now be set to use an arbitrary pool of OSTs.
1184 Severity : enhancement
1186 Description: Export bytes_read/bytes_write count on OSC/OST.
1190 Description: Early reply size mismatch, MGC loses connection
1191 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1192 the connect flags are properly negotiated.
1197 Description: quotas are not honored with O_DIRECT
1198 Details : all writes with the flag O_DIRECT will use grants which leads to
1199 this problem. Now using OBD_BRW_SYNC to guard this.
1203 Description: add quota statistics
1204 Details : 1. sort out quota proc entries and proc code.
1205 2. add quota statistics
1207 Severity : enhancement
1209 Description: enable quota support for HEAD.
1213 Description: Properly propagate oinfo flags from lov to osc for statfs
1214 Details : restore missing copy oi_flags to lov requests.
1216 Severity : enhancement
1218 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1219 : l_getgroups(8), lst(8), routerstat(8)
1220 Details : included man pages for llobdstat(8), llstat(8),
1221 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1223 Severity : enhancement
1225 Description: configure's --enable-quota should check the
1226 : kernel .config for CONFIG_QUOTA
1227 Details : configure is terminated if --enable-quota is passed but
1228 : no quota support is in kernel
1232 Description: Remove portals compatibility
1233 Details : Remove portals compatibility, not interoperable with releases
1238 Description: Resolve device initialization race
1239 Details : Prevent proc handler from accessing devices added to the
1240 obd_devs array but yet be intialized.
1242 Severity : enhancement
1244 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1246 Severity : enhancement
1248 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1253 Description: Using wrong pointer in osc_brw_prep_request
1254 Details : Access to array[-1] can produce panic if kernel compiled with
1255 CONFIG_PAGE_ALLOC enabled
1259 Description: Client runs out of low memory
1260 Details : Consider only lowmem when counting initial number of llap pages
1264 Description: *optional* service tags registration
1265 Details : if the "service tags" package is installed on a Lustre node
1266 When the filesystem is mounted, a local-node service tag will
1267 be created. See http://inventory.sun.com/ for more information
1268 about the Service Tags asset management system.
1272 Description: Kernel BUG tries to release flock
1273 Details : Lustre does not destroy flock lock before last reference goes
1274 away. So always drop flock locks when client is evicted and
1275 perform unlock regardless of successfulness of speaking to MDS.
1279 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1283 Description: sanity test 65a fails if stripecount of -1 is set
1284 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1289 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1290 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1293 Severity : enhancement
1295 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1299 Description: do not process already freed flock
1300 Details : flock can possibly be freed by another thread before it reaches
1301 to ldlm_flock_completion_ast.
1305 Description: LBUG during stress test
1306 Details : Need properly lock accesses the flock deadlock detection list.
1310 Description: oops in page fault handler
1311 Details : kernel page fault handler can return two special 'pages' in error
1312 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1316 Description: timeout with invalidate import.
1317 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1318 should be handled by ptlrpcd. This produce long age waiting and
1319 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1321 Severity : enhancement
1323 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1327 Description: enable MGS and MDT services start separately
1328 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1329 a co-located MGS without starting the MGS, which is a complement
1330 to 'nosvc' mount option.
1334 Frequency : after recovery
1335 Description: precreate to many object's after del orphan.
1336 Details : del orphan st in oscc last_id == next_id and this triger growing
1337 count of precreated objects. Set flag LOW to skip increase count
1338 of precreated objects.
1342 Frequency : rare, on clear nid stats
1343 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1344 Details : when clean nid stats sometimes try destroy live entry,
1345 and this produce panic in free.
1349 Description: Stack overflow during MDS log replay
1350 ease stack pressure by using a thread dealing llog_process.
1354 Description: wait until IO finished before start new when do lock cancel.
1355 Details : VM protocol want old IO finished before start new, in this case
1356 need wait until PG_writeback is cleared until check dirty flag and
1357 call writepages in lock cancel callback.
1359 Severity : enhancement
1361 Description: using special macro for print time and cleanup in includes.
1365 Description: mds_mfd_close() ASSERTION(rc == 0)
1366 Details : In mds_mfd_close(), we need protect inode's writecount change
1367 within its orphan write semaphore to prevent possible races.
1371 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1375 Frequency : rare, on shutdown ost
1376 Description: don't hit live lock with umount ost.
1377 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1378 use shrink_dcache_sb instead.
1382 Description: don't panic with use echo client
1383 Details : echo client pass NULL as client nid pointer and this produce null
1384 pointer dereference.
1388 Description: fix build on ppc32
1389 Details : compile code with -m64 flag produce wrong object file for ppc32.
1393 Description: add message levels for liblustreapi
1397 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1398 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1399 will result in an -EAGAIN and in the exit call path we call
1400 alloc_rr() which will with increasing aggressiveness attempt to
1401 aquire precreated objects on the minimum number of required OSCs.
1405 Description: Use old size assignment to avoid deadlock
1406 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1407 the scheduling while holding a spinlock. We do not need locking
1408 for size in ll_update_inode() because size is only updated from
1409 the MDS for directories or files without objects, so there is no
1410 other place to do the update, and concurrent access to such inodes
1411 are protected by the inode lock.
1415 Description: resolve "_IOWR redefined" build error on SLES10
1419 Description: dump the memory debugging after all modules are unloaded to
1420 suppress false negative in conf_sanity test 39
1422 Severity : enhancement
1424 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1429 Description: fix SLES kernel versioning
1430 Details : the kernel version for our SLES 10 kernel did not include a "-"
1431 before the "smp" at the end. while this was not a problem in
1432 general, it did mean that software trying to use the kernel
1433 version to try to detect a vendor specific kernel would fail.
1434 this was most evident by the OFED build scripts.
1438 Description: Don't update lov_desc members until making sure they are valid
1439 Details : When updating lov_desc members via proc fs, need fix their
1440 validities before doing the real update.
1444 Description: don't put request into delay list while invalidate in flight.
1445 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1446 import in flight. this produce timeout for invalidate and sometimes
1447 can cause stale data.
1451 Frequency : on ppc only
1452 Description: not convert ost objects for directory because it's not exist.
1453 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1457 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1459 Severity : enhancement
1461 Description: Update to support 2.6.22.14 vanilla kernel.
1465 Frequency : rare, on recovery
1466 Description: read procfs can produce deadlock in some situation
1467 Details : Holding lprocfs lock which send rpc can produce block for destroy
1468 obd objects and this also block reconnect with -EALREADY. This
1469 isn't fix all lprocfs bugs - but make it rare.
1471 Severity : enhancement
1473 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1476 Frequency : frequent on X2 node
1478 Description: mdc_set_open_replay_data LBUG
1479 Details : Set replay data for requests that are eligible for replay.
1483 Description: lustre_mgs: operation 101 on unconnected MGS
1484 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1485 MGC, and late on MGC cannot successfully connect to MGS and a lot
1486 of the error messages complaining that MGS is not connected.
1490 Frequency : on network error
1491 Description: panic with double free request if network error
1492 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1493 but it's true only for synchronus enqueue, for async enqueue
1494 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1497 Severity : enhancement
1499 Description: client-side metadata stat-ahead during readdir(directory readahead)
1500 Details : perform client-side metadata stat-ahead when the client detects
1501 readdir and sequential stat of dir entries therein
1504 Frequency : on start mds
1506 Description: Implement get_info(last_id) in obdfilter.
1509 Frequency : occasional
1511 Description: Correctly check stale fid, not start epoch if ost not support SOM
1512 Details : open with flag O_CREATE need set old fid in op_fid3 because
1513 op_fid2 was overwritten with new generated fid, but mds can answer
1514 with one of these two fids and both is not stale. Setattr
1515 incorrectly started an epoch and assume will be called
1516 done_writing, but without SOM done_writing ever being called.
1519 Frequency : rare, depends on device drivers and load
1521 Description: MDS or OSS nodes crash due to stack overflow
1522 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1523 In some cases, in conjunction with device drivers that use a lot
1524 of stack the MDS (or possibly OSS) service threads could overflow
1525 the stack. One change which was identified to consume additional
1526 stack has been reworked to avoid the extra stack usage.
1529 Frequency : occasional
1531 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1532 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1533 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1534 oscc_has_objects and see OSC has no objests and call
1535 oscc_internal_create to resend create request.
1537 Severity : enhancement
1539 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1541 Severity : enhancement
1543 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1546 Frequency : very rare
1548 Description: Fix replay if there is an un-replied request and open
1549 Details : In some cases, older replay request will revert the
1550 mcd->mcd_last_xid on MDS which is used to record the client's
1551 latest sent request.
1553 Severity : enhancement
1555 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1557 Severity : enhancement
1559 Description: Add rhel5 support to HEAD.
1561 Serverity : enhancement
1563 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1568 Description: Don't allow skipping OSTs if index has been specified.
1569 Details : Don't allow skipping OSTs if index has been specified, make
1570 locking in internal create lots better.
1574 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1575 Details : access to bitfield in structure is always rounded to long
1576 and this produce problem with not atomic change any bit.
1580 Description: Lustre make rpms failed.
1581 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1582 builds Lustre rpms from tarball.
1585 Frequency : rare on shutdown ost
1587 Description: If llog cancel was not send before clean_exports phase, this can
1588 produce deadlock in llog code.
1589 Details : If llog thread has last reference to obd and call class_import_put
1590 this produce deadlock because llog_cleanup_commit_master wait when
1591 last llog_commit_thread exited, but this never success because was
1592 called from llog_commit_thread.
1596 Description: allow userland application know is lost one of stripes.
1597 Details : fill lvb_blocks with error code on ost and return it to
1598 application if error flag found.
1602 Description: NULL lov_tgts causing MDS oops
1603 Details : more safe checks for NULL lov_tgts for avoid oops.
1605 Severity : enhancement
1607 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1611 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1612 Details : If an OST is down the MDS will hang indefinitely in
1613 obd_statfs_rqset() waiting for the statfs data. While for
1614 MDS QOS usage of statfs, it should not stuck in waiting.
1616 Severity : enhancement
1618 Description: remote_acl support
1619 Details : Support ACL-based permission check for remote user.
1620 Support setfacl/getfacl for remote user with the utils
1621 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1622 the system "{s,g}etfacl" utils.
1624 Severity : enhancement
1626 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1628 Severity : enhancement
1630 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1634 Description: make mgs_setparam() handle fsname containing dash
1635 Details : fsname containing a dash does not work with lctl conf_param
1637 Severity : enhancement
1639 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1641 Severity : enhancement
1643 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1647 Frequency : rare, at shutdown
1648 Description: access already free / zero obd_namespace.
1649 Details : if client_disconnect_export was called without force flag set,
1650 and exist connect request in flight, this can produce access to
1651 NULL pointer (or already free pointer) when connect_interpret
1652 store ocd flags in obd_namespace.
1656 Frequency : only at startup
1657 Description: not alloc memory with spinlock held.
1658 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1659 if any spinlock held.
1661 Severity : enhancement
1663 Description: make lustre randomly fail allocating memory
1664 Details : Make lustre randomly failed allocating memory for testing purpose.
1666 Severity : enhancement
1668 Description: lost problems with lov objid file
1669 Details : Fixes some scability and access to not inited memory problems
1670 in work with lov objdid file.
1675 Description: lfs find does not continue on file error
1676 Details : Continue other files processing when a file/dir is absent.
1680 Description: Inconsistent usage of lustre_pack_reply()
1681 Details : Standardize the usage of lustre_pack_reply() such that it
1682 always generate a CERROR on failure.
1685 Frequency : occasional
1687 Description: MDS hang or stay in waiting lock
1688 Details : If client receive lock with CBPENDING flag ldlm need send lock
1689 cancel as separate rpc, to avoid situation when cancel request
1690 can't processed due all i/o threads stay in wait lock.
1694 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1698 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1702 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1706 Description: Update to RHEL4 latest kernel.
1708 Severity : enhancement
1710 Description: Build SLES10 patchless client fails
1711 Details : The configure was broken by run ./configure with
1712 --with-linux-obj=.... argument for patchless client. When the
1713 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1714 can't search header adequately. Use absolute path such as
1715 -I($LINUX)/include instead.
1719 Description: interrupt oig_wait produce painc on resend.
1720 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1721 private set, and this produce situation when rq_ptlrpcd_data not
1722 copyed to new allocated request and triggered LBUG on assert
1723 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1724 ptlrpcd set if request is changed and can be safety changed to use
1729 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1730 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1731 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1732 these macros and replace them with their difinitions instead.
1736 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1737 Details : Update which_patch & target file for SLES10 latest kernel.
1740 Description: Remove client patches from SLES 10 kernel.
1741 Details : This causes SLES 10 clients to behave as patchless clients
1742 even on a Lustre-patched (server) kernel.
1744 Severity : enhancement
1746 Description: self-adjustable client's lru lists
1747 Details : use adaptive algorithm for managing client cached locks lru
1748 lists according to current server load, other client's work
1749 pattern, memory activities, etc. Both, server and client
1750 side namespaces provide number of proc tunables for controlling
1753 Severity : enhancement
1755 Description: light-weight GSS support
1756 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1757 and reduce performance overhead.
1759 Severity : enhancement
1761 Description: Linux keyring support
1762 Details : Support using service of Linux keyring for Lustre GSS internal
1763 context refresh/cache mechanism.
1767 Description: Fix errors in lfs documentation
1768 Details : Fixes man pages
1772 Description: don't use GFP_* in generic Lustre code.
1773 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1777 Description: obdclass is limited by single OBD_ALLOC(idarray)
1778 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1782 Description: warnings with build patchless client with vanila 2.6.19 and up
1783 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1784 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1788 Frequency : Only for SLES
1789 Description: sanity_quota fail test_1
1790 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1791 file, which makes TSTID[2] inunique.
1795 Frequency : Only for SLES9
1796 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1797 Details : Update target/ChangeLog/which_patch .
1801 Frequency : Only for test_10 in sanity-quota.sh
1802 Description: a bug in quota test code
1803 Details : checking a test flag in a wrong place
1807 Frequency : Only for test_13 in sanity-quota.sh
1808 Description: a bug in quota test script
1809 Details : shouldn't check overfull allocated quota
1813 Description: jbd statistics
1814 Details : Port older jbd statistics patch for sles10
1816 Severity : enhancement
1818 Description: Add printing inode info into message about error in writepage.
1823 Description: reply_lock_interpret crash due to race with it and lock cancel.
1824 Details : Do not replay locks that are being cancelled. Do not reference
1825 locks by their address during replay, just by their handle.
1829 Frequency : When flocks are used.
1830 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1831 Details : Flock locks might destroy just granted lock if it could be merged
1832 with another existing flock, this is done in completion handler,
1833 so teach ldlm_cli_enquque_fini that this is a valid case for
1839 Description: Oops in read and write path when failing to allocate lock.
1840 Details : Check if lock allocation failed and return error back.
1844 Description: lstripe command fails for valid OST index
1845 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1846 instead of lov->desc.ld_active_tgt_count.
1850 Description: sanity.sh failed test 103
1851 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1856 Description: df doesn't work properly if diskfs blocksize != 4K
1857 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1861 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1862 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1864 Severity : enhancement
1866 Description: Adaptive timeouts
1867 Details : RPC timeouts adapt to changing server load and network
1868 conditions to reduce resend attempts and improve recovery time.
1872 Description: llapi_file_create() does not allow some changes
1873 Details : add llapi_file_open() that allows specifying the mode and
1874 open flags, and also returns an open file handle.
1878 Description: merge and cleanup kernel patches.
1879 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1883 Description: Add journal checksum support.(Kernel part)
1884 Details : The journal checksum feature adds two new flags i.e
1885 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1886 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1887 indicates that the commit block contains the checksum for
1888 the blocks described by the descriptor blocks. Now commit
1889 record can be sent to disk without waiting for descriptor
1890 blocks to be written to disk. This behavior is controlled
1891 using JBD2_FEATURE_ASYNC_COMMIT flag.
1895 Description: OSS needs mutliple precreate threads
1896 Details : Add ability to start more than one create thread per OSS.
1900 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1901 Details : execute lfs setstripe on client
1905 Description: mds_obd_create error creating tmp object
1906 Details : When the user sets quota on root, llog will be affected and can't
1907 create files and write files.
1910 Frequency : Always on ia64 patchless client, and possibly others.
1912 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1913 Details : This allows the patchless client to be loaded on architectures
1914 without this export.
1918 Description: RedHat Update kernel for RHEL5
1919 Details : Add the kernel config file for RHEL5.
1923 Description: RedHat Update kernel for RHEL5
1924 Details : Modify the kernel config file more closer RHEL5.
1928 Description: Build failure against Centos5 (RHEL5)
1929 Details : Define PAGE_SIZE when it isn't present.
1933 Description: test_7 on sanity-quota.sh can't be used on separate machines
1934 Details : add facet to handle it
1938 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1939 Details : with NFS, the anon dentry's parent was set to itself in
1940 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1941 corresponding dentry other than use rec->ur_name.
1943 Severity : enhancement
1945 Description: lfs setstripe enhancement
1946 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1950 Description: enable data checksumming by default
1951 Details : enable checksum by default, allow --disable-checksum
1952 configure option and "-o nochecksum" mount option. Checksums
1953 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1957 Description: lustre support for RHEL5
1958 Details : Add support for RHEL5.
1962 Description: Client eviction due to failover config
1963 Details : after a connection loss, the lustre client should attempt to
1964 reconnect to the last active server first before trying the
1965 other potential connections.
1969 Description: when mds and osts use different quota unit(32bit and 64bit),
1970 quota will be released repeatly.
1971 Details : void sending multiple quota reqs to mds, which will keep the
1972 status between the reqs.
1976 Description: osts not allocated evenly to files
1977 Details : change the condition to increase offset_idx
1980 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1982 Description: Data corruption for OSTs that are formatted larger than 2TB
1984 Details : When generating the bio request for lustre file writes the
1985 sector number would overflow a temporary variable before being
1986 used for the IO. The data reads correctly from Lustre (which
1987 will overflow in a similar manner) but other file data or
1988 filesystem metadata may be corrupted in some cases.
1992 Description: Tune the kernel for good SCSI performance.
1993 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1994 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1999 Description: rewrite ext2-derived code in obdclass/uuid.c
2000 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2001 from scratch preserving functionality.
2005 Description: "lfs find -obd UUID" prints directories
2006 Details : "lfs find -obd UUID" will return all directory names instead
2007 of just file names. It is incorrect because the directories
2008 do not reside on the OSTs.
2012 Description: use i_size_read and i_size_write in 2.6 port
2013 Details : replace inode->i_size access with i_size_read/write()
2016 Frequency : when removing large files
2018 Description: scheduling issue during removal of large Lustre files
2019 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2020 It causes scheduling issues when removing large files (17TB in the
2024 Frequency : only with liblustre clients on XT3
2026 Description: evictions taking too long
2027 Details : allow llrd to evict clients directly on OSTs
2030 Frequency : only on ppc
2032 Description: /proc/fs/lustre/devices broken on ppc
2033 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2034 We should make sure the type of 'index' is loff_t and then cast
2035 explicitly as needed below. Do not assign an explicitly cast
2039 Frequency : only for rhel5
2041 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2042 Details : Modify the target file & which_kernel.
2044 Severity : enhancement
2046 Description: omit set fsid for export NFS
2047 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2048 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2052 Frequency : Always, for kernels after 2.6.16
2053 Description: Fix warning idr_remove called for id=.. which is not allocated.
2054 Details : Last kernels save old s_dev before kill super and not allow
2055 to restore from callback - restore it before call kill_anon_super.
2059 Description: buffer overruns could theoretically occur
2060 Details : llapi_semantic_traverse() modifies the "path" argument by
2061 appending values to the end of the origin string, and a
2062 overrun may occur. Adding buffer overrun check in liblustreapi.
2066 Description: Fix error on 'ls .' at the top of the Lustre mount.
2067 Details : Don't revalidate dentry if it is a root dentry.
2071 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2072 Details : Modify vm-tunables-rhel4.patch.
2076 Description: Kernel config for 2.6.18-vanilla.
2077 Details : Modify targets/2.6-vanilla.target.in.
2078 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2079 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2080 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2081 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2085 Description: improve handling recoverable errors
2086 Details : if request processig with error which can be recoverable on server
2087 request should be resend, otherwise page released from cache and
2092 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2093 Details : Modify target file & which_patch.
2094 A flaw was found in the IA32 system call emulation provided
2095 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2096 value could be stored in the %RAX register, which could trigger an
2097 out-of-bounds system call table access. An untrusted local user
2098 could exploit this flaw to run code in the kernel
2099 (ie a root privilege escalation). (CVE-2007-4573).
2103 Description: O_DIRECT bypasses client statistics.
2104 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2105 read_ahead_stats were updated. copied the stats section from
2106 osc_send_oap_rpc() into async_internal().
2110 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2114 Description: change order of libsysio includes
2115 Details : '#include sysio.h' should always come before '#include xtio.h'
2119 Description: handle "serious error: objid * already exists" more gracefully
2120 Details : If LAST_ID value on disk is smaller than the objects existing in
2121 the O/0/d* directories, it indicates disk corruption and causes an
2122 LBUG(). If the object is 0-length, then we should use the existing
2123 object. This will help to avoid a full fsck in most cases.
2125 Severity : enhancement
2127 Description: adapt the lustre_config script to support the upgrade case
2128 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2129 server targets to 1.6 in parallel.
2133 Description: To avoid grant space > avaible space when the disk is almost
2134 full. Without this patch you might see the error "grant XXXX >
2135 available" or some LBUG about grant, when the disk is almost
2137 Details : In filter_check_grant, for non_grant cache write, we should
2138 check the left space by if (*left > ungranted + bytes), instead
2139 of (*left > ungranted), because only we are sure the left space
2140 is enough for another "bytes", then the ungrant space should be
2141 increase. In client, we should update cl_avail_grant only there
2142 is OBD_MD_FLGRANT in the reply.
2146 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2147 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2148 described in CVE-2007-4573. This problem could allow untrusted
2149 local users to gain root access.
2152 Frequency : when using O_DIRECT and quotas
2154 Description: Incorrect file ownership on O_DIRECT output files
2155 Details : block usage reported by 'lfs quota' does not take into account
2156 files that have been written with O_DIRECT.
2161 Description: touch file failed when fs is not full
2162 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2163 otherwise we can get ENOSP while fs is not full.
2167 Description: parallel lock callbacks
2168 Details : Instead of sending blocking and completion callbacks as separated
2169 requests, adding them to a set and sending in parallel.
2172 Frequency : only for Cray XT3
2173 Bugzilla : 12829/13455
2174 Description: Changing primary group doesn't change the group lustre assigns to
2176 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2177 overridden with the primary group provided by the group upcall,
2178 whereas the supplied fsgid can be trusted if it is in the list of
2179 supplementary groups returned by the group upcall.
2181 Severity : enhancement
2183 Description: Allow masking D_WARNING, D_ERROR messages from console
2184 Details : Console messages can now be disabled via lnet.printk.
2188 Description: User code with malformed file open parameter crashes client node
2189 Details : Before packing join_file req, all the related reference should be
2190 checked carefully in case some malformed flags cause fake
2191 join_file req on client.
2195 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2196 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2197 mode is equal to the granted mode. However, it should also check
2198 that LDLM_FL_CP_REQD is not set.
2202 Description: Heavy nfs access might result in deadlocks
2203 Details : After ELC code landed, it is now improper to enqueue any mds
2204 locks under och_sem, because enqueue might want to decide to
2205 cancel open locks for same inode we are holding och_sem for.
2209 Description: Client eviction while running blogbench
2210 Details : A lot of unlink operations with concurrent I/O can lead to a
2211 deadlock causing evictions. To address the problem, the number of
2212 oustanding OST_DESTROY requests is now throttled to
2213 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2214 callbacks are processed in priority.
2218 Description: enable ACLs on MDS by default
2219 Details : ACLs must be enabled on MDS by default.
2222 Frequency : PPC/PPC64 only
2224 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2225 Details : fix duplicated definitions between asm-ppc64/types.h and
2226 lustre_types.h on PPC.
2229 Frequency : PPC/PPC64 only
2231 Description: asm-ppc/segment.h does not exist
2232 Details : fix compile issue on PPC.
2236 Description: better handle error messages in extents code
2239 Frequency : RHEL4 only
2241 Description: mkfs is very slow on IA64/RHEL4
2242 Details : A performance regression has been discovered in the MPT Fusion
2243 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2244 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2245 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2247 Severity : enhancement
2249 Description: SNMP support enhancement
2250 Details : Adding total number of sampled request for an MDS node in snmp
2253 Severity : enhancement
2255 Description: Optimize ldlm waiting list processing for PR extent locks
2256 Details : When processing waiting list for read extent lock and meeting read
2257 lock that is same or wider to it that is not contended, skip
2258 processing rest of the list and immediatelly return current
2259 status of conflictness, since we are guaranteed there are no
2260 conflicting locks in the rest of the list.
2264 Description: Time out and refuse to reconnect
2265 Details : When the failover node is the primary node, it is possible
2266 to have two identical connections in imp_conn_list. We must
2267 compare not conn's pointers but NIDs, otherwise we can defeat
2268 connection throttling.
2272 Description: port llog fixes from b1_6 into HEAD
2273 Details : Port llog reference couting and some llog cleanups from b1_6
2274 (bug 10800) into HEAD, for protect from panic and access to already
2275 free llog structures.
2279 Description: Detect stride IO mode in read-ahead
2280 Details : When a client does stride read, read-ahead should detect that and
2281 read-ahead pages according to the detected stride pattern.
2285 Description: data checksumming impacts single node performance
2286 Details : add support for several checksum algorithm. Currently, only CRC32
2287 and Adler-32 are supported. The checksum type can be changed on
2288 the fly via /proc/fs/lustre/osc/*/checksum_type.
2292 Description: use adler32 for page checksums
2293 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2298 Description: build for x2 fails
2299 Details : fix compile issue on Cray systems.
2303 Description: Properly match for duplicate locks
2304 Details : Due to different lock order from skiplists code, we need to
2305 traverse entire list for now
2308 Frequency : only on PPC/SLES10
2310 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2311 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2312 be defined. Add a hack in mkfs_lustre.c to work around this compile
2317 Description: LASSERT on MDS when client holding flock lock dies
2318 Details : ldlm pool logic depends on number of granted locks equal to
2319 number of released locks which is not true for flock locks, so
2320 just exclude such locks from consideration.
2324 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2325 Details : Use fsfilt_sync() for both the whole filesystem sync and
2326 individual file sync to eliminate dangerous inode locking
2327 with I_LOCK that can lead to a deadlock.
2331 Description: performance in 1.6.3
2332 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2333 sure that 1MB requests are not fragmented by the block layer.
2335 Severity : enhancement
2337 Description: organize the server-side client stats on per-nid basis
2338 Details : Change the structure of stats under obdfilter and mds to
2346 The "uuid"s file would list the uuids of _active_ exports.
2347 And the clear entry is to clear all stats and stale nids.
2349 Severity : enhancement
2351 Description: eliminate client locks in face of contention
2352 Details : file contention detection and lockless i/o implementation
2353 for contended files.
2357 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2360 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2362 Description: MDS may lose file striping (and hence file data) in some cases
2363 Details : If there are additional extended attributes stored on the MDS,
2364 in particular ACLs, SELinux, or user attributes (if user_xattr
2365 is specified for the client mount options) then there is a risk
2366 of attribute loss. Additionally, the Lustre file striping
2367 needs to be larger than default (e.g. striped over all OSTs),
2368 and an additional attribute must be stored initially in the
2369 inode and then increase in size enough to be moved to the
2370 external attribute block (e.g. ACL growing in size) for file
2375 Description: skiplist implementation simplification
2376 Details : skiplists are used to group compatible locks on granted list
2377 that was implemented as tracking first and last lock of each
2378 lock group the patch changes that to using doubly linked lists
2382 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2383 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2384 because LASSERT on both the data supplied by a client, and the
2385 data on disk is dangerous and incorrect.
2387 Severity : enhancement
2389 Description: Slow truncate/writes to huge files at high offsets.
2390 Details : Directly associate cached pages to lock that protect those pages,
2391 this allows us to quickly find what pages to write and remove
2392 once lock callback is received.
2396 Description: more ldlm soft lockups
2397 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2398 starve other threads from the resource lock for a long time in
2399 case of long waiting queue, so change the debug level from
2400 D_OTHER to the less frequently used D_INFO.
2402 Severity : enhancement
2404 Description: add -gid, -group, -uid, -user options to lfs find
2408 Description: Hung threads in invalidate_inode_pages2_range
2409 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2410 once one is completed. As a result, some RPCs are stuck in the
2411 queue and are never sent.
2415 Description: filter threads hungs on waiting journal commit
2416 Details : Cleanup filter group llog code, then only filter group llog will
2417 be only created in the MDS/OST syncing process.
2421 Description: Procfs and llog threads access destoryed import sometimes.
2422 Details : Sync the import destoryed process with procfs and llog threads by
2423 the import refcount and semaphore.
2425 Severity : enhancement
2427 Description: openlock cache of b1_6 port to HEAD
2432 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2433 Details : If insertion of an extent fails, then discard the inode
2434 preallocation and free data blocks else it can lead to duplicate
2439 Description: don't always update ctime in ext3_xattr_set_handle()
2440 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2441 In some cases the ctime should not be updated, for example for
2442 2.0->1.8 compatibility it is necessary to delete an xattr and it
2443 should not update the ctime.
2447 Bugzilla : 15713/16362
2448 Description: Assertion in iopen_connect_dentry in 1.6.3
2449 Details : looking up an inode via iopen with the wrong generation number can
2450 populate the dcache with a disconneced dentry while the inode
2451 number is in the process of being reallocated. This causes an
2452 assertion failure in iopen since the inode's dentry list contains
2453 both a connected and disconnected dentry.
2457 Description: assertion failure in ldlm_handle2lock()
2458 Details : fix a race between class_handle_unhash() and class_handle2object()
2459 introduced in lustre 1.6.5 by bug 13622.
2464 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2465 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2470 Description: oops in sd_iostats_seq_show()
2471 Details : unloading/reloading the scsi low level driver triggers a kernel
2472 bug when trying to access the sd iostat file.
2477 Description: Kernel panics during QLogic driver reload
2478 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2479 patch, causing memory corruption.
2484 Description: journal_dev option does not work in b1_6
2485 Details : pass mount option during pre-mount.
2487 Severity : enhancement
2489 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2490 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2491 extent information of a file. It can be used to map logical blocks
2492 in a file to physical blocks in the block device.
2496 Description: LDLM soft lockups - improvement
2497 Details : It is be possible to send the lock handle along with each read
2498 or write request because the client is already doing a lock match
2499 itself so there isn't any reason the OST should have to re-do that
2503 Frequency : only on Cray X2
2505 Description: X2 build failures
2506 Details : fix build failures on Cray X2.
2510 Description: xid & resent requests
2511 Details : Initialize RPC XID from clock at startup (randomly if clock is
2514 Severity : enhancement
2516 Description: Add lustre_start utility to start or stop multiple Lustre servers
2521 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2522 Details : In case of memory pressure, list_del() can be called twice on
2523 req->rq_history_list, causing a kernel oops.
2527 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2528 Details : fix stack overflow in the distributed lock manager by defering
2529 export eviction after a failed AST to the elt thread instead of
2530 handling it in the dlm interpret routine.
2534 Description: Convert some comments to new format.
2535 Details : Update documenting comments to match doxygen conventions.
2539 Description: Grammar fixes.
2540 Details : A couple of trivial spelling fixes.
2544 Description: OSD_COUNTERS-mandatory
2545 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2546 to simplify the code.
2550 Description: simplify cmm_device freeing logic.
2551 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2555 Description: Add lockdep support to dt_object_operations locking interface.
2556 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2557 parameter indicating lock ordering. Update mdd code to use new
2562 Description: Introduce failloc constants for lockless IO tests.
2563 Details : Add two new failloc constants to test lockless IO. Only one of
2564 them in implemented---another is checked in yet to be landed
2569 Description: Add lockdep support for inode mutex.
2570 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2571 in the situations where Lustre has to lock more than one inode
2576 Description: Add optional invariants checking support.
2577 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2578 configure switch. This macro is to be used for consistency and
2579 sanity checks that are too expensive to be left in `production'
2584 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2585 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2590 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2591 Details : Use spin_lock_nested() in (the only) situation where more than
2592 one ldlm_lock is locked simultaneously. Also, fix possible
2593 dead-lock in ldlm_lock_change_resource() by enforcing particular
2598 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2599 Details : Instead of specifying each ldlm_lock call-back through separate
2600 parameter, wrap them into struct ldlm_callback_suite.
2604 Description: Kill join_lru obd method and its callers.
2605 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2606 in memory---a requirement implemented through ->o_join_lru() obd
2607 method in HEAD. Remove this method and its users.
2611 Description: Add asynchronous ldlm ENQUEUE completion handler.
2612 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2613 case that stock ldlm_completion_ast() cannot handle as it waits
2614 until lock is granted. Introduce new ldlm_completion_ast_async()
2615 for this. Also comment ldlm_completion_ast().
2619 Description: ldlm_error <-> errno conversion.
2620 Details : Add functions to map (rather arbitrary) between LDLM error codes
2621 and standard errno values. CLIO needs this to prevent LDLM specific
2622 constants from escaping ldlm and osc.
2626 Description: Kill unused ldlm_handle2lock_ns() function.
2627 Details : Kill unused ldlm_handle2lock_ns() function.
2631 Description: Add lu_ref support to ldlm_lock
2632 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2633 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2634 and ldlm_resource. LDLM interface has to be changed, because code that
2635 releases a reference on a lock, has to "know" what reference this is.
2636 In the most frequent case
2638 lock = ldlm_handle2lock(handle);
2640 LDLM_LOCK_PUT(lock);
2642 no changes are required. When any other reference (received _not_ from
2643 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2644 instead of LDLM_LOCK_PUT().
2646 Arguably, changes are pervasive, and interface requires some discipline
2647 for proper use. On the other hand, it was very instrumental in finding
2648 a few leaked lock references.
2652 Description: Add ldlm_lock_addref_try().
2653 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2654 attempts to addref a lock that might be being canceled
2659 Description: Add ldlm_weigh_callback().
2660 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2661 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2662 hard-coded `number of pages' logic.
2666 Description: Add lockdep annotations to llog code.
2667 Details : Use appropriately tagged _nested() locking calls in the places
2668 where llog takes more than one ->lgh_lock lock.
2672 Description: Add loi_kms_set().
2673 Details : Wrap kms updates into a helper function.
2677 Description: Constify instances of struct lsm_operations.
2678 Details : Constify instances of struct lsm_operations.
2682 Description: lu_conf support.
2683 Details : On a server, a file system object is uniquely identified
2684 by a fid, which is sufficient to locate and load all object
2685 state (inode). On a client, on the other hand, more data are
2686 necessary instantiate an object. Change lu_object_find() and
2687 friends to take additional `lu_conf' argument describing object.
2688 Typically this includes layout information.
2692 Description: lu_context fixes.
2693 Details : Introduce new lu_context functions that are needed on the client
2694 side, where some system threads (ptlrpcd) are shared by multiple
2695 modules, and so cannot be stopped during module shutdown.
2699 Description: Add start and stop methods to lu_device_type_operations.
2700 Details : Introduce two new methods in lu_device_type_operations, that are
2701 invoked when first instance of a given type is created and last one
2702 is destroyed respectively. This is need by CLIO.
2706 Description: Add lu_ref support to struct lu_device.
2707 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2708 track leaked references.
2712 Description: Introduce lu_kmem_descr.
2713 Details : lu_kmem_descr and its companion interface allow to create
2714 and destroy a number of kmem caches at once.
2718 Description: Fix lu_object finalization race.
2719 Details : Fix a race between lu_object_find() finding an object and its
2720 concurrent finalization. This race is (most likely) not possible
2721 on the server, but might happen on the client.
2725 Description: Introduce lu_ref interface.
2726 Details : lu_ref is a debugging module allowing to track references to
2727 a given object. It is quite cpu expensive, and has to be
2728 explicitly enabled with --enable-lu_ref. See usage description
2733 Description: Factor lu_site procfs stats into a separate function.
2734 Details : Separate lu_site stats printing code into a separate function
2735 that can be reused on a client.
2739 Description: Constify instances of struct {lu,dt,md}_device_operations.
2740 Details : Constify instances of struct {lu,dt,md}_device_operations.
2744 Description: Introduce struct md_site and move meta-data specific parts of
2745 struct lu_site here.
2746 Details : Move md-specific fields out of struct lu_site into special struct
2747 md_site, so that lu_site can be used on a client.
2751 Description: Kill mdd_lov_destroy().
2752 Details : Remove unused mdd code.
2756 Description: Add st_block checking to multistat.c.
2757 Details : Add st_block checking to multistat.c.
2761 Description: Add lu_ref support to struct obd_device.
2762 Details : Add lu_ref tracking to obd_device.
2766 Description: Kill obd_set_fail_loc().
2767 Details : Remove unused code.
2771 Description: Add special type for ptlrpc_request interpret functions.
2772 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2773 there. Actual usage will be in CLIO.
2777 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2778 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2779 doesn't work with lockdep.
2783 Description: Add rwv.c test program.
2784 Details : New testing program exercising readv(2) and writev(2) (Qian).
2788 Description: Add sendfile.c test program.
2789 Details : New testing program exercising sendfile(2) (Jay).
2793 Description: Ratelimit a message that can be very frequent.
2794 Details : Ratelimit a memory allocation failure message that can
2799 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2800 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2804 Description: Liblustre build fixes.
2805 Details : Liblustre build fixes.
2809 Description: libcfs: add cfs_{need,cond}_resched() interface.
2810 Details : libcfs: add cfs_{need,cond}_resched() definition and
2811 implementations for Linux, NT, and liblustre.
2813 Severity : enhancement
2815 Description: More exported tunables for mballoc
2816 Details : Add support for tunable preallocation window and new tunables for
2817 large/small requests
2821 Description: Detect corruption of block bitmap and checking for preallocations
2822 Details : Checks validity of on-disk block bitmap. Also it does better
2823 checking of number of applied preallocations. When corruption is
2824 found, it turns filesystem readonly to prevent further corruptions.
2828 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2829 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2830 window to 0. If it does hit cache miss, and read-pattern is still
2831 stride-io mode, does not reset the stride window, but also does not
2832 increase the stride window length in this case.
2836 Frequency : only for big-endian servers
2837 Description: Check if system is big-endian while mounting fs with extents feature
2838 Details : Mounting a filesystem with extents feature will fail on big-endian
2839 systems since ext3-based ldiskfs is not supported on big-endian
2840 systems. This can be over-riden with "bigendian_extents" mount option.
2842 Severity : enhancement
2844 Description: The root squash functionality
2845 Details : A security feature, which is to prevent users from being able
2846 to mount lustre on their desktop, run as root, and delete
2847 all of the files in the filesystem. The goal is accomplished by
2848 remapping user id (UID) and group id (GID) of the root user to
2849 a UID and GID specified by the system administartor via Lustre
2850 configuration management server (MGS). The functionality also
2851 allows to specify sets of clients for which the remapping does
2856 Description: Excessive recovery window
2857 Details : With AT enabled, the recovery window can be excessively long (6000+
2858 seconds). To address this problem, we no longer use
2859 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2860 timeout no longer depends on the service time, it is set to
2861 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2862 time via pb_service_time.
2866 Description: Watchdog triggered on MDS failover
2867 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2868 the OSTs know that the MDS "UUID" can be reused for the same export
2869 from a different NID, so we do not need to wait for the export to be
2873 Frequency : rare, only if using MMP with Linux RAID
2875 Description: MMP doesn't work with Linux RAID
2876 Details : While using HA for Lustre servers with Linux RAID, it is possible
2877 that MMP will not detect multiple mounts. To make this work we
2878 need to unplug the device queue in RAID when the MMP block is being
2879 written. Also while reading the MMP block, we should read it from
2880 disk and not the cached one.
2882 Severity : enhancement
2884 Description: open file using fid
2885 Details : A file can be opened using just its fid, like
2886 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2889 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2892 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2893 Details : add proper locking for creation of jbd_4k slab cache
2897 Description: MMP check in ext3_remount() fails without displaying any error
2898 Details : When multiple mount protection fails during remount, proper error
2901 Severity : enhancement
2903 Description: Allow stripe size to be up to 4G-64k
2904 Details : Fix math logic to allow large stripe sizes.
2908 Description: add check for >8TB ldiskfs filesystems
2909 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2910 Don't allow >8TB ldiskfs filesystems to be mounted without
2911 force_over_8tb mount option
2916 Descriptoin: OSS complains "too many blocks" and crashes
2917 Details : In filter_commitrw_write():
2918 1. lquota_chkquota();
2919 2. obdo_from_inode();
2920 3. lquota_pending_commit();
2921 "2." has changed uid and gid in this case, then caused
2922 lquota_chkquota() and lquota_pending_commit() to use
2923 different uid and gid. That is the root of the bug.
2927 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2928 Details : Detect and print out the 'stripe count = -1' case properly.
2932 Description: Out or order replies might be lost on replay
2933 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2934 request with tid smaller than one currently being added, add it
2935 to the start, not end of the list.
2937 --------------------------------------------------------------------------------
2939 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2941 * Support for kernels:
2942 2.6.5-7.283 (SLES 9),
2943 2.6.9-55.EL (RHEL 4),
2944 2.6.16.46-0.14 (SLES 10),
2945 2.6.18.8 vanilla (kernel.org)
2946 * Client support for unpatched kernels:
2947 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2948 2.6.16 - 2.6.22 vanilla (kernel.org)
2949 * Due to recently discovered recovery problems, we do not recommend
2950 using patchless RHEL 4 clients with this or any earlier release.
2951 * Recommended e2fsprogs version: 1.39.cfs8
2952 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2953 * Starting with this release, the ldiskfs backing filesystem required
2954 by Lustre is now in its own package, lustre-ldiskfs. This package
2955 should be installed. It is versioned separately from Lustre and
2956 may be released separately in future.
2960 Description: block reactivating mgc import until all deactivates complete
2961 Details : Fix race when failing back MDT/MGS to itself (testing)
2964 Frequency : at statup only
2966 Description: mds_lov_synchronize race leads to various problems
2967 Details : simultaneous MDT->OST connections at startup can cause the
2968 sync to abort, leaving the OSC in a bad state.
2970 Severity : enhancement
2972 Description: add optional extra BUILD_VERSION info
2973 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2974 to override the lustre version.
2977 Frequency : 2.6.18 servers only
2979 Description: ll_kern_mount() doesn't release the module reference
2980 Details : The ldiskfs module reference count never drops down to 0
2981 because ll_kern_mount() doesn't release the module reference.
2986 Description: server LBUG when using old ost_num_threads parameter
2987 Details : Accept the old ost_num_threads parameter but warn that it
2988 is deprecated, and fix an off-by-one error that caused an LBUG.
2993 Description: Transient SCSI error results in persistent IO issue
2994 Details : iobuf->dr_error is not reinitialized to 0 between two
2998 Frequency : sometimes when underlying device returns I/O errors
3000 Description: OSTs not going read-only during write failures
3001 Details : OSTs are not remounted read-only when the journal commit threads
3002 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3003 instead of the ext3 wrappers.
3007 Description: poor connect scaling with increasing client count
3008 Details : Don't run filter_grant_sanity_check for more than 100 exports
3009 to improve scaling for large numbers of clients.
3012 Frequency : SLES10 only
3014 Description: sanity-quota.sh quotacheck failed: rc = -22
3015 Details : Quotas cannot be enabled on SLES10.
3018 Frequency : liblustre clients only
3020 Description: getdirentries does not give error when run on compute nodes
3021 Details : getdirentries does not fail when the size specified as an argument
3022 is too small to contain at least one entry
3024 Severity : enhancement
3026 Description: Add LNET router traceability for debug purposes
3027 Details : If a checksum failure occurs with a router as part of the
3028 IO path, the NID of the last router that forwarded the bulk data
3029 is printed so it can be identified.
3034 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3035 Details : Due to a race condition, liblustre clients were occasionally
3036 evicted incorrectly.
3038 Severity : enhancement
3040 Description: lfs setstripe use optional parameters instead of postional
3043 Severity : enhancement
3045 Description: Nanosecond timestamp support for ldiskfs
3046 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3047 resolution timestamps. There is not yet support for this at
3048 the Lustre filesystem level.
3051 Frequency : during server recovery
3053 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3054 Details : request with rq_no_resend flag not awake l_wait_event if they get
3058 Frequency : nfs export on patchless client
3060 Description: connectathon hang when test nfs export over patchless client
3061 Details : Disconnected dentry cannot be found with lookup, so we do not need
3062 to unhash it or make it invalid
3065 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3066 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3067 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3072 Description: open req refcounting wrong on reconnect
3073 Details : If reconnect happened between getting open reply from server and
3074 call to mdc_set_replay_data in ll_file_open, we will schedule
3075 replay for unreferenced request that we are about to free.
3076 Subsequent close will crash in variety of ways.
3077 Check that request is still eligible for replay in
3078 mdc_set_replay_data().
3083 Description: disable writes to filesystem when reading health_check file
3084 Details : the default for reading the health_check proc file has changed
3085 to NOT do a journal transaction and write to disk, because this
3086 can cause reads of the /proc file to hang and block HA state
3087 checking on a healthy but otherwise heavily loaded system. It
3088 is possible to return to the previous behaviour during configure
3089 with --enable-health-write.
3091 Severity : enhancement
3093 Description: 64-bit inode version
3094 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3095 to the inode. This will be required for version-based recovery.
3100 Description: MDS fails to start if a duplicate client export is detected
3101 Details : in some rare cases it was possible for a client to connect to
3102 an MDS multiple times. Upon recovery the MDS would detect this
3103 and fail during startup. Handle this more gracefully.
3105 Severity : enhancement
3107 Description: Add -o localflock option to simulate old noflock
3109 Details : This will achieve local-only flock/fcntl locks
3115 Description: log_commit_thread vs filter_destroy race leads to crash
3116 Details : Take import reference before releasing llog record semaphore
3121 Description: Wrong request locking in request set processing
3122 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3123 imp_delayed_list, in this place should be used imp_lock.
3126 Frequency : when reconnection
3128 Description: Grant Leak when osc reconnect to OST
3129 Details : When osc reconnect ost, OST(filter) should check whether it
3130 should grant more space to client by comparing fed_grant and
3131 cl_avail_grant, and return the granted space to client instead
3132 of "new granted" space, because client will call osc_init_grant
3133 to update the client grant space info.
3136 Frequency : when client reconnect to OST
3138 Description: Grant Leak when osc do resend and replay bulk write
3139 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3140 bulk write request, because the grant info while be sync between
3141 OSC and OST when reconnect, and we should ignore the grant info
3142 these of resend/replay write req.
3147 Description: Grant space more than available left space sometimes.
3148 Details : When then OST is about to be full, if two bulk writing from
3149 different clients came to OST. Accord the avaliable space of the
3150 OST, the first req should be permitted, and the second one
3151 should be denied by ENOSPC. But if the seconde arrived before
3152 the first one is committed. The OST might wrongly permit second
3153 writing, which will cause grant space > available space.
3156 Frequency : when client is evicted
3158 Description: Grant might be wrongly erased when osc is evicted by OST
3159 Details : when the import is evicted by server, it will fork another
3160 thread ptlrpc_invalidate_import_thread to invalidate the
3161 import, where the grant will be set to 0. While the original
3162 thread will update the grant it got when connecting. So if
3163 the former happened latter, the grant will be wrongly errased
3164 because of this race.
3169 Description: Checking Stale with correct fid
3170 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3171 to check whether it is stale, because sometimes, we want the
3172 enqueue happened anyway, and op_data.fid2 will not be initialized.
3174 Severity : enhancement
3176 Description: update patchless client
3177 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3180 Frequency : only with 2.4 kernel
3182 Description: random memory corruption
3183 Details : size of struct ll_inode_info is to big for union inode.u and this
3184 can be cause of random memory corruption.
3189 Description: Memory leak in recovery
3190 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3191 It should also check obd_fail before fsfilt_start, otherwise if
3192 fsfilt_start return -EROFS,(failover mds during mds recovery).
3193 then the req will return with repmsg->transno = 0 and rc = EROFS.
3194 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3195 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3196 be freed no matter whether fsfilt_commit success or not.
3199 Frequency : only with huge count clients
3201 Description: Prevents from taking the superblock lock in llap_from_page for
3203 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3204 for llap_from_page call in ll_removepage prevents from taking the
3205 superblock lock for a soon died page.
3210 Description: Not check open intent error before release open handle
3211 Details : in some rare cases, the open intent error is not checked before
3212 release open handle, which may cause
3213 ASSERTION(open_req->rq_transno != 0), because it tries to release
3214 the failed open handle.
3219 Description: Set cat log bitmap only after create log success.
3220 Details : in some rare cases, the cat log bitmap is set too early. and it
3221 should be set only after create log success.
3225 Description: Accessing a block bevice can re-enable I/O when Lustre is
3226 tearing down a device.
3227 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3232 Description: service threads may hog cpus when there are a lot of requests
3233 Details : Insert cond_resched to give other threads a chance to use some CPU
3238 Description: the cat log was not initialized in recovery
3239 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3240 unlink log on mds will not be initialized until mds post
3241 recovery. And also in mds post recovery, the unlink log will
3242 initialization will be done asynchronausly, so there will be race
3243 between add unlink log and unlink log initialization.
3247 Description: brw_stats were being printed incorrectly
3248 Details : brw_stats were being printed as log2 but all of them were not
3249 recorded as log2. Also remove some code duplication arising from
3250 filter_tally_{read,write}.
3254 Frequency : rare, only in recovery.
3255 Description: ASSERTION(req->rq_type != LI_POISON) failed
3256 Details : imp_lock should be held while iterating over imp_sending_list for
3257 prevent destroy request after get timeout in ptlrpc_queue_wait.
3261 Description: replay-single.sh test 52 fails
3262 Details : A lock's skiplist need to be cleanup when it being unlinked
3263 from its resource list.
3267 Description: Short directio read returns full requested size rather than
3269 Details : Direct I/O operations should return actual amount of bytes
3270 transferred rather than requested size.
3272 Severity : enhancement
3274 Description: metadata RPC reduction (e.g. for rm performance)
3275 Details : decrease the amount of synchronous RPC between clients and servers
3276 by canceling conflicing lock before the operation on the client
3277 and packing thier handles into the main operation RPC to server.
3279 Severity : enhancement
3281 Description: Async OSC create to avoid the blocking unnecessarily.
3282 Details : If an OST has no remaining object, system will block on the
3283 creation when it needs to create a new object on this OST. Now,
3284 ways use pre-created objects when available, instead of blocking on
3285 an empty osc while others are not empty. If we must block, we
3286 block for the shortest possible period of time.
3290 Description: improve handling recoverable errors
3291 Details : if request processig with error which can be recoverable on server
3292 request should be resend, otherwise page released from cache and
3295 Severity : enhancement
3297 Description: refine locking for avoid write wrong info into lov_objid file
3298 Details : fix possible races with add new target and write/update data in
3301 --------------------------------------------------------------------------------
3303 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3308 Frequency : on some architectures
3310 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3311 Details : Uninitialized flags sometimes cause configuration commands to
3315 Frequency : patchless clients only
3317 Description: missing __iget() symbol export
3318 Details : The __iget() symbol export is missing. To avoid the need for
3319 this on patchless clients the deathrow inode reaper is turned
3320 off, and we depend on the VM to clean up old inodes. This
3321 dependency was during via the fix for bug 12181.
3323 --------------------------------------------------------------------------------
3325 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3327 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3328 INTEROPERATE with older versions automatically. In many cases a
3329 special upgrade step is needed. Please read the
3330 user documentation before upgrading any part of a live system.
3331 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3332 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3333 * WARNING: Lustre configuration and startup changes are required with
3334 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3338 Severity : enhancement
3340 Description: MountConf
3341 Details : Lustre configuration is now managed via mkfs and mount
3342 commands instead of lmc and lconf. New obd types (MGS, MGC)
3343 are added for dynamic configuration management. See
3344 https://mail.clusterfs.com/wikis/lustre/MountConf for
3347 Severity : enhancement
3349 Description: dynamic OST addition
3350 Details : OSTs can now be added to a live filesystem
3352 Severity : enhancement
3354 Description: startup order invariance
3355 Details : MDTs and OSTs can be started in any order. Clients only
3356 require the MDT to complete startup.
3358 Severity : enhancement
3360 Description: parallel, asynchronous orphan cleanup
3361 Details : orphan cleanup is now performed in separate threads for each
3362 OST, allowing parallel non-blocking operation.
3364 Severity : enhancement
3366 Description: optimized stripe assignment
3367 Details : stripe assignments are now made based on ost space available,
3368 ost previous usage, and OSS previous usage, in order to try
3369 to optimize storage space and networking resources.
3371 Severity : enhancement
3373 Description: Permanently set tunables
3374 Details : All writable /proc/fs/lustre tunables can now be permanently
3375 set on a per-server basis, at mkfs time or on a live system.
3377 Severity : enhancement
3379 Description: Lustre message v2
3380 Details : Add lustre message format v2.
3382 Severity : enhancement
3384 Description: client OST exclusion list
3385 Details : Clients can be started with a list of OSTs that should be
3386 declared "inactive" for known non-responsive OSTs.
3390 Description: SPEC SFS validation failure on NFS v2 over lustre.
3391 Details : Changes the blocksize for regular files to be 2x RPC size,
3392 and not depend on stripe size.
3394 Severity : enhancement
3396 Description: Multiple MD RPCs in flight.
3397 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3398 To avoid overly-overloading MDS, introduce a limit on number of
3399 MDS RPCs in flight for a single client and add /proc controls
3400 to adjust this limit.
3402 Severity : enhancement
3404 Description: client read/write statistics
3405 Details : Add client read/write call usage stats for performance
3406 analysis of user processes.
3407 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3408 file access. extents_stats shows chunk size distribution.
3409 extents_stats_per_process show chunk size distribution per
3412 Severity : enhancement
3414 Description: mds statistics
3415 Details : Add detailed mds operations statistics in
3416 /proc/fs/lustre/mds/*/stats.
3420 Description: Failure of copying files with lustre special EAs.
3421 Details : Client side always return success for setxattr call for lustre
3422 special xattr (currently only "trusted.lov").
3427 Description: Refcount LNET uuids
3428 Details : The global LNET uuid list grew linearly with every startup;
3429 refcount repeated list entries instead of always adding to
3432 Severity : enhancement
3434 Description: Dynamic service threads
3435 Details : Within a small range, start extra service threads
3436 automatically when the request queue builds up.
3439 Frequency : mixed-endian client/server environments
3441 Description: mixed-endian crashes
3442 Details : The new msg_v2 system had some failures in mixed-endian
3445 Severity : enhancement
3447 Description: Easy OST removal
3448 Details : OSTs can be permanently deactivated with e.g. 'lctl
3449 conf_param lustre-OST0001.osc.active=0'
3451 Severity : enhancement
3453 Description: MGS proc entries
3454 Details : Added basic proc entries for the MGS showing what filesystems
3457 Severity : enhancement
3459 Description: provide MGS failover
3460 Details : Added config lock reacquisition after MGS server failover.
3462 Severity : enhancement
3464 Description: add Linux 2.4 support
3465 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3469 Description: a large application tries to do I/O to the same resource and dies
3470 in the middle of it.
3471 Details : Check the req->rq_arrival time after the call to
3472 ost_brw_lock_get(), but before we do anything about
3473 processing it & sending the BULK transfer request. This
3474 should help move old stale pending locks off the queue as
3475 quickly as obd_timeout.
3478 Frequency : when an incorrect nid is specified during startup
3480 Description: ptlrpc connect to non-existant node causes kernel crash
3481 Details : LNET can't be re-entered from an event callback, which
3482 happened when we expire a message after the export has been
3483 cleaned up. Instead, hand the zombie cleanup off to another
3486 Severity : enhancement
3488 Description: plain/inodebits lock performance improvement
3489 Details : Grouping plain/inodebits in granted list by their request modes
3490 and bits policy, thus improving the performance of search through
3494 Frequency : only if OST filesystem is corrupted
3496 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3497 Details : for a short time RPCs with bulk IO are in the replay list,
3498 but replay of bulk IOs is unimplemented. If the OST filesystem
3499 is corrupted due to disk cache incoherency and then replay is
3500 started it is possible to trip an assertion. Avoid putting
3501 committed RPCs into the replay list at all to avoid this issue.
3504 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3506 Description: System hang on startup
3507 Details : This bug allowed the liblustre (e.g. catamount) client to
3508 return to the app before handling all startup RPCs. This
3509 could leave the node unresponsive to lustre network traffic
3510 and manifested as a server ptllnd timeout.
3512 Severity : enhancement
3514 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3515 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3516 boolean to control whether to print peer debug info when a
3517 client's RPC times out.
3520 Frequency : only for kernels with patches from Lustre below 1.4.3
3522 Description: Remove old rdonly API
3523 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3526 Frequency : only for devices with external journals
3528 Description: Set external device read-only also
3529 Details : During a commanded failover stop, we set the disk device
3530 read-only while the server shuts down. We now also set any
3531 external journal device read-only at the same time.
3534 Frequency : when upgrading from 1.4 while trying to change parameters
3536 Description: The wrong (new) MDC name was used when setting parameters for
3537 upgraded MDT's. Also allows changing of OSC (and MDC)
3538 parameters if --writeconf is specified at tunefs upgrade time.
3541 Frequency : when setting specific ost indicies
3543 Description: QOS code breaks on skipped indicies
3544 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3545 created with --index need not be sequential.
3547 Severity : enhancement
3549 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3550 Details : The uninit_groups feature works in conjunction with the kernel
3551 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3552 up the pass1 processing of e2fsck. This is a read-only feature
3553 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3554 mount filesystems that have had this feature enabled.
3556 Severity : enhancement
3558 Description: Improve multi-block allocation algorithm to avoid fragmentation
3559 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3560 allocation locality and avoid filesystem fragmentation.
3562 ------------------------------------------------------------------------------
3564 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3566 * Support for kernels:
3567 2.4.21-47.0.1.EL (RHEL 3)
3568 2.6.5-7.283 (SLES 9)
3569 2.6.9-42.0.10.EL (RHEL 4)
3570 2.6.12.6 vanilla (kernel.org)
3571 2.6.16.27-0.9 (SLES 10)
3572 * Recommended e2fsprogs version: 1.39.cfs5
3574 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3578 Frequency : occasional, depends on client load and configuration
3579 Bugzilla : 12181, 12203
3580 Description: data loss for recently-modified files
3582 Details : In some cases it is possible that recently written or created
3583 files may not be written to disk in a timely manner (this should
3584 normally be within 30s unless client IO load is very high).
3585 The problem appears as zero-length files or files that are a
3586 multiple of 1MB in size after a client crash or client eviction
3587 that are missing data at the end of the file.
3589 This problem is more likely to be hit on clients where files are
3590 repeatedly created and unlinked in the same directory, clients
3591 have a large amount of RAM, have many CPUs, the filesystem has
3592 many OSTs, the clients are rebooted frequently, and/or the files
3593 are not accessed by other nodes after being written.
3595 The presence of the problem can be detected by looking at
3596 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3597 smaller than the second (nr_unused) then dirty files will not
3598 be flushed automatically to disk. "sync; sleep 10" should be
3599 run several times on the node before unmounting it to update
3600 Lustre (this is also safe to run on nodes without this problem).
3602 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3603 that can cause this same problem, so customers using that kernel
3604 also need to update the kernel in addition to Lustre. In order
3605 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3607 It is normal that files written just before a client crash (less
3608 than 30s) may not yet have been flushed to disk, even for local
3612 Frequency : frequent on thin XT3 nodes
3614 Description: UUID collision on thin XT3 Linux nodes
3615 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3616 randomly, since we relied on an insufficiently-seeded PRNG.
3621 Description: OSS hangs after "All ost request buffers busy"
3622 Details : A deadlock between quota and journal operations caused OSS
3623 hangs after printing "All ost request buffers busy."
3626 Frequency : always on liblustre builds
3628 Description: Cleanup compiler warnings on liblustre
3631 Frequency : always on liblustre builds on XT3
3633 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3637 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3638 > 32000 subdirectories
3639 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3640 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3641 subdirectory count crosses 32000. This will aid e2fsck to
3642 correctly handle more than 32000 subdirectories.
3645 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3647 Description: System hang on startup
3648 Details : This bug allowed the liblustre (e.g. catamount) client to
3649 return to the app before handling all startup RPCs. This
3650 could leave the node unresponsive to lustre network traffic
3651 and manifested as a server ptllnd timeout.
3653 Severity : enhancement
3655 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3656 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3657 boolean to control whether to print peer debug info when a
3658 client's RPC times out.
3663 Description: make O_SYNC working on 2.6 kernels
3664 Details : 2.6 kernels use different method for mark pages for write,
3665 so need add a code to lustre for O_SYNC work.
3670 Description: Failure to close file and release space on NFS
3671 Details : Put inode details into lock acquired in ll_intent_file_open.
3672 Use mdc_intent_lock in ll_intent_open to properly
3673 detect all kind of errors unhandled by mdc_enqueue.
3678 Description: proc file read during shutdown sometimes raced obd removal,
3680 Details : Add lock to prevent obd access after proc file removal.
3683 Frequency : Only for files larger than 4GB on 32-bit clients.
3685 Description: improperly doing page alignment of locks
3686 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3687 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3692 Description: RPCs being resent when they shouldn't be
3693 Details : Some RPCs that should not be resent are being resent. This
3694 can cause inconsistencies in the RPC state machine. Do not
3695 resend such requests.
3698 Frequency : rare, only with NFS export
3700 Description: Crash on NFS re-export node
3701 Details : under very unusual load conditions an assertion is hit in
3702 ll_intent_file_open()
3705 Frequency : only if OST filesystem is corrupted
3707 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3708 Details : for a short time RPCs with bulk IO are in the replay list,
3709 but replay of bulk IOs is unimplemented. If the OST filesystem
3710 is corrupted due to disk cache incoherency and then replay is
3711 started it is possible to trip an assertion. Avoid putting
3712 committed RPCs into the replay list at all to avoid this issue.
3717 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3718 Details : Large single O_DIRECT read and write calls can fail to allocate
3719 a sufficiently large buffer to process the request. In case of
3720 allocation failure the allocation is retried with a smaller
3721 buffer and broken into smaller requests.
3723 Severity : enhancement
3725 Description: Add -o localflock option to simulate old noflock behaviour.
3726 Details : This will achieve local-only flock/fcntl locks coherentness.
3731 Description: versioning check is incomplete
3732 Details : Checking the version difference of client vs. server, report
3733 error if the gap is too big.
3738 Description: add support PG_writeback bit
3739 Details : add support for PG_writeback bit for Lustre, for more carefull
3740 work with page cache in 2.6 kernel. This also fix some deadlocks
3741 and remove hack for work O_SYNC with 2.6 kernel.
3743 Severity : enhancement
3745 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3746 Details : The uninit_groups feature works in conjunction with the kernel
3747 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3748 up the pass1 processing of e2fsck. This is a read-only feature
3749 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3750 mount filesystems that have had this feature enabled.
3752 Severity : enhancement
3754 Description: Improve multi-block allocation algorithm to avoid fragmentation
3755 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3756 allocation locality and avoid filesystem fragmentation.
3758 ------------------------------------------------------------------------------
3760 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3762 * Support for kernels:
3763 2.6.9-42.0.3.EL (RHEL 4)
3764 2.6.5-7.276 (SLES 9)
3765 2.4.21-47.0.1.EL (RHEL 3)
3766 2.6.12.6 vanilla (kernel.org)
3767 2.6.16.21-0.8 (SLES10)
3768 * Recommended e2fsprogs version: 1.39.cfs2-0
3770 * The backwards-compatible /proc/sys/portals symlink has been removed
3771 in this release. Before upgrading, please ensure that you change
3772 any configuration scripts or /etc/sysctl.conf files that access
3773 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3774 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3775 in advance of the upgrade on any system running Lustre 1.4.6 or
3776 newer, since /proc/sys/lnet was added in that version.
3777 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3781 Frequency : only when quota is used
3783 Description: avoid scanning export list for quota master
3784 Details : Change the algorithms to avoid scanning export list in order
3785 to improve the efficiency.
3788 Frequency : MDS failover only, very rarely
3790 Description: "went back in time" messages on mds failover
3791 Details : The greatest transno may be lost when the current operation
3792 finishes with an error (transno==0) and the client's last_rcvd
3793 record is over-written. Save the greatest transno in the
3794 mds_last_transno for this case.
3797 Frequency : always for specific kernels and striping counts
3799 Description: client may get "Matching packet too big" without ACL support
3800 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3801 when trying to access files in certain configurations. The
3802 clients should in fact be denied when mounting because they do
3803 not understand ACLs.
3806 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3808 Description: many clients connecting with IO in progress causes connect timeouts
3809 Details : Avoid synchronous journal commits to avoid delays caused by many
3810 clients connecting/disconnecting when bulk IO is in progress.
3811 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3812 OST_IO_PORTAL to avoid delays behind potentially many pending
3816 Frequency : occasionally with multiple writers to a single file
3818 Description: shared writes to file may result in wrong size reported by stat()
3819 Details : Allow growing of kms when extent lock is cancelled
3822 Frequency : always with random mmap IO to multi-striped file
3824 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3825 Details : Only if the hole is at the end of OST object so that kms is too
3826 small. Fix is to increase kms accordingly in ll_nopage.
3829 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3831 Description: writes to a missing object would leak memory on the OST
3832 Details : If there is an inconsistency between the MDS and OST filesystems,
3833 such that the MDS references an object that doesn't exist, writes
3834 to that object will leak memory due to incorrect cleanup in the
3835 error handling path, eventually running out of memory on the OST.
3840 Description: Creating too long symlink causes lustre errors
3841 Details : Check symlink and name lengths before sending requests to MDS.
3844 Frequency : only if flock is enabled (not on by default)
3846 Description: posix locks not released on fd closure on 2.6.9+
3847 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3848 kernels, this caused such locks not to be released on fd close and
3849 then assertions on fs unmount about still used locks.
3852 Frequency : MDS failover only, very rarely
3854 Description: clients may get ASSERTION(granted_lock != NULL)
3855 Details : When request was taking a long time, and a client was resending
3856 a getattr by name lock request. The were multiple lock requests
3857 with the same client lock handle and
3858 mds_getattr_name->fixup_handle_for_resent_request found one of the
3859 lock handles but later failed with ASSERTION(granted_lock != NULL).
3864 Description: handle->h_buffer_credits > 0, assertion failure
3865 Details : h_buffer_credits is zero after truncate, causing assertion
3866 failure. This patch extends the transaction or creates a new
3870 Frequency : NFS re-export or patchless client
3871 Bugzilla : 11179, 10796
3872 Description: Crash on NFS re-export node (__d_move)
3873 Details : We do not want to hash the dentry if we don't have a lock.
3874 But if this dentry is later used in d_move, we'd hit uninitialised
3875 list head d_hash, so we just do this to init d_hash field but
3876 leave dentry unhashed.
3879 Frequency : NFS re-export or patchless client
3881 Description: NFS exports has problem with symbolic link
3882 Details : lustre client didn't properly install dentry when re-exported
3883 to NFS or running patchless client.
3886 Frequency : NFS re-export or patchless client
3888 Description: Various nfs/patchless fixes.
3889 Details : fixes reuse disconected alias for lookup process - this fixes
3890 warning "find_exported_dentry: npd != pd",
3891 fix permission error with open files at nfs.
3892 fix apply umask when do revalidate.
3895 Frequency : occasional
3897 Description: Crash on NFS re-export node
3898 Details : calling clear_page() on the wrong pointer triggered oops in
3899 generic_mapping_read().
3902 Frequency : rarely, using O_DIRECT IO
3904 Description: unaligned directio crashes client with LASSERT
3905 Details : check for unaligned buffers before trying any requests.
3908 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3910 Description: stale data returned from RAID cache
3911 Details : If only a small amount of IO is done to the RAID device before
3912 reading it again it is possible to get stale data from the RAID
3913 cache instead of reading it from disk.
3916 Frequency : always for sles10 kernel
3918 Description: sles10 support
3919 Details : ll_follow_link: compile fixes and using of nd_set_link
3920 under newer kernels.
3923 Frequency : depends on arch, kernel and compiler version, always on sles10
3926 Description: recursive or deep enough symlinks cause stack overflow
3927 Details : getting rid of large stack-allocated variable in
3931 Frequency : depends on hardware
3933 Description: lustre write performance loss in the SLES10 kernel
3934 Details : the performance loss is caused by using of write barriers in the
3935 ext3 code. The SLES10 kernel turns barrier support on by
3936 default. The fix is to undo that change for ldiskfs.
3938 ------------------------------------------------------------------------------
3940 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3942 * Support for kernels:
3943 2.6.9-42.0.3EL (RHEL 4)
3944 2.6.5-7.276 (SLES 9)
3945 2.4.21-47.0.1.EL (RHEL 3)
3946 2.6.12.6 vanilla (kernel.org)
3950 Frequency : quota enabled and large files being deleted
3952 Description: releasing more than 4GB of quota at once hangs OST
3953 Details : If a user deletes more than 4GB of files on a single OST it
3954 will cause the OST to spin in an infinite loop. Release
3955 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3960 Description: statfs data retrieved from /proc may be stale or zero
3961 Details : When reading per-device statfs data from /proc, in the
3962 {kbytes,files}_{total,free,avail} files, it may appear
3963 as zero or be out of date.
3966 Frequency : systems with MD RAID1 external journal devices
3968 Description: lconf's call to blkid is confused by RAID1 journal devices
3969 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3970 of returning all block devices that match the journal UUID.
3973 Frequency : always, for aggregate stripe size over 4GB
3975 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3976 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3977 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3980 Frequency : always if "lfs find" used on a local file/directory
3982 Description: "lfs find" segfaults if used on a local file/directory
3983 Details : The case where a directory component was not specified wasn't
3984 handled correctly. Handle this properly.
3987 Frequency : always on ppc64
3989 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3990 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3991 perform endianness conversion.
3994 Frequency : rarely (truncate to non-zero file size after write under load)
3995 Bugzilla : 10730, 10687
3996 Description: Files padded with zeros to next 4K multiple
3997 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3998 it is possible that files that are truncated to a non-zero size
3999 immediately after being written are filled with zero bytes beyond
4000 the truncated size. No file data is lost.
4002 Severity : enhancement
4004 Description: Allow recovery/failover for liblustre clients.
4005 Details : liblustre clients were unaware of failover configurations until
4008 Severity : enhancement
4010 Description: user file locks should fail when not mounting with flock option
4011 Details : Set up an error-returning stub in ll_file_operations.lock field
4012 to prevent incorrect behaviour when client is mounted without
4013 flock option. Also, set up properly f_op->flock field for
4017 Frequency : always on ia64
4019 Description: "lfs df" loops on printing out MDS statfs information
4020 Details : The obd_ioctl_data was not initialized and in some systems
4021 this caused a failure during the ioctl that did not return
4022 an error. Initialize the struct and return an error on failure.
4025 Frequency : SLES 9 only
4027 Description: Error of copying files with lustre special EAs as root
4028 Details : Client side always return success for setxattr call for lustre
4029 special xattr (currently only "trusted.lov").
4032 Frequency : rarely on clusters with both ia64+i386 clients
4034 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4035 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4036 clients are doing concurrent writes to the same file it is
4037 possible that the ia64 clients may LASSERT because the OST
4038 extent locks are not PAGE_SIZE aligned. Ensure that grown
4039 locks are always aligned on the request boundary.
4042 Frequency : specific use, occasional
4044 Description: Overwriting in use executable truncates on-disk binary image
4045 Details : If one node attempts to overwrite an executable in use by
4046 another node, we now correctly return ETXTBSY instead of
4047 truncating the file.
4049 Severity : enhancement
4051 Description: Async OSC create to avoid the blocking unnecessarily.
4052 Details : If an OST has no remaining object, system will block on the creating
4053 when need to create a new object on this OST. Now, ways use
4054 pre-created objects when available, instead of blocking on an
4055 empty osc while others are not empty. If we must block, we block
4056 for the shortest possible period of time.
4061 Description: chmod on Lustre root is propagated to other clients
4062 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4063 invalid by the follow_mount time.
4066 Frequency : liblustre clients only
4068 Description: Race in 'instant cancel' lock handling could lead to such locks
4069 never to be granted in case of SMP MDS
4070 Details : Do not destroy not yet granted but cbpending locks in
4074 Frequency : replay/resend of open
4076 Description: non null lock assetion failure in mds_intent_policy
4077 Details : Trying to replay/resend lockless open requests resulted in
4078 mds_open() returning 0 with no lock. Now it sets a flag if
4079 a lock is going to be returned.
4081 Severity : enhancement
4083 Description: Checksum enhancements
4084 Details : New checksum enhancements allow for resending RPCs that failed
4087 Severity : enhancement
4089 Description: Tunables on number of dirty pages in cacche
4090 Details : Allow to set limit on number of dirty pages cached.
4095 Description: client crash on unmount - lock still has references
4096 Details : In some error handling cases it was possible to leak a lock
4097 reference on a client while accessing a file. This was not
4098 harmful to the client during operation, but would cause the
4099 client to crash when the filesystem is unmounted.
4102 Frequency : specific case, rare
4104 Description: ETXTBSY on mds though file not in use by client
4105 Details : ETXTBSY is no longer incorrectly returned when attempting to
4106 chmod or chown a directory that the user previously tried to
4107 execute or a currently-executing binary.
4110 Frequency : extremely rare except on liblustre-based clients
4112 Description: Lustre space not freed when files are deleted
4113 Details : Clean up open-unlinked files after client eviction. Previously
4114 the unlink was skipped and the files remained as orphans.
4119 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4120 Details : In some cases it was possible to send a blocking callback to
4121 a client doing a glimpse, even though that client didn't get
4122 a lock granted. When the glimpse lock is cancelled on the OST
4123 the freed lock is left on the waiting list and corrupted the list.
4126 Frequency : all core dumps
4128 Description: Broke core dumps to lustre
4129 Details : Negative dentry may be unhashed if parent does not have UPDATE
4130 lock, but some callers, e.g. do_coredump, expect dentry to be
4131 hashed after successful create, hash it in ll_create_it.
4133 ------------------------------------------------------------------------------
4135 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4137 * Support for kernels:
4138 2.6.9-42.0.2.EL (RHEL 4)
4139 2.6.5-7.276 (SLES 9)
4140 2.4.21-40.EL (RHEL 3)
4141 2.6.12.6 vanilla (kernel.org)
4145 Frequency : always on RHEL 3
4147 Description: Number of open files grows over time
4148 Details : The number of open files grows over time, whether or not
4149 Lustre is started. This was due to a filp leak introduced
4150 by one of our kernel patches.
4152 ------------------------------------------------------------------------------
4154 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4156 * Support for kernels:
4157 2.6.9-42.EL (RHEL 4)
4158 2.6.5-7.276 (SLES 9)
4159 2.4.21-40.EL (RHEL 3)
4160 2.6.12.6 vanilla (kernel.org)
4165 Bugzilla : 5719, 9635, 9792, 9684
4166 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4167 Details : If a server is under heavy load and cannot reply to new
4168 connection requests before the client resends the (re)connect,
4169 the connection handling code can behave badly if two service
4170 threads are concurrently handing separate (re)connections from
4171 the same client. Add better locking to the connection handling
4172 code, and ensure that only a single connection will be processed
4173 for a given client UUID, even if the lock is dropped.
4175 Severity : enhancement
4177 Description: add TCP zero-copy support to kernel
4178 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4179 sends if the hardware supports scatter-gather and checksumming.
4180 This allows socklnd to do client-write and server-read more
4181 efficiently and reduce CPU utilization from skbuf copying.
4184 Frequency : only if NFS exporting from client
4186 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4187 Details : If setting attributes on a file created under NFS that had
4188 never been opened it would be possible to oops the client
4189 if the file had no objects.
4192 Frequency : always for liblustre
4194 Description: liblustre client does MDS+OSTs setattr RPC for each write
4195 Details : When doing a write from a liblustre client, the client
4196 incorrectly issued an RPC to the MDS and each OST the file was
4197 striped over in order to update the timestamps. When writing
4198 with small chunks and many clients this could overwhelm the MDS
4199 with RPCs. In all cases it would slow down the write because
4200 these RPCs are unnecessary.
4202 Severity : enhancement
4204 Description: allow number of MDS service threads to be changed at module load
4205 Details : It is now possible to change the number of MDS service threads
4206 running. Adding "options mds mds_num_threads={N}" to the MDS's
4207 /etc/modprobe.conf will set the number of threads for the next
4208 time Lustre is restarted (assuming the "mds" module is also
4209 reloaded at that time). The default number of threads will
4210 stay the same, 32 for most systems.
4215 Description: OST crash if filesystem is unformatted or corrupt
4216 Details : If an OST is started on a device that has never been formatted
4217 or if the filesystem is corrupt and cannot even mount then the
4218 error handling cleanup routines would dereference a NULL pointer.
4223 Description: NULL pointer deref in llap_from_page.
4224 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4225 incorrect) mapping if the page was truncated/reclaimed while it was
4226 searched for. Check for this condition and skip such pages when
4227 doing readahead. Introduce extra check to llap_from_page() to
4228 verify page->mapping->host is non-NULL (so page is not anonymous).
4231 Frequency : Sometimes when using sys_sendfile
4233 Description: "page not covered by a lock" warnings from ll_readpage
4234 Details : sendfile called ll_readpage without right page locks present.
4235 Now we introduced ll_file_sendfile that does necessary locking
4236 around call to generic_file_sendfile() much like we do in
4240 Frequency : with certain MDS communication failures at client mount time
4242 Description: NULL pointer deref after failed client mount
4243 Details : a client connection request may delayed by the network layer
4244 and not be sent until after the PTLRPC layer has timed out the
4245 request. If the client fails the mount immediately it will try
4246 to clean up before the network times out the request. Add a
4247 reference from the request import to the obd device and delay
4248 the cleanup until the network drops the request.
4251 Frequency : occasionally during client (re)connect
4253 Description: assertion failure during client (re)connect
4254 Details : processing a client connection request may be delayed by the
4255 client or server longer than the client connect timeout. This
4256 causes the client to resend the connection request. If the
4257 original connection request is replied in this interval, the
4258 client may trip an assertion failure in ptlrpc_connect_interpret()
4259 which thought it would be the only running connect process.
4262 Frequency : only with obd_echo servers and clients that are rebooted
4264 Description: kernel BUG accessing uninitialized data structure
4265 Details : When running an obd_echo server it did not start the ping_evictor
4266 thread, and when a client was evicted an uninitialized data
4267 structure was accessed. Start the ping_evictor in the RPC
4268 service startup instead of the OBD startup.
4270 Severity : enhancement
4271 Bugzilla : 10193 (patchless)
4272 Description: Remove dependency on various unexported kernel interfaces.
4273 Details : No longer need reparent_to_init, exit_mm, exit_files,
4274 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4277 Frequency : rare (only users of deprecated and unsupported LDAP config)
4279 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4280 Details : LDAP apparently contains 'lustreName' attributes instead of
4281 'name'. A simple remapping of the name is sufficient.
4284 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4286 Description: waiting_locks_callback trips kernel BUG if client is evicted
4287 Details : Running with the dump_on_timeout debug flag turned on makes
4288 it possible that the waiting_locks_callback() can try to dump
4289 the Lustre kernel debug logs from an interrupt handler. Defer
4290 this log dumping to the expired_lock_main() thread.
4292 Severity : enhancement
4294 Description: Support NFS exporting on 2.6 kernels.
4295 Details : Implement non-rawops metadata methods for NFS server to use without
4296 changing NFS server code.
4299 Frequency : very rare (synthetic metadata workload only)
4301 Description: two racing renames might cause an MDS thread to deadlock
4302 Details : Running the "racer" program may cause one MDS thread to rename
4303 a file from being the source of a rename to being the target of
4304 a rename at exactly the same time that another thread is doing
4305 so, and the second thread has already enqueued these locks after
4306 doing a lookup of the target and is trying to relock them in
4307 order. Ensure that we don't try to re-lock the same resource.
4310 Frequency : only very large systems with liblustre clients
4312 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4313 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4314 all OSTs in parallel. This allows the eviction of stale liblustre
4315 clients to proceed much faster than if they were done in series,
4316 and also offers similar improvements for other set_info RPCs.
4321 Description: excessive CPU usage during initial read phase on client
4322 Details : During the initial read phase on a client, it would agressively
4323 retry readahead on the file, consuming too much CPU and impacting
4324 performance (since 1.4.5.8). Improve the readahead algorithm
4325 to avoid this, and also improve some other common cases (read
4326 of small files in particular, where "small" is files smaller than
4327 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4332 Description: MDS crash when receiving packet with unknown intent.
4333 Details : Do not LBUG in unknown intent case, just return -EFAULT
4335 Severity : enhancement
4336 Bugzilla : 9293, 9385
4337 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4338 Details : Do not serialize getattr (non-intent version) and statfs.
4341 Frequency : occasional, when OST network is overloaded/intermittent
4343 Description: client evicted by OST after bulk IO timeout
4344 Details : If a client sends a bulk IO request (read or write) the OST
4345 may evict the client if it is unresposive to its data GET/PUT
4346 request. This is incorrect if the network is overloaded (takes
4347 too long to transfer the RPC data) or dropped the OST GET/PUT
4348 request. There is no need to evict the client at all, since
4349 the pinger and/or lock callbacks will handle this, and the
4350 client can restart the bulk request.
4353 Frequency : Always when mmapping file with no objects
4355 Description: client crashes when mmapping file with no objects
4356 Details : Check that we actually have objects in a file before doing any
4357 operations on objects in ll_vm_open, ll_vm_close and
4363 Description: Request leak when working with deleted CWD
4364 Details : Introduce advanced request refcount tracking for requests
4365 referenced from lustre intent.
4367 Severity : Enhancement
4369 Description: Cache open file handles on client.
4370 Details : MDS now will return special lock along with openhandle, if
4371 requested and client is allowed to hold openhandle, even if unused,
4372 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4373 closing files for every read/write openration.
4375 Severity : Enhancement
4377 Description: Cache open negative dentries on client when possible.
4378 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4379 negative dentries on lock revocation.
4384 Description: Remounting a client read-only wasn't possible with a zconf mount
4385 Details : It wasn't possible to remount a client read-only with llmount.
4387 Severity : enhancement
4388 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4389 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4390 MPICH is included to add Lustre-specific ADIO interfaces.
4391 This is based closely on the UFS ADIO layer and only differs
4392 in file creation, in order to allow the OST striping to be set.
4393 This is user-contributed code and not supported by CFS.
4398 Description: extended inode attributes (immutable, append-only) work improperly
4399 when 2.4 and 2.6 kernels are used on client/server or vice versa
4400 Details : Introduce kernel-independent values for these flags.
4402 Severity : enhancement
4405 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4406 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4407 granularity. Now, Fractional MB granularity is be supported,
4408 this is very useful for low memory system.
4410 Severity : enhancement
4412 Description: Getattr by fid
4413 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4414 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4415 have locks and avoids getattr rpc requests for it on every stat.
4418 Frequency : Always, for filesystems larger than 2TB
4420 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4421 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4422 filesystems larger than 2TB to be mounted with mballoc enabled.
4425 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4427 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4428 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4429 filesystems larger than 2TB is mounted then the kernel will
4430 silently corrupt the start of the filesystem. CONFIG_LBD is
4431 enabled for all CFS-supported kernels, but the possibility of
4432 this happening with a modified kernel config exists.
4434 Severity : enhancement
4436 Description: add client O_DIRECT support for 2.6 kernels
4437 Details : It is now possible to do O_DIRECT reads and writes to files
4438 in the Lustre client mountpoint on 2.6 kernel clients.
4440 Severity : enhancement
4442 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4443 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4444 parallel, not waiting for response from every OST before sending
4445 a rpc to the next OST.
4450 Description: setattr vs write race when updating file timestamps
4451 Details : Client processes that update a file timestamp into the past
4452 right after writing to the file (e.g. tar) it is possible that
4453 the updated file modification time can be reset to the current
4454 time due to a race between processing the setattr and write RPC.
4456 Severity : enhancement
4458 Description: Bring 'lfs find' closer in line with regular Linux find.
4459 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4460 -print0 options and obtains all the needed info through the lustre
4463 Severity : enhancement
4465 Description: support up to 1024 configured devices on one node
4466 Details : change obd_dev array from statically allocated to dynamically
4467 allocated structs as they are first used to reduce memory usage
4472 Description: Flush dirty partially truncated pages during truncate
4473 Details : Immediatelly flush partially truncated pages in filter_setattr,
4474 this way we completely avoid having any pages in page cache on OST
4475 and can retire ugly workarounds during writes to flush such pages.
4480 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4481 Details : protect inode from truncation within vfs_unlink() context
4482 just take a reference before calling vfs_unlink() and release it
4483 when parent's i_sem is free.
4488 Description: last_id value checked outside lock on OST caused LASSERT failure
4489 Details : If there were multiple MDS->OST object precreate requests in
4490 flight, it was possible that the OST's last object id was checked
4491 outside a lock and incorrectly tripped an assertion. Move checks
4492 inside locks, and discard old precreate requests.
4495 Frequency : always, if extents are used on OSTs
4497 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4498 Details : OSTs using the extents format would not zero the high 16 bits of
4499 the index physical block number. This is not a problem for any
4500 OST filesystems smaller than 16TB, and no kernels support ext3
4501 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4502 new/modified files) and can be fixed for existing filesystems
4503 with e2fsprogs-1.39-cfs1.
4508 Description: import connection selection may be incorrect if timer wraps
4509 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4510 import connections to be ignored if the 32-bit jiffies counter
4511 wraps. Use a 64-bit jiffies counter.
4514 Frequency : very large clusters immediately after boot
4516 Description: LNET request buffers exhausted under heavy short-term load
4517 Details : If a large number of client requests are generated on a service
4518 that has previously never seen so many requests it is possible
4519 that the request buffer growth cannot keep up with the spike in
4520 demand. Instead of dropping incoming requests, they are held in
4521 the LND until the RPC service can accept more requests.
4524 Frequency : Sometimes during replay
4526 Description: Assertion failure in ll_local_open after replay.
4527 Details : If replay happened on an open request reply before we were able
4528 to set replay handler, reply will become not swabbed tripping the
4529 assertion in ll_local_open. Now we set the handler right after
4530 recognising of open request
4533 Frequency : very rare
4535 Description: kernel reports "badness in vsnprintf"
4536 Details : Reading from the "recovery_status" /proc file in small chunks
4537 may cause a negative length in lprocfs_obd_rd_recovery_status()
4538 call to vsnprintf() (which is otherwise harmless). Exit early
4539 if there is no more space in the output buffer.
4541 Severity : enhancement
4543 Description: clear OBD RPC statistics by writing to them
4544 Details : It is now possible to clear the OBD RPC statistics by writing
4545 to the "stats" file.
4550 Description: Client mtime is not the same on different clients after utimes
4551 Details : In some cases, the client was using the utimes() syscall on
4552 a file cached on another node. The clients now validate the
4553 ctime from the MDS + OSTs to determine which one is right.
4558 Description: Inability to activate failout mode
4559 Details : lconf script incorrectly assumed that in pythong string's numeric
4560 value is used in comparisons.
4563 Frequency : always with multiple stripes per file
4565 Description: Inefficient object allocation for mutli-stripe files
4566 Details : When selecting which OSTs to stripe files over, for files with
4567 a stripe count that divides evenly into the number of OSTs,
4568 the MDS is always picking the same starting OST for each file.
4569 Return the OST selection heuristic to the original design.
4574 Description: mount failures may take full timeout to return an error
4575 Details : Under some heavy load conditions it is possible that a
4576 failed mount can wait for the full obd_timeout interval,
4577 possibly several minutes, before reporting an error.
4578 Instead return an error as soon as the status is known.
4580 Frequency : quota enabled and large files being deleted
4582 Description: releasing more than 4GB of quota at once hangs OST
4583 Details : If a user deletes more than 4GB of files on a single OST it
4584 will cause the OST to spin in an infinite loop. Release
4585 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4590 Description: statfs data retrieved from /proc may be stale or zero
4591 Details : When reading per-device statfs data from /proc, in the
4592 {kbytes,files}_{total,free,avail} files, it may appear
4593 as zero or be out of date.
4596 Frequency : systems with MD RAID1 external journal devices
4598 Description: lconf's call to blkid is confused by RAID1 journal devices
4599 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4600 of returning all block devices that match the journal UUID.
4603 Frequency : always, for aggregate stripe size over 4GB
4605 Description: assertion fails when trying to use 4GB stripe size
4606 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4607 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4610 Frequency : always on ppc64
4612 Description: the first write on an ext3 filesystem with mballoc got stuck
4613 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4614 perform endianness conversion.
4616 ------------------------------------------------------------------------------
4618 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4620 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4621 INTEROPERATE with older versions automatically. Please read the
4622 user documentation before upgrading any part of a live system.
4623 * WARNING: Lustre networking configuration changes are required with
4624 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4627 * Support for kernels:
4628 2.6.9-22.0.2.EL (RHEL 4)
4629 2.6.5-7.244 (SLES 9)
4630 2.6.12.6 vanilla (kernel.org)
4633 Severity : enhancement
4634 Bugzilla : 7981/8208
4635 Description: Introduced Lustre Networking (LNET)
4636 Details : LNET is new networking infrastructure for Lustre, it includes
4637 a reorganized network configuration mode (see the user
4638 documentation for full details) as well as support for routing
4639 between different network fabrics. Lustre Networking Devices
4640 (LNDs) for the supported network fabrics have also been
4641 created for this new infrastructure.
4643 Severity : enhancement
4644 Description: Introduced Access control lists
4645 Details : clients can set ACLs on files and directories in order to have
4646 more fine-grained permissions than the standard Unix UGO+RWX.
4647 The MDS must be started with the "-o acl" mount option.
4649 Severity : enhancement
4650 Description: Introduced filesystem quotas
4651 Details : Administrators may now establish per-user quotas on the
4654 Severity : enhancement
4656 Description: Configuration change for the XT3
4657 The PTLLND is now used to run Lustre over Portals on the XT3
4658 The configure option(s) --with-cray-portals are no longer used.
4659 Rather --with-portals=<path-to-portals-includes> is used to
4660 enable building on the XT3. In addition to enable XT3 specific
4661 features the option --enable-cray-xt3 must be used.
4666 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4667 Details : A race between allocating a new llcd and re-getting the llcd_lock
4668 allowed another thread to grab newly-allocated llcd.
4670 Severity : enhancement
4672 Description: 2.6 OST async journal commit and locking fix to improve performance
4673 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4674 2.6 kernels are now async as they already were in 2.4 kernels so
4675 that they can commit concurrently with the network bulk transfer.
4676 For block-allocated files the filter allocation semaphore is held
4677 to avoid filesystem fragmentation during allocation. BKL lock
4678 removed for 2.6 xattr operations where it is no longer needed.
4683 Description: lconf incorrectly determined whether two IP networks could talk
4684 Details : In some more complicated routing and multiple-network
4685 configurations, lconf will avoid trying to make a network
4686 connection to a disjoint part of the IP space. It was doing the
4687 math incorrectly for one set of cases.
4692 Description: Fix for potential infinite loop processing records in an llog.
4693 Details : If an llog record is corrupted/zeroed, it is possible to loop
4694 forever in llog_process(). Validate the llog record length
4695 and skip the remainder of the block on error.
4698 Frequency : occasional (liblustre only)
4700 Description: liblustre could not open files whose last component is a symlink
4701 Details : sysio_path_walk() would incorrectly pass the open intent to
4702 intermediate path components.
4705 Frequency : rare (liblustre only with non-standard tuning)
4706 Bugzilla : 7201 (7350)
4707 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4708 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4709 after it has been released, only do this for referenced locks
4711 Severity : enhancement
4713 Description: specifying an (invalid) directory default stripe_size of -1
4714 would reset the directory default striping
4715 Details : stripe_size -1 was used internally to signal directory stripe
4716 removal, now use "all default" to signal dir stripe removal
4717 as a directory striping of "all default" is not useful
4720 Frequency : common for large clusters running liblustre clients
4722 Description: doing an ls when liblustre clients are running is slow
4723 Details : sending a glimpse AST to a liblustre client waits for every AST
4724 to time out, as liblustre clients will not respond. Since they
4725 cannot cache data we refresh the OST lock LVB from disk instead.
4727 Severity : enhancement
4729 Description: doing an ls at the same time as file IO can be slow
4730 Details : enqueue and other "small" requests can be blocked behind many
4731 large IO requests. Create a new OST IO portal for non-IO
4732 requests so they can be processed faster.
4735 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4737 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4738 created file with mode 000
4739 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4740 allow an ftruncate() as a non-root user to a file with mode 000.
4741 Lustre now respects this flag to disable mode checks when
4742 truncating a file owned by the user
4745 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4748 Description: Revoking locks from clients that went dead or catatonic might take
4750 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4751 cancellation of such locks instant on servers without waiting for
4752 any reply from clients. Clients drops these locks when cancel
4753 notification from server is received without replying.
4756 Frequency : liblustre-only, when liblustre client dies or becomes busy
4758 Description: Doing ls on Linux clients can take a long time with active
4760 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4761 granting such locks to it in the first place if possible. Locks
4762 are taken by proxy on the OST during the read or write and
4763 dropped immediately afterward. Add connect flags handling, do
4764 not grant locks to liblustre clients for glimpse ASTs.
4766 Severity : enhancement
4768 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4769 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4770 sequential reads and behaves badly with applications doing random
4771 reads. Improve it by reading ahead at least read region, and
4772 avoiding excessive large RPC for small reads.
4774 Severity : enhancement
4776 Description: Creating more than 1000 files for a single job may cause a load
4777 imbalance on the OSTs if there are also a large number of OSTs.
4778 Details : qos_prep_create() uses an OST index reseed value that is an
4779 even multiple of the number of available OSTs so that if the
4780 reseed happens in the middle of the object allocation it will
4781 still utilize the OSTs as uniformly as possible.
4786 Description: OST or MDS may oops in ping_evictor_main()
4787 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4788 but doesn't restart at beginning of obd_exports_timed list
4791 Severity : enhancement
4793 Description: improve by-nid export eviction on the MDS and OST
4794 Details : allow multiple exports with the same NID to be evicted at one
4795 time without re-searching the exports list.
4798 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4800 Description: MDS may oops in groups_free()
4801 Details : in rare race conditions a newly allocated group_info struct is
4802 freed again, and this can be NULL. The 2.4 compatibility code
4803 for groups_free() checked for a NULL pointer, but 2.6 did not.
4806 Frequency : common for liblustre clients doing little filesystem IO
4807 Bugzilla : 9352, 7313
4808 Description: server may evict liblustre clients accessing contended locks
4809 Details : if a client is granted a lock or receives a completion AST
4810 with a blocking AST already set it would not reply to the AST
4811 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4814 Frequency : lfs setstripe, only systems with more than 160 OSTs
4816 Description: unable to set striping with a starting offset beyond OST 160
4817 Details : llapi_create_file() incorrectly limited the starting stripe
4818 index to the maximum single-file stripe count.
4821 Frequency : LDAP users only
4823 Description: lconf did not handle in-kernel recovery with LDAP properly
4824 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4826 Severity : enhancement
4828 Description: bind OST threads to NUMA nodes to improve performance
4829 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4830 node and do their allocations there to localize memory access
4832 Severity : enhancement
4834 Description: llmount can determine client NID directly from Myrinet (GM)
4835 Details : the client NID code from gmnalnid was moved directly into
4836 llmount, removing the need to use this or specifying the
4837 client NID explicitly when mounting GM clients with zeroconf
4840 Frequency : if client is started with down MDS
4842 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4843 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4844 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4845 ptlrpc_import_delayed_req() because we didn't check if the
4846 request was interrupted, and we also didn't break out of the
4847 event loop if there was no timeout
4852 Description: data loss during non-page-aligned writes to a single file from
4853 both multiple nodes and multiple threads on one node at same time
4854 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4855 inconsistency led to false short-reads, that were cached and later
4856 used by ->prepare_write() to fill in partially written page,
4857 leading to data loss.
4860 Frequency : always, if lconf --abort_recovery used
4862 Description: lconf --abort_recovery fails with 'Operation not supported'
4863 Details : lconf was attempting to abort recovery on the MDT device and not
4866 Severity : enhancement
4868 Description: remove cleanup logs
4869 Details : replace lconf-generated cleanup logs with lustre internal
4870 cleanup routines. Eliminates the need for client-cleanup and
4873 Severity : enhancement
4875 Description: add support for EAs (user and system) on lustre filesystems
4876 Details : it is now possible to store extended attributes in the Lustre
4877 client filesystem, and with the user_xattr mount option it
4878 is possible to allow users to store EAs on their files also
4880 Severity : enhancement
4882 Description: Add possibility (config option) to show minimal available OST free
4884 Details : When compiled with --enable-mindf configure option, statfs(2)
4885 (and so, df) will return least minimal free space available from
4886 all OSTs as amount of free space on FS, instead of summary of
4887 free spaces of all OSTs.
4889 Severity : enhancement
4891 Description: do not expand extent locks acquired on OST-side
4892 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4893 by server: they are not cached anyway.
4896 Frequency : when mmap is used/binaries executed from Lustre
4898 Description: Unmmap pages before throwing them away from read cache.
4899 Details : llap_shrink cache now attempts to unmap pages before discarding
4900 them (if unmapping failed - do not discard). SLES9 kernel has
4901 extra checks that trigger if this unmapping is not done first.
4906 Description: lconf didn't resolve symlinks before checking to see whether a
4907 given mountpoint was already in use
4910 Frequency : when migrating failover services
4911 Bugzilla : 6395, 9514
4912 Description: When migrating a subset of services from a node (e.g. failback
4913 from a failover service node) the remaining services would
4914 time out and evict clients.
4915 Details : lconf --force (implied by --failover) sets the global obd_timeout
4916 to 5 seconds in order to quickly disconnect, but this caused
4917 other RPCs to time out too quickly. Do not change the global
4918 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4920 Severity : enhancement
4921 Frequency : if MDS is started with down OST
4922 Bugzilla : 9439,5706
4923 Description: Allow startup/shutdown of an MDS without depending on the
4924 availability of the OSTs.
4925 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4926 Add appropriate locking and lov-osc refcounts for safe
4927 cleaning. Add osc abort_inflight calls in case the
4928 synchronize never started.
4931 Frequency : occasional (Cray XT3 only)
4933 Description: root not authorized to access files in CRAY_PORTALS environment
4934 Details : The client process capabilities were not honoured on the MDS in
4935 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4936 been accessed by an authorized user then root was able to access
4937 the file on the local client also. The root user capabilities
4938 are now allowed on the MDS, as this environment has secure UID.
4941 Frequency : occasional
4943 Description: ldiskfs "too long searching" message happens too often
4944 Details : A debugging message (otherwise harmless) prints too often on
4945 the OST console. This has been reduced to only happen when
4946 there are fragmentation problems on the filesystem.
4951 Description: Division by zero in statfs when all OSCs are inactive
4952 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4953 lov_statfs divides by value returned by lov_get_stripecnt().
4957 Bugzilla : 9489, 3273
4958 Description: First write from each client to each OST was only 4kB in size,
4959 to initialize client writeback cache, which caused sub-optimal
4960 RPCs and poor layout on disk for the first written file.
4961 Details : Clients now request an initial cache grant at (re)connect time
4962 and so that they can start streaming writes to the cache right
4963 away and always do full-sized RPCs if there is enough data.
4964 If the OST is rebooted the client also re-establishes its grant
4965 so that client cached writes will be honoured under the grant.
4970 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4971 Details : Now I/O RPCs go to different portal number and (presumably) fast
4972 lock requests (and glimses) and other RPCs get their own service
4973 threads pool that should be able to service those RPCs
4976 Severity : enhancement
4978 Description: Ability to exchange lustre version between client and servers and
4979 issue warnings at client side if client is too old. Also for
4980 liblustre clients there is ability to refuse connection of too old
4982 Details : New 'version' field is added to connect data structure that is
4983 filled with version info. That info is later checked by server and
4987 Frequency : rare, liblustre only.
4988 Bugzilla : 9296, 9581
4989 Description: Two simultaneous writes from liblustre at offset within same page
4990 might proceed at the same time overwriting eachother with stale
4992 Details : I/O lock withing llu_file_prwv was released too early, before data
4993 actually was hitting the wire. Extended lock-holding time until
4994 server acknowledges receiving data.
4997 Frequency : extremely rare. Never observed in practice.
4999 Description: avoid generating lustre_handle cookie of 0.
5000 Details : class_handle_hash() generates handle cookies by incrementing
5001 global counter, and can hit 0 occasionaly (this is unlikely, but
5002 not impossible, because initial value of cookie counter is
5003 selected randonly). Value of 0 is used as a sentinel meaning
5004 "unassigned handle" --- avoid it. Also coalesce two critical
5005 sections in this function into one.
5007 Severity : enhancement
5009 Description: allow liblustre clients to delegate truncate locking to OST
5010 Details : To avoid overhead of locking, liblustre client instructs OST to
5011 take extent lock in ost_punch() on client's behalf. New connection
5012 flag is added to handle backward compatibility.
5014 Severity : enhancement
5015 Bugzilla : 4928, 7341, 9758
5016 Description: allow number of OST service threads to be specified
5017 Details : a module parameter allows the number of OST service threads
5018 to be specified via "options ost ost_num_threads={N}" in the
5019 OSS's /etc/modules.conf or /etc/modprobe.conf.
5023 Bugzilla : 6146, 9635, 9895
5024 Description: servers crash with bad pointer in target_handle_connect()
5025 Details : In rare cases when a client is reconnecting it was possible that
5026 the connection request was the last reference for that export.
5027 We would temporarily drop the export reference and get a new
5028 one, but this may have been the last reference and the export
5029 was just destroyed. Get new reference before dropping old one.
5031 Severity : enhancement
5032 Frequency : if client is started with failover MDS
5034 Description: Allow multiple MDS hostnames in the mount command
5035 Details : Try to read the configuration from all specified MDS
5036 hostnames during a client mount in case the "primary"
5039 Severity : enhancement
5041 Description: Stop sending data to evicted clients as soon as possible.
5042 Details : Check if the client we are about to send or are sending data to
5043 was evicted already. (Check is done every second of waiting,
5044 for which l_wait_event interface was extended to allow checking
5045 of exit condition at specified intervals).
5048 Frequency : rare, normally only when NFS exporting is done from client
5050 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5052 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5055 Frequency : always, if srand() is called before liblustre initialization
5057 Description: Liblustre uses system PRNG disturbing its usage by user application
5058 Details : Introduce internal to lustre fast and high-quality PRNG for
5059 lustre usage and make liblustre and some other places in generic
5060 lustre code to use it.
5062 Severity : enhancement
5063 Bugzilla : 9477, 9557, 9870
5064 Description: Verify that the MDS configuration logs are updated when xml is
5065 Details : Check if the .xml configuration logs are newer than the config
5066 logs stored on the MDS and report an error if this is the case.
5067 Request --write-conf, or allow starting with --old_conf.
5069 Severity : enhancement
5071 Description: Handle symlinks in the path when checking if Lustre is mounted.
5072 Details : Resolve intermediate symlinks when checking if a client has
5073 mounted a filesystem to avoid duplicate client mounts.
5078 Description: lconf can hit an error exception but still return success.
5079 Details : The lconf command catches the Command error exception at the top
5080 level script context and will exit with the associated exit
5081 status, but doesn't ensure that this exit status is non-zero.
5086 Description: failure of ptlrpc thread startup can cause oops
5087 Details : Starting a ptlrpc service thread can fail if there are a large
5088 number of threads or the server memory is very fragmented.
5089 Handle this without oopsing.
5092 Frequency : always, only if liblustre and non-default acceptor port was used
5094 Description: liblustre cannot connect to servers with non-default acceptor port
5095 Details : tcpnal_set_default_params() was not called and was therefore
5096 ignoring the environment varaible TCPNAL_PORT, as well as other
5097 TCPNAL_ environment variables
5102 Description: two objects could be created on the same OST for a single file
5103 Details : If an OST is down, in some cases it was possible to create two
5104 objects on a single OST for a single file. No problems other
5105 than potential performance impact and spurious error messages.
5109 Bugzilla : 5681, 9562
5110 Description: Client may oops in ll_unhash_aliases
5111 Details : Client dcache may become inconsistent in race condition.
5112 In some cases "getcwd" can fail if the current directory is
5118 Description: Inode refcounting problems in NFS export code
5119 Details : link_raw functions used to call d_instantiate without obtaining
5120 extra inode reference first.
5124 Bugzilla : 9942, 9903
5125 Description: Referencing freed requests leading to crash, memleaks with NFS.
5126 Details : We used to require that call to ll_revalidate_it was always
5127 followed by ll_lookup_it. Also with revalidate_special() it is
5128 possible to call ll_revalidate_it() twice for the same dentry
5129 even if first occurence returned success. This fix changes semantic
5130 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5131 reference on a request referred from the intent.
5132 ll_intent_release() then releases such a request.
5135 Frequency : rare, normally benchmark loads only
5137 Description: unlinked inodes were kept in memory on the client
5138 Details : If a client is repeatedly creating and unlinking files it
5139 can accumulate a lot of stale inodes in the inode slab cache.
5140 If there is no other client load running this can cause the
5141 client node to run out of memory. Instead flush old inodes
5142 from client cache that have the same inode number as a new inode.
5145 Frequency : SLES9 2.6.5 kernel and long filenames only
5146 Bugzilla : 9969, 10379
5147 Description: utime reports stale NFS file handle
5148 Details : SLES9 uses out-of-dentry names in some cases, which confused
5149 the lustre dentry revalidation. Change it to always use the
5153 Frequency : rare, unless heavy write-truncate concurrency is continuous
5154 Bugzilla : 4180, 6984, 7171, 9963, 9331
5155 Description: OST becomes very slow and/or deadlocked during object unlink
5156 Details : filter_destroy() was holding onto the parent directory lock
5157 while truncating+unlinking objects. For very large objects this
5158 may block other threads for a long time and slow overall OST
5159 responsiveness. It may also be possible to get a lock ordering
5160 deadlock in this case, or run out of journal credits because of
5161 the combined truncate+unlink. Solution is to do object truncate
5162 first in one transaction without parent lock, and then do the
5163 final unlink in a new transaction with the parent lock. This
5164 reduces the lock hold time dramatically.
5167 Frequency : rare, 2.4 kernels only
5169 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5170 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5171 before unmounting their filesystem in order to avoid deadlock.
5172 The kernel_locked() function in 2.4 kernels only checks whether
5173 the kernel lock is held, not whether it is this process that is
5174 holding it as 2.6 kernels do.
5179 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5180 Details : The client ptlrpc code may be trying to reconnect to a down
5181 server before a previous connection attempt has timed out.
5182 Increase the reconnect interval to be longer than the connection
5183 timeout interval to avoid sending duplicate connections to
5187 Frequency : echo_client brw_test command
5189 Description: fix echo_client to work with OST preallocated code
5190 Details : OST preallocation code (5137) didn't take echo_client IO path
5191 into account: echo_client calls filter methods outside of any
5192 OST thread and, hence, there is no per-thread preallocated
5193 pages and buffers to use. Solution: hijack pga pages for IO. As
5194 a byproduct, this avoids unnecessary data copying.
5198 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5199 Description: Client can oops in mdc_commit_close() after open replay
5200 Details : It was possible for the MDS to return an open request with no
5201 transaction number in mds_finish_transno() if the client was
5202 evicted, but without actually returning an error. Clients
5203 would later try to replay that open and may trip an assertion
5204 Simplify the client close codepath, and always return an error
5205 from the MDS in case the open is not successful.
5208 Frequency : rare, 2.6 OSTs only
5210 Description: OST may deadlock under high load on fragmented files
5211 Details : If there was a heavy load and highly-fragmented OST filesystems
5212 it was possible to have all the OST threads deadlock waiting on
5213 allocation of biovecs, because the biovecs were not released
5214 until the entire RPC IO was completed. Instead, release biovecs
5215 as soon as they are complete to ensure forward IO progress.
5217 Severity : enhancement
5219 Description: Support for specifying external journal device at mount
5220 Details : If an OST or MDS device is formatted with an external journal
5221 device, this device major/minor is stored in the ext3 superblock
5222 and may not be valid for failover. Allow detecting and
5223 specifying the external journal at mount time.
5228 Description: Mounting an MDS with pending unlinked files may cause oops
5229 Details : target_finish_recovery() calls mds_postrecov() which returned
5230 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5231 considers this an error and immediately begins cleaning up the
5232 lov, just after starting the mds_lov process
5234 Severity : enhancement
5236 Description: Implement 'lfs df' to report actual free space on per-OST basis
5237 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5238 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5239 usage of objects; '-h' to report in human readable format.
5241 ------------------------------------------------------------------------------
5243 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5250 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5251 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5252 at startup during orphan unlinking. Instead, skip buddy update
5253 before setup, as it will be regenerated after recovery is complete.
5258 Description: If an OST is inactive, its locks might reference stale inodes.
5259 Details : lov_change_cbdata() must iterate over all namespaces, even if
5260 they are inactive to clear inode references from the lock.
5262 Severity : enhancement
5263 Frequency : occasional, if non-standard max_dirty_mb used
5265 Description: Client will block write RPCs if not enough grant
5266 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5267 then the client will block writes while waiting for another RPC
5268 to complete instead of consuming its dirty limit. With change
5269 we get improved performance when max_dirty_mb is small.
5271 Severity : enhancement
5272 Bugzilla : 3389, 6253
5273 Description: Add support for supplementary groups on the MDS.
5274 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5275 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5276 lookups for user supplementary groups into a cache.
5280 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5281 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5282 for liblustre O_EXCL is enforced by MDS.
5286 Description: readdir never returns NULL in liblustre.
5287 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5288 dentry in struct dirent.
5292 Frequency : liblustre only, depends on application IO pattern
5293 Description: liblustre clients evicted if not contacting servers
5294 Details : Don't put liblustre clients into the ping_evictor list, so
5295 they will not be evicted by the pinger ever.
5297 Severity : enhancement
5299 Description: Add ability to evict clients by NID from MDS.
5300 Details : By echoing "nid:$NID" string into
5301 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5302 $NID would be instantly evicted from this MDS and from all active
5303 OSTs connected to it.
5307 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5308 Details : lookup_it_finish() used to query file size from OSTs that was not
5313 Description: service threads change working directory to that of init
5314 Details : Starting lustre service threads may pin the working directory
5315 of the parent thread, making that filesystem busy. Threads
5316 now change to the working directory of init to avoid this.
5320 Frequency : during shutdown only
5321 Description: shutdown with a failed MDS or OST can cause unmount to hang
5322 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5327 Frequency : 2.6 only
5328 Description: chmod/chown may include an extra supplementary group
5329 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5330 supplementary group and if none is specified this is used.
5333 Bugzilla : 5479 (6816)
5335 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5336 Details : If lookup is in progress on a file that is unlinked we might try
5337 to revalidate the inode and fail in revalidate after lookup is
5338 complete and ll_file_open() enqueues the open again but
5339 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5342 Frequency : always, if lconf --abort_recovery used
5344 Description: lconf --abort_recovery fails with 'Operation not supported'
5345 Details : lconf was attempting to abort recovery on the MDT device and not
5348 ------------------------------------------------------------------------------
5350 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5355 Frequency : rare (only unsupported configurations with a node running as an
5357 Bugzilla : 6514, 5137
5358 Description: Mounting a Lustre file system on a node running as an OST could
5360 Details : OSTs now preallocates memory needed to write out data at
5361 startup, instead of when needed, to avoid having to
5362 allocate memory in possibly low memory situations.
5363 Specifically, if the file system is mounted on on OST,
5364 memory pressure could force it to try to write out data,
5365 which it needed to allocate memory to do. Due to the low
5366 memory, it would be unable to do so and the node would
5367 become unresponsive.
5369 Severity : enhancement
5371 Description: Addition of lconf --service command line option
5372 Details : lconf now accepts a '--service <arg>' option, which is
5373 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5375 Severity : enhancement
5377 Description: Failover mode is now the default for OSTs.
5378 Details : By default, OSTs will now run in failover mode. To return to
5379 the old behaviour, add '--failout' to the lmc line for OSTs.
5381 Severity : enhancement
5383 Description: Health checks are now provided for MDS and OSTs
5384 Details : Additional detailed health check information on MSD and OSTs
5385 is now provided through the procfs health_check value.
5388 Frequency : occasional, depends on IO load
5390 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5391 after numerous create/delete cycles
5392 Details : The ext3 inode allocation policy would not allocate new inodes
5393 very well on the OSTs because there are no new directories
5394 being created. Instead we look for groups with free space if
5395 the parent directories are nearly full.
5400 Description: Network or server problems during mount may cause partially
5401 mounted clients instead of returning an error.
5402 Details : The config llog parsing code may overwrite the error return
5403 code during mount error handling, returning success instead
5409 Description: MDS can fail to allocate large reply buffers
5410 Details : After long uptimes the MDS can fail to allocate large reply
5411 buffers (e.g. zconf client mount config records) due to memory
5412 fragmentation or consumption by the buffer cache. Preallocate
5413 some large reply buffers so that these replies can be sent even
5414 under memory pressure.
5418 Frequency : rare (liblustre)
5419 Description: fsx running with liblustre complained that using truncate() to
5420 extend the file doesn't work. This patch corrects that issue.
5421 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5422 ATTR_SIZE and lsm use in llu_setattr_raw.
5426 Frequency : rare, only 2.6 kernels
5427 Description: Unusual file access patterns on the MDS may result in inode
5428 data being lost in very rare circumstances.
5429 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5430 code in the 2.6 kernel caused the inode and/or EA data not to
5431 be read from disk, causing single-file corruption.
5435 Frequency : rare, only 2.6 filesystems using extents
5436 Description: Heavy concurrent write and delete load may cause data corruption.
5437 Details : It was possible under high-load situations to have an extent
5438 metadata block in the block device cache from a just-unlinked
5439 file overwrite a newly-allocated data block. We now unmap any
5440 metadata buffers that alias just-allocated data blocks.
5444 Frequency : filesystems with default stripe_count larger than 77
5445 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5446 Details : lconf specifies an inode size of 4096 bytes when the default
5447 stripe_count is larger than 77. This conflicts with the default
5448 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5449 this case to avoid pinning too much memory for large EAs.
5451 ------------------------------------------------------------------------------
5453 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5458 Frequency : rare (extremely heavy IO load with hundreds of clients)
5460 Description: Client is evicted, gets IO error writing to file
5461 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5462 caused clients to be evicted for AST timeouts. The fixes in
5463 bug 5192 mean we no longer need to have such short AST timeouts
5464 so ldlm_timeout has been increased.
5467 Frequency : occasional during --force or --failover shutdown under load
5468 Bugzilla : 5949, 4834
5469 Description: Server oops/LBUG if stopped with --force or --failover under load
5470 Details : a collection of import/export refcount and cleanup ordering
5471 issues fixed for safer force cleanup
5474 Frequency : only filesystems larger than 120 OSTs
5475 Bugzilla : 5990, 6223
5476 Description: lfs getstripe would oops on a very large filesystem
5477 Details : lov_getconfig used kfree on vmalloc'd memory
5480 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5481 Bugzilla : 6242, 6243
5482 Description: reading from files that had been truncated to a non-zero size
5483 but never opened returned no data
5484 Details : ll_file_read() reads zeros from no-object files to EOF
5489 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5490 Details : There's little harm in aborting MDS/OSS recovery and letting it
5491 try again, so I removed the LASSERT and return an error instead.
5493 Severity : enhancement
5495 Description: New debugging infrastructure for tracking down data corruption
5496 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5497 (b) cover more of the client-side code path, and (c) try to narrow
5498 down where problems occurred
5502 Bugzilla : 3819, 4364, 4397, 6313
5503 Description: Racing close and eviction MDS could cause assertion in mds_close
5504 Details : It was possible to get multiple mfd references during close and
5505 client eviction, leading to one thread referencing a freed mfd.
5507 Severity: : enhancement
5508 Bugzilla : 3262, 6359
5509 Description: Attempts to reconnect to servers are now more aggressive.
5510 Details : This builds on the enhanced upcall-less recovery that was added
5511 in 1.4.2. When trying to reconnect to servers, clients will
5512 now try each server in the failover group every 10 seconds. By
5513 default, clients would previously try one server every 25 seconds.
5518 Description: After recovery, certain operations trigger a failed
5519 assertion on a client.
5520 Details : Failing over an mds, using lconf -d --failover, while a
5521 client was doing a readdir() call would cause the client to
5522 LBUG after recovery completed and the readdir() was resent.
5524 Severity : enhancement
5526 Description: Default groups are now added by lconf
5527 Details : You can now run lconf --group <servicename> without having to
5528 manually add groups with lmc.
5531 Frequency : occasional
5533 Description: Nodes with an elan id of 0 trigger a failed assertion
5536 Frequency : always when accessing e.g. tty/console device nodes
5538 Description: tty and some other devices nodes cannot be used on lustre
5539 Details : file's private_data field is used by device data and lustre
5540 values in there got lost. New field was added to struct file to
5541 store fs-specific private data.
5544 Frequency : when exporting Lustre via NFS
5546 Description: NFSD failed occasionally when looking up a path component
5547 Details : NFSD is looking up ".." which was broken in ext3 directories
5548 that had grown large enough to become hashed.
5551 Frequency : Clusters with multiple interfaces not on the same subnet
5553 Description: Nodes will repeatedly try to reconnect to an interface which it
5554 cannot reach and report an error to the log.
5555 Details : Extra peer list entries will be created by lconf with some peers
5556 unreachable. lconf now validates the peer before adding it.
5559 Frequency : Only if a default stripe is set on the filesystem root.
5561 Description: Setting a default stripe on the filesystem root prevented the
5562 filesystem from being remounted.
5563 Details : The client was sending extra request flags in the root getattr
5564 request and did not allocate a reply buffer for the dir EA.
5567 Frequency : occasional, higher if lots of files are accessed by one client
5568 Bugzilla : 6159, 6097
5569 Description: Client trips assertion regarding lsm mismatch/magic
5570 Details : While revalidating inodes the VFS looks up inodes with ifind()
5571 and in rare cases can find an inode that is being freed.
5572 The ll_test_inode() code will free the lsm during ifind()
5573 when it finds an existing inode and then the VFS later attaches
5574 this free lsm to a new inode.
5578 Bugzilla : 6422, 7030
5579 Description: MDS deadlock between mkdir and client eviction
5580 Details : Creating a new file via mkdir or mknod (starting a transaction
5581 and getting the ns lock) can deadlock with client eviction
5582 (gets ns lock and trying to finish a synchronous transaction).
5585 Frequency : occasional
5586 Description: While starting a server, the fsfilt_ext3 module could not be
5588 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5589 kernels. Previously, lconf would still use the ext3 name
5590 when trying to load modules. Now, it will correctly use
5591 ext3 on 2.4 and ldiskfs on 2.6.
5593 Severity : enhancement
5594 Description: The default stripe count has been changed to 1
5595 Details : The interpretation of the default stripe count (0, to lfs
5596 or lmc) has been changed to mean striping across a single
5597 OST, rather than all available. For general usage we have
5598 found a stripe count of 1 or 2 works best.
5600 Severity : enhancement
5601 Description: Add support for compiling against Cray portals.
5602 Details : Conditional compiling for some areas that are different
5606 Frequency : occasional
5607 Bugzilla : 6409, 6834
5608 Description: Creating files with an explicit stripe count may lead to
5609 a failed assertion on the MDS
5610 Details : If some OSTs are full or unavailable, creating files may
5611 trigger a failed assertion on the MDS. Now, Lustre will
5612 try to use other servers or return an error to the
5616 Frequency : occasional
5618 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5619 caused lock timeout during readahead (since 1.4.2).
5620 Details : Processes doing readahead might match a lock that hasn't been
5621 granted yet if there are overlapping and conflicting lock
5622 requests. The readahead process waits on ungranted lock
5623 (original lock is CBPENDING), while OST waits for that process
5624 to cancel CBPENDING read lock and eventually evicts client.
5626 Severity : enhancement
5628 Description: Initial enabling of flock support for clients
5629 Details : Implements fcntl advisory locking and file status functions.
5630 This feature is provided as an optional mount flag (default
5631 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5632 locking are implemented yet, and those that are are not guaranteed
5633 to be completely correct in production environments.
5634 mount -t lustre -o [flock|noflock] ...
5637 Frequency : occasional
5639 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5640 assertion in the ext3 JBD (journaling) layer.
5641 Details : The b_committed_data struct is protected by the big kernel lock
5642 in 2.4 kernels, serializing journal_commit_transaction() and
5643 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5644 access to this struct. In 2.6 kernels there is finer grained
5645 locking to improve SMP performance of the JBD layer.
5649 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5654 Description: merge and cleanup kernel patches.
5657 -----------------------------------------------------------------------------
5659 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5661 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5662 versions of Lustre. Please update all servers and clients to
5663 version 1.4.2 or later at the same time. You must also run
5664 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5665 to update the configuration logs.
5667 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5668 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5669 - swab llog records of type '0' so we get proper header size/idx (5861)
5670 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5671 - fix rename of one directory over another leaking an inode (5953)
5672 - avoid SetPageDirty on 2.6 (5981)
5673 - don't re-add just-being-destroyed locks to the waiting list (5653)
5674 - when creating new directories, inherit the parent's custom
5675 striping settings if present parent (3048)
5676 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5677 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5678 - don't hold client locks on temporary worklist from l_lru (5666)
5679 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5680 - automatically evict dead clients (5921)
5681 - Update file size properly in create+truncate+fstat case (6196)
5682 - Do not unhash mountpoint dentries, do not allow removal of
5684 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5685 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5686 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5687 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5688 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5689 - Correctly handle reads of files with no objects (6243)
5690 - lctl recover will also mark a device active if deactivate used (5933)
5692 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5693 - allow --write-conf on an MDS with different nettype than client (5619)
5694 - don't write config llogs to MDS for mounts not from that MDS (5617)
5695 - lconf should create multiple TCP connections from a client (5201)
5696 - init scripts are now turned off by default; run chkconfig --on
5697 lustre and chkconfig --on lustrefs to use them
5698 - upcalls are no longer needed for clients to recover to failover
5700 - add --abort-recovery option to lconf to abort recovery on device
5702 - add support for an arbitrary number of OSTs (3026)
5703 - Quota support protocol changes.
5704 - forward compatibility changes to wire structs (6007)
5705 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5706 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5707 - improved "lustre status" script
5708 - initialize blocksize for non-regular files (6062)
5709 - added --disable-server and --disable-client configure options (5782)
5710 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5711 - Vanilla 2.4.29 support
5712 - increase maximum number of obd devices to 520 (6242)
5713 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5714 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5715 SLES 9 2.6.5 kernels
5716 - limit stripes per file to 160 (the maximum EA size) (6093)
5718 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5721 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5722 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5723 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5724 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5725 - small ext3 extents cleanups and fixes (5733)
5726 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5727 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5728 - avoid needless client->OST connect, fix handle mismatch (5317)
5729 - fix DLM error path that led to out-of-sync client, long delays (5779)
5730 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5731 - fix several locking issues related to i_size (5492,5624,5654,5672)
5732 - don't move pending lock onto export if it is already evicted (5683)
5733 - fix kernel oops when creating .foo in unlinked directory (5548)
5734 - fix deadlock in obdfilter statistics vs. object create (5811)
5735 - use time_{before,after} to avoid timer jiffies wrap (5882)
5736 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5737 - Do not leak request if server was not able to process it (5154)
5738 - If mds_open unable to find parent dir, make that negative lookup(5154)
5739 - don't create new directories with extent-mapping (5909, 5936)
5741 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5742 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5743 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5744 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5745 - support for RHEL4 kernel on i686 (5773)
5746 - provide error messages when incompatible logs are encountered (5898)
5748 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5749 * version 1.4.0.10 (1.4.1 release candidate 1)
5751 - don't keep a lock reference when lock is not granted (4238)
5752 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5753 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5754 - drop import inflight refcount on signal_completed_replay error (5255)
5755 - unlock page after async write error during send (3677)
5756 - handle missing objects in filter_preprw_read properly (5265)
5757 - no transno return for symlink open, don't save no-trasno open (3440)
5758 - don't try to complete elan receive that already failed (4012)
5759 - free RPC server reply state on error (5406)
5760 - clean up thread from ptlrpc_start_thread() on error (5160)
5761 - readahead could read extra page into cache that wasn't ejected (5388)
5762 - prevent races in class_attach/setup/cleanup/detach (5260)
5763 - don't dereference de->d_inode after l_dput of de (5458)
5764 - use "int" for stripe value returned from lock_to_stripe (5544)
5765 - mballoc allocation and error-checking fixes in 2.6 (5504)
5766 - block device patches to fix I/O request sizes in 2.6 (5482)
5767 - look up hostnames for IB nals (5602)
5768 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5769 - don't start multiple acceptors for the same port (5277)
5770 - fix incorrect LASSERT in mds_getattr_name (5635)
5771 - export a proc file for general "ping" checking (5628)
5772 - fix "lfs check" to not block when the MDS is down (5628)
5774 - service request history (4965)
5775 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5776 - create an "evict_client" /proc entry on OSTs, like the MDS has
5777 - fix mount usage message, return errors per mount(8) (5168)
5778 - change grep [] to grep "[]" in tests so they work in more UMLs
5779 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5780 - remove ancient LOV_MAGIC_V0 EA support (5047)
5781 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5782 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5783 - allow readahead allocations to fail when low on memory (5383)
5784 - mmap locking landed again, after considerable improvement (2828)
5785 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5787 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5790 - send OST transaction number in read/write reply to free req (4966)
5791 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5792 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5793 - the watchdog thread now runs as interruptible (5246)
5794 - flock/lockf fixes (but it's still disabled, pending 5135)
5795 - don't use EXT3 constants in llite code (5094)
5796 - memory shortage at startup could cause assertion (5176)
5798 - reorganization of lov code
5799 - single portals codebase
5801 - add extents/mballoc support (5025)
5802 - direct I/O reads in the obdfilter (4048)
5803 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5805 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5808 - send OST transaction number in read/write reply to free req (4966)
5809 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5810 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5811 - the watchdog thread now runs as interruptible (5246)
5812 - handle missing objects in filter_preprw_read properly (5265)
5813 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5814 - drop import inflight refcount on signal_completed_replay error (5255)
5815 - unlock page after async write error during send (3677)
5816 - return original error code on reconstructed replies (3761)
5817 - no transno return for symlink open, don't save no-trasno open (3440)
5819 - add pid to ldlm debugging output (4922)
5820 - bump the watchdog timeouts -- we can't handle 30sec yet
5821 - extra debugging for orphan dentry/inode bug (5259)
5823 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5826 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5827 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5828 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5829 - fix NULL dereference in /proc/sys/portals/routes (4827)
5830 - allow failed mdc_close() operations to be interrupted (4561)
5831 - stop precreate on OST before MDS would time out on it (4778)
5832 - don't send partial-page writes before EOF from client (4410)
5833 - discard client grant for sub-page writes on large-page clients (4520)
5834 - don't free dentries not owned by NFS code, check generation (4806)
5835 - fix lsm leak if mds_create_objects() fails (4801)
5836 - limit debug_daemon file size, always print CERROR messages (4789)
5837 - use transno after validating reply (3892)
5838 - process timed out requests if import state changes (3754)
5839 - update mtime on OST during writes, return in glimpse (4829)
5840 - add mkfsoptions to LDAP (4679)
5841 - use ->max_readahead method instead of zapping global ra (5039)
5842 - don't interrupt __l_wait_event() during strace
5844 - add software watchdogs to catch hung threads quickly (4941)
5845 - make lustrefs init script start after nfs is mounted
5846 - fix CWARN/ERROR duplication (4930)
5847 - return async write errors to application if possible (2248)
5848 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5849 - print NAL number in %x format (4645)
5850 - update barely-supported suse-2.4.21-171 series (4842)
5851 - support for sles 9 %post scripts
5852 - support for building 2.6 kernel-source packages
5853 - support for sles km_* packages
5855 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5858 - ignore -ENOENT errors in osc_destroy (3639)
5859 - notify osc create thread that OSC is being cleaned up (4600)
5860 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5861 - reconstruct ost_handle() like mds_handle() (4657)
5862 - create a new thread to do import eviction to avoid deadlock (3969)
5863 - let lconf resolve symlinked-to devices (4629)
5864 - don't unlink "objects" from directory with default EA (4554)
5865 - hold socknal file ref over connect in case target is down (4394)
5866 - allow more than 32000 subdirectories in a single directory (3244)
5867 - fix blocks count for O_DIRECT writes (3751)
5868 - OST returns ENOSPC from object create when no space left (4539)
5869 - don't send truncate RPC if file size isn't changing (4410)
5870 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5871 - bind to privileged port in socknal and tcpnal (3689)
5873 - rate limit CERROR/CWARN console message to avoid overload (4519)
5874 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5875 - basic mmap support (3918)
5876 - kernel patch series update from b1_4 (4711)
5878 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5881 - avoid crash during MDS cleanup with OST shut down (2775)
5882 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5883 - don't use bad inodes on the MDS (3744)
5884 - dynamic object preallocation to improve recovery speed (4236)
5885 - don't hold spinlock over lock dumping or change debug flags (4401)
5886 - don't zero obd_dev when it is force cleaned (3651)
5887 - print grants to console if they go negative (4431)
5888 - "lctl deactivate" will stop automatic recovery attempts (3406)
5889 - look for existing locks in ldlm_handle_enqueue() (3764)
5890 - don't resolve lock handle twice in recovery avoiding race (4401)
5891 - revalidate should check working dir is a directory (4134)
5893 - don't always mark "slow" obdfilter messages as errors (4418)
5895 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5898 - don't close LustreDB during write_conf until it is done (3860)
5899 - fix typo in lconf for_each_profile (3821)
5900 - allow dumping logs from multiple threads at one time (3820)
5901 - don't allow multiple threads in OSC recovery (3812)
5902 - fix debug_size parameters (3864)
5903 - fix mds_postrecov to initialize import for llog ctxt (3121)
5904 - replace config semaphore with spinlock (3306)
5905 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5906 - don't allow enqueue to complete on a destroyed export (3822)
5907 - down write_lock before checking llog header bitmap (3825)
5908 - recover from lock replay timeout (3764)
5909 - up llog sem before sending rpc (3652)
5910 - reduce ns lock hold times when setting kms (3267)
5911 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5912 - fix NULL deref and obd_dev leak on setup error (3312)
5913 - replace some LBUG about llog ops with error handling (3841)
5914 - don't match INVALID dentries from d_lookup and spin (3784)
5915 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5916 - fix invalid assertion in ptlrpc_set_wait (3880)
5918 - add libwrap support for the TCP acceptor (3996)
5919 - add /proc/sys/portals/routes for non-root route listing (3994)
5920 - allow setting MDS UUID in .xml (2580)
5921 - print the stack of a process that LBUGs (4228)
5923 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5926 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5927 - make sure to unset replay flag from failed open requests (3440)
5928 - if default stripe count is 0, use OST count for inode size (3636)
5929 - update parent mtime/ctime on client for create/unlink (2611)
5930 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5931 - free recovery state on server during a forced cleanup (3571)
5932 - unregister_reply for resent reqs (3063)
5933 - loop back devices mounting and status check on 2.6 (3563)
5934 - fix resource-creation race that can provoke i_size == 0 (3513)
5935 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5936 - more debugging for page-accounting assertion (3746)
5937 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5938 - avoid deadlock after precreation failure (3758)
5939 - fix race and lock order deadlock in orphan handling (3450, 3750)
5940 - add validity checks when grabbing inodes from l_ast_data (3599)
5942 - add /proc/.../recovery_status to obdfilter (3428)
5943 - lightweight CDEBUG infrastructure, debug daemon (3668)
5944 - change default OSC RPC parameters to be better on small clusters
5945 - turn off OST read cache for files smaller than 32MB
5946 - install man pages and include them in rpms (3100)
5947 - add new init script for (un)mounting lustre filesystems (2593)
5948 - run chkconfig in %post for init scripts (3701)
5949 - drop scimac NAL (unmaintained)
5951 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5954 - clean kiobufs before and after use (3485)
5955 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5956 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5957 - change init script to fail more clearly if not run as root (1528)
5958 - allow clients to reconnect during replay (1742)
5959 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5960 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5961 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5962 - fix cond_resched (3554)
5963 - only evict unfinished clients after recovery (3515)
5964 - allow bulk resend, prevent data loss (3570)
5965 - dynamic ptlrpc request buffer allocation (2102)
5966 - don't allow unlinking open directory if it isn't empty (2904)
5967 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5968 - remove extraneous obd dereference causing LASSERT failure (3334)
5969 - don't use get_cycles() when creating temp. files on the mds (3156)
5970 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5971 - handle EEXIST for set-stripe, set proper directory name (3336)
5973 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5974 - fix ksocknal_fmb_callback() error messages (2918)
5976 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5979 - don't copy lvb into (possibly NULL) reply on error (2983)
5980 - don't deref dentry after dput, don't free lvb on error (2922)
5981 - use the kms to determine writeback rpc length (2947)
5982 - increment oti_logcookies when osc is inactive (2948)
5983 - update client's i_blocks count via lvb messages (2543)
5984 - handle intent open/close of special files properly (1557)
5985 - mount MDS with errors=remount-ro, like obdfilter (2009)
5986 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5987 - don't use cancelling-locks' kms values (2947)
5988 - use highest lock extent for kms, not last one (2925)
5989 - don't dereference ERR_PTR() dentry in error handling path (3107)
5990 - fix thread race in portals_debug_dumplog() (3122)
5991 - create lprocfs device entries at setup instead of at attach (1519)
5992 - common AST error handler, don't evict client on completion race (3145)
5993 - zero nameidata in detach_mnt in 2.6 (3118)
5994 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5995 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5996 - initialize RPC timeout timer earlier for 2.6 (3219)
5997 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5998 - print nal/nid for unknown nid (3258)
5999 - additional checks for oscc recovery before doing precreate (3284)
6000 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6001 - don't crash in mdc_close for bad permissions on open (3285)
6002 - zero i_rdev for non-device files (3147)
6003 - clear page->private before handing to FS, better assertion (3119)
6004 - tune the read pipeline (3236)
6005 - fix incorrect decref of invalidated dentry (2350)
6006 - provide read-ahead stats and refine rpc in flight stats (3328)
6007 - don't hold journal transaction open across create RPC (3313)
6008 - update atime on MDS at close time (3265)
6009 - close LDAP connection when recovering to avoid server load (3315)
6010 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6011 - don't leak open file on MDS after open resend (3325)
6012 - serialize filter_precreate and filter_destroy_precreated (3329)
6013 - loop device shouldn't call sync_dev() for nul device (3092)
6014 - clear page cache after eviction (2766)
6015 - resynchronize MDS->OST in background (2824)
6016 - refuse to mount the same filesystem twice on same mountpoint (3394)
6017 - allow llmount to create routes for mounting behind routers (3320)
6018 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6019 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6020 - fix rare problem with rename on htree directories (3417)
6022 - allow default OST striping configuration per directory (1414)
6023 - fix compilation for qswnal for 2.6 kernels (3125)
6024 - increase maximum number of MDS request buffers for large systems
6025 - change liblustreapi to be useful for external progs like lfsck (3098)
6026 - increase local configuration timeout for slow disks (3353)
6027 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6029 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6032 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6033 - don't overwrite extent policy data in reply if lock was blocked (2901)
6034 - drop filter export grants atomically with removal from device (2663)
6035 - del obd_self_export from work_list in class_disconnect_exports (2908)
6036 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6037 - swab reply message in mdc_close, other PPC fixes (2464)
6038 - fix destroying of named logs (2325)
6039 - overwrite old logs when running lconf --write_conf (2264)
6040 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6041 - fix race in target_handle_connect (2898)
6042 - mds_reint_create() should take same inode create lock (2926)
6043 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6044 - don't close files for self_export to avoid uninitialized obd (2936)
6045 - allow MDS with the same name as client node (2939)
6046 - hold dentry reference for closed log files for unlink (2325)
6047 - reserve space for all logs during transactions (2059)
6048 - don't evict page beyond end of stripe extent (2925)
6049 - don't oops on a deleted current working directory (2399)
6050 - handle hard links to targets without a parent properly (2517)
6051 - don't dereference NULL lock when racing during eviction (2867)
6052 - don't grow lock extents when lots of conflicting locks (2919)
6054 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6057 - account for cache space usage on clients to avoid data loss (974)
6058 - lfsck support in lustre kernel code (2349)
6059 - reduce journal credits needed for BRW writes (2370)
6060 - orphan handling to avoid losing space on client/server crashes
6061 - ptlrpcd can be blocked, stopping ALL progress (2477)
6062 - use lock value blocks to assist in proper KMS, faster stat (1021)
6063 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6064 - recovery for initial connections (2355)
6065 - fixes for mds_cleanup_orphans (1934)
6066 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6067 - block all file creations until orphan recovery completes (1901)
6068 - client remove rq_connection from request struct (2423)
6069 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6070 - recovery timer race (2670)
6071 - mdc_close recovey bug (2532)
6072 - ptlrpc cleanup bug (2710)
6073 - mds timeout on local locks (2588)
6074 - namespace lock held during RPCs (2431)
6075 - handle interrupted sync write properly (2503)
6076 - don't try to handle a message that hasn't been replied to (2699)
6077 - client assert failure during cleanup after abort recovery (2701)
6078 - leak mdc device after failed mount (2712)
6079 - ptlrpc_check_set allows timedout requests to complete (2714)
6080 - wait for inflight reqs when ptlrpcd finishes (2710)
6081 - make sure unregistered services are removed from the srv_list
6082 - reset bulk XID's when resending them (caught by 1138 test)
6083 - unregister_bulk after timeout
6084 - fix lconf error (2694)
6085 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6086 - readahead locks pages, leaves pending causing memory pressure (2673)
6087 - increase OST request buffers to 4096 on large machines (2729)
6088 - fix up permission of existing directories in simple_mkdir (2661)
6089 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6090 - don't assign transno to errored transactions (2742)
6091 - don't delete objects on OST if given a bogus objid from MDS (2751)
6092 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6093 - if rq_no_resend, then timeout request after recovery (2432)
6094 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6095 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6096 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6097 - MDS recovery shouldn't delete objects that it has given out (2730)
6098 - if enqueue arrives after completion, don't clobber LVB (2819)
6099 - don't unlock pages twice when trigger_group_io returns error (2814)
6100 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6101 - don't write pages to disk if there was an error (1450)
6102 - don't ping imports that have recovery disabled (2676)
6103 - take buffered bytes into account when balancing socknal conn (2817)
6104 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6105 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6106 - remove little-endian swabbing of llog records (1987)
6107 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6108 - retry reposting request buffers if they fail (1191)
6109 - grow extent at grant time to avoid granting a revoked lock (2809)
6110 - lock revoke doesn't evict page if covered by a second lock (2765)
6111 - disable VM readahead to avoid reading outside lock extents (2805)
6113 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6114 - updated kernel patches for hp-2.4.20 kernel (2681)
6116 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6119 - fix truncated write corruption (2366)
6120 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6122 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6123 - ensure len(uuid) < 37 in lmc (1171)
6124 - fix ia64 OOPS in llog_test (2255)
6125 - zero end of page at obdfilter for partial page writes (2648)
6126 - don't leave stale dentries around after renames (bug 2428)
6127 - fix timeouts when evicting a client with a single lock held (2642)
6128 - set deadline for the initial HELLO message to drain (2634)
6129 - print out dotted-quad IP addresses in the socknal (2302)
6131 - additional debugging for MDS client eviction problem (2443)
6132 - fix mkfsoptions support for osts (2603, 2604)
6134 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6137 - add series for the vanilla 2.6.0 kernel
6138 - add series for the vanilla 2.4.24 kernel
6139 - add series for a cray x86/64 UL kernel drop
6140 - fix xattr patches for the vanilla 2.4.19 series
6142 - generate true UUIDs in lmc (1171)
6143 - have portals stack dumping break in UML (2466)
6144 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6145 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6146 - wake callback waiting threads on client eviction (2460)
6147 - Add --ptldebug and --subsystem to lmc (1719)
6148 - update assertion to allow safe interrupt allocation
6149 - set rq_no_resend for cancel requests (2432)
6150 - recalculate ptlrpcd timeout after resend (2494)
6151 - call vfs_rmdir when removing pending directories (2368)
6152 - fix renaming a file to itself (2429)
6153 - lmc creates a default one-stripe lov (2454)
6154 - expand procfs space to handle large clusters (2326)
6155 - increase UML stack to avoid overflow
6156 - update lconf's list of debug and subsystem masks
6157 - fix lfs find --obd (2510)
6158 - /proc tunable for disabling filter read caching (2591)
6159 - stop rpm packages from altering slapd.conf (2301)
6160 - disable nagle in the socknal under 0conf (2578)
6161 - choose mds inode size based on stripe count (2572)
6162 - fix kernel-source rpm problems (2516)
6164 - add --disable-doc to avoid pdf generation (2421)
6165 - update documentation, tests, type-os, comments
6166 - avoid format warnings on ia64
6167 - remove the TOE NAL
6168 - tiny code cleanups by removing unused fields
6170 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6173 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6174 - remove the most blatant lies from BUILDING file (2371)
6175 - change default debug level to reasonable production setting
6176 - reduce client side cache size to reduce cache flush time
6177 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6178 - make TCP zerocopy and pinger support enabled by default (2476)
6179 - sync writes completed after process exits caused crashes (2319)
6180 - maintain correct mount count on the MDS (2356)
6181 - backout 1557, because 2316 wasn't really fixed
6182 - better file I/O statistics gathering in /proc
6183 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6184 - another kernel patch to fix zero-copy TCP function export
6185 - don't take duplicate lock when processing re-sent getattr (2420)
6186 - lctl uses obd_self_export instead of creating new conn (2353)
6187 - MDS/OST recovery case which requires object creation asserted (2425)
6188 - move lfs from /usr/sbin to /usr/bin in packages
6189 - fix race between mds_client_add and mds_client_free (2417)
6190 - use kmalloc instead of slabs in portals (2430)
6191 - don't create duplicate records when a failover MDS is present (2442)
6192 - remove unnecessary mount age check (2332)
6193 - don't remove directory inodes from locks prematurely (2451)
6194 - don't break if MDS service name is the same as hostname (2103)
6195 - fix races in client write RPC generation when cache full (2482)
6197 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6200 - remove now-unused request->rq_obd (278)
6201 - if an allocation fails, print out how much memory we've used (1933)
6202 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6203 - add the "configurable stack size" patch to most series files (1256)
6204 - ability to write large log records, for 100+ OST configs (2306)
6205 - fix NULL deref when filter_prep fails (2314)
6206 - fix operator precedence error in filter_sync
6207 - dynamic allocation of socknal TX descriptors (2315)
6208 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6209 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6210 - ensure that CWARN messages in Portals always get to the syslog
6211 - __init/__exit are not for prototype decls (ldlm_init/exit)
6212 - x86-64 compile warning fixes
6213 - fix gateway LMC keyword conflict (2318)
6214 - fix MDS lock inversions in getattr/reint paths (1844)
6215 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6216 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6217 - DLM race condition prevented some lock evictions (2328)
6218 - ENOMEM detection and retry on socknal sends (2230)
6219 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6220 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6221 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6222 - better error messages when a client is rejected during recovery (1505)
6223 - avoid cancelling locks which were never granted, after failure (2330)
6224 - fix i_sem/journal inversion in mds_client_add (2333)
6225 - fix truncate/getattr lock cycle deadlock (2334)
6226 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6227 - fix two rare exit paths which could leak an l_lock() ref (2321)
6228 - fencepost error in MDS/OST orphan recovery (2226)
6229 - make log record alignment 8 bytes (1988)
6230 - lstripe now fails when requested offset > ost_count (2237)
6231 - ensure that all kernel series have a complete list.h (1607)
6232 - fix crashes in special-file operations (2316)
6233 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6235 - allow configurable automake binary, for testing new versions
6236 - small update to the lfs documentation
6238 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6240 * fix negative export reference count in fsfilt_sync (2312)
6242 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6243 * release candidate 0.9.1
6245 - orphans are moved into the PENDING directory for possible recovery
6246 - replayed opens now open by fid for orphan/rename safety (1042)
6247 - last close of an orphan inode generates a transno (683)
6248 - chdir() and mount() now pin the directory entry (1020)
6249 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6250 - discard very old requests without processing them (1502)
6251 - remove some common, well-understood CERRORs (1505)
6252 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6253 - clear "grant" flags in OST replies until OST grant code lands (1644)
6254 - fix read performance by not clobbering i_blksize on client (1598)
6255 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6256 - make LRU size a /proc tunable, clears locks when reduced (707)
6257 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6258 - server locks take references on exports now (1558)
6259 - build fixes for 2.4.20-rh trees (1663)
6260 - return an error from lov_create if all OSCs are inactive (1751)
6261 - fix import levels when a reconnect happens without a timeout (1597)
6262 - exit early from mds_open if we get a lookup error (1749)
6263 - partial page read at EOF wouldn't wait for disk before sending (1642)
6264 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6265 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6266 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6267 - write/truncate lock inversion (1639)
6268 - Don't auto-load obdclass, portals modules during cleanup (1495)
6269 - fix timestamps from jumping to "now" (1763)
6270 - extra journal assertions (1648)
6271 - add an extra multiunlink test (1771)
6272 - fix read_record/write_record API (1776)
6273 - fix leak of offset_extent, possible incorrect i_size later (1772)
6274 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6275 - Debugging for the kqswnal_get_idle_tx problems (1820)
6276 - Allow recovery to be attempted multiple times (1536)
6277 - Write out MDS last_rcvd file after it is first created (1600)
6278 - Fix tx_descriptor leak in failed transmit situations (1827)
6279 - ext3 journaling fixes for assertion failure after IO error (1871)
6280 - class_export_put() on freed export after completion AST error (1896)
6281 - Fix revalidate looping in VFS (1322)
6282 - Don't access a freed export during MDS_REINT timeout (1521)
6283 - Add open-unlink recovery support on the MDS (1673,1764)
6284 - Return an error if no MDS data was read from last_rcvd (1946)
6285 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6286 - Don't setup a disk device that doesn't match exported UUID (317)
6287 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6288 - avoid committing NULL handle in force close
6289 - local.sh is now a one-stripe LOV configuration
6290 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6291 - fix ext3 htree duplicate directory entry corruption (1516)
6292 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6293 - update to new LOV EA format (2097)
6294 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6296 2003-06-15 Phil Schwan <phil@clusterfs.com>
6299 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6300 - per-import/export recovery handling (958, 931, 959)
6301 - multiple last-rcvd slots, for serving multiple FSes (949)
6302 - connections are again shared between multiple imp/exports (963, 964)
6303 - "umount -f" would hang if any requests needed to be sent (393, 978)
6304 - avoid pinning large req buffer by copying for queued messages (989)
6305 - add "uuid" to "lctl device" command to help upcalls (991)
6306 - "open" RPCs with transnos would confuse recovery counters (1037)
6307 - do proper endian conversion of all wire messages (288, 340, 891)
6308 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6309 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6310 - fix ptlrpc_connection leak in target_handle_connect (1174)
6311 - fix import refcounting bug in OST and MDS cleanup (1134)
6312 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6313 - fix very unlikely obd_types race condition (501)
6314 - remove osc_open hack for echo_client (1187)
6315 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6316 - a failure in read_inode2 leads to deadlock (1139)
6317 - cancel ack-locks as soon as transaction is committed (1072)
6318 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6319 - make sure to commitrw after any preprw to avoid deadlock (1162)
6320 - failing to execute a file in a lustre FS would lock inode (1203)
6321 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6322 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6323 - obd_uuid2tgt would walk off the end of the list (1255)
6324 - on IA64 the osc would give portals incorrect bulk size (1258)
6325 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6326 - fix lock inversion caused by new llite matching code (1282)
6327 - limit the number of dirty pages on a client to 10MB (1286)
6328 - timed out locks were not being corrected cancelled (1289)
6329 - fix O_DIRECT above 4GB on IA-32 (1292)
6330 * major user-visible changes
6331 - fail out/fail over policy now controlled by the upcall (993)
6333 - add OBD_PING to check server availability and failure (954)
6334 - lustre messages are now sent in sending host order (288, 340, 891)
6335 - add eadatalen to MDS getattr reply (340)
6336 - OST read replies may contain second buffer, with per-page status (593)
6338 2003-03-11 Phil Schwan <phil@clusterfs.com>
6341 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6342 - failed open()s could cause deadlock; fixed (867, 869)
6343 - stop cancelling OST locks when files are closed (481)
6344 - overlapping XID spaces caused network corruption (851, 853)
6345 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6346 - setattr_raw added, to do single-RPC, server-side setattrs
6347 - lmc/lconf syntax change for OST UUIDs
6348 - fix crashy race condition between ptlrpc_free_req and osc_close
6349 - don't use request in mdc_enqueue if we hit a timeout (889)
6350 - don't set the inode i_size for regular files from the MDS (896)
6351 - handle out of order completion AST (842)
6352 - don't LBUG if a lock request times out after receiving AST (913)
6353 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6354 - if a bad lock AST arrives, send an error instead of dropping entirely
6355 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6356 - fix leak in bulk IO when only partially completed (899, 900, 926)
6357 - fix O_DIRECT for ia64 (55)
6358 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6359 - C-z after timeout could hang a process forever; fixed (977)
6361 - client-side I/O cache (678, 924, 929, 941, 970)
6363 - READPAGE and SETATTRs which don't take server-side locks get
6366 2003-02-11 Phil Schwan <phil@clusterfs.com>
6369 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6370 - Fix fsfilt_extN_readpage() to read a full page of directory
6371 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6372 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6373 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6374 - Most intent operations were removed, in favour of a new RPC mode
6375 that does a single RPC to the server and bypasses most of the VFS
6376 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6377 - Aggressively cancel local locks on DLM servers
6378 - mds_reint_unlink sends EA to the client if it's the last nlink.
6379 client uses that EA to unlink OST objects.
6380 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6381 - recursive symlinks were fixed (439)
6382 - fixed NULL deref in DEBUG_REQ
6383 - filter_update_lastobjid no longer calls sync, which annoyed extN
6384 - fixed multi-client small-writes to a single file problem (445)
6385 - fixed mtime updates during file writes (607)
6386 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6387 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6388 - fixed Linux OST size reporting problem (444, 656)
6389 - OST now updates object mtime with writes or setattr (607, 619)
6390 - client verifies file size before zeroing page past EOF (445)
6391 - OST now writes last allocated objid to disk with allocation (108)
6392 - LOV on echo now works (409)
6394 - mds_reint_unlink sends a new buffer, with the EA included. this
6395 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6396 set if a regular file was being unlinked, and it was the last link
6397 - use PtlGet from the target for bulk writes (315)
6398 - OST now updates object mtime with writes or setattr (607, 619)
6399 - LDLM now has a grant-time callback to revalidate locked items, if
6401 - Many MDS operations were reorganized to combat race conditions
6403 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6404 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6406 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6409 - Fully reactivate OST imports after reconnection (512, others)
6410 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6411 - More graceful error handling for truncating on dead OST (515)
6412 - Don't error out unless we're actually accessing dead stripes (474)
6413 - Fix garbage sizes when stripes are missing (410)
6414 - LRU counters were broken, causing constant lock purge (433, 432)
6415 - garbage on read from stripes with failed OSTs (441)
6416 - mark OSCs as active before reconnecting during recovery (438)
6417 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6418 - lfind did not preserve OST order in output (443)
6419 - symlinks cause hung clients, incorrect data (439)
6420 - stop dereferencing request after dropping refcount (457)
6421 - don't LASSERT(spin_is_locked) on non-SMP (455)
6422 - fixes for many rename() bugs
6423 - fstat didn't correctly synchronize attributes (399)
6424 - server must handle lock cancellation during blocking AST prep (487)
6425 - bulk descriptors were free()d too soon (511)
6426 - fix paths in lconf, which would load incorrect modules (451, 507)
6427 - fix confusing lconf 'host not found' error message (386)
6428 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6429 - fix race condition in mdc_blocking_ast() for inode access (526)
6430 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6431 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6432 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6433 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6434 - fix lstripe in several ways (kernel side) (527)
6435 - fix request leak in ldlm_cli_enqueue (262)
6436 - incorrect OSC was marked inactive after OST failure
6437 - call mds_fs_cleanup before unmounting filesystem (524)
6438 - fix races between taking ns_lock and ldlm_lock_change_resource
6439 - fix races updating LOV export open file list
6440 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6441 - fix recovery NULL deref in ldlm_cli_cancel_unused
6442 - fix some DLM races by using new hash table for lock handles (419)
6443 - permit the client to specify desired inodes, at replay
6444 - duplicate requests when we queue them for replay reintegration
6445 - fix last_rcvd offset calculation
6446 - sync after each recovered transaction, so we always make progress
6447 - never, not always, ERESTART requests without transnos
6448 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6450 - skip replay if the MDS says that the client is already connected
6451 - don't check for a recovery-enabled export to match lctl's UUID
6452 - don't INC_USE_COUNT for phantom exports
6453 - don't crash when cleaning up phantom exports (567)
6454 - don't double-finish or set replay data for errored mdc_open requests
6455 - abort requests when they time out, so we don't get old replies
6456 - send/receive replies for AST messages again
6457 - if the client says that it doesn't have the lock, cancel it on the
6459 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6460 mark it as destroyed, it will all work out when decref is called
6461 - fix module use counts (22, 581)
6463 - ASTs now expect a reply (server cancels lock on error reply)
6465 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6468 - fix many simultaneous client startup (392)
6469 - fix dentry->d_it clobbering
6470 - credentials weren't being shipped for readdir/getattr operations
6471 - remove invalid assertions triggered during some concurrent MD
6473 - proper Lustre versions added (336, 389)
6474 - fix memory leak for create error case (398)
6475 - fix LOV locking bug that would get cli/srv out of sync
6476 - fix echo client over LOV (409)
6477 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6478 - fix double-O_EXCL intent crash (424)
6479 - avoid sending multiple lock CANCELs (352)
6481 - MDS can do multi-client recovery (modulo bugs in new code)
6483 - many updates, edits, cleanups
6485 2002-11-18 Phil Schwan <phil@clusterfs.com>
6488 - fix null d_it dereference (346)
6489 - fix full OST/dbench hang (333)
6490 - fix permission problem with file removal (286)
6491 - fix removal of OSCs from LOV when they fail
6492 - fix NULL deref during bulk timeout (214)
6493 - fix problems related to multiple filesystems on one MDS (241)
6494 - fixed serious subtle metadata locking bugs
6495 - free locks on clients when inodes are removed due to memory
6497 - fix inode pointer in lock data (285)
6498 - partial support for multiple MDS on a single host (241)
6499 - data locks weren't cancelled at clear_inode time (290, 311)
6500 - intent locks could lead to unbounded lock growth (205)
6501 - added a maximum lock count, an LRU list, and a flusher
6502 - fix multiple rename (365)
6503 - properly abstracted the echo client
6504 - OSC locked 1 byte too many; fixed
6505 - rewrote brw callback code:
6506 - fixed recovery bugs related to LOVs (306)
6507 - fixed too-many-pages-in-one-write crash (191)
6508 - fixed (again) crash in sync_io_timeout (214)
6509 - probably fixed callback-related race (385)
6511 - Add capability to MDS protocol
6512 - LDLM cancellations and callbacks on different portals
6514 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6517 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6518 - defer open object destruction to close time (601981, 138)
6519 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6520 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6521 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6522 - don't call obd_setattr in ll_file_release for destroyed objects
6523 * protocol change to lustre_msg: move |version| and add |flags|
6524 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6525 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6526 * added replay of create, unlink, link and rename operations during
6527 MDS failover; recovery should be much more robust now
6528 * remove failed OSCs from LOVs (only lov_create uses this so far)
6529 * the lustre-HOWTO was brought (more) up to date (582544)
6531 2002-10-23 Phil Schwan <phil@clusterfs.com>
6534 - in-use dentries weren't being reused properly (617851)
6535 - prevent multiple LDLM setup (599178)
6536 - fix LOV size calculations for truncate (617853)
6537 - fix client handling of MDS intent errors (POSIX)
6538 - fix permission bug in lovstripe.c test (624321)
6539 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6540 - truncate past end of file could corrupt data
6541 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6542 - a race in recovery could return ETIMEDOUT to apps (623947)
6543 - building outside the source directory was fixed
6544 * the lustre-HOWTO was brought (more) up to date (582544)
6545 * major progress was made on recovery functionality
6547 2002-10-10 Phil Schwan <phil@clusterfs.com>
6550 - recovery deadlock fix
6551 - rm -rf causes LBUG fix (617817)
6552 - file open by multiple tasks fix (618962)
6553 - directory permissions bugs (602707 and 620007)
6554 - journal_stop fixed with locking (611313)
6555 - O_APPEND failures resolved (618273, perhaps 614459)
6556 - lconf PATH fix (619770)
6557 - IA64 build fix (621450)
6558 - RPC buffer sizes scale with amount of memory
6560 2002-10-01 Phil Schwan <phil@clusterfs.com>
6563 - locks would be cancelled without throwing away data pages,
6564 resulting in inconsistent data (605627)
6565 - inode attributes were not always being refreshed (605627, 612449)
6566 - lconf now continues to cleanup after lctl reports an error
6567 - MDS now enforces user permissions (602707)
6568 - lprocfs cleanup fixed, but not yet enabled (614157)
6569 - fixed infinite server hang, should a client not respond to an AST
6570 - avoid going into recovery if user calls readlink() with a buffer
6571 that's too small (613941)
6572 - AST RPCs no longer require replies (614867) -- this may be changed
6573 - don't crash server if client sends an IOV that's too big (611336)
6574 - fixed lock conversion deadlock (611892)
6575 - fixed the following of symlinks (614622)
6576 * recovery: the server can remove locks from a client that dies, other
6577 clients can make progress
6578 * more extN patch fixes
6579 * compile-time configurable ptlrpc buffer allocations
6581 - collaborative read cache document
6582 - Lustre Lite Performance CDR document-in-progress
6584 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6587 - fix typo in patch-2.4.18
6589 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6592 - clear ptlrpc request each time in handle_incoming_request()
6593 - unlink of files now destroys the object on the OST
6595 2002-09-19 Peter Braam <braam@clusterfs.com>
6597 * add hard link support
6598 * change obdfile creation method
6599 * kernel patch changed
6601 2002-09-19 Peter Braam <braam@clusterfs.com>
6604 - stack overflow bug in extN fixed
6606 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6608 * documentation updates
6609 - add man pages for config tools
6610 - update tests/README to describe testing with new config tools
6611 - finish metadata API descriptions
6612 * bug fixes and cleanups
6613 - statfs workaround for 16TB limit
6614 - LOV stripe allocation improved, can stripe on subset of OSTs
6615 - LOV file size/IO offset was wrong for files > 4GB in size
6616 - object EA data was being dropped, caused files to be unreadable
6617 - memory overflow with non-LOV OST caused memory corruption
6618 - fixed regression tests to work with new config tools, obdfilter
6619 - fixed bug when directory size became larger than 1 block
6620 - fixed bug (for single client case) when PWD was deleted
6621 - invalidate local directory pages when doing intent-based ops
6622 - avoid LDLM oops when lock callback contained bad data
6624 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6626 * documentation updates
6627 * bug fixes and cleanups
6628 - configuration tools
6631 - 64-bit compile warnings
6632 - 64-bit internal statfs data
6634 * test_brw on persistent OST devices
6636 * lprocfs (disabled)
6638 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6640 * documentation updates
6641 * bug fixes and cleanups
6642 * configuration tools
6644 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6647 * many small fixes to 0_5_4
6648 * io/network handling
6649 * thinkos in MDS operations
6651 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6654 * crucial basic fixes to 0.5.3
6655 * IOR, Iozone work over Elan
6658 2002-08-07 Phil Schwan <phil@clusterfs.com>
6659 * version 0_5_3, our first alpha
6660 * we use the new Portals iovs
6661 * documentation updates
6662 * bug fixes and cleanups
6663 * small changes in the DLM wire protocol
6665 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6666 * version 0_5_1 with some initial stability,
6667 * locking on MD and file I/O.
6668 * documentation updates
6669 * several bug fixes since 0.5.0
6670 * small changes in wire protocol
6672 2002-07-18 Phil Schwan <phil@clusterfs.com>
6674 * delivered as Lustre Light Alpha
6675 * fixed a crash after handling invalid MDS requests
6676 * fixed directory pages for architectures with non-4k pages sizes
6678 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6679 * release version v0_4_4
6680 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6682 * Fixes a number of interruption problems with OST operations.
6683 * Update documentation for portals header changes
6684 * Move all wire protocol structs/defines to lustre_idl.h
6685 * Fixes symlink length bug.
6686 * Add tcpdump to repository.
6688 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6689 * release version v0_4_3
6690 * Fixes statfs for inodes on extN.
6691 * Fixes bug in runtests which would delete /etc/hosts.
6692 * Use 64-bit object IDs wherever possible (not into VFS though)
6693 Remove ost_get_info, which is unused by lustre, and out of date.
6695 2002-07-03 Peter Braam <braam@clusterfs.com>
6696 * release version v0_4_2 Fixes a lookup error (type not passed)
6697 * move forward to head of Portals
6698 * move forward to latest Lustre kernel
6700 2002-06-25 Peter Braam <braam@clusterfs.com>
6701 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob