3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on 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 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-862.9.1.el7 (RHEL7.5)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.143-94.47 (SLES12 SP3)
17 3.13.0-101 (Ubuntu 14.04, ZFS only)
18 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
19 4.4.0-131 (Ubuntu 16.04)
20 vanilla linux 4.6.7 (ZFS only)
21 * Client known to build on unpatched kernels:
22 2.6.32-431.29.2.el6 (RHEL6.5)
23 2.6.32-504.30.3.el6 (RHEL6.6)
24 2.6.32-573.26.1.el6 (RHEL6.7)
25 2.6.32-642.15.1.el6 (RHEL6.8)
26 2.6.32-696.18.7.el6 (RHEL6.9)
27 3.10.0-862.9.1.el7 (RHEL7.5)
28 3.0.101-0.47.71 (SLES11 SP3)
29 3.0.101-107 (SLES11 SP4)
30 3.12.74-60.64.40 (SLES12 SP1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.143-94.47 (SLES12 SP3)
33 3.13.0-101 (Ubuntu 14.04)
34 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
35 4.4.0-131 (Ubuntu 16.04)
36 4.15.0-32 (Ubuntu 18.04)
38 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
39 * Recommended ZFS / SPL version: 0.7.9
40 * Tested with ZFS / SPL version: 0.7.9
41 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
42 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
43 stack overflow. For more information, please refer to bugzilla 17630.
44 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
45 the re-exporting nodes to be mounted with "32bitapi" mount option
46 * Note for Ubuntu 14.04.5:
47 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
48 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
50 --------------------------------------------------------------------------------
53 2018-04-03 Intel Corporation
55 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
56 for currently supported client and server kernel versions.
57 * Server known to build on patched kernels:
58 2.6.32-431.29.2.el6 (RHEL6.5)
59 2.6.32-504.30.3.el6 (RHEL6.6)
60 2.6.32-573.26.1.el6 (RHEL6.7)
61 2.6.32-642.15.1.el6 (RHEL6.8)
62 2.6.32-696.18.7.el6 (RHEL6.9)
63 3.10.0-693.21.1.el7 (RHEL7.4)
64 3.0.101-0.47.71 (SLES11 SP3)
65 3.0.101-107 (SLES11 SP4)
66 3.12.74-60.64.40 (SLES12 SP1)
67 4.4.114-92.64 (SLES12 SP2)
68 4.4.114-94.11 (SLES12 SP3)
69 3.13.0-101 (Ubuntu 14.04, ZFS only)
70 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
71 4.4.0-116 (Ubuntu 16.04)
72 vanilla linux 4.6.7 (ZFS only)
73 * Client known to build on unpatched kernels:
74 2.6.32-431.29.2.el6 (RHEL6.5)
75 2.6.32-504.30.3.el6 (RHEL6.6)
76 2.6.32-573.26.1.el6 (RHEL6.7)
77 2.6.32-642.15.1.el6 (RHEL6.8)
78 2.6.32-696.18.7.el6 (RHEL6.9)
79 3.10.0-693.21.1.el7 (RHEL7.4)
80 3.0.101-0.47.71 (SLES11 SP3)
81 3.0.101-107 (SLES11 SP4)
82 3.12.74-60.64.40 (SLES12 SP1)
83 4.4.114-92.64 (SLES12 SP2)
84 4.4.114-94.11 (SLES12 SP3)
85 3.13.0-101 (Ubuntu 14.04)
86 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
87 4.4.0-116 (Ubuntu 16.04)
89 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
90 * Recommended ZFS / SPL version: 0.7.6
91 * Tested with ZFS / SPL version: 0.7.6
92 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
93 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
94 stack overflow. For more information, please refer to bugzilla 17630.
95 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
96 the re-exporting nodes to be mounted with "32bitapi" mount option
97 * Note for Ubuntu 14.04.5:
98 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
99 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
101 --------------------------------------------------------------------------------
103 2017-07-05 Intel Corporation
105 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
106 for currently supported client and server kernel versions.
107 * Server known to build on patched kernels:
108 2.6.32-431.29.2.el6 (RHEL6.5)
109 2.6.32-504.30.3.el6 (RHEL6.6)
110 2.6.32-573.26.1.el6 (RHEL6.7)
111 2.6.32-642.15.1.el6 (RHEL6.8)
112 2.6.32-696.3.1.el6 (RHEL6.9)
113 3.10.0-514.21.1.el7 (RHEL7.3)
114 3.0.101-0.47.71 (SLES11 SP3)
115 3.0.101-100 (SLES11 SP4)
116 3.12.74-60.64.40 (SLES12 SP1)
117 4.4.59-92.17 (SLES12 SP2)
118 vanilla linux 4.6.7 (ZFS only)
119 * Client known to build on unpatched kernels:
120 2.6.32-431.29.2.el6 (RHEL6.5)
121 2.6.32-504.30.3.el6 (RHEL6.6)
122 2.6.32-573.26.1.el6 (RHEL6.7)
123 2.6.32-642.15.1.el6 (RHEL6.8)
124 2.6.32-696.3.1.el6 (RHEL6.9)
125 3.10.0-514.21.1.el7 (RHEL7.3)
126 3.0.101-0.47.71 (SLES11 SP3)
127 3.0.101-100 (SLES11 SP4)
128 3.12.74-60.64.40 (SLES12 SP1)
129 4.4.59-92.17 (SLES12 SP2)
131 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
132 * Recommended ZFS / SPL version: 0.6.5.9
133 * Tested with ZFS / SPL version: 0.6.5.9
134 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
135 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
136 stack overflow. For more information, please refer to bugzilla 17630.
137 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
138 the re-exporting nodes to be mounted with "32bitapi" mount option
140 --------------------------------------------------------------------------------
142 12-07-2016 Intel Corporation
144 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
145 for currently supported client and server kernel versions.
146 * Server known to build on patched kernels:
147 2.6.32-431.29.2.el6 (RHEL6.5)
148 2.6.32-504.30.3.el6 (RHEL6.6)
149 2.6.32-573.26.1.el6 (RHEL6.7)
150 2.6.32-642.11.1.el6 (RHEL6.8)
151 3.10.0-514.el7 (RHEL7.3)
152 3.0.101-0.47.71 (SLES11 SP3)
153 3.0.101-80 (SLES11 SP4)
154 3.12.62-60.64.8 (SLES12 SP1)
155 vanilla linux 4.5.7 (ZFS only)
156 * Client known to build on unpatched kernels:
157 2.6.32-431.29.2.el6 (RHEL6.5)
158 2.6.32-504.30.3.el6 (RHEL6.6)
159 2.6.32-573.26.1.el6 (RHEL6.7)
160 2.6.32-642.11.1.el6 (RHEL6.8)
161 3.10.0-514.el7 (RHEL7.3)
162 3.0.101-0.47.71 (SLES11 SP3)
163 3.0.101-80 (SLES11 SP4)
164 3.12.62-60.64.8 (SLES12 SP1)
166 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
167 * Recommended ZFS / SPL version: 0.6.5.8
168 * Tested with ZFS / SPL version: 0.6.5.8
169 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
170 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
171 stack overflow. For more information, please refer to bugzilla 17630.
172 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
173 the re-exporting nodes to be mounted with "32bitapi" mount option
175 --------------------------------------------------------------------------------
177 02-29-2016 Intel Corporation
179 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
180 for currently supported client and server kernel versions.
181 * Server known to build on patched kernels:
182 2.6.32-431.29.2.el6 (RHEL6.5)
183 2.6.32-504.30.3.el6 (RHEL6.6)
184 2.6.32-573.12.1.el6 (RHEL6.7)
185 3.10.0-327.3.1.el7 (RHEL7.2)
186 3.0.101-0.47.71 (SLES11 SP3)
187 3.0.101-68 (SLES11 SP4)
188 vanilla linux 4.2.1 (ZFS only)
189 * Client known to build on unpatched kernels:
190 2.6.32-431.29.2.el6 (RHEL6.5)
191 2.6.32-504.30.3.el6 (RHEL6.6)
192 2.6.32-573.12.1.el6 (RHEL6.7)
193 3.10.0-327.3.1.el7 (RHEL7.2)
194 3.0.101-0.47.71 (SLES11 SP3)
195 3.0.101-68 (SLES11 SP4)
198 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
199 * Recommended ZFS / SPL version: 0.6.4.2
200 * Tested with ZFS / SPL version: 0.6.4.2
201 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
202 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
203 stack overflow. For more information, please refer to bugzilla 17630.
204 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
205 the re-exporting nodes to be mounted with "32bitapi" mount option
207 --------------------------------------------------------------------------------
209 03-10-2015 Intel Corporation
211 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
212 for currently supported client and server kernel versions.
213 * Server known to build on patched kernels:
214 2.6.32-431.29.2.el6 (RHEL6.5)
215 2.6.32-504.8.1.el6 (RHEL6.6)
216 3.0.101-0.46 (SLES11 SP3)
217 * Client known to build on unpatched kernels:
218 2.6.32-431.29.2.el6 (RHEL6.5)
219 2.6.32-504.8.1.el6 (RHEL6.6)
220 3.10.0-123.20.1.el7 (RHEL7)
221 3.0.101-0.46 (SLES11 SP3)
222 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
223 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
224 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
225 stack overflow. For more information, please refer to bugzilla 17630.
226 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
227 the re-exporting nodes to be mounted with "32bitapi" mount option
229 Severity : enhancement
231 Description: control OST-index in IDIF via ROCOMPAT flag.
232 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
233 last_rcvd file. For new formatted OST device, it will be auto set;
234 for the case of upgrading from old OST device, you can enable it
235 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
236 enabled, for new created OST-object, its IDIF-in-LMA will contain
237 the OST-index; for the existing OST-object, the OSD will convert
238 old format IDIF as new format IDIF with OST-index stored in the
239 LMA EA when accessing such OST-object or via OI scrub. Once such
240 flag is enabled, it cannot be reverted back, so the system cannot
241 be downgraded to the orignal incompatible version.
243 --------------------------------------------------------------------------------
245 07-30-2014 Intel Corporation
247 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
248 for currently supported client and server kernel versions.
249 * Server known to build on patched kernels:
250 2.6.32-431.20.3.el6 (RHEL6.5)
251 3.0.101-0.31 (SLES11 SP3)
252 * Client known to build on unpatched kernels:
253 2.6.32-431.20.3.el6 (RHEL6.5)
254 3.0.101-0.31 (SLES11 SP3)
257 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
258 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
259 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
260 stack overflow. For more information, please refer to bugzilla 17630.
261 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
262 the re-exporting nodes to be mounted with "32bitapi" mount option
264 --------------------------------------------------------------------------------
266 10-31-2013 Intel Corporation
268 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
269 * Server builds for kernels:
270 2.6.32-358.18.1.el6 (RHEL6.4)
271 * Client builds for unpatched kernels:
272 2.6.32-358.18.1.el6 (RHEL6.4)
273 3.0.92-0.8 (SLES11 SP3)
275 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
276 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
277 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
278 stack overflow. For more information, please refer to bugzilla 17630.
279 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
280 the re-exporting nodes to be mounted with "32bitapi" mount option
282 --------------------------------------------------------------------------------
284 05-30-2013 Intel Corporation
286 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
287 * Server builds for kernels:
288 2.6.32-358.6.2.el6 (RHEL6.4)
289 * Client builds for unpatched kernels:
290 2.6.32-358.6.2.el6 (RHEL6.4)
291 3.0.74-0.6.6 (SLES11 SP2)
293 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
294 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
295 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
296 stack overflow. For more information, please refer to bugzilla 17630.
297 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
298 the re-exporting nodes to be mounted with "32bitapi" mount option
300 --------------------------------------------------------------------------------
302 9-30-2012 Whamcloud, Inc.
304 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
305 * Server support for kernels:
306 2.6.32-279.5.1.el6 (RHEL6)
307 * Client support for unpatched kernels:
308 2.6.32-279.5.1.el6 (RHEL6)
310 2.6.18-238.19.1.el5 (RHEL5)
311 2.6.16.60-0.69.1 (SLES10)
312 * Recommended e2fsprogs version: 1.41.90.wc4
313 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
314 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
315 stack overflow. For more information, please refer to bugzilla 17630.
316 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
317 the re-exporting nodes to be mounted with "32bitapi" mount option
319 Severity : enhancement
321 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
323 Severity : enhancement
325 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
327 Severity : enhancement
329 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
331 Severity : enhancement
333 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
335 Severity : enhancement
337 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
339 --------------------------------------------------------------------------------
341 04-30-2012 Whamcloud, Inc.
343 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
344 * Server support for kernels:
345 2.6.32-220.4.2.el6 (RHEL6)
346 * Client support for unpatched kernels:
347 2.6.32-220.4.2.el6 (RHEL6)
349 2.6.18-238.19.1.el5 (RHEL5)
350 2.6.16.60-0.69.1 (SLES10)
351 * Recommended e2fsprogs version: 1.41.90.wc4
352 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
353 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
354 stack overflow. For more information, please refer to bugzilla 17630.
355 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
356 the re-exporting nodes to be mounted with "32bitapi" mount option
358 Severity : enhancement
360 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
362 Severity : enhancement
364 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
366 Severity : enhancement
368 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
370 Severity : enhancement
372 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
374 --------------------------------------------------------------------------------
376 09-30-2011 Whamcloud, Inc.
378 * Server support for kernels:
379 2.6.18-238.19.1.el5 (RHEL5)
380 2.6.32-131.6.1.el6 (RHEL6)
381 * Client support for unpatched kernels:
382 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
383 2.6.18-238.19.1.el5 (RHEL5)
384 2.6.32-131.6.1.el6 (RHEL6)
385 2.6.29.4-167.fc11 (FC11)
386 2.6.16.60-0.69.1 (SLES10)
388 * Recommended e2fsprogs version: 1.41.90.wc3
389 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
390 Lustre filesystem with 4K stack may cause a stack overflow. For more
391 information, please refer to bugzilla 17630.
392 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
393 mounted with "32bitapi" mount option
395 Severity : enhancement
396 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
398 Severity : enhancement
399 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
401 Severity : enhancement
402 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
405 Frequency : only with SLES10
407 Description: Use OFED "KMP" provided by Novell
408 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
409 Lustre is now built against this package. That means you need to
410 install the ofed-kmp package from Novell for the patchless client.
411 Note that the ofed-kmp that Novell ships may not exactly match the
412 kernel version but should still be compatible.
414 Severity : enhancement
416 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
417 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
419 --------------------------------------------------------------------------------
421 2010-07-15 Oracle, Inc.
423 * Server support for kernels:
424 2.6.18-164.11.1.el5 (RHEL 5)
425 2.6.18-164.11.1.0.1.el5 (OEL 5)
426 * Client support for unpatched kernels:
427 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
428 2.6.18-164.11.1.el5 (RHEL 5),
429 2.6.18-164.11.1.0.1.el5 (OEL 5)
430 2.6.16.60-0.42.8 (SLES 10),
432 2.6.29.4-167.fc11 (FC11)
433 * Recommended e2fsprogs version: 1.41.10-sun2
435 Severity : enhancement
437 Description: Distribute lfs_migrate script with lustre
438 Details : The lfs_migrate mentioned in the manual is now included in the
439 lustre tree in lustre/scripts and will distributed in lustre RPMs.
441 Severity : enhancement
443 Description: Failover nid list added to import proc file.
445 Severity : enhancement
447 Description: "som_preview" mount option required to use SOM
448 Details : The clients will need to use som_preview mount option
452 Description: Lustre does not do 1MB IOs to HW RAID
453 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
455 Severity : enhancement
457 Description: lfs quota output is not very convenient for awk/sed-parsing
458 Details : Some positions in lfs quota output table could be empty or
459 non-empty which made it hard to parse it with scripts, now
460 a dash is put instead of space where there is not supposed
465 Description: Update to OFED 1.5.1
466 Details : OFED 1.5.1 will be used in environments that don't use
467 vendor supplied OFED.
471 Description: Remove lnet. prefix from lctl params display
472 Details : Also renamed lnet.memused to lnet_memused to
477 Description: Use conf_param -d to remove permanant settings.
481 Description: Add list_param -R
482 Details : lctl list_param now supports recursive mode.
485 Bugzilla : 21829, 22312
486 Description: Fix llobdstat.
487 Details : 1. Now updating the write_bytes counter
488 2. Port script fixes from 1.8.
492 Description: decrease the usage of memory on clients.
493 Details : 1. On clients, recycle dentries and inodes unused.
494 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
499 Description: llobdstat fix and enhancement
500 Details : add a counter to set a limit to how many samples will be returned
501 fix a wildcard in the path to limit to obdfilter stats only
505 Description: lfs check only allowed for root
506 Details : Code cleanups and added a regresion test of non-root use of
511 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
512 Details : Fix hr->hr_index overload.
516 Description: Fix FIEMAP functionality.
520 Description: FC11 patchless client support added
524 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
525 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
529 Description: New acc-sm script and various test-framework enhancments.
533 Description: Previously landed patch was reverted.
537 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
541 Description: Add ldiskfs maxdirsize mount option.
545 Description: Fix mkfs.lustre support for 16TB LUNs.
549 Description: Print mmp_check_interval
550 Details : print mmp_check_interval and make it possible to abort
551 mount operation in case it takes too long.
555 Description: Add cascading_rw to lustre mpi tests.
559 Description: Store SOM data in EA.
560 Details : Also several SOM related cleanups.
564 Description: Add refcount to md_open_data
568 Description: Use refcount on lov_request_set to prevent memory corruption.
572 Description: Statahead: use parent dir's i_mutex for synchronization.
576 Description: loadgen improvements.
580 Description: Enable peer health detection in ksocklnd and ko2iblnd.
584 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
588 Description: Fix race on obd_devcie:md_stats
592 Description: Fix SLES 11 client builds.
593 Details : Resolved uninitalized pointer gcc warnings.
597 Description: Allow non-root access for "lfs check".
598 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
602 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
604 Severity : enhancement
606 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
607 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
611 Description: Sanity test_27p fails.
612 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
613 Sanity tests 27 were cleaned up and improved.
617 Description: Unify req format on client/servers
618 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
619 buffers and swabbers approach..
624 Description: drop join file stuff.
626 Severity : enhancement
628 Description: Add 2.6.27 and SLES11 (patchless client) support
632 Description: Fix in ptlrpc_expire_one_request() to print the signed time
635 Severity : enhancement
637 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
640 Severity : enhancement
642 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
644 Severity : enhancement
646 Description: correctly handle big reply message.
647 Details : send LNet event if reply is bigger than buffer and adjust this buffer
652 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
654 Severity : enhancement
656 Description: Add support for OEL5.
660 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
662 Severity : enhancement
664 Description: "weak-modules" support
665 Details : Implement "weak-modules" support which enables kernel modules
666 to be used with any kernel that implements the same kABI. In
667 order to achieve this modules are now installed in
668 /lib/modules/$(uname -r)/updates/kernel on all distributions.
672 Description: client could not reconnect to OST because of an active request.
673 Details : The client now retries to reconnect to the same server, if a
674 connect request failed with EBUSY or -EAGAIN. Test case 35b
675 conf-sanity.sh has been created.
679 Description: Don't allow make backward step on assignin osc next id.
680 Details : race between allocation next id and ll_sync thread can be cause
681 of set wrong osc next id and can be kill valid ost objects.
685 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
689 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
693 Description: Error handling in osc_statfs_interpret() has been improved.
694 Details : Check in osc_statfs_interpret() for EBADR.
698 Description: Do not send statfs() requests to OSTs disabled by administrator.
699 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
701 Severity : enhancement
703 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
705 Severity : enhancement
707 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
709 Severity : enhancement
711 Description: Enhancement for lfs(1) command to use numeric uid/gid.
715 Description: Improved cleanup procedure for conf-sanity test_46a.
719 Description: Do not start bulk transfer, if client tried to reconnect.
720 Details : Do not start bulk transfer, if client tried to reconnect,
721 but failed due to active rpcs. Check for exp_abort_active_req
722 has been implemented.
726 Description: Prevent inconsistences between linux and lustre mount structures.
727 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
728 Make the sleep interruptible.
730 Severity : enhancement
732 Description: provide server to client comms path
733 Details : server to client communications path via new LDLM_SET_INFO rpc
737 Description: truncate starts GFP_FS allocation under transaction causing deadlock
738 Details : ldiskfs_truncate calls grab_cache_page which may start page
739 allocation under an open transaction. This may lead to
740 calling prune_icache with consequent lustre reentrance.
744 Description: send correctly lsm on open replay
745 Details : MDS is trust to LSM size on replay open, but client can set wrong size
750 Description: enable client interface failover
751 Details : When a child reconnects from another NID, properly update export
752 nid hash position and ldlm reverse import.
754 Severity : enhancement
756 Description: Communicate OST degraded/readonly state via statfs to MDS
757 Details : Flags in the statfs returned from OSTs indicate whether the
758 OST is in a degraded RAID state, or if the filesystem has
759 turned read-only after a filesystem error is detected.
764 Description: don't panic if EPROTO was hit when reading symlink
765 Details : correctly handling request reference in error cases.
767 Severity : enhancement
769 Description: Add LustreNetLink, a kernel-userspace communcation path.
771 Severity : enhancement
773 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
778 Description: access to llog context before init.
779 Details : move handling CATALOGS file at osc layer and forbid access to llog
783 Frequency : rare, only when client is writing to OST on local node
785 Description: Avoid deadlock for local client writes
786 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
787 memory freeing context. This allows OST threads to set the
788 PF_MEMALLOC flag on task structures in order to allocate memory
789 from reserved pools and complete IO.
790 Use GFP_HIGHUSER for OST allocations for non-local client writes,
791 so that the OST threads generate memory pressure and allow
792 inactive pages to be reclaimed.
794 Severity : enhancement
796 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
801 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
802 Details : move ldlm namespace creation in setup phase to avoid grab
803 _lprocfs_lock with cli_sem held.
807 Description: Temporarily disable grant shrink.
808 Details : Disable the feature for debugging.
811 Frequency : only when formatting test filesystems
813 Description: Unable to run several mkfs.lustre on loop devices at the same time
814 Details : mkfs.lustre returns error 256 on the concurrent loop devices
815 formatting. The solution is to proper handle the error.
817 Severity : enhancement
819 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
821 Severity : enhancement
823 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
828 Description: resolve race between obd_disconnect and class_disconnect_exports
829 Details : if obd_disconnect will be called to already disconnected export he
830 forget release one reference and osc module can't unloaded.
832 Severity : enhancement
834 Description: Allow tuning service thread via /proc
835 Details : For each service a new
836 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
837 created that can be used to set min/max thread counts, and get the
838 current number of running threads.
843 Descriptoin: don't return error if have partially created objects for file.
844 Details : lov_update_create_set uses set->set_success as index for created
845 objects, so if some requests will be failed, they will have hole
846 at end of array and qos_shrink_lsm for allocate correct lsm.
848 Severity : enhancement
850 Description: Update OFED support to 1.4.1
855 Description: Reduce small size read RPC
856 Details : Set read-ahead limit for every file and only do read-ahead when
857 available read-ahead pages are bigger than 1M to avoid small size
860 Severity : enhancement
862 Description: move AT tunable parameters for more consistent usage
863 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
865 Severity : enhancement
867 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
868 Details : allow skip disconnected ost for send statfs request and hide error
874 Description: fix racy locking of mballoc block bitmaps causing BUG
875 Details : The locking of the mballoc buddy bitmap and the in-memory
876 block bitmap was using two different spin locks in some
877 cases. This made it possible to incorrectly access the
878 mballoc bitmap while another process was modifying it,
879 causing a sanity assertion to fail. While no on-disk corruption
880 was reported, there was some risk of this happening.
883 Frequency : rare, on llog test 6
885 Descriptoin: don't allow connect to already connected import
886 Details : allowing connect to already connected import is hide connecting problem.
891 Descriptoin: allow kill process which wait statahead result
892 Details : for some reasons 'ls' can stick in waiting result from statahead,
893 in this case need way for kill this process.
895 Severity : enhancement
897 Description: Add state history info file, enhance import info file
898 Details : Track import connection state changes in a new osc/mdc proc file;
899 add overview-type data to the osc/mdc import proc file.
901 Severity : enhancement
903 Description: MDS create should not wait for statfs RPC while holding DLM lock.
905 Severity : enhancement
907 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
912 Descriptoin: OST grant too much space to client even there are not enough space.
913 Details : Client will shrink its grant cache to OST if there are no write
914 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
915 this grant cache if there are already not enough avaible space
916 (left_space < total_clients * 32M).
919 Frequency : start MDS on uncleanly shutdowned MDS device
921 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
922 Details : stay in waiting mds<>ost recovery finished produce random bugs
923 due race between two ll_sync thread for one lov target. send
924 ACTIVATE event only if connect realy finished and import have
928 Frequency : rare, connect and disconnect target at same time
930 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
931 Details : don't call obd_disconnect under lov_lock. this long time
932 operation and can block ptlrpcd which answer to connect request.
937 Descriptoin: don't lose wakeup for imp_recovery_waitq
938 Details : recover_import_no_retry or invalidate_import and import_close can
939 both sleep on imp_recovery_waitq, but we was send only one wakeup
943 Frequency : always with long access acl
945 Descriptoin: mds can't pack reply with long acl.
946 Details : mds don't control size of acl but they limited by reint/getattr
949 Severity : enhancement
951 Description: Update to SLES10 kernel-2.6.16.60-0.33.
953 Severity : enhancement
955 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
958 Frequency : start MDS on uncleanly shutdowned MDS device
960 Descriptoin: aborting recovery hang on MDS
961 Details : don't throttle destroy RPCs for the MDT.
964 Frequency : on remount
966 Description: external journal device not working after the remount
967 Details : clear dev_rdonly flag for external journal devices in
973 Description: shutdown vs evict race
974 Details : client_disconnect_export vs connect request race.
975 if client will evicted at this time - we start invalidate
976 thread without referece to import and import can be freed
980 Frequency : rare, need acl's on inode.
982 Description: client can't handle ost additional correctly
983 Details : if ost was added after client connected to mds client can have
984 hit lnet_try_match_md ... to big messages to wide striped files.
985 in this case need teach client to handle config events about add
986 lov target and update client max ea size at that event.
988 Severity : enhancement
990 Description: Changelogs
991 Details : Changelogs are a lightweight mechanism to track filesystem
992 metadata and namespace changes. The changelog is recorded
993 permanently on the MDTs, and is periodically "consumed" / purged
994 when records are no longer needed.
996 Severity : enhancement
998 Description: compact fld format with extents
999 Details : Store range of seq rather than every seq in FLD. Seq
1000 controller update FLD rather than clients. In Case of CMD, mdt0
1001 has FLD, all other metadata server act as non persistent proxy
1002 for FLD queries and cache fld entries in fld cache.
1007 Description: don't skip ost target if they assigned to file
1008 Details : Drop slow OSCs if we can, but not for requested start idx.
1009 This means "if OSC is slow and it is not the requested
1010 start OST, then it can be skipped, otherwise skip it only
1011 if it is inactive/recovering/out-of-space.
1015 Description: more cleanup in mds_lov
1016 Details : not send LOV EA under replay, we can't know about they size at this
1017 time. Don't allow client connect to mds before any ost connected,
1018 for avoid problems with LOV EA size and returning EIO to client.
1020 Severity : enhancement
1022 Description: Interoperability at server side (Disk interoperability)
1024 Severity : enhancement
1026 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1028 Severity : enhancement
1030 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1032 Severity : enhancement
1034 Description: New client IO stack (CLIO).
1036 Severity : enhancement
1038 Description: Commit on sharing. Eliminate inter-client dependencies between
1039 uncommitted transactions by doing transaction commits.
1040 Thereby clients may recovery independently.
1043 Frequency : Create a symlink file with a very long name
1045 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1046 Details : If there is no extra space in the request for early cancels,
1047 ldlm_req_handles_avail() returns 0 instead of a negative value.
1049 Severity : enhancement
1051 Description: Add /proc entry for import status
1052 Details : The mdc, osc, and mgc import directories now have
1053 an import directory that contains useful import data for debugging
1054 connection problems.
1056 Severity : enhancement
1058 Description: Re-disable certain /proc logging
1059 Details : Enable and disable client's offset_stats, extents_stats and
1060 extents_stats_per_process stats logging on the fly.
1063 Frequency : Only on FC kernels 2.6.22+
1065 Description: oops in statahead
1066 Details : Do not drop reference count for the dentry from VFS when lookup,
1067 VFS will do that by itself.
1069 Severity : enhancement
1071 Description: Generic /proc file permissions
1072 Details : Set /Proc file permissions in a more generic way to enable non-
1073 root users operate on some /proc files.
1077 Description: Hitting mdc_commit_close() ASSERTION
1078 Details : Properly handle request reference release in
1079 ll_release_openhandle().
1083 Description: quota recovery deadlock during mds failover
1084 Details : This patch includes att18982, att18236, att18237 in bz14840.
1086 1. fix osts hang when mds does failover with quotaon
1087 2. prevent watchdog storm when osts threads wait for the
1092 Frequency : only patchless client
1093 Description: add workaround for race between add/remove dentry from hash
1095 Severity : enhancement
1097 Description: Allow OST glimpses to return PW locks
1101 Description: LBUG when llog conf file is full
1102 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1106 Description: Prevent import from entering FULL state when server in recovery
1110 Description: service mount cannot take device name with ":"
1111 Details : Only when device name contains ":/" will mount treat it as
1117 Description: replace ptlrpcd with the statahead thread to interpret the async
1118 statahead RPC callback
1122 Frequency : on recovery
1123 Description: I/O failures after umount during fail back
1124 Details : if client reconnected to restarted server we need join to recovery
1125 instead of find server handler is changed and process self
1126 eviction with cancel all locks.
1128 Severity : enhancement
1130 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1132 Severity : enhancement
1134 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1136 Severity : enhancement
1138 Description: Upcall on Lustre log has been dumped
1139 Details : Allow for a user mode script to be called once a Lustre log has
1140 been dumped. It passes the filename of the dumped log to the
1141 script, the location of the script can be specified via
1142 /proc/sys/lnet/debug_log_upcall.
1147 Description: avoid idr_remove called for id which is not allocated.
1148 Details : Move assigment s_dev for clustered nfs to end of initialization,
1149 to avoid problem with error handling.
1154 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1155 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1156 into CONN_USED_HASH and this prodice warning when put connection
1157 again in unused hash.
1160 Severity : enhancement
1162 Description: File striping can now be set to use an arbitrary pool of OSTs.
1164 Severity : enhancement
1166 Description: Export bytes_read/bytes_write count on OSC/OST.
1170 Description: Early reply size mismatch, MGC loses connection
1171 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1172 the connect flags are properly negotiated.
1177 Description: quotas are not honored with O_DIRECT
1178 Details : all writes with the flag O_DIRECT will use grants which leads to
1179 this problem. Now using OBD_BRW_SYNC to guard this.
1183 Description: add quota statistics
1184 Details : 1. sort out quota proc entries and proc code.
1185 2. add quota statistics
1187 Severity : enhancement
1189 Description: enable quota support for HEAD.
1193 Description: Properly propagate oinfo flags from lov to osc for statfs
1194 Details : restore missing copy oi_flags to lov requests.
1196 Severity : enhancement
1198 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1199 : l_getgroups(8), lst(8), routerstat(8)
1200 Details : included man pages for llobdstat(8), llstat(8),
1201 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1203 Severity : enhancement
1205 Description: configure's --enable-quota should check the
1206 : kernel .config for CONFIG_QUOTA
1207 Details : configure is terminated if --enable-quota is passed but
1208 : no quota support is in kernel
1212 Description: Remove portals compatibility
1213 Details : Remove portals compatibility, not interoperable with releases
1218 Description: Resolve device initialization race
1219 Details : Prevent proc handler from accessing devices added to the
1220 obd_devs array but yet be intialized.
1222 Severity : enhancement
1224 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1226 Severity : enhancement
1228 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1233 Description: Using wrong pointer in osc_brw_prep_request
1234 Details : Access to array[-1] can produce panic if kernel compiled with
1235 CONFIG_PAGE_ALLOC enabled
1239 Description: Client runs out of low memory
1240 Details : Consider only lowmem when counting initial number of llap pages
1244 Description: *optional* service tags registration
1245 Details : if the "service tags" package is installed on a Lustre node
1246 When the filesystem is mounted, a local-node service tag will
1247 be created. See http://inventory.sun.com/ for more information
1248 about the Service Tags asset management system.
1252 Description: Kernel BUG tries to release flock
1253 Details : Lustre does not destroy flock lock before last reference goes
1254 away. So always drop flock locks when client is evicted and
1255 perform unlock regardless of successfulness of speaking to MDS.
1259 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1263 Description: sanity test 65a fails if stripecount of -1 is set
1264 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1269 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1270 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1273 Severity : enhancement
1275 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1279 Description: do not process already freed flock
1280 Details : flock can possibly be freed by another thread before it reaches
1281 to ldlm_flock_completion_ast.
1285 Description: LBUG during stress test
1286 Details : Need properly lock accesses the flock deadlock detection list.
1290 Description: oops in page fault handler
1291 Details : kernel page fault handler can return two special 'pages' in error
1292 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1296 Description: timeout with invalidate import.
1297 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1298 should be handled by ptlrpcd. This produce long age waiting and
1299 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1301 Severity : enhancement
1303 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1307 Description: enable MGS and MDT services start separately
1308 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1309 a co-located MGS without starting the MGS, which is a complement
1310 to 'nosvc' mount option.
1314 Frequency : after recovery
1315 Description: precreate to many object's after del orphan.
1316 Details : del orphan st in oscc last_id == next_id and this triger growing
1317 count of precreated objects. Set flag LOW to skip increase count
1318 of precreated objects.
1322 Frequency : rare, on clear nid stats
1323 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1324 Details : when clean nid stats sometimes try destroy live entry,
1325 and this produce panic in free.
1329 Description: Stack overflow during MDS log replay
1330 ease stack pressure by using a thread dealing llog_process.
1334 Description: wait until IO finished before start new when do lock cancel.
1335 Details : VM protocol want old IO finished before start new, in this case
1336 need wait until PG_writeback is cleared until check dirty flag and
1337 call writepages in lock cancel callback.
1339 Severity : enhancement
1341 Description: using special macro for print time and cleanup in includes.
1345 Description: mds_mfd_close() ASSERTION(rc == 0)
1346 Details : In mds_mfd_close(), we need protect inode's writecount change
1347 within its orphan write semaphore to prevent possible races.
1351 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1355 Frequency : rare, on shutdown ost
1356 Description: don't hit live lock with umount ost.
1357 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1358 use shrink_dcache_sb instead.
1362 Description: don't panic with use echo client
1363 Details : echo client pass NULL as client nid pointer and this produce null
1364 pointer dereference.
1368 Description: fix build on ppc32
1369 Details : compile code with -m64 flag produce wrong object file for ppc32.
1373 Description: add message levels for liblustreapi
1377 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1378 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1379 will result in an -EAGAIN and in the exit call path we call
1380 alloc_rr() which will with increasing aggressiveness attempt to
1381 aquire precreated objects on the minimum number of required OSCs.
1385 Description: Use old size assignment to avoid deadlock
1386 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1387 the scheduling while holding a spinlock. We do not need locking
1388 for size in ll_update_inode() because size is only updated from
1389 the MDS for directories or files without objects, so there is no
1390 other place to do the update, and concurrent access to such inodes
1391 are protected by the inode lock.
1395 Description: resolve "_IOWR redefined" build error on SLES10
1399 Description: dump the memory debugging after all modules are unloaded to
1400 suppress false negative in conf_sanity test 39
1402 Severity : enhancement
1404 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1409 Description: fix SLES kernel versioning
1410 Details : the kernel version for our SLES 10 kernel did not include a "-"
1411 before the "smp" at the end. while this was not a problem in
1412 general, it did mean that software trying to use the kernel
1413 version to try to detect a vendor specific kernel would fail.
1414 this was most evident by the OFED build scripts.
1418 Description: Don't update lov_desc members until making sure they are valid
1419 Details : When updating lov_desc members via proc fs, need fix their
1420 validities before doing the real update.
1424 Description: don't put request into delay list while invalidate in flight.
1425 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1426 import in flight. this produce timeout for invalidate and sometimes
1427 can cause stale data.
1431 Frequency : on ppc only
1432 Description: not convert ost objects for directory because it's not exist.
1433 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1437 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1439 Severity : enhancement
1441 Description: Update to support 2.6.22.14 vanilla kernel.
1445 Frequency : rare, on recovery
1446 Description: read procfs can produce deadlock in some situation
1447 Details : Holding lprocfs lock which send rpc can produce block for destroy
1448 obd objects and this also block reconnect with -EALREADY. This
1449 isn't fix all lprocfs bugs - but make it rare.
1451 Severity : enhancement
1453 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1456 Frequency : frequent on X2 node
1458 Description: mdc_set_open_replay_data LBUG
1459 Details : Set replay data for requests that are eligible for replay.
1463 Description: lustre_mgs: operation 101 on unconnected MGS
1464 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1465 MGC, and late on MGC cannot successfully connect to MGS and a lot
1466 of the error messages complaining that MGS is not connected.
1470 Frequency : on network error
1471 Description: panic with double free request if network error
1472 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1473 but it's true only for synchronus enqueue, for async enqueue
1474 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1477 Severity : enhancement
1479 Description: client-side metadata stat-ahead during readdir(directory readahead)
1480 Details : perform client-side metadata stat-ahead when the client detects
1481 readdir and sequential stat of dir entries therein
1484 Frequency : on start mds
1486 Description: Implement get_info(last_id) in obdfilter.
1489 Frequency : occasional
1491 Description: Correctly check stale fid, not start epoch if ost not support SOM
1492 Details : open with flag O_CREATE need set old fid in op_fid3 because
1493 op_fid2 was overwritten with new generated fid, but mds can answer
1494 with one of these two fids and both is not stale. Setattr
1495 incorrectly started an epoch and assume will be called
1496 done_writing, but without SOM done_writing ever being called.
1499 Frequency : rare, depends on device drivers and load
1501 Description: MDS or OSS nodes crash due to stack overflow
1502 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1503 In some cases, in conjunction with device drivers that use a lot
1504 of stack the MDS (or possibly OSS) service threads could overflow
1505 the stack. One change which was identified to consume additional
1506 stack has been reworked to avoid the extra stack usage.
1509 Frequency : occasional
1511 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1512 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1513 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1514 oscc_has_objects and see OSC has no objests and call
1515 oscc_internal_create to resend create request.
1517 Severity : enhancement
1519 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1521 Severity : enhancement
1523 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1526 Frequency : very rare
1528 Description: Fix replay if there is an un-replied request and open
1529 Details : In some cases, older replay request will revert the
1530 mcd->mcd_last_xid on MDS which is used to record the client's
1531 latest sent request.
1533 Severity : enhancement
1535 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1537 Severity : enhancement
1539 Description: Add rhel5 support to HEAD.
1541 Serverity : enhancement
1543 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1548 Description: Don't allow skipping OSTs if index has been specified.
1549 Details : Don't allow skipping OSTs if index has been specified, make
1550 locking in internal create lots better.
1554 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1555 Details : access to bitfield in structure is always rounded to long
1556 and this produce problem with not atomic change any bit.
1560 Description: Lustre make rpms failed.
1561 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1562 builds Lustre rpms from tarball.
1565 Frequency : rare on shutdown ost
1567 Description: If llog cancel was not send before clean_exports phase, this can
1568 produce deadlock in llog code.
1569 Details : If llog thread has last reference to obd and call class_import_put
1570 this produce deadlock because llog_cleanup_commit_master wait when
1571 last llog_commit_thread exited, but this never success because was
1572 called from llog_commit_thread.
1576 Description: allow userland application know is lost one of stripes.
1577 Details : fill lvb_blocks with error code on ost and return it to
1578 application if error flag found.
1582 Description: NULL lov_tgts causing MDS oops
1583 Details : more safe checks for NULL lov_tgts for avoid oops.
1585 Severity : enhancement
1587 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1591 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1592 Details : If an OST is down the MDS will hang indefinitely in
1593 obd_statfs_rqset() waiting for the statfs data. While for
1594 MDS QOS usage of statfs, it should not stuck in waiting.
1596 Severity : enhancement
1598 Description: remote_acl support
1599 Details : Support ACL-based permission check for remote user.
1600 Support setfacl/getfacl for remote user with the utils
1601 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1602 the system "{s,g}etfacl" utils.
1604 Severity : enhancement
1606 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1608 Severity : enhancement
1610 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1614 Description: make mgs_setparam() handle fsname containing dash
1615 Details : fsname containing a dash does not work with lctl conf_param
1617 Severity : enhancement
1619 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1621 Severity : enhancement
1623 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1627 Frequency : rare, at shutdown
1628 Description: access already free / zero obd_namespace.
1629 Details : if client_disconnect_export was called without force flag set,
1630 and exist connect request in flight, this can produce access to
1631 NULL pointer (or already free pointer) when connect_interpret
1632 store ocd flags in obd_namespace.
1636 Frequency : only at startup
1637 Description: not alloc memory with spinlock held.
1638 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1639 if any spinlock held.
1641 Severity : enhancement
1643 Description: make lustre randomly fail allocating memory
1644 Details : Make lustre randomly failed allocating memory for testing purpose.
1646 Severity : enhancement
1648 Description: lost problems with lov objid file
1649 Details : Fixes some scability and access to not inited memory problems
1650 in work with lov objdid file.
1655 Description: lfs find does not continue on file error
1656 Details : Continue other files processing when a file/dir is absent.
1660 Description: Inconsistent usage of lustre_pack_reply()
1661 Details : Standardize the usage of lustre_pack_reply() such that it
1662 always generate a CERROR on failure.
1665 Frequency : occasional
1667 Description: MDS hang or stay in waiting lock
1668 Details : If client receive lock with CBPENDING flag ldlm need send lock
1669 cancel as separate rpc, to avoid situation when cancel request
1670 can't processed due all i/o threads stay in wait lock.
1674 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1678 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1682 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1686 Description: Update to RHEL4 latest kernel.
1688 Severity : enhancement
1690 Description: Build SLES10 patchless client fails
1691 Details : The configure was broken by run ./configure with
1692 --with-linux-obj=.... argument for patchless client. When the
1693 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1694 can't search header adequately. Use absolute path such as
1695 -I($LINUX)/include instead.
1699 Description: interrupt oig_wait produce painc on resend.
1700 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1701 private set, and this produce situation when rq_ptlrpcd_data not
1702 copyed to new allocated request and triggered LBUG on assert
1703 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1704 ptlrpcd set if request is changed and can be safety changed to use
1709 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1710 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1711 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1712 these macros and replace them with their difinitions instead.
1716 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1717 Details : Update which_patch & target file for SLES10 latest kernel.
1720 Description: Remove client patches from SLES 10 kernel.
1721 Details : This causes SLES 10 clients to behave as patchless clients
1722 even on a Lustre-patched (server) kernel.
1724 Severity : enhancement
1726 Description: self-adjustable client's lru lists
1727 Details : use adaptive algorithm for managing client cached locks lru
1728 lists according to current server load, other client's work
1729 pattern, memory activities, etc. Both, server and client
1730 side namespaces provide number of proc tunables for controlling
1733 Severity : enhancement
1735 Description: light-weight GSS support
1736 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1737 and reduce performance overhead.
1739 Severity : enhancement
1741 Description: Linux keyring support
1742 Details : Support using service of Linux keyring for Lustre GSS internal
1743 context refresh/cache mechanism.
1747 Description: Fix errors in lfs documentation
1748 Details : Fixes man pages
1752 Description: don't use GFP_* in generic Lustre code.
1753 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1757 Description: obdclass is limited by single OBD_ALLOC(idarray)
1758 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1762 Description: warnings with build patchless client with vanila 2.6.19 and up
1763 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1764 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1768 Frequency : Only for SLES
1769 Description: sanity_quota fail test_1
1770 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1771 file, which makes TSTID[2] inunique.
1775 Frequency : Only for SLES9
1776 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1777 Details : Update target/ChangeLog/which_patch .
1781 Frequency : Only for test_10 in sanity-quota.sh
1782 Description: a bug in quota test code
1783 Details : checking a test flag in a wrong place
1787 Frequency : Only for test_13 in sanity-quota.sh
1788 Description: a bug in quota test script
1789 Details : shouldn't check overfull allocated quota
1793 Description: jbd statistics
1794 Details : Port older jbd statistics patch for sles10
1796 Severity : enhancement
1798 Description: Add printing inode info into message about error in writepage.
1803 Description: reply_lock_interpret crash due to race with it and lock cancel.
1804 Details : Do not replay locks that are being cancelled. Do not reference
1805 locks by their address during replay, just by their handle.
1809 Frequency : When flocks are used.
1810 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1811 Details : Flock locks might destroy just granted lock if it could be merged
1812 with another existing flock, this is done in completion handler,
1813 so teach ldlm_cli_enquque_fini that this is a valid case for
1819 Description: Oops in read and write path when failing to allocate lock.
1820 Details : Check if lock allocation failed and return error back.
1824 Description: lstripe command fails for valid OST index
1825 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1826 instead of lov->desc.ld_active_tgt_count.
1830 Description: sanity.sh failed test 103
1831 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1836 Description: df doesn't work properly if diskfs blocksize != 4K
1837 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1841 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1842 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1844 Severity : enhancement
1846 Description: Adaptive timeouts
1847 Details : RPC timeouts adapt to changing server load and network
1848 conditions to reduce resend attempts and improve recovery time.
1852 Description: llapi_file_create() does not allow some changes
1853 Details : add llapi_file_open() that allows specifying the mode and
1854 open flags, and also returns an open file handle.
1858 Description: merge and cleanup kernel patches.
1859 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1863 Description: Add journal checksum support.(Kernel part)
1864 Details : The journal checksum feature adds two new flags i.e
1865 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1866 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1867 indicates that the commit block contains the checksum for
1868 the blocks described by the descriptor blocks. Now commit
1869 record can be sent to disk without waiting for descriptor
1870 blocks to be written to disk. This behavior is controlled
1871 using JBD2_FEATURE_ASYNC_COMMIT flag.
1875 Description: OSS needs mutliple precreate threads
1876 Details : Add ability to start more than one create thread per OSS.
1880 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1881 Details : execute lfs setstripe on client
1885 Description: mds_obd_create error creating tmp object
1886 Details : When the user sets quota on root, llog will be affected and can't
1887 create files and write files.
1890 Frequency : Always on ia64 patchless client, and possibly others.
1892 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1893 Details : This allows the patchless client to be loaded on architectures
1894 without this export.
1898 Description: RedHat Update kernel for RHEL5
1899 Details : Add the kernel config file for RHEL5.
1903 Description: RedHat Update kernel for RHEL5
1904 Details : Modify the kernel config file more closer RHEL5.
1908 Description: Build failure against Centos5 (RHEL5)
1909 Details : Define PAGE_SIZE when it isn't present.
1913 Description: test_7 on sanity-quota.sh can't be used on separate machines
1914 Details : add facet to handle it
1918 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1919 Details : with NFS, the anon dentry's parent was set to itself in
1920 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1921 corresponding dentry other than use rec->ur_name.
1923 Severity : enhancement
1925 Description: lfs setstripe enhancement
1926 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1930 Description: enable data checksumming by default
1931 Details : enable checksum by default, allow --disable-checksum
1932 configure option and "-o nochecksum" mount option. Checksums
1933 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1937 Description: lustre support for RHEL5
1938 Details : Add support for RHEL5.
1942 Description: Client eviction due to failover config
1943 Details : after a connection loss, the lustre client should attempt to
1944 reconnect to the last active server first before trying the
1945 other potential connections.
1949 Description: when mds and osts use different quota unit(32bit and 64bit),
1950 quota will be released repeatly.
1951 Details : void sending multiple quota reqs to mds, which will keep the
1952 status between the reqs.
1956 Description: osts not allocated evenly to files
1957 Details : change the condition to increase offset_idx
1960 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1962 Description: Data corruption for OSTs that are formatted larger than 2TB
1964 Details : When generating the bio request for lustre file writes the
1965 sector number would overflow a temporary variable before being
1966 used for the IO. The data reads correctly from Lustre (which
1967 will overflow in a similar manner) but other file data or
1968 filesystem metadata may be corrupted in some cases.
1972 Description: Tune the kernel for good SCSI performance.
1973 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1974 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1979 Description: rewrite ext2-derived code in obdclass/uuid.c
1980 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1981 from scratch preserving functionality.
1985 Description: "lfs find -obd UUID" prints directories
1986 Details : "lfs find -obd UUID" will return all directory names instead
1987 of just file names. It is incorrect because the directories
1988 do not reside on the OSTs.
1992 Description: use i_size_read and i_size_write in 2.6 port
1993 Details : replace inode->i_size access with i_size_read/write()
1996 Frequency : when removing large files
1998 Description: scheduling issue during removal of large Lustre files
1999 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2000 It causes scheduling issues when removing large files (17TB in the
2004 Frequency : only with liblustre clients on XT3
2006 Description: evictions taking too long
2007 Details : allow llrd to evict clients directly on OSTs
2010 Frequency : only on ppc
2012 Description: /proc/fs/lustre/devices broken on ppc
2013 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2014 We should make sure the type of 'index' is loff_t and then cast
2015 explicitly as needed below. Do not assign an explicitly cast
2019 Frequency : only for rhel5
2021 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2022 Details : Modify the target file & which_kernel.
2024 Severity : enhancement
2026 Description: omit set fsid for export NFS
2027 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2028 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2032 Frequency : Always, for kernels after 2.6.16
2033 Description: Fix warning idr_remove called for id=.. which is not allocated.
2034 Details : Last kernels save old s_dev before kill super and not allow
2035 to restore from callback - restore it before call kill_anon_super.
2039 Description: buffer overruns could theoretically occur
2040 Details : llapi_semantic_traverse() modifies the "path" argument by
2041 appending values to the end of the origin string, and a
2042 overrun may occur. Adding buffer overrun check in liblustreapi.
2046 Description: Fix error on 'ls .' at the top of the Lustre mount.
2047 Details : Don't revalidate dentry if it is a root dentry.
2051 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2052 Details : Modify vm-tunables-rhel4.patch.
2056 Description: Kernel config for 2.6.18-vanilla.
2057 Details : Modify targets/2.6-vanilla.target.in.
2058 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2059 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2060 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2061 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2065 Description: improve handling recoverable errors
2066 Details : if request processig with error which can be recoverable on server
2067 request should be resend, otherwise page released from cache and
2072 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2073 Details : Modify target file & which_patch.
2074 A flaw was found in the IA32 system call emulation provided
2075 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2076 value could be stored in the %RAX register, which could trigger an
2077 out-of-bounds system call table access. An untrusted local user
2078 could exploit this flaw to run code in the kernel
2079 (ie a root privilege escalation). (CVE-2007-4573).
2083 Description: O_DIRECT bypasses client statistics.
2084 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2085 read_ahead_stats were updated. copied the stats section from
2086 osc_send_oap_rpc() into async_internal().
2090 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2094 Description: change order of libsysio includes
2095 Details : '#include sysio.h' should always come before '#include xtio.h'
2099 Description: handle "serious error: objid * already exists" more gracefully
2100 Details : If LAST_ID value on disk is smaller than the objects existing in
2101 the O/0/d* directories, it indicates disk corruption and causes an
2102 LBUG(). If the object is 0-length, then we should use the existing
2103 object. This will help to avoid a full fsck in most cases.
2105 Severity : enhancement
2107 Description: adapt the lustre_config script to support the upgrade case
2108 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2109 server targets to 1.6 in parallel.
2113 Description: To avoid grant space > avaible space when the disk is almost
2114 full. Without this patch you might see the error "grant XXXX >
2115 available" or some LBUG about grant, when the disk is almost
2117 Details : In filter_check_grant, for non_grant cache write, we should
2118 check the left space by if (*left > ungranted + bytes), instead
2119 of (*left > ungranted), because only we are sure the left space
2120 is enough for another "bytes", then the ungrant space should be
2121 increase. In client, we should update cl_avail_grant only there
2122 is OBD_MD_FLGRANT in the reply.
2126 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2127 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2128 described in CVE-2007-4573. This problem could allow untrusted
2129 local users to gain root access.
2132 Frequency : when using O_DIRECT and quotas
2134 Description: Incorrect file ownership on O_DIRECT output files
2135 Details : block usage reported by 'lfs quota' does not take into account
2136 files that have been written with O_DIRECT.
2141 Description: touch file failed when fs is not full
2142 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2143 otherwise we can get ENOSP while fs is not full.
2147 Description: parallel lock callbacks
2148 Details : Instead of sending blocking and completion callbacks as separated
2149 requests, adding them to a set and sending in parallel.
2152 Frequency : only for Cray XT3
2153 Bugzilla : 12829/13455
2154 Description: Changing primary group doesn't change the group lustre assigns to
2156 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2157 overridden with the primary group provided by the group upcall,
2158 whereas the supplied fsgid can be trusted if it is in the list of
2159 supplementary groups returned by the group upcall.
2161 Severity : enhancement
2163 Description: Allow masking D_WARNING, D_ERROR messages from console
2164 Details : Console messages can now be disabled via lnet.printk.
2168 Description: User code with malformed file open parameter crashes client node
2169 Details : Before packing join_file req, all the related reference should be
2170 checked carefully in case some malformed flags cause fake
2171 join_file req on client.
2175 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2176 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2177 mode is equal to the granted mode. However, it should also check
2178 that LDLM_FL_CP_REQD is not set.
2182 Description: Heavy nfs access might result in deadlocks
2183 Details : After ELC code landed, it is now improper to enqueue any mds
2184 locks under och_sem, because enqueue might want to decide to
2185 cancel open locks for same inode we are holding och_sem for.
2189 Description: Client eviction while running blogbench
2190 Details : A lot of unlink operations with concurrent I/O can lead to a
2191 deadlock causing evictions. To address the problem, the number of
2192 oustanding OST_DESTROY requests is now throttled to
2193 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2194 callbacks are processed in priority.
2198 Description: enable ACLs on MDS by default
2199 Details : ACLs must be enabled on MDS by default.
2202 Frequency : PPC/PPC64 only
2204 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2205 Details : fix duplicated definitions between asm-ppc64/types.h and
2206 lustre_types.h on PPC.
2209 Frequency : PPC/PPC64 only
2211 Description: asm-ppc/segment.h does not exist
2212 Details : fix compile issue on PPC.
2216 Description: better handle error messages in extents code
2219 Frequency : RHEL4 only
2221 Description: mkfs is very slow on IA64/RHEL4
2222 Details : A performance regression has been discovered in the MPT Fusion
2223 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2224 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2225 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2227 Severity : enhancement
2229 Description: SNMP support enhancement
2230 Details : Adding total number of sampled request for an MDS node in snmp
2233 Severity : enhancement
2235 Description: Optimize ldlm waiting list processing for PR extent locks
2236 Details : When processing waiting list for read extent lock and meeting read
2237 lock that is same or wider to it that is not contended, skip
2238 processing rest of the list and immediatelly return current
2239 status of conflictness, since we are guaranteed there are no
2240 conflicting locks in the rest of the list.
2244 Description: Time out and refuse to reconnect
2245 Details : When the failover node is the primary node, it is possible
2246 to have two identical connections in imp_conn_list. We must
2247 compare not conn's pointers but NIDs, otherwise we can defeat
2248 connection throttling.
2252 Description: port llog fixes from b1_6 into HEAD
2253 Details : Port llog reference couting and some llog cleanups from b1_6
2254 (bug 10800) into HEAD, for protect from panic and access to already
2255 free llog structures.
2259 Description: Detect stride IO mode in read-ahead
2260 Details : When a client does stride read, read-ahead should detect that and
2261 read-ahead pages according to the detected stride pattern.
2265 Description: data checksumming impacts single node performance
2266 Details : add support for several checksum algorithm. Currently, only CRC32
2267 and Adler-32 are supported. The checksum type can be changed on
2268 the fly via /proc/fs/lustre/osc/*/checksum_type.
2272 Description: use adler32 for page checksums
2273 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2278 Description: build for x2 fails
2279 Details : fix compile issue on Cray systems.
2283 Description: Properly match for duplicate locks
2284 Details : Due to different lock order from skiplists code, we need to
2285 traverse entire list for now
2288 Frequency : only on PPC/SLES10
2290 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2291 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2292 be defined. Add a hack in mkfs_lustre.c to work around this compile
2297 Description: LASSERT on MDS when client holding flock lock dies
2298 Details : ldlm pool logic depends on number of granted locks equal to
2299 number of released locks which is not true for flock locks, so
2300 just exclude such locks from consideration.
2304 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2305 Details : Use fsfilt_sync() for both the whole filesystem sync and
2306 individual file sync to eliminate dangerous inode locking
2307 with I_LOCK that can lead to a deadlock.
2311 Description: performance in 1.6.3
2312 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2313 sure that 1MB requests are not fragmented by the block layer.
2315 Severity : enhancement
2317 Description: organize the server-side client stats on per-nid basis
2318 Details : Change the structure of stats under obdfilter and mds to
2326 The "uuid"s file would list the uuids of _active_ exports.
2327 And the clear entry is to clear all stats and stale nids.
2329 Severity : enhancement
2331 Description: eliminate client locks in face of contention
2332 Details : file contention detection and lockless i/o implementation
2333 for contended files.
2337 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2340 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2342 Description: MDS may lose file striping (and hence file data) in some cases
2343 Details : If there are additional extended attributes stored on the MDS,
2344 in particular ACLs, SELinux, or user attributes (if user_xattr
2345 is specified for the client mount options) then there is a risk
2346 of attribute loss. Additionally, the Lustre file striping
2347 needs to be larger than default (e.g. striped over all OSTs),
2348 and an additional attribute must be stored initially in the
2349 inode and then increase in size enough to be moved to the
2350 external attribute block (e.g. ACL growing in size) for file
2355 Description: skiplist implementation simplification
2356 Details : skiplists are used to group compatible locks on granted list
2357 that was implemented as tracking first and last lock of each
2358 lock group the patch changes that to using doubly linked lists
2362 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2363 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2364 because LASSERT on both the data supplied by a client, and the
2365 data on disk is dangerous and incorrect.
2367 Severity : enhancement
2369 Description: Slow truncate/writes to huge files at high offsets.
2370 Details : Directly associate cached pages to lock that protect those pages,
2371 this allows us to quickly find what pages to write and remove
2372 once lock callback is received.
2376 Description: more ldlm soft lockups
2377 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2378 starve other threads from the resource lock for a long time in
2379 case of long waiting queue, so change the debug level from
2380 D_OTHER to the less frequently used D_INFO.
2382 Severity : enhancement
2384 Description: add -gid, -group, -uid, -user options to lfs find
2388 Description: Hung threads in invalidate_inode_pages2_range
2389 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2390 once one is completed. As a result, some RPCs are stuck in the
2391 queue and are never sent.
2395 Description: filter threads hungs on waiting journal commit
2396 Details : Cleanup filter group llog code, then only filter group llog will
2397 be only created in the MDS/OST syncing process.
2401 Description: Procfs and llog threads access destoryed import sometimes.
2402 Details : Sync the import destoryed process with procfs and llog threads by
2403 the import refcount and semaphore.
2405 Severity : enhancement
2407 Description: openlock cache of b1_6 port to HEAD
2412 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2413 Details : If insertion of an extent fails, then discard the inode
2414 preallocation and free data blocks else it can lead to duplicate
2419 Description: don't always update ctime in ext3_xattr_set_handle()
2420 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2421 In some cases the ctime should not be updated, for example for
2422 2.0->1.8 compatibility it is necessary to delete an xattr and it
2423 should not update the ctime.
2427 Bugzilla : 15713/16362
2428 Description: Assertion in iopen_connect_dentry in 1.6.3
2429 Details : looking up an inode via iopen with the wrong generation number can
2430 populate the dcache with a disconneced dentry while the inode
2431 number is in the process of being reallocated. This causes an
2432 assertion failure in iopen since the inode's dentry list contains
2433 both a connected and disconnected dentry.
2437 Description: assertion failure in ldlm_handle2lock()
2438 Details : fix a race between class_handle_unhash() and class_handle2object()
2439 introduced in lustre 1.6.5 by bug 13622.
2444 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2445 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2450 Description: oops in sd_iostats_seq_show()
2451 Details : unloading/reloading the scsi low level driver triggers a kernel
2452 bug when trying to access the sd iostat file.
2457 Description: Kernel panics during QLogic driver reload
2458 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2459 patch, causing memory corruption.
2464 Description: journal_dev option does not work in b1_6
2465 Details : pass mount option during pre-mount.
2467 Severity : enhancement
2469 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2470 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2471 extent information of a file. It can be used to map logical blocks
2472 in a file to physical blocks in the block device.
2476 Description: LDLM soft lockups - improvement
2477 Details : It is be possible to send the lock handle along with each read
2478 or write request because the client is already doing a lock match
2479 itself so there isn't any reason the OST should have to re-do that
2483 Frequency : only on Cray X2
2485 Description: X2 build failures
2486 Details : fix build failures on Cray X2.
2490 Description: xid & resent requests
2491 Details : Initialize RPC XID from clock at startup (randomly if clock is
2494 Severity : enhancement
2496 Description: Add lustre_start utility to start or stop multiple Lustre servers
2501 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2502 Details : In case of memory pressure, list_del() can be called twice on
2503 req->rq_history_list, causing a kernel oops.
2507 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2508 Details : fix stack overflow in the distributed lock manager by defering
2509 export eviction after a failed AST to the elt thread instead of
2510 handling it in the dlm interpret routine.
2514 Description: Convert some comments to new format.
2515 Details : Update documenting comments to match doxygen conventions.
2519 Description: Grammar fixes.
2520 Details : A couple of trivial spelling fixes.
2524 Description: OSD_COUNTERS-mandatory
2525 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2526 to simplify the code.
2530 Description: simplify cmm_device freeing logic.
2531 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2535 Description: Add lockdep support to dt_object_operations locking interface.
2536 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2537 parameter indicating lock ordering. Update mdd code to use new
2542 Description: Introduce failloc constants for lockless IO tests.
2543 Details : Add two new failloc constants to test lockless IO. Only one of
2544 them in implemented---another is checked in yet to be landed
2549 Description: Add lockdep support for inode mutex.
2550 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2551 in the situations where Lustre has to lock more than one inode
2556 Description: Add optional invariants checking support.
2557 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2558 configure switch. This macro is to be used for consistency and
2559 sanity checks that are too expensive to be left in `production'
2564 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2565 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2570 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2571 Details : Use spin_lock_nested() in (the only) situation where more than
2572 one ldlm_lock is locked simultaneously. Also, fix possible
2573 dead-lock in ldlm_lock_change_resource() by enforcing particular
2578 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2579 Details : Instead of specifying each ldlm_lock call-back through separate
2580 parameter, wrap them into struct ldlm_callback_suite.
2584 Description: Kill join_lru obd method and its callers.
2585 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2586 in memory---a requirement implemented through ->o_join_lru() obd
2587 method in HEAD. Remove this method and its users.
2591 Description: Add asynchronous ldlm ENQUEUE completion handler.
2592 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2593 case that stock ldlm_completion_ast() cannot handle as it waits
2594 until lock is granted. Introduce new ldlm_completion_ast_async()
2595 for this. Also comment ldlm_completion_ast().
2599 Description: ldlm_error <-> errno conversion.
2600 Details : Add functions to map (rather arbitrary) between LDLM error codes
2601 and standard errno values. CLIO needs this to prevent LDLM specific
2602 constants from escaping ldlm and osc.
2606 Description: Kill unused ldlm_handle2lock_ns() function.
2607 Details : Kill unused ldlm_handle2lock_ns() function.
2611 Description: Add lu_ref support to ldlm_lock
2612 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2613 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2614 and ldlm_resource. LDLM interface has to be changed, because code that
2615 releases a reference on a lock, has to "know" what reference this is.
2616 In the most frequent case
2618 lock = ldlm_handle2lock(handle);
2620 LDLM_LOCK_PUT(lock);
2622 no changes are required. When any other reference (received _not_ from
2623 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2624 instead of LDLM_LOCK_PUT().
2626 Arguably, changes are pervasive, and interface requires some discipline
2627 for proper use. On the other hand, it was very instrumental in finding
2628 a few leaked lock references.
2632 Description: Add ldlm_lock_addref_try().
2633 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2634 attempts to addref a lock that might be being canceled
2639 Description: Add ldlm_weigh_callback().
2640 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2641 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2642 hard-coded `number of pages' logic.
2646 Description: Add lockdep annotations to llog code.
2647 Details : Use appropriately tagged _nested() locking calls in the places
2648 where llog takes more than one ->lgh_lock lock.
2652 Description: Add loi_kms_set().
2653 Details : Wrap kms updates into a helper function.
2657 Description: Constify instances of struct lsm_operations.
2658 Details : Constify instances of struct lsm_operations.
2662 Description: lu_conf support.
2663 Details : On a server, a file system object is uniquely identified
2664 by a fid, which is sufficient to locate and load all object
2665 state (inode). On a client, on the other hand, more data are
2666 necessary instantiate an object. Change lu_object_find() and
2667 friends to take additional `lu_conf' argument describing object.
2668 Typically this includes layout information.
2672 Description: lu_context fixes.
2673 Details : Introduce new lu_context functions that are needed on the client
2674 side, where some system threads (ptlrpcd) are shared by multiple
2675 modules, and so cannot be stopped during module shutdown.
2679 Description: Add start and stop methods to lu_device_type_operations.
2680 Details : Introduce two new methods in lu_device_type_operations, that are
2681 invoked when first instance of a given type is created and last one
2682 is destroyed respectively. This is need by CLIO.
2686 Description: Add lu_ref support to struct lu_device.
2687 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2688 track leaked references.
2692 Description: Introduce lu_kmem_descr.
2693 Details : lu_kmem_descr and its companion interface allow to create
2694 and destroy a number of kmem caches at once.
2698 Description: Fix lu_object finalization race.
2699 Details : Fix a race between lu_object_find() finding an object and its
2700 concurrent finalization. This race is (most likely) not possible
2701 on the server, but might happen on the client.
2705 Description: Introduce lu_ref interface.
2706 Details : lu_ref is a debugging module allowing to track references to
2707 a given object. It is quite cpu expensive, and has to be
2708 explicitly enabled with --enable-lu_ref. See usage description
2713 Description: Factor lu_site procfs stats into a separate function.
2714 Details : Separate lu_site stats printing code into a separate function
2715 that can be reused on a client.
2719 Description: Constify instances of struct {lu,dt,md}_device_operations.
2720 Details : Constify instances of struct {lu,dt,md}_device_operations.
2724 Description: Introduce struct md_site and move meta-data specific parts of
2725 struct lu_site here.
2726 Details : Move md-specific fields out of struct lu_site into special struct
2727 md_site, so that lu_site can be used on a client.
2731 Description: Kill mdd_lov_destroy().
2732 Details : Remove unused mdd code.
2736 Description: Add st_block checking to multistat.c.
2737 Details : Add st_block checking to multistat.c.
2741 Description: Add lu_ref support to struct obd_device.
2742 Details : Add lu_ref tracking to obd_device.
2746 Description: Kill obd_set_fail_loc().
2747 Details : Remove unused code.
2751 Description: Add special type for ptlrpc_request interpret functions.
2752 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2753 there. Actual usage will be in CLIO.
2757 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2758 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2759 doesn't work with lockdep.
2763 Description: Add rwv.c test program.
2764 Details : New testing program exercising readv(2) and writev(2) (Qian).
2768 Description: Add sendfile.c test program.
2769 Details : New testing program exercising sendfile(2) (Jay).
2773 Description: Ratelimit a message that can be very frequent.
2774 Details : Ratelimit a memory allocation failure message that can
2779 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2780 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2784 Description: Liblustre build fixes.
2785 Details : Liblustre build fixes.
2789 Description: libcfs: add cfs_{need,cond}_resched() interface.
2790 Details : libcfs: add cfs_{need,cond}_resched() definition and
2791 implementations for Linux, NT, and liblustre.
2793 Severity : enhancement
2795 Description: More exported tunables for mballoc
2796 Details : Add support for tunable preallocation window and new tunables for
2797 large/small requests
2801 Description: Detect corruption of block bitmap and checking for preallocations
2802 Details : Checks validity of on-disk block bitmap. Also it does better
2803 checking of number of applied preallocations. When corruption is
2804 found, it turns filesystem readonly to prevent further corruptions.
2808 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2809 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2810 window to 0. If it does hit cache miss, and read-pattern is still
2811 stride-io mode, does not reset the stride window, but also does not
2812 increase the stride window length in this case.
2816 Frequency : only for big-endian servers
2817 Description: Check if system is big-endian while mounting fs with extents feature
2818 Details : Mounting a filesystem with extents feature will fail on big-endian
2819 systems since ext3-based ldiskfs is not supported on big-endian
2820 systems. This can be over-riden with "bigendian_extents" mount option.
2822 Severity : enhancement
2824 Description: The root squash functionality
2825 Details : A security feature, which is to prevent users from being able
2826 to mount lustre on their desktop, run as root, and delete
2827 all of the files in the filesystem. The goal is accomplished by
2828 remapping user id (UID) and group id (GID) of the root user to
2829 a UID and GID specified by the system administartor via Lustre
2830 configuration management server (MGS). The functionality also
2831 allows to specify sets of clients for which the remapping does
2836 Description: Excessive recovery window
2837 Details : With AT enabled, the recovery window can be excessively long (6000+
2838 seconds). To address this problem, we no longer use
2839 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2840 timeout no longer depends on the service time, it is set to
2841 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2842 time via pb_service_time.
2846 Description: Watchdog triggered on MDS failover
2847 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2848 the OSTs know that the MDS "UUID" can be reused for the same export
2849 from a different NID, so we do not need to wait for the export to be
2853 Frequency : rare, only if using MMP with Linux RAID
2855 Description: MMP doesn't work with Linux RAID
2856 Details : While using HA for Lustre servers with Linux RAID, it is possible
2857 that MMP will not detect multiple mounts. To make this work we
2858 need to unplug the device queue in RAID when the MMP block is being
2859 written. Also while reading the MMP block, we should read it from
2860 disk and not the cached one.
2862 Severity : enhancement
2864 Description: open file using fid
2865 Details : A file can be opened using just its fid, like
2866 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2869 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2872 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2873 Details : add proper locking for creation of jbd_4k slab cache
2877 Description: MMP check in ext3_remount() fails without displaying any error
2878 Details : When multiple mount protection fails during remount, proper error
2881 Severity : enhancement
2883 Description: Allow stripe size to be up to 4G-64k
2884 Details : Fix math logic to allow large stripe sizes.
2888 Description: add check for >8TB ldiskfs filesystems
2889 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2890 Don't allow >8TB ldiskfs filesystems to be mounted without
2891 force_over_8tb mount option
2896 Descriptoin: OSS complains "too many blocks" and crashes
2897 Details : In filter_commitrw_write():
2898 1. lquota_chkquota();
2899 2. obdo_from_inode();
2900 3. lquota_pending_commit();
2901 "2." has changed uid and gid in this case, then caused
2902 lquota_chkquota() and lquota_pending_commit() to use
2903 different uid and gid. That is the root of the bug.
2907 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2908 Details : Detect and print out the 'stripe count = -1' case properly.
2912 Description: Out or order replies might be lost on replay
2913 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2914 request with tid smaller than one currently being added, add it
2915 to the start, not end of the list.
2917 --------------------------------------------------------------------------------
2919 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2921 * Support for kernels:
2922 2.6.5-7.283 (SLES 9),
2923 2.6.9-55.EL (RHEL 4),
2924 2.6.16.46-0.14 (SLES 10),
2925 2.6.18.8 vanilla (kernel.org)
2926 * Client support for unpatched kernels:
2927 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2928 2.6.16 - 2.6.22 vanilla (kernel.org)
2929 * Due to recently discovered recovery problems, we do not recommend
2930 using patchless RHEL 4 clients with this or any earlier release.
2931 * Recommended e2fsprogs version: 1.39.cfs8
2932 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2933 * Starting with this release, the ldiskfs backing filesystem required
2934 by Lustre is now in its own package, lustre-ldiskfs. This package
2935 should be installed. It is versioned separately from Lustre and
2936 may be released separately in future.
2940 Description: block reactivating mgc import until all deactivates complete
2941 Details : Fix race when failing back MDT/MGS to itself (testing)
2944 Frequency : at statup only
2946 Description: mds_lov_synchronize race leads to various problems
2947 Details : simultaneous MDT->OST connections at startup can cause the
2948 sync to abort, leaving the OSC in a bad state.
2950 Severity : enhancement
2952 Description: add optional extra BUILD_VERSION info
2953 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2954 to override the lustre version.
2957 Frequency : 2.6.18 servers only
2959 Description: ll_kern_mount() doesn't release the module reference
2960 Details : The ldiskfs module reference count never drops down to 0
2961 because ll_kern_mount() doesn't release the module reference.
2966 Description: server LBUG when using old ost_num_threads parameter
2967 Details : Accept the old ost_num_threads parameter but warn that it
2968 is deprecated, and fix an off-by-one error that caused an LBUG.
2973 Description: Transient SCSI error results in persistent IO issue
2974 Details : iobuf->dr_error is not reinitialized to 0 between two
2978 Frequency : sometimes when underlying device returns I/O errors
2980 Description: OSTs not going read-only during write failures
2981 Details : OSTs are not remounted read-only when the journal commit threads
2982 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2983 instead of the ext3 wrappers.
2987 Description: poor connect scaling with increasing client count
2988 Details : Don't run filter_grant_sanity_check for more than 100 exports
2989 to improve scaling for large numbers of clients.
2992 Frequency : SLES10 only
2994 Description: sanity-quota.sh quotacheck failed: rc = -22
2995 Details : Quotas cannot be enabled on SLES10.
2998 Frequency : liblustre clients only
3000 Description: getdirentries does not give error when run on compute nodes
3001 Details : getdirentries does not fail when the size specified as an argument
3002 is too small to contain at least one entry
3004 Severity : enhancement
3006 Description: Add LNET router traceability for debug purposes
3007 Details : If a checksum failure occurs with a router as part of the
3008 IO path, the NID of the last router that forwarded the bulk data
3009 is printed so it can be identified.
3014 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3015 Details : Due to a race condition, liblustre clients were occasionally
3016 evicted incorrectly.
3018 Severity : enhancement
3020 Description: lfs setstripe use optional parameters instead of postional
3023 Severity : enhancement
3025 Description: Nanosecond timestamp support for ldiskfs
3026 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3027 resolution timestamps. There is not yet support for this at
3028 the Lustre filesystem level.
3031 Frequency : during server recovery
3033 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3034 Details : request with rq_no_resend flag not awake l_wait_event if they get
3038 Frequency : nfs export on patchless client
3040 Description: connectathon hang when test nfs export over patchless client
3041 Details : Disconnected dentry cannot be found with lookup, so we do not need
3042 to unhash it or make it invalid
3045 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3046 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3047 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3052 Description: open req refcounting wrong on reconnect
3053 Details : If reconnect happened between getting open reply from server and
3054 call to mdc_set_replay_data in ll_file_open, we will schedule
3055 replay for unreferenced request that we are about to free.
3056 Subsequent close will crash in variety of ways.
3057 Check that request is still eligible for replay in
3058 mdc_set_replay_data().
3063 Description: disable writes to filesystem when reading health_check file
3064 Details : the default for reading the health_check proc file has changed
3065 to NOT do a journal transaction and write to disk, because this
3066 can cause reads of the /proc file to hang and block HA state
3067 checking on a healthy but otherwise heavily loaded system. It
3068 is possible to return to the previous behaviour during configure
3069 with --enable-health-write.
3071 Severity : enhancement
3073 Description: 64-bit inode version
3074 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3075 to the inode. This will be required for version-based recovery.
3080 Description: MDS fails to start if a duplicate client export is detected
3081 Details : in some rare cases it was possible for a client to connect to
3082 an MDS multiple times. Upon recovery the MDS would detect this
3083 and fail during startup. Handle this more gracefully.
3085 Severity : enhancement
3087 Description: Add -o localflock option to simulate old noflock
3089 Details : This will achieve local-only flock/fcntl locks
3095 Description: log_commit_thread vs filter_destroy race leads to crash
3096 Details : Take import reference before releasing llog record semaphore
3101 Description: Wrong request locking in request set processing
3102 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3103 imp_delayed_list, in this place should be used imp_lock.
3106 Frequency : when reconnection
3108 Description: Grant Leak when osc reconnect to OST
3109 Details : When osc reconnect ost, OST(filter) should check whether it
3110 should grant more space to client by comparing fed_grant and
3111 cl_avail_grant, and return the granted space to client instead
3112 of "new granted" space, because client will call osc_init_grant
3113 to update the client grant space info.
3116 Frequency : when client reconnect to OST
3118 Description: Grant Leak when osc do resend and replay bulk write
3119 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3120 bulk write request, because the grant info while be sync between
3121 OSC and OST when reconnect, and we should ignore the grant info
3122 these of resend/replay write req.
3127 Description: Grant space more than available left space sometimes.
3128 Details : When then OST is about to be full, if two bulk writing from
3129 different clients came to OST. Accord the avaliable space of the
3130 OST, the first req should be permitted, and the second one
3131 should be denied by ENOSPC. But if the seconde arrived before
3132 the first one is committed. The OST might wrongly permit second
3133 writing, which will cause grant space > available space.
3136 Frequency : when client is evicted
3138 Description: Grant might be wrongly erased when osc is evicted by OST
3139 Details : when the import is evicted by server, it will fork another
3140 thread ptlrpc_invalidate_import_thread to invalidate the
3141 import, where the grant will be set to 0. While the original
3142 thread will update the grant it got when connecting. So if
3143 the former happened latter, the grant will be wrongly errased
3144 because of this race.
3149 Description: Checking Stale with correct fid
3150 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3151 to check whether it is stale, because sometimes, we want the
3152 enqueue happened anyway, and op_data.fid2 will not be initialized.
3154 Severity : enhancement
3156 Description: update patchless client
3157 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3160 Frequency : only with 2.4 kernel
3162 Description: random memory corruption
3163 Details : size of struct ll_inode_info is to big for union inode.u and this
3164 can be cause of random memory corruption.
3169 Description: Memory leak in recovery
3170 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3171 It should also check obd_fail before fsfilt_start, otherwise if
3172 fsfilt_start return -EROFS,(failover mds during mds recovery).
3173 then the req will return with repmsg->transno = 0 and rc = EROFS.
3174 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3175 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3176 be freed no matter whether fsfilt_commit success or not.
3179 Frequency : only with huge count clients
3181 Description: Prevents from taking the superblock lock in llap_from_page for
3183 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3184 for llap_from_page call in ll_removepage prevents from taking the
3185 superblock lock for a soon died page.
3190 Description: Not check open intent error before release open handle
3191 Details : in some rare cases, the open intent error is not checked before
3192 release open handle, which may cause
3193 ASSERTION(open_req->rq_transno != 0), because it tries to release
3194 the failed open handle.
3199 Description: Set cat log bitmap only after create log success.
3200 Details : in some rare cases, the cat log bitmap is set too early. and it
3201 should be set only after create log success.
3205 Description: Accessing a block bevice can re-enable I/O when Lustre is
3206 tearing down a device.
3207 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3212 Description: service threads may hog cpus when there are a lot of requests
3213 Details : Insert cond_resched to give other threads a chance to use some CPU
3218 Description: the cat log was not initialized in recovery
3219 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3220 unlink log on mds will not be initialized until mds post
3221 recovery. And also in mds post recovery, the unlink log will
3222 initialization will be done asynchronausly, so there will be race
3223 between add unlink log and unlink log initialization.
3227 Description: brw_stats were being printed incorrectly
3228 Details : brw_stats were being printed as log2 but all of them were not
3229 recorded as log2. Also remove some code duplication arising from
3230 filter_tally_{read,write}.
3234 Frequency : rare, only in recovery.
3235 Description: ASSERTION(req->rq_type != LI_POISON) failed
3236 Details : imp_lock should be held while iterating over imp_sending_list for
3237 prevent destroy request after get timeout in ptlrpc_queue_wait.
3241 Description: replay-single.sh test 52 fails
3242 Details : A lock's skiplist need to be cleanup when it being unlinked
3243 from its resource list.
3247 Description: Short directio read returns full requested size rather than
3249 Details : Direct I/O operations should return actual amount of bytes
3250 transferred rather than requested size.
3252 Severity : enhancement
3254 Description: metadata RPC reduction (e.g. for rm performance)
3255 Details : decrease the amount of synchronous RPC between clients and servers
3256 by canceling conflicing lock before the operation on the client
3257 and packing thier handles into the main operation RPC to server.
3259 Severity : enhancement
3261 Description: Async OSC create to avoid the blocking unnecessarily.
3262 Details : If an OST has no remaining object, system will block on the
3263 creation when it needs to create a new object on this OST. Now,
3264 ways use pre-created objects when available, instead of blocking on
3265 an empty osc while others are not empty. If we must block, we
3266 block for the shortest possible period of time.
3270 Description: improve handling recoverable errors
3271 Details : if request processig with error which can be recoverable on server
3272 request should be resend, otherwise page released from cache and
3275 Severity : enhancement
3277 Description: refine locking for avoid write wrong info into lov_objid file
3278 Details : fix possible races with add new target and write/update data in
3281 --------------------------------------------------------------------------------
3283 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3288 Frequency : on some architectures
3290 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3291 Details : Uninitialized flags sometimes cause configuration commands to
3295 Frequency : patchless clients only
3297 Description: missing __iget() symbol export
3298 Details : The __iget() symbol export is missing. To avoid the need for
3299 this on patchless clients the deathrow inode reaper is turned
3300 off, and we depend on the VM to clean up old inodes. This
3301 dependency was during via the fix for bug 12181.
3303 --------------------------------------------------------------------------------
3305 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3307 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3308 INTEROPERATE with older versions automatically. In many cases a
3309 special upgrade step is needed. Please read the
3310 user documentation before upgrading any part of a live system.
3311 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3312 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3313 * WARNING: Lustre configuration and startup changes are required with
3314 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3318 Severity : enhancement
3320 Description: MountConf
3321 Details : Lustre configuration is now managed via mkfs and mount
3322 commands instead of lmc and lconf. New obd types (MGS, MGC)
3323 are added for dynamic configuration management. See
3324 https://mail.clusterfs.com/wikis/lustre/MountConf for
3327 Severity : enhancement
3329 Description: dynamic OST addition
3330 Details : OSTs can now be added to a live filesystem
3332 Severity : enhancement
3334 Description: startup order invariance
3335 Details : MDTs and OSTs can be started in any order. Clients only
3336 require the MDT to complete startup.
3338 Severity : enhancement
3340 Description: parallel, asynchronous orphan cleanup
3341 Details : orphan cleanup is now performed in separate threads for each
3342 OST, allowing parallel non-blocking operation.
3344 Severity : enhancement
3346 Description: optimized stripe assignment
3347 Details : stripe assignments are now made based on ost space available,
3348 ost previous usage, and OSS previous usage, in order to try
3349 to optimize storage space and networking resources.
3351 Severity : enhancement
3353 Description: Permanently set tunables
3354 Details : All writable /proc/fs/lustre tunables can now be permanently
3355 set on a per-server basis, at mkfs time or on a live system.
3357 Severity : enhancement
3359 Description: Lustre message v2
3360 Details : Add lustre message format v2.
3362 Severity : enhancement
3364 Description: client OST exclusion list
3365 Details : Clients can be started with a list of OSTs that should be
3366 declared "inactive" for known non-responsive OSTs.
3370 Description: SPEC SFS validation failure on NFS v2 over lustre.
3371 Details : Changes the blocksize for regular files to be 2x RPC size,
3372 and not depend on stripe size.
3374 Severity : enhancement
3376 Description: Multiple MD RPCs in flight.
3377 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3378 To avoid overly-overloading MDS, introduce a limit on number of
3379 MDS RPCs in flight for a single client and add /proc controls
3380 to adjust this limit.
3382 Severity : enhancement
3384 Description: client read/write statistics
3385 Details : Add client read/write call usage stats for performance
3386 analysis of user processes.
3387 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3388 file access. extents_stats shows chunk size distribution.
3389 extents_stats_per_process show chunk size distribution per
3392 Severity : enhancement
3394 Description: mds statistics
3395 Details : Add detailed mds operations statistics in
3396 /proc/fs/lustre/mds/*/stats.
3400 Description: Failure of copying files with lustre special EAs.
3401 Details : Client side always return success for setxattr call for lustre
3402 special xattr (currently only "trusted.lov").
3407 Description: Refcount LNET uuids
3408 Details : The global LNET uuid list grew linearly with every startup;
3409 refcount repeated list entries instead of always adding to
3412 Severity : enhancement
3414 Description: Dynamic service threads
3415 Details : Within a small range, start extra service threads
3416 automatically when the request queue builds up.
3419 Frequency : mixed-endian client/server environments
3421 Description: mixed-endian crashes
3422 Details : The new msg_v2 system had some failures in mixed-endian
3425 Severity : enhancement
3427 Description: Easy OST removal
3428 Details : OSTs can be permanently deactivated with e.g. 'lctl
3429 conf_param lustre-OST0001.osc.active=0'
3431 Severity : enhancement
3433 Description: MGS proc entries
3434 Details : Added basic proc entries for the MGS showing what filesystems
3437 Severity : enhancement
3439 Description: provide MGS failover
3440 Details : Added config lock reacquisition after MGS server failover.
3442 Severity : enhancement
3444 Description: add Linux 2.4 support
3445 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3449 Description: a large application tries to do I/O to the same resource and dies
3450 in the middle of it.
3451 Details : Check the req->rq_arrival time after the call to
3452 ost_brw_lock_get(), but before we do anything about
3453 processing it & sending the BULK transfer request. This
3454 should help move old stale pending locks off the queue as
3455 quickly as obd_timeout.
3458 Frequency : when an incorrect nid is specified during startup
3460 Description: ptlrpc connect to non-existant node causes kernel crash
3461 Details : LNET can't be re-entered from an event callback, which
3462 happened when we expire a message after the export has been
3463 cleaned up. Instead, hand the zombie cleanup off to another
3466 Severity : enhancement
3468 Description: plain/inodebits lock performance improvement
3469 Details : Grouping plain/inodebits in granted list by their request modes
3470 and bits policy, thus improving the performance of search through
3474 Frequency : only if OST filesystem is corrupted
3476 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3477 Details : for a short time RPCs with bulk IO are in the replay list,
3478 but replay of bulk IOs is unimplemented. If the OST filesystem
3479 is corrupted due to disk cache incoherency and then replay is
3480 started it is possible to trip an assertion. Avoid putting
3481 committed RPCs into the replay list at all to avoid this issue.
3484 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3486 Description: System hang on startup
3487 Details : This bug allowed the liblustre (e.g. catamount) client to
3488 return to the app before handling all startup RPCs. This
3489 could leave the node unresponsive to lustre network traffic
3490 and manifested as a server ptllnd timeout.
3492 Severity : enhancement
3494 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3495 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3496 boolean to control whether to print peer debug info when a
3497 client's RPC times out.
3500 Frequency : only for kernels with patches from Lustre below 1.4.3
3502 Description: Remove old rdonly API
3503 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3506 Frequency : only for devices with external journals
3508 Description: Set external device read-only also
3509 Details : During a commanded failover stop, we set the disk device
3510 read-only while the server shuts down. We now also set any
3511 external journal device read-only at the same time.
3514 Frequency : when upgrading from 1.4 while trying to change parameters
3516 Description: The wrong (new) MDC name was used when setting parameters for
3517 upgraded MDT's. Also allows changing of OSC (and MDC)
3518 parameters if --writeconf is specified at tunefs upgrade time.
3521 Frequency : when setting specific ost indicies
3523 Description: QOS code breaks on skipped indicies
3524 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3525 created with --index need not be sequential.
3527 Severity : enhancement
3529 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3530 Details : The uninit_groups feature works in conjunction with the kernel
3531 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3532 up the pass1 processing of e2fsck. This is a read-only feature
3533 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3534 mount filesystems that have had this feature enabled.
3536 Severity : enhancement
3538 Description: Improve multi-block allocation algorithm to avoid fragmentation
3539 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3540 allocation locality and avoid filesystem fragmentation.
3542 ------------------------------------------------------------------------------
3544 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3546 * Support for kernels:
3547 2.4.21-47.0.1.EL (RHEL 3)
3548 2.6.5-7.283 (SLES 9)
3549 2.6.9-42.0.10.EL (RHEL 4)
3550 2.6.12.6 vanilla (kernel.org)
3551 2.6.16.27-0.9 (SLES 10)
3552 * Recommended e2fsprogs version: 1.39.cfs5
3554 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3558 Frequency : occasional, depends on client load and configuration
3559 Bugzilla : 12181, 12203
3560 Description: data loss for recently-modified files
3562 Details : In some cases it is possible that recently written or created
3563 files may not be written to disk in a timely manner (this should
3564 normally be within 30s unless client IO load is very high).
3565 The problem appears as zero-length files or files that are a
3566 multiple of 1MB in size after a client crash or client eviction
3567 that are missing data at the end of the file.
3569 This problem is more likely to be hit on clients where files are
3570 repeatedly created and unlinked in the same directory, clients
3571 have a large amount of RAM, have many CPUs, the filesystem has
3572 many OSTs, the clients are rebooted frequently, and/or the files
3573 are not accessed by other nodes after being written.
3575 The presence of the problem can be detected by looking at
3576 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3577 smaller than the second (nr_unused) then dirty files will not
3578 be flushed automatically to disk. "sync; sleep 10" should be
3579 run several times on the node before unmounting it to update
3580 Lustre (this is also safe to run on nodes without this problem).
3582 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3583 that can cause this same problem, so customers using that kernel
3584 also need to update the kernel in addition to Lustre. In order
3585 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3587 It is normal that files written just before a client crash (less
3588 than 30s) may not yet have been flushed to disk, even for local
3592 Frequency : frequent on thin XT3 nodes
3594 Description: UUID collision on thin XT3 Linux nodes
3595 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3596 randomly, since we relied on an insufficiently-seeded PRNG.
3601 Description: OSS hangs after "All ost request buffers busy"
3602 Details : A deadlock between quota and journal operations caused OSS
3603 hangs after printing "All ost request buffers busy."
3606 Frequency : always on liblustre builds
3608 Description: Cleanup compiler warnings on liblustre
3611 Frequency : always on liblustre builds on XT3
3613 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3617 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3618 > 32000 subdirectories
3619 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3620 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3621 subdirectory count crosses 32000. This will aid e2fsck to
3622 correctly handle more than 32000 subdirectories.
3625 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3627 Description: System hang on startup
3628 Details : This bug allowed the liblustre (e.g. catamount) client to
3629 return to the app before handling all startup RPCs. This
3630 could leave the node unresponsive to lustre network traffic
3631 and manifested as a server ptllnd timeout.
3633 Severity : enhancement
3635 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3636 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3637 boolean to control whether to print peer debug info when a
3638 client's RPC times out.
3643 Description: make O_SYNC working on 2.6 kernels
3644 Details : 2.6 kernels use different method for mark pages for write,
3645 so need add a code to lustre for O_SYNC work.
3650 Description: Failure to close file and release space on NFS
3651 Details : Put inode details into lock acquired in ll_intent_file_open.
3652 Use mdc_intent_lock in ll_intent_open to properly
3653 detect all kind of errors unhandled by mdc_enqueue.
3658 Description: proc file read during shutdown sometimes raced obd removal,
3660 Details : Add lock to prevent obd access after proc file removal.
3663 Frequency : Only for files larger than 4GB on 32-bit clients.
3665 Description: improperly doing page alignment of locks
3666 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3667 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3672 Description: RPCs being resent when they shouldn't be
3673 Details : Some RPCs that should not be resent are being resent. This
3674 can cause inconsistencies in the RPC state machine. Do not
3675 resend such requests.
3678 Frequency : rare, only with NFS export
3680 Description: Crash on NFS re-export node
3681 Details : under very unusual load conditions an assertion is hit in
3682 ll_intent_file_open()
3685 Frequency : only if OST filesystem is corrupted
3687 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3688 Details : for a short time RPCs with bulk IO are in the replay list,
3689 but replay of bulk IOs is unimplemented. If the OST filesystem
3690 is corrupted due to disk cache incoherency and then replay is
3691 started it is possible to trip an assertion. Avoid putting
3692 committed RPCs into the replay list at all to avoid this issue.
3697 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3698 Details : Large single O_DIRECT read and write calls can fail to allocate
3699 a sufficiently large buffer to process the request. In case of
3700 allocation failure the allocation is retried with a smaller
3701 buffer and broken into smaller requests.
3703 Severity : enhancement
3705 Description: Add -o localflock option to simulate old noflock behaviour.
3706 Details : This will achieve local-only flock/fcntl locks coherentness.
3711 Description: versioning check is incomplete
3712 Details : Checking the version difference of client vs. server, report
3713 error if the gap is too big.
3718 Description: add support PG_writeback bit
3719 Details : add support for PG_writeback bit for Lustre, for more carefull
3720 work with page cache in 2.6 kernel. This also fix some deadlocks
3721 and remove hack for work O_SYNC with 2.6 kernel.
3723 Severity : enhancement
3725 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3726 Details : The uninit_groups feature works in conjunction with the kernel
3727 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3728 up the pass1 processing of e2fsck. This is a read-only feature
3729 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3730 mount filesystems that have had this feature enabled.
3732 Severity : enhancement
3734 Description: Improve multi-block allocation algorithm to avoid fragmentation
3735 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3736 allocation locality and avoid filesystem fragmentation.
3738 ------------------------------------------------------------------------------
3740 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3742 * Support for kernels:
3743 2.6.9-42.0.3.EL (RHEL 4)
3744 2.6.5-7.276 (SLES 9)
3745 2.4.21-47.0.1.EL (RHEL 3)
3746 2.6.12.6 vanilla (kernel.org)
3747 2.6.16.21-0.8 (SLES10)
3748 * Recommended e2fsprogs version: 1.39.cfs2-0
3750 * The backwards-compatible /proc/sys/portals symlink has been removed
3751 in this release. Before upgrading, please ensure that you change
3752 any configuration scripts or /etc/sysctl.conf files that access
3753 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3754 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3755 in advance of the upgrade on any system running Lustre 1.4.6 or
3756 newer, since /proc/sys/lnet was added in that version.
3757 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3761 Frequency : only when quota is used
3763 Description: avoid scanning export list for quota master
3764 Details : Change the algorithms to avoid scanning export list in order
3765 to improve the efficiency.
3768 Frequency : MDS failover only, very rarely
3770 Description: "went back in time" messages on mds failover
3771 Details : The greatest transno may be lost when the current operation
3772 finishes with an error (transno==0) and the client's last_rcvd
3773 record is over-written. Save the greatest transno in the
3774 mds_last_transno for this case.
3777 Frequency : always for specific kernels and striping counts
3779 Description: client may get "Matching packet too big" without ACL support
3780 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3781 when trying to access files in certain configurations. The
3782 clients should in fact be denied when mounting because they do
3783 not understand ACLs.
3786 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3788 Description: many clients connecting with IO in progress causes connect timeouts
3789 Details : Avoid synchronous journal commits to avoid delays caused by many
3790 clients connecting/disconnecting when bulk IO is in progress.
3791 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3792 OST_IO_PORTAL to avoid delays behind potentially many pending
3796 Frequency : occasionally with multiple writers to a single file
3798 Description: shared writes to file may result in wrong size reported by stat()
3799 Details : Allow growing of kms when extent lock is cancelled
3802 Frequency : always with random mmap IO to multi-striped file
3804 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3805 Details : Only if the hole is at the end of OST object so that kms is too
3806 small. Fix is to increase kms accordingly in ll_nopage.
3809 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3811 Description: writes to a missing object would leak memory on the OST
3812 Details : If there is an inconsistency between the MDS and OST filesystems,
3813 such that the MDS references an object that doesn't exist, writes
3814 to that object will leak memory due to incorrect cleanup in the
3815 error handling path, eventually running out of memory on the OST.
3820 Description: Creating too long symlink causes lustre errors
3821 Details : Check symlink and name lengths before sending requests to MDS.
3824 Frequency : only if flock is enabled (not on by default)
3826 Description: posix locks not released on fd closure on 2.6.9+
3827 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3828 kernels, this caused such locks not to be released on fd close and
3829 then assertions on fs unmount about still used locks.
3832 Frequency : MDS failover only, very rarely
3834 Description: clients may get ASSERTION(granted_lock != NULL)
3835 Details : When request was taking a long time, and a client was resending
3836 a getattr by name lock request. The were multiple lock requests
3837 with the same client lock handle and
3838 mds_getattr_name->fixup_handle_for_resent_request found one of the
3839 lock handles but later failed with ASSERTION(granted_lock != NULL).
3844 Description: handle->h_buffer_credits > 0, assertion failure
3845 Details : h_buffer_credits is zero after truncate, causing assertion
3846 failure. This patch extends the transaction or creates a new
3850 Frequency : NFS re-export or patchless client
3851 Bugzilla : 11179, 10796
3852 Description: Crash on NFS re-export node (__d_move)
3853 Details : We do not want to hash the dentry if we don't have a lock.
3854 But if this dentry is later used in d_move, we'd hit uninitialised
3855 list head d_hash, so we just do this to init d_hash field but
3856 leave dentry unhashed.
3859 Frequency : NFS re-export or patchless client
3861 Description: NFS exports has problem with symbolic link
3862 Details : lustre client didn't properly install dentry when re-exported
3863 to NFS or running patchless client.
3866 Frequency : NFS re-export or patchless client
3868 Description: Various nfs/patchless fixes.
3869 Details : fixes reuse disconected alias for lookup process - this fixes
3870 warning "find_exported_dentry: npd != pd",
3871 fix permission error with open files at nfs.
3872 fix apply umask when do revalidate.
3875 Frequency : occasional
3877 Description: Crash on NFS re-export node
3878 Details : calling clear_page() on the wrong pointer triggered oops in
3879 generic_mapping_read().
3882 Frequency : rarely, using O_DIRECT IO
3884 Description: unaligned directio crashes client with LASSERT
3885 Details : check for unaligned buffers before trying any requests.
3888 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3890 Description: stale data returned from RAID cache
3891 Details : If only a small amount of IO is done to the RAID device before
3892 reading it again it is possible to get stale data from the RAID
3893 cache instead of reading it from disk.
3896 Frequency : always for sles10 kernel
3898 Description: sles10 support
3899 Details : ll_follow_link: compile fixes and using of nd_set_link
3900 under newer kernels.
3903 Frequency : depends on arch, kernel and compiler version, always on sles10
3906 Description: recursive or deep enough symlinks cause stack overflow
3907 Details : getting rid of large stack-allocated variable in
3911 Frequency : depends on hardware
3913 Description: lustre write performance loss in the SLES10 kernel
3914 Details : the performance loss is caused by using of write barriers in the
3915 ext3 code. The SLES10 kernel turns barrier support on by
3916 default. The fix is to undo that change for ldiskfs.
3918 ------------------------------------------------------------------------------
3920 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3922 * Support for kernels:
3923 2.6.9-42.0.3EL (RHEL 4)
3924 2.6.5-7.276 (SLES 9)
3925 2.4.21-47.0.1.EL (RHEL 3)
3926 2.6.12.6 vanilla (kernel.org)
3930 Frequency : quota enabled and large files being deleted
3932 Description: releasing more than 4GB of quota at once hangs OST
3933 Details : If a user deletes more than 4GB of files on a single OST it
3934 will cause the OST to spin in an infinite loop. Release
3935 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3940 Description: statfs data retrieved from /proc may be stale or zero
3941 Details : When reading per-device statfs data from /proc, in the
3942 {kbytes,files}_{total,free,avail} files, it may appear
3943 as zero or be out of date.
3946 Frequency : systems with MD RAID1 external journal devices
3948 Description: lconf's call to blkid is confused by RAID1 journal devices
3949 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3950 of returning all block devices that match the journal UUID.
3953 Frequency : always, for aggregate stripe size over 4GB
3955 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3956 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3957 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3960 Frequency : always if "lfs find" used on a local file/directory
3962 Description: "lfs find" segfaults if used on a local file/directory
3963 Details : The case where a directory component was not specified wasn't
3964 handled correctly. Handle this properly.
3967 Frequency : always on ppc64
3969 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3970 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3971 perform endianness conversion.
3974 Frequency : rarely (truncate to non-zero file size after write under load)
3975 Bugzilla : 10730, 10687
3976 Description: Files padded with zeros to next 4K multiple
3977 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3978 it is possible that files that are truncated to a non-zero size
3979 immediately after being written are filled with zero bytes beyond
3980 the truncated size. No file data is lost.
3982 Severity : enhancement
3984 Description: Allow recovery/failover for liblustre clients.
3985 Details : liblustre clients were unaware of failover configurations until
3988 Severity : enhancement
3990 Description: user file locks should fail when not mounting with flock option
3991 Details : Set up an error-returning stub in ll_file_operations.lock field
3992 to prevent incorrect behaviour when client is mounted without
3993 flock option. Also, set up properly f_op->flock field for
3997 Frequency : always on ia64
3999 Description: "lfs df" loops on printing out MDS statfs information
4000 Details : The obd_ioctl_data was not initialized and in some systems
4001 this caused a failure during the ioctl that did not return
4002 an error. Initialize the struct and return an error on failure.
4005 Frequency : SLES 9 only
4007 Description: Error of copying files with lustre special EAs as root
4008 Details : Client side always return success for setxattr call for lustre
4009 special xattr (currently only "trusted.lov").
4012 Frequency : rarely on clusters with both ia64+i386 clients
4014 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4015 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4016 clients are doing concurrent writes to the same file it is
4017 possible that the ia64 clients may LASSERT because the OST
4018 extent locks are not PAGE_SIZE aligned. Ensure that grown
4019 locks are always aligned on the request boundary.
4022 Frequency : specific use, occasional
4024 Description: Overwriting in use executable truncates on-disk binary image
4025 Details : If one node attempts to overwrite an executable in use by
4026 another node, we now correctly return ETXTBSY instead of
4027 truncating the file.
4029 Severity : enhancement
4031 Description: Async OSC create to avoid the blocking unnecessarily.
4032 Details : If an OST has no remaining object, system will block on the creating
4033 when need to create a new object on this OST. Now, ways use
4034 pre-created objects when available, instead of blocking on an
4035 empty osc while others are not empty. If we must block, we block
4036 for the shortest possible period of time.
4041 Description: chmod on Lustre root is propagated to other clients
4042 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4043 invalid by the follow_mount time.
4046 Frequency : liblustre clients only
4048 Description: Race in 'instant cancel' lock handling could lead to such locks
4049 never to be granted in case of SMP MDS
4050 Details : Do not destroy not yet granted but cbpending locks in
4054 Frequency : replay/resend of open
4056 Description: non null lock assetion failure in mds_intent_policy
4057 Details : Trying to replay/resend lockless open requests resulted in
4058 mds_open() returning 0 with no lock. Now it sets a flag if
4059 a lock is going to be returned.
4061 Severity : enhancement
4063 Description: Checksum enhancements
4064 Details : New checksum enhancements allow for resending RPCs that failed
4067 Severity : enhancement
4069 Description: Tunables on number of dirty pages in cacche
4070 Details : Allow to set limit on number of dirty pages cached.
4075 Description: client crash on unmount - lock still has references
4076 Details : In some error handling cases it was possible to leak a lock
4077 reference on a client while accessing a file. This was not
4078 harmful to the client during operation, but would cause the
4079 client to crash when the filesystem is unmounted.
4082 Frequency : specific case, rare
4084 Description: ETXTBSY on mds though file not in use by client
4085 Details : ETXTBSY is no longer incorrectly returned when attempting to
4086 chmod or chown a directory that the user previously tried to
4087 execute or a currently-executing binary.
4090 Frequency : extremely rare except on liblustre-based clients
4092 Description: Lustre space not freed when files are deleted
4093 Details : Clean up open-unlinked files after client eviction. Previously
4094 the unlink was skipped and the files remained as orphans.
4099 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4100 Details : In some cases it was possible to send a blocking callback to
4101 a client doing a glimpse, even though that client didn't get
4102 a lock granted. When the glimpse lock is cancelled on the OST
4103 the freed lock is left on the waiting list and corrupted the list.
4106 Frequency : all core dumps
4108 Description: Broke core dumps to lustre
4109 Details : Negative dentry may be unhashed if parent does not have UPDATE
4110 lock, but some callers, e.g. do_coredump, expect dentry to be
4111 hashed after successful create, hash it in ll_create_it.
4113 ------------------------------------------------------------------------------
4115 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4117 * Support for kernels:
4118 2.6.9-42.0.2.EL (RHEL 4)
4119 2.6.5-7.276 (SLES 9)
4120 2.4.21-40.EL (RHEL 3)
4121 2.6.12.6 vanilla (kernel.org)
4125 Frequency : always on RHEL 3
4127 Description: Number of open files grows over time
4128 Details : The number of open files grows over time, whether or not
4129 Lustre is started. This was due to a filp leak introduced
4130 by one of our kernel patches.
4132 ------------------------------------------------------------------------------
4134 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4136 * Support for kernels:
4137 2.6.9-42.EL (RHEL 4)
4138 2.6.5-7.276 (SLES 9)
4139 2.4.21-40.EL (RHEL 3)
4140 2.6.12.6 vanilla (kernel.org)
4145 Bugzilla : 5719, 9635, 9792, 9684
4146 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4147 Details : If a server is under heavy load and cannot reply to new
4148 connection requests before the client resends the (re)connect,
4149 the connection handling code can behave badly if two service
4150 threads are concurrently handing separate (re)connections from
4151 the same client. Add better locking to the connection handling
4152 code, and ensure that only a single connection will be processed
4153 for a given client UUID, even if the lock is dropped.
4155 Severity : enhancement
4157 Description: add TCP zero-copy support to kernel
4158 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4159 sends if the hardware supports scatter-gather and checksumming.
4160 This allows socklnd to do client-write and server-read more
4161 efficiently and reduce CPU utilization from skbuf copying.
4164 Frequency : only if NFS exporting from client
4166 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4167 Details : If setting attributes on a file created under NFS that had
4168 never been opened it would be possible to oops the client
4169 if the file had no objects.
4172 Frequency : always for liblustre
4174 Description: liblustre client does MDS+OSTs setattr RPC for each write
4175 Details : When doing a write from a liblustre client, the client
4176 incorrectly issued an RPC to the MDS and each OST the file was
4177 striped over in order to update the timestamps. When writing
4178 with small chunks and many clients this could overwhelm the MDS
4179 with RPCs. In all cases it would slow down the write because
4180 these RPCs are unnecessary.
4182 Severity : enhancement
4184 Description: allow number of MDS service threads to be changed at module load
4185 Details : It is now possible to change the number of MDS service threads
4186 running. Adding "options mds mds_num_threads={N}" to the MDS's
4187 /etc/modprobe.conf will set the number of threads for the next
4188 time Lustre is restarted (assuming the "mds" module is also
4189 reloaded at that time). The default number of threads will
4190 stay the same, 32 for most systems.
4195 Description: OST crash if filesystem is unformatted or corrupt
4196 Details : If an OST is started on a device that has never been formatted
4197 or if the filesystem is corrupt and cannot even mount then the
4198 error handling cleanup routines would dereference a NULL pointer.
4203 Description: NULL pointer deref in llap_from_page.
4204 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4205 incorrect) mapping if the page was truncated/reclaimed while it was
4206 searched for. Check for this condition and skip such pages when
4207 doing readahead. Introduce extra check to llap_from_page() to
4208 verify page->mapping->host is non-NULL (so page is not anonymous).
4211 Frequency : Sometimes when using sys_sendfile
4213 Description: "page not covered by a lock" warnings from ll_readpage
4214 Details : sendfile called ll_readpage without right page locks present.
4215 Now we introduced ll_file_sendfile that does necessary locking
4216 around call to generic_file_sendfile() much like we do in
4220 Frequency : with certain MDS communication failures at client mount time
4222 Description: NULL pointer deref after failed client mount
4223 Details : a client connection request may delayed by the network layer
4224 and not be sent until after the PTLRPC layer has timed out the
4225 request. If the client fails the mount immediately it will try
4226 to clean up before the network times out the request. Add a
4227 reference from the request import to the obd device and delay
4228 the cleanup until the network drops the request.
4231 Frequency : occasionally during client (re)connect
4233 Description: assertion failure during client (re)connect
4234 Details : processing a client connection request may be delayed by the
4235 client or server longer than the client connect timeout. This
4236 causes the client to resend the connection request. If the
4237 original connection request is replied in this interval, the
4238 client may trip an assertion failure in ptlrpc_connect_interpret()
4239 which thought it would be the only running connect process.
4242 Frequency : only with obd_echo servers and clients that are rebooted
4244 Description: kernel BUG accessing uninitialized data structure
4245 Details : When running an obd_echo server it did not start the ping_evictor
4246 thread, and when a client was evicted an uninitialized data
4247 structure was accessed. Start the ping_evictor in the RPC
4248 service startup instead of the OBD startup.
4250 Severity : enhancement
4251 Bugzilla : 10193 (patchless)
4252 Description: Remove dependency on various unexported kernel interfaces.
4253 Details : No longer need reparent_to_init, exit_mm, exit_files,
4254 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4257 Frequency : rare (only users of deprecated and unsupported LDAP config)
4259 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4260 Details : LDAP apparently contains 'lustreName' attributes instead of
4261 'name'. A simple remapping of the name is sufficient.
4264 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4266 Description: waiting_locks_callback trips kernel BUG if client is evicted
4267 Details : Running with the dump_on_timeout debug flag turned on makes
4268 it possible that the waiting_locks_callback() can try to dump
4269 the Lustre kernel debug logs from an interrupt handler. Defer
4270 this log dumping to the expired_lock_main() thread.
4272 Severity : enhancement
4274 Description: Support NFS exporting on 2.6 kernels.
4275 Details : Implement non-rawops metadata methods for NFS server to use without
4276 changing NFS server code.
4279 Frequency : very rare (synthetic metadata workload only)
4281 Description: two racing renames might cause an MDS thread to deadlock
4282 Details : Running the "racer" program may cause one MDS thread to rename
4283 a file from being the source of a rename to being the target of
4284 a rename at exactly the same time that another thread is doing
4285 so, and the second thread has already enqueued these locks after
4286 doing a lookup of the target and is trying to relock them in
4287 order. Ensure that we don't try to re-lock the same resource.
4290 Frequency : only very large systems with liblustre clients
4292 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4293 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4294 all OSTs in parallel. This allows the eviction of stale liblustre
4295 clients to proceed much faster than if they were done in series,
4296 and also offers similar improvements for other set_info RPCs.
4301 Description: excessive CPU usage during initial read phase on client
4302 Details : During the initial read phase on a client, it would agressively
4303 retry readahead on the file, consuming too much CPU and impacting
4304 performance (since 1.4.5.8). Improve the readahead algorithm
4305 to avoid this, and also improve some other common cases (read
4306 of small files in particular, where "small" is files smaller than
4307 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4312 Description: MDS crash when receiving packet with unknown intent.
4313 Details : Do not LBUG in unknown intent case, just return -EFAULT
4315 Severity : enhancement
4316 Bugzilla : 9293, 9385
4317 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4318 Details : Do not serialize getattr (non-intent version) and statfs.
4321 Frequency : occasional, when OST network is overloaded/intermittent
4323 Description: client evicted by OST after bulk IO timeout
4324 Details : If a client sends a bulk IO request (read or write) the OST
4325 may evict the client if it is unresposive to its data GET/PUT
4326 request. This is incorrect if the network is overloaded (takes
4327 too long to transfer the RPC data) or dropped the OST GET/PUT
4328 request. There is no need to evict the client at all, since
4329 the pinger and/or lock callbacks will handle this, and the
4330 client can restart the bulk request.
4333 Frequency : Always when mmapping file with no objects
4335 Description: client crashes when mmapping file with no objects
4336 Details : Check that we actually have objects in a file before doing any
4337 operations on objects in ll_vm_open, ll_vm_close and
4343 Description: Request leak when working with deleted CWD
4344 Details : Introduce advanced request refcount tracking for requests
4345 referenced from lustre intent.
4347 Severity : Enhancement
4349 Description: Cache open file handles on client.
4350 Details : MDS now will return special lock along with openhandle, if
4351 requested and client is allowed to hold openhandle, even if unused,
4352 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4353 closing files for every read/write openration.
4355 Severity : Enhancement
4357 Description: Cache open negative dentries on client when possible.
4358 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4359 negative dentries on lock revocation.
4364 Description: Remounting a client read-only wasn't possible with a zconf mount
4365 Details : It wasn't possible to remount a client read-only with llmount.
4367 Severity : enhancement
4368 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4369 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4370 MPICH is included to add Lustre-specific ADIO interfaces.
4371 This is based closely on the UFS ADIO layer and only differs
4372 in file creation, in order to allow the OST striping to be set.
4373 This is user-contributed code and not supported by CFS.
4378 Description: extended inode attributes (immutable, append-only) work improperly
4379 when 2.4 and 2.6 kernels are used on client/server or vice versa
4380 Details : Introduce kernel-independent values for these flags.
4382 Severity : enhancement
4385 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4386 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4387 granularity. Now, Fractional MB granularity is be supported,
4388 this is very useful for low memory system.
4390 Severity : enhancement
4392 Description: Getattr by fid
4393 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4394 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4395 have locks and avoids getattr rpc requests for it on every stat.
4398 Frequency : Always, for filesystems larger than 2TB
4400 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4401 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4402 filesystems larger than 2TB to be mounted with mballoc enabled.
4405 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4407 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4408 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4409 filesystems larger than 2TB is mounted then the kernel will
4410 silently corrupt the start of the filesystem. CONFIG_LBD is
4411 enabled for all CFS-supported kernels, but the possibility of
4412 this happening with a modified kernel config exists.
4414 Severity : enhancement
4416 Description: add client O_DIRECT support for 2.6 kernels
4417 Details : It is now possible to do O_DIRECT reads and writes to files
4418 in the Lustre client mountpoint on 2.6 kernel clients.
4420 Severity : enhancement
4422 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4423 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4424 parallel, not waiting for response from every OST before sending
4425 a rpc to the next OST.
4430 Description: setattr vs write race when updating file timestamps
4431 Details : Client processes that update a file timestamp into the past
4432 right after writing to the file (e.g. tar) it is possible that
4433 the updated file modification time can be reset to the current
4434 time due to a race between processing the setattr and write RPC.
4436 Severity : enhancement
4438 Description: Bring 'lfs find' closer in line with regular Linux find.
4439 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4440 -print0 options and obtains all the needed info through the lustre
4443 Severity : enhancement
4445 Description: support up to 1024 configured devices on one node
4446 Details : change obd_dev array from statically allocated to dynamically
4447 allocated structs as they are first used to reduce memory usage
4452 Description: Flush dirty partially truncated pages during truncate
4453 Details : Immediatelly flush partially truncated pages in filter_setattr,
4454 this way we completely avoid having any pages in page cache on OST
4455 and can retire ugly workarounds during writes to flush such pages.
4460 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4461 Details : protect inode from truncation within vfs_unlink() context
4462 just take a reference before calling vfs_unlink() and release it
4463 when parent's i_sem is free.
4468 Description: last_id value checked outside lock on OST caused LASSERT failure
4469 Details : If there were multiple MDS->OST object precreate requests in
4470 flight, it was possible that the OST's last object id was checked
4471 outside a lock and incorrectly tripped an assertion. Move checks
4472 inside locks, and discard old precreate requests.
4475 Frequency : always, if extents are used on OSTs
4477 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4478 Details : OSTs using the extents format would not zero the high 16 bits of
4479 the index physical block number. This is not a problem for any
4480 OST filesystems smaller than 16TB, and no kernels support ext3
4481 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4482 new/modified files) and can be fixed for existing filesystems
4483 with e2fsprogs-1.39-cfs1.
4488 Description: import connection selection may be incorrect if timer wraps
4489 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4490 import connections to be ignored if the 32-bit jiffies counter
4491 wraps. Use a 64-bit jiffies counter.
4494 Frequency : very large clusters immediately after boot
4496 Description: LNET request buffers exhausted under heavy short-term load
4497 Details : If a large number of client requests are generated on a service
4498 that has previously never seen so many requests it is possible
4499 that the request buffer growth cannot keep up with the spike in
4500 demand. Instead of dropping incoming requests, they are held in
4501 the LND until the RPC service can accept more requests.
4504 Frequency : Sometimes during replay
4506 Description: Assertion failure in ll_local_open after replay.
4507 Details : If replay happened on an open request reply before we were able
4508 to set replay handler, reply will become not swabbed tripping the
4509 assertion in ll_local_open. Now we set the handler right after
4510 recognising of open request
4513 Frequency : very rare
4515 Description: kernel reports "badness in vsnprintf"
4516 Details : Reading from the "recovery_status" /proc file in small chunks
4517 may cause a negative length in lprocfs_obd_rd_recovery_status()
4518 call to vsnprintf() (which is otherwise harmless). Exit early
4519 if there is no more space in the output buffer.
4521 Severity : enhancement
4523 Description: clear OBD RPC statistics by writing to them
4524 Details : It is now possible to clear the OBD RPC statistics by writing
4525 to the "stats" file.
4530 Description: Client mtime is not the same on different clients after utimes
4531 Details : In some cases, the client was using the utimes() syscall on
4532 a file cached on another node. The clients now validate the
4533 ctime from the MDS + OSTs to determine which one is right.
4538 Description: Inability to activate failout mode
4539 Details : lconf script incorrectly assumed that in pythong string's numeric
4540 value is used in comparisons.
4543 Frequency : always with multiple stripes per file
4545 Description: Inefficient object allocation for mutli-stripe files
4546 Details : When selecting which OSTs to stripe files over, for files with
4547 a stripe count that divides evenly into the number of OSTs,
4548 the MDS is always picking the same starting OST for each file.
4549 Return the OST selection heuristic to the original design.
4554 Description: mount failures may take full timeout to return an error
4555 Details : Under some heavy load conditions it is possible that a
4556 failed mount can wait for the full obd_timeout interval,
4557 possibly several minutes, before reporting an error.
4558 Instead return an error as soon as the status is known.
4560 Frequency : quota enabled and large files being deleted
4562 Description: releasing more than 4GB of quota at once hangs OST
4563 Details : If a user deletes more than 4GB of files on a single OST it
4564 will cause the OST to spin in an infinite loop. Release
4565 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4570 Description: statfs data retrieved from /proc may be stale or zero
4571 Details : When reading per-device statfs data from /proc, in the
4572 {kbytes,files}_{total,free,avail} files, it may appear
4573 as zero or be out of date.
4576 Frequency : systems with MD RAID1 external journal devices
4578 Description: lconf's call to blkid is confused by RAID1 journal devices
4579 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4580 of returning all block devices that match the journal UUID.
4583 Frequency : always, for aggregate stripe size over 4GB
4585 Description: assertion fails when trying to use 4GB stripe size
4586 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4587 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4590 Frequency : always on ppc64
4592 Description: the first write on an ext3 filesystem with mballoc got stuck
4593 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4594 perform endianness conversion.
4596 ------------------------------------------------------------------------------
4598 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4600 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4601 INTEROPERATE with older versions automatically. Please read the
4602 user documentation before upgrading any part of a live system.
4603 * WARNING: Lustre networking configuration changes are required with
4604 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4607 * Support for kernels:
4608 2.6.9-22.0.2.EL (RHEL 4)
4609 2.6.5-7.244 (SLES 9)
4610 2.6.12.6 vanilla (kernel.org)
4613 Severity : enhancement
4614 Bugzilla : 7981/8208
4615 Description: Introduced Lustre Networking (LNET)
4616 Details : LNET is new networking infrastructure for Lustre, it includes
4617 a reorganized network configuration mode (see the user
4618 documentation for full details) as well as support for routing
4619 between different network fabrics. Lustre Networking Devices
4620 (LNDs) for the supported network fabrics have also been
4621 created for this new infrastructure.
4623 Severity : enhancement
4624 Description: Introduced Access control lists
4625 Details : clients can set ACLs on files and directories in order to have
4626 more fine-grained permissions than the standard Unix UGO+RWX.
4627 The MDS must be started with the "-o acl" mount option.
4629 Severity : enhancement
4630 Description: Introduced filesystem quotas
4631 Details : Administrators may now establish per-user quotas on the
4634 Severity : enhancement
4636 Description: Configuration change for the XT3
4637 The PTLLND is now used to run Lustre over Portals on the XT3
4638 The configure option(s) --with-cray-portals are no longer used.
4639 Rather --with-portals=<path-to-portals-includes> is used to
4640 enable building on the XT3. In addition to enable XT3 specific
4641 features the option --enable-cray-xt3 must be used.
4646 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4647 Details : A race between allocating a new llcd and re-getting the llcd_lock
4648 allowed another thread to grab newly-allocated llcd.
4650 Severity : enhancement
4652 Description: 2.6 OST async journal commit and locking fix to improve performance
4653 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4654 2.6 kernels are now async as they already were in 2.4 kernels so
4655 that they can commit concurrently with the network bulk transfer.
4656 For block-allocated files the filter allocation semaphore is held
4657 to avoid filesystem fragmentation during allocation. BKL lock
4658 removed for 2.6 xattr operations where it is no longer needed.
4663 Description: lconf incorrectly determined whether two IP networks could talk
4664 Details : In some more complicated routing and multiple-network
4665 configurations, lconf will avoid trying to make a network
4666 connection to a disjoint part of the IP space. It was doing the
4667 math incorrectly for one set of cases.
4672 Description: Fix for potential infinite loop processing records in an llog.
4673 Details : If an llog record is corrupted/zeroed, it is possible to loop
4674 forever in llog_process(). Validate the llog record length
4675 and skip the remainder of the block on error.
4678 Frequency : occasional (liblustre only)
4680 Description: liblustre could not open files whose last component is a symlink
4681 Details : sysio_path_walk() would incorrectly pass the open intent to
4682 intermediate path components.
4685 Frequency : rare (liblustre only with non-standard tuning)
4686 Bugzilla : 7201 (7350)
4687 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4688 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4689 after it has been released, only do this for referenced locks
4691 Severity : enhancement
4693 Description: specifying an (invalid) directory default stripe_size of -1
4694 would reset the directory default striping
4695 Details : stripe_size -1 was used internally to signal directory stripe
4696 removal, now use "all default" to signal dir stripe removal
4697 as a directory striping of "all default" is not useful
4700 Frequency : common for large clusters running liblustre clients
4702 Description: doing an ls when liblustre clients are running is slow
4703 Details : sending a glimpse AST to a liblustre client waits for every AST
4704 to time out, as liblustre clients will not respond. Since they
4705 cannot cache data we refresh the OST lock LVB from disk instead.
4707 Severity : enhancement
4709 Description: doing an ls at the same time as file IO can be slow
4710 Details : enqueue and other "small" requests can be blocked behind many
4711 large IO requests. Create a new OST IO portal for non-IO
4712 requests so they can be processed faster.
4715 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4717 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4718 created file with mode 000
4719 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4720 allow an ftruncate() as a non-root user to a file with mode 000.
4721 Lustre now respects this flag to disable mode checks when
4722 truncating a file owned by the user
4725 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4728 Description: Revoking locks from clients that went dead or catatonic might take
4730 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4731 cancellation of such locks instant on servers without waiting for
4732 any reply from clients. Clients drops these locks when cancel
4733 notification from server is received without replying.
4736 Frequency : liblustre-only, when liblustre client dies or becomes busy
4738 Description: Doing ls on Linux clients can take a long time with active
4740 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4741 granting such locks to it in the first place if possible. Locks
4742 are taken by proxy on the OST during the read or write and
4743 dropped immediately afterward. Add connect flags handling, do
4744 not grant locks to liblustre clients for glimpse ASTs.
4746 Severity : enhancement
4748 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4749 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4750 sequential reads and behaves badly with applications doing random
4751 reads. Improve it by reading ahead at least read region, and
4752 avoiding excessive large RPC for small reads.
4754 Severity : enhancement
4756 Description: Creating more than 1000 files for a single job may cause a load
4757 imbalance on the OSTs if there are also a large number of OSTs.
4758 Details : qos_prep_create() uses an OST index reseed value that is an
4759 even multiple of the number of available OSTs so that if the
4760 reseed happens in the middle of the object allocation it will
4761 still utilize the OSTs as uniformly as possible.
4766 Description: OST or MDS may oops in ping_evictor_main()
4767 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4768 but doesn't restart at beginning of obd_exports_timed list
4771 Severity : enhancement
4773 Description: improve by-nid export eviction on the MDS and OST
4774 Details : allow multiple exports with the same NID to be evicted at one
4775 time without re-searching the exports list.
4778 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4780 Description: MDS may oops in groups_free()
4781 Details : in rare race conditions a newly allocated group_info struct is
4782 freed again, and this can be NULL. The 2.4 compatibility code
4783 for groups_free() checked for a NULL pointer, but 2.6 did not.
4786 Frequency : common for liblustre clients doing little filesystem IO
4787 Bugzilla : 9352, 7313
4788 Description: server may evict liblustre clients accessing contended locks
4789 Details : if a client is granted a lock or receives a completion AST
4790 with a blocking AST already set it would not reply to the AST
4791 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4794 Frequency : lfs setstripe, only systems with more than 160 OSTs
4796 Description: unable to set striping with a starting offset beyond OST 160
4797 Details : llapi_create_file() incorrectly limited the starting stripe
4798 index to the maximum single-file stripe count.
4801 Frequency : LDAP users only
4803 Description: lconf did not handle in-kernel recovery with LDAP properly
4804 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4806 Severity : enhancement
4808 Description: bind OST threads to NUMA nodes to improve performance
4809 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4810 node and do their allocations there to localize memory access
4812 Severity : enhancement
4814 Description: llmount can determine client NID directly from Myrinet (GM)
4815 Details : the client NID code from gmnalnid was moved directly into
4816 llmount, removing the need to use this or specifying the
4817 client NID explicitly when mounting GM clients with zeroconf
4820 Frequency : if client is started with down MDS
4822 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4823 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4824 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4825 ptlrpc_import_delayed_req() because we didn't check if the
4826 request was interrupted, and we also didn't break out of the
4827 event loop if there was no timeout
4832 Description: data loss during non-page-aligned writes to a single file from
4833 both multiple nodes and multiple threads on one node at same time
4834 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4835 inconsistency led to false short-reads, that were cached and later
4836 used by ->prepare_write() to fill in partially written page,
4837 leading to data loss.
4840 Frequency : always, if lconf --abort_recovery used
4842 Description: lconf --abort_recovery fails with 'Operation not supported'
4843 Details : lconf was attempting to abort recovery on the MDT device and not
4846 Severity : enhancement
4848 Description: remove cleanup logs
4849 Details : replace lconf-generated cleanup logs with lustre internal
4850 cleanup routines. Eliminates the need for client-cleanup and
4853 Severity : enhancement
4855 Description: add support for EAs (user and system) on lustre filesystems
4856 Details : it is now possible to store extended attributes in the Lustre
4857 client filesystem, and with the user_xattr mount option it
4858 is possible to allow users to store EAs on their files also
4860 Severity : enhancement
4862 Description: Add possibility (config option) to show minimal available OST free
4864 Details : When compiled with --enable-mindf configure option, statfs(2)
4865 (and so, df) will return least minimal free space available from
4866 all OSTs as amount of free space on FS, instead of summary of
4867 free spaces of all OSTs.
4869 Severity : enhancement
4871 Description: do not expand extent locks acquired on OST-side
4872 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4873 by server: they are not cached anyway.
4876 Frequency : when mmap is used/binaries executed from Lustre
4878 Description: Unmmap pages before throwing them away from read cache.
4879 Details : llap_shrink cache now attempts to unmap pages before discarding
4880 them (if unmapping failed - do not discard). SLES9 kernel has
4881 extra checks that trigger if this unmapping is not done first.
4886 Description: lconf didn't resolve symlinks before checking to see whether a
4887 given mountpoint was already in use
4890 Frequency : when migrating failover services
4891 Bugzilla : 6395, 9514
4892 Description: When migrating a subset of services from a node (e.g. failback
4893 from a failover service node) the remaining services would
4894 time out and evict clients.
4895 Details : lconf --force (implied by --failover) sets the global obd_timeout
4896 to 5 seconds in order to quickly disconnect, but this caused
4897 other RPCs to time out too quickly. Do not change the global
4898 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4900 Severity : enhancement
4901 Frequency : if MDS is started with down OST
4902 Bugzilla : 9439,5706
4903 Description: Allow startup/shutdown of an MDS without depending on the
4904 availability of the OSTs.
4905 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4906 Add appropriate locking and lov-osc refcounts for safe
4907 cleaning. Add osc abort_inflight calls in case the
4908 synchronize never started.
4911 Frequency : occasional (Cray XT3 only)
4913 Description: root not authorized to access files in CRAY_PORTALS environment
4914 Details : The client process capabilities were not honoured on the MDS in
4915 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4916 been accessed by an authorized user then root was able to access
4917 the file on the local client also. The root user capabilities
4918 are now allowed on the MDS, as this environment has secure UID.
4921 Frequency : occasional
4923 Description: ldiskfs "too long searching" message happens too often
4924 Details : A debugging message (otherwise harmless) prints too often on
4925 the OST console. This has been reduced to only happen when
4926 there are fragmentation problems on the filesystem.
4931 Description: Division by zero in statfs when all OSCs are inactive
4932 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4933 lov_statfs divides by value returned by lov_get_stripecnt().
4937 Bugzilla : 9489, 3273
4938 Description: First write from each client to each OST was only 4kB in size,
4939 to initialize client writeback cache, which caused sub-optimal
4940 RPCs and poor layout on disk for the first written file.
4941 Details : Clients now request an initial cache grant at (re)connect time
4942 and so that they can start streaming writes to the cache right
4943 away and always do full-sized RPCs if there is enough data.
4944 If the OST is rebooted the client also re-establishes its grant
4945 so that client cached writes will be honoured under the grant.
4950 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4951 Details : Now I/O RPCs go to different portal number and (presumably) fast
4952 lock requests (and glimses) and other RPCs get their own service
4953 threads pool that should be able to service those RPCs
4956 Severity : enhancement
4958 Description: Ability to exchange lustre version between client and servers and
4959 issue warnings at client side if client is too old. Also for
4960 liblustre clients there is ability to refuse connection of too old
4962 Details : New 'version' field is added to connect data structure that is
4963 filled with version info. That info is later checked by server and
4967 Frequency : rare, liblustre only.
4968 Bugzilla : 9296, 9581
4969 Description: Two simultaneous writes from liblustre at offset within same page
4970 might proceed at the same time overwriting eachother with stale
4972 Details : I/O lock withing llu_file_prwv was released too early, before data
4973 actually was hitting the wire. Extended lock-holding time until
4974 server acknowledges receiving data.
4977 Frequency : extremely rare. Never observed in practice.
4979 Description: avoid generating lustre_handle cookie of 0.
4980 Details : class_handle_hash() generates handle cookies by incrementing
4981 global counter, and can hit 0 occasionaly (this is unlikely, but
4982 not impossible, because initial value of cookie counter is
4983 selected randonly). Value of 0 is used as a sentinel meaning
4984 "unassigned handle" --- avoid it. Also coalesce two critical
4985 sections in this function into one.
4987 Severity : enhancement
4989 Description: allow liblustre clients to delegate truncate locking to OST
4990 Details : To avoid overhead of locking, liblustre client instructs OST to
4991 take extent lock in ost_punch() on client's behalf. New connection
4992 flag is added to handle backward compatibility.
4994 Severity : enhancement
4995 Bugzilla : 4928, 7341, 9758
4996 Description: allow number of OST service threads to be specified
4997 Details : a module parameter allows the number of OST service threads
4998 to be specified via "options ost ost_num_threads={N}" in the
4999 OSS's /etc/modules.conf or /etc/modprobe.conf.
5003 Bugzilla : 6146, 9635, 9895
5004 Description: servers crash with bad pointer in target_handle_connect()
5005 Details : In rare cases when a client is reconnecting it was possible that
5006 the connection request was the last reference for that export.
5007 We would temporarily drop the export reference and get a new
5008 one, but this may have been the last reference and the export
5009 was just destroyed. Get new reference before dropping old one.
5011 Severity : enhancement
5012 Frequency : if client is started with failover MDS
5014 Description: Allow multiple MDS hostnames in the mount command
5015 Details : Try to read the configuration from all specified MDS
5016 hostnames during a client mount in case the "primary"
5019 Severity : enhancement
5021 Description: Stop sending data to evicted clients as soon as possible.
5022 Details : Check if the client we are about to send or are sending data to
5023 was evicted already. (Check is done every second of waiting,
5024 for which l_wait_event interface was extended to allow checking
5025 of exit condition at specified intervals).
5028 Frequency : rare, normally only when NFS exporting is done from client
5030 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5032 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5035 Frequency : always, if srand() is called before liblustre initialization
5037 Description: Liblustre uses system PRNG disturbing its usage by user application
5038 Details : Introduce internal to lustre fast and high-quality PRNG for
5039 lustre usage and make liblustre and some other places in generic
5040 lustre code to use it.
5042 Severity : enhancement
5043 Bugzilla : 9477, 9557, 9870
5044 Description: Verify that the MDS configuration logs are updated when xml is
5045 Details : Check if the .xml configuration logs are newer than the config
5046 logs stored on the MDS and report an error if this is the case.
5047 Request --write-conf, or allow starting with --old_conf.
5049 Severity : enhancement
5051 Description: Handle symlinks in the path when checking if Lustre is mounted.
5052 Details : Resolve intermediate symlinks when checking if a client has
5053 mounted a filesystem to avoid duplicate client mounts.
5058 Description: lconf can hit an error exception but still return success.
5059 Details : The lconf command catches the Command error exception at the top
5060 level script context and will exit with the associated exit
5061 status, but doesn't ensure that this exit status is non-zero.
5066 Description: failure of ptlrpc thread startup can cause oops
5067 Details : Starting a ptlrpc service thread can fail if there are a large
5068 number of threads or the server memory is very fragmented.
5069 Handle this without oopsing.
5072 Frequency : always, only if liblustre and non-default acceptor port was used
5074 Description: liblustre cannot connect to servers with non-default acceptor port
5075 Details : tcpnal_set_default_params() was not called and was therefore
5076 ignoring the environment varaible TCPNAL_PORT, as well as other
5077 TCPNAL_ environment variables
5082 Description: two objects could be created on the same OST for a single file
5083 Details : If an OST is down, in some cases it was possible to create two
5084 objects on a single OST for a single file. No problems other
5085 than potential performance impact and spurious error messages.
5089 Bugzilla : 5681, 9562
5090 Description: Client may oops in ll_unhash_aliases
5091 Details : Client dcache may become inconsistent in race condition.
5092 In some cases "getcwd" can fail if the current directory is
5098 Description: Inode refcounting problems in NFS export code
5099 Details : link_raw functions used to call d_instantiate without obtaining
5100 extra inode reference first.
5104 Bugzilla : 9942, 9903
5105 Description: Referencing freed requests leading to crash, memleaks with NFS.
5106 Details : We used to require that call to ll_revalidate_it was always
5107 followed by ll_lookup_it. Also with revalidate_special() it is
5108 possible to call ll_revalidate_it() twice for the same dentry
5109 even if first occurence returned success. This fix changes semantic
5110 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5111 reference on a request referred from the intent.
5112 ll_intent_release() then releases such a request.
5115 Frequency : rare, normally benchmark loads only
5117 Description: unlinked inodes were kept in memory on the client
5118 Details : If a client is repeatedly creating and unlinking files it
5119 can accumulate a lot of stale inodes in the inode slab cache.
5120 If there is no other client load running this can cause the
5121 client node to run out of memory. Instead flush old inodes
5122 from client cache that have the same inode number as a new inode.
5125 Frequency : SLES9 2.6.5 kernel and long filenames only
5126 Bugzilla : 9969, 10379
5127 Description: utime reports stale NFS file handle
5128 Details : SLES9 uses out-of-dentry names in some cases, which confused
5129 the lustre dentry revalidation. Change it to always use the
5133 Frequency : rare, unless heavy write-truncate concurrency is continuous
5134 Bugzilla : 4180, 6984, 7171, 9963, 9331
5135 Description: OST becomes very slow and/or deadlocked during object unlink
5136 Details : filter_destroy() was holding onto the parent directory lock
5137 while truncating+unlinking objects. For very large objects this
5138 may block other threads for a long time and slow overall OST
5139 responsiveness. It may also be possible to get a lock ordering
5140 deadlock in this case, or run out of journal credits because of
5141 the combined truncate+unlink. Solution is to do object truncate
5142 first in one transaction without parent lock, and then do the
5143 final unlink in a new transaction with the parent lock. This
5144 reduces the lock hold time dramatically.
5147 Frequency : rare, 2.4 kernels only
5149 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5150 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5151 before unmounting their filesystem in order to avoid deadlock.
5152 The kernel_locked() function in 2.4 kernels only checks whether
5153 the kernel lock is held, not whether it is this process that is
5154 holding it as 2.6 kernels do.
5159 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5160 Details : The client ptlrpc code may be trying to reconnect to a down
5161 server before a previous connection attempt has timed out.
5162 Increase the reconnect interval to be longer than the connection
5163 timeout interval to avoid sending duplicate connections to
5167 Frequency : echo_client brw_test command
5169 Description: fix echo_client to work with OST preallocated code
5170 Details : OST preallocation code (5137) didn't take echo_client IO path
5171 into account: echo_client calls filter methods outside of any
5172 OST thread and, hence, there is no per-thread preallocated
5173 pages and buffers to use. Solution: hijack pga pages for IO. As
5174 a byproduct, this avoids unnecessary data copying.
5178 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5179 Description: Client can oops in mdc_commit_close() after open replay
5180 Details : It was possible for the MDS to return an open request with no
5181 transaction number in mds_finish_transno() if the client was
5182 evicted, but without actually returning an error. Clients
5183 would later try to replay that open and may trip an assertion
5184 Simplify the client close codepath, and always return an error
5185 from the MDS in case the open is not successful.
5188 Frequency : rare, 2.6 OSTs only
5190 Description: OST may deadlock under high load on fragmented files
5191 Details : If there was a heavy load and highly-fragmented OST filesystems
5192 it was possible to have all the OST threads deadlock waiting on
5193 allocation of biovecs, because the biovecs were not released
5194 until the entire RPC IO was completed. Instead, release biovecs
5195 as soon as they are complete to ensure forward IO progress.
5197 Severity : enhancement
5199 Description: Support for specifying external journal device at mount
5200 Details : If an OST or MDS device is formatted with an external journal
5201 device, this device major/minor is stored in the ext3 superblock
5202 and may not be valid for failover. Allow detecting and
5203 specifying the external journal at mount time.
5208 Description: Mounting an MDS with pending unlinked files may cause oops
5209 Details : target_finish_recovery() calls mds_postrecov() which returned
5210 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5211 considers this an error and immediately begins cleaning up the
5212 lov, just after starting the mds_lov process
5214 Severity : enhancement
5216 Description: Implement 'lfs df' to report actual free space on per-OST basis
5217 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5218 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5219 usage of objects; '-h' to report in human readable format.
5221 ------------------------------------------------------------------------------
5223 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5230 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5231 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5232 at startup during orphan unlinking. Instead, skip buddy update
5233 before setup, as it will be regenerated after recovery is complete.
5238 Description: If an OST is inactive, its locks might reference stale inodes.
5239 Details : lov_change_cbdata() must iterate over all namespaces, even if
5240 they are inactive to clear inode references from the lock.
5242 Severity : enhancement
5243 Frequency : occasional, if non-standard max_dirty_mb used
5245 Description: Client will block write RPCs if not enough grant
5246 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5247 then the client will block writes while waiting for another RPC
5248 to complete instead of consuming its dirty limit. With change
5249 we get improved performance when max_dirty_mb is small.
5251 Severity : enhancement
5252 Bugzilla : 3389, 6253
5253 Description: Add support for supplementary groups on the MDS.
5254 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5255 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5256 lookups for user supplementary groups into a cache.
5260 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5261 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5262 for liblustre O_EXCL is enforced by MDS.
5266 Description: readdir never returns NULL in liblustre.
5267 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5268 dentry in struct dirent.
5272 Frequency : liblustre only, depends on application IO pattern
5273 Description: liblustre clients evicted if not contacting servers
5274 Details : Don't put liblustre clients into the ping_evictor list, so
5275 they will not be evicted by the pinger ever.
5277 Severity : enhancement
5279 Description: Add ability to evict clients by NID from MDS.
5280 Details : By echoing "nid:$NID" string into
5281 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5282 $NID would be instantly evicted from this MDS and from all active
5283 OSTs connected to it.
5287 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5288 Details : lookup_it_finish() used to query file size from OSTs that was not
5293 Description: service threads change working directory to that of init
5294 Details : Starting lustre service threads may pin the working directory
5295 of the parent thread, making that filesystem busy. Threads
5296 now change to the working directory of init to avoid this.
5300 Frequency : during shutdown only
5301 Description: shutdown with a failed MDS or OST can cause unmount to hang
5302 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5307 Frequency : 2.6 only
5308 Description: chmod/chown may include an extra supplementary group
5309 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5310 supplementary group and if none is specified this is used.
5313 Bugzilla : 5479 (6816)
5315 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5316 Details : If lookup is in progress on a file that is unlinked we might try
5317 to revalidate the inode and fail in revalidate after lookup is
5318 complete and ll_file_open() enqueues the open again but
5319 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5322 Frequency : always, if lconf --abort_recovery used
5324 Description: lconf --abort_recovery fails with 'Operation not supported'
5325 Details : lconf was attempting to abort recovery on the MDT device and not
5328 ------------------------------------------------------------------------------
5330 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5335 Frequency : rare (only unsupported configurations with a node running as an
5337 Bugzilla : 6514, 5137
5338 Description: Mounting a Lustre file system on a node running as an OST could
5340 Details : OSTs now preallocates memory needed to write out data at
5341 startup, instead of when needed, to avoid having to
5342 allocate memory in possibly low memory situations.
5343 Specifically, if the file system is mounted on on OST,
5344 memory pressure could force it to try to write out data,
5345 which it needed to allocate memory to do. Due to the low
5346 memory, it would be unable to do so and the node would
5347 become unresponsive.
5349 Severity : enhancement
5351 Description: Addition of lconf --service command line option
5352 Details : lconf now accepts a '--service <arg>' option, which is
5353 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5355 Severity : enhancement
5357 Description: Failover mode is now the default for OSTs.
5358 Details : By default, OSTs will now run in failover mode. To return to
5359 the old behaviour, add '--failout' to the lmc line for OSTs.
5361 Severity : enhancement
5363 Description: Health checks are now provided for MDS and OSTs
5364 Details : Additional detailed health check information on MSD and OSTs
5365 is now provided through the procfs health_check value.
5368 Frequency : occasional, depends on IO load
5370 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5371 after numerous create/delete cycles
5372 Details : The ext3 inode allocation policy would not allocate new inodes
5373 very well on the OSTs because there are no new directories
5374 being created. Instead we look for groups with free space if
5375 the parent directories are nearly full.
5380 Description: Network or server problems during mount may cause partially
5381 mounted clients instead of returning an error.
5382 Details : The config llog parsing code may overwrite the error return
5383 code during mount error handling, returning success instead
5389 Description: MDS can fail to allocate large reply buffers
5390 Details : After long uptimes the MDS can fail to allocate large reply
5391 buffers (e.g. zconf client mount config records) due to memory
5392 fragmentation or consumption by the buffer cache. Preallocate
5393 some large reply buffers so that these replies can be sent even
5394 under memory pressure.
5398 Frequency : rare (liblustre)
5399 Description: fsx running with liblustre complained that using truncate() to
5400 extend the file doesn't work. This patch corrects that issue.
5401 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5402 ATTR_SIZE and lsm use in llu_setattr_raw.
5406 Frequency : rare, only 2.6 kernels
5407 Description: Unusual file access patterns on the MDS may result in inode
5408 data being lost in very rare circumstances.
5409 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5410 code in the 2.6 kernel caused the inode and/or EA data not to
5411 be read from disk, causing single-file corruption.
5415 Frequency : rare, only 2.6 filesystems using extents
5416 Description: Heavy concurrent write and delete load may cause data corruption.
5417 Details : It was possible under high-load situations to have an extent
5418 metadata block in the block device cache from a just-unlinked
5419 file overwrite a newly-allocated data block. We now unmap any
5420 metadata buffers that alias just-allocated data blocks.
5424 Frequency : filesystems with default stripe_count larger than 77
5425 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5426 Details : lconf specifies an inode size of 4096 bytes when the default
5427 stripe_count is larger than 77. This conflicts with the default
5428 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5429 this case to avoid pinning too much memory for large EAs.
5431 ------------------------------------------------------------------------------
5433 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5438 Frequency : rare (extremely heavy IO load with hundreds of clients)
5440 Description: Client is evicted, gets IO error writing to file
5441 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5442 caused clients to be evicted for AST timeouts. The fixes in
5443 bug 5192 mean we no longer need to have such short AST timeouts
5444 so ldlm_timeout has been increased.
5447 Frequency : occasional during --force or --failover shutdown under load
5448 Bugzilla : 5949, 4834
5449 Description: Server oops/LBUG if stopped with --force or --failover under load
5450 Details : a collection of import/export refcount and cleanup ordering
5451 issues fixed for safer force cleanup
5454 Frequency : only filesystems larger than 120 OSTs
5455 Bugzilla : 5990, 6223
5456 Description: lfs getstripe would oops on a very large filesystem
5457 Details : lov_getconfig used kfree on vmalloc'd memory
5460 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5461 Bugzilla : 6242, 6243
5462 Description: reading from files that had been truncated to a non-zero size
5463 but never opened returned no data
5464 Details : ll_file_read() reads zeros from no-object files to EOF
5469 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5470 Details : There's little harm in aborting MDS/OSS recovery and letting it
5471 try again, so I removed the LASSERT and return an error instead.
5473 Severity : enhancement
5475 Description: New debugging infrastructure for tracking down data corruption
5476 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5477 (b) cover more of the client-side code path, and (c) try to narrow
5478 down where problems occurred
5482 Bugzilla : 3819, 4364, 4397, 6313
5483 Description: Racing close and eviction MDS could cause assertion in mds_close
5484 Details : It was possible to get multiple mfd references during close and
5485 client eviction, leading to one thread referencing a freed mfd.
5487 Severity: : enhancement
5488 Bugzilla : 3262, 6359
5489 Description: Attempts to reconnect to servers are now more aggressive.
5490 Details : This builds on the enhanced upcall-less recovery that was added
5491 in 1.4.2. When trying to reconnect to servers, clients will
5492 now try each server in the failover group every 10 seconds. By
5493 default, clients would previously try one server every 25 seconds.
5498 Description: After recovery, certain operations trigger a failed
5499 assertion on a client.
5500 Details : Failing over an mds, using lconf -d --failover, while a
5501 client was doing a readdir() call would cause the client to
5502 LBUG after recovery completed and the readdir() was resent.
5504 Severity : enhancement
5506 Description: Default groups are now added by lconf
5507 Details : You can now run lconf --group <servicename> without having to
5508 manually add groups with lmc.
5511 Frequency : occasional
5513 Description: Nodes with an elan id of 0 trigger a failed assertion
5516 Frequency : always when accessing e.g. tty/console device nodes
5518 Description: tty and some other devices nodes cannot be used on lustre
5519 Details : file's private_data field is used by device data and lustre
5520 values in there got lost. New field was added to struct file to
5521 store fs-specific private data.
5524 Frequency : when exporting Lustre via NFS
5526 Description: NFSD failed occasionally when looking up a path component
5527 Details : NFSD is looking up ".." which was broken in ext3 directories
5528 that had grown large enough to become hashed.
5531 Frequency : Clusters with multiple interfaces not on the same subnet
5533 Description: Nodes will repeatedly try to reconnect to an interface which it
5534 cannot reach and report an error to the log.
5535 Details : Extra peer list entries will be created by lconf with some peers
5536 unreachable. lconf now validates the peer before adding it.
5539 Frequency : Only if a default stripe is set on the filesystem root.
5541 Description: Setting a default stripe on the filesystem root prevented the
5542 filesystem from being remounted.
5543 Details : The client was sending extra request flags in the root getattr
5544 request and did not allocate a reply buffer for the dir EA.
5547 Frequency : occasional, higher if lots of files are accessed by one client
5548 Bugzilla : 6159, 6097
5549 Description: Client trips assertion regarding lsm mismatch/magic
5550 Details : While revalidating inodes the VFS looks up inodes with ifind()
5551 and in rare cases can find an inode that is being freed.
5552 The ll_test_inode() code will free the lsm during ifind()
5553 when it finds an existing inode and then the VFS later attaches
5554 this free lsm to a new inode.
5558 Bugzilla : 6422, 7030
5559 Description: MDS deadlock between mkdir and client eviction
5560 Details : Creating a new file via mkdir or mknod (starting a transaction
5561 and getting the ns lock) can deadlock with client eviction
5562 (gets ns lock and trying to finish a synchronous transaction).
5565 Frequency : occasional
5566 Description: While starting a server, the fsfilt_ext3 module could not be
5568 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5569 kernels. Previously, lconf would still use the ext3 name
5570 when trying to load modules. Now, it will correctly use
5571 ext3 on 2.4 and ldiskfs on 2.6.
5573 Severity : enhancement
5574 Description: The default stripe count has been changed to 1
5575 Details : The interpretation of the default stripe count (0, to lfs
5576 or lmc) has been changed to mean striping across a single
5577 OST, rather than all available. For general usage we have
5578 found a stripe count of 1 or 2 works best.
5580 Severity : enhancement
5581 Description: Add support for compiling against Cray portals.
5582 Details : Conditional compiling for some areas that are different
5586 Frequency : occasional
5587 Bugzilla : 6409, 6834
5588 Description: Creating files with an explicit stripe count may lead to
5589 a failed assertion on the MDS
5590 Details : If some OSTs are full or unavailable, creating files may
5591 trigger a failed assertion on the MDS. Now, Lustre will
5592 try to use other servers or return an error to the
5596 Frequency : occasional
5598 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5599 caused lock timeout during readahead (since 1.4.2).
5600 Details : Processes doing readahead might match a lock that hasn't been
5601 granted yet if there are overlapping and conflicting lock
5602 requests. The readahead process waits on ungranted lock
5603 (original lock is CBPENDING), while OST waits for that process
5604 to cancel CBPENDING read lock and eventually evicts client.
5606 Severity : enhancement
5608 Description: Initial enabling of flock support for clients
5609 Details : Implements fcntl advisory locking and file status functions.
5610 This feature is provided as an optional mount flag (default
5611 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5612 locking are implemented yet, and those that are are not guaranteed
5613 to be completely correct in production environments.
5614 mount -t lustre -o [flock|noflock] ...
5617 Frequency : occasional
5619 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5620 assertion in the ext3 JBD (journaling) layer.
5621 Details : The b_committed_data struct is protected by the big kernel lock
5622 in 2.4 kernels, serializing journal_commit_transaction() and
5623 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5624 access to this struct. In 2.6 kernels there is finer grained
5625 locking to improve SMP performance of the JBD layer.
5629 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5634 Description: merge and cleanup kernel patches.
5637 -----------------------------------------------------------------------------
5639 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5641 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5642 versions of Lustre. Please update all servers and clients to
5643 version 1.4.2 or later at the same time. You must also run
5644 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5645 to update the configuration logs.
5647 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5648 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5649 - swab llog records of type '0' so we get proper header size/idx (5861)
5650 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5651 - fix rename of one directory over another leaking an inode (5953)
5652 - avoid SetPageDirty on 2.6 (5981)
5653 - don't re-add just-being-destroyed locks to the waiting list (5653)
5654 - when creating new directories, inherit the parent's custom
5655 striping settings if present parent (3048)
5656 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5657 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5658 - don't hold client locks on temporary worklist from l_lru (5666)
5659 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5660 - automatically evict dead clients (5921)
5661 - Update file size properly in create+truncate+fstat case (6196)
5662 - Do not unhash mountpoint dentries, do not allow removal of
5664 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5665 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5666 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5667 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5668 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5669 - Correctly handle reads of files with no objects (6243)
5670 - lctl recover will also mark a device active if deactivate used (5933)
5672 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5673 - allow --write-conf on an MDS with different nettype than client (5619)
5674 - don't write config llogs to MDS for mounts not from that MDS (5617)
5675 - lconf should create multiple TCP connections from a client (5201)
5676 - init scripts are now turned off by default; run chkconfig --on
5677 lustre and chkconfig --on lustrefs to use them
5678 - upcalls are no longer needed for clients to recover to failover
5680 - add --abort-recovery option to lconf to abort recovery on device
5682 - add support for an arbitrary number of OSTs (3026)
5683 - Quota support protocol changes.
5684 - forward compatibility changes to wire structs (6007)
5685 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5686 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5687 - improved "lustre status" script
5688 - initialize blocksize for non-regular files (6062)
5689 - added --disable-server and --disable-client configure options (5782)
5690 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5691 - Vanilla 2.4.29 support
5692 - increase maximum number of obd devices to 520 (6242)
5693 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5694 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5695 SLES 9 2.6.5 kernels
5696 - limit stripes per file to 160 (the maximum EA size) (6093)
5698 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5701 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5702 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5703 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5704 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5705 - small ext3 extents cleanups and fixes (5733)
5706 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5707 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5708 - avoid needless client->OST connect, fix handle mismatch (5317)
5709 - fix DLM error path that led to out-of-sync client, long delays (5779)
5710 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5711 - fix several locking issues related to i_size (5492,5624,5654,5672)
5712 - don't move pending lock onto export if it is already evicted (5683)
5713 - fix kernel oops when creating .foo in unlinked directory (5548)
5714 - fix deadlock in obdfilter statistics vs. object create (5811)
5715 - use time_{before,after} to avoid timer jiffies wrap (5882)
5716 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5717 - Do not leak request if server was not able to process it (5154)
5718 - If mds_open unable to find parent dir, make that negative lookup(5154)
5719 - don't create new directories with extent-mapping (5909, 5936)
5721 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5722 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5723 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5724 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5725 - support for RHEL4 kernel on i686 (5773)
5726 - provide error messages when incompatible logs are encountered (5898)
5728 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5729 * version 1.4.0.10 (1.4.1 release candidate 1)
5731 - don't keep a lock reference when lock is not granted (4238)
5732 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5733 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5734 - drop import inflight refcount on signal_completed_replay error (5255)
5735 - unlock page after async write error during send (3677)
5736 - handle missing objects in filter_preprw_read properly (5265)
5737 - no transno return for symlink open, don't save no-trasno open (3440)
5738 - don't try to complete elan receive that already failed (4012)
5739 - free RPC server reply state on error (5406)
5740 - clean up thread from ptlrpc_start_thread() on error (5160)
5741 - readahead could read extra page into cache that wasn't ejected (5388)
5742 - prevent races in class_attach/setup/cleanup/detach (5260)
5743 - don't dereference de->d_inode after l_dput of de (5458)
5744 - use "int" for stripe value returned from lock_to_stripe (5544)
5745 - mballoc allocation and error-checking fixes in 2.6 (5504)
5746 - block device patches to fix I/O request sizes in 2.6 (5482)
5747 - look up hostnames for IB nals (5602)
5748 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5749 - don't start multiple acceptors for the same port (5277)
5750 - fix incorrect LASSERT in mds_getattr_name (5635)
5751 - export a proc file for general "ping" checking (5628)
5752 - fix "lfs check" to not block when the MDS is down (5628)
5754 - service request history (4965)
5755 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5756 - create an "evict_client" /proc entry on OSTs, like the MDS has
5757 - fix mount usage message, return errors per mount(8) (5168)
5758 - change grep [] to grep "[]" in tests so they work in more UMLs
5759 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5760 - remove ancient LOV_MAGIC_V0 EA support (5047)
5761 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5762 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5763 - allow readahead allocations to fail when low on memory (5383)
5764 - mmap locking landed again, after considerable improvement (2828)
5765 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5767 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5770 - send OST transaction number in read/write reply to free req (4966)
5771 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5772 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5773 - the watchdog thread now runs as interruptible (5246)
5774 - flock/lockf fixes (but it's still disabled, pending 5135)
5775 - don't use EXT3 constants in llite code (5094)
5776 - memory shortage at startup could cause assertion (5176)
5778 - reorganization of lov code
5779 - single portals codebase
5781 - add extents/mballoc support (5025)
5782 - direct I/O reads in the obdfilter (4048)
5783 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5785 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5788 - send OST transaction number in read/write reply to free req (4966)
5789 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5790 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5791 - the watchdog thread now runs as interruptible (5246)
5792 - handle missing objects in filter_preprw_read properly (5265)
5793 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5794 - drop import inflight refcount on signal_completed_replay error (5255)
5795 - unlock page after async write error during send (3677)
5796 - return original error code on reconstructed replies (3761)
5797 - no transno return for symlink open, don't save no-trasno open (3440)
5799 - add pid to ldlm debugging output (4922)
5800 - bump the watchdog timeouts -- we can't handle 30sec yet
5801 - extra debugging for orphan dentry/inode bug (5259)
5803 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5806 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5807 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5808 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5809 - fix NULL dereference in /proc/sys/portals/routes (4827)
5810 - allow failed mdc_close() operations to be interrupted (4561)
5811 - stop precreate on OST before MDS would time out on it (4778)
5812 - don't send partial-page writes before EOF from client (4410)
5813 - discard client grant for sub-page writes on large-page clients (4520)
5814 - don't free dentries not owned by NFS code, check generation (4806)
5815 - fix lsm leak if mds_create_objects() fails (4801)
5816 - limit debug_daemon file size, always print CERROR messages (4789)
5817 - use transno after validating reply (3892)
5818 - process timed out requests if import state changes (3754)
5819 - update mtime on OST during writes, return in glimpse (4829)
5820 - add mkfsoptions to LDAP (4679)
5821 - use ->max_readahead method instead of zapping global ra (5039)
5822 - don't interrupt __l_wait_event() during strace
5824 - add software watchdogs to catch hung threads quickly (4941)
5825 - make lustrefs init script start after nfs is mounted
5826 - fix CWARN/ERROR duplication (4930)
5827 - return async write errors to application if possible (2248)
5828 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5829 - print NAL number in %x format (4645)
5830 - update barely-supported suse-2.4.21-171 series (4842)
5831 - support for sles 9 %post scripts
5832 - support for building 2.6 kernel-source packages
5833 - support for sles km_* packages
5835 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5838 - ignore -ENOENT errors in osc_destroy (3639)
5839 - notify osc create thread that OSC is being cleaned up (4600)
5840 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5841 - reconstruct ost_handle() like mds_handle() (4657)
5842 - create a new thread to do import eviction to avoid deadlock (3969)
5843 - let lconf resolve symlinked-to devices (4629)
5844 - don't unlink "objects" from directory with default EA (4554)
5845 - hold socknal file ref over connect in case target is down (4394)
5846 - allow more than 32000 subdirectories in a single directory (3244)
5847 - fix blocks count for O_DIRECT writes (3751)
5848 - OST returns ENOSPC from object create when no space left (4539)
5849 - don't send truncate RPC if file size isn't changing (4410)
5850 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5851 - bind to privileged port in socknal and tcpnal (3689)
5853 - rate limit CERROR/CWARN console message to avoid overload (4519)
5854 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5855 - basic mmap support (3918)
5856 - kernel patch series update from b1_4 (4711)
5858 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5861 - avoid crash during MDS cleanup with OST shut down (2775)
5862 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5863 - don't use bad inodes on the MDS (3744)
5864 - dynamic object preallocation to improve recovery speed (4236)
5865 - don't hold spinlock over lock dumping or change debug flags (4401)
5866 - don't zero obd_dev when it is force cleaned (3651)
5867 - print grants to console if they go negative (4431)
5868 - "lctl deactivate" will stop automatic recovery attempts (3406)
5869 - look for existing locks in ldlm_handle_enqueue() (3764)
5870 - don't resolve lock handle twice in recovery avoiding race (4401)
5871 - revalidate should check working dir is a directory (4134)
5873 - don't always mark "slow" obdfilter messages as errors (4418)
5875 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5878 - don't close LustreDB during write_conf until it is done (3860)
5879 - fix typo in lconf for_each_profile (3821)
5880 - allow dumping logs from multiple threads at one time (3820)
5881 - don't allow multiple threads in OSC recovery (3812)
5882 - fix debug_size parameters (3864)
5883 - fix mds_postrecov to initialize import for llog ctxt (3121)
5884 - replace config semaphore with spinlock (3306)
5885 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5886 - don't allow enqueue to complete on a destroyed export (3822)
5887 - down write_lock before checking llog header bitmap (3825)
5888 - recover from lock replay timeout (3764)
5889 - up llog sem before sending rpc (3652)
5890 - reduce ns lock hold times when setting kms (3267)
5891 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5892 - fix NULL deref and obd_dev leak on setup error (3312)
5893 - replace some LBUG about llog ops with error handling (3841)
5894 - don't match INVALID dentries from d_lookup and spin (3784)
5895 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5896 - fix invalid assertion in ptlrpc_set_wait (3880)
5898 - add libwrap support for the TCP acceptor (3996)
5899 - add /proc/sys/portals/routes for non-root route listing (3994)
5900 - allow setting MDS UUID in .xml (2580)
5901 - print the stack of a process that LBUGs (4228)
5903 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5906 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5907 - make sure to unset replay flag from failed open requests (3440)
5908 - if default stripe count is 0, use OST count for inode size (3636)
5909 - update parent mtime/ctime on client for create/unlink (2611)
5910 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5911 - free recovery state on server during a forced cleanup (3571)
5912 - unregister_reply for resent reqs (3063)
5913 - loop back devices mounting and status check on 2.6 (3563)
5914 - fix resource-creation race that can provoke i_size == 0 (3513)
5915 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5916 - more debugging for page-accounting assertion (3746)
5917 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5918 - avoid deadlock after precreation failure (3758)
5919 - fix race and lock order deadlock in orphan handling (3450, 3750)
5920 - add validity checks when grabbing inodes from l_ast_data (3599)
5922 - add /proc/.../recovery_status to obdfilter (3428)
5923 - lightweight CDEBUG infrastructure, debug daemon (3668)
5924 - change default OSC RPC parameters to be better on small clusters
5925 - turn off OST read cache for files smaller than 32MB
5926 - install man pages and include them in rpms (3100)
5927 - add new init script for (un)mounting lustre filesystems (2593)
5928 - run chkconfig in %post for init scripts (3701)
5929 - drop scimac NAL (unmaintained)
5931 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5934 - clean kiobufs before and after use (3485)
5935 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5936 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5937 - change init script to fail more clearly if not run as root (1528)
5938 - allow clients to reconnect during replay (1742)
5939 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5940 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5941 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5942 - fix cond_resched (3554)
5943 - only evict unfinished clients after recovery (3515)
5944 - allow bulk resend, prevent data loss (3570)
5945 - dynamic ptlrpc request buffer allocation (2102)
5946 - don't allow unlinking open directory if it isn't empty (2904)
5947 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5948 - remove extraneous obd dereference causing LASSERT failure (3334)
5949 - don't use get_cycles() when creating temp. files on the mds (3156)
5950 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5951 - handle EEXIST for set-stripe, set proper directory name (3336)
5953 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5954 - fix ksocknal_fmb_callback() error messages (2918)
5956 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5959 - don't copy lvb into (possibly NULL) reply on error (2983)
5960 - don't deref dentry after dput, don't free lvb on error (2922)
5961 - use the kms to determine writeback rpc length (2947)
5962 - increment oti_logcookies when osc is inactive (2948)
5963 - update client's i_blocks count via lvb messages (2543)
5964 - handle intent open/close of special files properly (1557)
5965 - mount MDS with errors=remount-ro, like obdfilter (2009)
5966 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5967 - don't use cancelling-locks' kms values (2947)
5968 - use highest lock extent for kms, not last one (2925)
5969 - don't dereference ERR_PTR() dentry in error handling path (3107)
5970 - fix thread race in portals_debug_dumplog() (3122)
5971 - create lprocfs device entries at setup instead of at attach (1519)
5972 - common AST error handler, don't evict client on completion race (3145)
5973 - zero nameidata in detach_mnt in 2.6 (3118)
5974 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5975 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5976 - initialize RPC timeout timer earlier for 2.6 (3219)
5977 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5978 - print nal/nid for unknown nid (3258)
5979 - additional checks for oscc recovery before doing precreate (3284)
5980 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5981 - don't crash in mdc_close for bad permissions on open (3285)
5982 - zero i_rdev for non-device files (3147)
5983 - clear page->private before handing to FS, better assertion (3119)
5984 - tune the read pipeline (3236)
5985 - fix incorrect decref of invalidated dentry (2350)
5986 - provide read-ahead stats and refine rpc in flight stats (3328)
5987 - don't hold journal transaction open across create RPC (3313)
5988 - update atime on MDS at close time (3265)
5989 - close LDAP connection when recovering to avoid server load (3315)
5990 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5991 - don't leak open file on MDS after open resend (3325)
5992 - serialize filter_precreate and filter_destroy_precreated (3329)
5993 - loop device shouldn't call sync_dev() for nul device (3092)
5994 - clear page cache after eviction (2766)
5995 - resynchronize MDS->OST in background (2824)
5996 - refuse to mount the same filesystem twice on same mountpoint (3394)
5997 - allow llmount to create routes for mounting behind routers (3320)
5998 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5999 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6000 - fix rare problem with rename on htree directories (3417)
6002 - allow default OST striping configuration per directory (1414)
6003 - fix compilation for qswnal for 2.6 kernels (3125)
6004 - increase maximum number of MDS request buffers for large systems
6005 - change liblustreapi to be useful for external progs like lfsck (3098)
6006 - increase local configuration timeout for slow disks (3353)
6007 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6009 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6012 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6013 - don't overwrite extent policy data in reply if lock was blocked (2901)
6014 - drop filter export grants atomically with removal from device (2663)
6015 - del obd_self_export from work_list in class_disconnect_exports (2908)
6016 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6017 - swab reply message in mdc_close, other PPC fixes (2464)
6018 - fix destroying of named logs (2325)
6019 - overwrite old logs when running lconf --write_conf (2264)
6020 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6021 - fix race in target_handle_connect (2898)
6022 - mds_reint_create() should take same inode create lock (2926)
6023 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6024 - don't close files for self_export to avoid uninitialized obd (2936)
6025 - allow MDS with the same name as client node (2939)
6026 - hold dentry reference for closed log files for unlink (2325)
6027 - reserve space for all logs during transactions (2059)
6028 - don't evict page beyond end of stripe extent (2925)
6029 - don't oops on a deleted current working directory (2399)
6030 - handle hard links to targets without a parent properly (2517)
6031 - don't dereference NULL lock when racing during eviction (2867)
6032 - don't grow lock extents when lots of conflicting locks (2919)
6034 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6037 - account for cache space usage on clients to avoid data loss (974)
6038 - lfsck support in lustre kernel code (2349)
6039 - reduce journal credits needed for BRW writes (2370)
6040 - orphan handling to avoid losing space on client/server crashes
6041 - ptlrpcd can be blocked, stopping ALL progress (2477)
6042 - use lock value blocks to assist in proper KMS, faster stat (1021)
6043 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6044 - recovery for initial connections (2355)
6045 - fixes for mds_cleanup_orphans (1934)
6046 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6047 - block all file creations until orphan recovery completes (1901)
6048 - client remove rq_connection from request struct (2423)
6049 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6050 - recovery timer race (2670)
6051 - mdc_close recovey bug (2532)
6052 - ptlrpc cleanup bug (2710)
6053 - mds timeout on local locks (2588)
6054 - namespace lock held during RPCs (2431)
6055 - handle interrupted sync write properly (2503)
6056 - don't try to handle a message that hasn't been replied to (2699)
6057 - client assert failure during cleanup after abort recovery (2701)
6058 - leak mdc device after failed mount (2712)
6059 - ptlrpc_check_set allows timedout requests to complete (2714)
6060 - wait for inflight reqs when ptlrpcd finishes (2710)
6061 - make sure unregistered services are removed from the srv_list
6062 - reset bulk XID's when resending them (caught by 1138 test)
6063 - unregister_bulk after timeout
6064 - fix lconf error (2694)
6065 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6066 - readahead locks pages, leaves pending causing memory pressure (2673)
6067 - increase OST request buffers to 4096 on large machines (2729)
6068 - fix up permission of existing directories in simple_mkdir (2661)
6069 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6070 - don't assign transno to errored transactions (2742)
6071 - don't delete objects on OST if given a bogus objid from MDS (2751)
6072 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6073 - if rq_no_resend, then timeout request after recovery (2432)
6074 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6075 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6076 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6077 - MDS recovery shouldn't delete objects that it has given out (2730)
6078 - if enqueue arrives after completion, don't clobber LVB (2819)
6079 - don't unlock pages twice when trigger_group_io returns error (2814)
6080 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6081 - don't write pages to disk if there was an error (1450)
6082 - don't ping imports that have recovery disabled (2676)
6083 - take buffered bytes into account when balancing socknal conn (2817)
6084 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6085 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6086 - remove little-endian swabbing of llog records (1987)
6087 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6088 - retry reposting request buffers if they fail (1191)
6089 - grow extent at grant time to avoid granting a revoked lock (2809)
6090 - lock revoke doesn't evict page if covered by a second lock (2765)
6091 - disable VM readahead to avoid reading outside lock extents (2805)
6093 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6094 - updated kernel patches for hp-2.4.20 kernel (2681)
6096 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6099 - fix truncated write corruption (2366)
6100 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6102 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6103 - ensure len(uuid) < 37 in lmc (1171)
6104 - fix ia64 OOPS in llog_test (2255)
6105 - zero end of page at obdfilter for partial page writes (2648)
6106 - don't leave stale dentries around after renames (bug 2428)
6107 - fix timeouts when evicting a client with a single lock held (2642)
6108 - set deadline for the initial HELLO message to drain (2634)
6109 - print out dotted-quad IP addresses in the socknal (2302)
6111 - additional debugging for MDS client eviction problem (2443)
6112 - fix mkfsoptions support for osts (2603, 2604)
6114 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6117 - add series for the vanilla 2.6.0 kernel
6118 - add series for the vanilla 2.4.24 kernel
6119 - add series for a cray x86/64 UL kernel drop
6120 - fix xattr patches for the vanilla 2.4.19 series
6122 - generate true UUIDs in lmc (1171)
6123 - have portals stack dumping break in UML (2466)
6124 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6125 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6126 - wake callback waiting threads on client eviction (2460)
6127 - Add --ptldebug and --subsystem to lmc (1719)
6128 - update assertion to allow safe interrupt allocation
6129 - set rq_no_resend for cancel requests (2432)
6130 - recalculate ptlrpcd timeout after resend (2494)
6131 - call vfs_rmdir when removing pending directories (2368)
6132 - fix renaming a file to itself (2429)
6133 - lmc creates a default one-stripe lov (2454)
6134 - expand procfs space to handle large clusters (2326)
6135 - increase UML stack to avoid overflow
6136 - update lconf's list of debug and subsystem masks
6137 - fix lfs find --obd (2510)
6138 - /proc tunable for disabling filter read caching (2591)
6139 - stop rpm packages from altering slapd.conf (2301)
6140 - disable nagle in the socknal under 0conf (2578)
6141 - choose mds inode size based on stripe count (2572)
6142 - fix kernel-source rpm problems (2516)
6144 - add --disable-doc to avoid pdf generation (2421)
6145 - update documentation, tests, type-os, comments
6146 - avoid format warnings on ia64
6147 - remove the TOE NAL
6148 - tiny code cleanups by removing unused fields
6150 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6153 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6154 - remove the most blatant lies from BUILDING file (2371)
6155 - change default debug level to reasonable production setting
6156 - reduce client side cache size to reduce cache flush time
6157 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6158 - make TCP zerocopy and pinger support enabled by default (2476)
6159 - sync writes completed after process exits caused crashes (2319)
6160 - maintain correct mount count on the MDS (2356)
6161 - backout 1557, because 2316 wasn't really fixed
6162 - better file I/O statistics gathering in /proc
6163 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6164 - another kernel patch to fix zero-copy TCP function export
6165 - don't take duplicate lock when processing re-sent getattr (2420)
6166 - lctl uses obd_self_export instead of creating new conn (2353)
6167 - MDS/OST recovery case which requires object creation asserted (2425)
6168 - move lfs from /usr/sbin to /usr/bin in packages
6169 - fix race between mds_client_add and mds_client_free (2417)
6170 - use kmalloc instead of slabs in portals (2430)
6171 - don't create duplicate records when a failover MDS is present (2442)
6172 - remove unnecessary mount age check (2332)
6173 - don't remove directory inodes from locks prematurely (2451)
6174 - don't break if MDS service name is the same as hostname (2103)
6175 - fix races in client write RPC generation when cache full (2482)
6177 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6180 - remove now-unused request->rq_obd (278)
6181 - if an allocation fails, print out how much memory we've used (1933)
6182 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6183 - add the "configurable stack size" patch to most series files (1256)
6184 - ability to write large log records, for 100+ OST configs (2306)
6185 - fix NULL deref when filter_prep fails (2314)
6186 - fix operator precedence error in filter_sync
6187 - dynamic allocation of socknal TX descriptors (2315)
6188 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6189 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6190 - ensure that CWARN messages in Portals always get to the syslog
6191 - __init/__exit are not for prototype decls (ldlm_init/exit)
6192 - x86-64 compile warning fixes
6193 - fix gateway LMC keyword conflict (2318)
6194 - fix MDS lock inversions in getattr/reint paths (1844)
6195 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6196 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6197 - DLM race condition prevented some lock evictions (2328)
6198 - ENOMEM detection and retry on socknal sends (2230)
6199 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6200 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6201 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6202 - better error messages when a client is rejected during recovery (1505)
6203 - avoid cancelling locks which were never granted, after failure (2330)
6204 - fix i_sem/journal inversion in mds_client_add (2333)
6205 - fix truncate/getattr lock cycle deadlock (2334)
6206 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6207 - fix two rare exit paths which could leak an l_lock() ref (2321)
6208 - fencepost error in MDS/OST orphan recovery (2226)
6209 - make log record alignment 8 bytes (1988)
6210 - lstripe now fails when requested offset > ost_count (2237)
6211 - ensure that all kernel series have a complete list.h (1607)
6212 - fix crashes in special-file operations (2316)
6213 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6215 - allow configurable automake binary, for testing new versions
6216 - small update to the lfs documentation
6218 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6220 * fix negative export reference count in fsfilt_sync (2312)
6222 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6223 * release candidate 0.9.1
6225 - orphans are moved into the PENDING directory for possible recovery
6226 - replayed opens now open by fid for orphan/rename safety (1042)
6227 - last close of an orphan inode generates a transno (683)
6228 - chdir() and mount() now pin the directory entry (1020)
6229 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6230 - discard very old requests without processing them (1502)
6231 - remove some common, well-understood CERRORs (1505)
6232 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6233 - clear "grant" flags in OST replies until OST grant code lands (1644)
6234 - fix read performance by not clobbering i_blksize on client (1598)
6235 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6236 - make LRU size a /proc tunable, clears locks when reduced (707)
6237 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6238 - server locks take references on exports now (1558)
6239 - build fixes for 2.4.20-rh trees (1663)
6240 - return an error from lov_create if all OSCs are inactive (1751)
6241 - fix import levels when a reconnect happens without a timeout (1597)
6242 - exit early from mds_open if we get a lookup error (1749)
6243 - partial page read at EOF wouldn't wait for disk before sending (1642)
6244 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6245 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6246 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6247 - write/truncate lock inversion (1639)
6248 - Don't auto-load obdclass, portals modules during cleanup (1495)
6249 - fix timestamps from jumping to "now" (1763)
6250 - extra journal assertions (1648)
6251 - add an extra multiunlink test (1771)
6252 - fix read_record/write_record API (1776)
6253 - fix leak of offset_extent, possible incorrect i_size later (1772)
6254 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6255 - Debugging for the kqswnal_get_idle_tx problems (1820)
6256 - Allow recovery to be attempted multiple times (1536)
6257 - Write out MDS last_rcvd file after it is first created (1600)
6258 - Fix tx_descriptor leak in failed transmit situations (1827)
6259 - ext3 journaling fixes for assertion failure after IO error (1871)
6260 - class_export_put() on freed export after completion AST error (1896)
6261 - Fix revalidate looping in VFS (1322)
6262 - Don't access a freed export during MDS_REINT timeout (1521)
6263 - Add open-unlink recovery support on the MDS (1673,1764)
6264 - Return an error if no MDS data was read from last_rcvd (1946)
6265 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6266 - Don't setup a disk device that doesn't match exported UUID (317)
6267 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6268 - avoid committing NULL handle in force close
6269 - local.sh is now a one-stripe LOV configuration
6270 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6271 - fix ext3 htree duplicate directory entry corruption (1516)
6272 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6273 - update to new LOV EA format (2097)
6274 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6276 2003-06-15 Phil Schwan <phil@clusterfs.com>
6279 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6280 - per-import/export recovery handling (958, 931, 959)
6281 - multiple last-rcvd slots, for serving multiple FSes (949)
6282 - connections are again shared between multiple imp/exports (963, 964)
6283 - "umount -f" would hang if any requests needed to be sent (393, 978)
6284 - avoid pinning large req buffer by copying for queued messages (989)
6285 - add "uuid" to "lctl device" command to help upcalls (991)
6286 - "open" RPCs with transnos would confuse recovery counters (1037)
6287 - do proper endian conversion of all wire messages (288, 340, 891)
6288 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6289 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6290 - fix ptlrpc_connection leak in target_handle_connect (1174)
6291 - fix import refcounting bug in OST and MDS cleanup (1134)
6292 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6293 - fix very unlikely obd_types race condition (501)
6294 - remove osc_open hack for echo_client (1187)
6295 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6296 - a failure in read_inode2 leads to deadlock (1139)
6297 - cancel ack-locks as soon as transaction is committed (1072)
6298 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6299 - make sure to commitrw after any preprw to avoid deadlock (1162)
6300 - failing to execute a file in a lustre FS would lock inode (1203)
6301 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6302 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6303 - obd_uuid2tgt would walk off the end of the list (1255)
6304 - on IA64 the osc would give portals incorrect bulk size (1258)
6305 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6306 - fix lock inversion caused by new llite matching code (1282)
6307 - limit the number of dirty pages on a client to 10MB (1286)
6308 - timed out locks were not being corrected cancelled (1289)
6309 - fix O_DIRECT above 4GB on IA-32 (1292)
6310 * major user-visible changes
6311 - fail out/fail over policy now controlled by the upcall (993)
6313 - add OBD_PING to check server availability and failure (954)
6314 - lustre messages are now sent in sending host order (288, 340, 891)
6315 - add eadatalen to MDS getattr reply (340)
6316 - OST read replies may contain second buffer, with per-page status (593)
6318 2003-03-11 Phil Schwan <phil@clusterfs.com>
6321 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6322 - failed open()s could cause deadlock; fixed (867, 869)
6323 - stop cancelling OST locks when files are closed (481)
6324 - overlapping XID spaces caused network corruption (851, 853)
6325 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6326 - setattr_raw added, to do single-RPC, server-side setattrs
6327 - lmc/lconf syntax change for OST UUIDs
6328 - fix crashy race condition between ptlrpc_free_req and osc_close
6329 - don't use request in mdc_enqueue if we hit a timeout (889)
6330 - don't set the inode i_size for regular files from the MDS (896)
6331 - handle out of order completion AST (842)
6332 - don't LBUG if a lock request times out after receiving AST (913)
6333 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6334 - if a bad lock AST arrives, send an error instead of dropping entirely
6335 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6336 - fix leak in bulk IO when only partially completed (899, 900, 926)
6337 - fix O_DIRECT for ia64 (55)
6338 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6339 - C-z after timeout could hang a process forever; fixed (977)
6341 - client-side I/O cache (678, 924, 929, 941, 970)
6343 - READPAGE and SETATTRs which don't take server-side locks get
6346 2003-02-11 Phil Schwan <phil@clusterfs.com>
6349 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6350 - Fix fsfilt_extN_readpage() to read a full page of directory
6351 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6352 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6353 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6354 - Most intent operations were removed, in favour of a new RPC mode
6355 that does a single RPC to the server and bypasses most of the VFS
6356 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6357 - Aggressively cancel local locks on DLM servers
6358 - mds_reint_unlink sends EA to the client if it's the last nlink.
6359 client uses that EA to unlink OST objects.
6360 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6361 - recursive symlinks were fixed (439)
6362 - fixed NULL deref in DEBUG_REQ
6363 - filter_update_lastobjid no longer calls sync, which annoyed extN
6364 - fixed multi-client small-writes to a single file problem (445)
6365 - fixed mtime updates during file writes (607)
6366 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6367 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6368 - fixed Linux OST size reporting problem (444, 656)
6369 - OST now updates object mtime with writes or setattr (607, 619)
6370 - client verifies file size before zeroing page past EOF (445)
6371 - OST now writes last allocated objid to disk with allocation (108)
6372 - LOV on echo now works (409)
6374 - mds_reint_unlink sends a new buffer, with the EA included. this
6375 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6376 set if a regular file was being unlinked, and it was the last link
6377 - use PtlGet from the target for bulk writes (315)
6378 - OST now updates object mtime with writes or setattr (607, 619)
6379 - LDLM now has a grant-time callback to revalidate locked items, if
6381 - Many MDS operations were reorganized to combat race conditions
6383 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6384 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6386 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6389 - Fully reactivate OST imports after reconnection (512, others)
6390 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6391 - More graceful error handling for truncating on dead OST (515)
6392 - Don't error out unless we're actually accessing dead stripes (474)
6393 - Fix garbage sizes when stripes are missing (410)
6394 - LRU counters were broken, causing constant lock purge (433, 432)
6395 - garbage on read from stripes with failed OSTs (441)
6396 - mark OSCs as active before reconnecting during recovery (438)
6397 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6398 - lfind did not preserve OST order in output (443)
6399 - symlinks cause hung clients, incorrect data (439)
6400 - stop dereferencing request after dropping refcount (457)
6401 - don't LASSERT(spin_is_locked) on non-SMP (455)
6402 - fixes for many rename() bugs
6403 - fstat didn't correctly synchronize attributes (399)
6404 - server must handle lock cancellation during blocking AST prep (487)
6405 - bulk descriptors were free()d too soon (511)
6406 - fix paths in lconf, which would load incorrect modules (451, 507)
6407 - fix confusing lconf 'host not found' error message (386)
6408 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6409 - fix race condition in mdc_blocking_ast() for inode access (526)
6410 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6411 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6412 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6413 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6414 - fix lstripe in several ways (kernel side) (527)
6415 - fix request leak in ldlm_cli_enqueue (262)
6416 - incorrect OSC was marked inactive after OST failure
6417 - call mds_fs_cleanup before unmounting filesystem (524)
6418 - fix races between taking ns_lock and ldlm_lock_change_resource
6419 - fix races updating LOV export open file list
6420 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6421 - fix recovery NULL deref in ldlm_cli_cancel_unused
6422 - fix some DLM races by using new hash table for lock handles (419)
6423 - permit the client to specify desired inodes, at replay
6424 - duplicate requests when we queue them for replay reintegration
6425 - fix last_rcvd offset calculation
6426 - sync after each recovered transaction, so we always make progress
6427 - never, not always, ERESTART requests without transnos
6428 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6430 - skip replay if the MDS says that the client is already connected
6431 - don't check for a recovery-enabled export to match lctl's UUID
6432 - don't INC_USE_COUNT for phantom exports
6433 - don't crash when cleaning up phantom exports (567)
6434 - don't double-finish or set replay data for errored mdc_open requests
6435 - abort requests when they time out, so we don't get old replies
6436 - send/receive replies for AST messages again
6437 - if the client says that it doesn't have the lock, cancel it on the
6439 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6440 mark it as destroyed, it will all work out when decref is called
6441 - fix module use counts (22, 581)
6443 - ASTs now expect a reply (server cancels lock on error reply)
6445 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6448 - fix many simultaneous client startup (392)
6449 - fix dentry->d_it clobbering
6450 - credentials weren't being shipped for readdir/getattr operations
6451 - remove invalid assertions triggered during some concurrent MD
6453 - proper Lustre versions added (336, 389)
6454 - fix memory leak for create error case (398)
6455 - fix LOV locking bug that would get cli/srv out of sync
6456 - fix echo client over LOV (409)
6457 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6458 - fix double-O_EXCL intent crash (424)
6459 - avoid sending multiple lock CANCELs (352)
6461 - MDS can do multi-client recovery (modulo bugs in new code)
6463 - many updates, edits, cleanups
6465 2002-11-18 Phil Schwan <phil@clusterfs.com>
6468 - fix null d_it dereference (346)
6469 - fix full OST/dbench hang (333)
6470 - fix permission problem with file removal (286)
6471 - fix removal of OSCs from LOV when they fail
6472 - fix NULL deref during bulk timeout (214)
6473 - fix problems related to multiple filesystems on one MDS (241)
6474 - fixed serious subtle metadata locking bugs
6475 - free locks on clients when inodes are removed due to memory
6477 - fix inode pointer in lock data (285)
6478 - partial support for multiple MDS on a single host (241)
6479 - data locks weren't cancelled at clear_inode time (290, 311)
6480 - intent locks could lead to unbounded lock growth (205)
6481 - added a maximum lock count, an LRU list, and a flusher
6482 - fix multiple rename (365)
6483 - properly abstracted the echo client
6484 - OSC locked 1 byte too many; fixed
6485 - rewrote brw callback code:
6486 - fixed recovery bugs related to LOVs (306)
6487 - fixed too-many-pages-in-one-write crash (191)
6488 - fixed (again) crash in sync_io_timeout (214)
6489 - probably fixed callback-related race (385)
6491 - Add capability to MDS protocol
6492 - LDLM cancellations and callbacks on different portals
6494 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6497 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6498 - defer open object destruction to close time (601981, 138)
6499 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6500 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6501 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6502 - don't call obd_setattr in ll_file_release for destroyed objects
6503 * protocol change to lustre_msg: move |version| and add |flags|
6504 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6505 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6506 * added replay of create, unlink, link and rename operations during
6507 MDS failover; recovery should be much more robust now
6508 * remove failed OSCs from LOVs (only lov_create uses this so far)
6509 * the lustre-HOWTO was brought (more) up to date (582544)
6511 2002-10-23 Phil Schwan <phil@clusterfs.com>
6514 - in-use dentries weren't being reused properly (617851)
6515 - prevent multiple LDLM setup (599178)
6516 - fix LOV size calculations for truncate (617853)
6517 - fix client handling of MDS intent errors (POSIX)
6518 - fix permission bug in lovstripe.c test (624321)
6519 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6520 - truncate past end of file could corrupt data
6521 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6522 - a race in recovery could return ETIMEDOUT to apps (623947)
6523 - building outside the source directory was fixed
6524 * the lustre-HOWTO was brought (more) up to date (582544)
6525 * major progress was made on recovery functionality
6527 2002-10-10 Phil Schwan <phil@clusterfs.com>
6530 - recovery deadlock fix
6531 - rm -rf causes LBUG fix (617817)
6532 - file open by multiple tasks fix (618962)
6533 - directory permissions bugs (602707 and 620007)
6534 - journal_stop fixed with locking (611313)
6535 - O_APPEND failures resolved (618273, perhaps 614459)
6536 - lconf PATH fix (619770)
6537 - IA64 build fix (621450)
6538 - RPC buffer sizes scale with amount of memory
6540 2002-10-01 Phil Schwan <phil@clusterfs.com>
6543 - locks would be cancelled without throwing away data pages,
6544 resulting in inconsistent data (605627)
6545 - inode attributes were not always being refreshed (605627, 612449)
6546 - lconf now continues to cleanup after lctl reports an error
6547 - MDS now enforces user permissions (602707)
6548 - lprocfs cleanup fixed, but not yet enabled (614157)
6549 - fixed infinite server hang, should a client not respond to an AST
6550 - avoid going into recovery if user calls readlink() with a buffer
6551 that's too small (613941)
6552 - AST RPCs no longer require replies (614867) -- this may be changed
6553 - don't crash server if client sends an IOV that's too big (611336)
6554 - fixed lock conversion deadlock (611892)
6555 - fixed the following of symlinks (614622)
6556 * recovery: the server can remove locks from a client that dies, other
6557 clients can make progress
6558 * more extN patch fixes
6559 * compile-time configurable ptlrpc buffer allocations
6561 - collaborative read cache document
6562 - Lustre Lite Performance CDR document-in-progress
6564 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6567 - fix typo in patch-2.4.18
6569 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6572 - clear ptlrpc request each time in handle_incoming_request()
6573 - unlink of files now destroys the object on the OST
6575 2002-09-19 Peter Braam <braam@clusterfs.com>
6577 * add hard link support
6578 * change obdfile creation method
6579 * kernel patch changed
6581 2002-09-19 Peter Braam <braam@clusterfs.com>
6584 - stack overflow bug in extN fixed
6586 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6588 * documentation updates
6589 - add man pages for config tools
6590 - update tests/README to describe testing with new config tools
6591 - finish metadata API descriptions
6592 * bug fixes and cleanups
6593 - statfs workaround for 16TB limit
6594 - LOV stripe allocation improved, can stripe on subset of OSTs
6595 - LOV file size/IO offset was wrong for files > 4GB in size
6596 - object EA data was being dropped, caused files to be unreadable
6597 - memory overflow with non-LOV OST caused memory corruption
6598 - fixed regression tests to work with new config tools, obdfilter
6599 - fixed bug when directory size became larger than 1 block
6600 - fixed bug (for single client case) when PWD was deleted
6601 - invalidate local directory pages when doing intent-based ops
6602 - avoid LDLM oops when lock callback contained bad data
6604 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6606 * documentation updates
6607 * bug fixes and cleanups
6608 - configuration tools
6611 - 64-bit compile warnings
6612 - 64-bit internal statfs data
6614 * test_brw on persistent OST devices
6616 * lprocfs (disabled)
6618 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6620 * documentation updates
6621 * bug fixes and cleanups
6622 * configuration tools
6624 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6627 * many small fixes to 0_5_4
6628 * io/network handling
6629 * thinkos in MDS operations
6631 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6634 * crucial basic fixes to 0.5.3
6635 * IOR, Iozone work over Elan
6638 2002-08-07 Phil Schwan <phil@clusterfs.com>
6639 * version 0_5_3, our first alpha
6640 * we use the new Portals iovs
6641 * documentation updates
6642 * bug fixes and cleanups
6643 * small changes in the DLM wire protocol
6645 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6646 * version 0_5_1 with some initial stability,
6647 * locking on MD and file I/O.
6648 * documentation updates
6649 * several bug fixes since 0.5.0
6650 * small changes in wire protocol
6652 2002-07-18 Phil Schwan <phil@clusterfs.com>
6654 * delivered as Lustre Light Alpha
6655 * fixed a crash after handling invalid MDS requests
6656 * fixed directory pages for architectures with non-4k pages sizes
6658 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6659 * release version v0_4_4
6660 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6662 * Fixes a number of interruption problems with OST operations.
6663 * Update documentation for portals header changes
6664 * Move all wire protocol structs/defines to lustre_idl.h
6665 * Fixes symlink length bug.
6666 * Add tcpdump to repository.
6668 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6669 * release version v0_4_3
6670 * Fixes statfs for inodes on extN.
6671 * Fixes bug in runtests which would delete /etc/hosts.
6672 * Use 64-bit object IDs wherever possible (not into VFS though)
6673 Remove ost_get_info, which is unused by lustre, and out of date.
6675 2002-07-03 Peter Braam <braam@clusterfs.com>
6676 * release version v0_4_2 Fixes a lookup error (type not passed)
6677 * move forward to head of Portals
6678 * move forward to latest Lustre kernel
6680 2002-06-25 Peter Braam <braam@clusterfs.com>
6681 * release version v0_4_1. Hopefully stable on single node use.