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 3.10.0-862.14.4.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 4.4.155-94.50 (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 3.10.0-862.14.4.el7 (RHEL7.5)
28 4.14.0-49.13.1.el7a (RHEL7.5)
29 3.0.101-0.47.71 (SLES11 SP3)
30 3.0.101-107 (SLES11 SP4)
31 3.12.74-60.64.40 (SLES12 SP1)
32 4.4.120-92.70 (SLES12 SP2)
33 4.4.155-94.50 (SLES12 SP3)
34 3.13.0-101 (Ubuntu 14.04)
35 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
36 4.4.0-131 (Ubuntu 16.04)
37 4.15.0-32 (Ubuntu 18.04)
39 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
40 * Recommended ZFS / SPL version: 0.7.9
41 * Tested with ZFS / SPL version: 0.7.9
42 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
43 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
44 stack overflow. For more information, please refer to bugzilla 17630.
45 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
46 the re-exporting nodes to be mounted with "32bitapi" mount option
47 * Note for Ubuntu 14.04.5:
48 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
49 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
51 --------------------------------------------------------------------------------
54 2018-04-03 Intel Corporation
56 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
57 for currently supported client and server kernel versions.
58 * Server known to build on patched kernels:
59 2.6.32-431.29.2.el6 (RHEL6.5)
60 2.6.32-504.30.3.el6 (RHEL6.6)
61 2.6.32-573.26.1.el6 (RHEL6.7)
62 2.6.32-642.15.1.el6 (RHEL6.8)
63 2.6.32-696.18.7.el6 (RHEL6.9)
64 3.10.0-693.21.1.el7 (RHEL7.4)
65 3.0.101-0.47.71 (SLES11 SP3)
66 3.0.101-107 (SLES11 SP4)
67 3.12.74-60.64.40 (SLES12 SP1)
68 4.4.114-92.64 (SLES12 SP2)
69 4.4.114-94.11 (SLES12 SP3)
70 3.13.0-101 (Ubuntu 14.04, ZFS only)
71 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
72 4.4.0-116 (Ubuntu 16.04)
73 vanilla linux 4.6.7 (ZFS only)
74 * Client known to build on unpatched kernels:
75 2.6.32-431.29.2.el6 (RHEL6.5)
76 2.6.32-504.30.3.el6 (RHEL6.6)
77 2.6.32-573.26.1.el6 (RHEL6.7)
78 2.6.32-642.15.1.el6 (RHEL6.8)
79 2.6.32-696.18.7.el6 (RHEL6.9)
80 3.10.0-693.21.1.el7 (RHEL7.4)
81 3.0.101-0.47.71 (SLES11 SP3)
82 3.0.101-107 (SLES11 SP4)
83 3.12.74-60.64.40 (SLES12 SP1)
84 4.4.114-92.64 (SLES12 SP2)
85 4.4.114-94.11 (SLES12 SP3)
86 3.13.0-101 (Ubuntu 14.04)
87 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
88 4.4.0-116 (Ubuntu 16.04)
90 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
91 * Recommended ZFS / SPL version: 0.7.6
92 * Tested with ZFS / SPL version: 0.7.6
93 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
94 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
95 stack overflow. For more information, please refer to bugzilla 17630.
96 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
97 the re-exporting nodes to be mounted with "32bitapi" mount option
98 * Note for Ubuntu 14.04.5:
99 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
100 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
102 --------------------------------------------------------------------------------
104 2017-07-05 Intel Corporation
106 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
107 for currently supported client and server kernel versions.
108 * Server known to build on patched kernels:
109 2.6.32-431.29.2.el6 (RHEL6.5)
110 2.6.32-504.30.3.el6 (RHEL6.6)
111 2.6.32-573.26.1.el6 (RHEL6.7)
112 2.6.32-642.15.1.el6 (RHEL6.8)
113 2.6.32-696.3.1.el6 (RHEL6.9)
114 3.10.0-514.21.1.el7 (RHEL7.3)
115 3.0.101-0.47.71 (SLES11 SP3)
116 3.0.101-100 (SLES11 SP4)
117 3.12.74-60.64.40 (SLES12 SP1)
118 4.4.59-92.17 (SLES12 SP2)
119 vanilla linux 4.6.7 (ZFS only)
120 * Client known to build on unpatched kernels:
121 2.6.32-431.29.2.el6 (RHEL6.5)
122 2.6.32-504.30.3.el6 (RHEL6.6)
123 2.6.32-573.26.1.el6 (RHEL6.7)
124 2.6.32-642.15.1.el6 (RHEL6.8)
125 2.6.32-696.3.1.el6 (RHEL6.9)
126 3.10.0-514.21.1.el7 (RHEL7.3)
127 3.0.101-0.47.71 (SLES11 SP3)
128 3.0.101-100 (SLES11 SP4)
129 3.12.74-60.64.40 (SLES12 SP1)
130 4.4.59-92.17 (SLES12 SP2)
132 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
133 * Recommended ZFS / SPL version: 0.6.5.9
134 * Tested with ZFS / SPL version: 0.6.5.9
135 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
136 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
137 stack overflow. For more information, please refer to bugzilla 17630.
138 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
139 the re-exporting nodes to be mounted with "32bitapi" mount option
141 --------------------------------------------------------------------------------
143 12-07-2016 Intel Corporation
145 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
146 for currently supported client and server kernel versions.
147 * Server known to build on patched kernels:
148 2.6.32-431.29.2.el6 (RHEL6.5)
149 2.6.32-504.30.3.el6 (RHEL6.6)
150 2.6.32-573.26.1.el6 (RHEL6.7)
151 2.6.32-642.11.1.el6 (RHEL6.8)
152 3.10.0-514.el7 (RHEL7.3)
153 3.0.101-0.47.71 (SLES11 SP3)
154 3.0.101-80 (SLES11 SP4)
155 3.12.62-60.64.8 (SLES12 SP1)
156 vanilla linux 4.5.7 (ZFS only)
157 * Client known to build on unpatched kernels:
158 2.6.32-431.29.2.el6 (RHEL6.5)
159 2.6.32-504.30.3.el6 (RHEL6.6)
160 2.6.32-573.26.1.el6 (RHEL6.7)
161 2.6.32-642.11.1.el6 (RHEL6.8)
162 3.10.0-514.el7 (RHEL7.3)
163 3.0.101-0.47.71 (SLES11 SP3)
164 3.0.101-80 (SLES11 SP4)
165 3.12.62-60.64.8 (SLES12 SP1)
167 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
168 * Recommended ZFS / SPL version: 0.6.5.8
169 * Tested with ZFS / SPL version: 0.6.5.8
170 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
171 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
172 stack overflow. For more information, please refer to bugzilla 17630.
173 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
174 the re-exporting nodes to be mounted with "32bitapi" mount option
176 --------------------------------------------------------------------------------
178 02-29-2016 Intel Corporation
180 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
181 for currently supported client and server kernel versions.
182 * Server known to build on patched kernels:
183 2.6.32-431.29.2.el6 (RHEL6.5)
184 2.6.32-504.30.3.el6 (RHEL6.6)
185 2.6.32-573.12.1.el6 (RHEL6.7)
186 3.10.0-327.3.1.el7 (RHEL7.2)
187 3.0.101-0.47.71 (SLES11 SP3)
188 3.0.101-68 (SLES11 SP4)
189 vanilla linux 4.2.1 (ZFS only)
190 * Client known to build on unpatched kernels:
191 2.6.32-431.29.2.el6 (RHEL6.5)
192 2.6.32-504.30.3.el6 (RHEL6.6)
193 2.6.32-573.12.1.el6 (RHEL6.7)
194 3.10.0-327.3.1.el7 (RHEL7.2)
195 3.0.101-0.47.71 (SLES11 SP3)
196 3.0.101-68 (SLES11 SP4)
199 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
200 * Recommended ZFS / SPL version: 0.6.4.2
201 * Tested with ZFS / SPL version: 0.6.4.2
202 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
203 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
204 stack overflow. For more information, please refer to bugzilla 17630.
205 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
206 the re-exporting nodes to be mounted with "32bitapi" mount option
208 --------------------------------------------------------------------------------
210 03-10-2015 Intel Corporation
212 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
213 for currently supported client and server kernel versions.
214 * Server known to build on patched kernels:
215 2.6.32-431.29.2.el6 (RHEL6.5)
216 2.6.32-504.8.1.el6 (RHEL6.6)
217 3.0.101-0.46 (SLES11 SP3)
218 * Client known to build on unpatched kernels:
219 2.6.32-431.29.2.el6 (RHEL6.5)
220 2.6.32-504.8.1.el6 (RHEL6.6)
221 3.10.0-123.20.1.el7 (RHEL7)
222 3.0.101-0.46 (SLES11 SP3)
223 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
224 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
225 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
226 stack overflow. For more information, please refer to bugzilla 17630.
227 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
228 the re-exporting nodes to be mounted with "32bitapi" mount option
230 Severity : enhancement
232 Description: control OST-index in IDIF via ROCOMPAT flag.
233 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
234 last_rcvd file. For new formatted OST device, it will be auto set;
235 for the case of upgrading from old OST device, you can enable it
236 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
237 enabled, for new created OST-object, its IDIF-in-LMA will contain
238 the OST-index; for the existing OST-object, the OSD will convert
239 old format IDIF as new format IDIF with OST-index stored in the
240 LMA EA when accessing such OST-object or via OI scrub. Once such
241 flag is enabled, it cannot be reverted back, so the system cannot
242 be downgraded to the orignal incompatible version.
244 --------------------------------------------------------------------------------
246 07-30-2014 Intel Corporation
248 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
249 for currently supported client and server kernel versions.
250 * Server known to build on patched kernels:
251 2.6.32-431.20.3.el6 (RHEL6.5)
252 3.0.101-0.31 (SLES11 SP3)
253 * Client known to build on unpatched kernels:
254 2.6.32-431.20.3.el6 (RHEL6.5)
255 3.0.101-0.31 (SLES11 SP3)
258 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
259 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
260 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
261 stack overflow. For more information, please refer to bugzilla 17630.
262 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
263 the re-exporting nodes to be mounted with "32bitapi" mount option
265 --------------------------------------------------------------------------------
267 10-31-2013 Intel Corporation
269 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
270 * Server builds for kernels:
271 2.6.32-358.18.1.el6 (RHEL6.4)
272 * Client builds for unpatched kernels:
273 2.6.32-358.18.1.el6 (RHEL6.4)
274 3.0.92-0.8 (SLES11 SP3)
276 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
277 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
278 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
279 stack overflow. For more information, please refer to bugzilla 17630.
280 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
281 the re-exporting nodes to be mounted with "32bitapi" mount option
283 --------------------------------------------------------------------------------
285 05-30-2013 Intel Corporation
287 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
288 * Server builds for kernels:
289 2.6.32-358.6.2.el6 (RHEL6.4)
290 * Client builds for unpatched kernels:
291 2.6.32-358.6.2.el6 (RHEL6.4)
292 3.0.74-0.6.6 (SLES11 SP2)
294 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
295 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
296 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
297 stack overflow. For more information, please refer to bugzilla 17630.
298 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
299 the re-exporting nodes to be mounted with "32bitapi" mount option
301 --------------------------------------------------------------------------------
303 9-30-2012 Whamcloud, Inc.
305 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
306 * Server support for kernels:
307 2.6.32-279.5.1.el6 (RHEL6)
308 * Client support for unpatched kernels:
309 2.6.32-279.5.1.el6 (RHEL6)
311 2.6.18-238.19.1.el5 (RHEL5)
312 2.6.16.60-0.69.1 (SLES10)
313 * Recommended e2fsprogs version: 1.41.90.wc4
314 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
315 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
316 stack overflow. For more information, please refer to bugzilla 17630.
317 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
318 the re-exporting nodes to be mounted with "32bitapi" mount option
320 Severity : enhancement
322 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
324 Severity : enhancement
326 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
328 Severity : enhancement
330 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
332 Severity : enhancement
334 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
336 Severity : enhancement
338 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
340 --------------------------------------------------------------------------------
342 04-30-2012 Whamcloud, Inc.
344 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
345 * Server support for kernels:
346 2.6.32-220.4.2.el6 (RHEL6)
347 * Client support for unpatched kernels:
348 2.6.32-220.4.2.el6 (RHEL6)
350 2.6.18-238.19.1.el5 (RHEL5)
351 2.6.16.60-0.69.1 (SLES10)
352 * Recommended e2fsprogs version: 1.41.90.wc4
353 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
354 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
355 stack overflow. For more information, please refer to bugzilla 17630.
356 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
357 the re-exporting nodes to be mounted with "32bitapi" mount option
359 Severity : enhancement
361 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
363 Severity : enhancement
365 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
367 Severity : enhancement
369 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
371 Severity : enhancement
373 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
375 --------------------------------------------------------------------------------
377 09-30-2011 Whamcloud, Inc.
379 * Server support for kernels:
380 2.6.18-238.19.1.el5 (RHEL5)
381 2.6.32-131.6.1.el6 (RHEL6)
382 * Client support for unpatched kernels:
383 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
384 2.6.18-238.19.1.el5 (RHEL5)
385 2.6.32-131.6.1.el6 (RHEL6)
386 2.6.29.4-167.fc11 (FC11)
387 2.6.16.60-0.69.1 (SLES10)
389 * Recommended e2fsprogs version: 1.41.90.wc3
390 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
391 Lustre filesystem with 4K stack may cause a stack overflow. For more
392 information, please refer to bugzilla 17630.
393 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
394 mounted with "32bitapi" mount option
396 Severity : enhancement
397 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
399 Severity : enhancement
400 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
402 Severity : enhancement
403 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
406 Frequency : only with SLES10
408 Description: Use OFED "KMP" provided by Novell
409 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
410 Lustre is now built against this package. That means you need to
411 install the ofed-kmp package from Novell for the patchless client.
412 Note that the ofed-kmp that Novell ships may not exactly match the
413 kernel version but should still be compatible.
415 Severity : enhancement
417 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
418 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
420 --------------------------------------------------------------------------------
422 2010-07-15 Oracle, Inc.
424 * Server support for kernels:
425 2.6.18-164.11.1.el5 (RHEL 5)
426 2.6.18-164.11.1.0.1.el5 (OEL 5)
427 * Client support for unpatched kernels:
428 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
429 2.6.18-164.11.1.el5 (RHEL 5),
430 2.6.18-164.11.1.0.1.el5 (OEL 5)
431 2.6.16.60-0.42.8 (SLES 10),
433 2.6.29.4-167.fc11 (FC11)
434 * Recommended e2fsprogs version: 1.41.10-sun2
436 Severity : enhancement
438 Description: Distribute lfs_migrate script with lustre
439 Details : The lfs_migrate mentioned in the manual is now included in the
440 lustre tree in lustre/scripts and will distributed in lustre RPMs.
442 Severity : enhancement
444 Description: Failover nid list added to import proc file.
446 Severity : enhancement
448 Description: "som_preview" mount option required to use SOM
449 Details : The clients will need to use som_preview mount option
453 Description: Lustre does not do 1MB IOs to HW RAID
454 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
456 Severity : enhancement
458 Description: lfs quota output is not very convenient for awk/sed-parsing
459 Details : Some positions in lfs quota output table could be empty or
460 non-empty which made it hard to parse it with scripts, now
461 a dash is put instead of space where there is not supposed
466 Description: Update to OFED 1.5.1
467 Details : OFED 1.5.1 will be used in environments that don't use
468 vendor supplied OFED.
472 Description: Remove lnet. prefix from lctl params display
473 Details : Also renamed lnet.memused to lnet_memused to
478 Description: Use conf_param -d to remove permanant settings.
482 Description: Add list_param -R
483 Details : lctl list_param now supports recursive mode.
486 Bugzilla : 21829, 22312
487 Description: Fix llobdstat.
488 Details : 1. Now updating the write_bytes counter
489 2. Port script fixes from 1.8.
493 Description: decrease the usage of memory on clients.
494 Details : 1. On clients, recycle dentries and inodes unused.
495 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
500 Description: llobdstat fix and enhancement
501 Details : add a counter to set a limit to how many samples will be returned
502 fix a wildcard in the path to limit to obdfilter stats only
506 Description: lfs check only allowed for root
507 Details : Code cleanups and added a regresion test of non-root use of
512 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
513 Details : Fix hr->hr_index overload.
517 Description: Fix FIEMAP functionality.
521 Description: FC11 patchless client support added
525 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
526 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
530 Description: New acc-sm script and various test-framework enhancments.
534 Description: Previously landed patch was reverted.
538 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
542 Description: Add ldiskfs maxdirsize mount option.
546 Description: Fix mkfs.lustre support for 16TB LUNs.
550 Description: Print mmp_check_interval
551 Details : print mmp_check_interval and make it possible to abort
552 mount operation in case it takes too long.
556 Description: Add cascading_rw to lustre mpi tests.
560 Description: Store SOM data in EA.
561 Details : Also several SOM related cleanups.
565 Description: Add refcount to md_open_data
569 Description: Use refcount on lov_request_set to prevent memory corruption.
573 Description: Statahead: use parent dir's i_mutex for synchronization.
577 Description: loadgen improvements.
581 Description: Enable peer health detection in ksocklnd and ko2iblnd.
585 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
589 Description: Fix race on obd_devcie:md_stats
593 Description: Fix SLES 11 client builds.
594 Details : Resolved uninitalized pointer gcc warnings.
598 Description: Allow non-root access for "lfs check".
599 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
603 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
605 Severity : enhancement
607 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
608 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
612 Description: Sanity test_27p fails.
613 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
614 Sanity tests 27 were cleaned up and improved.
618 Description: Unify req format on client/servers
619 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
620 buffers and swabbers approach..
625 Description: drop join file stuff.
627 Severity : enhancement
629 Description: Add 2.6.27 and SLES11 (patchless client) support
633 Description: Fix in ptlrpc_expire_one_request() to print the signed time
636 Severity : enhancement
638 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
641 Severity : enhancement
643 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
645 Severity : enhancement
647 Description: correctly handle big reply message.
648 Details : send LNet event if reply is bigger than buffer and adjust this buffer
653 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
655 Severity : enhancement
657 Description: Add support for OEL5.
661 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
663 Severity : enhancement
665 Description: "weak-modules" support
666 Details : Implement "weak-modules" support which enables kernel modules
667 to be used with any kernel that implements the same kABI. In
668 order to achieve this modules are now installed in
669 /lib/modules/$(uname -r)/updates/kernel on all distributions.
673 Description: client could not reconnect to OST because of an active request.
674 Details : The client now retries to reconnect to the same server, if a
675 connect request failed with EBUSY or -EAGAIN. Test case 35b
676 conf-sanity.sh has been created.
680 Description: Don't allow make backward step on assignin osc next id.
681 Details : race between allocation next id and ll_sync thread can be cause
682 of set wrong osc next id and can be kill valid ost objects.
686 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
690 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
694 Description: Error handling in osc_statfs_interpret() has been improved.
695 Details : Check in osc_statfs_interpret() for EBADR.
699 Description: Do not send statfs() requests to OSTs disabled by administrator.
700 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
702 Severity : enhancement
704 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
706 Severity : enhancement
708 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
710 Severity : enhancement
712 Description: Enhancement for lfs(1) command to use numeric uid/gid.
716 Description: Improved cleanup procedure for conf-sanity test_46a.
720 Description: Do not start bulk transfer, if client tried to reconnect.
721 Details : Do not start bulk transfer, if client tried to reconnect,
722 but failed due to active rpcs. Check for exp_abort_active_req
723 has been implemented.
727 Description: Prevent inconsistences between linux and lustre mount structures.
728 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
729 Make the sleep interruptible.
731 Severity : enhancement
733 Description: provide server to client comms path
734 Details : server to client communications path via new LDLM_SET_INFO rpc
738 Description: truncate starts GFP_FS allocation under transaction causing deadlock
739 Details : ldiskfs_truncate calls grab_cache_page which may start page
740 allocation under an open transaction. This may lead to
741 calling prune_icache with consequent lustre reentrance.
745 Description: send correctly lsm on open replay
746 Details : MDS is trust to LSM size on replay open, but client can set wrong size
751 Description: enable client interface failover
752 Details : When a child reconnects from another NID, properly update export
753 nid hash position and ldlm reverse import.
755 Severity : enhancement
757 Description: Communicate OST degraded/readonly state via statfs to MDS
758 Details : Flags in the statfs returned from OSTs indicate whether the
759 OST is in a degraded RAID state, or if the filesystem has
760 turned read-only after a filesystem error is detected.
765 Description: don't panic if EPROTO was hit when reading symlink
766 Details : correctly handling request reference in error cases.
768 Severity : enhancement
770 Description: Add LustreNetLink, a kernel-userspace communcation path.
772 Severity : enhancement
774 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
779 Description: access to llog context before init.
780 Details : move handling CATALOGS file at osc layer and forbid access to llog
784 Frequency : rare, only when client is writing to OST on local node
786 Description: Avoid deadlock for local client writes
787 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
788 memory freeing context. This allows OST threads to set the
789 PF_MEMALLOC flag on task structures in order to allocate memory
790 from reserved pools and complete IO.
791 Use GFP_HIGHUSER for OST allocations for non-local client writes,
792 so that the OST threads generate memory pressure and allow
793 inactive pages to be reclaimed.
795 Severity : enhancement
797 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
802 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
803 Details : move ldlm namespace creation in setup phase to avoid grab
804 _lprocfs_lock with cli_sem held.
808 Description: Temporarily disable grant shrink.
809 Details : Disable the feature for debugging.
812 Frequency : only when formatting test filesystems
814 Description: Unable to run several mkfs.lustre on loop devices at the same time
815 Details : mkfs.lustre returns error 256 on the concurrent loop devices
816 formatting. The solution is to proper handle the error.
818 Severity : enhancement
820 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
822 Severity : enhancement
824 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
829 Description: resolve race between obd_disconnect and class_disconnect_exports
830 Details : if obd_disconnect will be called to already disconnected export he
831 forget release one reference and osc module can't unloaded.
833 Severity : enhancement
835 Description: Allow tuning service thread via /proc
836 Details : For each service a new
837 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
838 created that can be used to set min/max thread counts, and get the
839 current number of running threads.
844 Descriptoin: don't return error if have partially created objects for file.
845 Details : lov_update_create_set uses set->set_success as index for created
846 objects, so if some requests will be failed, they will have hole
847 at end of array and qos_shrink_lsm for allocate correct lsm.
849 Severity : enhancement
851 Description: Update OFED support to 1.4.1
856 Description: Reduce small size read RPC
857 Details : Set read-ahead limit for every file and only do read-ahead when
858 available read-ahead pages are bigger than 1M to avoid small size
861 Severity : enhancement
863 Description: move AT tunable parameters for more consistent usage
864 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
866 Severity : enhancement
868 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
869 Details : allow skip disconnected ost for send statfs request and hide error
875 Description: fix racy locking of mballoc block bitmaps causing BUG
876 Details : The locking of the mballoc buddy bitmap and the in-memory
877 block bitmap was using two different spin locks in some
878 cases. This made it possible to incorrectly access the
879 mballoc bitmap while another process was modifying it,
880 causing a sanity assertion to fail. While no on-disk corruption
881 was reported, there was some risk of this happening.
884 Frequency : rare, on llog test 6
886 Descriptoin: don't allow connect to already connected import
887 Details : allowing connect to already connected import is hide connecting problem.
892 Descriptoin: allow kill process which wait statahead result
893 Details : for some reasons 'ls' can stick in waiting result from statahead,
894 in this case need way for kill this process.
896 Severity : enhancement
898 Description: Add state history info file, enhance import info file
899 Details : Track import connection state changes in a new osc/mdc proc file;
900 add overview-type data to the osc/mdc import proc file.
902 Severity : enhancement
904 Description: MDS create should not wait for statfs RPC while holding DLM lock.
906 Severity : enhancement
908 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
913 Descriptoin: OST grant too much space to client even there are not enough space.
914 Details : Client will shrink its grant cache to OST if there are no write
915 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
916 this grant cache if there are already not enough avaible space
917 (left_space < total_clients * 32M).
920 Frequency : start MDS on uncleanly shutdowned MDS device
922 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
923 Details : stay in waiting mds<>ost recovery finished produce random bugs
924 due race between two ll_sync thread for one lov target. send
925 ACTIVATE event only if connect realy finished and import have
929 Frequency : rare, connect and disconnect target at same time
931 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
932 Details : don't call obd_disconnect under lov_lock. this long time
933 operation and can block ptlrpcd which answer to connect request.
938 Descriptoin: don't lose wakeup for imp_recovery_waitq
939 Details : recover_import_no_retry or invalidate_import and import_close can
940 both sleep on imp_recovery_waitq, but we was send only one wakeup
944 Frequency : always with long access acl
946 Descriptoin: mds can't pack reply with long acl.
947 Details : mds don't control size of acl but they limited by reint/getattr
950 Severity : enhancement
952 Description: Update to SLES10 kernel-2.6.16.60-0.33.
954 Severity : enhancement
956 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
959 Frequency : start MDS on uncleanly shutdowned MDS device
961 Descriptoin: aborting recovery hang on MDS
962 Details : don't throttle destroy RPCs for the MDT.
965 Frequency : on remount
967 Description: external journal device not working after the remount
968 Details : clear dev_rdonly flag for external journal devices in
974 Description: shutdown vs evict race
975 Details : client_disconnect_export vs connect request race.
976 if client will evicted at this time - we start invalidate
977 thread without referece to import and import can be freed
981 Frequency : rare, need acl's on inode.
983 Description: client can't handle ost additional correctly
984 Details : if ost was added after client connected to mds client can have
985 hit lnet_try_match_md ... to big messages to wide striped files.
986 in this case need teach client to handle config events about add
987 lov target and update client max ea size at that event.
989 Severity : enhancement
991 Description: Changelogs
992 Details : Changelogs are a lightweight mechanism to track filesystem
993 metadata and namespace changes. The changelog is recorded
994 permanently on the MDTs, and is periodically "consumed" / purged
995 when records are no longer needed.
997 Severity : enhancement
999 Description: compact fld format with extents
1000 Details : Store range of seq rather than every seq in FLD. Seq
1001 controller update FLD rather than clients. In Case of CMD, mdt0
1002 has FLD, all other metadata server act as non persistent proxy
1003 for FLD queries and cache fld entries in fld cache.
1008 Description: don't skip ost target if they assigned to file
1009 Details : Drop slow OSCs if we can, but not for requested start idx.
1010 This means "if OSC is slow and it is not the requested
1011 start OST, then it can be skipped, otherwise skip it only
1012 if it is inactive/recovering/out-of-space.
1016 Description: more cleanup in mds_lov
1017 Details : not send LOV EA under replay, we can't know about they size at this
1018 time. Don't allow client connect to mds before any ost connected,
1019 for avoid problems with LOV EA size and returning EIO to client.
1021 Severity : enhancement
1023 Description: Interoperability at server side (Disk interoperability)
1025 Severity : enhancement
1027 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1029 Severity : enhancement
1031 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1033 Severity : enhancement
1035 Description: New client IO stack (CLIO).
1037 Severity : enhancement
1039 Description: Commit on sharing. Eliminate inter-client dependencies between
1040 uncommitted transactions by doing transaction commits.
1041 Thereby clients may recovery independently.
1044 Frequency : Create a symlink file with a very long name
1046 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1047 Details : If there is no extra space in the request for early cancels,
1048 ldlm_req_handles_avail() returns 0 instead of a negative value.
1050 Severity : enhancement
1052 Description: Add /proc entry for import status
1053 Details : The mdc, osc, and mgc import directories now have
1054 an import directory that contains useful import data for debugging
1055 connection problems.
1057 Severity : enhancement
1059 Description: Re-disable certain /proc logging
1060 Details : Enable and disable client's offset_stats, extents_stats and
1061 extents_stats_per_process stats logging on the fly.
1064 Frequency : Only on FC kernels 2.6.22+
1066 Description: oops in statahead
1067 Details : Do not drop reference count for the dentry from VFS when lookup,
1068 VFS will do that by itself.
1070 Severity : enhancement
1072 Description: Generic /proc file permissions
1073 Details : Set /Proc file permissions in a more generic way to enable non-
1074 root users operate on some /proc files.
1078 Description: Hitting mdc_commit_close() ASSERTION
1079 Details : Properly handle request reference release in
1080 ll_release_openhandle().
1084 Description: quota recovery deadlock during mds failover
1085 Details : This patch includes att18982, att18236, att18237 in bz14840.
1087 1. fix osts hang when mds does failover with quotaon
1088 2. prevent watchdog storm when osts threads wait for the
1093 Frequency : only patchless client
1094 Description: add workaround for race between add/remove dentry from hash
1096 Severity : enhancement
1098 Description: Allow OST glimpses to return PW locks
1102 Description: LBUG when llog conf file is full
1103 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1107 Description: Prevent import from entering FULL state when server in recovery
1111 Description: service mount cannot take device name with ":"
1112 Details : Only when device name contains ":/" will mount treat it as
1118 Description: replace ptlrpcd with the statahead thread to interpret the async
1119 statahead RPC callback
1123 Frequency : on recovery
1124 Description: I/O failures after umount during fail back
1125 Details : if client reconnected to restarted server we need join to recovery
1126 instead of find server handler is changed and process self
1127 eviction with cancel all locks.
1129 Severity : enhancement
1131 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1133 Severity : enhancement
1135 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1137 Severity : enhancement
1139 Description: Upcall on Lustre log has been dumped
1140 Details : Allow for a user mode script to be called once a Lustre log has
1141 been dumped. It passes the filename of the dumped log to the
1142 script, the location of the script can be specified via
1143 /proc/sys/lnet/debug_log_upcall.
1148 Description: avoid idr_remove called for id which is not allocated.
1149 Details : Move assigment s_dev for clustered nfs to end of initialization,
1150 to avoid problem with error handling.
1155 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1156 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1157 into CONN_USED_HASH and this prodice warning when put connection
1158 again in unused hash.
1161 Severity : enhancement
1163 Description: File striping can now be set to use an arbitrary pool of OSTs.
1165 Severity : enhancement
1167 Description: Export bytes_read/bytes_write count on OSC/OST.
1171 Description: Early reply size mismatch, MGC loses connection
1172 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1173 the connect flags are properly negotiated.
1178 Description: quotas are not honored with O_DIRECT
1179 Details : all writes with the flag O_DIRECT will use grants which leads to
1180 this problem. Now using OBD_BRW_SYNC to guard this.
1184 Description: add quota statistics
1185 Details : 1. sort out quota proc entries and proc code.
1186 2. add quota statistics
1188 Severity : enhancement
1190 Description: enable quota support for HEAD.
1194 Description: Properly propagate oinfo flags from lov to osc for statfs
1195 Details : restore missing copy oi_flags to lov requests.
1197 Severity : enhancement
1199 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1200 : l_getgroups(8), lst(8), routerstat(8)
1201 Details : included man pages for llobdstat(8), llstat(8),
1202 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1204 Severity : enhancement
1206 Description: configure's --enable-quota should check the
1207 : kernel .config for CONFIG_QUOTA
1208 Details : configure is terminated if --enable-quota is passed but
1209 : no quota support is in kernel
1213 Description: Remove portals compatibility
1214 Details : Remove portals compatibility, not interoperable with releases
1219 Description: Resolve device initialization race
1220 Details : Prevent proc handler from accessing devices added to the
1221 obd_devs array but yet be intialized.
1223 Severity : enhancement
1225 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1227 Severity : enhancement
1229 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1234 Description: Using wrong pointer in osc_brw_prep_request
1235 Details : Access to array[-1] can produce panic if kernel compiled with
1236 CONFIG_PAGE_ALLOC enabled
1240 Description: Client runs out of low memory
1241 Details : Consider only lowmem when counting initial number of llap pages
1245 Description: *optional* service tags registration
1246 Details : if the "service tags" package is installed on a Lustre node
1247 When the filesystem is mounted, a local-node service tag will
1248 be created. See http://inventory.sun.com/ for more information
1249 about the Service Tags asset management system.
1253 Description: Kernel BUG tries to release flock
1254 Details : Lustre does not destroy flock lock before last reference goes
1255 away. So always drop flock locks when client is evicted and
1256 perform unlock regardless of successfulness of speaking to MDS.
1260 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1264 Description: sanity test 65a fails if stripecount of -1 is set
1265 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1270 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1271 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1274 Severity : enhancement
1276 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1280 Description: do not process already freed flock
1281 Details : flock can possibly be freed by another thread before it reaches
1282 to ldlm_flock_completion_ast.
1286 Description: LBUG during stress test
1287 Details : Need properly lock accesses the flock deadlock detection list.
1291 Description: oops in page fault handler
1292 Details : kernel page fault handler can return two special 'pages' in error
1293 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1297 Description: timeout with invalidate import.
1298 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1299 should be handled by ptlrpcd. This produce long age waiting and
1300 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1302 Severity : enhancement
1304 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1308 Description: enable MGS and MDT services start separately
1309 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1310 a co-located MGS without starting the MGS, which is a complement
1311 to 'nosvc' mount option.
1315 Frequency : after recovery
1316 Description: precreate to many object's after del orphan.
1317 Details : del orphan st in oscc last_id == next_id and this triger growing
1318 count of precreated objects. Set flag LOW to skip increase count
1319 of precreated objects.
1323 Frequency : rare, on clear nid stats
1324 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1325 Details : when clean nid stats sometimes try destroy live entry,
1326 and this produce panic in free.
1330 Description: Stack overflow during MDS log replay
1331 ease stack pressure by using a thread dealing llog_process.
1335 Description: wait until IO finished before start new when do lock cancel.
1336 Details : VM protocol want old IO finished before start new, in this case
1337 need wait until PG_writeback is cleared until check dirty flag and
1338 call writepages in lock cancel callback.
1340 Severity : enhancement
1342 Description: using special macro for print time and cleanup in includes.
1346 Description: mds_mfd_close() ASSERTION(rc == 0)
1347 Details : In mds_mfd_close(), we need protect inode's writecount change
1348 within its orphan write semaphore to prevent possible races.
1352 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1356 Frequency : rare, on shutdown ost
1357 Description: don't hit live lock with umount ost.
1358 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1359 use shrink_dcache_sb instead.
1363 Description: don't panic with use echo client
1364 Details : echo client pass NULL as client nid pointer and this produce null
1365 pointer dereference.
1369 Description: fix build on ppc32
1370 Details : compile code with -m64 flag produce wrong object file for ppc32.
1374 Description: add message levels for liblustreapi
1378 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1379 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1380 will result in an -EAGAIN and in the exit call path we call
1381 alloc_rr() which will with increasing aggressiveness attempt to
1382 aquire precreated objects on the minimum number of required OSCs.
1386 Description: Use old size assignment to avoid deadlock
1387 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1388 the scheduling while holding a spinlock. We do not need locking
1389 for size in ll_update_inode() because size is only updated from
1390 the MDS for directories or files without objects, so there is no
1391 other place to do the update, and concurrent access to such inodes
1392 are protected by the inode lock.
1396 Description: resolve "_IOWR redefined" build error on SLES10
1400 Description: dump the memory debugging after all modules are unloaded to
1401 suppress false negative in conf_sanity test 39
1403 Severity : enhancement
1405 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1410 Description: fix SLES kernel versioning
1411 Details : the kernel version for our SLES 10 kernel did not include a "-"
1412 before the "smp" at the end. while this was not a problem in
1413 general, it did mean that software trying to use the kernel
1414 version to try to detect a vendor specific kernel would fail.
1415 this was most evident by the OFED build scripts.
1419 Description: Don't update lov_desc members until making sure they are valid
1420 Details : When updating lov_desc members via proc fs, need fix their
1421 validities before doing the real update.
1425 Description: don't put request into delay list while invalidate in flight.
1426 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1427 import in flight. this produce timeout for invalidate and sometimes
1428 can cause stale data.
1432 Frequency : on ppc only
1433 Description: not convert ost objects for directory because it's not exist.
1434 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1438 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1440 Severity : enhancement
1442 Description: Update to support 2.6.22.14 vanilla kernel.
1446 Frequency : rare, on recovery
1447 Description: read procfs can produce deadlock in some situation
1448 Details : Holding lprocfs lock which send rpc can produce block for destroy
1449 obd objects and this also block reconnect with -EALREADY. This
1450 isn't fix all lprocfs bugs - but make it rare.
1452 Severity : enhancement
1454 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1457 Frequency : frequent on X2 node
1459 Description: mdc_set_open_replay_data LBUG
1460 Details : Set replay data for requests that are eligible for replay.
1464 Description: lustre_mgs: operation 101 on unconnected MGS
1465 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1466 MGC, and late on MGC cannot successfully connect to MGS and a lot
1467 of the error messages complaining that MGS is not connected.
1471 Frequency : on network error
1472 Description: panic with double free request if network error
1473 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1474 but it's true only for synchronus enqueue, for async enqueue
1475 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1478 Severity : enhancement
1480 Description: client-side metadata stat-ahead during readdir(directory readahead)
1481 Details : perform client-side metadata stat-ahead when the client detects
1482 readdir and sequential stat of dir entries therein
1485 Frequency : on start mds
1487 Description: Implement get_info(last_id) in obdfilter.
1490 Frequency : occasional
1492 Description: Correctly check stale fid, not start epoch if ost not support SOM
1493 Details : open with flag O_CREATE need set old fid in op_fid3 because
1494 op_fid2 was overwritten with new generated fid, but mds can answer
1495 with one of these two fids and both is not stale. Setattr
1496 incorrectly started an epoch and assume will be called
1497 done_writing, but without SOM done_writing ever being called.
1500 Frequency : rare, depends on device drivers and load
1502 Description: MDS or OSS nodes crash due to stack overflow
1503 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1504 In some cases, in conjunction with device drivers that use a lot
1505 of stack the MDS (or possibly OSS) service threads could overflow
1506 the stack. One change which was identified to consume additional
1507 stack has been reworked to avoid the extra stack usage.
1510 Frequency : occasional
1512 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1513 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1514 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1515 oscc_has_objects and see OSC has no objests and call
1516 oscc_internal_create to resend create request.
1518 Severity : enhancement
1520 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1522 Severity : enhancement
1524 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1527 Frequency : very rare
1529 Description: Fix replay if there is an un-replied request and open
1530 Details : In some cases, older replay request will revert the
1531 mcd->mcd_last_xid on MDS which is used to record the client's
1532 latest sent request.
1534 Severity : enhancement
1536 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1538 Severity : enhancement
1540 Description: Add rhel5 support to HEAD.
1542 Serverity : enhancement
1544 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1549 Description: Don't allow skipping OSTs if index has been specified.
1550 Details : Don't allow skipping OSTs if index has been specified, make
1551 locking in internal create lots better.
1555 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1556 Details : access to bitfield in structure is always rounded to long
1557 and this produce problem with not atomic change any bit.
1561 Description: Lustre make rpms failed.
1562 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1563 builds Lustre rpms from tarball.
1566 Frequency : rare on shutdown ost
1568 Description: If llog cancel was not send before clean_exports phase, this can
1569 produce deadlock in llog code.
1570 Details : If llog thread has last reference to obd and call class_import_put
1571 this produce deadlock because llog_cleanup_commit_master wait when
1572 last llog_commit_thread exited, but this never success because was
1573 called from llog_commit_thread.
1577 Description: allow userland application know is lost one of stripes.
1578 Details : fill lvb_blocks with error code on ost and return it to
1579 application if error flag found.
1583 Description: NULL lov_tgts causing MDS oops
1584 Details : more safe checks for NULL lov_tgts for avoid oops.
1586 Severity : enhancement
1588 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1592 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1593 Details : If an OST is down the MDS will hang indefinitely in
1594 obd_statfs_rqset() waiting for the statfs data. While for
1595 MDS QOS usage of statfs, it should not stuck in waiting.
1597 Severity : enhancement
1599 Description: remote_acl support
1600 Details : Support ACL-based permission check for remote user.
1601 Support setfacl/getfacl for remote user with the utils
1602 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1603 the system "{s,g}etfacl" utils.
1605 Severity : enhancement
1607 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1609 Severity : enhancement
1611 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1615 Description: make mgs_setparam() handle fsname containing dash
1616 Details : fsname containing a dash does not work with lctl conf_param
1618 Severity : enhancement
1620 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1622 Severity : enhancement
1624 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1628 Frequency : rare, at shutdown
1629 Description: access already free / zero obd_namespace.
1630 Details : if client_disconnect_export was called without force flag set,
1631 and exist connect request in flight, this can produce access to
1632 NULL pointer (or already free pointer) when connect_interpret
1633 store ocd flags in obd_namespace.
1637 Frequency : only at startup
1638 Description: not alloc memory with spinlock held.
1639 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1640 if any spinlock held.
1642 Severity : enhancement
1644 Description: make lustre randomly fail allocating memory
1645 Details : Make lustre randomly failed allocating memory for testing purpose.
1647 Severity : enhancement
1649 Description: lost problems with lov objid file
1650 Details : Fixes some scability and access to not inited memory problems
1651 in work with lov objdid file.
1656 Description: lfs find does not continue on file error
1657 Details : Continue other files processing when a file/dir is absent.
1661 Description: Inconsistent usage of lustre_pack_reply()
1662 Details : Standardize the usage of lustre_pack_reply() such that it
1663 always generate a CERROR on failure.
1666 Frequency : occasional
1668 Description: MDS hang or stay in waiting lock
1669 Details : If client receive lock with CBPENDING flag ldlm need send lock
1670 cancel as separate rpc, to avoid situation when cancel request
1671 can't processed due all i/o threads stay in wait lock.
1675 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1679 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1683 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1687 Description: Update to RHEL4 latest kernel.
1689 Severity : enhancement
1691 Description: Build SLES10 patchless client fails
1692 Details : The configure was broken by run ./configure with
1693 --with-linux-obj=.... argument for patchless client. When the
1694 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1695 can't search header adequately. Use absolute path such as
1696 -I($LINUX)/include instead.
1700 Description: interrupt oig_wait produce painc on resend.
1701 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1702 private set, and this produce situation when rq_ptlrpcd_data not
1703 copyed to new allocated request and triggered LBUG on assert
1704 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1705 ptlrpcd set if request is changed and can be safety changed to use
1710 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1711 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1712 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1713 these macros and replace them with their difinitions instead.
1717 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1718 Details : Update which_patch & target file for SLES10 latest kernel.
1721 Description: Remove client patches from SLES 10 kernel.
1722 Details : This causes SLES 10 clients to behave as patchless clients
1723 even on a Lustre-patched (server) kernel.
1725 Severity : enhancement
1727 Description: self-adjustable client's lru lists
1728 Details : use adaptive algorithm for managing client cached locks lru
1729 lists according to current server load, other client's work
1730 pattern, memory activities, etc. Both, server and client
1731 side namespaces provide number of proc tunables for controlling
1734 Severity : enhancement
1736 Description: light-weight GSS support
1737 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1738 and reduce performance overhead.
1740 Severity : enhancement
1742 Description: Linux keyring support
1743 Details : Support using service of Linux keyring for Lustre GSS internal
1744 context refresh/cache mechanism.
1748 Description: Fix errors in lfs documentation
1749 Details : Fixes man pages
1753 Description: don't use GFP_* in generic Lustre code.
1754 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1758 Description: obdclass is limited by single OBD_ALLOC(idarray)
1759 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1763 Description: warnings with build patchless client with vanila 2.6.19 and up
1764 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1765 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1769 Frequency : Only for SLES
1770 Description: sanity_quota fail test_1
1771 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1772 file, which makes TSTID[2] inunique.
1776 Frequency : Only for SLES9
1777 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1778 Details : Update target/ChangeLog/which_patch .
1782 Frequency : Only for test_10 in sanity-quota.sh
1783 Description: a bug in quota test code
1784 Details : checking a test flag in a wrong place
1788 Frequency : Only for test_13 in sanity-quota.sh
1789 Description: a bug in quota test script
1790 Details : shouldn't check overfull allocated quota
1794 Description: jbd statistics
1795 Details : Port older jbd statistics patch for sles10
1797 Severity : enhancement
1799 Description: Add printing inode info into message about error in writepage.
1804 Description: reply_lock_interpret crash due to race with it and lock cancel.
1805 Details : Do not replay locks that are being cancelled. Do not reference
1806 locks by their address during replay, just by their handle.
1810 Frequency : When flocks are used.
1811 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1812 Details : Flock locks might destroy just granted lock if it could be merged
1813 with another existing flock, this is done in completion handler,
1814 so teach ldlm_cli_enquque_fini that this is a valid case for
1820 Description: Oops in read and write path when failing to allocate lock.
1821 Details : Check if lock allocation failed and return error back.
1825 Description: lstripe command fails for valid OST index
1826 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1827 instead of lov->desc.ld_active_tgt_count.
1831 Description: sanity.sh failed test 103
1832 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1837 Description: df doesn't work properly if diskfs blocksize != 4K
1838 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1842 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1843 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1845 Severity : enhancement
1847 Description: Adaptive timeouts
1848 Details : RPC timeouts adapt to changing server load and network
1849 conditions to reduce resend attempts and improve recovery time.
1853 Description: llapi_file_create() does not allow some changes
1854 Details : add llapi_file_open() that allows specifying the mode and
1855 open flags, and also returns an open file handle.
1859 Description: merge and cleanup kernel patches.
1860 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1864 Description: Add journal checksum support.(Kernel part)
1865 Details : The journal checksum feature adds two new flags i.e
1866 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1867 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1868 indicates that the commit block contains the checksum for
1869 the blocks described by the descriptor blocks. Now commit
1870 record can be sent to disk without waiting for descriptor
1871 blocks to be written to disk. This behavior is controlled
1872 using JBD2_FEATURE_ASYNC_COMMIT flag.
1876 Description: OSS needs mutliple precreate threads
1877 Details : Add ability to start more than one create thread per OSS.
1881 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1882 Details : execute lfs setstripe on client
1886 Description: mds_obd_create error creating tmp object
1887 Details : When the user sets quota on root, llog will be affected and can't
1888 create files and write files.
1891 Frequency : Always on ia64 patchless client, and possibly others.
1893 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1894 Details : This allows the patchless client to be loaded on architectures
1895 without this export.
1899 Description: RedHat Update kernel for RHEL5
1900 Details : Add the kernel config file for RHEL5.
1904 Description: RedHat Update kernel for RHEL5
1905 Details : Modify the kernel config file more closer RHEL5.
1909 Description: Build failure against Centos5 (RHEL5)
1910 Details : Define PAGE_SIZE when it isn't present.
1914 Description: test_7 on sanity-quota.sh can't be used on separate machines
1915 Details : add facet to handle it
1919 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1920 Details : with NFS, the anon dentry's parent was set to itself in
1921 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1922 corresponding dentry other than use rec->ur_name.
1924 Severity : enhancement
1926 Description: lfs setstripe enhancement
1927 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1931 Description: enable data checksumming by default
1932 Details : enable checksum by default, allow --disable-checksum
1933 configure option and "-o nochecksum" mount option. Checksums
1934 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1938 Description: lustre support for RHEL5
1939 Details : Add support for RHEL5.
1943 Description: Client eviction due to failover config
1944 Details : after a connection loss, the lustre client should attempt to
1945 reconnect to the last active server first before trying the
1946 other potential connections.
1950 Description: when mds and osts use different quota unit(32bit and 64bit),
1951 quota will be released repeatly.
1952 Details : void sending multiple quota reqs to mds, which will keep the
1953 status between the reqs.
1957 Description: osts not allocated evenly to files
1958 Details : change the condition to increase offset_idx
1961 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1963 Description: Data corruption for OSTs that are formatted larger than 2TB
1965 Details : When generating the bio request for lustre file writes the
1966 sector number would overflow a temporary variable before being
1967 used for the IO. The data reads correctly from Lustre (which
1968 will overflow in a similar manner) but other file data or
1969 filesystem metadata may be corrupted in some cases.
1973 Description: Tune the kernel for good SCSI performance.
1974 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1975 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1980 Description: rewrite ext2-derived code in obdclass/uuid.c
1981 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1982 from scratch preserving functionality.
1986 Description: "lfs find -obd UUID" prints directories
1987 Details : "lfs find -obd UUID" will return all directory names instead
1988 of just file names. It is incorrect because the directories
1989 do not reside on the OSTs.
1993 Description: use i_size_read and i_size_write in 2.6 port
1994 Details : replace inode->i_size access with i_size_read/write()
1997 Frequency : when removing large files
1999 Description: scheduling issue during removal of large Lustre files
2000 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2001 It causes scheduling issues when removing large files (17TB in the
2005 Frequency : only with liblustre clients on XT3
2007 Description: evictions taking too long
2008 Details : allow llrd to evict clients directly on OSTs
2011 Frequency : only on ppc
2013 Description: /proc/fs/lustre/devices broken on ppc
2014 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2015 We should make sure the type of 'index' is loff_t and then cast
2016 explicitly as needed below. Do not assign an explicitly cast
2020 Frequency : only for rhel5
2022 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2023 Details : Modify the target file & which_kernel.
2025 Severity : enhancement
2027 Description: omit set fsid for export NFS
2028 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2029 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2033 Frequency : Always, for kernels after 2.6.16
2034 Description: Fix warning idr_remove called for id=.. which is not allocated.
2035 Details : Last kernels save old s_dev before kill super and not allow
2036 to restore from callback - restore it before call kill_anon_super.
2040 Description: buffer overruns could theoretically occur
2041 Details : llapi_semantic_traverse() modifies the "path" argument by
2042 appending values to the end of the origin string, and a
2043 overrun may occur. Adding buffer overrun check in liblustreapi.
2047 Description: Fix error on 'ls .' at the top of the Lustre mount.
2048 Details : Don't revalidate dentry if it is a root dentry.
2052 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2053 Details : Modify vm-tunables-rhel4.patch.
2057 Description: Kernel config for 2.6.18-vanilla.
2058 Details : Modify targets/2.6-vanilla.target.in.
2059 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2060 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2061 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2062 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2066 Description: improve handling recoverable errors
2067 Details : if request processig with error which can be recoverable on server
2068 request should be resend, otherwise page released from cache and
2073 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2074 Details : Modify target file & which_patch.
2075 A flaw was found in the IA32 system call emulation provided
2076 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2077 value could be stored in the %RAX register, which could trigger an
2078 out-of-bounds system call table access. An untrusted local user
2079 could exploit this flaw to run code in the kernel
2080 (ie a root privilege escalation). (CVE-2007-4573).
2084 Description: O_DIRECT bypasses client statistics.
2085 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2086 read_ahead_stats were updated. copied the stats section from
2087 osc_send_oap_rpc() into async_internal().
2091 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2095 Description: change order of libsysio includes
2096 Details : '#include sysio.h' should always come before '#include xtio.h'
2100 Description: handle "serious error: objid * already exists" more gracefully
2101 Details : If LAST_ID value on disk is smaller than the objects existing in
2102 the O/0/d* directories, it indicates disk corruption and causes an
2103 LBUG(). If the object is 0-length, then we should use the existing
2104 object. This will help to avoid a full fsck in most cases.
2106 Severity : enhancement
2108 Description: adapt the lustre_config script to support the upgrade case
2109 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2110 server targets to 1.6 in parallel.
2114 Description: To avoid grant space > avaible space when the disk is almost
2115 full. Without this patch you might see the error "grant XXXX >
2116 available" or some LBUG about grant, when the disk is almost
2118 Details : In filter_check_grant, for non_grant cache write, we should
2119 check the left space by if (*left > ungranted + bytes), instead
2120 of (*left > ungranted), because only we are sure the left space
2121 is enough for another "bytes", then the ungrant space should be
2122 increase. In client, we should update cl_avail_grant only there
2123 is OBD_MD_FLGRANT in the reply.
2127 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2128 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2129 described in CVE-2007-4573. This problem could allow untrusted
2130 local users to gain root access.
2133 Frequency : when using O_DIRECT and quotas
2135 Description: Incorrect file ownership on O_DIRECT output files
2136 Details : block usage reported by 'lfs quota' does not take into account
2137 files that have been written with O_DIRECT.
2142 Description: touch file failed when fs is not full
2143 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2144 otherwise we can get ENOSP while fs is not full.
2148 Description: parallel lock callbacks
2149 Details : Instead of sending blocking and completion callbacks as separated
2150 requests, adding them to a set and sending in parallel.
2153 Frequency : only for Cray XT3
2154 Bugzilla : 12829/13455
2155 Description: Changing primary group doesn't change the group lustre assigns to
2157 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2158 overridden with the primary group provided by the group upcall,
2159 whereas the supplied fsgid can be trusted if it is in the list of
2160 supplementary groups returned by the group upcall.
2162 Severity : enhancement
2164 Description: Allow masking D_WARNING, D_ERROR messages from console
2165 Details : Console messages can now be disabled via lnet.printk.
2169 Description: User code with malformed file open parameter crashes client node
2170 Details : Before packing join_file req, all the related reference should be
2171 checked carefully in case some malformed flags cause fake
2172 join_file req on client.
2176 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2177 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2178 mode is equal to the granted mode. However, it should also check
2179 that LDLM_FL_CP_REQD is not set.
2183 Description: Heavy nfs access might result in deadlocks
2184 Details : After ELC code landed, it is now improper to enqueue any mds
2185 locks under och_sem, because enqueue might want to decide to
2186 cancel open locks for same inode we are holding och_sem for.
2190 Description: Client eviction while running blogbench
2191 Details : A lot of unlink operations with concurrent I/O can lead to a
2192 deadlock causing evictions. To address the problem, the number of
2193 oustanding OST_DESTROY requests is now throttled to
2194 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2195 callbacks are processed in priority.
2199 Description: enable ACLs on MDS by default
2200 Details : ACLs must be enabled on MDS by default.
2203 Frequency : PPC/PPC64 only
2205 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2206 Details : fix duplicated definitions between asm-ppc64/types.h and
2207 lustre_types.h on PPC.
2210 Frequency : PPC/PPC64 only
2212 Description: asm-ppc/segment.h does not exist
2213 Details : fix compile issue on PPC.
2217 Description: better handle error messages in extents code
2220 Frequency : RHEL4 only
2222 Description: mkfs is very slow on IA64/RHEL4
2223 Details : A performance regression has been discovered in the MPT Fusion
2224 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2225 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2226 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2228 Severity : enhancement
2230 Description: SNMP support enhancement
2231 Details : Adding total number of sampled request for an MDS node in snmp
2234 Severity : enhancement
2236 Description: Optimize ldlm waiting list processing for PR extent locks
2237 Details : When processing waiting list for read extent lock and meeting read
2238 lock that is same or wider to it that is not contended, skip
2239 processing rest of the list and immediatelly return current
2240 status of conflictness, since we are guaranteed there are no
2241 conflicting locks in the rest of the list.
2245 Description: Time out and refuse to reconnect
2246 Details : When the failover node is the primary node, it is possible
2247 to have two identical connections in imp_conn_list. We must
2248 compare not conn's pointers but NIDs, otherwise we can defeat
2249 connection throttling.
2253 Description: port llog fixes from b1_6 into HEAD
2254 Details : Port llog reference couting and some llog cleanups from b1_6
2255 (bug 10800) into HEAD, for protect from panic and access to already
2256 free llog structures.
2260 Description: Detect stride IO mode in read-ahead
2261 Details : When a client does stride read, read-ahead should detect that and
2262 read-ahead pages according to the detected stride pattern.
2266 Description: data checksumming impacts single node performance
2267 Details : add support for several checksum algorithm. Currently, only CRC32
2268 and Adler-32 are supported. The checksum type can be changed on
2269 the fly via /proc/fs/lustre/osc/*/checksum_type.
2273 Description: use adler32 for page checksums
2274 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2279 Description: build for x2 fails
2280 Details : fix compile issue on Cray systems.
2284 Description: Properly match for duplicate locks
2285 Details : Due to different lock order from skiplists code, we need to
2286 traverse entire list for now
2289 Frequency : only on PPC/SLES10
2291 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2292 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2293 be defined. Add a hack in mkfs_lustre.c to work around this compile
2298 Description: LASSERT on MDS when client holding flock lock dies
2299 Details : ldlm pool logic depends on number of granted locks equal to
2300 number of released locks which is not true for flock locks, so
2301 just exclude such locks from consideration.
2305 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2306 Details : Use fsfilt_sync() for both the whole filesystem sync and
2307 individual file sync to eliminate dangerous inode locking
2308 with I_LOCK that can lead to a deadlock.
2312 Description: performance in 1.6.3
2313 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2314 sure that 1MB requests are not fragmented by the block layer.
2316 Severity : enhancement
2318 Description: organize the server-side client stats on per-nid basis
2319 Details : Change the structure of stats under obdfilter and mds to
2327 The "uuid"s file would list the uuids of _active_ exports.
2328 And the clear entry is to clear all stats and stale nids.
2330 Severity : enhancement
2332 Description: eliminate client locks in face of contention
2333 Details : file contention detection and lockless i/o implementation
2334 for contended files.
2338 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2341 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2343 Description: MDS may lose file striping (and hence file data) in some cases
2344 Details : If there are additional extended attributes stored on the MDS,
2345 in particular ACLs, SELinux, or user attributes (if user_xattr
2346 is specified for the client mount options) then there is a risk
2347 of attribute loss. Additionally, the Lustre file striping
2348 needs to be larger than default (e.g. striped over all OSTs),
2349 and an additional attribute must be stored initially in the
2350 inode and then increase in size enough to be moved to the
2351 external attribute block (e.g. ACL growing in size) for file
2356 Description: skiplist implementation simplification
2357 Details : skiplists are used to group compatible locks on granted list
2358 that was implemented as tracking first and last lock of each
2359 lock group the patch changes that to using doubly linked lists
2363 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2364 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2365 because LASSERT on both the data supplied by a client, and the
2366 data on disk is dangerous and incorrect.
2368 Severity : enhancement
2370 Description: Slow truncate/writes to huge files at high offsets.
2371 Details : Directly associate cached pages to lock that protect those pages,
2372 this allows us to quickly find what pages to write and remove
2373 once lock callback is received.
2377 Description: more ldlm soft lockups
2378 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2379 starve other threads from the resource lock for a long time in
2380 case of long waiting queue, so change the debug level from
2381 D_OTHER to the less frequently used D_INFO.
2383 Severity : enhancement
2385 Description: add -gid, -group, -uid, -user options to lfs find
2389 Description: Hung threads in invalidate_inode_pages2_range
2390 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2391 once one is completed. As a result, some RPCs are stuck in the
2392 queue and are never sent.
2396 Description: filter threads hungs on waiting journal commit
2397 Details : Cleanup filter group llog code, then only filter group llog will
2398 be only created in the MDS/OST syncing process.
2402 Description: Procfs and llog threads access destoryed import sometimes.
2403 Details : Sync the import destoryed process with procfs and llog threads by
2404 the import refcount and semaphore.
2406 Severity : enhancement
2408 Description: openlock cache of b1_6 port to HEAD
2413 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2414 Details : If insertion of an extent fails, then discard the inode
2415 preallocation and free data blocks else it can lead to duplicate
2420 Description: don't always update ctime in ext3_xattr_set_handle()
2421 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2422 In some cases the ctime should not be updated, for example for
2423 2.0->1.8 compatibility it is necessary to delete an xattr and it
2424 should not update the ctime.
2428 Bugzilla : 15713/16362
2429 Description: Assertion in iopen_connect_dentry in 1.6.3
2430 Details : looking up an inode via iopen with the wrong generation number can
2431 populate the dcache with a disconneced dentry while the inode
2432 number is in the process of being reallocated. This causes an
2433 assertion failure in iopen since the inode's dentry list contains
2434 both a connected and disconnected dentry.
2438 Description: assertion failure in ldlm_handle2lock()
2439 Details : fix a race between class_handle_unhash() and class_handle2object()
2440 introduced in lustre 1.6.5 by bug 13622.
2445 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2446 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2451 Description: oops in sd_iostats_seq_show()
2452 Details : unloading/reloading the scsi low level driver triggers a kernel
2453 bug when trying to access the sd iostat file.
2458 Description: Kernel panics during QLogic driver reload
2459 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2460 patch, causing memory corruption.
2465 Description: journal_dev option does not work in b1_6
2466 Details : pass mount option during pre-mount.
2468 Severity : enhancement
2470 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2471 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2472 extent information of a file. It can be used to map logical blocks
2473 in a file to physical blocks in the block device.
2477 Description: LDLM soft lockups - improvement
2478 Details : It is be possible to send the lock handle along with each read
2479 or write request because the client is already doing a lock match
2480 itself so there isn't any reason the OST should have to re-do that
2484 Frequency : only on Cray X2
2486 Description: X2 build failures
2487 Details : fix build failures on Cray X2.
2491 Description: xid & resent requests
2492 Details : Initialize RPC XID from clock at startup (randomly if clock is
2495 Severity : enhancement
2497 Description: Add lustre_start utility to start or stop multiple Lustre servers
2502 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2503 Details : In case of memory pressure, list_del() can be called twice on
2504 req->rq_history_list, causing a kernel oops.
2508 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2509 Details : fix stack overflow in the distributed lock manager by defering
2510 export eviction after a failed AST to the elt thread instead of
2511 handling it in the dlm interpret routine.
2515 Description: Convert some comments to new format.
2516 Details : Update documenting comments to match doxygen conventions.
2520 Description: Grammar fixes.
2521 Details : A couple of trivial spelling fixes.
2525 Description: OSD_COUNTERS-mandatory
2526 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2527 to simplify the code.
2531 Description: simplify cmm_device freeing logic.
2532 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2536 Description: Add lockdep support to dt_object_operations locking interface.
2537 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2538 parameter indicating lock ordering. Update mdd code to use new
2543 Description: Introduce failloc constants for lockless IO tests.
2544 Details : Add two new failloc constants to test lockless IO. Only one of
2545 them in implemented---another is checked in yet to be landed
2550 Description: Add lockdep support for inode mutex.
2551 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2552 in the situations where Lustre has to lock more than one inode
2557 Description: Add optional invariants checking support.
2558 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2559 configure switch. This macro is to be used for consistency and
2560 sanity checks that are too expensive to be left in `production'
2565 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2566 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2571 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2572 Details : Use spin_lock_nested() in (the only) situation where more than
2573 one ldlm_lock is locked simultaneously. Also, fix possible
2574 dead-lock in ldlm_lock_change_resource() by enforcing particular
2579 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2580 Details : Instead of specifying each ldlm_lock call-back through separate
2581 parameter, wrap them into struct ldlm_callback_suite.
2585 Description: Kill join_lru obd method and its callers.
2586 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2587 in memory---a requirement implemented through ->o_join_lru() obd
2588 method in HEAD. Remove this method and its users.
2592 Description: Add asynchronous ldlm ENQUEUE completion handler.
2593 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2594 case that stock ldlm_completion_ast() cannot handle as it waits
2595 until lock is granted. Introduce new ldlm_completion_ast_async()
2596 for this. Also comment ldlm_completion_ast().
2600 Description: ldlm_error <-> errno conversion.
2601 Details : Add functions to map (rather arbitrary) between LDLM error codes
2602 and standard errno values. CLIO needs this to prevent LDLM specific
2603 constants from escaping ldlm and osc.
2607 Description: Kill unused ldlm_handle2lock_ns() function.
2608 Details : Kill unused ldlm_handle2lock_ns() function.
2612 Description: Add lu_ref support to ldlm_lock
2613 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2614 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2615 and ldlm_resource. LDLM interface has to be changed, because code that
2616 releases a reference on a lock, has to "know" what reference this is.
2617 In the most frequent case
2619 lock = ldlm_handle2lock(handle);
2621 LDLM_LOCK_PUT(lock);
2623 no changes are required. When any other reference (received _not_ from
2624 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2625 instead of LDLM_LOCK_PUT().
2627 Arguably, changes are pervasive, and interface requires some discipline
2628 for proper use. On the other hand, it was very instrumental in finding
2629 a few leaked lock references.
2633 Description: Add ldlm_lock_addref_try().
2634 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2635 attempts to addref a lock that might be being canceled
2640 Description: Add ldlm_weigh_callback().
2641 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2642 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2643 hard-coded `number of pages' logic.
2647 Description: Add lockdep annotations to llog code.
2648 Details : Use appropriately tagged _nested() locking calls in the places
2649 where llog takes more than one ->lgh_lock lock.
2653 Description: Add loi_kms_set().
2654 Details : Wrap kms updates into a helper function.
2658 Description: Constify instances of struct lsm_operations.
2659 Details : Constify instances of struct lsm_operations.
2663 Description: lu_conf support.
2664 Details : On a server, a file system object is uniquely identified
2665 by a fid, which is sufficient to locate and load all object
2666 state (inode). On a client, on the other hand, more data are
2667 necessary instantiate an object. Change lu_object_find() and
2668 friends to take additional `lu_conf' argument describing object.
2669 Typically this includes layout information.
2673 Description: lu_context fixes.
2674 Details : Introduce new lu_context functions that are needed on the client
2675 side, where some system threads (ptlrpcd) are shared by multiple
2676 modules, and so cannot be stopped during module shutdown.
2680 Description: Add start and stop methods to lu_device_type_operations.
2681 Details : Introduce two new methods in lu_device_type_operations, that are
2682 invoked when first instance of a given type is created and last one
2683 is destroyed respectively. This is need by CLIO.
2687 Description: Add lu_ref support to struct lu_device.
2688 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2689 track leaked references.
2693 Description: Introduce lu_kmem_descr.
2694 Details : lu_kmem_descr and its companion interface allow to create
2695 and destroy a number of kmem caches at once.
2699 Description: Fix lu_object finalization race.
2700 Details : Fix a race between lu_object_find() finding an object and its
2701 concurrent finalization. This race is (most likely) not possible
2702 on the server, but might happen on the client.
2706 Description: Introduce lu_ref interface.
2707 Details : lu_ref is a debugging module allowing to track references to
2708 a given object. It is quite cpu expensive, and has to be
2709 explicitly enabled with --enable-lu_ref. See usage description
2714 Description: Factor lu_site procfs stats into a separate function.
2715 Details : Separate lu_site stats printing code into a separate function
2716 that can be reused on a client.
2720 Description: Constify instances of struct {lu,dt,md}_device_operations.
2721 Details : Constify instances of struct {lu,dt,md}_device_operations.
2725 Description: Introduce struct md_site and move meta-data specific parts of
2726 struct lu_site here.
2727 Details : Move md-specific fields out of struct lu_site into special struct
2728 md_site, so that lu_site can be used on a client.
2732 Description: Kill mdd_lov_destroy().
2733 Details : Remove unused mdd code.
2737 Description: Add st_block checking to multistat.c.
2738 Details : Add st_block checking to multistat.c.
2742 Description: Add lu_ref support to struct obd_device.
2743 Details : Add lu_ref tracking to obd_device.
2747 Description: Kill obd_set_fail_loc().
2748 Details : Remove unused code.
2752 Description: Add special type for ptlrpc_request interpret functions.
2753 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2754 there. Actual usage will be in CLIO.
2758 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2759 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2760 doesn't work with lockdep.
2764 Description: Add rwv.c test program.
2765 Details : New testing program exercising readv(2) and writev(2) (Qian).
2769 Description: Add sendfile.c test program.
2770 Details : New testing program exercising sendfile(2) (Jay).
2774 Description: Ratelimit a message that can be very frequent.
2775 Details : Ratelimit a memory allocation failure message that can
2780 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2781 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2785 Description: Liblustre build fixes.
2786 Details : Liblustre build fixes.
2790 Description: libcfs: add cfs_{need,cond}_resched() interface.
2791 Details : libcfs: add cfs_{need,cond}_resched() definition and
2792 implementations for Linux, NT, and liblustre.
2794 Severity : enhancement
2796 Description: More exported tunables for mballoc
2797 Details : Add support for tunable preallocation window and new tunables for
2798 large/small requests
2802 Description: Detect corruption of block bitmap and checking for preallocations
2803 Details : Checks validity of on-disk block bitmap. Also it does better
2804 checking of number of applied preallocations. When corruption is
2805 found, it turns filesystem readonly to prevent further corruptions.
2809 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2810 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2811 window to 0. If it does hit cache miss, and read-pattern is still
2812 stride-io mode, does not reset the stride window, but also does not
2813 increase the stride window length in this case.
2817 Frequency : only for big-endian servers
2818 Description: Check if system is big-endian while mounting fs with extents feature
2819 Details : Mounting a filesystem with extents feature will fail on big-endian
2820 systems since ext3-based ldiskfs is not supported on big-endian
2821 systems. This can be over-riden with "bigendian_extents" mount option.
2823 Severity : enhancement
2825 Description: The root squash functionality
2826 Details : A security feature, which is to prevent users from being able
2827 to mount lustre on their desktop, run as root, and delete
2828 all of the files in the filesystem. The goal is accomplished by
2829 remapping user id (UID) and group id (GID) of the root user to
2830 a UID and GID specified by the system administartor via Lustre
2831 configuration management server (MGS). The functionality also
2832 allows to specify sets of clients for which the remapping does
2837 Description: Excessive recovery window
2838 Details : With AT enabled, the recovery window can be excessively long (6000+
2839 seconds). To address this problem, we no longer use
2840 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2841 timeout no longer depends on the service time, it is set to
2842 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2843 time via pb_service_time.
2847 Description: Watchdog triggered on MDS failover
2848 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2849 the OSTs know that the MDS "UUID" can be reused for the same export
2850 from a different NID, so we do not need to wait for the export to be
2854 Frequency : rare, only if using MMP with Linux RAID
2856 Description: MMP doesn't work with Linux RAID
2857 Details : While using HA for Lustre servers with Linux RAID, it is possible
2858 that MMP will not detect multiple mounts. To make this work we
2859 need to unplug the device queue in RAID when the MMP block is being
2860 written. Also while reading the MMP block, we should read it from
2861 disk and not the cached one.
2863 Severity : enhancement
2865 Description: open file using fid
2866 Details : A file can be opened using just its fid, like
2867 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2870 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2873 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2874 Details : add proper locking for creation of jbd_4k slab cache
2878 Description: MMP check in ext3_remount() fails without displaying any error
2879 Details : When multiple mount protection fails during remount, proper error
2882 Severity : enhancement
2884 Description: Allow stripe size to be up to 4G-64k
2885 Details : Fix math logic to allow large stripe sizes.
2889 Description: add check for >8TB ldiskfs filesystems
2890 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2891 Don't allow >8TB ldiskfs filesystems to be mounted without
2892 force_over_8tb mount option
2897 Descriptoin: OSS complains "too many blocks" and crashes
2898 Details : In filter_commitrw_write():
2899 1. lquota_chkquota();
2900 2. obdo_from_inode();
2901 3. lquota_pending_commit();
2902 "2." has changed uid and gid in this case, then caused
2903 lquota_chkquota() and lquota_pending_commit() to use
2904 different uid and gid. That is the root of the bug.
2908 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2909 Details : Detect and print out the 'stripe count = -1' case properly.
2913 Description: Out or order replies might be lost on replay
2914 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2915 request with tid smaller than one currently being added, add it
2916 to the start, not end of the list.
2918 --------------------------------------------------------------------------------
2920 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2922 * Support for kernels:
2923 2.6.5-7.283 (SLES 9),
2924 2.6.9-55.EL (RHEL 4),
2925 2.6.16.46-0.14 (SLES 10),
2926 2.6.18.8 vanilla (kernel.org)
2927 * Client support for unpatched kernels:
2928 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2929 2.6.16 - 2.6.22 vanilla (kernel.org)
2930 * Due to recently discovered recovery problems, we do not recommend
2931 using patchless RHEL 4 clients with this or any earlier release.
2932 * Recommended e2fsprogs version: 1.39.cfs8
2933 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2934 * Starting with this release, the ldiskfs backing filesystem required
2935 by Lustre is now in its own package, lustre-ldiskfs. This package
2936 should be installed. It is versioned separately from Lustre and
2937 may be released separately in future.
2941 Description: block reactivating mgc import until all deactivates complete
2942 Details : Fix race when failing back MDT/MGS to itself (testing)
2945 Frequency : at statup only
2947 Description: mds_lov_synchronize race leads to various problems
2948 Details : simultaneous MDT->OST connections at startup can cause the
2949 sync to abort, leaving the OSC in a bad state.
2951 Severity : enhancement
2953 Description: add optional extra BUILD_VERSION info
2954 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2955 to override the lustre version.
2958 Frequency : 2.6.18 servers only
2960 Description: ll_kern_mount() doesn't release the module reference
2961 Details : The ldiskfs module reference count never drops down to 0
2962 because ll_kern_mount() doesn't release the module reference.
2967 Description: server LBUG when using old ost_num_threads parameter
2968 Details : Accept the old ost_num_threads parameter but warn that it
2969 is deprecated, and fix an off-by-one error that caused an LBUG.
2974 Description: Transient SCSI error results in persistent IO issue
2975 Details : iobuf->dr_error is not reinitialized to 0 between two
2979 Frequency : sometimes when underlying device returns I/O errors
2981 Description: OSTs not going read-only during write failures
2982 Details : OSTs are not remounted read-only when the journal commit threads
2983 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2984 instead of the ext3 wrappers.
2988 Description: poor connect scaling with increasing client count
2989 Details : Don't run filter_grant_sanity_check for more than 100 exports
2990 to improve scaling for large numbers of clients.
2993 Frequency : SLES10 only
2995 Description: sanity-quota.sh quotacheck failed: rc = -22
2996 Details : Quotas cannot be enabled on SLES10.
2999 Frequency : liblustre clients only
3001 Description: getdirentries does not give error when run on compute nodes
3002 Details : getdirentries does not fail when the size specified as an argument
3003 is too small to contain at least one entry
3005 Severity : enhancement
3007 Description: Add LNET router traceability for debug purposes
3008 Details : If a checksum failure occurs with a router as part of the
3009 IO path, the NID of the last router that forwarded the bulk data
3010 is printed so it can be identified.
3015 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3016 Details : Due to a race condition, liblustre clients were occasionally
3017 evicted incorrectly.
3019 Severity : enhancement
3021 Description: lfs setstripe use optional parameters instead of postional
3024 Severity : enhancement
3026 Description: Nanosecond timestamp support for ldiskfs
3027 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3028 resolution timestamps. There is not yet support for this at
3029 the Lustre filesystem level.
3032 Frequency : during server recovery
3034 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3035 Details : request with rq_no_resend flag not awake l_wait_event if they get
3039 Frequency : nfs export on patchless client
3041 Description: connectathon hang when test nfs export over patchless client
3042 Details : Disconnected dentry cannot be found with lookup, so we do not need
3043 to unhash it or make it invalid
3046 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3047 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3048 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3053 Description: open req refcounting wrong on reconnect
3054 Details : If reconnect happened between getting open reply from server and
3055 call to mdc_set_replay_data in ll_file_open, we will schedule
3056 replay for unreferenced request that we are about to free.
3057 Subsequent close will crash in variety of ways.
3058 Check that request is still eligible for replay in
3059 mdc_set_replay_data().
3064 Description: disable writes to filesystem when reading health_check file
3065 Details : the default for reading the health_check proc file has changed
3066 to NOT do a journal transaction and write to disk, because this
3067 can cause reads of the /proc file to hang and block HA state
3068 checking on a healthy but otherwise heavily loaded system. It
3069 is possible to return to the previous behaviour during configure
3070 with --enable-health-write.
3072 Severity : enhancement
3074 Description: 64-bit inode version
3075 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3076 to the inode. This will be required for version-based recovery.
3081 Description: MDS fails to start if a duplicate client export is detected
3082 Details : in some rare cases it was possible for a client to connect to
3083 an MDS multiple times. Upon recovery the MDS would detect this
3084 and fail during startup. Handle this more gracefully.
3086 Severity : enhancement
3088 Description: Add -o localflock option to simulate old noflock
3090 Details : This will achieve local-only flock/fcntl locks
3096 Description: log_commit_thread vs filter_destroy race leads to crash
3097 Details : Take import reference before releasing llog record semaphore
3102 Description: Wrong request locking in request set processing
3103 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3104 imp_delayed_list, in this place should be used imp_lock.
3107 Frequency : when reconnection
3109 Description: Grant Leak when osc reconnect to OST
3110 Details : When osc reconnect ost, OST(filter) should check whether it
3111 should grant more space to client by comparing fed_grant and
3112 cl_avail_grant, and return the granted space to client instead
3113 of "new granted" space, because client will call osc_init_grant
3114 to update the client grant space info.
3117 Frequency : when client reconnect to OST
3119 Description: Grant Leak when osc do resend and replay bulk write
3120 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3121 bulk write request, because the grant info while be sync between
3122 OSC and OST when reconnect, and we should ignore the grant info
3123 these of resend/replay write req.
3128 Description: Grant space more than available left space sometimes.
3129 Details : When then OST is about to be full, if two bulk writing from
3130 different clients came to OST. Accord the avaliable space of the
3131 OST, the first req should be permitted, and the second one
3132 should be denied by ENOSPC. But if the seconde arrived before
3133 the first one is committed. The OST might wrongly permit second
3134 writing, which will cause grant space > available space.
3137 Frequency : when client is evicted
3139 Description: Grant might be wrongly erased when osc is evicted by OST
3140 Details : when the import is evicted by server, it will fork another
3141 thread ptlrpc_invalidate_import_thread to invalidate the
3142 import, where the grant will be set to 0. While the original
3143 thread will update the grant it got when connecting. So if
3144 the former happened latter, the grant will be wrongly errased
3145 because of this race.
3150 Description: Checking Stale with correct fid
3151 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3152 to check whether it is stale, because sometimes, we want the
3153 enqueue happened anyway, and op_data.fid2 will not be initialized.
3155 Severity : enhancement
3157 Description: update patchless client
3158 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3161 Frequency : only with 2.4 kernel
3163 Description: random memory corruption
3164 Details : size of struct ll_inode_info is to big for union inode.u and this
3165 can be cause of random memory corruption.
3170 Description: Memory leak in recovery
3171 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3172 It should also check obd_fail before fsfilt_start, otherwise if
3173 fsfilt_start return -EROFS,(failover mds during mds recovery).
3174 then the req will return with repmsg->transno = 0 and rc = EROFS.
3175 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3176 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3177 be freed no matter whether fsfilt_commit success or not.
3180 Frequency : only with huge count clients
3182 Description: Prevents from taking the superblock lock in llap_from_page for
3184 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3185 for llap_from_page call in ll_removepage prevents from taking the
3186 superblock lock for a soon died page.
3191 Description: Not check open intent error before release open handle
3192 Details : in some rare cases, the open intent error is not checked before
3193 release open handle, which may cause
3194 ASSERTION(open_req->rq_transno != 0), because it tries to release
3195 the failed open handle.
3200 Description: Set cat log bitmap only after create log success.
3201 Details : in some rare cases, the cat log bitmap is set too early. and it
3202 should be set only after create log success.
3206 Description: Accessing a block bevice can re-enable I/O when Lustre is
3207 tearing down a device.
3208 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3213 Description: service threads may hog cpus when there are a lot of requests
3214 Details : Insert cond_resched to give other threads a chance to use some CPU
3219 Description: the cat log was not initialized in recovery
3220 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3221 unlink log on mds will not be initialized until mds post
3222 recovery. And also in mds post recovery, the unlink log will
3223 initialization will be done asynchronausly, so there will be race
3224 between add unlink log and unlink log initialization.
3228 Description: brw_stats were being printed incorrectly
3229 Details : brw_stats were being printed as log2 but all of them were not
3230 recorded as log2. Also remove some code duplication arising from
3231 filter_tally_{read,write}.
3235 Frequency : rare, only in recovery.
3236 Description: ASSERTION(req->rq_type != LI_POISON) failed
3237 Details : imp_lock should be held while iterating over imp_sending_list for
3238 prevent destroy request after get timeout in ptlrpc_queue_wait.
3242 Description: replay-single.sh test 52 fails
3243 Details : A lock's skiplist need to be cleanup when it being unlinked
3244 from its resource list.
3248 Description: Short directio read returns full requested size rather than
3250 Details : Direct I/O operations should return actual amount of bytes
3251 transferred rather than requested size.
3253 Severity : enhancement
3255 Description: metadata RPC reduction (e.g. for rm performance)
3256 Details : decrease the amount of synchronous RPC between clients and servers
3257 by canceling conflicing lock before the operation on the client
3258 and packing thier handles into the main operation RPC to server.
3260 Severity : enhancement
3262 Description: Async OSC create to avoid the blocking unnecessarily.
3263 Details : If an OST has no remaining object, system will block on the
3264 creation when it needs to create a new object on this OST. Now,
3265 ways use pre-created objects when available, instead of blocking on
3266 an empty osc while others are not empty. If we must block, we
3267 block for the shortest possible period of time.
3271 Description: improve handling recoverable errors
3272 Details : if request processig with error which can be recoverable on server
3273 request should be resend, otherwise page released from cache and
3276 Severity : enhancement
3278 Description: refine locking for avoid write wrong info into lov_objid file
3279 Details : fix possible races with add new target and write/update data in
3282 --------------------------------------------------------------------------------
3284 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3289 Frequency : on some architectures
3291 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3292 Details : Uninitialized flags sometimes cause configuration commands to
3296 Frequency : patchless clients only
3298 Description: missing __iget() symbol export
3299 Details : The __iget() symbol export is missing. To avoid the need for
3300 this on patchless clients the deathrow inode reaper is turned
3301 off, and we depend on the VM to clean up old inodes. This
3302 dependency was during via the fix for bug 12181.
3304 --------------------------------------------------------------------------------
3306 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3308 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3309 INTEROPERATE with older versions automatically. In many cases a
3310 special upgrade step is needed. Please read the
3311 user documentation before upgrading any part of a live system.
3312 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3313 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3314 * WARNING: Lustre configuration and startup changes are required with
3315 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3319 Severity : enhancement
3321 Description: MountConf
3322 Details : Lustre configuration is now managed via mkfs and mount
3323 commands instead of lmc and lconf. New obd types (MGS, MGC)
3324 are added for dynamic configuration management. See
3325 https://mail.clusterfs.com/wikis/lustre/MountConf for
3328 Severity : enhancement
3330 Description: dynamic OST addition
3331 Details : OSTs can now be added to a live filesystem
3333 Severity : enhancement
3335 Description: startup order invariance
3336 Details : MDTs and OSTs can be started in any order. Clients only
3337 require the MDT to complete startup.
3339 Severity : enhancement
3341 Description: parallel, asynchronous orphan cleanup
3342 Details : orphan cleanup is now performed in separate threads for each
3343 OST, allowing parallel non-blocking operation.
3345 Severity : enhancement
3347 Description: optimized stripe assignment
3348 Details : stripe assignments are now made based on ost space available,
3349 ost previous usage, and OSS previous usage, in order to try
3350 to optimize storage space and networking resources.
3352 Severity : enhancement
3354 Description: Permanently set tunables
3355 Details : All writable /proc/fs/lustre tunables can now be permanently
3356 set on a per-server basis, at mkfs time or on a live system.
3358 Severity : enhancement
3360 Description: Lustre message v2
3361 Details : Add lustre message format v2.
3363 Severity : enhancement
3365 Description: client OST exclusion list
3366 Details : Clients can be started with a list of OSTs that should be
3367 declared "inactive" for known non-responsive OSTs.
3371 Description: SPEC SFS validation failure on NFS v2 over lustre.
3372 Details : Changes the blocksize for regular files to be 2x RPC size,
3373 and not depend on stripe size.
3375 Severity : enhancement
3377 Description: Multiple MD RPCs in flight.
3378 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3379 To avoid overly-overloading MDS, introduce a limit on number of
3380 MDS RPCs in flight for a single client and add /proc controls
3381 to adjust this limit.
3383 Severity : enhancement
3385 Description: client read/write statistics
3386 Details : Add client read/write call usage stats for performance
3387 analysis of user processes.
3388 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3389 file access. extents_stats shows chunk size distribution.
3390 extents_stats_per_process show chunk size distribution per
3393 Severity : enhancement
3395 Description: mds statistics
3396 Details : Add detailed mds operations statistics in
3397 /proc/fs/lustre/mds/*/stats.
3401 Description: Failure of copying files with lustre special EAs.
3402 Details : Client side always return success for setxattr call for lustre
3403 special xattr (currently only "trusted.lov").
3408 Description: Refcount LNET uuids
3409 Details : The global LNET uuid list grew linearly with every startup;
3410 refcount repeated list entries instead of always adding to
3413 Severity : enhancement
3415 Description: Dynamic service threads
3416 Details : Within a small range, start extra service threads
3417 automatically when the request queue builds up.
3420 Frequency : mixed-endian client/server environments
3422 Description: mixed-endian crashes
3423 Details : The new msg_v2 system had some failures in mixed-endian
3426 Severity : enhancement
3428 Description: Easy OST removal
3429 Details : OSTs can be permanently deactivated with e.g. 'lctl
3430 conf_param lustre-OST0001.osc.active=0'
3432 Severity : enhancement
3434 Description: MGS proc entries
3435 Details : Added basic proc entries for the MGS showing what filesystems
3438 Severity : enhancement
3440 Description: provide MGS failover
3441 Details : Added config lock reacquisition after MGS server failover.
3443 Severity : enhancement
3445 Description: add Linux 2.4 support
3446 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3450 Description: a large application tries to do I/O to the same resource and dies
3451 in the middle of it.
3452 Details : Check the req->rq_arrival time after the call to
3453 ost_brw_lock_get(), but before we do anything about
3454 processing it & sending the BULK transfer request. This
3455 should help move old stale pending locks off the queue as
3456 quickly as obd_timeout.
3459 Frequency : when an incorrect nid is specified during startup
3461 Description: ptlrpc connect to non-existant node causes kernel crash
3462 Details : LNET can't be re-entered from an event callback, which
3463 happened when we expire a message after the export has been
3464 cleaned up. Instead, hand the zombie cleanup off to another
3467 Severity : enhancement
3469 Description: plain/inodebits lock performance improvement
3470 Details : Grouping plain/inodebits in granted list by their request modes
3471 and bits policy, thus improving the performance of search through
3475 Frequency : only if OST filesystem is corrupted
3477 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3478 Details : for a short time RPCs with bulk IO are in the replay list,
3479 but replay of bulk IOs is unimplemented. If the OST filesystem
3480 is corrupted due to disk cache incoherency and then replay is
3481 started it is possible to trip an assertion. Avoid putting
3482 committed RPCs into the replay list at all to avoid this issue.
3485 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3487 Description: System hang on startup
3488 Details : This bug allowed the liblustre (e.g. catamount) client to
3489 return to the app before handling all startup RPCs. This
3490 could leave the node unresponsive to lustre network traffic
3491 and manifested as a server ptllnd timeout.
3493 Severity : enhancement
3495 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3496 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3497 boolean to control whether to print peer debug info when a
3498 client's RPC times out.
3501 Frequency : only for kernels with patches from Lustre below 1.4.3
3503 Description: Remove old rdonly API
3504 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3507 Frequency : only for devices with external journals
3509 Description: Set external device read-only also
3510 Details : During a commanded failover stop, we set the disk device
3511 read-only while the server shuts down. We now also set any
3512 external journal device read-only at the same time.
3515 Frequency : when upgrading from 1.4 while trying to change parameters
3517 Description: The wrong (new) MDC name was used when setting parameters for
3518 upgraded MDT's. Also allows changing of OSC (and MDC)
3519 parameters if --writeconf is specified at tunefs upgrade time.
3522 Frequency : when setting specific ost indicies
3524 Description: QOS code breaks on skipped indicies
3525 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3526 created with --index need not be sequential.
3528 Severity : enhancement
3530 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3531 Details : The uninit_groups feature works in conjunction with the kernel
3532 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3533 up the pass1 processing of e2fsck. This is a read-only feature
3534 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3535 mount filesystems that have had this feature enabled.
3537 Severity : enhancement
3539 Description: Improve multi-block allocation algorithm to avoid fragmentation
3540 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3541 allocation locality and avoid filesystem fragmentation.
3543 ------------------------------------------------------------------------------
3545 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3547 * Support for kernels:
3548 2.4.21-47.0.1.EL (RHEL 3)
3549 2.6.5-7.283 (SLES 9)
3550 2.6.9-42.0.10.EL (RHEL 4)
3551 2.6.12.6 vanilla (kernel.org)
3552 2.6.16.27-0.9 (SLES 10)
3553 * Recommended e2fsprogs version: 1.39.cfs5
3555 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3559 Frequency : occasional, depends on client load and configuration
3560 Bugzilla : 12181, 12203
3561 Description: data loss for recently-modified files
3563 Details : In some cases it is possible that recently written or created
3564 files may not be written to disk in a timely manner (this should
3565 normally be within 30s unless client IO load is very high).
3566 The problem appears as zero-length files or files that are a
3567 multiple of 1MB in size after a client crash or client eviction
3568 that are missing data at the end of the file.
3570 This problem is more likely to be hit on clients where files are
3571 repeatedly created and unlinked in the same directory, clients
3572 have a large amount of RAM, have many CPUs, the filesystem has
3573 many OSTs, the clients are rebooted frequently, and/or the files
3574 are not accessed by other nodes after being written.
3576 The presence of the problem can be detected by looking at
3577 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3578 smaller than the second (nr_unused) then dirty files will not
3579 be flushed automatically to disk. "sync; sleep 10" should be
3580 run several times on the node before unmounting it to update
3581 Lustre (this is also safe to run on nodes without this problem).
3583 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3584 that can cause this same problem, so customers using that kernel
3585 also need to update the kernel in addition to Lustre. In order
3586 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3588 It is normal that files written just before a client crash (less
3589 than 30s) may not yet have been flushed to disk, even for local
3593 Frequency : frequent on thin XT3 nodes
3595 Description: UUID collision on thin XT3 Linux nodes
3596 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3597 randomly, since we relied on an insufficiently-seeded PRNG.
3602 Description: OSS hangs after "All ost request buffers busy"
3603 Details : A deadlock between quota and journal operations caused OSS
3604 hangs after printing "All ost request buffers busy."
3607 Frequency : always on liblustre builds
3609 Description: Cleanup compiler warnings on liblustre
3612 Frequency : always on liblustre builds on XT3
3614 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3618 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3619 > 32000 subdirectories
3620 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3621 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3622 subdirectory count crosses 32000. This will aid e2fsck to
3623 correctly handle more than 32000 subdirectories.
3626 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3628 Description: System hang on startup
3629 Details : This bug allowed the liblustre (e.g. catamount) client to
3630 return to the app before handling all startup RPCs. This
3631 could leave the node unresponsive to lustre network traffic
3632 and manifested as a server ptllnd timeout.
3634 Severity : enhancement
3636 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3637 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3638 boolean to control whether to print peer debug info when a
3639 client's RPC times out.
3644 Description: make O_SYNC working on 2.6 kernels
3645 Details : 2.6 kernels use different method for mark pages for write,
3646 so need add a code to lustre for O_SYNC work.
3651 Description: Failure to close file and release space on NFS
3652 Details : Put inode details into lock acquired in ll_intent_file_open.
3653 Use mdc_intent_lock in ll_intent_open to properly
3654 detect all kind of errors unhandled by mdc_enqueue.
3659 Description: proc file read during shutdown sometimes raced obd removal,
3661 Details : Add lock to prevent obd access after proc file removal.
3664 Frequency : Only for files larger than 4GB on 32-bit clients.
3666 Description: improperly doing page alignment of locks
3667 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3668 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3673 Description: RPCs being resent when they shouldn't be
3674 Details : Some RPCs that should not be resent are being resent. This
3675 can cause inconsistencies in the RPC state machine. Do not
3676 resend such requests.
3679 Frequency : rare, only with NFS export
3681 Description: Crash on NFS re-export node
3682 Details : under very unusual load conditions an assertion is hit in
3683 ll_intent_file_open()
3686 Frequency : only if OST filesystem is corrupted
3688 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3689 Details : for a short time RPCs with bulk IO are in the replay list,
3690 but replay of bulk IOs is unimplemented. If the OST filesystem
3691 is corrupted due to disk cache incoherency and then replay is
3692 started it is possible to trip an assertion. Avoid putting
3693 committed RPCs into the replay list at all to avoid this issue.
3698 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3699 Details : Large single O_DIRECT read and write calls can fail to allocate
3700 a sufficiently large buffer to process the request. In case of
3701 allocation failure the allocation is retried with a smaller
3702 buffer and broken into smaller requests.
3704 Severity : enhancement
3706 Description: Add -o localflock option to simulate old noflock behaviour.
3707 Details : This will achieve local-only flock/fcntl locks coherentness.
3712 Description: versioning check is incomplete
3713 Details : Checking the version difference of client vs. server, report
3714 error if the gap is too big.
3719 Description: add support PG_writeback bit
3720 Details : add support for PG_writeback bit for Lustre, for more carefull
3721 work with page cache in 2.6 kernel. This also fix some deadlocks
3722 and remove hack for work O_SYNC with 2.6 kernel.
3724 Severity : enhancement
3726 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3727 Details : The uninit_groups feature works in conjunction with the kernel
3728 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3729 up the pass1 processing of e2fsck. This is a read-only feature
3730 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3731 mount filesystems that have had this feature enabled.
3733 Severity : enhancement
3735 Description: Improve multi-block allocation algorithm to avoid fragmentation
3736 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3737 allocation locality and avoid filesystem fragmentation.
3739 ------------------------------------------------------------------------------
3741 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3743 * Support for kernels:
3744 2.6.9-42.0.3.EL (RHEL 4)
3745 2.6.5-7.276 (SLES 9)
3746 2.4.21-47.0.1.EL (RHEL 3)
3747 2.6.12.6 vanilla (kernel.org)
3748 2.6.16.21-0.8 (SLES10)
3749 * Recommended e2fsprogs version: 1.39.cfs2-0
3751 * The backwards-compatible /proc/sys/portals symlink has been removed
3752 in this release. Before upgrading, please ensure that you change
3753 any configuration scripts or /etc/sysctl.conf files that access
3754 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3755 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3756 in advance of the upgrade on any system running Lustre 1.4.6 or
3757 newer, since /proc/sys/lnet was added in that version.
3758 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3762 Frequency : only when quota is used
3764 Description: avoid scanning export list for quota master
3765 Details : Change the algorithms to avoid scanning export list in order
3766 to improve the efficiency.
3769 Frequency : MDS failover only, very rarely
3771 Description: "went back in time" messages on mds failover
3772 Details : The greatest transno may be lost when the current operation
3773 finishes with an error (transno==0) and the client's last_rcvd
3774 record is over-written. Save the greatest transno in the
3775 mds_last_transno for this case.
3778 Frequency : always for specific kernels and striping counts
3780 Description: client may get "Matching packet too big" without ACL support
3781 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3782 when trying to access files in certain configurations. The
3783 clients should in fact be denied when mounting because they do
3784 not understand ACLs.
3787 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3789 Description: many clients connecting with IO in progress causes connect timeouts
3790 Details : Avoid synchronous journal commits to avoid delays caused by many
3791 clients connecting/disconnecting when bulk IO is in progress.
3792 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3793 OST_IO_PORTAL to avoid delays behind potentially many pending
3797 Frequency : occasionally with multiple writers to a single file
3799 Description: shared writes to file may result in wrong size reported by stat()
3800 Details : Allow growing of kms when extent lock is cancelled
3803 Frequency : always with random mmap IO to multi-striped file
3805 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3806 Details : Only if the hole is at the end of OST object so that kms is too
3807 small. Fix is to increase kms accordingly in ll_nopage.
3810 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3812 Description: writes to a missing object would leak memory on the OST
3813 Details : If there is an inconsistency between the MDS and OST filesystems,
3814 such that the MDS references an object that doesn't exist, writes
3815 to that object will leak memory due to incorrect cleanup in the
3816 error handling path, eventually running out of memory on the OST.
3821 Description: Creating too long symlink causes lustre errors
3822 Details : Check symlink and name lengths before sending requests to MDS.
3825 Frequency : only if flock is enabled (not on by default)
3827 Description: posix locks not released on fd closure on 2.6.9+
3828 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3829 kernels, this caused such locks not to be released on fd close and
3830 then assertions on fs unmount about still used locks.
3833 Frequency : MDS failover only, very rarely
3835 Description: clients may get ASSERTION(granted_lock != NULL)
3836 Details : When request was taking a long time, and a client was resending
3837 a getattr by name lock request. The were multiple lock requests
3838 with the same client lock handle and
3839 mds_getattr_name->fixup_handle_for_resent_request found one of the
3840 lock handles but later failed with ASSERTION(granted_lock != NULL).
3845 Description: handle->h_buffer_credits > 0, assertion failure
3846 Details : h_buffer_credits is zero after truncate, causing assertion
3847 failure. This patch extends the transaction or creates a new
3851 Frequency : NFS re-export or patchless client
3852 Bugzilla : 11179, 10796
3853 Description: Crash on NFS re-export node (__d_move)
3854 Details : We do not want to hash the dentry if we don't have a lock.
3855 But if this dentry is later used in d_move, we'd hit uninitialised
3856 list head d_hash, so we just do this to init d_hash field but
3857 leave dentry unhashed.
3860 Frequency : NFS re-export or patchless client
3862 Description: NFS exports has problem with symbolic link
3863 Details : lustre client didn't properly install dentry when re-exported
3864 to NFS or running patchless client.
3867 Frequency : NFS re-export or patchless client
3869 Description: Various nfs/patchless fixes.
3870 Details : fixes reuse disconected alias for lookup process - this fixes
3871 warning "find_exported_dentry: npd != pd",
3872 fix permission error with open files at nfs.
3873 fix apply umask when do revalidate.
3876 Frequency : occasional
3878 Description: Crash on NFS re-export node
3879 Details : calling clear_page() on the wrong pointer triggered oops in
3880 generic_mapping_read().
3883 Frequency : rarely, using O_DIRECT IO
3885 Description: unaligned directio crashes client with LASSERT
3886 Details : check for unaligned buffers before trying any requests.
3889 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3891 Description: stale data returned from RAID cache
3892 Details : If only a small amount of IO is done to the RAID device before
3893 reading it again it is possible to get stale data from the RAID
3894 cache instead of reading it from disk.
3897 Frequency : always for sles10 kernel
3899 Description: sles10 support
3900 Details : ll_follow_link: compile fixes and using of nd_set_link
3901 under newer kernels.
3904 Frequency : depends on arch, kernel and compiler version, always on sles10
3907 Description: recursive or deep enough symlinks cause stack overflow
3908 Details : getting rid of large stack-allocated variable in
3912 Frequency : depends on hardware
3914 Description: lustre write performance loss in the SLES10 kernel
3915 Details : the performance loss is caused by using of write barriers in the
3916 ext3 code. The SLES10 kernel turns barrier support on by
3917 default. The fix is to undo that change for ldiskfs.
3919 ------------------------------------------------------------------------------
3921 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3923 * Support for kernels:
3924 2.6.9-42.0.3EL (RHEL 4)
3925 2.6.5-7.276 (SLES 9)
3926 2.4.21-47.0.1.EL (RHEL 3)
3927 2.6.12.6 vanilla (kernel.org)
3931 Frequency : quota enabled and large files being deleted
3933 Description: releasing more than 4GB of quota at once hangs OST
3934 Details : If a user deletes more than 4GB of files on a single OST it
3935 will cause the OST to spin in an infinite loop. Release
3936 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3941 Description: statfs data retrieved from /proc may be stale or zero
3942 Details : When reading per-device statfs data from /proc, in the
3943 {kbytes,files}_{total,free,avail} files, it may appear
3944 as zero or be out of date.
3947 Frequency : systems with MD RAID1 external journal devices
3949 Description: lconf's call to blkid is confused by RAID1 journal devices
3950 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3951 of returning all block devices that match the journal UUID.
3954 Frequency : always, for aggregate stripe size over 4GB
3956 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3957 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3958 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3961 Frequency : always if "lfs find" used on a local file/directory
3963 Description: "lfs find" segfaults if used on a local file/directory
3964 Details : The case where a directory component was not specified wasn't
3965 handled correctly. Handle this properly.
3968 Frequency : always on ppc64
3970 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3971 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3972 perform endianness conversion.
3975 Frequency : rarely (truncate to non-zero file size after write under load)
3976 Bugzilla : 10730, 10687
3977 Description: Files padded with zeros to next 4K multiple
3978 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3979 it is possible that files that are truncated to a non-zero size
3980 immediately after being written are filled with zero bytes beyond
3981 the truncated size. No file data is lost.
3983 Severity : enhancement
3985 Description: Allow recovery/failover for liblustre clients.
3986 Details : liblustre clients were unaware of failover configurations until
3989 Severity : enhancement
3991 Description: user file locks should fail when not mounting with flock option
3992 Details : Set up an error-returning stub in ll_file_operations.lock field
3993 to prevent incorrect behaviour when client is mounted without
3994 flock option. Also, set up properly f_op->flock field for
3998 Frequency : always on ia64
4000 Description: "lfs df" loops on printing out MDS statfs information
4001 Details : The obd_ioctl_data was not initialized and in some systems
4002 this caused a failure during the ioctl that did not return
4003 an error. Initialize the struct and return an error on failure.
4006 Frequency : SLES 9 only
4008 Description: Error of copying files with lustre special EAs as root
4009 Details : Client side always return success for setxattr call for lustre
4010 special xattr (currently only "trusted.lov").
4013 Frequency : rarely on clusters with both ia64+i386 clients
4015 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4016 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4017 clients are doing concurrent writes to the same file it is
4018 possible that the ia64 clients may LASSERT because the OST
4019 extent locks are not PAGE_SIZE aligned. Ensure that grown
4020 locks are always aligned on the request boundary.
4023 Frequency : specific use, occasional
4025 Description: Overwriting in use executable truncates on-disk binary image
4026 Details : If one node attempts to overwrite an executable in use by
4027 another node, we now correctly return ETXTBSY instead of
4028 truncating the file.
4030 Severity : enhancement
4032 Description: Async OSC create to avoid the blocking unnecessarily.
4033 Details : If an OST has no remaining object, system will block on the creating
4034 when need to create a new object on this OST. Now, ways use
4035 pre-created objects when available, instead of blocking on an
4036 empty osc while others are not empty. If we must block, we block
4037 for the shortest possible period of time.
4042 Description: chmod on Lustre root is propagated to other clients
4043 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4044 invalid by the follow_mount time.
4047 Frequency : liblustre clients only
4049 Description: Race in 'instant cancel' lock handling could lead to such locks
4050 never to be granted in case of SMP MDS
4051 Details : Do not destroy not yet granted but cbpending locks in
4055 Frequency : replay/resend of open
4057 Description: non null lock assetion failure in mds_intent_policy
4058 Details : Trying to replay/resend lockless open requests resulted in
4059 mds_open() returning 0 with no lock. Now it sets a flag if
4060 a lock is going to be returned.
4062 Severity : enhancement
4064 Description: Checksum enhancements
4065 Details : New checksum enhancements allow for resending RPCs that failed
4068 Severity : enhancement
4070 Description: Tunables on number of dirty pages in cacche
4071 Details : Allow to set limit on number of dirty pages cached.
4076 Description: client crash on unmount - lock still has references
4077 Details : In some error handling cases it was possible to leak a lock
4078 reference on a client while accessing a file. This was not
4079 harmful to the client during operation, but would cause the
4080 client to crash when the filesystem is unmounted.
4083 Frequency : specific case, rare
4085 Description: ETXTBSY on mds though file not in use by client
4086 Details : ETXTBSY is no longer incorrectly returned when attempting to
4087 chmod or chown a directory that the user previously tried to
4088 execute or a currently-executing binary.
4091 Frequency : extremely rare except on liblustre-based clients
4093 Description: Lustre space not freed when files are deleted
4094 Details : Clean up open-unlinked files after client eviction. Previously
4095 the unlink was skipped and the files remained as orphans.
4100 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4101 Details : In some cases it was possible to send a blocking callback to
4102 a client doing a glimpse, even though that client didn't get
4103 a lock granted. When the glimpse lock is cancelled on the OST
4104 the freed lock is left on the waiting list and corrupted the list.
4107 Frequency : all core dumps
4109 Description: Broke core dumps to lustre
4110 Details : Negative dentry may be unhashed if parent does not have UPDATE
4111 lock, but some callers, e.g. do_coredump, expect dentry to be
4112 hashed after successful create, hash it in ll_create_it.
4114 ------------------------------------------------------------------------------
4116 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4118 * Support for kernels:
4119 2.6.9-42.0.2.EL (RHEL 4)
4120 2.6.5-7.276 (SLES 9)
4121 2.4.21-40.EL (RHEL 3)
4122 2.6.12.6 vanilla (kernel.org)
4126 Frequency : always on RHEL 3
4128 Description: Number of open files grows over time
4129 Details : The number of open files grows over time, whether or not
4130 Lustre is started. This was due to a filp leak introduced
4131 by one of our kernel patches.
4133 ------------------------------------------------------------------------------
4135 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4137 * Support for kernels:
4138 2.6.9-42.EL (RHEL 4)
4139 2.6.5-7.276 (SLES 9)
4140 2.4.21-40.EL (RHEL 3)
4141 2.6.12.6 vanilla (kernel.org)
4146 Bugzilla : 5719, 9635, 9792, 9684
4147 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4148 Details : If a server is under heavy load and cannot reply to new
4149 connection requests before the client resends the (re)connect,
4150 the connection handling code can behave badly if two service
4151 threads are concurrently handing separate (re)connections from
4152 the same client. Add better locking to the connection handling
4153 code, and ensure that only a single connection will be processed
4154 for a given client UUID, even if the lock is dropped.
4156 Severity : enhancement
4158 Description: add TCP zero-copy support to kernel
4159 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4160 sends if the hardware supports scatter-gather and checksumming.
4161 This allows socklnd to do client-write and server-read more
4162 efficiently and reduce CPU utilization from skbuf copying.
4165 Frequency : only if NFS exporting from client
4167 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4168 Details : If setting attributes on a file created under NFS that had
4169 never been opened it would be possible to oops the client
4170 if the file had no objects.
4173 Frequency : always for liblustre
4175 Description: liblustre client does MDS+OSTs setattr RPC for each write
4176 Details : When doing a write from a liblustre client, the client
4177 incorrectly issued an RPC to the MDS and each OST the file was
4178 striped over in order to update the timestamps. When writing
4179 with small chunks and many clients this could overwhelm the MDS
4180 with RPCs. In all cases it would slow down the write because
4181 these RPCs are unnecessary.
4183 Severity : enhancement
4185 Description: allow number of MDS service threads to be changed at module load
4186 Details : It is now possible to change the number of MDS service threads
4187 running. Adding "options mds mds_num_threads={N}" to the MDS's
4188 /etc/modprobe.conf will set the number of threads for the next
4189 time Lustre is restarted (assuming the "mds" module is also
4190 reloaded at that time). The default number of threads will
4191 stay the same, 32 for most systems.
4196 Description: OST crash if filesystem is unformatted or corrupt
4197 Details : If an OST is started on a device that has never been formatted
4198 or if the filesystem is corrupt and cannot even mount then the
4199 error handling cleanup routines would dereference a NULL pointer.
4204 Description: NULL pointer deref in llap_from_page.
4205 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4206 incorrect) mapping if the page was truncated/reclaimed while it was
4207 searched for. Check for this condition and skip such pages when
4208 doing readahead. Introduce extra check to llap_from_page() to
4209 verify page->mapping->host is non-NULL (so page is not anonymous).
4212 Frequency : Sometimes when using sys_sendfile
4214 Description: "page not covered by a lock" warnings from ll_readpage
4215 Details : sendfile called ll_readpage without right page locks present.
4216 Now we introduced ll_file_sendfile that does necessary locking
4217 around call to generic_file_sendfile() much like we do in
4221 Frequency : with certain MDS communication failures at client mount time
4223 Description: NULL pointer deref after failed client mount
4224 Details : a client connection request may delayed by the network layer
4225 and not be sent until after the PTLRPC layer has timed out the
4226 request. If the client fails the mount immediately it will try
4227 to clean up before the network times out the request. Add a
4228 reference from the request import to the obd device and delay
4229 the cleanup until the network drops the request.
4232 Frequency : occasionally during client (re)connect
4234 Description: assertion failure during client (re)connect
4235 Details : processing a client connection request may be delayed by the
4236 client or server longer than the client connect timeout. This
4237 causes the client to resend the connection request. If the
4238 original connection request is replied in this interval, the
4239 client may trip an assertion failure in ptlrpc_connect_interpret()
4240 which thought it would be the only running connect process.
4243 Frequency : only with obd_echo servers and clients that are rebooted
4245 Description: kernel BUG accessing uninitialized data structure
4246 Details : When running an obd_echo server it did not start the ping_evictor
4247 thread, and when a client was evicted an uninitialized data
4248 structure was accessed. Start the ping_evictor in the RPC
4249 service startup instead of the OBD startup.
4251 Severity : enhancement
4252 Bugzilla : 10193 (patchless)
4253 Description: Remove dependency on various unexported kernel interfaces.
4254 Details : No longer need reparent_to_init, exit_mm, exit_files,
4255 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4258 Frequency : rare (only users of deprecated and unsupported LDAP config)
4260 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4261 Details : LDAP apparently contains 'lustreName' attributes instead of
4262 'name'. A simple remapping of the name is sufficient.
4265 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4267 Description: waiting_locks_callback trips kernel BUG if client is evicted
4268 Details : Running with the dump_on_timeout debug flag turned on makes
4269 it possible that the waiting_locks_callback() can try to dump
4270 the Lustre kernel debug logs from an interrupt handler. Defer
4271 this log dumping to the expired_lock_main() thread.
4273 Severity : enhancement
4275 Description: Support NFS exporting on 2.6 kernels.
4276 Details : Implement non-rawops metadata methods for NFS server to use without
4277 changing NFS server code.
4280 Frequency : very rare (synthetic metadata workload only)
4282 Description: two racing renames might cause an MDS thread to deadlock
4283 Details : Running the "racer" program may cause one MDS thread to rename
4284 a file from being the source of a rename to being the target of
4285 a rename at exactly the same time that another thread is doing
4286 so, and the second thread has already enqueued these locks after
4287 doing a lookup of the target and is trying to relock them in
4288 order. Ensure that we don't try to re-lock the same resource.
4291 Frequency : only very large systems with liblustre clients
4293 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4294 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4295 all OSTs in parallel. This allows the eviction of stale liblustre
4296 clients to proceed much faster than if they were done in series,
4297 and also offers similar improvements for other set_info RPCs.
4302 Description: excessive CPU usage during initial read phase on client
4303 Details : During the initial read phase on a client, it would agressively
4304 retry readahead on the file, consuming too much CPU and impacting
4305 performance (since 1.4.5.8). Improve the readahead algorithm
4306 to avoid this, and also improve some other common cases (read
4307 of small files in particular, where "small" is files smaller than
4308 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4313 Description: MDS crash when receiving packet with unknown intent.
4314 Details : Do not LBUG in unknown intent case, just return -EFAULT
4316 Severity : enhancement
4317 Bugzilla : 9293, 9385
4318 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4319 Details : Do not serialize getattr (non-intent version) and statfs.
4322 Frequency : occasional, when OST network is overloaded/intermittent
4324 Description: client evicted by OST after bulk IO timeout
4325 Details : If a client sends a bulk IO request (read or write) the OST
4326 may evict the client if it is unresposive to its data GET/PUT
4327 request. This is incorrect if the network is overloaded (takes
4328 too long to transfer the RPC data) or dropped the OST GET/PUT
4329 request. There is no need to evict the client at all, since
4330 the pinger and/or lock callbacks will handle this, and the
4331 client can restart the bulk request.
4334 Frequency : Always when mmapping file with no objects
4336 Description: client crashes when mmapping file with no objects
4337 Details : Check that we actually have objects in a file before doing any
4338 operations on objects in ll_vm_open, ll_vm_close and
4344 Description: Request leak when working with deleted CWD
4345 Details : Introduce advanced request refcount tracking for requests
4346 referenced from lustre intent.
4348 Severity : Enhancement
4350 Description: Cache open file handles on client.
4351 Details : MDS now will return special lock along with openhandle, if
4352 requested and client is allowed to hold openhandle, even if unused,
4353 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4354 closing files for every read/write openration.
4356 Severity : Enhancement
4358 Description: Cache open negative dentries on client when possible.
4359 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4360 negative dentries on lock revocation.
4365 Description: Remounting a client read-only wasn't possible with a zconf mount
4366 Details : It wasn't possible to remount a client read-only with llmount.
4368 Severity : enhancement
4369 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4370 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4371 MPICH is included to add Lustre-specific ADIO interfaces.
4372 This is based closely on the UFS ADIO layer and only differs
4373 in file creation, in order to allow the OST striping to be set.
4374 This is user-contributed code and not supported by CFS.
4379 Description: extended inode attributes (immutable, append-only) work improperly
4380 when 2.4 and 2.6 kernels are used on client/server or vice versa
4381 Details : Introduce kernel-independent values for these flags.
4383 Severity : enhancement
4386 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4387 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4388 granularity. Now, Fractional MB granularity is be supported,
4389 this is very useful for low memory system.
4391 Severity : enhancement
4393 Description: Getattr by fid
4394 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4395 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4396 have locks and avoids getattr rpc requests for it on every stat.
4399 Frequency : Always, for filesystems larger than 2TB
4401 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4402 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4403 filesystems larger than 2TB to be mounted with mballoc enabled.
4406 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4408 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4409 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4410 filesystems larger than 2TB is mounted then the kernel will
4411 silently corrupt the start of the filesystem. CONFIG_LBD is
4412 enabled for all CFS-supported kernels, but the possibility of
4413 this happening with a modified kernel config exists.
4415 Severity : enhancement
4417 Description: add client O_DIRECT support for 2.6 kernels
4418 Details : It is now possible to do O_DIRECT reads and writes to files
4419 in the Lustre client mountpoint on 2.6 kernel clients.
4421 Severity : enhancement
4423 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4424 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4425 parallel, not waiting for response from every OST before sending
4426 a rpc to the next OST.
4431 Description: setattr vs write race when updating file timestamps
4432 Details : Client processes that update a file timestamp into the past
4433 right after writing to the file (e.g. tar) it is possible that
4434 the updated file modification time can be reset to the current
4435 time due to a race between processing the setattr and write RPC.
4437 Severity : enhancement
4439 Description: Bring 'lfs find' closer in line with regular Linux find.
4440 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4441 -print0 options and obtains all the needed info through the lustre
4444 Severity : enhancement
4446 Description: support up to 1024 configured devices on one node
4447 Details : change obd_dev array from statically allocated to dynamically
4448 allocated structs as they are first used to reduce memory usage
4453 Description: Flush dirty partially truncated pages during truncate
4454 Details : Immediatelly flush partially truncated pages in filter_setattr,
4455 this way we completely avoid having any pages in page cache on OST
4456 and can retire ugly workarounds during writes to flush such pages.
4461 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4462 Details : protect inode from truncation within vfs_unlink() context
4463 just take a reference before calling vfs_unlink() and release it
4464 when parent's i_sem is free.
4469 Description: last_id value checked outside lock on OST caused LASSERT failure
4470 Details : If there were multiple MDS->OST object precreate requests in
4471 flight, it was possible that the OST's last object id was checked
4472 outside a lock and incorrectly tripped an assertion. Move checks
4473 inside locks, and discard old precreate requests.
4476 Frequency : always, if extents are used on OSTs
4478 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4479 Details : OSTs using the extents format would not zero the high 16 bits of
4480 the index physical block number. This is not a problem for any
4481 OST filesystems smaller than 16TB, and no kernels support ext3
4482 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4483 new/modified files) and can be fixed for existing filesystems
4484 with e2fsprogs-1.39-cfs1.
4489 Description: import connection selection may be incorrect if timer wraps
4490 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4491 import connections to be ignored if the 32-bit jiffies counter
4492 wraps. Use a 64-bit jiffies counter.
4495 Frequency : very large clusters immediately after boot
4497 Description: LNET request buffers exhausted under heavy short-term load
4498 Details : If a large number of client requests are generated on a service
4499 that has previously never seen so many requests it is possible
4500 that the request buffer growth cannot keep up with the spike in
4501 demand. Instead of dropping incoming requests, they are held in
4502 the LND until the RPC service can accept more requests.
4505 Frequency : Sometimes during replay
4507 Description: Assertion failure in ll_local_open after replay.
4508 Details : If replay happened on an open request reply before we were able
4509 to set replay handler, reply will become not swabbed tripping the
4510 assertion in ll_local_open. Now we set the handler right after
4511 recognising of open request
4514 Frequency : very rare
4516 Description: kernel reports "badness in vsnprintf"
4517 Details : Reading from the "recovery_status" /proc file in small chunks
4518 may cause a negative length in lprocfs_obd_rd_recovery_status()
4519 call to vsnprintf() (which is otherwise harmless). Exit early
4520 if there is no more space in the output buffer.
4522 Severity : enhancement
4524 Description: clear OBD RPC statistics by writing to them
4525 Details : It is now possible to clear the OBD RPC statistics by writing
4526 to the "stats" file.
4531 Description: Client mtime is not the same on different clients after utimes
4532 Details : In some cases, the client was using the utimes() syscall on
4533 a file cached on another node. The clients now validate the
4534 ctime from the MDS + OSTs to determine which one is right.
4539 Description: Inability to activate failout mode
4540 Details : lconf script incorrectly assumed that in pythong string's numeric
4541 value is used in comparisons.
4544 Frequency : always with multiple stripes per file
4546 Description: Inefficient object allocation for mutli-stripe files
4547 Details : When selecting which OSTs to stripe files over, for files with
4548 a stripe count that divides evenly into the number of OSTs,
4549 the MDS is always picking the same starting OST for each file.
4550 Return the OST selection heuristic to the original design.
4555 Description: mount failures may take full timeout to return an error
4556 Details : Under some heavy load conditions it is possible that a
4557 failed mount can wait for the full obd_timeout interval,
4558 possibly several minutes, before reporting an error.
4559 Instead return an error as soon as the status is known.
4561 Frequency : quota enabled and large files being deleted
4563 Description: releasing more than 4GB of quota at once hangs OST
4564 Details : If a user deletes more than 4GB of files on a single OST it
4565 will cause the OST to spin in an infinite loop. Release
4566 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4571 Description: statfs data retrieved from /proc may be stale or zero
4572 Details : When reading per-device statfs data from /proc, in the
4573 {kbytes,files}_{total,free,avail} files, it may appear
4574 as zero or be out of date.
4577 Frequency : systems with MD RAID1 external journal devices
4579 Description: lconf's call to blkid is confused by RAID1 journal devices
4580 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4581 of returning all block devices that match the journal UUID.
4584 Frequency : always, for aggregate stripe size over 4GB
4586 Description: assertion fails when trying to use 4GB stripe size
4587 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4588 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4591 Frequency : always on ppc64
4593 Description: the first write on an ext3 filesystem with mballoc got stuck
4594 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4595 perform endianness conversion.
4597 ------------------------------------------------------------------------------
4599 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4601 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4602 INTEROPERATE with older versions automatically. Please read the
4603 user documentation before upgrading any part of a live system.
4604 * WARNING: Lustre networking configuration changes are required with
4605 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4608 * Support for kernels:
4609 2.6.9-22.0.2.EL (RHEL 4)
4610 2.6.5-7.244 (SLES 9)
4611 2.6.12.6 vanilla (kernel.org)
4614 Severity : enhancement
4615 Bugzilla : 7981/8208
4616 Description: Introduced Lustre Networking (LNET)
4617 Details : LNET is new networking infrastructure for Lustre, it includes
4618 a reorganized network configuration mode (see the user
4619 documentation for full details) as well as support for routing
4620 between different network fabrics. Lustre Networking Devices
4621 (LNDs) for the supported network fabrics have also been
4622 created for this new infrastructure.
4624 Severity : enhancement
4625 Description: Introduced Access control lists
4626 Details : clients can set ACLs on files and directories in order to have
4627 more fine-grained permissions than the standard Unix UGO+RWX.
4628 The MDS must be started with the "-o acl" mount option.
4630 Severity : enhancement
4631 Description: Introduced filesystem quotas
4632 Details : Administrators may now establish per-user quotas on the
4635 Severity : enhancement
4637 Description: Configuration change for the XT3
4638 The PTLLND is now used to run Lustre over Portals on the XT3
4639 The configure option(s) --with-cray-portals are no longer used.
4640 Rather --with-portals=<path-to-portals-includes> is used to
4641 enable building on the XT3. In addition to enable XT3 specific
4642 features the option --enable-cray-xt3 must be used.
4647 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4648 Details : A race between allocating a new llcd and re-getting the llcd_lock
4649 allowed another thread to grab newly-allocated llcd.
4651 Severity : enhancement
4653 Description: 2.6 OST async journal commit and locking fix to improve performance
4654 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4655 2.6 kernels are now async as they already were in 2.4 kernels so
4656 that they can commit concurrently with the network bulk transfer.
4657 For block-allocated files the filter allocation semaphore is held
4658 to avoid filesystem fragmentation during allocation. BKL lock
4659 removed for 2.6 xattr operations where it is no longer needed.
4664 Description: lconf incorrectly determined whether two IP networks could talk
4665 Details : In some more complicated routing and multiple-network
4666 configurations, lconf will avoid trying to make a network
4667 connection to a disjoint part of the IP space. It was doing the
4668 math incorrectly for one set of cases.
4673 Description: Fix for potential infinite loop processing records in an llog.
4674 Details : If an llog record is corrupted/zeroed, it is possible to loop
4675 forever in llog_process(). Validate the llog record length
4676 and skip the remainder of the block on error.
4679 Frequency : occasional (liblustre only)
4681 Description: liblustre could not open files whose last component is a symlink
4682 Details : sysio_path_walk() would incorrectly pass the open intent to
4683 intermediate path components.
4686 Frequency : rare (liblustre only with non-standard tuning)
4687 Bugzilla : 7201 (7350)
4688 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4689 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4690 after it has been released, only do this for referenced locks
4692 Severity : enhancement
4694 Description: specifying an (invalid) directory default stripe_size of -1
4695 would reset the directory default striping
4696 Details : stripe_size -1 was used internally to signal directory stripe
4697 removal, now use "all default" to signal dir stripe removal
4698 as a directory striping of "all default" is not useful
4701 Frequency : common for large clusters running liblustre clients
4703 Description: doing an ls when liblustre clients are running is slow
4704 Details : sending a glimpse AST to a liblustre client waits for every AST
4705 to time out, as liblustre clients will not respond. Since they
4706 cannot cache data we refresh the OST lock LVB from disk instead.
4708 Severity : enhancement
4710 Description: doing an ls at the same time as file IO can be slow
4711 Details : enqueue and other "small" requests can be blocked behind many
4712 large IO requests. Create a new OST IO portal for non-IO
4713 requests so they can be processed faster.
4716 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4718 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4719 created file with mode 000
4720 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4721 allow an ftruncate() as a non-root user to a file with mode 000.
4722 Lustre now respects this flag to disable mode checks when
4723 truncating a file owned by the user
4726 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4729 Description: Revoking locks from clients that went dead or catatonic might take
4731 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4732 cancellation of such locks instant on servers without waiting for
4733 any reply from clients. Clients drops these locks when cancel
4734 notification from server is received without replying.
4737 Frequency : liblustre-only, when liblustre client dies or becomes busy
4739 Description: Doing ls on Linux clients can take a long time with active
4741 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4742 granting such locks to it in the first place if possible. Locks
4743 are taken by proxy on the OST during the read or write and
4744 dropped immediately afterward. Add connect flags handling, do
4745 not grant locks to liblustre clients for glimpse ASTs.
4747 Severity : enhancement
4749 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4750 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4751 sequential reads and behaves badly with applications doing random
4752 reads. Improve it by reading ahead at least read region, and
4753 avoiding excessive large RPC for small reads.
4755 Severity : enhancement
4757 Description: Creating more than 1000 files for a single job may cause a load
4758 imbalance on the OSTs if there are also a large number of OSTs.
4759 Details : qos_prep_create() uses an OST index reseed value that is an
4760 even multiple of the number of available OSTs so that if the
4761 reseed happens in the middle of the object allocation it will
4762 still utilize the OSTs as uniformly as possible.
4767 Description: OST or MDS may oops in ping_evictor_main()
4768 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4769 but doesn't restart at beginning of obd_exports_timed list
4772 Severity : enhancement
4774 Description: improve by-nid export eviction on the MDS and OST
4775 Details : allow multiple exports with the same NID to be evicted at one
4776 time without re-searching the exports list.
4779 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4781 Description: MDS may oops in groups_free()
4782 Details : in rare race conditions a newly allocated group_info struct is
4783 freed again, and this can be NULL. The 2.4 compatibility code
4784 for groups_free() checked for a NULL pointer, but 2.6 did not.
4787 Frequency : common for liblustre clients doing little filesystem IO
4788 Bugzilla : 9352, 7313
4789 Description: server may evict liblustre clients accessing contended locks
4790 Details : if a client is granted a lock or receives a completion AST
4791 with a blocking AST already set it would not reply to the AST
4792 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4795 Frequency : lfs setstripe, only systems with more than 160 OSTs
4797 Description: unable to set striping with a starting offset beyond OST 160
4798 Details : llapi_create_file() incorrectly limited the starting stripe
4799 index to the maximum single-file stripe count.
4802 Frequency : LDAP users only
4804 Description: lconf did not handle in-kernel recovery with LDAP properly
4805 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4807 Severity : enhancement
4809 Description: bind OST threads to NUMA nodes to improve performance
4810 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4811 node and do their allocations there to localize memory access
4813 Severity : enhancement
4815 Description: llmount can determine client NID directly from Myrinet (GM)
4816 Details : the client NID code from gmnalnid was moved directly into
4817 llmount, removing the need to use this or specifying the
4818 client NID explicitly when mounting GM clients with zeroconf
4821 Frequency : if client is started with down MDS
4823 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4824 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4825 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4826 ptlrpc_import_delayed_req() because we didn't check if the
4827 request was interrupted, and we also didn't break out of the
4828 event loop if there was no timeout
4833 Description: data loss during non-page-aligned writes to a single file from
4834 both multiple nodes and multiple threads on one node at same time
4835 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4836 inconsistency led to false short-reads, that were cached and later
4837 used by ->prepare_write() to fill in partially written page,
4838 leading to data loss.
4841 Frequency : always, if lconf --abort_recovery used
4843 Description: lconf --abort_recovery fails with 'Operation not supported'
4844 Details : lconf was attempting to abort recovery on the MDT device and not
4847 Severity : enhancement
4849 Description: remove cleanup logs
4850 Details : replace lconf-generated cleanup logs with lustre internal
4851 cleanup routines. Eliminates the need for client-cleanup and
4854 Severity : enhancement
4856 Description: add support for EAs (user and system) on lustre filesystems
4857 Details : it is now possible to store extended attributes in the Lustre
4858 client filesystem, and with the user_xattr mount option it
4859 is possible to allow users to store EAs on their files also
4861 Severity : enhancement
4863 Description: Add possibility (config option) to show minimal available OST free
4865 Details : When compiled with --enable-mindf configure option, statfs(2)
4866 (and so, df) will return least minimal free space available from
4867 all OSTs as amount of free space on FS, instead of summary of
4868 free spaces of all OSTs.
4870 Severity : enhancement
4872 Description: do not expand extent locks acquired on OST-side
4873 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4874 by server: they are not cached anyway.
4877 Frequency : when mmap is used/binaries executed from Lustre
4879 Description: Unmmap pages before throwing them away from read cache.
4880 Details : llap_shrink cache now attempts to unmap pages before discarding
4881 them (if unmapping failed - do not discard). SLES9 kernel has
4882 extra checks that trigger if this unmapping is not done first.
4887 Description: lconf didn't resolve symlinks before checking to see whether a
4888 given mountpoint was already in use
4891 Frequency : when migrating failover services
4892 Bugzilla : 6395, 9514
4893 Description: When migrating a subset of services from a node (e.g. failback
4894 from a failover service node) the remaining services would
4895 time out and evict clients.
4896 Details : lconf --force (implied by --failover) sets the global obd_timeout
4897 to 5 seconds in order to quickly disconnect, but this caused
4898 other RPCs to time out too quickly. Do not change the global
4899 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4901 Severity : enhancement
4902 Frequency : if MDS is started with down OST
4903 Bugzilla : 9439,5706
4904 Description: Allow startup/shutdown of an MDS without depending on the
4905 availability of the OSTs.
4906 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4907 Add appropriate locking and lov-osc refcounts for safe
4908 cleaning. Add osc abort_inflight calls in case the
4909 synchronize never started.
4912 Frequency : occasional (Cray XT3 only)
4914 Description: root not authorized to access files in CRAY_PORTALS environment
4915 Details : The client process capabilities were not honoured on the MDS in
4916 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4917 been accessed by an authorized user then root was able to access
4918 the file on the local client also. The root user capabilities
4919 are now allowed on the MDS, as this environment has secure UID.
4922 Frequency : occasional
4924 Description: ldiskfs "too long searching" message happens too often
4925 Details : A debugging message (otherwise harmless) prints too often on
4926 the OST console. This has been reduced to only happen when
4927 there are fragmentation problems on the filesystem.
4932 Description: Division by zero in statfs when all OSCs are inactive
4933 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4934 lov_statfs divides by value returned by lov_get_stripecnt().
4938 Bugzilla : 9489, 3273
4939 Description: First write from each client to each OST was only 4kB in size,
4940 to initialize client writeback cache, which caused sub-optimal
4941 RPCs and poor layout on disk for the first written file.
4942 Details : Clients now request an initial cache grant at (re)connect time
4943 and so that they can start streaming writes to the cache right
4944 away and always do full-sized RPCs if there is enough data.
4945 If the OST is rebooted the client also re-establishes its grant
4946 so that client cached writes will be honoured under the grant.
4951 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4952 Details : Now I/O RPCs go to different portal number and (presumably) fast
4953 lock requests (and glimses) and other RPCs get their own service
4954 threads pool that should be able to service those RPCs
4957 Severity : enhancement
4959 Description: Ability to exchange lustre version between client and servers and
4960 issue warnings at client side if client is too old. Also for
4961 liblustre clients there is ability to refuse connection of too old
4963 Details : New 'version' field is added to connect data structure that is
4964 filled with version info. That info is later checked by server and
4968 Frequency : rare, liblustre only.
4969 Bugzilla : 9296, 9581
4970 Description: Two simultaneous writes from liblustre at offset within same page
4971 might proceed at the same time overwriting eachother with stale
4973 Details : I/O lock withing llu_file_prwv was released too early, before data
4974 actually was hitting the wire. Extended lock-holding time until
4975 server acknowledges receiving data.
4978 Frequency : extremely rare. Never observed in practice.
4980 Description: avoid generating lustre_handle cookie of 0.
4981 Details : class_handle_hash() generates handle cookies by incrementing
4982 global counter, and can hit 0 occasionaly (this is unlikely, but
4983 not impossible, because initial value of cookie counter is
4984 selected randonly). Value of 0 is used as a sentinel meaning
4985 "unassigned handle" --- avoid it. Also coalesce two critical
4986 sections in this function into one.
4988 Severity : enhancement
4990 Description: allow liblustre clients to delegate truncate locking to OST
4991 Details : To avoid overhead of locking, liblustre client instructs OST to
4992 take extent lock in ost_punch() on client's behalf. New connection
4993 flag is added to handle backward compatibility.
4995 Severity : enhancement
4996 Bugzilla : 4928, 7341, 9758
4997 Description: allow number of OST service threads to be specified
4998 Details : a module parameter allows the number of OST service threads
4999 to be specified via "options ost ost_num_threads={N}" in the
5000 OSS's /etc/modules.conf or /etc/modprobe.conf.
5004 Bugzilla : 6146, 9635, 9895
5005 Description: servers crash with bad pointer in target_handle_connect()
5006 Details : In rare cases when a client is reconnecting it was possible that
5007 the connection request was the last reference for that export.
5008 We would temporarily drop the export reference and get a new
5009 one, but this may have been the last reference and the export
5010 was just destroyed. Get new reference before dropping old one.
5012 Severity : enhancement
5013 Frequency : if client is started with failover MDS
5015 Description: Allow multiple MDS hostnames in the mount command
5016 Details : Try to read the configuration from all specified MDS
5017 hostnames during a client mount in case the "primary"
5020 Severity : enhancement
5022 Description: Stop sending data to evicted clients as soon as possible.
5023 Details : Check if the client we are about to send or are sending data to
5024 was evicted already. (Check is done every second of waiting,
5025 for which l_wait_event interface was extended to allow checking
5026 of exit condition at specified intervals).
5029 Frequency : rare, normally only when NFS exporting is done from client
5031 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5033 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5036 Frequency : always, if srand() is called before liblustre initialization
5038 Description: Liblustre uses system PRNG disturbing its usage by user application
5039 Details : Introduce internal to lustre fast and high-quality PRNG for
5040 lustre usage and make liblustre and some other places in generic
5041 lustre code to use it.
5043 Severity : enhancement
5044 Bugzilla : 9477, 9557, 9870
5045 Description: Verify that the MDS configuration logs are updated when xml is
5046 Details : Check if the .xml configuration logs are newer than the config
5047 logs stored on the MDS and report an error if this is the case.
5048 Request --write-conf, or allow starting with --old_conf.
5050 Severity : enhancement
5052 Description: Handle symlinks in the path when checking if Lustre is mounted.
5053 Details : Resolve intermediate symlinks when checking if a client has
5054 mounted a filesystem to avoid duplicate client mounts.
5059 Description: lconf can hit an error exception but still return success.
5060 Details : The lconf command catches the Command error exception at the top
5061 level script context and will exit with the associated exit
5062 status, but doesn't ensure that this exit status is non-zero.
5067 Description: failure of ptlrpc thread startup can cause oops
5068 Details : Starting a ptlrpc service thread can fail if there are a large
5069 number of threads or the server memory is very fragmented.
5070 Handle this without oopsing.
5073 Frequency : always, only if liblustre and non-default acceptor port was used
5075 Description: liblustre cannot connect to servers with non-default acceptor port
5076 Details : tcpnal_set_default_params() was not called and was therefore
5077 ignoring the environment varaible TCPNAL_PORT, as well as other
5078 TCPNAL_ environment variables
5083 Description: two objects could be created on the same OST for a single file
5084 Details : If an OST is down, in some cases it was possible to create two
5085 objects on a single OST for a single file. No problems other
5086 than potential performance impact and spurious error messages.
5090 Bugzilla : 5681, 9562
5091 Description: Client may oops in ll_unhash_aliases
5092 Details : Client dcache may become inconsistent in race condition.
5093 In some cases "getcwd" can fail if the current directory is
5099 Description: Inode refcounting problems in NFS export code
5100 Details : link_raw functions used to call d_instantiate without obtaining
5101 extra inode reference first.
5105 Bugzilla : 9942, 9903
5106 Description: Referencing freed requests leading to crash, memleaks with NFS.
5107 Details : We used to require that call to ll_revalidate_it was always
5108 followed by ll_lookup_it. Also with revalidate_special() it is
5109 possible to call ll_revalidate_it() twice for the same dentry
5110 even if first occurence returned success. This fix changes semantic
5111 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5112 reference on a request referred from the intent.
5113 ll_intent_release() then releases such a request.
5116 Frequency : rare, normally benchmark loads only
5118 Description: unlinked inodes were kept in memory on the client
5119 Details : If a client is repeatedly creating and unlinking files it
5120 can accumulate a lot of stale inodes in the inode slab cache.
5121 If there is no other client load running this can cause the
5122 client node to run out of memory. Instead flush old inodes
5123 from client cache that have the same inode number as a new inode.
5126 Frequency : SLES9 2.6.5 kernel and long filenames only
5127 Bugzilla : 9969, 10379
5128 Description: utime reports stale NFS file handle
5129 Details : SLES9 uses out-of-dentry names in some cases, which confused
5130 the lustre dentry revalidation. Change it to always use the
5134 Frequency : rare, unless heavy write-truncate concurrency is continuous
5135 Bugzilla : 4180, 6984, 7171, 9963, 9331
5136 Description: OST becomes very slow and/or deadlocked during object unlink
5137 Details : filter_destroy() was holding onto the parent directory lock
5138 while truncating+unlinking objects. For very large objects this
5139 may block other threads for a long time and slow overall OST
5140 responsiveness. It may also be possible to get a lock ordering
5141 deadlock in this case, or run out of journal credits because of
5142 the combined truncate+unlink. Solution is to do object truncate
5143 first in one transaction without parent lock, and then do the
5144 final unlink in a new transaction with the parent lock. This
5145 reduces the lock hold time dramatically.
5148 Frequency : rare, 2.4 kernels only
5150 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5151 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5152 before unmounting their filesystem in order to avoid deadlock.
5153 The kernel_locked() function in 2.4 kernels only checks whether
5154 the kernel lock is held, not whether it is this process that is
5155 holding it as 2.6 kernels do.
5160 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5161 Details : The client ptlrpc code may be trying to reconnect to a down
5162 server before a previous connection attempt has timed out.
5163 Increase the reconnect interval to be longer than the connection
5164 timeout interval to avoid sending duplicate connections to
5168 Frequency : echo_client brw_test command
5170 Description: fix echo_client to work with OST preallocated code
5171 Details : OST preallocation code (5137) didn't take echo_client IO path
5172 into account: echo_client calls filter methods outside of any
5173 OST thread and, hence, there is no per-thread preallocated
5174 pages and buffers to use. Solution: hijack pga pages for IO. As
5175 a byproduct, this avoids unnecessary data copying.
5179 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5180 Description: Client can oops in mdc_commit_close() after open replay
5181 Details : It was possible for the MDS to return an open request with no
5182 transaction number in mds_finish_transno() if the client was
5183 evicted, but without actually returning an error. Clients
5184 would later try to replay that open and may trip an assertion
5185 Simplify the client close codepath, and always return an error
5186 from the MDS in case the open is not successful.
5189 Frequency : rare, 2.6 OSTs only
5191 Description: OST may deadlock under high load on fragmented files
5192 Details : If there was a heavy load and highly-fragmented OST filesystems
5193 it was possible to have all the OST threads deadlock waiting on
5194 allocation of biovecs, because the biovecs were not released
5195 until the entire RPC IO was completed. Instead, release biovecs
5196 as soon as they are complete to ensure forward IO progress.
5198 Severity : enhancement
5200 Description: Support for specifying external journal device at mount
5201 Details : If an OST or MDS device is formatted with an external journal
5202 device, this device major/minor is stored in the ext3 superblock
5203 and may not be valid for failover. Allow detecting and
5204 specifying the external journal at mount time.
5209 Description: Mounting an MDS with pending unlinked files may cause oops
5210 Details : target_finish_recovery() calls mds_postrecov() which returned
5211 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5212 considers this an error and immediately begins cleaning up the
5213 lov, just after starting the mds_lov process
5215 Severity : enhancement
5217 Description: Implement 'lfs df' to report actual free space on per-OST basis
5218 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5219 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5220 usage of objects; '-h' to report in human readable format.
5222 ------------------------------------------------------------------------------
5224 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5231 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5232 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5233 at startup during orphan unlinking. Instead, skip buddy update
5234 before setup, as it will be regenerated after recovery is complete.
5239 Description: If an OST is inactive, its locks might reference stale inodes.
5240 Details : lov_change_cbdata() must iterate over all namespaces, even if
5241 they are inactive to clear inode references from the lock.
5243 Severity : enhancement
5244 Frequency : occasional, if non-standard max_dirty_mb used
5246 Description: Client will block write RPCs if not enough grant
5247 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5248 then the client will block writes while waiting for another RPC
5249 to complete instead of consuming its dirty limit. With change
5250 we get improved performance when max_dirty_mb is small.
5252 Severity : enhancement
5253 Bugzilla : 3389, 6253
5254 Description: Add support for supplementary groups on the MDS.
5255 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5256 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5257 lookups for user supplementary groups into a cache.
5261 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5262 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5263 for liblustre O_EXCL is enforced by MDS.
5267 Description: readdir never returns NULL in liblustre.
5268 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5269 dentry in struct dirent.
5273 Frequency : liblustre only, depends on application IO pattern
5274 Description: liblustre clients evicted if not contacting servers
5275 Details : Don't put liblustre clients into the ping_evictor list, so
5276 they will not be evicted by the pinger ever.
5278 Severity : enhancement
5280 Description: Add ability to evict clients by NID from MDS.
5281 Details : By echoing "nid:$NID" string into
5282 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5283 $NID would be instantly evicted from this MDS and from all active
5284 OSTs connected to it.
5288 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5289 Details : lookup_it_finish() used to query file size from OSTs that was not
5294 Description: service threads change working directory to that of init
5295 Details : Starting lustre service threads may pin the working directory
5296 of the parent thread, making that filesystem busy. Threads
5297 now change to the working directory of init to avoid this.
5301 Frequency : during shutdown only
5302 Description: shutdown with a failed MDS or OST can cause unmount to hang
5303 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5308 Frequency : 2.6 only
5309 Description: chmod/chown may include an extra supplementary group
5310 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5311 supplementary group and if none is specified this is used.
5314 Bugzilla : 5479 (6816)
5316 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5317 Details : If lookup is in progress on a file that is unlinked we might try
5318 to revalidate the inode and fail in revalidate after lookup is
5319 complete and ll_file_open() enqueues the open again but
5320 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5323 Frequency : always, if lconf --abort_recovery used
5325 Description: lconf --abort_recovery fails with 'Operation not supported'
5326 Details : lconf was attempting to abort recovery on the MDT device and not
5329 ------------------------------------------------------------------------------
5331 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5336 Frequency : rare (only unsupported configurations with a node running as an
5338 Bugzilla : 6514, 5137
5339 Description: Mounting a Lustre file system on a node running as an OST could
5341 Details : OSTs now preallocates memory needed to write out data at
5342 startup, instead of when needed, to avoid having to
5343 allocate memory in possibly low memory situations.
5344 Specifically, if the file system is mounted on on OST,
5345 memory pressure could force it to try to write out data,
5346 which it needed to allocate memory to do. Due to the low
5347 memory, it would be unable to do so and the node would
5348 become unresponsive.
5350 Severity : enhancement
5352 Description: Addition of lconf --service command line option
5353 Details : lconf now accepts a '--service <arg>' option, which is
5354 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5356 Severity : enhancement
5358 Description: Failover mode is now the default for OSTs.
5359 Details : By default, OSTs will now run in failover mode. To return to
5360 the old behaviour, add '--failout' to the lmc line for OSTs.
5362 Severity : enhancement
5364 Description: Health checks are now provided for MDS and OSTs
5365 Details : Additional detailed health check information on MSD and OSTs
5366 is now provided through the procfs health_check value.
5369 Frequency : occasional, depends on IO load
5371 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5372 after numerous create/delete cycles
5373 Details : The ext3 inode allocation policy would not allocate new inodes
5374 very well on the OSTs because there are no new directories
5375 being created. Instead we look for groups with free space if
5376 the parent directories are nearly full.
5381 Description: Network or server problems during mount may cause partially
5382 mounted clients instead of returning an error.
5383 Details : The config llog parsing code may overwrite the error return
5384 code during mount error handling, returning success instead
5390 Description: MDS can fail to allocate large reply buffers
5391 Details : After long uptimes the MDS can fail to allocate large reply
5392 buffers (e.g. zconf client mount config records) due to memory
5393 fragmentation or consumption by the buffer cache. Preallocate
5394 some large reply buffers so that these replies can be sent even
5395 under memory pressure.
5399 Frequency : rare (liblustre)
5400 Description: fsx running with liblustre complained that using truncate() to
5401 extend the file doesn't work. This patch corrects that issue.
5402 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5403 ATTR_SIZE and lsm use in llu_setattr_raw.
5407 Frequency : rare, only 2.6 kernels
5408 Description: Unusual file access patterns on the MDS may result in inode
5409 data being lost in very rare circumstances.
5410 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5411 code in the 2.6 kernel caused the inode and/or EA data not to
5412 be read from disk, causing single-file corruption.
5416 Frequency : rare, only 2.6 filesystems using extents
5417 Description: Heavy concurrent write and delete load may cause data corruption.
5418 Details : It was possible under high-load situations to have an extent
5419 metadata block in the block device cache from a just-unlinked
5420 file overwrite a newly-allocated data block. We now unmap any
5421 metadata buffers that alias just-allocated data blocks.
5425 Frequency : filesystems with default stripe_count larger than 77
5426 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5427 Details : lconf specifies an inode size of 4096 bytes when the default
5428 stripe_count is larger than 77. This conflicts with the default
5429 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5430 this case to avoid pinning too much memory for large EAs.
5432 ------------------------------------------------------------------------------
5434 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5439 Frequency : rare (extremely heavy IO load with hundreds of clients)
5441 Description: Client is evicted, gets IO error writing to file
5442 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5443 caused clients to be evicted for AST timeouts. The fixes in
5444 bug 5192 mean we no longer need to have such short AST timeouts
5445 so ldlm_timeout has been increased.
5448 Frequency : occasional during --force or --failover shutdown under load
5449 Bugzilla : 5949, 4834
5450 Description: Server oops/LBUG if stopped with --force or --failover under load
5451 Details : a collection of import/export refcount and cleanup ordering
5452 issues fixed for safer force cleanup
5455 Frequency : only filesystems larger than 120 OSTs
5456 Bugzilla : 5990, 6223
5457 Description: lfs getstripe would oops on a very large filesystem
5458 Details : lov_getconfig used kfree on vmalloc'd memory
5461 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5462 Bugzilla : 6242, 6243
5463 Description: reading from files that had been truncated to a non-zero size
5464 but never opened returned no data
5465 Details : ll_file_read() reads zeros from no-object files to EOF
5470 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5471 Details : There's little harm in aborting MDS/OSS recovery and letting it
5472 try again, so I removed the LASSERT and return an error instead.
5474 Severity : enhancement
5476 Description: New debugging infrastructure for tracking down data corruption
5477 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5478 (b) cover more of the client-side code path, and (c) try to narrow
5479 down where problems occurred
5483 Bugzilla : 3819, 4364, 4397, 6313
5484 Description: Racing close and eviction MDS could cause assertion in mds_close
5485 Details : It was possible to get multiple mfd references during close and
5486 client eviction, leading to one thread referencing a freed mfd.
5488 Severity: : enhancement
5489 Bugzilla : 3262, 6359
5490 Description: Attempts to reconnect to servers are now more aggressive.
5491 Details : This builds on the enhanced upcall-less recovery that was added
5492 in 1.4.2. When trying to reconnect to servers, clients will
5493 now try each server in the failover group every 10 seconds. By
5494 default, clients would previously try one server every 25 seconds.
5499 Description: After recovery, certain operations trigger a failed
5500 assertion on a client.
5501 Details : Failing over an mds, using lconf -d --failover, while a
5502 client was doing a readdir() call would cause the client to
5503 LBUG after recovery completed and the readdir() was resent.
5505 Severity : enhancement
5507 Description: Default groups are now added by lconf
5508 Details : You can now run lconf --group <servicename> without having to
5509 manually add groups with lmc.
5512 Frequency : occasional
5514 Description: Nodes with an elan id of 0 trigger a failed assertion
5517 Frequency : always when accessing e.g. tty/console device nodes
5519 Description: tty and some other devices nodes cannot be used on lustre
5520 Details : file's private_data field is used by device data and lustre
5521 values in there got lost. New field was added to struct file to
5522 store fs-specific private data.
5525 Frequency : when exporting Lustre via NFS
5527 Description: NFSD failed occasionally when looking up a path component
5528 Details : NFSD is looking up ".." which was broken in ext3 directories
5529 that had grown large enough to become hashed.
5532 Frequency : Clusters with multiple interfaces not on the same subnet
5534 Description: Nodes will repeatedly try to reconnect to an interface which it
5535 cannot reach and report an error to the log.
5536 Details : Extra peer list entries will be created by lconf with some peers
5537 unreachable. lconf now validates the peer before adding it.
5540 Frequency : Only if a default stripe is set on the filesystem root.
5542 Description: Setting a default stripe on the filesystem root prevented the
5543 filesystem from being remounted.
5544 Details : The client was sending extra request flags in the root getattr
5545 request and did not allocate a reply buffer for the dir EA.
5548 Frequency : occasional, higher if lots of files are accessed by one client
5549 Bugzilla : 6159, 6097
5550 Description: Client trips assertion regarding lsm mismatch/magic
5551 Details : While revalidating inodes the VFS looks up inodes with ifind()
5552 and in rare cases can find an inode that is being freed.
5553 The ll_test_inode() code will free the lsm during ifind()
5554 when it finds an existing inode and then the VFS later attaches
5555 this free lsm to a new inode.
5559 Bugzilla : 6422, 7030
5560 Description: MDS deadlock between mkdir and client eviction
5561 Details : Creating a new file via mkdir or mknod (starting a transaction
5562 and getting the ns lock) can deadlock with client eviction
5563 (gets ns lock and trying to finish a synchronous transaction).
5566 Frequency : occasional
5567 Description: While starting a server, the fsfilt_ext3 module could not be
5569 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5570 kernels. Previously, lconf would still use the ext3 name
5571 when trying to load modules. Now, it will correctly use
5572 ext3 on 2.4 and ldiskfs on 2.6.
5574 Severity : enhancement
5575 Description: The default stripe count has been changed to 1
5576 Details : The interpretation of the default stripe count (0, to lfs
5577 or lmc) has been changed to mean striping across a single
5578 OST, rather than all available. For general usage we have
5579 found a stripe count of 1 or 2 works best.
5581 Severity : enhancement
5582 Description: Add support for compiling against Cray portals.
5583 Details : Conditional compiling for some areas that are different
5587 Frequency : occasional
5588 Bugzilla : 6409, 6834
5589 Description: Creating files with an explicit stripe count may lead to
5590 a failed assertion on the MDS
5591 Details : If some OSTs are full or unavailable, creating files may
5592 trigger a failed assertion on the MDS. Now, Lustre will
5593 try to use other servers or return an error to the
5597 Frequency : occasional
5599 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5600 caused lock timeout during readahead (since 1.4.2).
5601 Details : Processes doing readahead might match a lock that hasn't been
5602 granted yet if there are overlapping and conflicting lock
5603 requests. The readahead process waits on ungranted lock
5604 (original lock is CBPENDING), while OST waits for that process
5605 to cancel CBPENDING read lock and eventually evicts client.
5607 Severity : enhancement
5609 Description: Initial enabling of flock support for clients
5610 Details : Implements fcntl advisory locking and file status functions.
5611 This feature is provided as an optional mount flag (default
5612 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5613 locking are implemented yet, and those that are are not guaranteed
5614 to be completely correct in production environments.
5615 mount -t lustre -o [flock|noflock] ...
5618 Frequency : occasional
5620 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5621 assertion in the ext3 JBD (journaling) layer.
5622 Details : The b_committed_data struct is protected by the big kernel lock
5623 in 2.4 kernels, serializing journal_commit_transaction() and
5624 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5625 access to this struct. In 2.6 kernels there is finer grained
5626 locking to improve SMP performance of the JBD layer.
5630 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5635 Description: merge and cleanup kernel patches.
5638 -----------------------------------------------------------------------------
5640 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5642 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5643 versions of Lustre. Please update all servers and clients to
5644 version 1.4.2 or later at the same time. You must also run
5645 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5646 to update the configuration logs.
5648 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5649 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5650 - swab llog records of type '0' so we get proper header size/idx (5861)
5651 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5652 - fix rename of one directory over another leaking an inode (5953)
5653 - avoid SetPageDirty on 2.6 (5981)
5654 - don't re-add just-being-destroyed locks to the waiting list (5653)
5655 - when creating new directories, inherit the parent's custom
5656 striping settings if present parent (3048)
5657 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5658 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5659 - don't hold client locks on temporary worklist from l_lru (5666)
5660 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5661 - automatically evict dead clients (5921)
5662 - Update file size properly in create+truncate+fstat case (6196)
5663 - Do not unhash mountpoint dentries, do not allow removal of
5665 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5666 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5667 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5668 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5669 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5670 - Correctly handle reads of files with no objects (6243)
5671 - lctl recover will also mark a device active if deactivate used (5933)
5673 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5674 - allow --write-conf on an MDS with different nettype than client (5619)
5675 - don't write config llogs to MDS for mounts not from that MDS (5617)
5676 - lconf should create multiple TCP connections from a client (5201)
5677 - init scripts are now turned off by default; run chkconfig --on
5678 lustre and chkconfig --on lustrefs to use them
5679 - upcalls are no longer needed for clients to recover to failover
5681 - add --abort-recovery option to lconf to abort recovery on device
5683 - add support for an arbitrary number of OSTs (3026)
5684 - Quota support protocol changes.
5685 - forward compatibility changes to wire structs (6007)
5686 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5687 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5688 - improved "lustre status" script
5689 - initialize blocksize for non-regular files (6062)
5690 - added --disable-server and --disable-client configure options (5782)
5691 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5692 - Vanilla 2.4.29 support
5693 - increase maximum number of obd devices to 520 (6242)
5694 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5695 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5696 SLES 9 2.6.5 kernels
5697 - limit stripes per file to 160 (the maximum EA size) (6093)
5699 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5702 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5703 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5704 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5705 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5706 - small ext3 extents cleanups and fixes (5733)
5707 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5708 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5709 - avoid needless client->OST connect, fix handle mismatch (5317)
5710 - fix DLM error path that led to out-of-sync client, long delays (5779)
5711 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5712 - fix several locking issues related to i_size (5492,5624,5654,5672)
5713 - don't move pending lock onto export if it is already evicted (5683)
5714 - fix kernel oops when creating .foo in unlinked directory (5548)
5715 - fix deadlock in obdfilter statistics vs. object create (5811)
5716 - use time_{before,after} to avoid timer jiffies wrap (5882)
5717 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5718 - Do not leak request if server was not able to process it (5154)
5719 - If mds_open unable to find parent dir, make that negative lookup(5154)
5720 - don't create new directories with extent-mapping (5909, 5936)
5722 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5723 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5724 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5725 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5726 - support for RHEL4 kernel on i686 (5773)
5727 - provide error messages when incompatible logs are encountered (5898)
5729 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5730 * version 1.4.0.10 (1.4.1 release candidate 1)
5732 - don't keep a lock reference when lock is not granted (4238)
5733 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5734 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5735 - drop import inflight refcount on signal_completed_replay error (5255)
5736 - unlock page after async write error during send (3677)
5737 - handle missing objects in filter_preprw_read properly (5265)
5738 - no transno return for symlink open, don't save no-trasno open (3440)
5739 - don't try to complete elan receive that already failed (4012)
5740 - free RPC server reply state on error (5406)
5741 - clean up thread from ptlrpc_start_thread() on error (5160)
5742 - readahead could read extra page into cache that wasn't ejected (5388)
5743 - prevent races in class_attach/setup/cleanup/detach (5260)
5744 - don't dereference de->d_inode after l_dput of de (5458)
5745 - use "int" for stripe value returned from lock_to_stripe (5544)
5746 - mballoc allocation and error-checking fixes in 2.6 (5504)
5747 - block device patches to fix I/O request sizes in 2.6 (5482)
5748 - look up hostnames for IB nals (5602)
5749 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5750 - don't start multiple acceptors for the same port (5277)
5751 - fix incorrect LASSERT in mds_getattr_name (5635)
5752 - export a proc file for general "ping" checking (5628)
5753 - fix "lfs check" to not block when the MDS is down (5628)
5755 - service request history (4965)
5756 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5757 - create an "evict_client" /proc entry on OSTs, like the MDS has
5758 - fix mount usage message, return errors per mount(8) (5168)
5759 - change grep [] to grep "[]" in tests so they work in more UMLs
5760 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5761 - remove ancient LOV_MAGIC_V0 EA support (5047)
5762 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5763 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5764 - allow readahead allocations to fail when low on memory (5383)
5765 - mmap locking landed again, after considerable improvement (2828)
5766 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5768 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5771 - send OST transaction number in read/write reply to free req (4966)
5772 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5773 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5774 - the watchdog thread now runs as interruptible (5246)
5775 - flock/lockf fixes (but it's still disabled, pending 5135)
5776 - don't use EXT3 constants in llite code (5094)
5777 - memory shortage at startup could cause assertion (5176)
5779 - reorganization of lov code
5780 - single portals codebase
5782 - add extents/mballoc support (5025)
5783 - direct I/O reads in the obdfilter (4048)
5784 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5786 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5789 - send OST transaction number in read/write reply to free req (4966)
5790 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5791 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5792 - the watchdog thread now runs as interruptible (5246)
5793 - handle missing objects in filter_preprw_read properly (5265)
5794 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5795 - drop import inflight refcount on signal_completed_replay error (5255)
5796 - unlock page after async write error during send (3677)
5797 - return original error code on reconstructed replies (3761)
5798 - no transno return for symlink open, don't save no-trasno open (3440)
5800 - add pid to ldlm debugging output (4922)
5801 - bump the watchdog timeouts -- we can't handle 30sec yet
5802 - extra debugging for orphan dentry/inode bug (5259)
5804 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5807 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5808 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5809 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5810 - fix NULL dereference in /proc/sys/portals/routes (4827)
5811 - allow failed mdc_close() operations to be interrupted (4561)
5812 - stop precreate on OST before MDS would time out on it (4778)
5813 - don't send partial-page writes before EOF from client (4410)
5814 - discard client grant for sub-page writes on large-page clients (4520)
5815 - don't free dentries not owned by NFS code, check generation (4806)
5816 - fix lsm leak if mds_create_objects() fails (4801)
5817 - limit debug_daemon file size, always print CERROR messages (4789)
5818 - use transno after validating reply (3892)
5819 - process timed out requests if import state changes (3754)
5820 - update mtime on OST during writes, return in glimpse (4829)
5821 - add mkfsoptions to LDAP (4679)
5822 - use ->max_readahead method instead of zapping global ra (5039)
5823 - don't interrupt __l_wait_event() during strace
5825 - add software watchdogs to catch hung threads quickly (4941)
5826 - make lustrefs init script start after nfs is mounted
5827 - fix CWARN/ERROR duplication (4930)
5828 - return async write errors to application if possible (2248)
5829 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5830 - print NAL number in %x format (4645)
5831 - update barely-supported suse-2.4.21-171 series (4842)
5832 - support for sles 9 %post scripts
5833 - support for building 2.6 kernel-source packages
5834 - support for sles km_* packages
5836 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5839 - ignore -ENOENT errors in osc_destroy (3639)
5840 - notify osc create thread that OSC is being cleaned up (4600)
5841 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5842 - reconstruct ost_handle() like mds_handle() (4657)
5843 - create a new thread to do import eviction to avoid deadlock (3969)
5844 - let lconf resolve symlinked-to devices (4629)
5845 - don't unlink "objects" from directory with default EA (4554)
5846 - hold socknal file ref over connect in case target is down (4394)
5847 - allow more than 32000 subdirectories in a single directory (3244)
5848 - fix blocks count for O_DIRECT writes (3751)
5849 - OST returns ENOSPC from object create when no space left (4539)
5850 - don't send truncate RPC if file size isn't changing (4410)
5851 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5852 - bind to privileged port in socknal and tcpnal (3689)
5854 - rate limit CERROR/CWARN console message to avoid overload (4519)
5855 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5856 - basic mmap support (3918)
5857 - kernel patch series update from b1_4 (4711)
5859 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5862 - avoid crash during MDS cleanup with OST shut down (2775)
5863 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5864 - don't use bad inodes on the MDS (3744)
5865 - dynamic object preallocation to improve recovery speed (4236)
5866 - don't hold spinlock over lock dumping or change debug flags (4401)
5867 - don't zero obd_dev when it is force cleaned (3651)
5868 - print grants to console if they go negative (4431)
5869 - "lctl deactivate" will stop automatic recovery attempts (3406)
5870 - look for existing locks in ldlm_handle_enqueue() (3764)
5871 - don't resolve lock handle twice in recovery avoiding race (4401)
5872 - revalidate should check working dir is a directory (4134)
5874 - don't always mark "slow" obdfilter messages as errors (4418)
5876 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5879 - don't close LustreDB during write_conf until it is done (3860)
5880 - fix typo in lconf for_each_profile (3821)
5881 - allow dumping logs from multiple threads at one time (3820)
5882 - don't allow multiple threads in OSC recovery (3812)
5883 - fix debug_size parameters (3864)
5884 - fix mds_postrecov to initialize import for llog ctxt (3121)
5885 - replace config semaphore with spinlock (3306)
5886 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5887 - don't allow enqueue to complete on a destroyed export (3822)
5888 - down write_lock before checking llog header bitmap (3825)
5889 - recover from lock replay timeout (3764)
5890 - up llog sem before sending rpc (3652)
5891 - reduce ns lock hold times when setting kms (3267)
5892 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5893 - fix NULL deref and obd_dev leak on setup error (3312)
5894 - replace some LBUG about llog ops with error handling (3841)
5895 - don't match INVALID dentries from d_lookup and spin (3784)
5896 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5897 - fix invalid assertion in ptlrpc_set_wait (3880)
5899 - add libwrap support for the TCP acceptor (3996)
5900 - add /proc/sys/portals/routes for non-root route listing (3994)
5901 - allow setting MDS UUID in .xml (2580)
5902 - print the stack of a process that LBUGs (4228)
5904 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5907 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5908 - make sure to unset replay flag from failed open requests (3440)
5909 - if default stripe count is 0, use OST count for inode size (3636)
5910 - update parent mtime/ctime on client for create/unlink (2611)
5911 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5912 - free recovery state on server during a forced cleanup (3571)
5913 - unregister_reply for resent reqs (3063)
5914 - loop back devices mounting and status check on 2.6 (3563)
5915 - fix resource-creation race that can provoke i_size == 0 (3513)
5916 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5917 - more debugging for page-accounting assertion (3746)
5918 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5919 - avoid deadlock after precreation failure (3758)
5920 - fix race and lock order deadlock in orphan handling (3450, 3750)
5921 - add validity checks when grabbing inodes from l_ast_data (3599)
5923 - add /proc/.../recovery_status to obdfilter (3428)
5924 - lightweight CDEBUG infrastructure, debug daemon (3668)
5925 - change default OSC RPC parameters to be better on small clusters
5926 - turn off OST read cache for files smaller than 32MB
5927 - install man pages and include them in rpms (3100)
5928 - add new init script for (un)mounting lustre filesystems (2593)
5929 - run chkconfig in %post for init scripts (3701)
5930 - drop scimac NAL (unmaintained)
5932 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5935 - clean kiobufs before and after use (3485)
5936 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5937 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5938 - change init script to fail more clearly if not run as root (1528)
5939 - allow clients to reconnect during replay (1742)
5940 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5941 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5942 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5943 - fix cond_resched (3554)
5944 - only evict unfinished clients after recovery (3515)
5945 - allow bulk resend, prevent data loss (3570)
5946 - dynamic ptlrpc request buffer allocation (2102)
5947 - don't allow unlinking open directory if it isn't empty (2904)
5948 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5949 - remove extraneous obd dereference causing LASSERT failure (3334)
5950 - don't use get_cycles() when creating temp. files on the mds (3156)
5951 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5952 - handle EEXIST for set-stripe, set proper directory name (3336)
5954 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5955 - fix ksocknal_fmb_callback() error messages (2918)
5957 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5960 - don't copy lvb into (possibly NULL) reply on error (2983)
5961 - don't deref dentry after dput, don't free lvb on error (2922)
5962 - use the kms to determine writeback rpc length (2947)
5963 - increment oti_logcookies when osc is inactive (2948)
5964 - update client's i_blocks count via lvb messages (2543)
5965 - handle intent open/close of special files properly (1557)
5966 - mount MDS with errors=remount-ro, like obdfilter (2009)
5967 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5968 - don't use cancelling-locks' kms values (2947)
5969 - use highest lock extent for kms, not last one (2925)
5970 - don't dereference ERR_PTR() dentry in error handling path (3107)
5971 - fix thread race in portals_debug_dumplog() (3122)
5972 - create lprocfs device entries at setup instead of at attach (1519)
5973 - common AST error handler, don't evict client on completion race (3145)
5974 - zero nameidata in detach_mnt in 2.6 (3118)
5975 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5976 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5977 - initialize RPC timeout timer earlier for 2.6 (3219)
5978 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5979 - print nal/nid for unknown nid (3258)
5980 - additional checks for oscc recovery before doing precreate (3284)
5981 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5982 - don't crash in mdc_close for bad permissions on open (3285)
5983 - zero i_rdev for non-device files (3147)
5984 - clear page->private before handing to FS, better assertion (3119)
5985 - tune the read pipeline (3236)
5986 - fix incorrect decref of invalidated dentry (2350)
5987 - provide read-ahead stats and refine rpc in flight stats (3328)
5988 - don't hold journal transaction open across create RPC (3313)
5989 - update atime on MDS at close time (3265)
5990 - close LDAP connection when recovering to avoid server load (3315)
5991 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5992 - don't leak open file on MDS after open resend (3325)
5993 - serialize filter_precreate and filter_destroy_precreated (3329)
5994 - loop device shouldn't call sync_dev() for nul device (3092)
5995 - clear page cache after eviction (2766)
5996 - resynchronize MDS->OST in background (2824)
5997 - refuse to mount the same filesystem twice on same mountpoint (3394)
5998 - allow llmount to create routes for mounting behind routers (3320)
5999 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6000 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6001 - fix rare problem with rename on htree directories (3417)
6003 - allow default OST striping configuration per directory (1414)
6004 - fix compilation for qswnal for 2.6 kernels (3125)
6005 - increase maximum number of MDS request buffers for large systems
6006 - change liblustreapi to be useful for external progs like lfsck (3098)
6007 - increase local configuration timeout for slow disks (3353)
6008 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6010 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6013 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6014 - don't overwrite extent policy data in reply if lock was blocked (2901)
6015 - drop filter export grants atomically with removal from device (2663)
6016 - del obd_self_export from work_list in class_disconnect_exports (2908)
6017 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6018 - swab reply message in mdc_close, other PPC fixes (2464)
6019 - fix destroying of named logs (2325)
6020 - overwrite old logs when running lconf --write_conf (2264)
6021 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6022 - fix race in target_handle_connect (2898)
6023 - mds_reint_create() should take same inode create lock (2926)
6024 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6025 - don't close files for self_export to avoid uninitialized obd (2936)
6026 - allow MDS with the same name as client node (2939)
6027 - hold dentry reference for closed log files for unlink (2325)
6028 - reserve space for all logs during transactions (2059)
6029 - don't evict page beyond end of stripe extent (2925)
6030 - don't oops on a deleted current working directory (2399)
6031 - handle hard links to targets without a parent properly (2517)
6032 - don't dereference NULL lock when racing during eviction (2867)
6033 - don't grow lock extents when lots of conflicting locks (2919)
6035 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6038 - account for cache space usage on clients to avoid data loss (974)
6039 - lfsck support in lustre kernel code (2349)
6040 - reduce journal credits needed for BRW writes (2370)
6041 - orphan handling to avoid losing space on client/server crashes
6042 - ptlrpcd can be blocked, stopping ALL progress (2477)
6043 - use lock value blocks to assist in proper KMS, faster stat (1021)
6044 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6045 - recovery for initial connections (2355)
6046 - fixes for mds_cleanup_orphans (1934)
6047 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6048 - block all file creations until orphan recovery completes (1901)
6049 - client remove rq_connection from request struct (2423)
6050 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6051 - recovery timer race (2670)
6052 - mdc_close recovey bug (2532)
6053 - ptlrpc cleanup bug (2710)
6054 - mds timeout on local locks (2588)
6055 - namespace lock held during RPCs (2431)
6056 - handle interrupted sync write properly (2503)
6057 - don't try to handle a message that hasn't been replied to (2699)
6058 - client assert failure during cleanup after abort recovery (2701)
6059 - leak mdc device after failed mount (2712)
6060 - ptlrpc_check_set allows timedout requests to complete (2714)
6061 - wait for inflight reqs when ptlrpcd finishes (2710)
6062 - make sure unregistered services are removed from the srv_list
6063 - reset bulk XID's when resending them (caught by 1138 test)
6064 - unregister_bulk after timeout
6065 - fix lconf error (2694)
6066 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6067 - readahead locks pages, leaves pending causing memory pressure (2673)
6068 - increase OST request buffers to 4096 on large machines (2729)
6069 - fix up permission of existing directories in simple_mkdir (2661)
6070 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6071 - don't assign transno to errored transactions (2742)
6072 - don't delete objects on OST if given a bogus objid from MDS (2751)
6073 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6074 - if rq_no_resend, then timeout request after recovery (2432)
6075 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6076 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6077 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6078 - MDS recovery shouldn't delete objects that it has given out (2730)
6079 - if enqueue arrives after completion, don't clobber LVB (2819)
6080 - don't unlock pages twice when trigger_group_io returns error (2814)
6081 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6082 - don't write pages to disk if there was an error (1450)
6083 - don't ping imports that have recovery disabled (2676)
6084 - take buffered bytes into account when balancing socknal conn (2817)
6085 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6086 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6087 - remove little-endian swabbing of llog records (1987)
6088 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6089 - retry reposting request buffers if they fail (1191)
6090 - grow extent at grant time to avoid granting a revoked lock (2809)
6091 - lock revoke doesn't evict page if covered by a second lock (2765)
6092 - disable VM readahead to avoid reading outside lock extents (2805)
6094 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6095 - updated kernel patches for hp-2.4.20 kernel (2681)
6097 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6100 - fix truncated write corruption (2366)
6101 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6103 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6104 - ensure len(uuid) < 37 in lmc (1171)
6105 - fix ia64 OOPS in llog_test (2255)
6106 - zero end of page at obdfilter for partial page writes (2648)
6107 - don't leave stale dentries around after renames (bug 2428)
6108 - fix timeouts when evicting a client with a single lock held (2642)
6109 - set deadline for the initial HELLO message to drain (2634)
6110 - print out dotted-quad IP addresses in the socknal (2302)
6112 - additional debugging for MDS client eviction problem (2443)
6113 - fix mkfsoptions support for osts (2603, 2604)
6115 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6118 - add series for the vanilla 2.6.0 kernel
6119 - add series for the vanilla 2.4.24 kernel
6120 - add series for a cray x86/64 UL kernel drop
6121 - fix xattr patches for the vanilla 2.4.19 series
6123 - generate true UUIDs in lmc (1171)
6124 - have portals stack dumping break in UML (2466)
6125 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6126 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6127 - wake callback waiting threads on client eviction (2460)
6128 - Add --ptldebug and --subsystem to lmc (1719)
6129 - update assertion to allow safe interrupt allocation
6130 - set rq_no_resend for cancel requests (2432)
6131 - recalculate ptlrpcd timeout after resend (2494)
6132 - call vfs_rmdir when removing pending directories (2368)
6133 - fix renaming a file to itself (2429)
6134 - lmc creates a default one-stripe lov (2454)
6135 - expand procfs space to handle large clusters (2326)
6136 - increase UML stack to avoid overflow
6137 - update lconf's list of debug and subsystem masks
6138 - fix lfs find --obd (2510)
6139 - /proc tunable for disabling filter read caching (2591)
6140 - stop rpm packages from altering slapd.conf (2301)
6141 - disable nagle in the socknal under 0conf (2578)
6142 - choose mds inode size based on stripe count (2572)
6143 - fix kernel-source rpm problems (2516)
6145 - add --disable-doc to avoid pdf generation (2421)
6146 - update documentation, tests, type-os, comments
6147 - avoid format warnings on ia64
6148 - remove the TOE NAL
6149 - tiny code cleanups by removing unused fields
6151 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6154 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6155 - remove the most blatant lies from BUILDING file (2371)
6156 - change default debug level to reasonable production setting
6157 - reduce client side cache size to reduce cache flush time
6158 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6159 - make TCP zerocopy and pinger support enabled by default (2476)
6160 - sync writes completed after process exits caused crashes (2319)
6161 - maintain correct mount count on the MDS (2356)
6162 - backout 1557, because 2316 wasn't really fixed
6163 - better file I/O statistics gathering in /proc
6164 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6165 - another kernel patch to fix zero-copy TCP function export
6166 - don't take duplicate lock when processing re-sent getattr (2420)
6167 - lctl uses obd_self_export instead of creating new conn (2353)
6168 - MDS/OST recovery case which requires object creation asserted (2425)
6169 - move lfs from /usr/sbin to /usr/bin in packages
6170 - fix race between mds_client_add and mds_client_free (2417)
6171 - use kmalloc instead of slabs in portals (2430)
6172 - don't create duplicate records when a failover MDS is present (2442)
6173 - remove unnecessary mount age check (2332)
6174 - don't remove directory inodes from locks prematurely (2451)
6175 - don't break if MDS service name is the same as hostname (2103)
6176 - fix races in client write RPC generation when cache full (2482)
6178 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6181 - remove now-unused request->rq_obd (278)
6182 - if an allocation fails, print out how much memory we've used (1933)
6183 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6184 - add the "configurable stack size" patch to most series files (1256)
6185 - ability to write large log records, for 100+ OST configs (2306)
6186 - fix NULL deref when filter_prep fails (2314)
6187 - fix operator precedence error in filter_sync
6188 - dynamic allocation of socknal TX descriptors (2315)
6189 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6190 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6191 - ensure that CWARN messages in Portals always get to the syslog
6192 - __init/__exit are not for prototype decls (ldlm_init/exit)
6193 - x86-64 compile warning fixes
6194 - fix gateway LMC keyword conflict (2318)
6195 - fix MDS lock inversions in getattr/reint paths (1844)
6196 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6197 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6198 - DLM race condition prevented some lock evictions (2328)
6199 - ENOMEM detection and retry on socknal sends (2230)
6200 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6201 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6202 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6203 - better error messages when a client is rejected during recovery (1505)
6204 - avoid cancelling locks which were never granted, after failure (2330)
6205 - fix i_sem/journal inversion in mds_client_add (2333)
6206 - fix truncate/getattr lock cycle deadlock (2334)
6207 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6208 - fix two rare exit paths which could leak an l_lock() ref (2321)
6209 - fencepost error in MDS/OST orphan recovery (2226)
6210 - make log record alignment 8 bytes (1988)
6211 - lstripe now fails when requested offset > ost_count (2237)
6212 - ensure that all kernel series have a complete list.h (1607)
6213 - fix crashes in special-file operations (2316)
6214 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6216 - allow configurable automake binary, for testing new versions
6217 - small update to the lfs documentation
6219 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6221 * fix negative export reference count in fsfilt_sync (2312)
6223 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6224 * release candidate 0.9.1
6226 - orphans are moved into the PENDING directory for possible recovery
6227 - replayed opens now open by fid for orphan/rename safety (1042)
6228 - last close of an orphan inode generates a transno (683)
6229 - chdir() and mount() now pin the directory entry (1020)
6230 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6231 - discard very old requests without processing them (1502)
6232 - remove some common, well-understood CERRORs (1505)
6233 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6234 - clear "grant" flags in OST replies until OST grant code lands (1644)
6235 - fix read performance by not clobbering i_blksize on client (1598)
6236 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6237 - make LRU size a /proc tunable, clears locks when reduced (707)
6238 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6239 - server locks take references on exports now (1558)
6240 - build fixes for 2.4.20-rh trees (1663)
6241 - return an error from lov_create if all OSCs are inactive (1751)
6242 - fix import levels when a reconnect happens without a timeout (1597)
6243 - exit early from mds_open if we get a lookup error (1749)
6244 - partial page read at EOF wouldn't wait for disk before sending (1642)
6245 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6246 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6247 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6248 - write/truncate lock inversion (1639)
6249 - Don't auto-load obdclass, portals modules during cleanup (1495)
6250 - fix timestamps from jumping to "now" (1763)
6251 - extra journal assertions (1648)
6252 - add an extra multiunlink test (1771)
6253 - fix read_record/write_record API (1776)
6254 - fix leak of offset_extent, possible incorrect i_size later (1772)
6255 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6256 - Debugging for the kqswnal_get_idle_tx problems (1820)
6257 - Allow recovery to be attempted multiple times (1536)
6258 - Write out MDS last_rcvd file after it is first created (1600)
6259 - Fix tx_descriptor leak in failed transmit situations (1827)
6260 - ext3 journaling fixes for assertion failure after IO error (1871)
6261 - class_export_put() on freed export after completion AST error (1896)
6262 - Fix revalidate looping in VFS (1322)
6263 - Don't access a freed export during MDS_REINT timeout (1521)
6264 - Add open-unlink recovery support on the MDS (1673,1764)
6265 - Return an error if no MDS data was read from last_rcvd (1946)
6266 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6267 - Don't setup a disk device that doesn't match exported UUID (317)
6268 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6269 - avoid committing NULL handle in force close
6270 - local.sh is now a one-stripe LOV configuration
6271 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6272 - fix ext3 htree duplicate directory entry corruption (1516)
6273 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6274 - update to new LOV EA format (2097)
6275 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6277 2003-06-15 Phil Schwan <phil@clusterfs.com>
6280 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6281 - per-import/export recovery handling (958, 931, 959)
6282 - multiple last-rcvd slots, for serving multiple FSes (949)
6283 - connections are again shared between multiple imp/exports (963, 964)
6284 - "umount -f" would hang if any requests needed to be sent (393, 978)
6285 - avoid pinning large req buffer by copying for queued messages (989)
6286 - add "uuid" to "lctl device" command to help upcalls (991)
6287 - "open" RPCs with transnos would confuse recovery counters (1037)
6288 - do proper endian conversion of all wire messages (288, 340, 891)
6289 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6290 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6291 - fix ptlrpc_connection leak in target_handle_connect (1174)
6292 - fix import refcounting bug in OST and MDS cleanup (1134)
6293 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6294 - fix very unlikely obd_types race condition (501)
6295 - remove osc_open hack for echo_client (1187)
6296 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6297 - a failure in read_inode2 leads to deadlock (1139)
6298 - cancel ack-locks as soon as transaction is committed (1072)
6299 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6300 - make sure to commitrw after any preprw to avoid deadlock (1162)
6301 - failing to execute a file in a lustre FS would lock inode (1203)
6302 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6303 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6304 - obd_uuid2tgt would walk off the end of the list (1255)
6305 - on IA64 the osc would give portals incorrect bulk size (1258)
6306 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6307 - fix lock inversion caused by new llite matching code (1282)
6308 - limit the number of dirty pages on a client to 10MB (1286)
6309 - timed out locks were not being corrected cancelled (1289)
6310 - fix O_DIRECT above 4GB on IA-32 (1292)
6311 * major user-visible changes
6312 - fail out/fail over policy now controlled by the upcall (993)
6314 - add OBD_PING to check server availability and failure (954)
6315 - lustre messages are now sent in sending host order (288, 340, 891)
6316 - add eadatalen to MDS getattr reply (340)
6317 - OST read replies may contain second buffer, with per-page status (593)
6319 2003-03-11 Phil Schwan <phil@clusterfs.com>
6322 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6323 - failed open()s could cause deadlock; fixed (867, 869)
6324 - stop cancelling OST locks when files are closed (481)
6325 - overlapping XID spaces caused network corruption (851, 853)
6326 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6327 - setattr_raw added, to do single-RPC, server-side setattrs
6328 - lmc/lconf syntax change for OST UUIDs
6329 - fix crashy race condition between ptlrpc_free_req and osc_close
6330 - don't use request in mdc_enqueue if we hit a timeout (889)
6331 - don't set the inode i_size for regular files from the MDS (896)
6332 - handle out of order completion AST (842)
6333 - don't LBUG if a lock request times out after receiving AST (913)
6334 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6335 - if a bad lock AST arrives, send an error instead of dropping entirely
6336 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6337 - fix leak in bulk IO when only partially completed (899, 900, 926)
6338 - fix O_DIRECT for ia64 (55)
6339 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6340 - C-z after timeout could hang a process forever; fixed (977)
6342 - client-side I/O cache (678, 924, 929, 941, 970)
6344 - READPAGE and SETATTRs which don't take server-side locks get
6347 2003-02-11 Phil Schwan <phil@clusterfs.com>
6350 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6351 - Fix fsfilt_extN_readpage() to read a full page of directory
6352 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6353 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6354 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6355 - Most intent operations were removed, in favour of a new RPC mode
6356 that does a single RPC to the server and bypasses most of the VFS
6357 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6358 - Aggressively cancel local locks on DLM servers
6359 - mds_reint_unlink sends EA to the client if it's the last nlink.
6360 client uses that EA to unlink OST objects.
6361 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6362 - recursive symlinks were fixed (439)
6363 - fixed NULL deref in DEBUG_REQ
6364 - filter_update_lastobjid no longer calls sync, which annoyed extN
6365 - fixed multi-client small-writes to a single file problem (445)
6366 - fixed mtime updates during file writes (607)
6367 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6368 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6369 - fixed Linux OST size reporting problem (444, 656)
6370 - OST now updates object mtime with writes or setattr (607, 619)
6371 - client verifies file size before zeroing page past EOF (445)
6372 - OST now writes last allocated objid to disk with allocation (108)
6373 - LOV on echo now works (409)
6375 - mds_reint_unlink sends a new buffer, with the EA included. this
6376 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6377 set if a regular file was being unlinked, and it was the last link
6378 - use PtlGet from the target for bulk writes (315)
6379 - OST now updates object mtime with writes or setattr (607, 619)
6380 - LDLM now has a grant-time callback to revalidate locked items, if
6382 - Many MDS operations were reorganized to combat race conditions
6384 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6385 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6387 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6390 - Fully reactivate OST imports after reconnection (512, others)
6391 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6392 - More graceful error handling for truncating on dead OST (515)
6393 - Don't error out unless we're actually accessing dead stripes (474)
6394 - Fix garbage sizes when stripes are missing (410)
6395 - LRU counters were broken, causing constant lock purge (433, 432)
6396 - garbage on read from stripes with failed OSTs (441)
6397 - mark OSCs as active before reconnecting during recovery (438)
6398 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6399 - lfind did not preserve OST order in output (443)
6400 - symlinks cause hung clients, incorrect data (439)
6401 - stop dereferencing request after dropping refcount (457)
6402 - don't LASSERT(spin_is_locked) on non-SMP (455)
6403 - fixes for many rename() bugs
6404 - fstat didn't correctly synchronize attributes (399)
6405 - server must handle lock cancellation during blocking AST prep (487)
6406 - bulk descriptors were free()d too soon (511)
6407 - fix paths in lconf, which would load incorrect modules (451, 507)
6408 - fix confusing lconf 'host not found' error message (386)
6409 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6410 - fix race condition in mdc_blocking_ast() for inode access (526)
6411 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6412 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6413 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6414 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6415 - fix lstripe in several ways (kernel side) (527)
6416 - fix request leak in ldlm_cli_enqueue (262)
6417 - incorrect OSC was marked inactive after OST failure
6418 - call mds_fs_cleanup before unmounting filesystem (524)
6419 - fix races between taking ns_lock and ldlm_lock_change_resource
6420 - fix races updating LOV export open file list
6421 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6422 - fix recovery NULL deref in ldlm_cli_cancel_unused
6423 - fix some DLM races by using new hash table for lock handles (419)
6424 - permit the client to specify desired inodes, at replay
6425 - duplicate requests when we queue them for replay reintegration
6426 - fix last_rcvd offset calculation
6427 - sync after each recovered transaction, so we always make progress
6428 - never, not always, ERESTART requests without transnos
6429 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6431 - skip replay if the MDS says that the client is already connected
6432 - don't check for a recovery-enabled export to match lctl's UUID
6433 - don't INC_USE_COUNT for phantom exports
6434 - don't crash when cleaning up phantom exports (567)
6435 - don't double-finish or set replay data for errored mdc_open requests
6436 - abort requests when they time out, so we don't get old replies
6437 - send/receive replies for AST messages again
6438 - if the client says that it doesn't have the lock, cancel it on the
6440 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6441 mark it as destroyed, it will all work out when decref is called
6442 - fix module use counts (22, 581)
6444 - ASTs now expect a reply (server cancels lock on error reply)
6446 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6449 - fix many simultaneous client startup (392)
6450 - fix dentry->d_it clobbering
6451 - credentials weren't being shipped for readdir/getattr operations
6452 - remove invalid assertions triggered during some concurrent MD
6454 - proper Lustre versions added (336, 389)
6455 - fix memory leak for create error case (398)
6456 - fix LOV locking bug that would get cli/srv out of sync
6457 - fix echo client over LOV (409)
6458 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6459 - fix double-O_EXCL intent crash (424)
6460 - avoid sending multiple lock CANCELs (352)
6462 - MDS can do multi-client recovery (modulo bugs in new code)
6464 - many updates, edits, cleanups
6466 2002-11-18 Phil Schwan <phil@clusterfs.com>
6469 - fix null d_it dereference (346)
6470 - fix full OST/dbench hang (333)
6471 - fix permission problem with file removal (286)
6472 - fix removal of OSCs from LOV when they fail
6473 - fix NULL deref during bulk timeout (214)
6474 - fix problems related to multiple filesystems on one MDS (241)
6475 - fixed serious subtle metadata locking bugs
6476 - free locks on clients when inodes are removed due to memory
6478 - fix inode pointer in lock data (285)
6479 - partial support for multiple MDS on a single host (241)
6480 - data locks weren't cancelled at clear_inode time (290, 311)
6481 - intent locks could lead to unbounded lock growth (205)
6482 - added a maximum lock count, an LRU list, and a flusher
6483 - fix multiple rename (365)
6484 - properly abstracted the echo client
6485 - OSC locked 1 byte too many; fixed
6486 - rewrote brw callback code:
6487 - fixed recovery bugs related to LOVs (306)
6488 - fixed too-many-pages-in-one-write crash (191)
6489 - fixed (again) crash in sync_io_timeout (214)
6490 - probably fixed callback-related race (385)
6492 - Add capability to MDS protocol
6493 - LDLM cancellations and callbacks on different portals
6495 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6498 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6499 - defer open object destruction to close time (601981, 138)
6500 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6501 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6502 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6503 - don't call obd_setattr in ll_file_release for destroyed objects
6504 * protocol change to lustre_msg: move |version| and add |flags|
6505 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6506 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6507 * added replay of create, unlink, link and rename operations during
6508 MDS failover; recovery should be much more robust now
6509 * remove failed OSCs from LOVs (only lov_create uses this so far)
6510 * the lustre-HOWTO was brought (more) up to date (582544)
6512 2002-10-23 Phil Schwan <phil@clusterfs.com>
6515 - in-use dentries weren't being reused properly (617851)
6516 - prevent multiple LDLM setup (599178)
6517 - fix LOV size calculations for truncate (617853)
6518 - fix client handling of MDS intent errors (POSIX)
6519 - fix permission bug in lovstripe.c test (624321)
6520 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6521 - truncate past end of file could corrupt data
6522 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6523 - a race in recovery could return ETIMEDOUT to apps (623947)
6524 - building outside the source directory was fixed
6525 * the lustre-HOWTO was brought (more) up to date (582544)
6526 * major progress was made on recovery functionality
6528 2002-10-10 Phil Schwan <phil@clusterfs.com>
6531 - recovery deadlock fix
6532 - rm -rf causes LBUG fix (617817)
6533 - file open by multiple tasks fix (618962)
6534 - directory permissions bugs (602707 and 620007)
6535 - journal_stop fixed with locking (611313)
6536 - O_APPEND failures resolved (618273, perhaps 614459)
6537 - lconf PATH fix (619770)
6538 - IA64 build fix (621450)
6539 - RPC buffer sizes scale with amount of memory
6541 2002-10-01 Phil Schwan <phil@clusterfs.com>
6544 - locks would be cancelled without throwing away data pages,
6545 resulting in inconsistent data (605627)
6546 - inode attributes were not always being refreshed (605627, 612449)
6547 - lconf now continues to cleanup after lctl reports an error
6548 - MDS now enforces user permissions (602707)
6549 - lprocfs cleanup fixed, but not yet enabled (614157)
6550 - fixed infinite server hang, should a client not respond to an AST
6551 - avoid going into recovery if user calls readlink() with a buffer
6552 that's too small (613941)
6553 - AST RPCs no longer require replies (614867) -- this may be changed
6554 - don't crash server if client sends an IOV that's too big (611336)
6555 - fixed lock conversion deadlock (611892)
6556 - fixed the following of symlinks (614622)
6557 * recovery: the server can remove locks from a client that dies, other
6558 clients can make progress
6559 * more extN patch fixes
6560 * compile-time configurable ptlrpc buffer allocations
6562 - collaborative read cache document
6563 - Lustre Lite Performance CDR document-in-progress
6565 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6568 - fix typo in patch-2.4.18
6570 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6573 - clear ptlrpc request each time in handle_incoming_request()
6574 - unlink of files now destroys the object on the OST
6576 2002-09-19 Peter Braam <braam@clusterfs.com>
6578 * add hard link support
6579 * change obdfile creation method
6580 * kernel patch changed
6582 2002-09-19 Peter Braam <braam@clusterfs.com>
6585 - stack overflow bug in extN fixed
6587 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6589 * documentation updates
6590 - add man pages for config tools
6591 - update tests/README to describe testing with new config tools
6592 - finish metadata API descriptions
6593 * bug fixes and cleanups
6594 - statfs workaround for 16TB limit
6595 - LOV stripe allocation improved, can stripe on subset of OSTs
6596 - LOV file size/IO offset was wrong for files > 4GB in size
6597 - object EA data was being dropped, caused files to be unreadable
6598 - memory overflow with non-LOV OST caused memory corruption
6599 - fixed regression tests to work with new config tools, obdfilter
6600 - fixed bug when directory size became larger than 1 block
6601 - fixed bug (for single client case) when PWD was deleted
6602 - invalidate local directory pages when doing intent-based ops
6603 - avoid LDLM oops when lock callback contained bad data
6605 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6607 * documentation updates
6608 * bug fixes and cleanups
6609 - configuration tools
6612 - 64-bit compile warnings
6613 - 64-bit internal statfs data
6615 * test_brw on persistent OST devices
6617 * lprocfs (disabled)
6619 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6621 * documentation updates
6622 * bug fixes and cleanups
6623 * configuration tools
6625 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6628 * many small fixes to 0_5_4
6629 * io/network handling
6630 * thinkos in MDS operations
6632 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6635 * crucial basic fixes to 0.5.3
6636 * IOR, Iozone work over Elan
6639 2002-08-07 Phil Schwan <phil@clusterfs.com>
6640 * version 0_5_3, our first alpha
6641 * we use the new Portals iovs
6642 * documentation updates
6643 * bug fixes and cleanups
6644 * small changes in the DLM wire protocol
6646 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6647 * version 0_5_1 with some initial stability,
6648 * locking on MD and file I/O.
6649 * documentation updates
6650 * several bug fixes since 0.5.0
6651 * small changes in wire protocol
6653 2002-07-18 Phil Schwan <phil@clusterfs.com>
6655 * delivered as Lustre Light Alpha
6656 * fixed a crash after handling invalid MDS requests
6657 * fixed directory pages for architectures with non-4k pages sizes
6659 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6660 * release version v0_4_4
6661 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6663 * Fixes a number of interruption problems with OST operations.
6664 * Update documentation for portals header changes
6665 * Move all wire protocol structs/defines to lustre_idl.h
6666 * Fixes symlink length bug.
6667 * Add tcpdump to repository.
6669 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6670 * release version v0_4_3
6671 * Fixes statfs for inodes on extN.
6672 * Fixes bug in runtests which would delete /etc/hosts.
6673 * Use 64-bit object IDs wherever possible (not into VFS though)
6674 Remove ost_get_info, which is unused by lustre, and out of date.
6676 2002-07-03 Peter Braam <braam@clusterfs.com>
6677 * release version v0_4_2 Fixes a lookup error (type not passed)
6678 * move forward to head of Portals
6679 * move forward to latest Lustre kernel
6681 2002-06-25 Peter Braam <braam@clusterfs.com>
6682 * release version v0_4_1. Hopefully stable on single node use.