3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-693.21.1.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.126-94.22 (SLES12 SP3)
17 3.13.0-101 (Ubuntu 14.04, ZFS only)
18 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
19 4.4.0-116 (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 3.10.0-693.21.1.el7 (RHEL7.4)
28 3.0.101-0.47.71 (SLES11 SP3)
29 3.0.101-107 (SLES11 SP4)
30 3.12.74-60.64.40 (SLES12 SP1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.126-94.22 (SLES12 SP3)
33 3.13.0-101 (Ubuntu 14.04)
34 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
35 4.4.0-116 (Ubuntu 16.04)
37 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
38 * Recommended ZFS / SPL version: 0.7.8
39 * Tested with ZFS / SPL version: 0.7.8
40 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
41 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
42 stack overflow. For more information, please refer to bugzilla 17630.
43 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
44 the re-exporting nodes to be mounted with "32bitapi" mount option
45 * Note for Ubuntu 14.04.5:
46 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
47 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
49 --------------------------------------------------------------------------------
52 2018-04-03 Intel Corporation
54 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
55 for currently supported client and server kernel versions.
56 * Server known to build on patched kernels:
57 2.6.32-431.29.2.el6 (RHEL6.5)
58 2.6.32-504.30.3.el6 (RHEL6.6)
59 2.6.32-573.26.1.el6 (RHEL6.7)
60 2.6.32-642.15.1.el6 (RHEL6.8)
61 2.6.32-696.18.7.el6 (RHEL6.9)
62 3.10.0-693.21.1.el7 (RHEL7.4)
63 3.0.101-0.47.71 (SLES11 SP3)
64 3.0.101-107 (SLES11 SP4)
65 3.12.74-60.64.40 (SLES12 SP1)
66 4.4.114-92.64 (SLES12 SP2)
67 4.4.114-94.11 (SLES12 SP3)
68 3.13.0-101 (Ubuntu 14.04, ZFS only)
69 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
70 4.4.0-116 (Ubuntu 16.04)
71 vanilla linux 4.6.7 (ZFS only)
72 * Client known to build on unpatched kernels:
73 2.6.32-431.29.2.el6 (RHEL6.5)
74 2.6.32-504.30.3.el6 (RHEL6.6)
75 2.6.32-573.26.1.el6 (RHEL6.7)
76 2.6.32-642.15.1.el6 (RHEL6.8)
77 2.6.32-696.18.7.el6 (RHEL6.9)
78 3.10.0-693.21.1.el7 (RHEL7.4)
79 3.0.101-0.47.71 (SLES11 SP3)
80 3.0.101-107 (SLES11 SP4)
81 3.12.74-60.64.40 (SLES12 SP1)
82 4.4.114-92.64 (SLES12 SP2)
83 4.4.114-94.11 (SLES12 SP3)
84 3.13.0-101 (Ubuntu 14.04)
85 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
86 4.4.0-116 (Ubuntu 16.04)
88 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
89 * Recommended ZFS / SPL version: 0.7.6
90 * Tested with ZFS / SPL version: 0.7.6
91 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
92 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
93 stack overflow. For more information, please refer to bugzilla 17630.
94 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
95 the re-exporting nodes to be mounted with "32bitapi" mount option
96 * Note for Ubuntu 14.04.5:
97 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
98 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
100 --------------------------------------------------------------------------------
102 2017-07-05 Intel Corporation
104 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
105 for currently supported client and server kernel versions.
106 * Server known to build on patched kernels:
107 2.6.32-431.29.2.el6 (RHEL6.5)
108 2.6.32-504.30.3.el6 (RHEL6.6)
109 2.6.32-573.26.1.el6 (RHEL6.7)
110 2.6.32-642.15.1.el6 (RHEL6.8)
111 2.6.32-696.3.1.el6 (RHEL6.9)
112 3.10.0-514.21.1.el7 (RHEL7.3)
113 3.0.101-0.47.71 (SLES11 SP3)
114 3.0.101-100 (SLES11 SP4)
115 3.12.74-60.64.40 (SLES12 SP1)
116 4.4.59-92.17 (SLES12 SP2)
117 vanilla linux 4.6.7 (ZFS only)
118 * Client known to build on unpatched kernels:
119 2.6.32-431.29.2.el6 (RHEL6.5)
120 2.6.32-504.30.3.el6 (RHEL6.6)
121 2.6.32-573.26.1.el6 (RHEL6.7)
122 2.6.32-642.15.1.el6 (RHEL6.8)
123 2.6.32-696.3.1.el6 (RHEL6.9)
124 3.10.0-514.21.1.el7 (RHEL7.3)
125 3.0.101-0.47.71 (SLES11 SP3)
126 3.0.101-100 (SLES11 SP4)
127 3.12.74-60.64.40 (SLES12 SP1)
128 4.4.59-92.17 (SLES12 SP2)
130 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
131 * Recommended ZFS / SPL version: 0.6.5.9
132 * Tested with ZFS / SPL version: 0.6.5.9
133 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
134 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
135 stack overflow. For more information, please refer to bugzilla 17630.
136 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
137 the re-exporting nodes to be mounted with "32bitapi" mount option
139 --------------------------------------------------------------------------------
141 12-07-2016 Intel Corporation
143 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
144 for currently supported client and server kernel versions.
145 * Server known to build on patched kernels:
146 2.6.32-431.29.2.el6 (RHEL6.5)
147 2.6.32-504.30.3.el6 (RHEL6.6)
148 2.6.32-573.26.1.el6 (RHEL6.7)
149 2.6.32-642.11.1.el6 (RHEL6.8)
150 3.10.0-514.el7 (RHEL7.3)
151 3.0.101-0.47.71 (SLES11 SP3)
152 3.0.101-80 (SLES11 SP4)
153 3.12.62-60.64.8 (SLES12 SP1)
154 vanilla linux 4.5.7 (ZFS only)
155 * Client known to build on unpatched kernels:
156 2.6.32-431.29.2.el6 (RHEL6.5)
157 2.6.32-504.30.3.el6 (RHEL6.6)
158 2.6.32-573.26.1.el6 (RHEL6.7)
159 2.6.32-642.11.1.el6 (RHEL6.8)
160 3.10.0-514.el7 (RHEL7.3)
161 3.0.101-0.47.71 (SLES11 SP3)
162 3.0.101-80 (SLES11 SP4)
163 3.12.62-60.64.8 (SLES12 SP1)
165 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
166 * Recommended ZFS / SPL version: 0.6.5.8
167 * Tested with ZFS / SPL version: 0.6.5.8
168 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
169 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
170 stack overflow. For more information, please refer to bugzilla 17630.
171 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
172 the re-exporting nodes to be mounted with "32bitapi" mount option
174 --------------------------------------------------------------------------------
176 02-29-2016 Intel Corporation
178 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
179 for currently supported client and server kernel versions.
180 * Server known to build on patched kernels:
181 2.6.32-431.29.2.el6 (RHEL6.5)
182 2.6.32-504.30.3.el6 (RHEL6.6)
183 2.6.32-573.12.1.el6 (RHEL6.7)
184 3.10.0-327.3.1.el7 (RHEL7.2)
185 3.0.101-0.47.71 (SLES11 SP3)
186 3.0.101-68 (SLES11 SP4)
187 vanilla linux 4.2.1 (ZFS only)
188 * Client known to build on unpatched kernels:
189 2.6.32-431.29.2.el6 (RHEL6.5)
190 2.6.32-504.30.3.el6 (RHEL6.6)
191 2.6.32-573.12.1.el6 (RHEL6.7)
192 3.10.0-327.3.1.el7 (RHEL7.2)
193 3.0.101-0.47.71 (SLES11 SP3)
194 3.0.101-68 (SLES11 SP4)
197 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
198 * Recommended ZFS / SPL version: 0.6.4.2
199 * Tested with ZFS / SPL version: 0.6.4.2
200 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
201 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
202 stack overflow. For more information, please refer to bugzilla 17630.
203 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
204 the re-exporting nodes to be mounted with "32bitapi" mount option
206 --------------------------------------------------------------------------------
208 03-10-2015 Intel Corporation
210 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
211 for currently supported client and server kernel versions.
212 * Server known to build on patched kernels:
213 2.6.32-431.29.2.el6 (RHEL6.5)
214 2.6.32-504.8.1.el6 (RHEL6.6)
215 3.0.101-0.46 (SLES11 SP3)
216 * Client known to build on unpatched kernels:
217 2.6.32-431.29.2.el6 (RHEL6.5)
218 2.6.32-504.8.1.el6 (RHEL6.6)
219 3.10.0-123.20.1.el7 (RHEL7)
220 3.0.101-0.46 (SLES11 SP3)
221 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
222 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
223 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
224 stack overflow. For more information, please refer to bugzilla 17630.
225 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
226 the re-exporting nodes to be mounted with "32bitapi" mount option
228 Severity : enhancement
230 Description: control OST-index in IDIF via ROCOMPAT flag.
231 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
232 last_rcvd file. For new formatted OST device, it will be auto set;
233 for the case of upgrading from old OST device, you can enable it
234 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
235 enabled, for new created OST-object, its IDIF-in-LMA will contain
236 the OST-index; for the existing OST-object, the OSD will convert
237 old format IDIF as new format IDIF with OST-index stored in the
238 LMA EA when accessing such OST-object or via OI scrub. Once such
239 flag is enabled, it cannot be reverted back, so the system cannot
240 be downgraded to the orignal incompatible version.
242 --------------------------------------------------------------------------------
244 07-30-2014 Intel Corporation
246 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
247 for currently supported client and server kernel versions.
248 * Server known to build on patched kernels:
249 2.6.32-431.20.3.el6 (RHEL6.5)
250 3.0.101-0.31 (SLES11 SP3)
251 * Client known to build on unpatched kernels:
252 2.6.32-431.20.3.el6 (RHEL6.5)
253 3.0.101-0.31 (SLES11 SP3)
256 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
257 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
258 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
259 stack overflow. For more information, please refer to bugzilla 17630.
260 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
261 the re-exporting nodes to be mounted with "32bitapi" mount option
263 --------------------------------------------------------------------------------
265 10-31-2013 Intel Corporation
267 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
268 * Server builds for kernels:
269 2.6.32-358.18.1.el6 (RHEL6.4)
270 * Client builds for unpatched kernels:
271 2.6.32-358.18.1.el6 (RHEL6.4)
272 3.0.92-0.8 (SLES11 SP3)
274 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
275 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
276 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
277 stack overflow. For more information, please refer to bugzilla 17630.
278 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
279 the re-exporting nodes to be mounted with "32bitapi" mount option
281 --------------------------------------------------------------------------------
283 05-30-2013 Intel Corporation
285 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
286 * Server builds for kernels:
287 2.6.32-358.6.2.el6 (RHEL6.4)
288 * Client builds for unpatched kernels:
289 2.6.32-358.6.2.el6 (RHEL6.4)
290 3.0.74-0.6.6 (SLES11 SP2)
292 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
293 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
294 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
295 stack overflow. For more information, please refer to bugzilla 17630.
296 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
297 the re-exporting nodes to be mounted with "32bitapi" mount option
299 --------------------------------------------------------------------------------
301 9-30-2012 Whamcloud, Inc.
303 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
304 * Server support for kernels:
305 2.6.32-279.5.1.el6 (RHEL6)
306 * Client support for unpatched kernels:
307 2.6.32-279.5.1.el6 (RHEL6)
309 2.6.18-238.19.1.el5 (RHEL5)
310 2.6.16.60-0.69.1 (SLES10)
311 * Recommended e2fsprogs version: 1.41.90.wc4
312 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
313 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
314 stack overflow. For more information, please refer to bugzilla 17630.
315 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
316 the re-exporting nodes to be mounted with "32bitapi" mount option
318 Severity : enhancement
320 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
322 Severity : enhancement
324 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
326 Severity : enhancement
328 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
330 Severity : enhancement
332 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
334 Severity : enhancement
336 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
338 --------------------------------------------------------------------------------
340 04-30-2012 Whamcloud, Inc.
342 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
343 * Server support for kernels:
344 2.6.32-220.4.2.el6 (RHEL6)
345 * Client support for unpatched kernels:
346 2.6.32-220.4.2.el6 (RHEL6)
348 2.6.18-238.19.1.el5 (RHEL5)
349 2.6.16.60-0.69.1 (SLES10)
350 * Recommended e2fsprogs version: 1.41.90.wc4
351 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
352 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
353 stack overflow. For more information, please refer to bugzilla 17630.
354 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
355 the re-exporting nodes to be mounted with "32bitapi" mount option
357 Severity : enhancement
359 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
361 Severity : enhancement
363 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
365 Severity : enhancement
367 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
369 Severity : enhancement
371 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
373 --------------------------------------------------------------------------------
375 09-30-2011 Whamcloud, Inc.
377 * Server support for kernels:
378 2.6.18-238.19.1.el5 (RHEL5)
379 2.6.32-131.6.1.el6 (RHEL6)
380 * Client support for unpatched kernels:
381 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
382 2.6.18-238.19.1.el5 (RHEL5)
383 2.6.32-131.6.1.el6 (RHEL6)
384 2.6.29.4-167.fc11 (FC11)
385 2.6.16.60-0.69.1 (SLES10)
387 * Recommended e2fsprogs version: 1.41.90.wc3
388 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
389 Lustre filesystem with 4K stack may cause a stack overflow. For more
390 information, please refer to bugzilla 17630.
391 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
392 mounted with "32bitapi" mount option
394 Severity : enhancement
395 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
397 Severity : enhancement
398 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
400 Severity : enhancement
401 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
404 Frequency : only with SLES10
406 Description: Use OFED "KMP" provided by Novell
407 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
408 Lustre is now built against this package. That means you need to
409 install the ofed-kmp package from Novell for the patchless client.
410 Note that the ofed-kmp that Novell ships may not exactly match the
411 kernel version but should still be compatible.
413 Severity : enhancement
415 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
416 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
418 --------------------------------------------------------------------------------
420 2010-07-15 Oracle, Inc.
422 * Server support for kernels:
423 2.6.18-164.11.1.el5 (RHEL 5)
424 2.6.18-164.11.1.0.1.el5 (OEL 5)
425 * Client support for unpatched kernels:
426 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
427 2.6.18-164.11.1.el5 (RHEL 5),
428 2.6.18-164.11.1.0.1.el5 (OEL 5)
429 2.6.16.60-0.42.8 (SLES 10),
431 2.6.29.4-167.fc11 (FC11)
432 * Recommended e2fsprogs version: 1.41.10-sun2
434 Severity : enhancement
436 Description: Distribute lfs_migrate script with lustre
437 Details : The lfs_migrate mentioned in the manual is now included in the
438 lustre tree in lustre/scripts and will distributed in lustre RPMs.
440 Severity : enhancement
442 Description: Failover nid list added to import proc file.
444 Severity : enhancement
446 Description: "som_preview" mount option required to use SOM
447 Details : The clients will need to use som_preview mount option
451 Description: Lustre does not do 1MB IOs to HW RAID
452 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
454 Severity : enhancement
456 Description: lfs quota output is not very convenient for awk/sed-parsing
457 Details : Some positions in lfs quota output table could be empty or
458 non-empty which made it hard to parse it with scripts, now
459 a dash is put instead of space where there is not supposed
464 Description: Update to OFED 1.5.1
465 Details : OFED 1.5.1 will be used in environments that don't use
466 vendor supplied OFED.
470 Description: Remove lnet. prefix from lctl params display
471 Details : Also renamed lnet.memused to lnet_memused to
476 Description: Use conf_param -d to remove permanant settings.
480 Description: Add list_param -R
481 Details : lctl list_param now supports recursive mode.
484 Bugzilla : 21829, 22312
485 Description: Fix llobdstat.
486 Details : 1. Now updating the write_bytes counter
487 2. Port script fixes from 1.8.
491 Description: decrease the usage of memory on clients.
492 Details : 1. On clients, recycle dentries and inodes unused.
493 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
498 Description: llobdstat fix and enhancement
499 Details : add a counter to set a limit to how many samples will be returned
500 fix a wildcard in the path to limit to obdfilter stats only
504 Description: lfs check only allowed for root
505 Details : Code cleanups and added a regresion test of non-root use of
510 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
511 Details : Fix hr->hr_index overload.
515 Description: Fix FIEMAP functionality.
519 Description: FC11 patchless client support added
523 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
524 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
528 Description: New acc-sm script and various test-framework enhancments.
532 Description: Previously landed patch was reverted.
536 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
540 Description: Add ldiskfs maxdirsize mount option.
544 Description: Fix mkfs.lustre support for 16TB LUNs.
548 Description: Print mmp_check_interval
549 Details : print mmp_check_interval and make it possible to abort
550 mount operation in case it takes too long.
554 Description: Add cascading_rw to lustre mpi tests.
558 Description: Store SOM data in EA.
559 Details : Also several SOM related cleanups.
563 Description: Add refcount to md_open_data
567 Description: Use refcount on lov_request_set to prevent memory corruption.
571 Description: Statahead: use parent dir's i_mutex for synchronization.
575 Description: loadgen improvements.
579 Description: Enable peer health detection in ksocklnd and ko2iblnd.
583 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
587 Description: Fix race on obd_devcie:md_stats
591 Description: Fix SLES 11 client builds.
592 Details : Resolved uninitalized pointer gcc warnings.
596 Description: Allow non-root access for "lfs check".
597 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
601 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
603 Severity : enhancement
605 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
606 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
610 Description: Sanity test_27p fails.
611 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
612 Sanity tests 27 were cleaned up and improved.
616 Description: Unify req format on client/servers
617 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
618 buffers and swabbers approach..
623 Description: drop join file stuff.
625 Severity : enhancement
627 Description: Add 2.6.27 and SLES11 (patchless client) support
631 Description: Fix in ptlrpc_expire_one_request() to print the signed time
634 Severity : enhancement
636 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
639 Severity : enhancement
641 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
643 Severity : enhancement
645 Description: correctly handle big reply message.
646 Details : send LNet event if reply is bigger than buffer and adjust this buffer
651 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
653 Severity : enhancement
655 Description: Add support for OEL5.
659 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
661 Severity : enhancement
663 Description: "weak-modules" support
664 Details : Implement "weak-modules" support which enables kernel modules
665 to be used with any kernel that implements the same kABI. In
666 order to achieve this modules are now installed in
667 /lib/modules/$(uname -r)/updates/kernel on all distributions.
671 Description: client could not reconnect to OST because of an active request.
672 Details : The client now retries to reconnect to the same server, if a
673 connect request failed with EBUSY or -EAGAIN. Test case 35b
674 conf-sanity.sh has been created.
678 Description: Don't allow make backward step on assignin osc next id.
679 Details : race between allocation next id and ll_sync thread can be cause
680 of set wrong osc next id and can be kill valid ost objects.
684 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
688 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
692 Description: Error handling in osc_statfs_interpret() has been improved.
693 Details : Check in osc_statfs_interpret() for EBADR.
697 Description: Do not send statfs() requests to OSTs disabled by administrator.
698 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
700 Severity : enhancement
702 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
704 Severity : enhancement
706 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
708 Severity : enhancement
710 Description: Enhancement for lfs(1) command to use numeric uid/gid.
714 Description: Improved cleanup procedure for conf-sanity test_46a.
718 Description: Do not start bulk transfer, if client tried to reconnect.
719 Details : Do not start bulk transfer, if client tried to reconnect,
720 but failed due to active rpcs. Check for exp_abort_active_req
721 has been implemented.
725 Description: Prevent inconsistences between linux and lustre mount structures.
726 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
727 Make the sleep interruptible.
729 Severity : enhancement
731 Description: provide server to client comms path
732 Details : server to client communications path via new LDLM_SET_INFO rpc
736 Description: truncate starts GFP_FS allocation under transaction causing deadlock
737 Details : ldiskfs_truncate calls grab_cache_page which may start page
738 allocation under an open transaction. This may lead to
739 calling prune_icache with consequent lustre reentrance.
743 Description: send correctly lsm on open replay
744 Details : MDS is trust to LSM size on replay open, but client can set wrong size
749 Description: enable client interface failover
750 Details : When a child reconnects from another NID, properly update export
751 nid hash position and ldlm reverse import.
753 Severity : enhancement
755 Description: Communicate OST degraded/readonly state via statfs to MDS
756 Details : Flags in the statfs returned from OSTs indicate whether the
757 OST is in a degraded RAID state, or if the filesystem has
758 turned read-only after a filesystem error is detected.
763 Description: don't panic if EPROTO was hit when reading symlink
764 Details : correctly handling request reference in error cases.
766 Severity : enhancement
768 Description: Add LustreNetLink, a kernel-userspace communcation path.
770 Severity : enhancement
772 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
777 Description: access to llog context before init.
778 Details : move handling CATALOGS file at osc layer and forbid access to llog
782 Frequency : rare, only when client is writing to OST on local node
784 Description: Avoid deadlock for local client writes
785 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
786 memory freeing context. This allows OST threads to set the
787 PF_MEMALLOC flag on task structures in order to allocate memory
788 from reserved pools and complete IO.
789 Use GFP_HIGHUSER for OST allocations for non-local client writes,
790 so that the OST threads generate memory pressure and allow
791 inactive pages to be reclaimed.
793 Severity : enhancement
795 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
800 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
801 Details : move ldlm namespace creation in setup phase to avoid grab
802 _lprocfs_lock with cli_sem held.
806 Description: Temporarily disable grant shrink.
807 Details : Disable the feature for debugging.
810 Frequency : only when formatting test filesystems
812 Description: Unable to run several mkfs.lustre on loop devices at the same time
813 Details : mkfs.lustre returns error 256 on the concurrent loop devices
814 formatting. The solution is to proper handle the error.
816 Severity : enhancement
818 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
820 Severity : enhancement
822 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
827 Description: resolve race between obd_disconnect and class_disconnect_exports
828 Details : if obd_disconnect will be called to already disconnected export he
829 forget release one reference and osc module can't unloaded.
831 Severity : enhancement
833 Description: Allow tuning service thread via /proc
834 Details : For each service a new
835 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
836 created that can be used to set min/max thread counts, and get the
837 current number of running threads.
842 Descriptoin: don't return error if have partially created objects for file.
843 Details : lov_update_create_set uses set->set_success as index for created
844 objects, so if some requests will be failed, they will have hole
845 at end of array and qos_shrink_lsm for allocate correct lsm.
847 Severity : enhancement
849 Description: Update OFED support to 1.4.1
854 Description: Reduce small size read RPC
855 Details : Set read-ahead limit for every file and only do read-ahead when
856 available read-ahead pages are bigger than 1M to avoid small size
859 Severity : enhancement
861 Description: move AT tunable parameters for more consistent usage
862 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
864 Severity : enhancement
866 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
867 Details : allow skip disconnected ost for send statfs request and hide error
873 Description: fix racy locking of mballoc block bitmaps causing BUG
874 Details : The locking of the mballoc buddy bitmap and the in-memory
875 block bitmap was using two different spin locks in some
876 cases. This made it possible to incorrectly access the
877 mballoc bitmap while another process was modifying it,
878 causing a sanity assertion to fail. While no on-disk corruption
879 was reported, there was some risk of this happening.
882 Frequency : rare, on llog test 6
884 Descriptoin: don't allow connect to already connected import
885 Details : allowing connect to already connected import is hide connecting problem.
890 Descriptoin: allow kill process which wait statahead result
891 Details : for some reasons 'ls' can stick in waiting result from statahead,
892 in this case need way for kill this process.
894 Severity : enhancement
896 Description: Add state history info file, enhance import info file
897 Details : Track import connection state changes in a new osc/mdc proc file;
898 add overview-type data to the osc/mdc import proc file.
900 Severity : enhancement
902 Description: MDS create should not wait for statfs RPC while holding DLM lock.
904 Severity : enhancement
906 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
911 Descriptoin: OST grant too much space to client even there are not enough space.
912 Details : Client will shrink its grant cache to OST if there are no write
913 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
914 this grant cache if there are already not enough avaible space
915 (left_space < total_clients * 32M).
918 Frequency : start MDS on uncleanly shutdowned MDS device
920 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
921 Details : stay in waiting mds<>ost recovery finished produce random bugs
922 due race between two ll_sync thread for one lov target. send
923 ACTIVATE event only if connect realy finished and import have
927 Frequency : rare, connect and disconnect target at same time
929 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
930 Details : don't call obd_disconnect under lov_lock. this long time
931 operation and can block ptlrpcd which answer to connect request.
936 Descriptoin: don't lose wakeup for imp_recovery_waitq
937 Details : recover_import_no_retry or invalidate_import and import_close can
938 both sleep on imp_recovery_waitq, but we was send only one wakeup
942 Frequency : always with long access acl
944 Descriptoin: mds can't pack reply with long acl.
945 Details : mds don't control size of acl but they limited by reint/getattr
948 Severity : enhancement
950 Description: Update to SLES10 kernel-2.6.16.60-0.33.
952 Severity : enhancement
954 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
957 Frequency : start MDS on uncleanly shutdowned MDS device
959 Descriptoin: aborting recovery hang on MDS
960 Details : don't throttle destroy RPCs for the MDT.
963 Frequency : on remount
965 Description: external journal device not working after the remount
966 Details : clear dev_rdonly flag for external journal devices in
972 Description: shutdown vs evict race
973 Details : client_disconnect_export vs connect request race.
974 if client will evicted at this time - we start invalidate
975 thread without referece to import and import can be freed
979 Frequency : rare, need acl's on inode.
981 Description: client can't handle ost additional correctly
982 Details : if ost was added after client connected to mds client can have
983 hit lnet_try_match_md ... to big messages to wide striped files.
984 in this case need teach client to handle config events about add
985 lov target and update client max ea size at that event.
987 Severity : enhancement
989 Description: Changelogs
990 Details : Changelogs are a lightweight mechanism to track filesystem
991 metadata and namespace changes. The changelog is recorded
992 permanently on the MDTs, and is periodically "consumed" / purged
993 when records are no longer needed.
995 Severity : enhancement
997 Description: compact fld format with extents
998 Details : Store range of seq rather than every seq in FLD. Seq
999 controller update FLD rather than clients. In Case of CMD, mdt0
1000 has FLD, all other metadata server act as non persistent proxy
1001 for FLD queries and cache fld entries in fld cache.
1006 Description: don't skip ost target if they assigned to file
1007 Details : Drop slow OSCs if we can, but not for requested start idx.
1008 This means "if OSC is slow and it is not the requested
1009 start OST, then it can be skipped, otherwise skip it only
1010 if it is inactive/recovering/out-of-space.
1014 Description: more cleanup in mds_lov
1015 Details : not send LOV EA under replay, we can't know about they size at this
1016 time. Don't allow client connect to mds before any ost connected,
1017 for avoid problems with LOV EA size and returning EIO to client.
1019 Severity : enhancement
1021 Description: Interoperability at server side (Disk interoperability)
1023 Severity : enhancement
1025 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1027 Severity : enhancement
1029 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1031 Severity : enhancement
1033 Description: New client IO stack (CLIO).
1035 Severity : enhancement
1037 Description: Commit on sharing. Eliminate inter-client dependencies between
1038 uncommitted transactions by doing transaction commits.
1039 Thereby clients may recovery independently.
1042 Frequency : Create a symlink file with a very long name
1044 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1045 Details : If there is no extra space in the request for early cancels,
1046 ldlm_req_handles_avail() returns 0 instead of a negative value.
1048 Severity : enhancement
1050 Description: Add /proc entry for import status
1051 Details : The mdc, osc, and mgc import directories now have
1052 an import directory that contains useful import data for debugging
1053 connection problems.
1055 Severity : enhancement
1057 Description: Re-disable certain /proc logging
1058 Details : Enable and disable client's offset_stats, extents_stats and
1059 extents_stats_per_process stats logging on the fly.
1062 Frequency : Only on FC kernels 2.6.22+
1064 Description: oops in statahead
1065 Details : Do not drop reference count for the dentry from VFS when lookup,
1066 VFS will do that by itself.
1068 Severity : enhancement
1070 Description: Generic /proc file permissions
1071 Details : Set /Proc file permissions in a more generic way to enable non-
1072 root users operate on some /proc files.
1076 Description: Hitting mdc_commit_close() ASSERTION
1077 Details : Properly handle request reference release in
1078 ll_release_openhandle().
1082 Description: quota recovery deadlock during mds failover
1083 Details : This patch includes att18982, att18236, att18237 in bz14840.
1085 1. fix osts hang when mds does failover with quotaon
1086 2. prevent watchdog storm when osts threads wait for the
1091 Frequency : only patchless client
1092 Description: add workaround for race between add/remove dentry from hash
1094 Severity : enhancement
1096 Description: Allow OST glimpses to return PW locks
1100 Description: LBUG when llog conf file is full
1101 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1105 Description: Prevent import from entering FULL state when server in recovery
1109 Description: service mount cannot take device name with ":"
1110 Details : Only when device name contains ":/" will mount treat it as
1116 Description: replace ptlrpcd with the statahead thread to interpret the async
1117 statahead RPC callback
1121 Frequency : on recovery
1122 Description: I/O failures after umount during fail back
1123 Details : if client reconnected to restarted server we need join to recovery
1124 instead of find server handler is changed and process self
1125 eviction with cancel all locks.
1127 Severity : enhancement
1129 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1131 Severity : enhancement
1133 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1135 Severity : enhancement
1137 Description: Upcall on Lustre log has been dumped
1138 Details : Allow for a user mode script to be called once a Lustre log has
1139 been dumped. It passes the filename of the dumped log to the
1140 script, the location of the script can be specified via
1141 /proc/sys/lnet/debug_log_upcall.
1146 Description: avoid idr_remove called for id which is not allocated.
1147 Details : Move assigment s_dev for clustered nfs to end of initialization,
1148 to avoid problem with error handling.
1153 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1154 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1155 into CONN_USED_HASH and this prodice warning when put connection
1156 again in unused hash.
1159 Severity : enhancement
1161 Description: File striping can now be set to use an arbitrary pool of OSTs.
1163 Severity : enhancement
1165 Description: Export bytes_read/bytes_write count on OSC/OST.
1169 Description: Early reply size mismatch, MGC loses connection
1170 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1171 the connect flags are properly negotiated.
1176 Description: quotas are not honored with O_DIRECT
1177 Details : all writes with the flag O_DIRECT will use grants which leads to
1178 this problem. Now using OBD_BRW_SYNC to guard this.
1182 Description: add quota statistics
1183 Details : 1. sort out quota proc entries and proc code.
1184 2. add quota statistics
1186 Severity : enhancement
1188 Description: enable quota support for HEAD.
1192 Description: Properly propagate oinfo flags from lov to osc for statfs
1193 Details : restore missing copy oi_flags to lov requests.
1195 Severity : enhancement
1197 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1198 : l_getgroups(8), lst(8), routerstat(8)
1199 Details : included man pages for llobdstat(8), llstat(8),
1200 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1202 Severity : enhancement
1204 Description: configure's --enable-quota should check the
1205 : kernel .config for CONFIG_QUOTA
1206 Details : configure is terminated if --enable-quota is passed but
1207 : no quota support is in kernel
1211 Description: Remove portals compatibility
1212 Details : Remove portals compatibility, not interoperable with releases
1217 Description: Resolve device initialization race
1218 Details : Prevent proc handler from accessing devices added to the
1219 obd_devs array but yet be intialized.
1221 Severity : enhancement
1223 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1225 Severity : enhancement
1227 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1232 Description: Using wrong pointer in osc_brw_prep_request
1233 Details : Access to array[-1] can produce panic if kernel compiled with
1234 CONFIG_PAGE_ALLOC enabled
1238 Description: Client runs out of low memory
1239 Details : Consider only lowmem when counting initial number of llap pages
1243 Description: *optional* service tags registration
1244 Details : if the "service tags" package is installed on a Lustre node
1245 When the filesystem is mounted, a local-node service tag will
1246 be created. See http://inventory.sun.com/ for more information
1247 about the Service Tags asset management system.
1251 Description: Kernel BUG tries to release flock
1252 Details : Lustre does not destroy flock lock before last reference goes
1253 away. So always drop flock locks when client is evicted and
1254 perform unlock regardless of successfulness of speaking to MDS.
1258 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1262 Description: sanity test 65a fails if stripecount of -1 is set
1263 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1268 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1269 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1272 Severity : enhancement
1274 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1278 Description: do not process already freed flock
1279 Details : flock can possibly be freed by another thread before it reaches
1280 to ldlm_flock_completion_ast.
1284 Description: LBUG during stress test
1285 Details : Need properly lock accesses the flock deadlock detection list.
1289 Description: oops in page fault handler
1290 Details : kernel page fault handler can return two special 'pages' in error
1291 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1295 Description: timeout with invalidate import.
1296 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1297 should be handled by ptlrpcd. This produce long age waiting and
1298 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1300 Severity : enhancement
1302 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1306 Description: enable MGS and MDT services start separately
1307 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1308 a co-located MGS without starting the MGS, which is a complement
1309 to 'nosvc' mount option.
1313 Frequency : after recovery
1314 Description: precreate to many object's after del orphan.
1315 Details : del orphan st in oscc last_id == next_id and this triger growing
1316 count of precreated objects. Set flag LOW to skip increase count
1317 of precreated objects.
1321 Frequency : rare, on clear nid stats
1322 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1323 Details : when clean nid stats sometimes try destroy live entry,
1324 and this produce panic in free.
1328 Description: Stack overflow during MDS log replay
1329 ease stack pressure by using a thread dealing llog_process.
1333 Description: wait until IO finished before start new when do lock cancel.
1334 Details : VM protocol want old IO finished before start new, in this case
1335 need wait until PG_writeback is cleared until check dirty flag and
1336 call writepages in lock cancel callback.
1338 Severity : enhancement
1340 Description: using special macro for print time and cleanup in includes.
1344 Description: mds_mfd_close() ASSERTION(rc == 0)
1345 Details : In mds_mfd_close(), we need protect inode's writecount change
1346 within its orphan write semaphore to prevent possible races.
1350 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1354 Frequency : rare, on shutdown ost
1355 Description: don't hit live lock with umount ost.
1356 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1357 use shrink_dcache_sb instead.
1361 Description: don't panic with use echo client
1362 Details : echo client pass NULL as client nid pointer and this produce null
1363 pointer dereference.
1367 Description: fix build on ppc32
1368 Details : compile code with -m64 flag produce wrong object file for ppc32.
1372 Description: add message levels for liblustreapi
1376 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1377 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1378 will result in an -EAGAIN and in the exit call path we call
1379 alloc_rr() which will with increasing aggressiveness attempt to
1380 aquire precreated objects on the minimum number of required OSCs.
1384 Description: Use old size assignment to avoid deadlock
1385 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1386 the scheduling while holding a spinlock. We do not need locking
1387 for size in ll_update_inode() because size is only updated from
1388 the MDS for directories or files without objects, so there is no
1389 other place to do the update, and concurrent access to such inodes
1390 are protected by the inode lock.
1394 Description: resolve "_IOWR redefined" build error on SLES10
1398 Description: dump the memory debugging after all modules are unloaded to
1399 suppress false negative in conf_sanity test 39
1401 Severity : enhancement
1403 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1408 Description: fix SLES kernel versioning
1409 Details : the kernel version for our SLES 10 kernel did not include a "-"
1410 before the "smp" at the end. while this was not a problem in
1411 general, it did mean that software trying to use the kernel
1412 version to try to detect a vendor specific kernel would fail.
1413 this was most evident by the OFED build scripts.
1417 Description: Don't update lov_desc members until making sure they are valid
1418 Details : When updating lov_desc members via proc fs, need fix their
1419 validities before doing the real update.
1423 Description: don't put request into delay list while invalidate in flight.
1424 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1425 import in flight. this produce timeout for invalidate and sometimes
1426 can cause stale data.
1430 Frequency : on ppc only
1431 Description: not convert ost objects for directory because it's not exist.
1432 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1436 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1438 Severity : enhancement
1440 Description: Update to support 2.6.22.14 vanilla kernel.
1444 Frequency : rare, on recovery
1445 Description: read procfs can produce deadlock in some situation
1446 Details : Holding lprocfs lock which send rpc can produce block for destroy
1447 obd objects and this also block reconnect with -EALREADY. This
1448 isn't fix all lprocfs bugs - but make it rare.
1450 Severity : enhancement
1452 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1455 Frequency : frequent on X2 node
1457 Description: mdc_set_open_replay_data LBUG
1458 Details : Set replay data for requests that are eligible for replay.
1462 Description: lustre_mgs: operation 101 on unconnected MGS
1463 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1464 MGC, and late on MGC cannot successfully connect to MGS and a lot
1465 of the error messages complaining that MGS is not connected.
1469 Frequency : on network error
1470 Description: panic with double free request if network error
1471 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1472 but it's true only for synchronus enqueue, for async enqueue
1473 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1476 Severity : enhancement
1478 Description: client-side metadata stat-ahead during readdir(directory readahead)
1479 Details : perform client-side metadata stat-ahead when the client detects
1480 readdir and sequential stat of dir entries therein
1483 Frequency : on start mds
1485 Description: Implement get_info(last_id) in obdfilter.
1488 Frequency : occasional
1490 Description: Correctly check stale fid, not start epoch if ost not support SOM
1491 Details : open with flag O_CREATE need set old fid in op_fid3 because
1492 op_fid2 was overwritten with new generated fid, but mds can answer
1493 with one of these two fids and both is not stale. Setattr
1494 incorrectly started an epoch and assume will be called
1495 done_writing, but without SOM done_writing ever being called.
1498 Frequency : rare, depends on device drivers and load
1500 Description: MDS or OSS nodes crash due to stack overflow
1501 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1502 In some cases, in conjunction with device drivers that use a lot
1503 of stack the MDS (or possibly OSS) service threads could overflow
1504 the stack. One change which was identified to consume additional
1505 stack has been reworked to avoid the extra stack usage.
1508 Frequency : occasional
1510 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1511 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1512 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1513 oscc_has_objects and see OSC has no objests and call
1514 oscc_internal_create to resend create request.
1516 Severity : enhancement
1518 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1520 Severity : enhancement
1522 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1525 Frequency : very rare
1527 Description: Fix replay if there is an un-replied request and open
1528 Details : In some cases, older replay request will revert the
1529 mcd->mcd_last_xid on MDS which is used to record the client's
1530 latest sent request.
1532 Severity : enhancement
1534 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1536 Severity : enhancement
1538 Description: Add rhel5 support to HEAD.
1540 Serverity : enhancement
1542 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1547 Description: Don't allow skipping OSTs if index has been specified.
1548 Details : Don't allow skipping OSTs if index has been specified, make
1549 locking in internal create lots better.
1553 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1554 Details : access to bitfield in structure is always rounded to long
1555 and this produce problem with not atomic change any bit.
1559 Description: Lustre make rpms failed.
1560 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1561 builds Lustre rpms from tarball.
1564 Frequency : rare on shutdown ost
1566 Description: If llog cancel was not send before clean_exports phase, this can
1567 produce deadlock in llog code.
1568 Details : If llog thread has last reference to obd and call class_import_put
1569 this produce deadlock because llog_cleanup_commit_master wait when
1570 last llog_commit_thread exited, but this never success because was
1571 called from llog_commit_thread.
1575 Description: allow userland application know is lost one of stripes.
1576 Details : fill lvb_blocks with error code on ost and return it to
1577 application if error flag found.
1581 Description: NULL lov_tgts causing MDS oops
1582 Details : more safe checks for NULL lov_tgts for avoid oops.
1584 Severity : enhancement
1586 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1590 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1591 Details : If an OST is down the MDS will hang indefinitely in
1592 obd_statfs_rqset() waiting for the statfs data. While for
1593 MDS QOS usage of statfs, it should not stuck in waiting.
1595 Severity : enhancement
1597 Description: remote_acl support
1598 Details : Support ACL-based permission check for remote user.
1599 Support setfacl/getfacl for remote user with the utils
1600 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1601 the system "{s,g}etfacl" utils.
1603 Severity : enhancement
1605 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1607 Severity : enhancement
1609 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1613 Description: make mgs_setparam() handle fsname containing dash
1614 Details : fsname containing a dash does not work with lctl conf_param
1616 Severity : enhancement
1618 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1620 Severity : enhancement
1622 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1626 Frequency : rare, at shutdown
1627 Description: access already free / zero obd_namespace.
1628 Details : if client_disconnect_export was called without force flag set,
1629 and exist connect request in flight, this can produce access to
1630 NULL pointer (or already free pointer) when connect_interpret
1631 store ocd flags in obd_namespace.
1635 Frequency : only at startup
1636 Description: not alloc memory with spinlock held.
1637 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1638 if any spinlock held.
1640 Severity : enhancement
1642 Description: make lustre randomly fail allocating memory
1643 Details : Make lustre randomly failed allocating memory for testing purpose.
1645 Severity : enhancement
1647 Description: lost problems with lov objid file
1648 Details : Fixes some scability and access to not inited memory problems
1649 in work with lov objdid file.
1654 Description: lfs find does not continue on file error
1655 Details : Continue other files processing when a file/dir is absent.
1659 Description: Inconsistent usage of lustre_pack_reply()
1660 Details : Standardize the usage of lustre_pack_reply() such that it
1661 always generate a CERROR on failure.
1664 Frequency : occasional
1666 Description: MDS hang or stay in waiting lock
1667 Details : If client receive lock with CBPENDING flag ldlm need send lock
1668 cancel as separate rpc, to avoid situation when cancel request
1669 can't processed due all i/o threads stay in wait lock.
1673 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1677 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1681 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1685 Description: Update to RHEL4 latest kernel.
1687 Severity : enhancement
1689 Description: Build SLES10 patchless client fails
1690 Details : The configure was broken by run ./configure with
1691 --with-linux-obj=.... argument for patchless client. When the
1692 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1693 can't search header adequately. Use absolute path such as
1694 -I($LINUX)/include instead.
1698 Description: interrupt oig_wait produce painc on resend.
1699 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1700 private set, and this produce situation when rq_ptlrpcd_data not
1701 copyed to new allocated request and triggered LBUG on assert
1702 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1703 ptlrpcd set if request is changed and can be safety changed to use
1708 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1709 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1710 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1711 these macros and replace them with their difinitions instead.
1715 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1716 Details : Update which_patch & target file for SLES10 latest kernel.
1719 Description: Remove client patches from SLES 10 kernel.
1720 Details : This causes SLES 10 clients to behave as patchless clients
1721 even on a Lustre-patched (server) kernel.
1723 Severity : enhancement
1725 Description: self-adjustable client's lru lists
1726 Details : use adaptive algorithm for managing client cached locks lru
1727 lists according to current server load, other client's work
1728 pattern, memory activities, etc. Both, server and client
1729 side namespaces provide number of proc tunables for controlling
1732 Severity : enhancement
1734 Description: light-weight GSS support
1735 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1736 and reduce performance overhead.
1738 Severity : enhancement
1740 Description: Linux keyring support
1741 Details : Support using service of Linux keyring for Lustre GSS internal
1742 context refresh/cache mechanism.
1746 Description: Fix errors in lfs documentation
1747 Details : Fixes man pages
1751 Description: don't use GFP_* in generic Lustre code.
1752 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1756 Description: obdclass is limited by single OBD_ALLOC(idarray)
1757 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1761 Description: warnings with build patchless client with vanila 2.6.19 and up
1762 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1763 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1767 Frequency : Only for SLES
1768 Description: sanity_quota fail test_1
1769 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1770 file, which makes TSTID[2] inunique.
1774 Frequency : Only for SLES9
1775 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1776 Details : Update target/ChangeLog/which_patch .
1780 Frequency : Only for test_10 in sanity-quota.sh
1781 Description: a bug in quota test code
1782 Details : checking a test flag in a wrong place
1786 Frequency : Only for test_13 in sanity-quota.sh
1787 Description: a bug in quota test script
1788 Details : shouldn't check overfull allocated quota
1792 Description: jbd statistics
1793 Details : Port older jbd statistics patch for sles10
1795 Severity : enhancement
1797 Description: Add printing inode info into message about error in writepage.
1802 Description: reply_lock_interpret crash due to race with it and lock cancel.
1803 Details : Do not replay locks that are being cancelled. Do not reference
1804 locks by their address during replay, just by their handle.
1808 Frequency : When flocks are used.
1809 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1810 Details : Flock locks might destroy just granted lock if it could be merged
1811 with another existing flock, this is done in completion handler,
1812 so teach ldlm_cli_enquque_fini that this is a valid case for
1818 Description: Oops in read and write path when failing to allocate lock.
1819 Details : Check if lock allocation failed and return error back.
1823 Description: lstripe command fails for valid OST index
1824 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1825 instead of lov->desc.ld_active_tgt_count.
1829 Description: sanity.sh failed test 103
1830 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1835 Description: df doesn't work properly if diskfs blocksize != 4K
1836 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1840 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1841 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1843 Severity : enhancement
1845 Description: Adaptive timeouts
1846 Details : RPC timeouts adapt to changing server load and network
1847 conditions to reduce resend attempts and improve recovery time.
1851 Description: llapi_file_create() does not allow some changes
1852 Details : add llapi_file_open() that allows specifying the mode and
1853 open flags, and also returns an open file handle.
1857 Description: merge and cleanup kernel patches.
1858 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1862 Description: Add journal checksum support.(Kernel part)
1863 Details : The journal checksum feature adds two new flags i.e
1864 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1865 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1866 indicates that the commit block contains the checksum for
1867 the blocks described by the descriptor blocks. Now commit
1868 record can be sent to disk without waiting for descriptor
1869 blocks to be written to disk. This behavior is controlled
1870 using JBD2_FEATURE_ASYNC_COMMIT flag.
1874 Description: OSS needs mutliple precreate threads
1875 Details : Add ability to start more than one create thread per OSS.
1879 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1880 Details : execute lfs setstripe on client
1884 Description: mds_obd_create error creating tmp object
1885 Details : When the user sets quota on root, llog will be affected and can't
1886 create files and write files.
1889 Frequency : Always on ia64 patchless client, and possibly others.
1891 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1892 Details : This allows the patchless client to be loaded on architectures
1893 without this export.
1897 Description: RedHat Update kernel for RHEL5
1898 Details : Add the kernel config file for RHEL5.
1902 Description: RedHat Update kernel for RHEL5
1903 Details : Modify the kernel config file more closer RHEL5.
1907 Description: Build failure against Centos5 (RHEL5)
1908 Details : Define PAGE_SIZE when it isn't present.
1912 Description: test_7 on sanity-quota.sh can't be used on separate machines
1913 Details : add facet to handle it
1917 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1918 Details : with NFS, the anon dentry's parent was set to itself in
1919 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1920 corresponding dentry other than use rec->ur_name.
1922 Severity : enhancement
1924 Description: lfs setstripe enhancement
1925 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1929 Description: enable data checksumming by default
1930 Details : enable checksum by default, allow --disable-checksum
1931 configure option and "-o nochecksum" mount option. Checksums
1932 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1936 Description: lustre support for RHEL5
1937 Details : Add support for RHEL5.
1941 Description: Client eviction due to failover config
1942 Details : after a connection loss, the lustre client should attempt to
1943 reconnect to the last active server first before trying the
1944 other potential connections.
1948 Description: when mds and osts use different quota unit(32bit and 64bit),
1949 quota will be released repeatly.
1950 Details : void sending multiple quota reqs to mds, which will keep the
1951 status between the reqs.
1955 Description: osts not allocated evenly to files
1956 Details : change the condition to increase offset_idx
1959 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1961 Description: Data corruption for OSTs that are formatted larger than 2TB
1963 Details : When generating the bio request for lustre file writes the
1964 sector number would overflow a temporary variable before being
1965 used for the IO. The data reads correctly from Lustre (which
1966 will overflow in a similar manner) but other file data or
1967 filesystem metadata may be corrupted in some cases.
1971 Description: Tune the kernel for good SCSI performance.
1972 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1973 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1978 Description: rewrite ext2-derived code in obdclass/uuid.c
1979 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1980 from scratch preserving functionality.
1984 Description: "lfs find -obd UUID" prints directories
1985 Details : "lfs find -obd UUID" will return all directory names instead
1986 of just file names. It is incorrect because the directories
1987 do not reside on the OSTs.
1991 Description: use i_size_read and i_size_write in 2.6 port
1992 Details : replace inode->i_size access with i_size_read/write()
1995 Frequency : when removing large files
1997 Description: scheduling issue during removal of large Lustre files
1998 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1999 It causes scheduling issues when removing large files (17TB in the
2003 Frequency : only with liblustre clients on XT3
2005 Description: evictions taking too long
2006 Details : allow llrd to evict clients directly on OSTs
2009 Frequency : only on ppc
2011 Description: /proc/fs/lustre/devices broken on ppc
2012 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2013 We should make sure the type of 'index' is loff_t and then cast
2014 explicitly as needed below. Do not assign an explicitly cast
2018 Frequency : only for rhel5
2020 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2021 Details : Modify the target file & which_kernel.
2023 Severity : enhancement
2025 Description: omit set fsid for export NFS
2026 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2027 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2031 Frequency : Always, for kernels after 2.6.16
2032 Description: Fix warning idr_remove called for id=.. which is not allocated.
2033 Details : Last kernels save old s_dev before kill super and not allow
2034 to restore from callback - restore it before call kill_anon_super.
2038 Description: buffer overruns could theoretically occur
2039 Details : llapi_semantic_traverse() modifies the "path" argument by
2040 appending values to the end of the origin string, and a
2041 overrun may occur. Adding buffer overrun check in liblustreapi.
2045 Description: Fix error on 'ls .' at the top of the Lustre mount.
2046 Details : Don't revalidate dentry if it is a root dentry.
2050 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2051 Details : Modify vm-tunables-rhel4.patch.
2055 Description: Kernel config for 2.6.18-vanilla.
2056 Details : Modify targets/2.6-vanilla.target.in.
2057 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2058 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2059 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2060 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2064 Description: improve handling recoverable errors
2065 Details : if request processig with error which can be recoverable on server
2066 request should be resend, otherwise page released from cache and
2071 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2072 Details : Modify target file & which_patch.
2073 A flaw was found in the IA32 system call emulation provided
2074 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2075 value could be stored in the %RAX register, which could trigger an
2076 out-of-bounds system call table access. An untrusted local user
2077 could exploit this flaw to run code in the kernel
2078 (ie a root privilege escalation). (CVE-2007-4573).
2082 Description: O_DIRECT bypasses client statistics.
2083 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2084 read_ahead_stats were updated. copied the stats section from
2085 osc_send_oap_rpc() into async_internal().
2089 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2093 Description: change order of libsysio includes
2094 Details : '#include sysio.h' should always come before '#include xtio.h'
2098 Description: handle "serious error: objid * already exists" more gracefully
2099 Details : If LAST_ID value on disk is smaller than the objects existing in
2100 the O/0/d* directories, it indicates disk corruption and causes an
2101 LBUG(). If the object is 0-length, then we should use the existing
2102 object. This will help to avoid a full fsck in most cases.
2104 Severity : enhancement
2106 Description: adapt the lustre_config script to support the upgrade case
2107 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2108 server targets to 1.6 in parallel.
2112 Description: To avoid grant space > avaible space when the disk is almost
2113 full. Without this patch you might see the error "grant XXXX >
2114 available" or some LBUG about grant, when the disk is almost
2116 Details : In filter_check_grant, for non_grant cache write, we should
2117 check the left space by if (*left > ungranted + bytes), instead
2118 of (*left > ungranted), because only we are sure the left space
2119 is enough for another "bytes", then the ungrant space should be
2120 increase. In client, we should update cl_avail_grant only there
2121 is OBD_MD_FLGRANT in the reply.
2125 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2126 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2127 described in CVE-2007-4573. This problem could allow untrusted
2128 local users to gain root access.
2131 Frequency : when using O_DIRECT and quotas
2133 Description: Incorrect file ownership on O_DIRECT output files
2134 Details : block usage reported by 'lfs quota' does not take into account
2135 files that have been written with O_DIRECT.
2140 Description: touch file failed when fs is not full
2141 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2142 otherwise we can get ENOSP while fs is not full.
2146 Description: parallel lock callbacks
2147 Details : Instead of sending blocking and completion callbacks as separated
2148 requests, adding them to a set and sending in parallel.
2151 Frequency : only for Cray XT3
2152 Bugzilla : 12829/13455
2153 Description: Changing primary group doesn't change the group lustre assigns to
2155 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2156 overridden with the primary group provided by the group upcall,
2157 whereas the supplied fsgid can be trusted if it is in the list of
2158 supplementary groups returned by the group upcall.
2160 Severity : enhancement
2162 Description: Allow masking D_WARNING, D_ERROR messages from console
2163 Details : Console messages can now be disabled via lnet.printk.
2167 Description: User code with malformed file open parameter crashes client node
2168 Details : Before packing join_file req, all the related reference should be
2169 checked carefully in case some malformed flags cause fake
2170 join_file req on client.
2174 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2175 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2176 mode is equal to the granted mode. However, it should also check
2177 that LDLM_FL_CP_REQD is not set.
2181 Description: Heavy nfs access might result in deadlocks
2182 Details : After ELC code landed, it is now improper to enqueue any mds
2183 locks under och_sem, because enqueue might want to decide to
2184 cancel open locks for same inode we are holding och_sem for.
2188 Description: Client eviction while running blogbench
2189 Details : A lot of unlink operations with concurrent I/O can lead to a
2190 deadlock causing evictions. To address the problem, the number of
2191 oustanding OST_DESTROY requests is now throttled to
2192 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2193 callbacks are processed in priority.
2197 Description: enable ACLs on MDS by default
2198 Details : ACLs must be enabled on MDS by default.
2201 Frequency : PPC/PPC64 only
2203 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2204 Details : fix duplicated definitions between asm-ppc64/types.h and
2205 lustre_types.h on PPC.
2208 Frequency : PPC/PPC64 only
2210 Description: asm-ppc/segment.h does not exist
2211 Details : fix compile issue on PPC.
2215 Description: better handle error messages in extents code
2218 Frequency : RHEL4 only
2220 Description: mkfs is very slow on IA64/RHEL4
2221 Details : A performance regression has been discovered in the MPT Fusion
2222 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2223 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2224 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2226 Severity : enhancement
2228 Description: SNMP support enhancement
2229 Details : Adding total number of sampled request for an MDS node in snmp
2232 Severity : enhancement
2234 Description: Optimize ldlm waiting list processing for PR extent locks
2235 Details : When processing waiting list for read extent lock and meeting read
2236 lock that is same or wider to it that is not contended, skip
2237 processing rest of the list and immediatelly return current
2238 status of conflictness, since we are guaranteed there are no
2239 conflicting locks in the rest of the list.
2243 Description: Time out and refuse to reconnect
2244 Details : When the failover node is the primary node, it is possible
2245 to have two identical connections in imp_conn_list. We must
2246 compare not conn's pointers but NIDs, otherwise we can defeat
2247 connection throttling.
2251 Description: port llog fixes from b1_6 into HEAD
2252 Details : Port llog reference couting and some llog cleanups from b1_6
2253 (bug 10800) into HEAD, for protect from panic and access to already
2254 free llog structures.
2258 Description: Detect stride IO mode in read-ahead
2259 Details : When a client does stride read, read-ahead should detect that and
2260 read-ahead pages according to the detected stride pattern.
2264 Description: data checksumming impacts single node performance
2265 Details : add support for several checksum algorithm. Currently, only CRC32
2266 and Adler-32 are supported. The checksum type can be changed on
2267 the fly via /proc/fs/lustre/osc/*/checksum_type.
2271 Description: use adler32 for page checksums
2272 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2277 Description: build for x2 fails
2278 Details : fix compile issue on Cray systems.
2282 Description: Properly match for duplicate locks
2283 Details : Due to different lock order from skiplists code, we need to
2284 traverse entire list for now
2287 Frequency : only on PPC/SLES10
2289 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2290 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2291 be defined. Add a hack in mkfs_lustre.c to work around this compile
2296 Description: LASSERT on MDS when client holding flock lock dies
2297 Details : ldlm pool logic depends on number of granted locks equal to
2298 number of released locks which is not true for flock locks, so
2299 just exclude such locks from consideration.
2303 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2304 Details : Use fsfilt_sync() for both the whole filesystem sync and
2305 individual file sync to eliminate dangerous inode locking
2306 with I_LOCK that can lead to a deadlock.
2310 Description: performance in 1.6.3
2311 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2312 sure that 1MB requests are not fragmented by the block layer.
2314 Severity : enhancement
2316 Description: organize the server-side client stats on per-nid basis
2317 Details : Change the structure of stats under obdfilter and mds to
2325 The "uuid"s file would list the uuids of _active_ exports.
2326 And the clear entry is to clear all stats and stale nids.
2328 Severity : enhancement
2330 Description: eliminate client locks in face of contention
2331 Details : file contention detection and lockless i/o implementation
2332 for contended files.
2336 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2339 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2341 Description: MDS may lose file striping (and hence file data) in some cases
2342 Details : If there are additional extended attributes stored on the MDS,
2343 in particular ACLs, SELinux, or user attributes (if user_xattr
2344 is specified for the client mount options) then there is a risk
2345 of attribute loss. Additionally, the Lustre file striping
2346 needs to be larger than default (e.g. striped over all OSTs),
2347 and an additional attribute must be stored initially in the
2348 inode and then increase in size enough to be moved to the
2349 external attribute block (e.g. ACL growing in size) for file
2354 Description: skiplist implementation simplification
2355 Details : skiplists are used to group compatible locks on granted list
2356 that was implemented as tracking first and last lock of each
2357 lock group the patch changes that to using doubly linked lists
2361 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2362 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2363 because LASSERT on both the data supplied by a client, and the
2364 data on disk is dangerous and incorrect.
2366 Severity : enhancement
2368 Description: Slow truncate/writes to huge files at high offsets.
2369 Details : Directly associate cached pages to lock that protect those pages,
2370 this allows us to quickly find what pages to write and remove
2371 once lock callback is received.
2375 Description: more ldlm soft lockups
2376 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2377 starve other threads from the resource lock for a long time in
2378 case of long waiting queue, so change the debug level from
2379 D_OTHER to the less frequently used D_INFO.
2381 Severity : enhancement
2383 Description: add -gid, -group, -uid, -user options to lfs find
2387 Description: Hung threads in invalidate_inode_pages2_range
2388 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2389 once one is completed. As a result, some RPCs are stuck in the
2390 queue and are never sent.
2394 Description: filter threads hungs on waiting journal commit
2395 Details : Cleanup filter group llog code, then only filter group llog will
2396 be only created in the MDS/OST syncing process.
2400 Description: Procfs and llog threads access destoryed import sometimes.
2401 Details : Sync the import destoryed process with procfs and llog threads by
2402 the import refcount and semaphore.
2404 Severity : enhancement
2406 Description: openlock cache of b1_6 port to HEAD
2411 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2412 Details : If insertion of an extent fails, then discard the inode
2413 preallocation and free data blocks else it can lead to duplicate
2418 Description: don't always update ctime in ext3_xattr_set_handle()
2419 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2420 In some cases the ctime should not be updated, for example for
2421 2.0->1.8 compatibility it is necessary to delete an xattr and it
2422 should not update the ctime.
2426 Bugzilla : 15713/16362
2427 Description: Assertion in iopen_connect_dentry in 1.6.3
2428 Details : looking up an inode via iopen with the wrong generation number can
2429 populate the dcache with a disconneced dentry while the inode
2430 number is in the process of being reallocated. This causes an
2431 assertion failure in iopen since the inode's dentry list contains
2432 both a connected and disconnected dentry.
2436 Description: assertion failure in ldlm_handle2lock()
2437 Details : fix a race between class_handle_unhash() and class_handle2object()
2438 introduced in lustre 1.6.5 by bug 13622.
2443 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2444 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2449 Description: oops in sd_iostats_seq_show()
2450 Details : unloading/reloading the scsi low level driver triggers a kernel
2451 bug when trying to access the sd iostat file.
2456 Description: Kernel panics during QLogic driver reload
2457 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2458 patch, causing memory corruption.
2463 Description: journal_dev option does not work in b1_6
2464 Details : pass mount option during pre-mount.
2466 Severity : enhancement
2468 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2469 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2470 extent information of a file. It can be used to map logical blocks
2471 in a file to physical blocks in the block device.
2475 Description: LDLM soft lockups - improvement
2476 Details : It is be possible to send the lock handle along with each read
2477 or write request because the client is already doing a lock match
2478 itself so there isn't any reason the OST should have to re-do that
2482 Frequency : only on Cray X2
2484 Description: X2 build failures
2485 Details : fix build failures on Cray X2.
2489 Description: xid & resent requests
2490 Details : Initialize RPC XID from clock at startup (randomly if clock is
2493 Severity : enhancement
2495 Description: Add lustre_start utility to start or stop multiple Lustre servers
2500 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2501 Details : In case of memory pressure, list_del() can be called twice on
2502 req->rq_history_list, causing a kernel oops.
2506 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2507 Details : fix stack overflow in the distributed lock manager by defering
2508 export eviction after a failed AST to the elt thread instead of
2509 handling it in the dlm interpret routine.
2513 Description: Convert some comments to new format.
2514 Details : Update documenting comments to match doxygen conventions.
2518 Description: Grammar fixes.
2519 Details : A couple of trivial spelling fixes.
2523 Description: OSD_COUNTERS-mandatory
2524 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2525 to simplify the code.
2529 Description: simplify cmm_device freeing logic.
2530 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2534 Description: Add lockdep support to dt_object_operations locking interface.
2535 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2536 parameter indicating lock ordering. Update mdd code to use new
2541 Description: Introduce failloc constants for lockless IO tests.
2542 Details : Add two new failloc constants to test lockless IO. Only one of
2543 them in implemented---another is checked in yet to be landed
2548 Description: Add lockdep support for inode mutex.
2549 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2550 in the situations where Lustre has to lock more than one inode
2555 Description: Add optional invariants checking support.
2556 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2557 configure switch. This macro is to be used for consistency and
2558 sanity checks that are too expensive to be left in `production'
2563 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2564 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2569 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2570 Details : Use spin_lock_nested() in (the only) situation where more than
2571 one ldlm_lock is locked simultaneously. Also, fix possible
2572 dead-lock in ldlm_lock_change_resource() by enforcing particular
2577 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2578 Details : Instead of specifying each ldlm_lock call-back through separate
2579 parameter, wrap them into struct ldlm_callback_suite.
2583 Description: Kill join_lru obd method and its callers.
2584 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2585 in memory---a requirement implemented through ->o_join_lru() obd
2586 method in HEAD. Remove this method and its users.
2590 Description: Add asynchronous ldlm ENQUEUE completion handler.
2591 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2592 case that stock ldlm_completion_ast() cannot handle as it waits
2593 until lock is granted. Introduce new ldlm_completion_ast_async()
2594 for this. Also comment ldlm_completion_ast().
2598 Description: ldlm_error <-> errno conversion.
2599 Details : Add functions to map (rather arbitrary) between LDLM error codes
2600 and standard errno values. CLIO needs this to prevent LDLM specific
2601 constants from escaping ldlm and osc.
2605 Description: Kill unused ldlm_handle2lock_ns() function.
2606 Details : Kill unused ldlm_handle2lock_ns() function.
2610 Description: Add lu_ref support to ldlm_lock
2611 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2612 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2613 and ldlm_resource. LDLM interface has to be changed, because code that
2614 releases a reference on a lock, has to "know" what reference this is.
2615 In the most frequent case
2617 lock = ldlm_handle2lock(handle);
2619 LDLM_LOCK_PUT(lock);
2621 no changes are required. When any other reference (received _not_ from
2622 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2623 instead of LDLM_LOCK_PUT().
2625 Arguably, changes are pervasive, and interface requires some discipline
2626 for proper use. On the other hand, it was very instrumental in finding
2627 a few leaked lock references.
2631 Description: Add ldlm_lock_addref_try().
2632 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2633 attempts to addref a lock that might be being canceled
2638 Description: Add ldlm_weigh_callback().
2639 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2640 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2641 hard-coded `number of pages' logic.
2645 Description: Add lockdep annotations to llog code.
2646 Details : Use appropriately tagged _nested() locking calls in the places
2647 where llog takes more than one ->lgh_lock lock.
2651 Description: Add loi_kms_set().
2652 Details : Wrap kms updates into a helper function.
2656 Description: Constify instances of struct lsm_operations.
2657 Details : Constify instances of struct lsm_operations.
2661 Description: lu_conf support.
2662 Details : On a server, a file system object is uniquely identified
2663 by a fid, which is sufficient to locate and load all object
2664 state (inode). On a client, on the other hand, more data are
2665 necessary instantiate an object. Change lu_object_find() and
2666 friends to take additional `lu_conf' argument describing object.
2667 Typically this includes layout information.
2671 Description: lu_context fixes.
2672 Details : Introduce new lu_context functions that are needed on the client
2673 side, where some system threads (ptlrpcd) are shared by multiple
2674 modules, and so cannot be stopped during module shutdown.
2678 Description: Add start and stop methods to lu_device_type_operations.
2679 Details : Introduce two new methods in lu_device_type_operations, that are
2680 invoked when first instance of a given type is created and last one
2681 is destroyed respectively. This is need by CLIO.
2685 Description: Add lu_ref support to struct lu_device.
2686 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2687 track leaked references.
2691 Description: Introduce lu_kmem_descr.
2692 Details : lu_kmem_descr and its companion interface allow to create
2693 and destroy a number of kmem caches at once.
2697 Description: Fix lu_object finalization race.
2698 Details : Fix a race between lu_object_find() finding an object and its
2699 concurrent finalization. This race is (most likely) not possible
2700 on the server, but might happen on the client.
2704 Description: Introduce lu_ref interface.
2705 Details : lu_ref is a debugging module allowing to track references to
2706 a given object. It is quite cpu expensive, and has to be
2707 explicitly enabled with --enable-lu_ref. See usage description
2712 Description: Factor lu_site procfs stats into a separate function.
2713 Details : Separate lu_site stats printing code into a separate function
2714 that can be reused on a client.
2718 Description: Constify instances of struct {lu,dt,md}_device_operations.
2719 Details : Constify instances of struct {lu,dt,md}_device_operations.
2723 Description: Introduce struct md_site and move meta-data specific parts of
2724 struct lu_site here.
2725 Details : Move md-specific fields out of struct lu_site into special struct
2726 md_site, so that lu_site can be used on a client.
2730 Description: Kill mdd_lov_destroy().
2731 Details : Remove unused mdd code.
2735 Description: Add st_block checking to multistat.c.
2736 Details : Add st_block checking to multistat.c.
2740 Description: Add lu_ref support to struct obd_device.
2741 Details : Add lu_ref tracking to obd_device.
2745 Description: Kill obd_set_fail_loc().
2746 Details : Remove unused code.
2750 Description: Add special type for ptlrpc_request interpret functions.
2751 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2752 there. Actual usage will be in CLIO.
2756 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2757 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2758 doesn't work with lockdep.
2762 Description: Add rwv.c test program.
2763 Details : New testing program exercising readv(2) and writev(2) (Qian).
2767 Description: Add sendfile.c test program.
2768 Details : New testing program exercising sendfile(2) (Jay).
2772 Description: Ratelimit a message that can be very frequent.
2773 Details : Ratelimit a memory allocation failure message that can
2778 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2779 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2783 Description: Liblustre build fixes.
2784 Details : Liblustre build fixes.
2788 Description: libcfs: add cfs_{need,cond}_resched() interface.
2789 Details : libcfs: add cfs_{need,cond}_resched() definition and
2790 implementations for Linux, NT, and liblustre.
2792 Severity : enhancement
2794 Description: More exported tunables for mballoc
2795 Details : Add support for tunable preallocation window and new tunables for
2796 large/small requests
2800 Description: Detect corruption of block bitmap and checking for preallocations
2801 Details : Checks validity of on-disk block bitmap. Also it does better
2802 checking of number of applied preallocations. When corruption is
2803 found, it turns filesystem readonly to prevent further corruptions.
2807 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2808 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2809 window to 0. If it does hit cache miss, and read-pattern is still
2810 stride-io mode, does not reset the stride window, but also does not
2811 increase the stride window length in this case.
2815 Frequency : only for big-endian servers
2816 Description: Check if system is big-endian while mounting fs with extents feature
2817 Details : Mounting a filesystem with extents feature will fail on big-endian
2818 systems since ext3-based ldiskfs is not supported on big-endian
2819 systems. This can be over-riden with "bigendian_extents" mount option.
2821 Severity : enhancement
2823 Description: The root squash functionality
2824 Details : A security feature, which is to prevent users from being able
2825 to mount lustre on their desktop, run as root, and delete
2826 all of the files in the filesystem. The goal is accomplished by
2827 remapping user id (UID) and group id (GID) of the root user to
2828 a UID and GID specified by the system administartor via Lustre
2829 configuration management server (MGS). The functionality also
2830 allows to specify sets of clients for which the remapping does
2835 Description: Excessive recovery window
2836 Details : With AT enabled, the recovery window can be excessively long (6000+
2837 seconds). To address this problem, we no longer use
2838 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2839 timeout no longer depends on the service time, it is set to
2840 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2841 time via pb_service_time.
2845 Description: Watchdog triggered on MDS failover
2846 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2847 the OSTs know that the MDS "UUID" can be reused for the same export
2848 from a different NID, so we do not need to wait for the export to be
2852 Frequency : rare, only if using MMP with Linux RAID
2854 Description: MMP doesn't work with Linux RAID
2855 Details : While using HA for Lustre servers with Linux RAID, it is possible
2856 that MMP will not detect multiple mounts. To make this work we
2857 need to unplug the device queue in RAID when the MMP block is being
2858 written. Also while reading the MMP block, we should read it from
2859 disk and not the cached one.
2861 Severity : enhancement
2863 Description: open file using fid
2864 Details : A file can be opened using just its fid, like
2865 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2868 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2871 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2872 Details : add proper locking for creation of jbd_4k slab cache
2876 Description: MMP check in ext3_remount() fails without displaying any error
2877 Details : When multiple mount protection fails during remount, proper error
2880 Severity : enhancement
2882 Description: Allow stripe size to be up to 4G-64k
2883 Details : Fix math logic to allow large stripe sizes.
2887 Description: add check for >8TB ldiskfs filesystems
2888 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2889 Don't allow >8TB ldiskfs filesystems to be mounted without
2890 force_over_8tb mount option
2895 Descriptoin: OSS complains "too many blocks" and crashes
2896 Details : In filter_commitrw_write():
2897 1. lquota_chkquota();
2898 2. obdo_from_inode();
2899 3. lquota_pending_commit();
2900 "2." has changed uid and gid in this case, then caused
2901 lquota_chkquota() and lquota_pending_commit() to use
2902 different uid and gid. That is the root of the bug.
2906 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2907 Details : Detect and print out the 'stripe count = -1' case properly.
2911 Description: Out or order replies might be lost on replay
2912 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2913 request with tid smaller than one currently being added, add it
2914 to the start, not end of the list.
2916 --------------------------------------------------------------------------------
2918 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2920 * Support for kernels:
2921 2.6.5-7.283 (SLES 9),
2922 2.6.9-55.EL (RHEL 4),
2923 2.6.16.46-0.14 (SLES 10),
2924 2.6.18.8 vanilla (kernel.org)
2925 * Client support for unpatched kernels:
2926 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2927 2.6.16 - 2.6.22 vanilla (kernel.org)
2928 * Due to recently discovered recovery problems, we do not recommend
2929 using patchless RHEL 4 clients with this or any earlier release.
2930 * Recommended e2fsprogs version: 1.39.cfs8
2931 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2932 * Starting with this release, the ldiskfs backing filesystem required
2933 by Lustre is now in its own package, lustre-ldiskfs. This package
2934 should be installed. It is versioned separately from Lustre and
2935 may be released separately in future.
2939 Description: block reactivating mgc import until all deactivates complete
2940 Details : Fix race when failing back MDT/MGS to itself (testing)
2943 Frequency : at statup only
2945 Description: mds_lov_synchronize race leads to various problems
2946 Details : simultaneous MDT->OST connections at startup can cause the
2947 sync to abort, leaving the OSC in a bad state.
2949 Severity : enhancement
2951 Description: add optional extra BUILD_VERSION info
2952 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2953 to override the lustre version.
2956 Frequency : 2.6.18 servers only
2958 Description: ll_kern_mount() doesn't release the module reference
2959 Details : The ldiskfs module reference count never drops down to 0
2960 because ll_kern_mount() doesn't release the module reference.
2965 Description: server LBUG when using old ost_num_threads parameter
2966 Details : Accept the old ost_num_threads parameter but warn that it
2967 is deprecated, and fix an off-by-one error that caused an LBUG.
2972 Description: Transient SCSI error results in persistent IO issue
2973 Details : iobuf->dr_error is not reinitialized to 0 between two
2977 Frequency : sometimes when underlying device returns I/O errors
2979 Description: OSTs not going read-only during write failures
2980 Details : OSTs are not remounted read-only when the journal commit threads
2981 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2982 instead of the ext3 wrappers.
2986 Description: poor connect scaling with increasing client count
2987 Details : Don't run filter_grant_sanity_check for more than 100 exports
2988 to improve scaling for large numbers of clients.
2991 Frequency : SLES10 only
2993 Description: sanity-quota.sh quotacheck failed: rc = -22
2994 Details : Quotas cannot be enabled on SLES10.
2997 Frequency : liblustre clients only
2999 Description: getdirentries does not give error when run on compute nodes
3000 Details : getdirentries does not fail when the size specified as an argument
3001 is too small to contain at least one entry
3003 Severity : enhancement
3005 Description: Add LNET router traceability for debug purposes
3006 Details : If a checksum failure occurs with a router as part of the
3007 IO path, the NID of the last router that forwarded the bulk data
3008 is printed so it can be identified.
3013 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3014 Details : Due to a race condition, liblustre clients were occasionally
3015 evicted incorrectly.
3017 Severity : enhancement
3019 Description: lfs setstripe use optional parameters instead of postional
3022 Severity : enhancement
3024 Description: Nanosecond timestamp support for ldiskfs
3025 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3026 resolution timestamps. There is not yet support for this at
3027 the Lustre filesystem level.
3030 Frequency : during server recovery
3032 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3033 Details : request with rq_no_resend flag not awake l_wait_event if they get
3037 Frequency : nfs export on patchless client
3039 Description: connectathon hang when test nfs export over patchless client
3040 Details : Disconnected dentry cannot be found with lookup, so we do not need
3041 to unhash it or make it invalid
3044 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3045 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3046 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3051 Description: open req refcounting wrong on reconnect
3052 Details : If reconnect happened between getting open reply from server and
3053 call to mdc_set_replay_data in ll_file_open, we will schedule
3054 replay for unreferenced request that we are about to free.
3055 Subsequent close will crash in variety of ways.
3056 Check that request is still eligible for replay in
3057 mdc_set_replay_data().
3062 Description: disable writes to filesystem when reading health_check file
3063 Details : the default for reading the health_check proc file has changed
3064 to NOT do a journal transaction and write to disk, because this
3065 can cause reads of the /proc file to hang and block HA state
3066 checking on a healthy but otherwise heavily loaded system. It
3067 is possible to return to the previous behaviour during configure
3068 with --enable-health-write.
3070 Severity : enhancement
3072 Description: 64-bit inode version
3073 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3074 to the inode. This will be required for version-based recovery.
3079 Description: MDS fails to start if a duplicate client export is detected
3080 Details : in some rare cases it was possible for a client to connect to
3081 an MDS multiple times. Upon recovery the MDS would detect this
3082 and fail during startup. Handle this more gracefully.
3084 Severity : enhancement
3086 Description: Add -o localflock option to simulate old noflock
3088 Details : This will achieve local-only flock/fcntl locks
3094 Description: log_commit_thread vs filter_destroy race leads to crash
3095 Details : Take import reference before releasing llog record semaphore
3100 Description: Wrong request locking in request set processing
3101 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3102 imp_delayed_list, in this place should be used imp_lock.
3105 Frequency : when reconnection
3107 Description: Grant Leak when osc reconnect to OST
3108 Details : When osc reconnect ost, OST(filter) should check whether it
3109 should grant more space to client by comparing fed_grant and
3110 cl_avail_grant, and return the granted space to client instead
3111 of "new granted" space, because client will call osc_init_grant
3112 to update the client grant space info.
3115 Frequency : when client reconnect to OST
3117 Description: Grant Leak when osc do resend and replay bulk write
3118 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3119 bulk write request, because the grant info while be sync between
3120 OSC and OST when reconnect, and we should ignore the grant info
3121 these of resend/replay write req.
3126 Description: Grant space more than available left space sometimes.
3127 Details : When then OST is about to be full, if two bulk writing from
3128 different clients came to OST. Accord the avaliable space of the
3129 OST, the first req should be permitted, and the second one
3130 should be denied by ENOSPC. But if the seconde arrived before
3131 the first one is committed. The OST might wrongly permit second
3132 writing, which will cause grant space > available space.
3135 Frequency : when client is evicted
3137 Description: Grant might be wrongly erased when osc is evicted by OST
3138 Details : when the import is evicted by server, it will fork another
3139 thread ptlrpc_invalidate_import_thread to invalidate the
3140 import, where the grant will be set to 0. While the original
3141 thread will update the grant it got when connecting. So if
3142 the former happened latter, the grant will be wrongly errased
3143 because of this race.
3148 Description: Checking Stale with correct fid
3149 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3150 to check whether it is stale, because sometimes, we want the
3151 enqueue happened anyway, and op_data.fid2 will not be initialized.
3153 Severity : enhancement
3155 Description: update patchless client
3156 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3159 Frequency : only with 2.4 kernel
3161 Description: random memory corruption
3162 Details : size of struct ll_inode_info is to big for union inode.u and this
3163 can be cause of random memory corruption.
3168 Description: Memory leak in recovery
3169 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3170 It should also check obd_fail before fsfilt_start, otherwise if
3171 fsfilt_start return -EROFS,(failover mds during mds recovery).
3172 then the req will return with repmsg->transno = 0 and rc = EROFS.
3173 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3174 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3175 be freed no matter whether fsfilt_commit success or not.
3178 Frequency : only with huge count clients
3180 Description: Prevents from taking the superblock lock in llap_from_page for
3182 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3183 for llap_from_page call in ll_removepage prevents from taking the
3184 superblock lock for a soon died page.
3189 Description: Not check open intent error before release open handle
3190 Details : in some rare cases, the open intent error is not checked before
3191 release open handle, which may cause
3192 ASSERTION(open_req->rq_transno != 0), because it tries to release
3193 the failed open handle.
3198 Description: Set cat log bitmap only after create log success.
3199 Details : in some rare cases, the cat log bitmap is set too early. and it
3200 should be set only after create log success.
3204 Description: Accessing a block bevice can re-enable I/O when Lustre is
3205 tearing down a device.
3206 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3211 Description: service threads may hog cpus when there are a lot of requests
3212 Details : Insert cond_resched to give other threads a chance to use some CPU
3217 Description: the cat log was not initialized in recovery
3218 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3219 unlink log on mds will not be initialized until mds post
3220 recovery. And also in mds post recovery, the unlink log will
3221 initialization will be done asynchronausly, so there will be race
3222 between add unlink log and unlink log initialization.
3226 Description: brw_stats were being printed incorrectly
3227 Details : brw_stats were being printed as log2 but all of them were not
3228 recorded as log2. Also remove some code duplication arising from
3229 filter_tally_{read,write}.
3233 Frequency : rare, only in recovery.
3234 Description: ASSERTION(req->rq_type != LI_POISON) failed
3235 Details : imp_lock should be held while iterating over imp_sending_list for
3236 prevent destroy request after get timeout in ptlrpc_queue_wait.
3240 Description: replay-single.sh test 52 fails
3241 Details : A lock's skiplist need to be cleanup when it being unlinked
3242 from its resource list.
3246 Description: Short directio read returns full requested size rather than
3248 Details : Direct I/O operations should return actual amount of bytes
3249 transferred rather than requested size.
3251 Severity : enhancement
3253 Description: metadata RPC reduction (e.g. for rm performance)
3254 Details : decrease the amount of synchronous RPC between clients and servers
3255 by canceling conflicing lock before the operation on the client
3256 and packing thier handles into the main operation RPC to server.
3258 Severity : enhancement
3260 Description: Async OSC create to avoid the blocking unnecessarily.
3261 Details : If an OST has no remaining object, system will block on the
3262 creation when it needs to create a new object on this OST. Now,
3263 ways use pre-created objects when available, instead of blocking on
3264 an empty osc while others are not empty. If we must block, we
3265 block for the shortest possible period of time.
3269 Description: improve handling recoverable errors
3270 Details : if request processig with error which can be recoverable on server
3271 request should be resend, otherwise page released from cache and
3274 Severity : enhancement
3276 Description: refine locking for avoid write wrong info into lov_objid file
3277 Details : fix possible races with add new target and write/update data in
3280 --------------------------------------------------------------------------------
3282 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3287 Frequency : on some architectures
3289 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3290 Details : Uninitialized flags sometimes cause configuration commands to
3294 Frequency : patchless clients only
3296 Description: missing __iget() symbol export
3297 Details : The __iget() symbol export is missing. To avoid the need for
3298 this on patchless clients the deathrow inode reaper is turned
3299 off, and we depend on the VM to clean up old inodes. This
3300 dependency was during via the fix for bug 12181.
3302 --------------------------------------------------------------------------------
3304 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3306 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3307 INTEROPERATE with older versions automatically. In many cases a
3308 special upgrade step is needed. Please read the
3309 user documentation before upgrading any part of a live system.
3310 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3311 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3312 * WARNING: Lustre configuration and startup changes are required with
3313 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3317 Severity : enhancement
3319 Description: MountConf
3320 Details : Lustre configuration is now managed via mkfs and mount
3321 commands instead of lmc and lconf. New obd types (MGS, MGC)
3322 are added for dynamic configuration management. See
3323 https://mail.clusterfs.com/wikis/lustre/MountConf for
3326 Severity : enhancement
3328 Description: dynamic OST addition
3329 Details : OSTs can now be added to a live filesystem
3331 Severity : enhancement
3333 Description: startup order invariance
3334 Details : MDTs and OSTs can be started in any order. Clients only
3335 require the MDT to complete startup.
3337 Severity : enhancement
3339 Description: parallel, asynchronous orphan cleanup
3340 Details : orphan cleanup is now performed in separate threads for each
3341 OST, allowing parallel non-blocking operation.
3343 Severity : enhancement
3345 Description: optimized stripe assignment
3346 Details : stripe assignments are now made based on ost space available,
3347 ost previous usage, and OSS previous usage, in order to try
3348 to optimize storage space and networking resources.
3350 Severity : enhancement
3352 Description: Permanently set tunables
3353 Details : All writable /proc/fs/lustre tunables can now be permanently
3354 set on a per-server basis, at mkfs time or on a live system.
3356 Severity : enhancement
3358 Description: Lustre message v2
3359 Details : Add lustre message format v2.
3361 Severity : enhancement
3363 Description: client OST exclusion list
3364 Details : Clients can be started with a list of OSTs that should be
3365 declared "inactive" for known non-responsive OSTs.
3369 Description: SPEC SFS validation failure on NFS v2 over lustre.
3370 Details : Changes the blocksize for regular files to be 2x RPC size,
3371 and not depend on stripe size.
3373 Severity : enhancement
3375 Description: Multiple MD RPCs in flight.
3376 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3377 To avoid overly-overloading MDS, introduce a limit on number of
3378 MDS RPCs in flight for a single client and add /proc controls
3379 to adjust this limit.
3381 Severity : enhancement
3383 Description: client read/write statistics
3384 Details : Add client read/write call usage stats for performance
3385 analysis of user processes.
3386 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3387 file access. extents_stats shows chunk size distribution.
3388 extents_stats_per_process show chunk size distribution per
3391 Severity : enhancement
3393 Description: mds statistics
3394 Details : Add detailed mds operations statistics in
3395 /proc/fs/lustre/mds/*/stats.
3399 Description: Failure of copying files with lustre special EAs.
3400 Details : Client side always return success for setxattr call for lustre
3401 special xattr (currently only "trusted.lov").
3406 Description: Refcount LNET uuids
3407 Details : The global LNET uuid list grew linearly with every startup;
3408 refcount repeated list entries instead of always adding to
3411 Severity : enhancement
3413 Description: Dynamic service threads
3414 Details : Within a small range, start extra service threads
3415 automatically when the request queue builds up.
3418 Frequency : mixed-endian client/server environments
3420 Description: mixed-endian crashes
3421 Details : The new msg_v2 system had some failures in mixed-endian
3424 Severity : enhancement
3426 Description: Easy OST removal
3427 Details : OSTs can be permanently deactivated with e.g. 'lctl
3428 conf_param lustre-OST0001.osc.active=0'
3430 Severity : enhancement
3432 Description: MGS proc entries
3433 Details : Added basic proc entries for the MGS showing what filesystems
3436 Severity : enhancement
3438 Description: provide MGS failover
3439 Details : Added config lock reacquisition after MGS server failover.
3441 Severity : enhancement
3443 Description: add Linux 2.4 support
3444 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3448 Description: a large application tries to do I/O to the same resource and dies
3449 in the middle of it.
3450 Details : Check the req->rq_arrival time after the call to
3451 ost_brw_lock_get(), but before we do anything about
3452 processing it & sending the BULK transfer request. This
3453 should help move old stale pending locks off the queue as
3454 quickly as obd_timeout.
3457 Frequency : when an incorrect nid is specified during startup
3459 Description: ptlrpc connect to non-existant node causes kernel crash
3460 Details : LNET can't be re-entered from an event callback, which
3461 happened when we expire a message after the export has been
3462 cleaned up. Instead, hand the zombie cleanup off to another
3465 Severity : enhancement
3467 Description: plain/inodebits lock performance improvement
3468 Details : Grouping plain/inodebits in granted list by their request modes
3469 and bits policy, thus improving the performance of search through
3473 Frequency : only if OST filesystem is corrupted
3475 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3476 Details : for a short time RPCs with bulk IO are in the replay list,
3477 but replay of bulk IOs is unimplemented. If the OST filesystem
3478 is corrupted due to disk cache incoherency and then replay is
3479 started it is possible to trip an assertion. Avoid putting
3480 committed RPCs into the replay list at all to avoid this issue.
3483 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3485 Description: System hang on startup
3486 Details : This bug allowed the liblustre (e.g. catamount) client to
3487 return to the app before handling all startup RPCs. This
3488 could leave the node unresponsive to lustre network traffic
3489 and manifested as a server ptllnd timeout.
3491 Severity : enhancement
3493 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3494 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3495 boolean to control whether to print peer debug info when a
3496 client's RPC times out.
3499 Frequency : only for kernels with patches from Lustre below 1.4.3
3501 Description: Remove old rdonly API
3502 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3505 Frequency : only for devices with external journals
3507 Description: Set external device read-only also
3508 Details : During a commanded failover stop, we set the disk device
3509 read-only while the server shuts down. We now also set any
3510 external journal device read-only at the same time.
3513 Frequency : when upgrading from 1.4 while trying to change parameters
3515 Description: The wrong (new) MDC name was used when setting parameters for
3516 upgraded MDT's. Also allows changing of OSC (and MDC)
3517 parameters if --writeconf is specified at tunefs upgrade time.
3520 Frequency : when setting specific ost indicies
3522 Description: QOS code breaks on skipped indicies
3523 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3524 created with --index need not be sequential.
3526 Severity : enhancement
3528 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3529 Details : The uninit_groups feature works in conjunction with the kernel
3530 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3531 up the pass1 processing of e2fsck. This is a read-only feature
3532 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3533 mount filesystems that have had this feature enabled.
3535 Severity : enhancement
3537 Description: Improve multi-block allocation algorithm to avoid fragmentation
3538 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3539 allocation locality and avoid filesystem fragmentation.
3541 ------------------------------------------------------------------------------
3543 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3545 * Support for kernels:
3546 2.4.21-47.0.1.EL (RHEL 3)
3547 2.6.5-7.283 (SLES 9)
3548 2.6.9-42.0.10.EL (RHEL 4)
3549 2.6.12.6 vanilla (kernel.org)
3550 2.6.16.27-0.9 (SLES 10)
3551 * Recommended e2fsprogs version: 1.39.cfs5
3553 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3557 Frequency : occasional, depends on client load and configuration
3558 Bugzilla : 12181, 12203
3559 Description: data loss for recently-modified files
3561 Details : In some cases it is possible that recently written or created
3562 files may not be written to disk in a timely manner (this should
3563 normally be within 30s unless client IO load is very high).
3564 The problem appears as zero-length files or files that are a
3565 multiple of 1MB in size after a client crash or client eviction
3566 that are missing data at the end of the file.
3568 This problem is more likely to be hit on clients where files are
3569 repeatedly created and unlinked in the same directory, clients
3570 have a large amount of RAM, have many CPUs, the filesystem has
3571 many OSTs, the clients are rebooted frequently, and/or the files
3572 are not accessed by other nodes after being written.
3574 The presence of the problem can be detected by looking at
3575 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3576 smaller than the second (nr_unused) then dirty files will not
3577 be flushed automatically to disk. "sync; sleep 10" should be
3578 run several times on the node before unmounting it to update
3579 Lustre (this is also safe to run on nodes without this problem).
3581 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3582 that can cause this same problem, so customers using that kernel
3583 also need to update the kernel in addition to Lustre. In order
3584 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3586 It is normal that files written just before a client crash (less
3587 than 30s) may not yet have been flushed to disk, even for local
3591 Frequency : frequent on thin XT3 nodes
3593 Description: UUID collision on thin XT3 Linux nodes
3594 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3595 randomly, since we relied on an insufficiently-seeded PRNG.
3600 Description: OSS hangs after "All ost request buffers busy"
3601 Details : A deadlock between quota and journal operations caused OSS
3602 hangs after printing "All ost request buffers busy."
3605 Frequency : always on liblustre builds
3607 Description: Cleanup compiler warnings on liblustre
3610 Frequency : always on liblustre builds on XT3
3612 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3616 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3617 > 32000 subdirectories
3618 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3619 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3620 subdirectory count crosses 32000. This will aid e2fsck to
3621 correctly handle more than 32000 subdirectories.
3624 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3626 Description: System hang on startup
3627 Details : This bug allowed the liblustre (e.g. catamount) client to
3628 return to the app before handling all startup RPCs. This
3629 could leave the node unresponsive to lustre network traffic
3630 and manifested as a server ptllnd timeout.
3632 Severity : enhancement
3634 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3635 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3636 boolean to control whether to print peer debug info when a
3637 client's RPC times out.
3642 Description: make O_SYNC working on 2.6 kernels
3643 Details : 2.6 kernels use different method for mark pages for write,
3644 so need add a code to lustre for O_SYNC work.
3649 Description: Failure to close file and release space on NFS
3650 Details : Put inode details into lock acquired in ll_intent_file_open.
3651 Use mdc_intent_lock in ll_intent_open to properly
3652 detect all kind of errors unhandled by mdc_enqueue.
3657 Description: proc file read during shutdown sometimes raced obd removal,
3659 Details : Add lock to prevent obd access after proc file removal.
3662 Frequency : Only for files larger than 4GB on 32-bit clients.
3664 Description: improperly doing page alignment of locks
3665 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3666 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3671 Description: RPCs being resent when they shouldn't be
3672 Details : Some RPCs that should not be resent are being resent. This
3673 can cause inconsistencies in the RPC state machine. Do not
3674 resend such requests.
3677 Frequency : rare, only with NFS export
3679 Description: Crash on NFS re-export node
3680 Details : under very unusual load conditions an assertion is hit in
3681 ll_intent_file_open()
3684 Frequency : only if OST filesystem is corrupted
3686 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3687 Details : for a short time RPCs with bulk IO are in the replay list,
3688 but replay of bulk IOs is unimplemented. If the OST filesystem
3689 is corrupted due to disk cache incoherency and then replay is
3690 started it is possible to trip an assertion. Avoid putting
3691 committed RPCs into the replay list at all to avoid this issue.
3696 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3697 Details : Large single O_DIRECT read and write calls can fail to allocate
3698 a sufficiently large buffer to process the request. In case of
3699 allocation failure the allocation is retried with a smaller
3700 buffer and broken into smaller requests.
3702 Severity : enhancement
3704 Description: Add -o localflock option to simulate old noflock behaviour.
3705 Details : This will achieve local-only flock/fcntl locks coherentness.
3710 Description: versioning check is incomplete
3711 Details : Checking the version difference of client vs. server, report
3712 error if the gap is too big.
3717 Description: add support PG_writeback bit
3718 Details : add support for PG_writeback bit for Lustre, for more carefull
3719 work with page cache in 2.6 kernel. This also fix some deadlocks
3720 and remove hack for work O_SYNC with 2.6 kernel.
3722 Severity : enhancement
3724 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3725 Details : The uninit_groups feature works in conjunction with the kernel
3726 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3727 up the pass1 processing of e2fsck. This is a read-only feature
3728 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3729 mount filesystems that have had this feature enabled.
3731 Severity : enhancement
3733 Description: Improve multi-block allocation algorithm to avoid fragmentation
3734 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3735 allocation locality and avoid filesystem fragmentation.
3737 ------------------------------------------------------------------------------
3739 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3741 * Support for kernels:
3742 2.6.9-42.0.3.EL (RHEL 4)
3743 2.6.5-7.276 (SLES 9)
3744 2.4.21-47.0.1.EL (RHEL 3)
3745 2.6.12.6 vanilla (kernel.org)
3746 2.6.16.21-0.8 (SLES10)
3747 * Recommended e2fsprogs version: 1.39.cfs2-0
3749 * The backwards-compatible /proc/sys/portals symlink has been removed
3750 in this release. Before upgrading, please ensure that you change
3751 any configuration scripts or /etc/sysctl.conf files that access
3752 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3753 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3754 in advance of the upgrade on any system running Lustre 1.4.6 or
3755 newer, since /proc/sys/lnet was added in that version.
3756 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3760 Frequency : only when quota is used
3762 Description: avoid scanning export list for quota master
3763 Details : Change the algorithms to avoid scanning export list in order
3764 to improve the efficiency.
3767 Frequency : MDS failover only, very rarely
3769 Description: "went back in time" messages on mds failover
3770 Details : The greatest transno may be lost when the current operation
3771 finishes with an error (transno==0) and the client's last_rcvd
3772 record is over-written. Save the greatest transno in the
3773 mds_last_transno for this case.
3776 Frequency : always for specific kernels and striping counts
3778 Description: client may get "Matching packet too big" without ACL support
3779 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3780 when trying to access files in certain configurations. The
3781 clients should in fact be denied when mounting because they do
3782 not understand ACLs.
3785 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3787 Description: many clients connecting with IO in progress causes connect timeouts
3788 Details : Avoid synchronous journal commits to avoid delays caused by many
3789 clients connecting/disconnecting when bulk IO is in progress.
3790 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3791 OST_IO_PORTAL to avoid delays behind potentially many pending
3795 Frequency : occasionally with multiple writers to a single file
3797 Description: shared writes to file may result in wrong size reported by stat()
3798 Details : Allow growing of kms when extent lock is cancelled
3801 Frequency : always with random mmap IO to multi-striped file
3803 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3804 Details : Only if the hole is at the end of OST object so that kms is too
3805 small. Fix is to increase kms accordingly in ll_nopage.
3808 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3810 Description: writes to a missing object would leak memory on the OST
3811 Details : If there is an inconsistency between the MDS and OST filesystems,
3812 such that the MDS references an object that doesn't exist, writes
3813 to that object will leak memory due to incorrect cleanup in the
3814 error handling path, eventually running out of memory on the OST.
3819 Description: Creating too long symlink causes lustre errors
3820 Details : Check symlink and name lengths before sending requests to MDS.
3823 Frequency : only if flock is enabled (not on by default)
3825 Description: posix locks not released on fd closure on 2.6.9+
3826 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3827 kernels, this caused such locks not to be released on fd close and
3828 then assertions on fs unmount about still used locks.
3831 Frequency : MDS failover only, very rarely
3833 Description: clients may get ASSERTION(granted_lock != NULL)
3834 Details : When request was taking a long time, and a client was resending
3835 a getattr by name lock request. The were multiple lock requests
3836 with the same client lock handle and
3837 mds_getattr_name->fixup_handle_for_resent_request found one of the
3838 lock handles but later failed with ASSERTION(granted_lock != NULL).
3843 Description: handle->h_buffer_credits > 0, assertion failure
3844 Details : h_buffer_credits is zero after truncate, causing assertion
3845 failure. This patch extends the transaction or creates a new
3849 Frequency : NFS re-export or patchless client
3850 Bugzilla : 11179, 10796
3851 Description: Crash on NFS re-export node (__d_move)
3852 Details : We do not want to hash the dentry if we don't have a lock.
3853 But if this dentry is later used in d_move, we'd hit uninitialised
3854 list head d_hash, so we just do this to init d_hash field but
3855 leave dentry unhashed.
3858 Frequency : NFS re-export or patchless client
3860 Description: NFS exports has problem with symbolic link
3861 Details : lustre client didn't properly install dentry when re-exported
3862 to NFS or running patchless client.
3865 Frequency : NFS re-export or patchless client
3867 Description: Various nfs/patchless fixes.
3868 Details : fixes reuse disconected alias for lookup process - this fixes
3869 warning "find_exported_dentry: npd != pd",
3870 fix permission error with open files at nfs.
3871 fix apply umask when do revalidate.
3874 Frequency : occasional
3876 Description: Crash on NFS re-export node
3877 Details : calling clear_page() on the wrong pointer triggered oops in
3878 generic_mapping_read().
3881 Frequency : rarely, using O_DIRECT IO
3883 Description: unaligned directio crashes client with LASSERT
3884 Details : check for unaligned buffers before trying any requests.
3887 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3889 Description: stale data returned from RAID cache
3890 Details : If only a small amount of IO is done to the RAID device before
3891 reading it again it is possible to get stale data from the RAID
3892 cache instead of reading it from disk.
3895 Frequency : always for sles10 kernel
3897 Description: sles10 support
3898 Details : ll_follow_link: compile fixes and using of nd_set_link
3899 under newer kernels.
3902 Frequency : depends on arch, kernel and compiler version, always on sles10
3905 Description: recursive or deep enough symlinks cause stack overflow
3906 Details : getting rid of large stack-allocated variable in
3910 Frequency : depends on hardware
3912 Description: lustre write performance loss in the SLES10 kernel
3913 Details : the performance loss is caused by using of write barriers in the
3914 ext3 code. The SLES10 kernel turns barrier support on by
3915 default. The fix is to undo that change for ldiskfs.
3917 ------------------------------------------------------------------------------
3919 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3921 * Support for kernels:
3922 2.6.9-42.0.3EL (RHEL 4)
3923 2.6.5-7.276 (SLES 9)
3924 2.4.21-47.0.1.EL (RHEL 3)
3925 2.6.12.6 vanilla (kernel.org)
3929 Frequency : quota enabled and large files being deleted
3931 Description: releasing more than 4GB of quota at once hangs OST
3932 Details : If a user deletes more than 4GB of files on a single OST it
3933 will cause the OST to spin in an infinite loop. Release
3934 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3939 Description: statfs data retrieved from /proc may be stale or zero
3940 Details : When reading per-device statfs data from /proc, in the
3941 {kbytes,files}_{total,free,avail} files, it may appear
3942 as zero or be out of date.
3945 Frequency : systems with MD RAID1 external journal devices
3947 Description: lconf's call to blkid is confused by RAID1 journal devices
3948 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3949 of returning all block devices that match the journal UUID.
3952 Frequency : always, for aggregate stripe size over 4GB
3954 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3955 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3956 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3959 Frequency : always if "lfs find" used on a local file/directory
3961 Description: "lfs find" segfaults if used on a local file/directory
3962 Details : The case where a directory component was not specified wasn't
3963 handled correctly. Handle this properly.
3966 Frequency : always on ppc64
3968 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3969 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3970 perform endianness conversion.
3973 Frequency : rarely (truncate to non-zero file size after write under load)
3974 Bugzilla : 10730, 10687
3975 Description: Files padded with zeros to next 4K multiple
3976 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3977 it is possible that files that are truncated to a non-zero size
3978 immediately after being written are filled with zero bytes beyond
3979 the truncated size. No file data is lost.
3981 Severity : enhancement
3983 Description: Allow recovery/failover for liblustre clients.
3984 Details : liblustre clients were unaware of failover configurations until
3987 Severity : enhancement
3989 Description: user file locks should fail when not mounting with flock option
3990 Details : Set up an error-returning stub in ll_file_operations.lock field
3991 to prevent incorrect behaviour when client is mounted without
3992 flock option. Also, set up properly f_op->flock field for
3996 Frequency : always on ia64
3998 Description: "lfs df" loops on printing out MDS statfs information
3999 Details : The obd_ioctl_data was not initialized and in some systems
4000 this caused a failure during the ioctl that did not return
4001 an error. Initialize the struct and return an error on failure.
4004 Frequency : SLES 9 only
4006 Description: Error of copying files with lustre special EAs as root
4007 Details : Client side always return success for setxattr call for lustre
4008 special xattr (currently only "trusted.lov").
4011 Frequency : rarely on clusters with both ia64+i386 clients
4013 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4014 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4015 clients are doing concurrent writes to the same file it is
4016 possible that the ia64 clients may LASSERT because the OST
4017 extent locks are not PAGE_SIZE aligned. Ensure that grown
4018 locks are always aligned on the request boundary.
4021 Frequency : specific use, occasional
4023 Description: Overwriting in use executable truncates on-disk binary image
4024 Details : If one node attempts to overwrite an executable in use by
4025 another node, we now correctly return ETXTBSY instead of
4026 truncating the file.
4028 Severity : enhancement
4030 Description: Async OSC create to avoid the blocking unnecessarily.
4031 Details : If an OST has no remaining object, system will block on the creating
4032 when need to create a new object on this OST. Now, ways use
4033 pre-created objects when available, instead of blocking on an
4034 empty osc while others are not empty. If we must block, we block
4035 for the shortest possible period of time.
4040 Description: chmod on Lustre root is propagated to other clients
4041 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4042 invalid by the follow_mount time.
4045 Frequency : liblustre clients only
4047 Description: Race in 'instant cancel' lock handling could lead to such locks
4048 never to be granted in case of SMP MDS
4049 Details : Do not destroy not yet granted but cbpending locks in
4053 Frequency : replay/resend of open
4055 Description: non null lock assetion failure in mds_intent_policy
4056 Details : Trying to replay/resend lockless open requests resulted in
4057 mds_open() returning 0 with no lock. Now it sets a flag if
4058 a lock is going to be returned.
4060 Severity : enhancement
4062 Description: Checksum enhancements
4063 Details : New checksum enhancements allow for resending RPCs that failed
4066 Severity : enhancement
4068 Description: Tunables on number of dirty pages in cacche
4069 Details : Allow to set limit on number of dirty pages cached.
4074 Description: client crash on unmount - lock still has references
4075 Details : In some error handling cases it was possible to leak a lock
4076 reference on a client while accessing a file. This was not
4077 harmful to the client during operation, but would cause the
4078 client to crash when the filesystem is unmounted.
4081 Frequency : specific case, rare
4083 Description: ETXTBSY on mds though file not in use by client
4084 Details : ETXTBSY is no longer incorrectly returned when attempting to
4085 chmod or chown a directory that the user previously tried to
4086 execute or a currently-executing binary.
4089 Frequency : extremely rare except on liblustre-based clients
4091 Description: Lustre space not freed when files are deleted
4092 Details : Clean up open-unlinked files after client eviction. Previously
4093 the unlink was skipped and the files remained as orphans.
4098 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4099 Details : In some cases it was possible to send a blocking callback to
4100 a client doing a glimpse, even though that client didn't get
4101 a lock granted. When the glimpse lock is cancelled on the OST
4102 the freed lock is left on the waiting list and corrupted the list.
4105 Frequency : all core dumps
4107 Description: Broke core dumps to lustre
4108 Details : Negative dentry may be unhashed if parent does not have UPDATE
4109 lock, but some callers, e.g. do_coredump, expect dentry to be
4110 hashed after successful create, hash it in ll_create_it.
4112 ------------------------------------------------------------------------------
4114 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4116 * Support for kernels:
4117 2.6.9-42.0.2.EL (RHEL 4)
4118 2.6.5-7.276 (SLES 9)
4119 2.4.21-40.EL (RHEL 3)
4120 2.6.12.6 vanilla (kernel.org)
4124 Frequency : always on RHEL 3
4126 Description: Number of open files grows over time
4127 Details : The number of open files grows over time, whether or not
4128 Lustre is started. This was due to a filp leak introduced
4129 by one of our kernel patches.
4131 ------------------------------------------------------------------------------
4133 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4135 * Support for kernels:
4136 2.6.9-42.EL (RHEL 4)
4137 2.6.5-7.276 (SLES 9)
4138 2.4.21-40.EL (RHEL 3)
4139 2.6.12.6 vanilla (kernel.org)
4144 Bugzilla : 5719, 9635, 9792, 9684
4145 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4146 Details : If a server is under heavy load and cannot reply to new
4147 connection requests before the client resends the (re)connect,
4148 the connection handling code can behave badly if two service
4149 threads are concurrently handing separate (re)connections from
4150 the same client. Add better locking to the connection handling
4151 code, and ensure that only a single connection will be processed
4152 for a given client UUID, even if the lock is dropped.
4154 Severity : enhancement
4156 Description: add TCP zero-copy support to kernel
4157 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4158 sends if the hardware supports scatter-gather and checksumming.
4159 This allows socklnd to do client-write and server-read more
4160 efficiently and reduce CPU utilization from skbuf copying.
4163 Frequency : only if NFS exporting from client
4165 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4166 Details : If setting attributes on a file created under NFS that had
4167 never been opened it would be possible to oops the client
4168 if the file had no objects.
4171 Frequency : always for liblustre
4173 Description: liblustre client does MDS+OSTs setattr RPC for each write
4174 Details : When doing a write from a liblustre client, the client
4175 incorrectly issued an RPC to the MDS and each OST the file was
4176 striped over in order to update the timestamps. When writing
4177 with small chunks and many clients this could overwhelm the MDS
4178 with RPCs. In all cases it would slow down the write because
4179 these RPCs are unnecessary.
4181 Severity : enhancement
4183 Description: allow number of MDS service threads to be changed at module load
4184 Details : It is now possible to change the number of MDS service threads
4185 running. Adding "options mds mds_num_threads={N}" to the MDS's
4186 /etc/modprobe.conf will set the number of threads for the next
4187 time Lustre is restarted (assuming the "mds" module is also
4188 reloaded at that time). The default number of threads will
4189 stay the same, 32 for most systems.
4194 Description: OST crash if filesystem is unformatted or corrupt
4195 Details : If an OST is started on a device that has never been formatted
4196 or if the filesystem is corrupt and cannot even mount then the
4197 error handling cleanup routines would dereference a NULL pointer.
4202 Description: NULL pointer deref in llap_from_page.
4203 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4204 incorrect) mapping if the page was truncated/reclaimed while it was
4205 searched for. Check for this condition and skip such pages when
4206 doing readahead. Introduce extra check to llap_from_page() to
4207 verify page->mapping->host is non-NULL (so page is not anonymous).
4210 Frequency : Sometimes when using sys_sendfile
4212 Description: "page not covered by a lock" warnings from ll_readpage
4213 Details : sendfile called ll_readpage without right page locks present.
4214 Now we introduced ll_file_sendfile that does necessary locking
4215 around call to generic_file_sendfile() much like we do in
4219 Frequency : with certain MDS communication failures at client mount time
4221 Description: NULL pointer deref after failed client mount
4222 Details : a client connection request may delayed by the network layer
4223 and not be sent until after the PTLRPC layer has timed out the
4224 request. If the client fails the mount immediately it will try
4225 to clean up before the network times out the request. Add a
4226 reference from the request import to the obd device and delay
4227 the cleanup until the network drops the request.
4230 Frequency : occasionally during client (re)connect
4232 Description: assertion failure during client (re)connect
4233 Details : processing a client connection request may be delayed by the
4234 client or server longer than the client connect timeout. This
4235 causes the client to resend the connection request. If the
4236 original connection request is replied in this interval, the
4237 client may trip an assertion failure in ptlrpc_connect_interpret()
4238 which thought it would be the only running connect process.
4241 Frequency : only with obd_echo servers and clients that are rebooted
4243 Description: kernel BUG accessing uninitialized data structure
4244 Details : When running an obd_echo server it did not start the ping_evictor
4245 thread, and when a client was evicted an uninitialized data
4246 structure was accessed. Start the ping_evictor in the RPC
4247 service startup instead of the OBD startup.
4249 Severity : enhancement
4250 Bugzilla : 10193 (patchless)
4251 Description: Remove dependency on various unexported kernel interfaces.
4252 Details : No longer need reparent_to_init, exit_mm, exit_files,
4253 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4256 Frequency : rare (only users of deprecated and unsupported LDAP config)
4258 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4259 Details : LDAP apparently contains 'lustreName' attributes instead of
4260 'name'. A simple remapping of the name is sufficient.
4263 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4265 Description: waiting_locks_callback trips kernel BUG if client is evicted
4266 Details : Running with the dump_on_timeout debug flag turned on makes
4267 it possible that the waiting_locks_callback() can try to dump
4268 the Lustre kernel debug logs from an interrupt handler. Defer
4269 this log dumping to the expired_lock_main() thread.
4271 Severity : enhancement
4273 Description: Support NFS exporting on 2.6 kernels.
4274 Details : Implement non-rawops metadata methods for NFS server to use without
4275 changing NFS server code.
4278 Frequency : very rare (synthetic metadata workload only)
4280 Description: two racing renames might cause an MDS thread to deadlock
4281 Details : Running the "racer" program may cause one MDS thread to rename
4282 a file from being the source of a rename to being the target of
4283 a rename at exactly the same time that another thread is doing
4284 so, and the second thread has already enqueued these locks after
4285 doing a lookup of the target and is trying to relock them in
4286 order. Ensure that we don't try to re-lock the same resource.
4289 Frequency : only very large systems with liblustre clients
4291 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4292 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4293 all OSTs in parallel. This allows the eviction of stale liblustre
4294 clients to proceed much faster than if they were done in series,
4295 and also offers similar improvements for other set_info RPCs.
4300 Description: excessive CPU usage during initial read phase on client
4301 Details : During the initial read phase on a client, it would agressively
4302 retry readahead on the file, consuming too much CPU and impacting
4303 performance (since 1.4.5.8). Improve the readahead algorithm
4304 to avoid this, and also improve some other common cases (read
4305 of small files in particular, where "small" is files smaller than
4306 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4311 Description: MDS crash when receiving packet with unknown intent.
4312 Details : Do not LBUG in unknown intent case, just return -EFAULT
4314 Severity : enhancement
4315 Bugzilla : 9293, 9385
4316 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4317 Details : Do not serialize getattr (non-intent version) and statfs.
4320 Frequency : occasional, when OST network is overloaded/intermittent
4322 Description: client evicted by OST after bulk IO timeout
4323 Details : If a client sends a bulk IO request (read or write) the OST
4324 may evict the client if it is unresposive to its data GET/PUT
4325 request. This is incorrect if the network is overloaded (takes
4326 too long to transfer the RPC data) or dropped the OST GET/PUT
4327 request. There is no need to evict the client at all, since
4328 the pinger and/or lock callbacks will handle this, and the
4329 client can restart the bulk request.
4332 Frequency : Always when mmapping file with no objects
4334 Description: client crashes when mmapping file with no objects
4335 Details : Check that we actually have objects in a file before doing any
4336 operations on objects in ll_vm_open, ll_vm_close and
4342 Description: Request leak when working with deleted CWD
4343 Details : Introduce advanced request refcount tracking for requests
4344 referenced from lustre intent.
4346 Severity : Enhancement
4348 Description: Cache open file handles on client.
4349 Details : MDS now will return special lock along with openhandle, if
4350 requested and client is allowed to hold openhandle, even if unused,
4351 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4352 closing files for every read/write openration.
4354 Severity : Enhancement
4356 Description: Cache open negative dentries on client when possible.
4357 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4358 negative dentries on lock revocation.
4363 Description: Remounting a client read-only wasn't possible with a zconf mount
4364 Details : It wasn't possible to remount a client read-only with llmount.
4366 Severity : enhancement
4367 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4368 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4369 MPICH is included to add Lustre-specific ADIO interfaces.
4370 This is based closely on the UFS ADIO layer and only differs
4371 in file creation, in order to allow the OST striping to be set.
4372 This is user-contributed code and not supported by CFS.
4377 Description: extended inode attributes (immutable, append-only) work improperly
4378 when 2.4 and 2.6 kernels are used on client/server or vice versa
4379 Details : Introduce kernel-independent values for these flags.
4381 Severity : enhancement
4384 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4385 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4386 granularity. Now, Fractional MB granularity is be supported,
4387 this is very useful for low memory system.
4389 Severity : enhancement
4391 Description: Getattr by fid
4392 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4393 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4394 have locks and avoids getattr rpc requests for it on every stat.
4397 Frequency : Always, for filesystems larger than 2TB
4399 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4400 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4401 filesystems larger than 2TB to be mounted with mballoc enabled.
4404 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4406 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4407 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4408 filesystems larger than 2TB is mounted then the kernel will
4409 silently corrupt the start of the filesystem. CONFIG_LBD is
4410 enabled for all CFS-supported kernels, but the possibility of
4411 this happening with a modified kernel config exists.
4413 Severity : enhancement
4415 Description: add client O_DIRECT support for 2.6 kernels
4416 Details : It is now possible to do O_DIRECT reads and writes to files
4417 in the Lustre client mountpoint on 2.6 kernel clients.
4419 Severity : enhancement
4421 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4422 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4423 parallel, not waiting for response from every OST before sending
4424 a rpc to the next OST.
4429 Description: setattr vs write race when updating file timestamps
4430 Details : Client processes that update a file timestamp into the past
4431 right after writing to the file (e.g. tar) it is possible that
4432 the updated file modification time can be reset to the current
4433 time due to a race between processing the setattr and write RPC.
4435 Severity : enhancement
4437 Description: Bring 'lfs find' closer in line with regular Linux find.
4438 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4439 -print0 options and obtains all the needed info through the lustre
4442 Severity : enhancement
4444 Description: support up to 1024 configured devices on one node
4445 Details : change obd_dev array from statically allocated to dynamically
4446 allocated structs as they are first used to reduce memory usage
4451 Description: Flush dirty partially truncated pages during truncate
4452 Details : Immediatelly flush partially truncated pages in filter_setattr,
4453 this way we completely avoid having any pages in page cache on OST
4454 and can retire ugly workarounds during writes to flush such pages.
4459 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4460 Details : protect inode from truncation within vfs_unlink() context
4461 just take a reference before calling vfs_unlink() and release it
4462 when parent's i_sem is free.
4467 Description: last_id value checked outside lock on OST caused LASSERT failure
4468 Details : If there were multiple MDS->OST object precreate requests in
4469 flight, it was possible that the OST's last object id was checked
4470 outside a lock and incorrectly tripped an assertion. Move checks
4471 inside locks, and discard old precreate requests.
4474 Frequency : always, if extents are used on OSTs
4476 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4477 Details : OSTs using the extents format would not zero the high 16 bits of
4478 the index physical block number. This is not a problem for any
4479 OST filesystems smaller than 16TB, and no kernels support ext3
4480 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4481 new/modified files) and can be fixed for existing filesystems
4482 with e2fsprogs-1.39-cfs1.
4487 Description: import connection selection may be incorrect if timer wraps
4488 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4489 import connections to be ignored if the 32-bit jiffies counter
4490 wraps. Use a 64-bit jiffies counter.
4493 Frequency : very large clusters immediately after boot
4495 Description: LNET request buffers exhausted under heavy short-term load
4496 Details : If a large number of client requests are generated on a service
4497 that has previously never seen so many requests it is possible
4498 that the request buffer growth cannot keep up with the spike in
4499 demand. Instead of dropping incoming requests, they are held in
4500 the LND until the RPC service can accept more requests.
4503 Frequency : Sometimes during replay
4505 Description: Assertion failure in ll_local_open after replay.
4506 Details : If replay happened on an open request reply before we were able
4507 to set replay handler, reply will become not swabbed tripping the
4508 assertion in ll_local_open. Now we set the handler right after
4509 recognising of open request
4512 Frequency : very rare
4514 Description: kernel reports "badness in vsnprintf"
4515 Details : Reading from the "recovery_status" /proc file in small chunks
4516 may cause a negative length in lprocfs_obd_rd_recovery_status()
4517 call to vsnprintf() (which is otherwise harmless). Exit early
4518 if there is no more space in the output buffer.
4520 Severity : enhancement
4522 Description: clear OBD RPC statistics by writing to them
4523 Details : It is now possible to clear the OBD RPC statistics by writing
4524 to the "stats" file.
4529 Description: Client mtime is not the same on different clients after utimes
4530 Details : In some cases, the client was using the utimes() syscall on
4531 a file cached on another node. The clients now validate the
4532 ctime from the MDS + OSTs to determine which one is right.
4537 Description: Inability to activate failout mode
4538 Details : lconf script incorrectly assumed that in pythong string's numeric
4539 value is used in comparisons.
4542 Frequency : always with multiple stripes per file
4544 Description: Inefficient object allocation for mutli-stripe files
4545 Details : When selecting which OSTs to stripe files over, for files with
4546 a stripe count that divides evenly into the number of OSTs,
4547 the MDS is always picking the same starting OST for each file.
4548 Return the OST selection heuristic to the original design.
4553 Description: mount failures may take full timeout to return an error
4554 Details : Under some heavy load conditions it is possible that a
4555 failed mount can wait for the full obd_timeout interval,
4556 possibly several minutes, before reporting an error.
4557 Instead return an error as soon as the status is known.
4559 Frequency : quota enabled and large files being deleted
4561 Description: releasing more than 4GB of quota at once hangs OST
4562 Details : If a user deletes more than 4GB of files on a single OST it
4563 will cause the OST to spin in an infinite loop. Release
4564 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4569 Description: statfs data retrieved from /proc may be stale or zero
4570 Details : When reading per-device statfs data from /proc, in the
4571 {kbytes,files}_{total,free,avail} files, it may appear
4572 as zero or be out of date.
4575 Frequency : systems with MD RAID1 external journal devices
4577 Description: lconf's call to blkid is confused by RAID1 journal devices
4578 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4579 of returning all block devices that match the journal UUID.
4582 Frequency : always, for aggregate stripe size over 4GB
4584 Description: assertion fails when trying to use 4GB stripe size
4585 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4586 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4589 Frequency : always on ppc64
4591 Description: the first write on an ext3 filesystem with mballoc got stuck
4592 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4593 perform endianness conversion.
4595 ------------------------------------------------------------------------------
4597 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4599 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4600 INTEROPERATE with older versions automatically. Please read the
4601 user documentation before upgrading any part of a live system.
4602 * WARNING: Lustre networking configuration changes are required with
4603 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4606 * Support for kernels:
4607 2.6.9-22.0.2.EL (RHEL 4)
4608 2.6.5-7.244 (SLES 9)
4609 2.6.12.6 vanilla (kernel.org)
4612 Severity : enhancement
4613 Bugzilla : 7981/8208
4614 Description: Introduced Lustre Networking (LNET)
4615 Details : LNET is new networking infrastructure for Lustre, it includes
4616 a reorganized network configuration mode (see the user
4617 documentation for full details) as well as support for routing
4618 between different network fabrics. Lustre Networking Devices
4619 (LNDs) for the supported network fabrics have also been
4620 created for this new infrastructure.
4622 Severity : enhancement
4623 Description: Introduced Access control lists
4624 Details : clients can set ACLs on files and directories in order to have
4625 more fine-grained permissions than the standard Unix UGO+RWX.
4626 The MDS must be started with the "-o acl" mount option.
4628 Severity : enhancement
4629 Description: Introduced filesystem quotas
4630 Details : Administrators may now establish per-user quotas on the
4633 Severity : enhancement
4635 Description: Configuration change for the XT3
4636 The PTLLND is now used to run Lustre over Portals on the XT3
4637 The configure option(s) --with-cray-portals are no longer used.
4638 Rather --with-portals=<path-to-portals-includes> is used to
4639 enable building on the XT3. In addition to enable XT3 specific
4640 features the option --enable-cray-xt3 must be used.
4645 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4646 Details : A race between allocating a new llcd and re-getting the llcd_lock
4647 allowed another thread to grab newly-allocated llcd.
4649 Severity : enhancement
4651 Description: 2.6 OST async journal commit and locking fix to improve performance
4652 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4653 2.6 kernels are now async as they already were in 2.4 kernels so
4654 that they can commit concurrently with the network bulk transfer.
4655 For block-allocated files the filter allocation semaphore is held
4656 to avoid filesystem fragmentation during allocation. BKL lock
4657 removed for 2.6 xattr operations where it is no longer needed.
4662 Description: lconf incorrectly determined whether two IP networks could talk
4663 Details : In some more complicated routing and multiple-network
4664 configurations, lconf will avoid trying to make a network
4665 connection to a disjoint part of the IP space. It was doing the
4666 math incorrectly for one set of cases.
4671 Description: Fix for potential infinite loop processing records in an llog.
4672 Details : If an llog record is corrupted/zeroed, it is possible to loop
4673 forever in llog_process(). Validate the llog record length
4674 and skip the remainder of the block on error.
4677 Frequency : occasional (liblustre only)
4679 Description: liblustre could not open files whose last component is a symlink
4680 Details : sysio_path_walk() would incorrectly pass the open intent to
4681 intermediate path components.
4684 Frequency : rare (liblustre only with non-standard tuning)
4685 Bugzilla : 7201 (7350)
4686 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4687 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4688 after it has been released, only do this for referenced locks
4690 Severity : enhancement
4692 Description: specifying an (invalid) directory default stripe_size of -1
4693 would reset the directory default striping
4694 Details : stripe_size -1 was used internally to signal directory stripe
4695 removal, now use "all default" to signal dir stripe removal
4696 as a directory striping of "all default" is not useful
4699 Frequency : common for large clusters running liblustre clients
4701 Description: doing an ls when liblustre clients are running is slow
4702 Details : sending a glimpse AST to a liblustre client waits for every AST
4703 to time out, as liblustre clients will not respond. Since they
4704 cannot cache data we refresh the OST lock LVB from disk instead.
4706 Severity : enhancement
4708 Description: doing an ls at the same time as file IO can be slow
4709 Details : enqueue and other "small" requests can be blocked behind many
4710 large IO requests. Create a new OST IO portal for non-IO
4711 requests so they can be processed faster.
4714 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4716 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4717 created file with mode 000
4718 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4719 allow an ftruncate() as a non-root user to a file with mode 000.
4720 Lustre now respects this flag to disable mode checks when
4721 truncating a file owned by the user
4724 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4727 Description: Revoking locks from clients that went dead or catatonic might take
4729 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4730 cancellation of such locks instant on servers without waiting for
4731 any reply from clients. Clients drops these locks when cancel
4732 notification from server is received without replying.
4735 Frequency : liblustre-only, when liblustre client dies or becomes busy
4737 Description: Doing ls on Linux clients can take a long time with active
4739 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4740 granting such locks to it in the first place if possible. Locks
4741 are taken by proxy on the OST during the read or write and
4742 dropped immediately afterward. Add connect flags handling, do
4743 not grant locks to liblustre clients for glimpse ASTs.
4745 Severity : enhancement
4747 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4748 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4749 sequential reads and behaves badly with applications doing random
4750 reads. Improve it by reading ahead at least read region, and
4751 avoiding excessive large RPC for small reads.
4753 Severity : enhancement
4755 Description: Creating more than 1000 files for a single job may cause a load
4756 imbalance on the OSTs if there are also a large number of OSTs.
4757 Details : qos_prep_create() uses an OST index reseed value that is an
4758 even multiple of the number of available OSTs so that if the
4759 reseed happens in the middle of the object allocation it will
4760 still utilize the OSTs as uniformly as possible.
4765 Description: OST or MDS may oops in ping_evictor_main()
4766 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4767 but doesn't restart at beginning of obd_exports_timed list
4770 Severity : enhancement
4772 Description: improve by-nid export eviction on the MDS and OST
4773 Details : allow multiple exports with the same NID to be evicted at one
4774 time without re-searching the exports list.
4777 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4779 Description: MDS may oops in groups_free()
4780 Details : in rare race conditions a newly allocated group_info struct is
4781 freed again, and this can be NULL. The 2.4 compatibility code
4782 for groups_free() checked for a NULL pointer, but 2.6 did not.
4785 Frequency : common for liblustre clients doing little filesystem IO
4786 Bugzilla : 9352, 7313
4787 Description: server may evict liblustre clients accessing contended locks
4788 Details : if a client is granted a lock or receives a completion AST
4789 with a blocking AST already set it would not reply to the AST
4790 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4793 Frequency : lfs setstripe, only systems with more than 160 OSTs
4795 Description: unable to set striping with a starting offset beyond OST 160
4796 Details : llapi_create_file() incorrectly limited the starting stripe
4797 index to the maximum single-file stripe count.
4800 Frequency : LDAP users only
4802 Description: lconf did not handle in-kernel recovery with LDAP properly
4803 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4805 Severity : enhancement
4807 Description: bind OST threads to NUMA nodes to improve performance
4808 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4809 node and do their allocations there to localize memory access
4811 Severity : enhancement
4813 Description: llmount can determine client NID directly from Myrinet (GM)
4814 Details : the client NID code from gmnalnid was moved directly into
4815 llmount, removing the need to use this or specifying the
4816 client NID explicitly when mounting GM clients with zeroconf
4819 Frequency : if client is started with down MDS
4821 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4822 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4823 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4824 ptlrpc_import_delayed_req() because we didn't check if the
4825 request was interrupted, and we also didn't break out of the
4826 event loop if there was no timeout
4831 Description: data loss during non-page-aligned writes to a single file from
4832 both multiple nodes and multiple threads on one node at same time
4833 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4834 inconsistency led to false short-reads, that were cached and later
4835 used by ->prepare_write() to fill in partially written page,
4836 leading to data loss.
4839 Frequency : always, if lconf --abort_recovery used
4841 Description: lconf --abort_recovery fails with 'Operation not supported'
4842 Details : lconf was attempting to abort recovery on the MDT device and not
4845 Severity : enhancement
4847 Description: remove cleanup logs
4848 Details : replace lconf-generated cleanup logs with lustre internal
4849 cleanup routines. Eliminates the need for client-cleanup and
4852 Severity : enhancement
4854 Description: add support for EAs (user and system) on lustre filesystems
4855 Details : it is now possible to store extended attributes in the Lustre
4856 client filesystem, and with the user_xattr mount option it
4857 is possible to allow users to store EAs on their files also
4859 Severity : enhancement
4861 Description: Add possibility (config option) to show minimal available OST free
4863 Details : When compiled with --enable-mindf configure option, statfs(2)
4864 (and so, df) will return least minimal free space available from
4865 all OSTs as amount of free space on FS, instead of summary of
4866 free spaces of all OSTs.
4868 Severity : enhancement
4870 Description: do not expand extent locks acquired on OST-side
4871 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4872 by server: they are not cached anyway.
4875 Frequency : when mmap is used/binaries executed from Lustre
4877 Description: Unmmap pages before throwing them away from read cache.
4878 Details : llap_shrink cache now attempts to unmap pages before discarding
4879 them (if unmapping failed - do not discard). SLES9 kernel has
4880 extra checks that trigger if this unmapping is not done first.
4885 Description: lconf didn't resolve symlinks before checking to see whether a
4886 given mountpoint was already in use
4889 Frequency : when migrating failover services
4890 Bugzilla : 6395, 9514
4891 Description: When migrating a subset of services from a node (e.g. failback
4892 from a failover service node) the remaining services would
4893 time out and evict clients.
4894 Details : lconf --force (implied by --failover) sets the global obd_timeout
4895 to 5 seconds in order to quickly disconnect, but this caused
4896 other RPCs to time out too quickly. Do not change the global
4897 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4899 Severity : enhancement
4900 Frequency : if MDS is started with down OST
4901 Bugzilla : 9439,5706
4902 Description: Allow startup/shutdown of an MDS without depending on the
4903 availability of the OSTs.
4904 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4905 Add appropriate locking and lov-osc refcounts for safe
4906 cleaning. Add osc abort_inflight calls in case the
4907 synchronize never started.
4910 Frequency : occasional (Cray XT3 only)
4912 Description: root not authorized to access files in CRAY_PORTALS environment
4913 Details : The client process capabilities were not honoured on the MDS in
4914 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4915 been accessed by an authorized user then root was able to access
4916 the file on the local client also. The root user capabilities
4917 are now allowed on the MDS, as this environment has secure UID.
4920 Frequency : occasional
4922 Description: ldiskfs "too long searching" message happens too often
4923 Details : A debugging message (otherwise harmless) prints too often on
4924 the OST console. This has been reduced to only happen when
4925 there are fragmentation problems on the filesystem.
4930 Description: Division by zero in statfs when all OSCs are inactive
4931 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4932 lov_statfs divides by value returned by lov_get_stripecnt().
4936 Bugzilla : 9489, 3273
4937 Description: First write from each client to each OST was only 4kB in size,
4938 to initialize client writeback cache, which caused sub-optimal
4939 RPCs and poor layout on disk for the first written file.
4940 Details : Clients now request an initial cache grant at (re)connect time
4941 and so that they can start streaming writes to the cache right
4942 away and always do full-sized RPCs if there is enough data.
4943 If the OST is rebooted the client also re-establishes its grant
4944 so that client cached writes will be honoured under the grant.
4949 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4950 Details : Now I/O RPCs go to different portal number and (presumably) fast
4951 lock requests (and glimses) and other RPCs get their own service
4952 threads pool that should be able to service those RPCs
4955 Severity : enhancement
4957 Description: Ability to exchange lustre version between client and servers and
4958 issue warnings at client side if client is too old. Also for
4959 liblustre clients there is ability to refuse connection of too old
4961 Details : New 'version' field is added to connect data structure that is
4962 filled with version info. That info is later checked by server and
4966 Frequency : rare, liblustre only.
4967 Bugzilla : 9296, 9581
4968 Description: Two simultaneous writes from liblustre at offset within same page
4969 might proceed at the same time overwriting eachother with stale
4971 Details : I/O lock withing llu_file_prwv was released too early, before data
4972 actually was hitting the wire. Extended lock-holding time until
4973 server acknowledges receiving data.
4976 Frequency : extremely rare. Never observed in practice.
4978 Description: avoid generating lustre_handle cookie of 0.
4979 Details : class_handle_hash() generates handle cookies by incrementing
4980 global counter, and can hit 0 occasionaly (this is unlikely, but
4981 not impossible, because initial value of cookie counter is
4982 selected randonly). Value of 0 is used as a sentinel meaning
4983 "unassigned handle" --- avoid it. Also coalesce two critical
4984 sections in this function into one.
4986 Severity : enhancement
4988 Description: allow liblustre clients to delegate truncate locking to OST
4989 Details : To avoid overhead of locking, liblustre client instructs OST to
4990 take extent lock in ost_punch() on client's behalf. New connection
4991 flag is added to handle backward compatibility.
4993 Severity : enhancement
4994 Bugzilla : 4928, 7341, 9758
4995 Description: allow number of OST service threads to be specified
4996 Details : a module parameter allows the number of OST service threads
4997 to be specified via "options ost ost_num_threads={N}" in the
4998 OSS's /etc/modules.conf or /etc/modprobe.conf.
5002 Bugzilla : 6146, 9635, 9895
5003 Description: servers crash with bad pointer in target_handle_connect()
5004 Details : In rare cases when a client is reconnecting it was possible that
5005 the connection request was the last reference for that export.
5006 We would temporarily drop the export reference and get a new
5007 one, but this may have been the last reference and the export
5008 was just destroyed. Get new reference before dropping old one.
5010 Severity : enhancement
5011 Frequency : if client is started with failover MDS
5013 Description: Allow multiple MDS hostnames in the mount command
5014 Details : Try to read the configuration from all specified MDS
5015 hostnames during a client mount in case the "primary"
5018 Severity : enhancement
5020 Description: Stop sending data to evicted clients as soon as possible.
5021 Details : Check if the client we are about to send or are sending data to
5022 was evicted already. (Check is done every second of waiting,
5023 for which l_wait_event interface was extended to allow checking
5024 of exit condition at specified intervals).
5027 Frequency : rare, normally only when NFS exporting is done from client
5029 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5031 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5034 Frequency : always, if srand() is called before liblustre initialization
5036 Description: Liblustre uses system PRNG disturbing its usage by user application
5037 Details : Introduce internal to lustre fast and high-quality PRNG for
5038 lustre usage and make liblustre and some other places in generic
5039 lustre code to use it.
5041 Severity : enhancement
5042 Bugzilla : 9477, 9557, 9870
5043 Description: Verify that the MDS configuration logs are updated when xml is
5044 Details : Check if the .xml configuration logs are newer than the config
5045 logs stored on the MDS and report an error if this is the case.
5046 Request --write-conf, or allow starting with --old_conf.
5048 Severity : enhancement
5050 Description: Handle symlinks in the path when checking if Lustre is mounted.
5051 Details : Resolve intermediate symlinks when checking if a client has
5052 mounted a filesystem to avoid duplicate client mounts.
5057 Description: lconf can hit an error exception but still return success.
5058 Details : The lconf command catches the Command error exception at the top
5059 level script context and will exit with the associated exit
5060 status, but doesn't ensure that this exit status is non-zero.
5065 Description: failure of ptlrpc thread startup can cause oops
5066 Details : Starting a ptlrpc service thread can fail if there are a large
5067 number of threads or the server memory is very fragmented.
5068 Handle this without oopsing.
5071 Frequency : always, only if liblustre and non-default acceptor port was used
5073 Description: liblustre cannot connect to servers with non-default acceptor port
5074 Details : tcpnal_set_default_params() was not called and was therefore
5075 ignoring the environment varaible TCPNAL_PORT, as well as other
5076 TCPNAL_ environment variables
5081 Description: two objects could be created on the same OST for a single file
5082 Details : If an OST is down, in some cases it was possible to create two
5083 objects on a single OST for a single file. No problems other
5084 than potential performance impact and spurious error messages.
5088 Bugzilla : 5681, 9562
5089 Description: Client may oops in ll_unhash_aliases
5090 Details : Client dcache may become inconsistent in race condition.
5091 In some cases "getcwd" can fail if the current directory is
5097 Description: Inode refcounting problems in NFS export code
5098 Details : link_raw functions used to call d_instantiate without obtaining
5099 extra inode reference first.
5103 Bugzilla : 9942, 9903
5104 Description: Referencing freed requests leading to crash, memleaks with NFS.
5105 Details : We used to require that call to ll_revalidate_it was always
5106 followed by ll_lookup_it. Also with revalidate_special() it is
5107 possible to call ll_revalidate_it() twice for the same dentry
5108 even if first occurence returned success. This fix changes semantic
5109 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5110 reference on a request referred from the intent.
5111 ll_intent_release() then releases such a request.
5114 Frequency : rare, normally benchmark loads only
5116 Description: unlinked inodes were kept in memory on the client
5117 Details : If a client is repeatedly creating and unlinking files it
5118 can accumulate a lot of stale inodes in the inode slab cache.
5119 If there is no other client load running this can cause the
5120 client node to run out of memory. Instead flush old inodes
5121 from client cache that have the same inode number as a new inode.
5124 Frequency : SLES9 2.6.5 kernel and long filenames only
5125 Bugzilla : 9969, 10379
5126 Description: utime reports stale NFS file handle
5127 Details : SLES9 uses out-of-dentry names in some cases, which confused
5128 the lustre dentry revalidation. Change it to always use the
5132 Frequency : rare, unless heavy write-truncate concurrency is continuous
5133 Bugzilla : 4180, 6984, 7171, 9963, 9331
5134 Description: OST becomes very slow and/or deadlocked during object unlink
5135 Details : filter_destroy() was holding onto the parent directory lock
5136 while truncating+unlinking objects. For very large objects this
5137 may block other threads for a long time and slow overall OST
5138 responsiveness. It may also be possible to get a lock ordering
5139 deadlock in this case, or run out of journal credits because of
5140 the combined truncate+unlink. Solution is to do object truncate
5141 first in one transaction without parent lock, and then do the
5142 final unlink in a new transaction with the parent lock. This
5143 reduces the lock hold time dramatically.
5146 Frequency : rare, 2.4 kernels only
5148 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5149 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5150 before unmounting their filesystem in order to avoid deadlock.
5151 The kernel_locked() function in 2.4 kernels only checks whether
5152 the kernel lock is held, not whether it is this process that is
5153 holding it as 2.6 kernels do.
5158 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5159 Details : The client ptlrpc code may be trying to reconnect to a down
5160 server before a previous connection attempt has timed out.
5161 Increase the reconnect interval to be longer than the connection
5162 timeout interval to avoid sending duplicate connections to
5166 Frequency : echo_client brw_test command
5168 Description: fix echo_client to work with OST preallocated code
5169 Details : OST preallocation code (5137) didn't take echo_client IO path
5170 into account: echo_client calls filter methods outside of any
5171 OST thread and, hence, there is no per-thread preallocated
5172 pages and buffers to use. Solution: hijack pga pages for IO. As
5173 a byproduct, this avoids unnecessary data copying.
5177 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5178 Description: Client can oops in mdc_commit_close() after open replay
5179 Details : It was possible for the MDS to return an open request with no
5180 transaction number in mds_finish_transno() if the client was
5181 evicted, but without actually returning an error. Clients
5182 would later try to replay that open and may trip an assertion
5183 Simplify the client close codepath, and always return an error
5184 from the MDS in case the open is not successful.
5187 Frequency : rare, 2.6 OSTs only
5189 Description: OST may deadlock under high load on fragmented files
5190 Details : If there was a heavy load and highly-fragmented OST filesystems
5191 it was possible to have all the OST threads deadlock waiting on
5192 allocation of biovecs, because the biovecs were not released
5193 until the entire RPC IO was completed. Instead, release biovecs
5194 as soon as they are complete to ensure forward IO progress.
5196 Severity : enhancement
5198 Description: Support for specifying external journal device at mount
5199 Details : If an OST or MDS device is formatted with an external journal
5200 device, this device major/minor is stored in the ext3 superblock
5201 and may not be valid for failover. Allow detecting and
5202 specifying the external journal at mount time.
5207 Description: Mounting an MDS with pending unlinked files may cause oops
5208 Details : target_finish_recovery() calls mds_postrecov() which returned
5209 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5210 considers this an error and immediately begins cleaning up the
5211 lov, just after starting the mds_lov process
5213 Severity : enhancement
5215 Description: Implement 'lfs df' to report actual free space on per-OST basis
5216 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5217 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5218 usage of objects; '-h' to report in human readable format.
5220 ------------------------------------------------------------------------------
5222 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5229 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5230 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5231 at startup during orphan unlinking. Instead, skip buddy update
5232 before setup, as it will be regenerated after recovery is complete.
5237 Description: If an OST is inactive, its locks might reference stale inodes.
5238 Details : lov_change_cbdata() must iterate over all namespaces, even if
5239 they are inactive to clear inode references from the lock.
5241 Severity : enhancement
5242 Frequency : occasional, if non-standard max_dirty_mb used
5244 Description: Client will block write RPCs if not enough grant
5245 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5246 then the client will block writes while waiting for another RPC
5247 to complete instead of consuming its dirty limit. With change
5248 we get improved performance when max_dirty_mb is small.
5250 Severity : enhancement
5251 Bugzilla : 3389, 6253
5252 Description: Add support for supplementary groups on the MDS.
5253 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5254 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5255 lookups for user supplementary groups into a cache.
5259 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5260 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5261 for liblustre O_EXCL is enforced by MDS.
5265 Description: readdir never returns NULL in liblustre.
5266 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5267 dentry in struct dirent.
5271 Frequency : liblustre only, depends on application IO pattern
5272 Description: liblustre clients evicted if not contacting servers
5273 Details : Don't put liblustre clients into the ping_evictor list, so
5274 they will not be evicted by the pinger ever.
5276 Severity : enhancement
5278 Description: Add ability to evict clients by NID from MDS.
5279 Details : By echoing "nid:$NID" string into
5280 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5281 $NID would be instantly evicted from this MDS and from all active
5282 OSTs connected to it.
5286 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5287 Details : lookup_it_finish() used to query file size from OSTs that was not
5292 Description: service threads change working directory to that of init
5293 Details : Starting lustre service threads may pin the working directory
5294 of the parent thread, making that filesystem busy. Threads
5295 now change to the working directory of init to avoid this.
5299 Frequency : during shutdown only
5300 Description: shutdown with a failed MDS or OST can cause unmount to hang
5301 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5306 Frequency : 2.6 only
5307 Description: chmod/chown may include an extra supplementary group
5308 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5309 supplementary group and if none is specified this is used.
5312 Bugzilla : 5479 (6816)
5314 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5315 Details : If lookup is in progress on a file that is unlinked we might try
5316 to revalidate the inode and fail in revalidate after lookup is
5317 complete and ll_file_open() enqueues the open again but
5318 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5321 Frequency : always, if lconf --abort_recovery used
5323 Description: lconf --abort_recovery fails with 'Operation not supported'
5324 Details : lconf was attempting to abort recovery on the MDT device and not
5327 ------------------------------------------------------------------------------
5329 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5334 Frequency : rare (only unsupported configurations with a node running as an
5336 Bugzilla : 6514, 5137
5337 Description: Mounting a Lustre file system on a node running as an OST could
5339 Details : OSTs now preallocates memory needed to write out data at
5340 startup, instead of when needed, to avoid having to
5341 allocate memory in possibly low memory situations.
5342 Specifically, if the file system is mounted on on OST,
5343 memory pressure could force it to try to write out data,
5344 which it needed to allocate memory to do. Due to the low
5345 memory, it would be unable to do so and the node would
5346 become unresponsive.
5348 Severity : enhancement
5350 Description: Addition of lconf --service command line option
5351 Details : lconf now accepts a '--service <arg>' option, which is
5352 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5354 Severity : enhancement
5356 Description: Failover mode is now the default for OSTs.
5357 Details : By default, OSTs will now run in failover mode. To return to
5358 the old behaviour, add '--failout' to the lmc line for OSTs.
5360 Severity : enhancement
5362 Description: Health checks are now provided for MDS and OSTs
5363 Details : Additional detailed health check information on MSD and OSTs
5364 is now provided through the procfs health_check value.
5367 Frequency : occasional, depends on IO load
5369 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5370 after numerous create/delete cycles
5371 Details : The ext3 inode allocation policy would not allocate new inodes
5372 very well on the OSTs because there are no new directories
5373 being created. Instead we look for groups with free space if
5374 the parent directories are nearly full.
5379 Description: Network or server problems during mount may cause partially
5380 mounted clients instead of returning an error.
5381 Details : The config llog parsing code may overwrite the error return
5382 code during mount error handling, returning success instead
5388 Description: MDS can fail to allocate large reply buffers
5389 Details : After long uptimes the MDS can fail to allocate large reply
5390 buffers (e.g. zconf client mount config records) due to memory
5391 fragmentation or consumption by the buffer cache. Preallocate
5392 some large reply buffers so that these replies can be sent even
5393 under memory pressure.
5397 Frequency : rare (liblustre)
5398 Description: fsx running with liblustre complained that using truncate() to
5399 extend the file doesn't work. This patch corrects that issue.
5400 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5401 ATTR_SIZE and lsm use in llu_setattr_raw.
5405 Frequency : rare, only 2.6 kernels
5406 Description: Unusual file access patterns on the MDS may result in inode
5407 data being lost in very rare circumstances.
5408 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5409 code in the 2.6 kernel caused the inode and/or EA data not to
5410 be read from disk, causing single-file corruption.
5414 Frequency : rare, only 2.6 filesystems using extents
5415 Description: Heavy concurrent write and delete load may cause data corruption.
5416 Details : It was possible under high-load situations to have an extent
5417 metadata block in the block device cache from a just-unlinked
5418 file overwrite a newly-allocated data block. We now unmap any
5419 metadata buffers that alias just-allocated data blocks.
5423 Frequency : filesystems with default stripe_count larger than 77
5424 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5425 Details : lconf specifies an inode size of 4096 bytes when the default
5426 stripe_count is larger than 77. This conflicts with the default
5427 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5428 this case to avoid pinning too much memory for large EAs.
5430 ------------------------------------------------------------------------------
5432 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5437 Frequency : rare (extremely heavy IO load with hundreds of clients)
5439 Description: Client is evicted, gets IO error writing to file
5440 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5441 caused clients to be evicted for AST timeouts. The fixes in
5442 bug 5192 mean we no longer need to have such short AST timeouts
5443 so ldlm_timeout has been increased.
5446 Frequency : occasional during --force or --failover shutdown under load
5447 Bugzilla : 5949, 4834
5448 Description: Server oops/LBUG if stopped with --force or --failover under load
5449 Details : a collection of import/export refcount and cleanup ordering
5450 issues fixed for safer force cleanup
5453 Frequency : only filesystems larger than 120 OSTs
5454 Bugzilla : 5990, 6223
5455 Description: lfs getstripe would oops on a very large filesystem
5456 Details : lov_getconfig used kfree on vmalloc'd memory
5459 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5460 Bugzilla : 6242, 6243
5461 Description: reading from files that had been truncated to a non-zero size
5462 but never opened returned no data
5463 Details : ll_file_read() reads zeros from no-object files to EOF
5468 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5469 Details : There's little harm in aborting MDS/OSS recovery and letting it
5470 try again, so I removed the LASSERT and return an error instead.
5472 Severity : enhancement
5474 Description: New debugging infrastructure for tracking down data corruption
5475 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5476 (b) cover more of the client-side code path, and (c) try to narrow
5477 down where problems occurred
5481 Bugzilla : 3819, 4364, 4397, 6313
5482 Description: Racing close and eviction MDS could cause assertion in mds_close
5483 Details : It was possible to get multiple mfd references during close and
5484 client eviction, leading to one thread referencing a freed mfd.
5486 Severity: : enhancement
5487 Bugzilla : 3262, 6359
5488 Description: Attempts to reconnect to servers are now more aggressive.
5489 Details : This builds on the enhanced upcall-less recovery that was added
5490 in 1.4.2. When trying to reconnect to servers, clients will
5491 now try each server in the failover group every 10 seconds. By
5492 default, clients would previously try one server every 25 seconds.
5497 Description: After recovery, certain operations trigger a failed
5498 assertion on a client.
5499 Details : Failing over an mds, using lconf -d --failover, while a
5500 client was doing a readdir() call would cause the client to
5501 LBUG after recovery completed and the readdir() was resent.
5503 Severity : enhancement
5505 Description: Default groups are now added by lconf
5506 Details : You can now run lconf --group <servicename> without having to
5507 manually add groups with lmc.
5510 Frequency : occasional
5512 Description: Nodes with an elan id of 0 trigger a failed assertion
5515 Frequency : always when accessing e.g. tty/console device nodes
5517 Description: tty and some other devices nodes cannot be used on lustre
5518 Details : file's private_data field is used by device data and lustre
5519 values in there got lost. New field was added to struct file to
5520 store fs-specific private data.
5523 Frequency : when exporting Lustre via NFS
5525 Description: NFSD failed occasionally when looking up a path component
5526 Details : NFSD is looking up ".." which was broken in ext3 directories
5527 that had grown large enough to become hashed.
5530 Frequency : Clusters with multiple interfaces not on the same subnet
5532 Description: Nodes will repeatedly try to reconnect to an interface which it
5533 cannot reach and report an error to the log.
5534 Details : Extra peer list entries will be created by lconf with some peers
5535 unreachable. lconf now validates the peer before adding it.
5538 Frequency : Only if a default stripe is set on the filesystem root.
5540 Description: Setting a default stripe on the filesystem root prevented the
5541 filesystem from being remounted.
5542 Details : The client was sending extra request flags in the root getattr
5543 request and did not allocate a reply buffer for the dir EA.
5546 Frequency : occasional, higher if lots of files are accessed by one client
5547 Bugzilla : 6159, 6097
5548 Description: Client trips assertion regarding lsm mismatch/magic
5549 Details : While revalidating inodes the VFS looks up inodes with ifind()
5550 and in rare cases can find an inode that is being freed.
5551 The ll_test_inode() code will free the lsm during ifind()
5552 when it finds an existing inode and then the VFS later attaches
5553 this free lsm to a new inode.
5557 Bugzilla : 6422, 7030
5558 Description: MDS deadlock between mkdir and client eviction
5559 Details : Creating a new file via mkdir or mknod (starting a transaction
5560 and getting the ns lock) can deadlock with client eviction
5561 (gets ns lock and trying to finish a synchronous transaction).
5564 Frequency : occasional
5565 Description: While starting a server, the fsfilt_ext3 module could not be
5567 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5568 kernels. Previously, lconf would still use the ext3 name
5569 when trying to load modules. Now, it will correctly use
5570 ext3 on 2.4 and ldiskfs on 2.6.
5572 Severity : enhancement
5573 Description: The default stripe count has been changed to 1
5574 Details : The interpretation of the default stripe count (0, to lfs
5575 or lmc) has been changed to mean striping across a single
5576 OST, rather than all available. For general usage we have
5577 found a stripe count of 1 or 2 works best.
5579 Severity : enhancement
5580 Description: Add support for compiling against Cray portals.
5581 Details : Conditional compiling for some areas that are different
5585 Frequency : occasional
5586 Bugzilla : 6409, 6834
5587 Description: Creating files with an explicit stripe count may lead to
5588 a failed assertion on the MDS
5589 Details : If some OSTs are full or unavailable, creating files may
5590 trigger a failed assertion on the MDS. Now, Lustre will
5591 try to use other servers or return an error to the
5595 Frequency : occasional
5597 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5598 caused lock timeout during readahead (since 1.4.2).
5599 Details : Processes doing readahead might match a lock that hasn't been
5600 granted yet if there are overlapping and conflicting lock
5601 requests. The readahead process waits on ungranted lock
5602 (original lock is CBPENDING), while OST waits for that process
5603 to cancel CBPENDING read lock and eventually evicts client.
5605 Severity : enhancement
5607 Description: Initial enabling of flock support for clients
5608 Details : Implements fcntl advisory locking and file status functions.
5609 This feature is provided as an optional mount flag (default
5610 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5611 locking are implemented yet, and those that are are not guaranteed
5612 to be completely correct in production environments.
5613 mount -t lustre -o [flock|noflock] ...
5616 Frequency : occasional
5618 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5619 assertion in the ext3 JBD (journaling) layer.
5620 Details : The b_committed_data struct is protected by the big kernel lock
5621 in 2.4 kernels, serializing journal_commit_transaction() and
5622 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5623 access to this struct. In 2.6 kernels there is finer grained
5624 locking to improve SMP performance of the JBD layer.
5628 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5633 Description: merge and cleanup kernel patches.
5636 -----------------------------------------------------------------------------
5638 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5640 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5641 versions of Lustre. Please update all servers and clients to
5642 version 1.4.2 or later at the same time. You must also run
5643 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5644 to update the configuration logs.
5646 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5647 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5648 - swab llog records of type '0' so we get proper header size/idx (5861)
5649 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5650 - fix rename of one directory over another leaking an inode (5953)
5651 - avoid SetPageDirty on 2.6 (5981)
5652 - don't re-add just-being-destroyed locks to the waiting list (5653)
5653 - when creating new directories, inherit the parent's custom
5654 striping settings if present parent (3048)
5655 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5656 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5657 - don't hold client locks on temporary worklist from l_lru (5666)
5658 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5659 - automatically evict dead clients (5921)
5660 - Update file size properly in create+truncate+fstat case (6196)
5661 - Do not unhash mountpoint dentries, do not allow removal of
5663 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5664 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5665 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5666 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5667 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5668 - Correctly handle reads of files with no objects (6243)
5669 - lctl recover will also mark a device active if deactivate used (5933)
5671 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5672 - allow --write-conf on an MDS with different nettype than client (5619)
5673 - don't write config llogs to MDS for mounts not from that MDS (5617)
5674 - lconf should create multiple TCP connections from a client (5201)
5675 - init scripts are now turned off by default; run chkconfig --on
5676 lustre and chkconfig --on lustrefs to use them
5677 - upcalls are no longer needed for clients to recover to failover
5679 - add --abort-recovery option to lconf to abort recovery on device
5681 - add support for an arbitrary number of OSTs (3026)
5682 - Quota support protocol changes.
5683 - forward compatibility changes to wire structs (6007)
5684 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5685 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5686 - improved "lustre status" script
5687 - initialize blocksize for non-regular files (6062)
5688 - added --disable-server and --disable-client configure options (5782)
5689 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5690 - Vanilla 2.4.29 support
5691 - increase maximum number of obd devices to 520 (6242)
5692 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5693 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5694 SLES 9 2.6.5 kernels
5695 - limit stripes per file to 160 (the maximum EA size) (6093)
5697 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5700 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5701 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5702 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5703 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5704 - small ext3 extents cleanups and fixes (5733)
5705 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5706 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5707 - avoid needless client->OST connect, fix handle mismatch (5317)
5708 - fix DLM error path that led to out-of-sync client, long delays (5779)
5709 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5710 - fix several locking issues related to i_size (5492,5624,5654,5672)
5711 - don't move pending lock onto export if it is already evicted (5683)
5712 - fix kernel oops when creating .foo in unlinked directory (5548)
5713 - fix deadlock in obdfilter statistics vs. object create (5811)
5714 - use time_{before,after} to avoid timer jiffies wrap (5882)
5715 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5716 - Do not leak request if server was not able to process it (5154)
5717 - If mds_open unable to find parent dir, make that negative lookup(5154)
5718 - don't create new directories with extent-mapping (5909, 5936)
5720 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5721 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5722 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5723 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5724 - support for RHEL4 kernel on i686 (5773)
5725 - provide error messages when incompatible logs are encountered (5898)
5727 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5728 * version 1.4.0.10 (1.4.1 release candidate 1)
5730 - don't keep a lock reference when lock is not granted (4238)
5731 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5732 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5733 - drop import inflight refcount on signal_completed_replay error (5255)
5734 - unlock page after async write error during send (3677)
5735 - handle missing objects in filter_preprw_read properly (5265)
5736 - no transno return for symlink open, don't save no-trasno open (3440)
5737 - don't try to complete elan receive that already failed (4012)
5738 - free RPC server reply state on error (5406)
5739 - clean up thread from ptlrpc_start_thread() on error (5160)
5740 - readahead could read extra page into cache that wasn't ejected (5388)
5741 - prevent races in class_attach/setup/cleanup/detach (5260)
5742 - don't dereference de->d_inode after l_dput of de (5458)
5743 - use "int" for stripe value returned from lock_to_stripe (5544)
5744 - mballoc allocation and error-checking fixes in 2.6 (5504)
5745 - block device patches to fix I/O request sizes in 2.6 (5482)
5746 - look up hostnames for IB nals (5602)
5747 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5748 - don't start multiple acceptors for the same port (5277)
5749 - fix incorrect LASSERT in mds_getattr_name (5635)
5750 - export a proc file for general "ping" checking (5628)
5751 - fix "lfs check" to not block when the MDS is down (5628)
5753 - service request history (4965)
5754 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5755 - create an "evict_client" /proc entry on OSTs, like the MDS has
5756 - fix mount usage message, return errors per mount(8) (5168)
5757 - change grep [] to grep "[]" in tests so they work in more UMLs
5758 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5759 - remove ancient LOV_MAGIC_V0 EA support (5047)
5760 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5761 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5762 - allow readahead allocations to fail when low on memory (5383)
5763 - mmap locking landed again, after considerable improvement (2828)
5764 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5766 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5769 - send OST transaction number in read/write reply to free req (4966)
5770 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5771 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5772 - the watchdog thread now runs as interruptible (5246)
5773 - flock/lockf fixes (but it's still disabled, pending 5135)
5774 - don't use EXT3 constants in llite code (5094)
5775 - memory shortage at startup could cause assertion (5176)
5777 - reorganization of lov code
5778 - single portals codebase
5780 - add extents/mballoc support (5025)
5781 - direct I/O reads in the obdfilter (4048)
5782 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5784 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5787 - send OST transaction number in read/write reply to free req (4966)
5788 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5789 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5790 - the watchdog thread now runs as interruptible (5246)
5791 - handle missing objects in filter_preprw_read properly (5265)
5792 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5793 - drop import inflight refcount on signal_completed_replay error (5255)
5794 - unlock page after async write error during send (3677)
5795 - return original error code on reconstructed replies (3761)
5796 - no transno return for symlink open, don't save no-trasno open (3440)
5798 - add pid to ldlm debugging output (4922)
5799 - bump the watchdog timeouts -- we can't handle 30sec yet
5800 - extra debugging for orphan dentry/inode bug (5259)
5802 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5805 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5806 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5807 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5808 - fix NULL dereference in /proc/sys/portals/routes (4827)
5809 - allow failed mdc_close() operations to be interrupted (4561)
5810 - stop precreate on OST before MDS would time out on it (4778)
5811 - don't send partial-page writes before EOF from client (4410)
5812 - discard client grant for sub-page writes on large-page clients (4520)
5813 - don't free dentries not owned by NFS code, check generation (4806)
5814 - fix lsm leak if mds_create_objects() fails (4801)
5815 - limit debug_daemon file size, always print CERROR messages (4789)
5816 - use transno after validating reply (3892)
5817 - process timed out requests if import state changes (3754)
5818 - update mtime on OST during writes, return in glimpse (4829)
5819 - add mkfsoptions to LDAP (4679)
5820 - use ->max_readahead method instead of zapping global ra (5039)
5821 - don't interrupt __l_wait_event() during strace
5823 - add software watchdogs to catch hung threads quickly (4941)
5824 - make lustrefs init script start after nfs is mounted
5825 - fix CWARN/ERROR duplication (4930)
5826 - return async write errors to application if possible (2248)
5827 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5828 - print NAL number in %x format (4645)
5829 - update barely-supported suse-2.4.21-171 series (4842)
5830 - support for sles 9 %post scripts
5831 - support for building 2.6 kernel-source packages
5832 - support for sles km_* packages
5834 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5837 - ignore -ENOENT errors in osc_destroy (3639)
5838 - notify osc create thread that OSC is being cleaned up (4600)
5839 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5840 - reconstruct ost_handle() like mds_handle() (4657)
5841 - create a new thread to do import eviction to avoid deadlock (3969)
5842 - let lconf resolve symlinked-to devices (4629)
5843 - don't unlink "objects" from directory with default EA (4554)
5844 - hold socknal file ref over connect in case target is down (4394)
5845 - allow more than 32000 subdirectories in a single directory (3244)
5846 - fix blocks count for O_DIRECT writes (3751)
5847 - OST returns ENOSPC from object create when no space left (4539)
5848 - don't send truncate RPC if file size isn't changing (4410)
5849 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5850 - bind to privileged port in socknal and tcpnal (3689)
5852 - rate limit CERROR/CWARN console message to avoid overload (4519)
5853 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5854 - basic mmap support (3918)
5855 - kernel patch series update from b1_4 (4711)
5857 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5860 - avoid crash during MDS cleanup with OST shut down (2775)
5861 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5862 - don't use bad inodes on the MDS (3744)
5863 - dynamic object preallocation to improve recovery speed (4236)
5864 - don't hold spinlock over lock dumping or change debug flags (4401)
5865 - don't zero obd_dev when it is force cleaned (3651)
5866 - print grants to console if they go negative (4431)
5867 - "lctl deactivate" will stop automatic recovery attempts (3406)
5868 - look for existing locks in ldlm_handle_enqueue() (3764)
5869 - don't resolve lock handle twice in recovery avoiding race (4401)
5870 - revalidate should check working dir is a directory (4134)
5872 - don't always mark "slow" obdfilter messages as errors (4418)
5874 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5877 - don't close LustreDB during write_conf until it is done (3860)
5878 - fix typo in lconf for_each_profile (3821)
5879 - allow dumping logs from multiple threads at one time (3820)
5880 - don't allow multiple threads in OSC recovery (3812)
5881 - fix debug_size parameters (3864)
5882 - fix mds_postrecov to initialize import for llog ctxt (3121)
5883 - replace config semaphore with spinlock (3306)
5884 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5885 - don't allow enqueue to complete on a destroyed export (3822)
5886 - down write_lock before checking llog header bitmap (3825)
5887 - recover from lock replay timeout (3764)
5888 - up llog sem before sending rpc (3652)
5889 - reduce ns lock hold times when setting kms (3267)
5890 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5891 - fix NULL deref and obd_dev leak on setup error (3312)
5892 - replace some LBUG about llog ops with error handling (3841)
5893 - don't match INVALID dentries from d_lookup and spin (3784)
5894 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5895 - fix invalid assertion in ptlrpc_set_wait (3880)
5897 - add libwrap support for the TCP acceptor (3996)
5898 - add /proc/sys/portals/routes for non-root route listing (3994)
5899 - allow setting MDS UUID in .xml (2580)
5900 - print the stack of a process that LBUGs (4228)
5902 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5905 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5906 - make sure to unset replay flag from failed open requests (3440)
5907 - if default stripe count is 0, use OST count for inode size (3636)
5908 - update parent mtime/ctime on client for create/unlink (2611)
5909 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5910 - free recovery state on server during a forced cleanup (3571)
5911 - unregister_reply for resent reqs (3063)
5912 - loop back devices mounting and status check on 2.6 (3563)
5913 - fix resource-creation race that can provoke i_size == 0 (3513)
5914 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5915 - more debugging for page-accounting assertion (3746)
5916 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5917 - avoid deadlock after precreation failure (3758)
5918 - fix race and lock order deadlock in orphan handling (3450, 3750)
5919 - add validity checks when grabbing inodes from l_ast_data (3599)
5921 - add /proc/.../recovery_status to obdfilter (3428)
5922 - lightweight CDEBUG infrastructure, debug daemon (3668)
5923 - change default OSC RPC parameters to be better on small clusters
5924 - turn off OST read cache for files smaller than 32MB
5925 - install man pages and include them in rpms (3100)
5926 - add new init script for (un)mounting lustre filesystems (2593)
5927 - run chkconfig in %post for init scripts (3701)
5928 - drop scimac NAL (unmaintained)
5930 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5933 - clean kiobufs before and after use (3485)
5934 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5935 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5936 - change init script to fail more clearly if not run as root (1528)
5937 - allow clients to reconnect during replay (1742)
5938 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5939 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5940 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5941 - fix cond_resched (3554)
5942 - only evict unfinished clients after recovery (3515)
5943 - allow bulk resend, prevent data loss (3570)
5944 - dynamic ptlrpc request buffer allocation (2102)
5945 - don't allow unlinking open directory if it isn't empty (2904)
5946 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5947 - remove extraneous obd dereference causing LASSERT failure (3334)
5948 - don't use get_cycles() when creating temp. files on the mds (3156)
5949 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5950 - handle EEXIST for set-stripe, set proper directory name (3336)
5952 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5953 - fix ksocknal_fmb_callback() error messages (2918)
5955 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5958 - don't copy lvb into (possibly NULL) reply on error (2983)
5959 - don't deref dentry after dput, don't free lvb on error (2922)
5960 - use the kms to determine writeback rpc length (2947)
5961 - increment oti_logcookies when osc is inactive (2948)
5962 - update client's i_blocks count via lvb messages (2543)
5963 - handle intent open/close of special files properly (1557)
5964 - mount MDS with errors=remount-ro, like obdfilter (2009)
5965 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5966 - don't use cancelling-locks' kms values (2947)
5967 - use highest lock extent for kms, not last one (2925)
5968 - don't dereference ERR_PTR() dentry in error handling path (3107)
5969 - fix thread race in portals_debug_dumplog() (3122)
5970 - create lprocfs device entries at setup instead of at attach (1519)
5971 - common AST error handler, don't evict client on completion race (3145)
5972 - zero nameidata in detach_mnt in 2.6 (3118)
5973 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5974 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5975 - initialize RPC timeout timer earlier for 2.6 (3219)
5976 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5977 - print nal/nid for unknown nid (3258)
5978 - additional checks for oscc recovery before doing precreate (3284)
5979 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5980 - don't crash in mdc_close for bad permissions on open (3285)
5981 - zero i_rdev for non-device files (3147)
5982 - clear page->private before handing to FS, better assertion (3119)
5983 - tune the read pipeline (3236)
5984 - fix incorrect decref of invalidated dentry (2350)
5985 - provide read-ahead stats and refine rpc in flight stats (3328)
5986 - don't hold journal transaction open across create RPC (3313)
5987 - update atime on MDS at close time (3265)
5988 - close LDAP connection when recovering to avoid server load (3315)
5989 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5990 - don't leak open file on MDS after open resend (3325)
5991 - serialize filter_precreate and filter_destroy_precreated (3329)
5992 - loop device shouldn't call sync_dev() for nul device (3092)
5993 - clear page cache after eviction (2766)
5994 - resynchronize MDS->OST in background (2824)
5995 - refuse to mount the same filesystem twice on same mountpoint (3394)
5996 - allow llmount to create routes for mounting behind routers (3320)
5997 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5998 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5999 - fix rare problem with rename on htree directories (3417)
6001 - allow default OST striping configuration per directory (1414)
6002 - fix compilation for qswnal for 2.6 kernels (3125)
6003 - increase maximum number of MDS request buffers for large systems
6004 - change liblustreapi to be useful for external progs like lfsck (3098)
6005 - increase local configuration timeout for slow disks (3353)
6006 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6008 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6012 - don't overwrite extent policy data in reply if lock was blocked (2901)
6013 - drop filter export grants atomically with removal from device (2663)
6014 - del obd_self_export from work_list in class_disconnect_exports (2908)
6015 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6016 - swab reply message in mdc_close, other PPC fixes (2464)
6017 - fix destroying of named logs (2325)
6018 - overwrite old logs when running lconf --write_conf (2264)
6019 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6020 - fix race in target_handle_connect (2898)
6021 - mds_reint_create() should take same inode create lock (2926)
6022 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6023 - don't close files for self_export to avoid uninitialized obd (2936)
6024 - allow MDS with the same name as client node (2939)
6025 - hold dentry reference for closed log files for unlink (2325)
6026 - reserve space for all logs during transactions (2059)
6027 - don't evict page beyond end of stripe extent (2925)
6028 - don't oops on a deleted current working directory (2399)
6029 - handle hard links to targets without a parent properly (2517)
6030 - don't dereference NULL lock when racing during eviction (2867)
6031 - don't grow lock extents when lots of conflicting locks (2919)
6033 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6036 - account for cache space usage on clients to avoid data loss (974)
6037 - lfsck support in lustre kernel code (2349)
6038 - reduce journal credits needed for BRW writes (2370)
6039 - orphan handling to avoid losing space on client/server crashes
6040 - ptlrpcd can be blocked, stopping ALL progress (2477)
6041 - use lock value blocks to assist in proper KMS, faster stat (1021)
6042 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6043 - recovery for initial connections (2355)
6044 - fixes for mds_cleanup_orphans (1934)
6045 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6046 - block all file creations until orphan recovery completes (1901)
6047 - client remove rq_connection from request struct (2423)
6048 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6049 - recovery timer race (2670)
6050 - mdc_close recovey bug (2532)
6051 - ptlrpc cleanup bug (2710)
6052 - mds timeout on local locks (2588)
6053 - namespace lock held during RPCs (2431)
6054 - handle interrupted sync write properly (2503)
6055 - don't try to handle a message that hasn't been replied to (2699)
6056 - client assert failure during cleanup after abort recovery (2701)
6057 - leak mdc device after failed mount (2712)
6058 - ptlrpc_check_set allows timedout requests to complete (2714)
6059 - wait for inflight reqs when ptlrpcd finishes (2710)
6060 - make sure unregistered services are removed from the srv_list
6061 - reset bulk XID's when resending them (caught by 1138 test)
6062 - unregister_bulk after timeout
6063 - fix lconf error (2694)
6064 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6065 - readahead locks pages, leaves pending causing memory pressure (2673)
6066 - increase OST request buffers to 4096 on large machines (2729)
6067 - fix up permission of existing directories in simple_mkdir (2661)
6068 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6069 - don't assign transno to errored transactions (2742)
6070 - don't delete objects on OST if given a bogus objid from MDS (2751)
6071 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6072 - if rq_no_resend, then timeout request after recovery (2432)
6073 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6074 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6075 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6076 - MDS recovery shouldn't delete objects that it has given out (2730)
6077 - if enqueue arrives after completion, don't clobber LVB (2819)
6078 - don't unlock pages twice when trigger_group_io returns error (2814)
6079 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6080 - don't write pages to disk if there was an error (1450)
6081 - don't ping imports that have recovery disabled (2676)
6082 - take buffered bytes into account when balancing socknal conn (2817)
6083 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6084 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6085 - remove little-endian swabbing of llog records (1987)
6086 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6087 - retry reposting request buffers if they fail (1191)
6088 - grow extent at grant time to avoid granting a revoked lock (2809)
6089 - lock revoke doesn't evict page if covered by a second lock (2765)
6090 - disable VM readahead to avoid reading outside lock extents (2805)
6092 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6093 - updated kernel patches for hp-2.4.20 kernel (2681)
6095 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6098 - fix truncated write corruption (2366)
6099 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6101 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6102 - ensure len(uuid) < 37 in lmc (1171)
6103 - fix ia64 OOPS in llog_test (2255)
6104 - zero end of page at obdfilter for partial page writes (2648)
6105 - don't leave stale dentries around after renames (bug 2428)
6106 - fix timeouts when evicting a client with a single lock held (2642)
6107 - set deadline for the initial HELLO message to drain (2634)
6108 - print out dotted-quad IP addresses in the socknal (2302)
6110 - additional debugging for MDS client eviction problem (2443)
6111 - fix mkfsoptions support for osts (2603, 2604)
6113 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6116 - add series for the vanilla 2.6.0 kernel
6117 - add series for the vanilla 2.4.24 kernel
6118 - add series for a cray x86/64 UL kernel drop
6119 - fix xattr patches for the vanilla 2.4.19 series
6121 - generate true UUIDs in lmc (1171)
6122 - have portals stack dumping break in UML (2466)
6123 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6124 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6125 - wake callback waiting threads on client eviction (2460)
6126 - Add --ptldebug and --subsystem to lmc (1719)
6127 - update assertion to allow safe interrupt allocation
6128 - set rq_no_resend for cancel requests (2432)
6129 - recalculate ptlrpcd timeout after resend (2494)
6130 - call vfs_rmdir when removing pending directories (2368)
6131 - fix renaming a file to itself (2429)
6132 - lmc creates a default one-stripe lov (2454)
6133 - expand procfs space to handle large clusters (2326)
6134 - increase UML stack to avoid overflow
6135 - update lconf's list of debug and subsystem masks
6136 - fix lfs find --obd (2510)
6137 - /proc tunable for disabling filter read caching (2591)
6138 - stop rpm packages from altering slapd.conf (2301)
6139 - disable nagle in the socknal under 0conf (2578)
6140 - choose mds inode size based on stripe count (2572)
6141 - fix kernel-source rpm problems (2516)
6143 - add --disable-doc to avoid pdf generation (2421)
6144 - update documentation, tests, type-os, comments
6145 - avoid format warnings on ia64
6146 - remove the TOE NAL
6147 - tiny code cleanups by removing unused fields
6149 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6152 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6153 - remove the most blatant lies from BUILDING file (2371)
6154 - change default debug level to reasonable production setting
6155 - reduce client side cache size to reduce cache flush time
6156 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6157 - make TCP zerocopy and pinger support enabled by default (2476)
6158 - sync writes completed after process exits caused crashes (2319)
6159 - maintain correct mount count on the MDS (2356)
6160 - backout 1557, because 2316 wasn't really fixed
6161 - better file I/O statistics gathering in /proc
6162 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6163 - another kernel patch to fix zero-copy TCP function export
6164 - don't take duplicate lock when processing re-sent getattr (2420)
6165 - lctl uses obd_self_export instead of creating new conn (2353)
6166 - MDS/OST recovery case which requires object creation asserted (2425)
6167 - move lfs from /usr/sbin to /usr/bin in packages
6168 - fix race between mds_client_add and mds_client_free (2417)
6169 - use kmalloc instead of slabs in portals (2430)
6170 - don't create duplicate records when a failover MDS is present (2442)
6171 - remove unnecessary mount age check (2332)
6172 - don't remove directory inodes from locks prematurely (2451)
6173 - don't break if MDS service name is the same as hostname (2103)
6174 - fix races in client write RPC generation when cache full (2482)
6176 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6179 - remove now-unused request->rq_obd (278)
6180 - if an allocation fails, print out how much memory we've used (1933)
6181 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6182 - add the "configurable stack size" patch to most series files (1256)
6183 - ability to write large log records, for 100+ OST configs (2306)
6184 - fix NULL deref when filter_prep fails (2314)
6185 - fix operator precedence error in filter_sync
6186 - dynamic allocation of socknal TX descriptors (2315)
6187 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6188 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6189 - ensure that CWARN messages in Portals always get to the syslog
6190 - __init/__exit are not for prototype decls (ldlm_init/exit)
6191 - x86-64 compile warning fixes
6192 - fix gateway LMC keyword conflict (2318)
6193 - fix MDS lock inversions in getattr/reint paths (1844)
6194 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6195 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6196 - DLM race condition prevented some lock evictions (2328)
6197 - ENOMEM detection and retry on socknal sends (2230)
6198 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6199 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6200 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6201 - better error messages when a client is rejected during recovery (1505)
6202 - avoid cancelling locks which were never granted, after failure (2330)
6203 - fix i_sem/journal inversion in mds_client_add (2333)
6204 - fix truncate/getattr lock cycle deadlock (2334)
6205 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6206 - fix two rare exit paths which could leak an l_lock() ref (2321)
6207 - fencepost error in MDS/OST orphan recovery (2226)
6208 - make log record alignment 8 bytes (1988)
6209 - lstripe now fails when requested offset > ost_count (2237)
6210 - ensure that all kernel series have a complete list.h (1607)
6211 - fix crashes in special-file operations (2316)
6212 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6214 - allow configurable automake binary, for testing new versions
6215 - small update to the lfs documentation
6217 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6219 * fix negative export reference count in fsfilt_sync (2312)
6221 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6222 * release candidate 0.9.1
6224 - orphans are moved into the PENDING directory for possible recovery
6225 - replayed opens now open by fid for orphan/rename safety (1042)
6226 - last close of an orphan inode generates a transno (683)
6227 - chdir() and mount() now pin the directory entry (1020)
6228 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6229 - discard very old requests without processing them (1502)
6230 - remove some common, well-understood CERRORs (1505)
6231 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6232 - clear "grant" flags in OST replies until OST grant code lands (1644)
6233 - fix read performance by not clobbering i_blksize on client (1598)
6234 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6235 - make LRU size a /proc tunable, clears locks when reduced (707)
6236 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6237 - server locks take references on exports now (1558)
6238 - build fixes for 2.4.20-rh trees (1663)
6239 - return an error from lov_create if all OSCs are inactive (1751)
6240 - fix import levels when a reconnect happens without a timeout (1597)
6241 - exit early from mds_open if we get a lookup error (1749)
6242 - partial page read at EOF wouldn't wait for disk before sending (1642)
6243 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6244 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6245 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6246 - write/truncate lock inversion (1639)
6247 - Don't auto-load obdclass, portals modules during cleanup (1495)
6248 - fix timestamps from jumping to "now" (1763)
6249 - extra journal assertions (1648)
6250 - add an extra multiunlink test (1771)
6251 - fix read_record/write_record API (1776)
6252 - fix leak of offset_extent, possible incorrect i_size later (1772)
6253 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6254 - Debugging for the kqswnal_get_idle_tx problems (1820)
6255 - Allow recovery to be attempted multiple times (1536)
6256 - Write out MDS last_rcvd file after it is first created (1600)
6257 - Fix tx_descriptor leak in failed transmit situations (1827)
6258 - ext3 journaling fixes for assertion failure after IO error (1871)
6259 - class_export_put() on freed export after completion AST error (1896)
6260 - Fix revalidate looping in VFS (1322)
6261 - Don't access a freed export during MDS_REINT timeout (1521)
6262 - Add open-unlink recovery support on the MDS (1673,1764)
6263 - Return an error if no MDS data was read from last_rcvd (1946)
6264 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6265 - Don't setup a disk device that doesn't match exported UUID (317)
6266 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6267 - avoid committing NULL handle in force close
6268 - local.sh is now a one-stripe LOV configuration
6269 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6270 - fix ext3 htree duplicate directory entry corruption (1516)
6271 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6272 - update to new LOV EA format (2097)
6273 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6275 2003-06-15 Phil Schwan <phil@clusterfs.com>
6278 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6279 - per-import/export recovery handling (958, 931, 959)
6280 - multiple last-rcvd slots, for serving multiple FSes (949)
6281 - connections are again shared between multiple imp/exports (963, 964)
6282 - "umount -f" would hang if any requests needed to be sent (393, 978)
6283 - avoid pinning large req buffer by copying for queued messages (989)
6284 - add "uuid" to "lctl device" command to help upcalls (991)
6285 - "open" RPCs with transnos would confuse recovery counters (1037)
6286 - do proper endian conversion of all wire messages (288, 340, 891)
6287 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6288 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6289 - fix ptlrpc_connection leak in target_handle_connect (1174)
6290 - fix import refcounting bug in OST and MDS cleanup (1134)
6291 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6292 - fix very unlikely obd_types race condition (501)
6293 - remove osc_open hack for echo_client (1187)
6294 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6295 - a failure in read_inode2 leads to deadlock (1139)
6296 - cancel ack-locks as soon as transaction is committed (1072)
6297 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6298 - make sure to commitrw after any preprw to avoid deadlock (1162)
6299 - failing to execute a file in a lustre FS would lock inode (1203)
6300 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6301 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6302 - obd_uuid2tgt would walk off the end of the list (1255)
6303 - on IA64 the osc would give portals incorrect bulk size (1258)
6304 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6305 - fix lock inversion caused by new llite matching code (1282)
6306 - limit the number of dirty pages on a client to 10MB (1286)
6307 - timed out locks were not being corrected cancelled (1289)
6308 - fix O_DIRECT above 4GB on IA-32 (1292)
6309 * major user-visible changes
6310 - fail out/fail over policy now controlled by the upcall (993)
6312 - add OBD_PING to check server availability and failure (954)
6313 - lustre messages are now sent in sending host order (288, 340, 891)
6314 - add eadatalen to MDS getattr reply (340)
6315 - OST read replies may contain second buffer, with per-page status (593)
6317 2003-03-11 Phil Schwan <phil@clusterfs.com>
6320 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6321 - failed open()s could cause deadlock; fixed (867, 869)
6322 - stop cancelling OST locks when files are closed (481)
6323 - overlapping XID spaces caused network corruption (851, 853)
6324 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6325 - setattr_raw added, to do single-RPC, server-side setattrs
6326 - lmc/lconf syntax change for OST UUIDs
6327 - fix crashy race condition between ptlrpc_free_req and osc_close
6328 - don't use request in mdc_enqueue if we hit a timeout (889)
6329 - don't set the inode i_size for regular files from the MDS (896)
6330 - handle out of order completion AST (842)
6331 - don't LBUG if a lock request times out after receiving AST (913)
6332 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6333 - if a bad lock AST arrives, send an error instead of dropping entirely
6334 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6335 - fix leak in bulk IO when only partially completed (899, 900, 926)
6336 - fix O_DIRECT for ia64 (55)
6337 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6338 - C-z after timeout could hang a process forever; fixed (977)
6340 - client-side I/O cache (678, 924, 929, 941, 970)
6342 - READPAGE and SETATTRs which don't take server-side locks get
6345 2003-02-11 Phil Schwan <phil@clusterfs.com>
6348 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6349 - Fix fsfilt_extN_readpage() to read a full page of directory
6350 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6351 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6352 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6353 - Most intent operations were removed, in favour of a new RPC mode
6354 that does a single RPC to the server and bypasses most of the VFS
6355 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6356 - Aggressively cancel local locks on DLM servers
6357 - mds_reint_unlink sends EA to the client if it's the last nlink.
6358 client uses that EA to unlink OST objects.
6359 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6360 - recursive symlinks were fixed (439)
6361 - fixed NULL deref in DEBUG_REQ
6362 - filter_update_lastobjid no longer calls sync, which annoyed extN
6363 - fixed multi-client small-writes to a single file problem (445)
6364 - fixed mtime updates during file writes (607)
6365 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6366 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6367 - fixed Linux OST size reporting problem (444, 656)
6368 - OST now updates object mtime with writes or setattr (607, 619)
6369 - client verifies file size before zeroing page past EOF (445)
6370 - OST now writes last allocated objid to disk with allocation (108)
6371 - LOV on echo now works (409)
6373 - mds_reint_unlink sends a new buffer, with the EA included. this
6374 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6375 set if a regular file was being unlinked, and it was the last link
6376 - use PtlGet from the target for bulk writes (315)
6377 - OST now updates object mtime with writes or setattr (607, 619)
6378 - LDLM now has a grant-time callback to revalidate locked items, if
6380 - Many MDS operations were reorganized to combat race conditions
6382 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6383 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6385 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6388 - Fully reactivate OST imports after reconnection (512, others)
6389 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6390 - More graceful error handling for truncating on dead OST (515)
6391 - Don't error out unless we're actually accessing dead stripes (474)
6392 - Fix garbage sizes when stripes are missing (410)
6393 - LRU counters were broken, causing constant lock purge (433, 432)
6394 - garbage on read from stripes with failed OSTs (441)
6395 - mark OSCs as active before reconnecting during recovery (438)
6396 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6397 - lfind did not preserve OST order in output (443)
6398 - symlinks cause hung clients, incorrect data (439)
6399 - stop dereferencing request after dropping refcount (457)
6400 - don't LASSERT(spin_is_locked) on non-SMP (455)
6401 - fixes for many rename() bugs
6402 - fstat didn't correctly synchronize attributes (399)
6403 - server must handle lock cancellation during blocking AST prep (487)
6404 - bulk descriptors were free()d too soon (511)
6405 - fix paths in lconf, which would load incorrect modules (451, 507)
6406 - fix confusing lconf 'host not found' error message (386)
6407 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6408 - fix race condition in mdc_blocking_ast() for inode access (526)
6409 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6410 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6411 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6412 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6413 - fix lstripe in several ways (kernel side) (527)
6414 - fix request leak in ldlm_cli_enqueue (262)
6415 - incorrect OSC was marked inactive after OST failure
6416 - call mds_fs_cleanup before unmounting filesystem (524)
6417 - fix races between taking ns_lock and ldlm_lock_change_resource
6418 - fix races updating LOV export open file list
6419 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6420 - fix recovery NULL deref in ldlm_cli_cancel_unused
6421 - fix some DLM races by using new hash table for lock handles (419)
6422 - permit the client to specify desired inodes, at replay
6423 - duplicate requests when we queue them for replay reintegration
6424 - fix last_rcvd offset calculation
6425 - sync after each recovered transaction, so we always make progress
6426 - never, not always, ERESTART requests without transnos
6427 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6429 - skip replay if the MDS says that the client is already connected
6430 - don't check for a recovery-enabled export to match lctl's UUID
6431 - don't INC_USE_COUNT for phantom exports
6432 - don't crash when cleaning up phantom exports (567)
6433 - don't double-finish or set replay data for errored mdc_open requests
6434 - abort requests when they time out, so we don't get old replies
6435 - send/receive replies for AST messages again
6436 - if the client says that it doesn't have the lock, cancel it on the
6438 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6439 mark it as destroyed, it will all work out when decref is called
6440 - fix module use counts (22, 581)
6442 - ASTs now expect a reply (server cancels lock on error reply)
6444 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6447 - fix many simultaneous client startup (392)
6448 - fix dentry->d_it clobbering
6449 - credentials weren't being shipped for readdir/getattr operations
6450 - remove invalid assertions triggered during some concurrent MD
6452 - proper Lustre versions added (336, 389)
6453 - fix memory leak for create error case (398)
6454 - fix LOV locking bug that would get cli/srv out of sync
6455 - fix echo client over LOV (409)
6456 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6457 - fix double-O_EXCL intent crash (424)
6458 - avoid sending multiple lock CANCELs (352)
6460 - MDS can do multi-client recovery (modulo bugs in new code)
6462 - many updates, edits, cleanups
6464 2002-11-18 Phil Schwan <phil@clusterfs.com>
6467 - fix null d_it dereference (346)
6468 - fix full OST/dbench hang (333)
6469 - fix permission problem with file removal (286)
6470 - fix removal of OSCs from LOV when they fail
6471 - fix NULL deref during bulk timeout (214)
6472 - fix problems related to multiple filesystems on one MDS (241)
6473 - fixed serious subtle metadata locking bugs
6474 - free locks on clients when inodes are removed due to memory
6476 - fix inode pointer in lock data (285)
6477 - partial support for multiple MDS on a single host (241)
6478 - data locks weren't cancelled at clear_inode time (290, 311)
6479 - intent locks could lead to unbounded lock growth (205)
6480 - added a maximum lock count, an LRU list, and a flusher
6481 - fix multiple rename (365)
6482 - properly abstracted the echo client
6483 - OSC locked 1 byte too many; fixed
6484 - rewrote brw callback code:
6485 - fixed recovery bugs related to LOVs (306)
6486 - fixed too-many-pages-in-one-write crash (191)
6487 - fixed (again) crash in sync_io_timeout (214)
6488 - probably fixed callback-related race (385)
6490 - Add capability to MDS protocol
6491 - LDLM cancellations and callbacks on different portals
6493 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6496 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6497 - defer open object destruction to close time (601981, 138)
6498 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6499 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6500 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6501 - don't call obd_setattr in ll_file_release for destroyed objects
6502 * protocol change to lustre_msg: move |version| and add |flags|
6503 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6504 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6505 * added replay of create, unlink, link and rename operations during
6506 MDS failover; recovery should be much more robust now
6507 * remove failed OSCs from LOVs (only lov_create uses this so far)
6508 * the lustre-HOWTO was brought (more) up to date (582544)
6510 2002-10-23 Phil Schwan <phil@clusterfs.com>
6513 - in-use dentries weren't being reused properly (617851)
6514 - prevent multiple LDLM setup (599178)
6515 - fix LOV size calculations for truncate (617853)
6516 - fix client handling of MDS intent errors (POSIX)
6517 - fix permission bug in lovstripe.c test (624321)
6518 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6519 - truncate past end of file could corrupt data
6520 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6521 - a race in recovery could return ETIMEDOUT to apps (623947)
6522 - building outside the source directory was fixed
6523 * the lustre-HOWTO was brought (more) up to date (582544)
6524 * major progress was made on recovery functionality
6526 2002-10-10 Phil Schwan <phil@clusterfs.com>
6529 - recovery deadlock fix
6530 - rm -rf causes LBUG fix (617817)
6531 - file open by multiple tasks fix (618962)
6532 - directory permissions bugs (602707 and 620007)
6533 - journal_stop fixed with locking (611313)
6534 - O_APPEND failures resolved (618273, perhaps 614459)
6535 - lconf PATH fix (619770)
6536 - IA64 build fix (621450)
6537 - RPC buffer sizes scale with amount of memory
6539 2002-10-01 Phil Schwan <phil@clusterfs.com>
6542 - locks would be cancelled without throwing away data pages,
6543 resulting in inconsistent data (605627)
6544 - inode attributes were not always being refreshed (605627, 612449)
6545 - lconf now continues to cleanup after lctl reports an error
6546 - MDS now enforces user permissions (602707)
6547 - lprocfs cleanup fixed, but not yet enabled (614157)
6548 - fixed infinite server hang, should a client not respond to an AST
6549 - avoid going into recovery if user calls readlink() with a buffer
6550 that's too small (613941)
6551 - AST RPCs no longer require replies (614867) -- this may be changed
6552 - don't crash server if client sends an IOV that's too big (611336)
6553 - fixed lock conversion deadlock (611892)
6554 - fixed the following of symlinks (614622)
6555 * recovery: the server can remove locks from a client that dies, other
6556 clients can make progress
6557 * more extN patch fixes
6558 * compile-time configurable ptlrpc buffer allocations
6560 - collaborative read cache document
6561 - Lustre Lite Performance CDR document-in-progress
6563 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6566 - fix typo in patch-2.4.18
6568 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6571 - clear ptlrpc request each time in handle_incoming_request()
6572 - unlink of files now destroys the object on the OST
6574 2002-09-19 Peter Braam <braam@clusterfs.com>
6576 * add hard link support
6577 * change obdfile creation method
6578 * kernel patch changed
6580 2002-09-19 Peter Braam <braam@clusterfs.com>
6583 - stack overflow bug in extN fixed
6585 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6587 * documentation updates
6588 - add man pages for config tools
6589 - update tests/README to describe testing with new config tools
6590 - finish metadata API descriptions
6591 * bug fixes and cleanups
6592 - statfs workaround for 16TB limit
6593 - LOV stripe allocation improved, can stripe on subset of OSTs
6594 - LOV file size/IO offset was wrong for files > 4GB in size
6595 - object EA data was being dropped, caused files to be unreadable
6596 - memory overflow with non-LOV OST caused memory corruption
6597 - fixed regression tests to work with new config tools, obdfilter
6598 - fixed bug when directory size became larger than 1 block
6599 - fixed bug (for single client case) when PWD was deleted
6600 - invalidate local directory pages when doing intent-based ops
6601 - avoid LDLM oops when lock callback contained bad data
6603 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6605 * documentation updates
6606 * bug fixes and cleanups
6607 - configuration tools
6610 - 64-bit compile warnings
6611 - 64-bit internal statfs data
6613 * test_brw on persistent OST devices
6615 * lprocfs (disabled)
6617 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6619 * documentation updates
6620 * bug fixes and cleanups
6621 * configuration tools
6623 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6626 * many small fixes to 0_5_4
6627 * io/network handling
6628 * thinkos in MDS operations
6630 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6633 * crucial basic fixes to 0.5.3
6634 * IOR, Iozone work over Elan
6637 2002-08-07 Phil Schwan <phil@clusterfs.com>
6638 * version 0_5_3, our first alpha
6639 * we use the new Portals iovs
6640 * documentation updates
6641 * bug fixes and cleanups
6642 * small changes in the DLM wire protocol
6644 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6645 * version 0_5_1 with some initial stability,
6646 * locking on MD and file I/O.
6647 * documentation updates
6648 * several bug fixes since 0.5.0
6649 * small changes in wire protocol
6651 2002-07-18 Phil Schwan <phil@clusterfs.com>
6653 * delivered as Lustre Light Alpha
6654 * fixed a crash after handling invalid MDS requests
6655 * fixed directory pages for architectures with non-4k pages sizes
6657 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6658 * release version v0_4_4
6659 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6661 * Fixes a number of interruption problems with OST operations.
6662 * Update documentation for portals header changes
6663 * Move all wire protocol structs/defines to lustre_idl.h
6664 * Fixes symlink length bug.
6665 * Add tcpdump to repository.
6667 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6668 * release version v0_4_3
6669 * Fixes statfs for inodes on extN.
6670 * Fixes bug in runtests which would delete /etc/hosts.
6671 * Use 64-bit object IDs wherever possible (not into VFS though)
6672 Remove ost_get_info, which is unused by lustre, and out of date.
6674 2002-07-03 Peter Braam <braam@clusterfs.com>
6675 * release version v0_4_2 Fixes a lookup error (type not passed)
6676 * move forward to head of Portals
6677 * move forward to latest Lustre kernel
6679 2002-06-25 Peter Braam <braam@clusterfs.com>
6680 * release version v0_4_1. Hopefully stable on single node use.