3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.162-94.69 (SLES12 SP3)
11 4.4.0-131 (Ubuntu 16.04)
12 4.15.0-32 (Ubuntu 18.04)
13 vanilla linux 4.6.7 (ZFS only)
14 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
15 * Client known to build on unpatched kernels (others may also work):
16 2.6.32-696.18.7.el6 (RHEL6.9)
17 2.6.32-754.6.3.el6 (RHEL6.10)
18 3.10.0-862.14.4.el7 (RHEL7.5)
19 3.10.0-957.el7 (RHEL7.6)
20 4.14.0-49.13.1.el7a (RHEL7.5)
21 4.14.0-115.2.2.el7a (RHEL7.6)
22 4.4.120-92.70 (SLES12 SP2)
23 4.4.162-94.69 (SLES12 SP3)
24 4.4.0-131 (Ubuntu 16.04)
25 4.15.0-32 (Ubuntu 18.04)
26 vanilla linux 4.15.0 (ZFS only)
27 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
28 * Recommended ZFS / SPL version: 0.7.9
29 * Tested with ZFS / SPL version: 0.7.9
30 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
31 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
32 stack overflow. For more information, please refer to bugzilla 17630.
33 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
34 the re-exporting nodes to be mounted with "32bitapi" mount option
35 * Note for Ubuntu 14.04.5:
36 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
37 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
39 --------------------------------------------------------------------------------
42 2018-04-03 Intel Corporation
44 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on patched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.26.1.el6 (RHEL6.7)
50 2.6.32-642.15.1.el6 (RHEL6.8)
51 2.6.32-696.18.7.el6 (RHEL6.9)
52 3.10.0-693.21.1.el7 (RHEL7.4)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-107 (SLES11 SP4)
55 3.12.74-60.64.40 (SLES12 SP1)
56 4.4.114-92.64 (SLES12 SP2)
57 4.4.114-94.11 (SLES12 SP3)
58 3.13.0-101 (Ubuntu 14.04, ZFS only)
59 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
60 4.4.0-116 (Ubuntu 16.04)
61 vanilla linux 4.6.7 (ZFS only)
62 * Client known to build on unpatched kernels:
63 2.6.32-431.29.2.el6 (RHEL6.5)
64 2.6.32-504.30.3.el6 (RHEL6.6)
65 2.6.32-573.26.1.el6 (RHEL6.7)
66 2.6.32-642.15.1.el6 (RHEL6.8)
67 2.6.32-696.18.7.el6 (RHEL6.9)
68 3.10.0-693.21.1.el7 (RHEL7.4)
69 3.0.101-0.47.71 (SLES11 SP3)
70 3.0.101-107 (SLES11 SP4)
71 3.12.74-60.64.40 (SLES12 SP1)
72 4.4.114-92.64 (SLES12 SP2)
73 4.4.114-94.11 (SLES12 SP3)
74 3.13.0-101 (Ubuntu 14.04)
75 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
76 4.4.0-116 (Ubuntu 16.04)
78 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
79 * Recommended ZFS / SPL version: 0.7.6
80 * Tested with ZFS / SPL version: 0.7.6
81 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
82 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
83 stack overflow. For more information, please refer to bugzilla 17630.
84 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
85 the re-exporting nodes to be mounted with "32bitapi" mount option
86 * Note for Ubuntu 14.04.5:
87 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
88 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
90 --------------------------------------------------------------------------------
92 2017-07-05 Intel Corporation
94 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
95 for currently supported client and server kernel versions.
96 * Server known to build on patched kernels:
97 2.6.32-431.29.2.el6 (RHEL6.5)
98 2.6.32-504.30.3.el6 (RHEL6.6)
99 2.6.32-573.26.1.el6 (RHEL6.7)
100 2.6.32-642.15.1.el6 (RHEL6.8)
101 2.6.32-696.3.1.el6 (RHEL6.9)
102 3.10.0-514.21.1.el7 (RHEL7.3)
103 3.0.101-0.47.71 (SLES11 SP3)
104 3.0.101-100 (SLES11 SP4)
105 3.12.74-60.64.40 (SLES12 SP1)
106 4.4.59-92.17 (SLES12 SP2)
107 vanilla linux 4.6.7 (ZFS only)
108 * Client known to build on unpatched kernels:
109 2.6.32-431.29.2.el6 (RHEL6.5)
110 2.6.32-504.30.3.el6 (RHEL6.6)
111 2.6.32-573.26.1.el6 (RHEL6.7)
112 2.6.32-642.15.1.el6 (RHEL6.8)
113 2.6.32-696.3.1.el6 (RHEL6.9)
114 3.10.0-514.21.1.el7 (RHEL7.3)
115 3.0.101-0.47.71 (SLES11 SP3)
116 3.0.101-100 (SLES11 SP4)
117 3.12.74-60.64.40 (SLES12 SP1)
118 4.4.59-92.17 (SLES12 SP2)
120 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
121 * Recommended ZFS / SPL version: 0.6.5.9
122 * Tested with ZFS / SPL version: 0.6.5.9
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
129 --------------------------------------------------------------------------------
131 12-07-2016 Intel Corporation
133 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
134 for currently supported client and server kernel versions.
135 * Server known to build on patched kernels:
136 2.6.32-431.29.2.el6 (RHEL6.5)
137 2.6.32-504.30.3.el6 (RHEL6.6)
138 2.6.32-573.26.1.el6 (RHEL6.7)
139 2.6.32-642.11.1.el6 (RHEL6.8)
140 3.10.0-514.el7 (RHEL7.3)
141 3.0.101-0.47.71 (SLES11 SP3)
142 3.0.101-80 (SLES11 SP4)
143 3.12.62-60.64.8 (SLES12 SP1)
144 vanilla linux 4.5.7 (ZFS only)
145 * Client known to build on unpatched 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)
155 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
156 * Recommended ZFS / SPL version: 0.6.5.8
157 * Tested with ZFS / SPL version: 0.6.5.8
158 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
159 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
160 stack overflow. For more information, please refer to bugzilla 17630.
161 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
162 the re-exporting nodes to be mounted with "32bitapi" mount option
164 --------------------------------------------------------------------------------
166 02-29-2016 Intel Corporation
168 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
169 for currently supported client and server kernel versions.
170 * Server known to build on patched kernels:
171 2.6.32-431.29.2.el6 (RHEL6.5)
172 2.6.32-504.30.3.el6 (RHEL6.6)
173 2.6.32-573.12.1.el6 (RHEL6.7)
174 3.10.0-327.3.1.el7 (RHEL7.2)
175 3.0.101-0.47.71 (SLES11 SP3)
176 3.0.101-68 (SLES11 SP4)
177 vanilla linux 4.2.1 (ZFS only)
178 * Client known to build on unpatched kernels:
179 2.6.32-431.29.2.el6 (RHEL6.5)
180 2.6.32-504.30.3.el6 (RHEL6.6)
181 2.6.32-573.12.1.el6 (RHEL6.7)
182 3.10.0-327.3.1.el7 (RHEL7.2)
183 3.0.101-0.47.71 (SLES11 SP3)
184 3.0.101-68 (SLES11 SP4)
187 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
188 * Recommended ZFS / SPL version: 0.6.4.2
189 * Tested with ZFS / SPL version: 0.6.4.2
190 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
191 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
192 stack overflow. For more information, please refer to bugzilla 17630.
193 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
194 the re-exporting nodes to be mounted with "32bitapi" mount option
196 --------------------------------------------------------------------------------
198 03-10-2015 Intel Corporation
200 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
201 for currently supported client and server kernel versions.
202 * Server known to build on patched kernels:
203 2.6.32-431.29.2.el6 (RHEL6.5)
204 2.6.32-504.8.1.el6 (RHEL6.6)
205 3.0.101-0.46 (SLES11 SP3)
206 * Client known to build on unpatched kernels:
207 2.6.32-431.29.2.el6 (RHEL6.5)
208 2.6.32-504.8.1.el6 (RHEL6.6)
209 3.10.0-123.20.1.el7 (RHEL7)
210 3.0.101-0.46 (SLES11 SP3)
211 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
212 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
213 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
214 stack overflow. For more information, please refer to bugzilla 17630.
215 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
216 the re-exporting nodes to be mounted with "32bitapi" mount option
218 Severity : enhancement
220 Description: control OST-index in IDIF via ROCOMPAT flag.
221 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
222 last_rcvd file. For new formatted OST device, it will be auto set;
223 for the case of upgrading from old OST device, you can enable it
224 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
225 enabled, for new created OST-object, its IDIF-in-LMA will contain
226 the OST-index; for the existing OST-object, the OSD will convert
227 old format IDIF as new format IDIF with OST-index stored in the
228 LMA EA when accessing such OST-object or via OI scrub. Once such
229 flag is enabled, it cannot be reverted back, so the system cannot
230 be downgraded to the orignal incompatible version.
232 --------------------------------------------------------------------------------
234 07-30-2014 Intel Corporation
236 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
237 for currently supported client and server kernel versions.
238 * Server known to build on patched kernels:
239 2.6.32-431.20.3.el6 (RHEL6.5)
240 3.0.101-0.31 (SLES11 SP3)
241 * Client known to build on unpatched kernels:
242 2.6.32-431.20.3.el6 (RHEL6.5)
243 3.0.101-0.31 (SLES11 SP3)
246 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
247 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
248 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
249 stack overflow. For more information, please refer to bugzilla 17630.
250 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
251 the re-exporting nodes to be mounted with "32bitapi" mount option
253 --------------------------------------------------------------------------------
255 10-31-2013 Intel Corporation
257 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
258 * Server builds for kernels:
259 2.6.32-358.18.1.el6 (RHEL6.4)
260 * Client builds for unpatched kernels:
261 2.6.32-358.18.1.el6 (RHEL6.4)
262 3.0.92-0.8 (SLES11 SP3)
264 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
265 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
266 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
267 stack overflow. For more information, please refer to bugzilla 17630.
268 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
269 the re-exporting nodes to be mounted with "32bitapi" mount option
271 --------------------------------------------------------------------------------
273 05-30-2013 Intel Corporation
275 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
276 * Server builds for kernels:
277 2.6.32-358.6.2.el6 (RHEL6.4)
278 * Client builds for unpatched kernels:
279 2.6.32-358.6.2.el6 (RHEL6.4)
280 3.0.74-0.6.6 (SLES11 SP2)
282 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
283 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
284 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
285 stack overflow. For more information, please refer to bugzilla 17630.
286 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
287 the re-exporting nodes to be mounted with "32bitapi" mount option
289 --------------------------------------------------------------------------------
291 9-30-2012 Whamcloud, Inc.
293 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
294 * Server support for kernels:
295 2.6.32-279.5.1.el6 (RHEL6)
296 * Client support for unpatched kernels:
297 2.6.32-279.5.1.el6 (RHEL6)
299 2.6.18-238.19.1.el5 (RHEL5)
300 2.6.16.60-0.69.1 (SLES10)
301 * Recommended e2fsprogs version: 1.41.90.wc4
302 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
303 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
304 stack overflow. For more information, please refer to bugzilla 17630.
305 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
306 the re-exporting nodes to be mounted with "32bitapi" mount option
308 Severity : enhancement
310 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
312 Severity : enhancement
314 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
316 Severity : enhancement
318 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
320 Severity : enhancement
322 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
324 Severity : enhancement
326 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
328 --------------------------------------------------------------------------------
330 04-30-2012 Whamcloud, Inc.
332 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
333 * Server support for kernels:
334 2.6.32-220.4.2.el6 (RHEL6)
335 * Client support for unpatched kernels:
336 2.6.32-220.4.2.el6 (RHEL6)
338 2.6.18-238.19.1.el5 (RHEL5)
339 2.6.16.60-0.69.1 (SLES10)
340 * Recommended e2fsprogs version: 1.41.90.wc4
341 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
342 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
343 stack overflow. For more information, please refer to bugzilla 17630.
344 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
345 the re-exporting nodes to be mounted with "32bitapi" mount option
347 Severity : enhancement
349 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
351 Severity : enhancement
353 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
355 Severity : enhancement
357 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
359 Severity : enhancement
361 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
363 --------------------------------------------------------------------------------
365 09-30-2011 Whamcloud, Inc.
367 * Server support for kernels:
368 2.6.18-238.19.1.el5 (RHEL5)
369 2.6.32-131.6.1.el6 (RHEL6)
370 * Client support for unpatched kernels:
371 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
372 2.6.18-238.19.1.el5 (RHEL5)
373 2.6.32-131.6.1.el6 (RHEL6)
374 2.6.29.4-167.fc11 (FC11)
375 2.6.16.60-0.69.1 (SLES10)
377 * Recommended e2fsprogs version: 1.41.90.wc3
378 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
379 Lustre filesystem with 4K stack may cause a stack overflow. For more
380 information, please refer to bugzilla 17630.
381 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
382 mounted with "32bitapi" mount option
384 Severity : enhancement
385 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
387 Severity : enhancement
388 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
390 Severity : enhancement
391 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
394 Frequency : only with SLES10
396 Description: Use OFED "KMP" provided by Novell
397 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
398 Lustre is now built against this package. That means you need to
399 install the ofed-kmp package from Novell for the patchless client.
400 Note that the ofed-kmp that Novell ships may not exactly match the
401 kernel version but should still be compatible.
403 Severity : enhancement
405 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
406 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
408 --------------------------------------------------------------------------------
410 2010-07-15 Oracle, Inc.
412 * Server support for kernels:
413 2.6.18-164.11.1.el5 (RHEL 5)
414 2.6.18-164.11.1.0.1.el5 (OEL 5)
415 * Client support for unpatched kernels:
416 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
417 2.6.18-164.11.1.el5 (RHEL 5),
418 2.6.18-164.11.1.0.1.el5 (OEL 5)
419 2.6.16.60-0.42.8 (SLES 10),
421 2.6.29.4-167.fc11 (FC11)
422 * Recommended e2fsprogs version: 1.41.10-sun2
424 Severity : enhancement
426 Description: Distribute lfs_migrate script with lustre
427 Details : The lfs_migrate mentioned in the manual is now included in the
428 lustre tree in lustre/scripts and will distributed in lustre RPMs.
430 Severity : enhancement
432 Description: Failover nid list added to import proc file.
434 Severity : enhancement
436 Description: "som_preview" mount option required to use SOM
437 Details : The clients will need to use som_preview mount option
441 Description: Lustre does not do 1MB IOs to HW RAID
442 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
444 Severity : enhancement
446 Description: lfs quota output is not very convenient for awk/sed-parsing
447 Details : Some positions in lfs quota output table could be empty or
448 non-empty which made it hard to parse it with scripts, now
449 a dash is put instead of space where there is not supposed
454 Description: Update to OFED 1.5.1
455 Details : OFED 1.5.1 will be used in environments that don't use
456 vendor supplied OFED.
460 Description: Remove lnet. prefix from lctl params display
461 Details : Also renamed lnet.memused to lnet_memused to
466 Description: Use conf_param -d to remove permanant settings.
470 Description: Add list_param -R
471 Details : lctl list_param now supports recursive mode.
474 Bugzilla : 21829, 22312
475 Description: Fix llobdstat.
476 Details : 1. Now updating the write_bytes counter
477 2. Port script fixes from 1.8.
481 Description: decrease the usage of memory on clients.
482 Details : 1. On clients, recycle dentries and inodes unused.
483 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
488 Description: llobdstat fix and enhancement
489 Details : add a counter to set a limit to how many samples will be returned
490 fix a wildcard in the path to limit to obdfilter stats only
494 Description: lfs check only allowed for root
495 Details : Code cleanups and added a regresion test of non-root use of
500 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
501 Details : Fix hr->hr_index overload.
505 Description: Fix FIEMAP functionality.
509 Description: FC11 patchless client support added
513 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
514 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
518 Description: New acc-sm script and various test-framework enhancments.
522 Description: Previously landed patch was reverted.
526 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
530 Description: Add ldiskfs maxdirsize mount option.
534 Description: Fix mkfs.lustre support for 16TB LUNs.
538 Description: Print mmp_check_interval
539 Details : print mmp_check_interval and make it possible to abort
540 mount operation in case it takes too long.
544 Description: Add cascading_rw to lustre mpi tests.
548 Description: Store SOM data in EA.
549 Details : Also several SOM related cleanups.
553 Description: Add refcount to md_open_data
557 Description: Use refcount on lov_request_set to prevent memory corruption.
561 Description: Statahead: use parent dir's i_mutex for synchronization.
565 Description: loadgen improvements.
569 Description: Enable peer health detection in ksocklnd and ko2iblnd.
573 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
577 Description: Fix race on obd_devcie:md_stats
581 Description: Fix SLES 11 client builds.
582 Details : Resolved uninitalized pointer gcc warnings.
586 Description: Allow non-root access for "lfs check".
587 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
591 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
593 Severity : enhancement
595 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
596 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
600 Description: Sanity test_27p fails.
601 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
602 Sanity tests 27 were cleaned up and improved.
606 Description: Unify req format on client/servers
607 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
608 buffers and swabbers approach..
613 Description: drop join file stuff.
615 Severity : enhancement
617 Description: Add 2.6.27 and SLES11 (patchless client) support
621 Description: Fix in ptlrpc_expire_one_request() to print the signed time
624 Severity : enhancement
626 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
629 Severity : enhancement
631 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
633 Severity : enhancement
635 Description: correctly handle big reply message.
636 Details : send LNet event if reply is bigger than buffer and adjust this buffer
641 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
643 Severity : enhancement
645 Description: Add support for OEL5.
649 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
651 Severity : enhancement
653 Description: "weak-modules" support
654 Details : Implement "weak-modules" support which enables kernel modules
655 to be used with any kernel that implements the same kABI. In
656 order to achieve this modules are now installed in
657 /lib/modules/$(uname -r)/updates/kernel on all distributions.
661 Description: client could not reconnect to OST because of an active request.
662 Details : The client now retries to reconnect to the same server, if a
663 connect request failed with EBUSY or -EAGAIN. Test case 35b
664 conf-sanity.sh has been created.
668 Description: Don't allow make backward step on assignin osc next id.
669 Details : race between allocation next id and ll_sync thread can be cause
670 of set wrong osc next id and can be kill valid ost objects.
674 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
678 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
682 Description: Error handling in osc_statfs_interpret() has been improved.
683 Details : Check in osc_statfs_interpret() for EBADR.
687 Description: Do not send statfs() requests to OSTs disabled by administrator.
688 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
690 Severity : enhancement
692 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
694 Severity : enhancement
696 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
698 Severity : enhancement
700 Description: Enhancement for lfs(1) command to use numeric uid/gid.
704 Description: Improved cleanup procedure for conf-sanity test_46a.
708 Description: Do not start bulk transfer, if client tried to reconnect.
709 Details : Do not start bulk transfer, if client tried to reconnect,
710 but failed due to active rpcs. Check for exp_abort_active_req
711 has been implemented.
715 Description: Prevent inconsistences between linux and lustre mount structures.
716 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
717 Make the sleep interruptible.
719 Severity : enhancement
721 Description: provide server to client comms path
722 Details : server to client communications path via new LDLM_SET_INFO rpc
726 Description: truncate starts GFP_FS allocation under transaction causing deadlock
727 Details : ldiskfs_truncate calls grab_cache_page which may start page
728 allocation under an open transaction. This may lead to
729 calling prune_icache with consequent lustre reentrance.
733 Description: send correctly lsm on open replay
734 Details : MDS is trust to LSM size on replay open, but client can set wrong size
739 Description: enable client interface failover
740 Details : When a child reconnects from another NID, properly update export
741 nid hash position and ldlm reverse import.
743 Severity : enhancement
745 Description: Communicate OST degraded/readonly state via statfs to MDS
746 Details : Flags in the statfs returned from OSTs indicate whether the
747 OST is in a degraded RAID state, or if the filesystem has
748 turned read-only after a filesystem error is detected.
753 Description: don't panic if EPROTO was hit when reading symlink
754 Details : correctly handling request reference in error cases.
756 Severity : enhancement
758 Description: Add LustreNetLink, a kernel-userspace communcation path.
760 Severity : enhancement
762 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
767 Description: access to llog context before init.
768 Details : move handling CATALOGS file at osc layer and forbid access to llog
772 Frequency : rare, only when client is writing to OST on local node
774 Description: Avoid deadlock for local client writes
775 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
776 memory freeing context. This allows OST threads to set the
777 PF_MEMALLOC flag on task structures in order to allocate memory
778 from reserved pools and complete IO.
779 Use GFP_HIGHUSER for OST allocations for non-local client writes,
780 so that the OST threads generate memory pressure and allow
781 inactive pages to be reclaimed.
783 Severity : enhancement
785 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
790 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
791 Details : move ldlm namespace creation in setup phase to avoid grab
792 _lprocfs_lock with cli_sem held.
796 Description: Temporarily disable grant shrink.
797 Details : Disable the feature for debugging.
800 Frequency : only when formatting test filesystems
802 Description: Unable to run several mkfs.lustre on loop devices at the same time
803 Details : mkfs.lustre returns error 256 on the concurrent loop devices
804 formatting. The solution is to proper handle the error.
806 Severity : enhancement
808 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
810 Severity : enhancement
812 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
817 Description: resolve race between obd_disconnect and class_disconnect_exports
818 Details : if obd_disconnect will be called to already disconnected export he
819 forget release one reference and osc module can't unloaded.
821 Severity : enhancement
823 Description: Allow tuning service thread via /proc
824 Details : For each service a new
825 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
826 created that can be used to set min/max thread counts, and get the
827 current number of running threads.
832 Descriptoin: don't return error if have partially created objects for file.
833 Details : lov_update_create_set uses set->set_success as index for created
834 objects, so if some requests will be failed, they will have hole
835 at end of array and qos_shrink_lsm for allocate correct lsm.
837 Severity : enhancement
839 Description: Update OFED support to 1.4.1
844 Description: Reduce small size read RPC
845 Details : Set read-ahead limit for every file and only do read-ahead when
846 available read-ahead pages are bigger than 1M to avoid small size
849 Severity : enhancement
851 Description: move AT tunable parameters for more consistent usage
852 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
854 Severity : enhancement
856 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
857 Details : allow skip disconnected ost for send statfs request and hide error
863 Description: fix racy locking of mballoc block bitmaps causing BUG
864 Details : The locking of the mballoc buddy bitmap and the in-memory
865 block bitmap was using two different spin locks in some
866 cases. This made it possible to incorrectly access the
867 mballoc bitmap while another process was modifying it,
868 causing a sanity assertion to fail. While no on-disk corruption
869 was reported, there was some risk of this happening.
872 Frequency : rare, on llog test 6
874 Descriptoin: don't allow connect to already connected import
875 Details : allowing connect to already connected import is hide connecting problem.
880 Descriptoin: allow kill process which wait statahead result
881 Details : for some reasons 'ls' can stick in waiting result from statahead,
882 in this case need way for kill this process.
884 Severity : enhancement
886 Description: Add state history info file, enhance import info file
887 Details : Track import connection state changes in a new osc/mdc proc file;
888 add overview-type data to the osc/mdc import proc file.
890 Severity : enhancement
892 Description: MDS create should not wait for statfs RPC while holding DLM lock.
894 Severity : enhancement
896 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
901 Descriptoin: OST grant too much space to client even there are not enough space.
902 Details : Client will shrink its grant cache to OST if there are no write
903 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
904 this grant cache if there are already not enough avaible space
905 (left_space < total_clients * 32M).
908 Frequency : start MDS on uncleanly shutdowned MDS device
910 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
911 Details : stay in waiting mds<>ost recovery finished produce random bugs
912 due race between two ll_sync thread for one lov target. send
913 ACTIVATE event only if connect realy finished and import have
917 Frequency : rare, connect and disconnect target at same time
919 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
920 Details : don't call obd_disconnect under lov_lock. this long time
921 operation and can block ptlrpcd which answer to connect request.
926 Descriptoin: don't lose wakeup for imp_recovery_waitq
927 Details : recover_import_no_retry or invalidate_import and import_close can
928 both sleep on imp_recovery_waitq, but we was send only one wakeup
932 Frequency : always with long access acl
934 Descriptoin: mds can't pack reply with long acl.
935 Details : mds don't control size of acl but they limited by reint/getattr
938 Severity : enhancement
940 Description: Update to SLES10 kernel-2.6.16.60-0.33.
942 Severity : enhancement
944 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
947 Frequency : start MDS on uncleanly shutdowned MDS device
949 Descriptoin: aborting recovery hang on MDS
950 Details : don't throttle destroy RPCs for the MDT.
953 Frequency : on remount
955 Description: external journal device not working after the remount
956 Details : clear dev_rdonly flag for external journal devices in
962 Description: shutdown vs evict race
963 Details : client_disconnect_export vs connect request race.
964 if client will evicted at this time - we start invalidate
965 thread without referece to import and import can be freed
969 Frequency : rare, need acl's on inode.
971 Description: client can't handle ost additional correctly
972 Details : if ost was added after client connected to mds client can have
973 hit lnet_try_match_md ... to big messages to wide striped files.
974 in this case need teach client to handle config events about add
975 lov target and update client max ea size at that event.
977 Severity : enhancement
979 Description: Changelogs
980 Details : Changelogs are a lightweight mechanism to track filesystem
981 metadata and namespace changes. The changelog is recorded
982 permanently on the MDTs, and is periodically "consumed" / purged
983 when records are no longer needed.
985 Severity : enhancement
987 Description: compact fld format with extents
988 Details : Store range of seq rather than every seq in FLD. Seq
989 controller update FLD rather than clients. In Case of CMD, mdt0
990 has FLD, all other metadata server act as non persistent proxy
991 for FLD queries and cache fld entries in fld cache.
996 Description: don't skip ost target if they assigned to file
997 Details : Drop slow OSCs if we can, but not for requested start idx.
998 This means "if OSC is slow and it is not the requested
999 start OST, then it can be skipped, otherwise skip it only
1000 if it is inactive/recovering/out-of-space.
1004 Description: more cleanup in mds_lov
1005 Details : not send LOV EA under replay, we can't know about they size at this
1006 time. Don't allow client connect to mds before any ost connected,
1007 for avoid problems with LOV EA size and returning EIO to client.
1009 Severity : enhancement
1011 Description: Interoperability at server side (Disk interoperability)
1013 Severity : enhancement
1015 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1017 Severity : enhancement
1019 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1021 Severity : enhancement
1023 Description: New client IO stack (CLIO).
1025 Severity : enhancement
1027 Description: Commit on sharing. Eliminate inter-client dependencies between
1028 uncommitted transactions by doing transaction commits.
1029 Thereby clients may recovery independently.
1032 Frequency : Create a symlink file with a very long name
1034 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1035 Details : If there is no extra space in the request for early cancels,
1036 ldlm_req_handles_avail() returns 0 instead of a negative value.
1038 Severity : enhancement
1040 Description: Add /proc entry for import status
1041 Details : The mdc, osc, and mgc import directories now have
1042 an import directory that contains useful import data for debugging
1043 connection problems.
1045 Severity : enhancement
1047 Description: Re-disable certain /proc logging
1048 Details : Enable and disable client's offset_stats, extents_stats and
1049 extents_stats_per_process stats logging on the fly.
1052 Frequency : Only on FC kernels 2.6.22+
1054 Description: oops in statahead
1055 Details : Do not drop reference count for the dentry from VFS when lookup,
1056 VFS will do that by itself.
1058 Severity : enhancement
1060 Description: Generic /proc file permissions
1061 Details : Set /Proc file permissions in a more generic way to enable non-
1062 root users operate on some /proc files.
1066 Description: Hitting mdc_commit_close() ASSERTION
1067 Details : Properly handle request reference release in
1068 ll_release_openhandle().
1072 Description: quota recovery deadlock during mds failover
1073 Details : This patch includes att18982, att18236, att18237 in bz14840.
1075 1. fix osts hang when mds does failover with quotaon
1076 2. prevent watchdog storm when osts threads wait for the
1081 Frequency : only patchless client
1082 Description: add workaround for race between add/remove dentry from hash
1084 Severity : enhancement
1086 Description: Allow OST glimpses to return PW locks
1090 Description: LBUG when llog conf file is full
1091 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1095 Description: Prevent import from entering FULL state when server in recovery
1099 Description: service mount cannot take device name with ":"
1100 Details : Only when device name contains ":/" will mount treat it as
1106 Description: replace ptlrpcd with the statahead thread to interpret the async
1107 statahead RPC callback
1111 Frequency : on recovery
1112 Description: I/O failures after umount during fail back
1113 Details : if client reconnected to restarted server we need join to recovery
1114 instead of find server handler is changed and process self
1115 eviction with cancel all locks.
1117 Severity : enhancement
1119 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1121 Severity : enhancement
1123 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1125 Severity : enhancement
1127 Description: Upcall on Lustre log has been dumped
1128 Details : Allow for a user mode script to be called once a Lustre log has
1129 been dumped. It passes the filename of the dumped log to the
1130 script, the location of the script can be specified via
1131 /proc/sys/lnet/debug_log_upcall.
1136 Description: avoid idr_remove called for id which is not allocated.
1137 Details : Move assigment s_dev for clustered nfs to end of initialization,
1138 to avoid problem with error handling.
1143 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1144 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1145 into CONN_USED_HASH and this prodice warning when put connection
1146 again in unused hash.
1149 Severity : enhancement
1151 Description: File striping can now be set to use an arbitrary pool of OSTs.
1153 Severity : enhancement
1155 Description: Export bytes_read/bytes_write count on OSC/OST.
1159 Description: Early reply size mismatch, MGC loses connection
1160 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1161 the connect flags are properly negotiated.
1166 Description: quotas are not honored with O_DIRECT
1167 Details : all writes with the flag O_DIRECT will use grants which leads to
1168 this problem. Now using OBD_BRW_SYNC to guard this.
1172 Description: add quota statistics
1173 Details : 1. sort out quota proc entries and proc code.
1174 2. add quota statistics
1176 Severity : enhancement
1178 Description: enable quota support for HEAD.
1182 Description: Properly propagate oinfo flags from lov to osc for statfs
1183 Details : restore missing copy oi_flags to lov requests.
1185 Severity : enhancement
1187 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1188 : l_getgroups(8), lst(8), routerstat(8)
1189 Details : included man pages for llobdstat(8), llstat(8),
1190 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1192 Severity : enhancement
1194 Description: configure's --enable-quota should check the
1195 : kernel .config for CONFIG_QUOTA
1196 Details : configure is terminated if --enable-quota is passed but
1197 : no quota support is in kernel
1201 Description: Remove portals compatibility
1202 Details : Remove portals compatibility, not interoperable with releases
1207 Description: Resolve device initialization race
1208 Details : Prevent proc handler from accessing devices added to the
1209 obd_devs array but yet be intialized.
1211 Severity : enhancement
1213 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1215 Severity : enhancement
1217 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1222 Description: Using wrong pointer in osc_brw_prep_request
1223 Details : Access to array[-1] can produce panic if kernel compiled with
1224 CONFIG_PAGE_ALLOC enabled
1228 Description: Client runs out of low memory
1229 Details : Consider only lowmem when counting initial number of llap pages
1233 Description: *optional* service tags registration
1234 Details : if the "service tags" package is installed on a Lustre node
1235 When the filesystem is mounted, a local-node service tag will
1236 be created. See http://inventory.sun.com/ for more information
1237 about the Service Tags asset management system.
1241 Description: Kernel BUG tries to release flock
1242 Details : Lustre does not destroy flock lock before last reference goes
1243 away. So always drop flock locks when client is evicted and
1244 perform unlock regardless of successfulness of speaking to MDS.
1248 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1252 Description: sanity test 65a fails if stripecount of -1 is set
1253 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1258 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1259 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1262 Severity : enhancement
1264 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1268 Description: do not process already freed flock
1269 Details : flock can possibly be freed by another thread before it reaches
1270 to ldlm_flock_completion_ast.
1274 Description: LBUG during stress test
1275 Details : Need properly lock accesses the flock deadlock detection list.
1279 Description: oops in page fault handler
1280 Details : kernel page fault handler can return two special 'pages' in error
1281 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1285 Description: timeout with invalidate import.
1286 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1287 should be handled by ptlrpcd. This produce long age waiting and
1288 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1290 Severity : enhancement
1292 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1296 Description: enable MGS and MDT services start separately
1297 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1298 a co-located MGS without starting the MGS, which is a complement
1299 to 'nosvc' mount option.
1303 Frequency : after recovery
1304 Description: precreate to many object's after del orphan.
1305 Details : del orphan st in oscc last_id == next_id and this triger growing
1306 count of precreated objects. Set flag LOW to skip increase count
1307 of precreated objects.
1311 Frequency : rare, on clear nid stats
1312 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1313 Details : when clean nid stats sometimes try destroy live entry,
1314 and this produce panic in free.
1318 Description: Stack overflow during MDS log replay
1319 ease stack pressure by using a thread dealing llog_process.
1323 Description: wait until IO finished before start new when do lock cancel.
1324 Details : VM protocol want old IO finished before start new, in this case
1325 need wait until PG_writeback is cleared until check dirty flag and
1326 call writepages in lock cancel callback.
1328 Severity : enhancement
1330 Description: using special macro for print time and cleanup in includes.
1334 Description: mds_mfd_close() ASSERTION(rc == 0)
1335 Details : In mds_mfd_close(), we need protect inode's writecount change
1336 within its orphan write semaphore to prevent possible races.
1340 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1344 Frequency : rare, on shutdown ost
1345 Description: don't hit live lock with umount ost.
1346 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1347 use shrink_dcache_sb instead.
1351 Description: don't panic with use echo client
1352 Details : echo client pass NULL as client nid pointer and this produce null
1353 pointer dereference.
1357 Description: fix build on ppc32
1358 Details : compile code with -m64 flag produce wrong object file for ppc32.
1362 Description: add message levels for liblustreapi
1366 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1367 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1368 will result in an -EAGAIN and in the exit call path we call
1369 alloc_rr() which will with increasing aggressiveness attempt to
1370 aquire precreated objects on the minimum number of required OSCs.
1374 Description: Use old size assignment to avoid deadlock
1375 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1376 the scheduling while holding a spinlock. We do not need locking
1377 for size in ll_update_inode() because size is only updated from
1378 the MDS for directories or files without objects, so there is no
1379 other place to do the update, and concurrent access to such inodes
1380 are protected by the inode lock.
1384 Description: resolve "_IOWR redefined" build error on SLES10
1388 Description: dump the memory debugging after all modules are unloaded to
1389 suppress false negative in conf_sanity test 39
1391 Severity : enhancement
1393 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1398 Description: fix SLES kernel versioning
1399 Details : the kernel version for our SLES 10 kernel did not include a "-"
1400 before the "smp" at the end. while this was not a problem in
1401 general, it did mean that software trying to use the kernel
1402 version to try to detect a vendor specific kernel would fail.
1403 this was most evident by the OFED build scripts.
1407 Description: Don't update lov_desc members until making sure they are valid
1408 Details : When updating lov_desc members via proc fs, need fix their
1409 validities before doing the real update.
1413 Description: don't put request into delay list while invalidate in flight.
1414 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1415 import in flight. this produce timeout for invalidate and sometimes
1416 can cause stale data.
1420 Frequency : on ppc only
1421 Description: not convert ost objects for directory because it's not exist.
1422 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1426 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1428 Severity : enhancement
1430 Description: Update to support 2.6.22.14 vanilla kernel.
1434 Frequency : rare, on recovery
1435 Description: read procfs can produce deadlock in some situation
1436 Details : Holding lprocfs lock which send rpc can produce block for destroy
1437 obd objects and this also block reconnect with -EALREADY. This
1438 isn't fix all lprocfs bugs - but make it rare.
1440 Severity : enhancement
1442 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1445 Frequency : frequent on X2 node
1447 Description: mdc_set_open_replay_data LBUG
1448 Details : Set replay data for requests that are eligible for replay.
1452 Description: lustre_mgs: operation 101 on unconnected MGS
1453 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1454 MGC, and late on MGC cannot successfully connect to MGS and a lot
1455 of the error messages complaining that MGS is not connected.
1459 Frequency : on network error
1460 Description: panic with double free request if network error
1461 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1462 but it's true only for synchronus enqueue, for async enqueue
1463 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1466 Severity : enhancement
1468 Description: client-side metadata stat-ahead during readdir(directory readahead)
1469 Details : perform client-side metadata stat-ahead when the client detects
1470 readdir and sequential stat of dir entries therein
1473 Frequency : on start mds
1475 Description: Implement get_info(last_id) in obdfilter.
1478 Frequency : occasional
1480 Description: Correctly check stale fid, not start epoch if ost not support SOM
1481 Details : open with flag O_CREATE need set old fid in op_fid3 because
1482 op_fid2 was overwritten with new generated fid, but mds can answer
1483 with one of these two fids and both is not stale. Setattr
1484 incorrectly started an epoch and assume will be called
1485 done_writing, but without SOM done_writing ever being called.
1488 Frequency : rare, depends on device drivers and load
1490 Description: MDS or OSS nodes crash due to stack overflow
1491 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1492 In some cases, in conjunction with device drivers that use a lot
1493 of stack the MDS (or possibly OSS) service threads could overflow
1494 the stack. One change which was identified to consume additional
1495 stack has been reworked to avoid the extra stack usage.
1498 Frequency : occasional
1500 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1501 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1502 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1503 oscc_has_objects and see OSC has no objests and call
1504 oscc_internal_create to resend create request.
1506 Severity : enhancement
1508 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1510 Severity : enhancement
1512 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1515 Frequency : very rare
1517 Description: Fix replay if there is an un-replied request and open
1518 Details : In some cases, older replay request will revert the
1519 mcd->mcd_last_xid on MDS which is used to record the client's
1520 latest sent request.
1522 Severity : enhancement
1524 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1526 Severity : enhancement
1528 Description: Add rhel5 support to HEAD.
1530 Serverity : enhancement
1532 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1537 Description: Don't allow skipping OSTs if index has been specified.
1538 Details : Don't allow skipping OSTs if index has been specified, make
1539 locking in internal create lots better.
1543 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1544 Details : access to bitfield in structure is always rounded to long
1545 and this produce problem with not atomic change any bit.
1549 Description: Lustre make rpms failed.
1550 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1551 builds Lustre rpms from tarball.
1554 Frequency : rare on shutdown ost
1556 Description: If llog cancel was not send before clean_exports phase, this can
1557 produce deadlock in llog code.
1558 Details : If llog thread has last reference to obd and call class_import_put
1559 this produce deadlock because llog_cleanup_commit_master wait when
1560 last llog_commit_thread exited, but this never success because was
1561 called from llog_commit_thread.
1565 Description: allow userland application know is lost one of stripes.
1566 Details : fill lvb_blocks with error code on ost and return it to
1567 application if error flag found.
1571 Description: NULL lov_tgts causing MDS oops
1572 Details : more safe checks for NULL lov_tgts for avoid oops.
1574 Severity : enhancement
1576 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1580 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1581 Details : If an OST is down the MDS will hang indefinitely in
1582 obd_statfs_rqset() waiting for the statfs data. While for
1583 MDS QOS usage of statfs, it should not stuck in waiting.
1585 Severity : enhancement
1587 Description: remote_acl support
1588 Details : Support ACL-based permission check for remote user.
1589 Support setfacl/getfacl for remote user with the utils
1590 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1591 the system "{s,g}etfacl" utils.
1593 Severity : enhancement
1595 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1597 Severity : enhancement
1599 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1603 Description: make mgs_setparam() handle fsname containing dash
1604 Details : fsname containing a dash does not work with lctl conf_param
1606 Severity : enhancement
1608 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1610 Severity : enhancement
1612 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1616 Frequency : rare, at shutdown
1617 Description: access already free / zero obd_namespace.
1618 Details : if client_disconnect_export was called without force flag set,
1619 and exist connect request in flight, this can produce access to
1620 NULL pointer (or already free pointer) when connect_interpret
1621 store ocd flags in obd_namespace.
1625 Frequency : only at startup
1626 Description: not alloc memory with spinlock held.
1627 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1628 if any spinlock held.
1630 Severity : enhancement
1632 Description: make lustre randomly fail allocating memory
1633 Details : Make lustre randomly failed allocating memory for testing purpose.
1635 Severity : enhancement
1637 Description: lost problems with lov objid file
1638 Details : Fixes some scability and access to not inited memory problems
1639 in work with lov objdid file.
1644 Description: lfs find does not continue on file error
1645 Details : Continue other files processing when a file/dir is absent.
1649 Description: Inconsistent usage of lustre_pack_reply()
1650 Details : Standardize the usage of lustre_pack_reply() such that it
1651 always generate a CERROR on failure.
1654 Frequency : occasional
1656 Description: MDS hang or stay in waiting lock
1657 Details : If client receive lock with CBPENDING flag ldlm need send lock
1658 cancel as separate rpc, to avoid situation when cancel request
1659 can't processed due all i/o threads stay in wait lock.
1663 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1667 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1671 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1675 Description: Update to RHEL4 latest kernel.
1677 Severity : enhancement
1679 Description: Build SLES10 patchless client fails
1680 Details : The configure was broken by run ./configure with
1681 --with-linux-obj=.... argument for patchless client. When the
1682 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1683 can't search header adequately. Use absolute path such as
1684 -I($LINUX)/include instead.
1688 Description: interrupt oig_wait produce painc on resend.
1689 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1690 private set, and this produce situation when rq_ptlrpcd_data not
1691 copyed to new allocated request and triggered LBUG on assert
1692 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1693 ptlrpcd set if request is changed and can be safety changed to use
1698 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1699 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1700 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1701 these macros and replace them with their difinitions instead.
1705 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1706 Details : Update which_patch & target file for SLES10 latest kernel.
1709 Description: Remove client patches from SLES 10 kernel.
1710 Details : This causes SLES 10 clients to behave as patchless clients
1711 even on a Lustre-patched (server) kernel.
1713 Severity : enhancement
1715 Description: self-adjustable client's lru lists
1716 Details : use adaptive algorithm for managing client cached locks lru
1717 lists according to current server load, other client's work
1718 pattern, memory activities, etc. Both, server and client
1719 side namespaces provide number of proc tunables for controlling
1722 Severity : enhancement
1724 Description: light-weight GSS support
1725 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1726 and reduce performance overhead.
1728 Severity : enhancement
1730 Description: Linux keyring support
1731 Details : Support using service of Linux keyring for Lustre GSS internal
1732 context refresh/cache mechanism.
1736 Description: Fix errors in lfs documentation
1737 Details : Fixes man pages
1741 Description: don't use GFP_* in generic Lustre code.
1742 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1746 Description: obdclass is limited by single OBD_ALLOC(idarray)
1747 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1751 Description: warnings with build patchless client with vanila 2.6.19 and up
1752 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1753 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1757 Frequency : Only for SLES
1758 Description: sanity_quota fail test_1
1759 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1760 file, which makes TSTID[2] inunique.
1764 Frequency : Only for SLES9
1765 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1766 Details : Update target/ChangeLog/which_patch .
1770 Frequency : Only for test_10 in sanity-quota.sh
1771 Description: a bug in quota test code
1772 Details : checking a test flag in a wrong place
1776 Frequency : Only for test_13 in sanity-quota.sh
1777 Description: a bug in quota test script
1778 Details : shouldn't check overfull allocated quota
1782 Description: jbd statistics
1783 Details : Port older jbd statistics patch for sles10
1785 Severity : enhancement
1787 Description: Add printing inode info into message about error in writepage.
1792 Description: reply_lock_interpret crash due to race with it and lock cancel.
1793 Details : Do not replay locks that are being cancelled. Do not reference
1794 locks by their address during replay, just by their handle.
1798 Frequency : When flocks are used.
1799 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1800 Details : Flock locks might destroy just granted lock if it could be merged
1801 with another existing flock, this is done in completion handler,
1802 so teach ldlm_cli_enquque_fini that this is a valid case for
1808 Description: Oops in read and write path when failing to allocate lock.
1809 Details : Check if lock allocation failed and return error back.
1813 Description: lstripe command fails for valid OST index
1814 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1815 instead of lov->desc.ld_active_tgt_count.
1819 Description: sanity.sh failed test 103
1820 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1825 Description: df doesn't work properly if diskfs blocksize != 4K
1826 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1830 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1831 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1833 Severity : enhancement
1835 Description: Adaptive timeouts
1836 Details : RPC timeouts adapt to changing server load and network
1837 conditions to reduce resend attempts and improve recovery time.
1841 Description: llapi_file_create() does not allow some changes
1842 Details : add llapi_file_open() that allows specifying the mode and
1843 open flags, and also returns an open file handle.
1847 Description: merge and cleanup kernel patches.
1848 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1852 Description: Add journal checksum support.(Kernel part)
1853 Details : The journal checksum feature adds two new flags i.e
1854 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1855 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1856 indicates that the commit block contains the checksum for
1857 the blocks described by the descriptor blocks. Now commit
1858 record can be sent to disk without waiting for descriptor
1859 blocks to be written to disk. This behavior is controlled
1860 using JBD2_FEATURE_ASYNC_COMMIT flag.
1864 Description: OSS needs mutliple precreate threads
1865 Details : Add ability to start more than one create thread per OSS.
1869 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1870 Details : execute lfs setstripe on client
1874 Description: mds_obd_create error creating tmp object
1875 Details : When the user sets quota on root, llog will be affected and can't
1876 create files and write files.
1879 Frequency : Always on ia64 patchless client, and possibly others.
1881 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1882 Details : This allows the patchless client to be loaded on architectures
1883 without this export.
1887 Description: RedHat Update kernel for RHEL5
1888 Details : Add the kernel config file for RHEL5.
1892 Description: RedHat Update kernel for RHEL5
1893 Details : Modify the kernel config file more closer RHEL5.
1897 Description: Build failure against Centos5 (RHEL5)
1898 Details : Define PAGE_SIZE when it isn't present.
1902 Description: test_7 on sanity-quota.sh can't be used on separate machines
1903 Details : add facet to handle it
1907 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1908 Details : with NFS, the anon dentry's parent was set to itself in
1909 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1910 corresponding dentry other than use rec->ur_name.
1912 Severity : enhancement
1914 Description: lfs setstripe enhancement
1915 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1919 Description: enable data checksumming by default
1920 Details : enable checksum by default, allow --disable-checksum
1921 configure option and "-o nochecksum" mount option. Checksums
1922 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1926 Description: lustre support for RHEL5
1927 Details : Add support for RHEL5.
1931 Description: Client eviction due to failover config
1932 Details : after a connection loss, the lustre client should attempt to
1933 reconnect to the last active server first before trying the
1934 other potential connections.
1938 Description: when mds and osts use different quota unit(32bit and 64bit),
1939 quota will be released repeatly.
1940 Details : void sending multiple quota reqs to mds, which will keep the
1941 status between the reqs.
1945 Description: osts not allocated evenly to files
1946 Details : change the condition to increase offset_idx
1949 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1951 Description: Data corruption for OSTs that are formatted larger than 2TB
1953 Details : When generating the bio request for lustre file writes the
1954 sector number would overflow a temporary variable before being
1955 used for the IO. The data reads correctly from Lustre (which
1956 will overflow in a similar manner) but other file data or
1957 filesystem metadata may be corrupted in some cases.
1961 Description: Tune the kernel for good SCSI performance.
1962 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1963 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1968 Description: rewrite ext2-derived code in obdclass/uuid.c
1969 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1970 from scratch preserving functionality.
1974 Description: "lfs find -obd UUID" prints directories
1975 Details : "lfs find -obd UUID" will return all directory names instead
1976 of just file names. It is incorrect because the directories
1977 do not reside on the OSTs.
1981 Description: use i_size_read and i_size_write in 2.6 port
1982 Details : replace inode->i_size access with i_size_read/write()
1985 Frequency : when removing large files
1987 Description: scheduling issue during removal of large Lustre files
1988 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1989 It causes scheduling issues when removing large files (17TB in the
1993 Frequency : only with liblustre clients on XT3
1995 Description: evictions taking too long
1996 Details : allow llrd to evict clients directly on OSTs
1999 Frequency : only on ppc
2001 Description: /proc/fs/lustre/devices broken on ppc
2002 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2003 We should make sure the type of 'index' is loff_t and then cast
2004 explicitly as needed below. Do not assign an explicitly cast
2008 Frequency : only for rhel5
2010 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2011 Details : Modify the target file & which_kernel.
2013 Severity : enhancement
2015 Description: omit set fsid for export NFS
2016 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2017 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2021 Frequency : Always, for kernels after 2.6.16
2022 Description: Fix warning idr_remove called for id=.. which is not allocated.
2023 Details : Last kernels save old s_dev before kill super and not allow
2024 to restore from callback - restore it before call kill_anon_super.
2028 Description: buffer overruns could theoretically occur
2029 Details : llapi_semantic_traverse() modifies the "path" argument by
2030 appending values to the end of the origin string, and a
2031 overrun may occur. Adding buffer overrun check in liblustreapi.
2035 Description: Fix error on 'ls .' at the top of the Lustre mount.
2036 Details : Don't revalidate dentry if it is a root dentry.
2040 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2041 Details : Modify vm-tunables-rhel4.patch.
2045 Description: Kernel config for 2.6.18-vanilla.
2046 Details : Modify targets/2.6-vanilla.target.in.
2047 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2048 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2049 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2050 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2054 Description: improve handling recoverable errors
2055 Details : if request processig with error which can be recoverable on server
2056 request should be resend, otherwise page released from cache and
2061 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2062 Details : Modify target file & which_patch.
2063 A flaw was found in the IA32 system call emulation provided
2064 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2065 value could be stored in the %RAX register, which could trigger an
2066 out-of-bounds system call table access. An untrusted local user
2067 could exploit this flaw to run code in the kernel
2068 (ie a root privilege escalation). (CVE-2007-4573).
2072 Description: O_DIRECT bypasses client statistics.
2073 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2074 read_ahead_stats were updated. copied the stats section from
2075 osc_send_oap_rpc() into async_internal().
2079 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2083 Description: change order of libsysio includes
2084 Details : '#include sysio.h' should always come before '#include xtio.h'
2088 Description: handle "serious error: objid * already exists" more gracefully
2089 Details : If LAST_ID value on disk is smaller than the objects existing in
2090 the O/0/d* directories, it indicates disk corruption and causes an
2091 LBUG(). If the object is 0-length, then we should use the existing
2092 object. This will help to avoid a full fsck in most cases.
2094 Severity : enhancement
2096 Description: adapt the lustre_config script to support the upgrade case
2097 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2098 server targets to 1.6 in parallel.
2102 Description: To avoid grant space > avaible space when the disk is almost
2103 full. Without this patch you might see the error "grant XXXX >
2104 available" or some LBUG about grant, when the disk is almost
2106 Details : In filter_check_grant, for non_grant cache write, we should
2107 check the left space by if (*left > ungranted + bytes), instead
2108 of (*left > ungranted), because only we are sure the left space
2109 is enough for another "bytes", then the ungrant space should be
2110 increase. In client, we should update cl_avail_grant only there
2111 is OBD_MD_FLGRANT in the reply.
2115 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2116 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2117 described in CVE-2007-4573. This problem could allow untrusted
2118 local users to gain root access.
2121 Frequency : when using O_DIRECT and quotas
2123 Description: Incorrect file ownership on O_DIRECT output files
2124 Details : block usage reported by 'lfs quota' does not take into account
2125 files that have been written with O_DIRECT.
2130 Description: touch file failed when fs is not full
2131 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2132 otherwise we can get ENOSP while fs is not full.
2136 Description: parallel lock callbacks
2137 Details : Instead of sending blocking and completion callbacks as separated
2138 requests, adding them to a set and sending in parallel.
2141 Frequency : only for Cray XT3
2142 Bugzilla : 12829/13455
2143 Description: Changing primary group doesn't change the group lustre assigns to
2145 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2146 overridden with the primary group provided by the group upcall,
2147 whereas the supplied fsgid can be trusted if it is in the list of
2148 supplementary groups returned by the group upcall.
2150 Severity : enhancement
2152 Description: Allow masking D_WARNING, D_ERROR messages from console
2153 Details : Console messages can now be disabled via lnet.printk.
2157 Description: User code with malformed file open parameter crashes client node
2158 Details : Before packing join_file req, all the related reference should be
2159 checked carefully in case some malformed flags cause fake
2160 join_file req on client.
2164 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2165 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2166 mode is equal to the granted mode. However, it should also check
2167 that LDLM_FL_CP_REQD is not set.
2171 Description: Heavy nfs access might result in deadlocks
2172 Details : After ELC code landed, it is now improper to enqueue any mds
2173 locks under och_sem, because enqueue might want to decide to
2174 cancel open locks for same inode we are holding och_sem for.
2178 Description: Client eviction while running blogbench
2179 Details : A lot of unlink operations with concurrent I/O can lead to a
2180 deadlock causing evictions. To address the problem, the number of
2181 oustanding OST_DESTROY requests is now throttled to
2182 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2183 callbacks are processed in priority.
2187 Description: enable ACLs on MDS by default
2188 Details : ACLs must be enabled on MDS by default.
2191 Frequency : PPC/PPC64 only
2193 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2194 Details : fix duplicated definitions between asm-ppc64/types.h and
2195 lustre_types.h on PPC.
2198 Frequency : PPC/PPC64 only
2200 Description: asm-ppc/segment.h does not exist
2201 Details : fix compile issue on PPC.
2205 Description: better handle error messages in extents code
2208 Frequency : RHEL4 only
2210 Description: mkfs is very slow on IA64/RHEL4
2211 Details : A performance regression has been discovered in the MPT Fusion
2212 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2213 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2214 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2216 Severity : enhancement
2218 Description: SNMP support enhancement
2219 Details : Adding total number of sampled request for an MDS node in snmp
2222 Severity : enhancement
2224 Description: Optimize ldlm waiting list processing for PR extent locks
2225 Details : When processing waiting list for read extent lock and meeting read
2226 lock that is same or wider to it that is not contended, skip
2227 processing rest of the list and immediatelly return current
2228 status of conflictness, since we are guaranteed there are no
2229 conflicting locks in the rest of the list.
2233 Description: Time out and refuse to reconnect
2234 Details : When the failover node is the primary node, it is possible
2235 to have two identical connections in imp_conn_list. We must
2236 compare not conn's pointers but NIDs, otherwise we can defeat
2237 connection throttling.
2241 Description: port llog fixes from b1_6 into HEAD
2242 Details : Port llog reference couting and some llog cleanups from b1_6
2243 (bug 10800) into HEAD, for protect from panic and access to already
2244 free llog structures.
2248 Description: Detect stride IO mode in read-ahead
2249 Details : When a client does stride read, read-ahead should detect that and
2250 read-ahead pages according to the detected stride pattern.
2254 Description: data checksumming impacts single node performance
2255 Details : add support for several checksum algorithm. Currently, only CRC32
2256 and Adler-32 are supported. The checksum type can be changed on
2257 the fly via /proc/fs/lustre/osc/*/checksum_type.
2261 Description: use adler32 for page checksums
2262 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2267 Description: build for x2 fails
2268 Details : fix compile issue on Cray systems.
2272 Description: Properly match for duplicate locks
2273 Details : Due to different lock order from skiplists code, we need to
2274 traverse entire list for now
2277 Frequency : only on PPC/SLES10
2279 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2280 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2281 be defined. Add a hack in mkfs_lustre.c to work around this compile
2286 Description: LASSERT on MDS when client holding flock lock dies
2287 Details : ldlm pool logic depends on number of granted locks equal to
2288 number of released locks which is not true for flock locks, so
2289 just exclude such locks from consideration.
2293 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2294 Details : Use fsfilt_sync() for both the whole filesystem sync and
2295 individual file sync to eliminate dangerous inode locking
2296 with I_LOCK that can lead to a deadlock.
2300 Description: performance in 1.6.3
2301 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2302 sure that 1MB requests are not fragmented by the block layer.
2304 Severity : enhancement
2306 Description: organize the server-side client stats on per-nid basis
2307 Details : Change the structure of stats under obdfilter and mds to
2315 The "uuid"s file would list the uuids of _active_ exports.
2316 And the clear entry is to clear all stats and stale nids.
2318 Severity : enhancement
2320 Description: eliminate client locks in face of contention
2321 Details : file contention detection and lockless i/o implementation
2322 for contended files.
2326 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2329 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2331 Description: MDS may lose file striping (and hence file data) in some cases
2332 Details : If there are additional extended attributes stored on the MDS,
2333 in particular ACLs, SELinux, or user attributes (if user_xattr
2334 is specified for the client mount options) then there is a risk
2335 of attribute loss. Additionally, the Lustre file striping
2336 needs to be larger than default (e.g. striped over all OSTs),
2337 and an additional attribute must be stored initially in the
2338 inode and then increase in size enough to be moved to the
2339 external attribute block (e.g. ACL growing in size) for file
2344 Description: skiplist implementation simplification
2345 Details : skiplists are used to group compatible locks on granted list
2346 that was implemented as tracking first and last lock of each
2347 lock group the patch changes that to using doubly linked lists
2351 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2352 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2353 because LASSERT on both the data supplied by a client, and the
2354 data on disk is dangerous and incorrect.
2356 Severity : enhancement
2358 Description: Slow truncate/writes to huge files at high offsets.
2359 Details : Directly associate cached pages to lock that protect those pages,
2360 this allows us to quickly find what pages to write and remove
2361 once lock callback is received.
2365 Description: more ldlm soft lockups
2366 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2367 starve other threads from the resource lock for a long time in
2368 case of long waiting queue, so change the debug level from
2369 D_OTHER to the less frequently used D_INFO.
2371 Severity : enhancement
2373 Description: add -gid, -group, -uid, -user options to lfs find
2377 Description: Hung threads in invalidate_inode_pages2_range
2378 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2379 once one is completed. As a result, some RPCs are stuck in the
2380 queue and are never sent.
2384 Description: filter threads hungs on waiting journal commit
2385 Details : Cleanup filter group llog code, then only filter group llog will
2386 be only created in the MDS/OST syncing process.
2390 Description: Procfs and llog threads access destoryed import sometimes.
2391 Details : Sync the import destoryed process with procfs and llog threads by
2392 the import refcount and semaphore.
2394 Severity : enhancement
2396 Description: openlock cache of b1_6 port to HEAD
2401 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2402 Details : If insertion of an extent fails, then discard the inode
2403 preallocation and free data blocks else it can lead to duplicate
2408 Description: don't always update ctime in ext3_xattr_set_handle()
2409 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2410 In some cases the ctime should not be updated, for example for
2411 2.0->1.8 compatibility it is necessary to delete an xattr and it
2412 should not update the ctime.
2416 Bugzilla : 15713/16362
2417 Description: Assertion in iopen_connect_dentry in 1.6.3
2418 Details : looking up an inode via iopen with the wrong generation number can
2419 populate the dcache with a disconneced dentry while the inode
2420 number is in the process of being reallocated. This causes an
2421 assertion failure in iopen since the inode's dentry list contains
2422 both a connected and disconnected dentry.
2426 Description: assertion failure in ldlm_handle2lock()
2427 Details : fix a race between class_handle_unhash() and class_handle2object()
2428 introduced in lustre 1.6.5 by bug 13622.
2433 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2434 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2439 Description: oops in sd_iostats_seq_show()
2440 Details : unloading/reloading the scsi low level driver triggers a kernel
2441 bug when trying to access the sd iostat file.
2446 Description: Kernel panics during QLogic driver reload
2447 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2448 patch, causing memory corruption.
2453 Description: journal_dev option does not work in b1_6
2454 Details : pass mount option during pre-mount.
2456 Severity : enhancement
2458 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2459 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2460 extent information of a file. It can be used to map logical blocks
2461 in a file to physical blocks in the block device.
2465 Description: LDLM soft lockups - improvement
2466 Details : It is be possible to send the lock handle along with each read
2467 or write request because the client is already doing a lock match
2468 itself so there isn't any reason the OST should have to re-do that
2472 Frequency : only on Cray X2
2474 Description: X2 build failures
2475 Details : fix build failures on Cray X2.
2479 Description: xid & resent requests
2480 Details : Initialize RPC XID from clock at startup (randomly if clock is
2483 Severity : enhancement
2485 Description: Add lustre_start utility to start or stop multiple Lustre servers
2490 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2491 Details : In case of memory pressure, list_del() can be called twice on
2492 req->rq_history_list, causing a kernel oops.
2496 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2497 Details : fix stack overflow in the distributed lock manager by defering
2498 export eviction after a failed AST to the elt thread instead of
2499 handling it in the dlm interpret routine.
2503 Description: Convert some comments to new format.
2504 Details : Update documenting comments to match doxygen conventions.
2508 Description: Grammar fixes.
2509 Details : A couple of trivial spelling fixes.
2513 Description: OSD_COUNTERS-mandatory
2514 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2515 to simplify the code.
2519 Description: simplify cmm_device freeing logic.
2520 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2524 Description: Add lockdep support to dt_object_operations locking interface.
2525 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2526 parameter indicating lock ordering. Update mdd code to use new
2531 Description: Introduce failloc constants for lockless IO tests.
2532 Details : Add two new failloc constants to test lockless IO. Only one of
2533 them in implemented---another is checked in yet to be landed
2538 Description: Add lockdep support for inode mutex.
2539 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2540 in the situations where Lustre has to lock more than one inode
2545 Description: Add optional invariants checking support.
2546 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2547 configure switch. This macro is to be used for consistency and
2548 sanity checks that are too expensive to be left in `production'
2553 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2554 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2559 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2560 Details : Use spin_lock_nested() in (the only) situation where more than
2561 one ldlm_lock is locked simultaneously. Also, fix possible
2562 dead-lock in ldlm_lock_change_resource() by enforcing particular
2567 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2568 Details : Instead of specifying each ldlm_lock call-back through separate
2569 parameter, wrap them into struct ldlm_callback_suite.
2573 Description: Kill join_lru obd method and its callers.
2574 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2575 in memory---a requirement implemented through ->o_join_lru() obd
2576 method in HEAD. Remove this method and its users.
2580 Description: Add asynchronous ldlm ENQUEUE completion handler.
2581 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2582 case that stock ldlm_completion_ast() cannot handle as it waits
2583 until lock is granted. Introduce new ldlm_completion_ast_async()
2584 for this. Also comment ldlm_completion_ast().
2588 Description: ldlm_error <-> errno conversion.
2589 Details : Add functions to map (rather arbitrary) between LDLM error codes
2590 and standard errno values. CLIO needs this to prevent LDLM specific
2591 constants from escaping ldlm and osc.
2595 Description: Kill unused ldlm_handle2lock_ns() function.
2596 Details : Kill unused ldlm_handle2lock_ns() function.
2600 Description: Add lu_ref support to ldlm_lock
2601 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2602 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2603 and ldlm_resource. LDLM interface has to be changed, because code that
2604 releases a reference on a lock, has to "know" what reference this is.
2605 In the most frequent case
2607 lock = ldlm_handle2lock(handle);
2609 LDLM_LOCK_PUT(lock);
2611 no changes are required. When any other reference (received _not_ from
2612 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2613 instead of LDLM_LOCK_PUT().
2615 Arguably, changes are pervasive, and interface requires some discipline
2616 for proper use. On the other hand, it was very instrumental in finding
2617 a few leaked lock references.
2621 Description: Add ldlm_lock_addref_try().
2622 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2623 attempts to addref a lock that might be being canceled
2628 Description: Add ldlm_weigh_callback().
2629 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2630 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2631 hard-coded `number of pages' logic.
2635 Description: Add lockdep annotations to llog code.
2636 Details : Use appropriately tagged _nested() locking calls in the places
2637 where llog takes more than one ->lgh_lock lock.
2641 Description: Add loi_kms_set().
2642 Details : Wrap kms updates into a helper function.
2646 Description: Constify instances of struct lsm_operations.
2647 Details : Constify instances of struct lsm_operations.
2651 Description: lu_conf support.
2652 Details : On a server, a file system object is uniquely identified
2653 by a fid, which is sufficient to locate and load all object
2654 state (inode). On a client, on the other hand, more data are
2655 necessary instantiate an object. Change lu_object_find() and
2656 friends to take additional `lu_conf' argument describing object.
2657 Typically this includes layout information.
2661 Description: lu_context fixes.
2662 Details : Introduce new lu_context functions that are needed on the client
2663 side, where some system threads (ptlrpcd) are shared by multiple
2664 modules, and so cannot be stopped during module shutdown.
2668 Description: Add start and stop methods to lu_device_type_operations.
2669 Details : Introduce two new methods in lu_device_type_operations, that are
2670 invoked when first instance of a given type is created and last one
2671 is destroyed respectively. This is need by CLIO.
2675 Description: Add lu_ref support to struct lu_device.
2676 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2677 track leaked references.
2681 Description: Introduce lu_kmem_descr.
2682 Details : lu_kmem_descr and its companion interface allow to create
2683 and destroy a number of kmem caches at once.
2687 Description: Fix lu_object finalization race.
2688 Details : Fix a race between lu_object_find() finding an object and its
2689 concurrent finalization. This race is (most likely) not possible
2690 on the server, but might happen on the client.
2694 Description: Introduce lu_ref interface.
2695 Details : lu_ref is a debugging module allowing to track references to
2696 a given object. It is quite cpu expensive, and has to be
2697 explicitly enabled with --enable-lu_ref. See usage description
2702 Description: Factor lu_site procfs stats into a separate function.
2703 Details : Separate lu_site stats printing code into a separate function
2704 that can be reused on a client.
2708 Description: Constify instances of struct {lu,dt,md}_device_operations.
2709 Details : Constify instances of struct {lu,dt,md}_device_operations.
2713 Description: Introduce struct md_site and move meta-data specific parts of
2714 struct lu_site here.
2715 Details : Move md-specific fields out of struct lu_site into special struct
2716 md_site, so that lu_site can be used on a client.
2720 Description: Kill mdd_lov_destroy().
2721 Details : Remove unused mdd code.
2725 Description: Add st_block checking to multistat.c.
2726 Details : Add st_block checking to multistat.c.
2730 Description: Add lu_ref support to struct obd_device.
2731 Details : Add lu_ref tracking to obd_device.
2735 Description: Kill obd_set_fail_loc().
2736 Details : Remove unused code.
2740 Description: Add special type for ptlrpc_request interpret functions.
2741 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2742 there. Actual usage will be in CLIO.
2746 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2747 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2748 doesn't work with lockdep.
2752 Description: Add rwv.c test program.
2753 Details : New testing program exercising readv(2) and writev(2) (Qian).
2757 Description: Add sendfile.c test program.
2758 Details : New testing program exercising sendfile(2) (Jay).
2762 Description: Ratelimit a message that can be very frequent.
2763 Details : Ratelimit a memory allocation failure message that can
2768 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2769 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2773 Description: Liblustre build fixes.
2774 Details : Liblustre build fixes.
2778 Description: libcfs: add cfs_{need,cond}_resched() interface.
2779 Details : libcfs: add cfs_{need,cond}_resched() definition and
2780 implementations for Linux, NT, and liblustre.
2782 Severity : enhancement
2784 Description: More exported tunables for mballoc
2785 Details : Add support for tunable preallocation window and new tunables for
2786 large/small requests
2790 Description: Detect corruption of block bitmap and checking for preallocations
2791 Details : Checks validity of on-disk block bitmap. Also it does better
2792 checking of number of applied preallocations. When corruption is
2793 found, it turns filesystem readonly to prevent further corruptions.
2797 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2798 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2799 window to 0. If it does hit cache miss, and read-pattern is still
2800 stride-io mode, does not reset the stride window, but also does not
2801 increase the stride window length in this case.
2805 Frequency : only for big-endian servers
2806 Description: Check if system is big-endian while mounting fs with extents feature
2807 Details : Mounting a filesystem with extents feature will fail on big-endian
2808 systems since ext3-based ldiskfs is not supported on big-endian
2809 systems. This can be over-riden with "bigendian_extents" mount option.
2811 Severity : enhancement
2813 Description: The root squash functionality
2814 Details : A security feature, which is to prevent users from being able
2815 to mount lustre on their desktop, run as root, and delete
2816 all of the files in the filesystem. The goal is accomplished by
2817 remapping user id (UID) and group id (GID) of the root user to
2818 a UID and GID specified by the system administartor via Lustre
2819 configuration management server (MGS). The functionality also
2820 allows to specify sets of clients for which the remapping does
2825 Description: Excessive recovery window
2826 Details : With AT enabled, the recovery window can be excessively long (6000+
2827 seconds). To address this problem, we no longer use
2828 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2829 timeout no longer depends on the service time, it is set to
2830 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2831 time via pb_service_time.
2835 Description: Watchdog triggered on MDS failover
2836 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2837 the OSTs know that the MDS "UUID" can be reused for the same export
2838 from a different NID, so we do not need to wait for the export to be
2842 Frequency : rare, only if using MMP with Linux RAID
2844 Description: MMP doesn't work with Linux RAID
2845 Details : While using HA for Lustre servers with Linux RAID, it is possible
2846 that MMP will not detect multiple mounts. To make this work we
2847 need to unplug the device queue in RAID when the MMP block is being
2848 written. Also while reading the MMP block, we should read it from
2849 disk and not the cached one.
2851 Severity : enhancement
2853 Description: open file using fid
2854 Details : A file can be opened using just its fid, like
2855 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2858 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2861 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2862 Details : add proper locking for creation of jbd_4k slab cache
2866 Description: MMP check in ext3_remount() fails without displaying any error
2867 Details : When multiple mount protection fails during remount, proper error
2870 Severity : enhancement
2872 Description: Allow stripe size to be up to 4G-64k
2873 Details : Fix math logic to allow large stripe sizes.
2877 Description: add check for >8TB ldiskfs filesystems
2878 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2879 Don't allow >8TB ldiskfs filesystems to be mounted without
2880 force_over_8tb mount option
2885 Descriptoin: OSS complains "too many blocks" and crashes
2886 Details : In filter_commitrw_write():
2887 1. lquota_chkquota();
2888 2. obdo_from_inode();
2889 3. lquota_pending_commit();
2890 "2." has changed uid and gid in this case, then caused
2891 lquota_chkquota() and lquota_pending_commit() to use
2892 different uid and gid. That is the root of the bug.
2896 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2897 Details : Detect and print out the 'stripe count = -1' case properly.
2901 Description: Out or order replies might be lost on replay
2902 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2903 request with tid smaller than one currently being added, add it
2904 to the start, not end of the list.
2906 --------------------------------------------------------------------------------
2908 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2910 * Support for kernels:
2911 2.6.5-7.283 (SLES 9),
2912 2.6.9-55.EL (RHEL 4),
2913 2.6.16.46-0.14 (SLES 10),
2914 2.6.18.8 vanilla (kernel.org)
2915 * Client support for unpatched kernels:
2916 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2917 2.6.16 - 2.6.22 vanilla (kernel.org)
2918 * Due to recently discovered recovery problems, we do not recommend
2919 using patchless RHEL 4 clients with this or any earlier release.
2920 * Recommended e2fsprogs version: 1.39.cfs8
2921 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2922 * Starting with this release, the ldiskfs backing filesystem required
2923 by Lustre is now in its own package, lustre-ldiskfs. This package
2924 should be installed. It is versioned separately from Lustre and
2925 may be released separately in future.
2929 Description: block reactivating mgc import until all deactivates complete
2930 Details : Fix race when failing back MDT/MGS to itself (testing)
2933 Frequency : at statup only
2935 Description: mds_lov_synchronize race leads to various problems
2936 Details : simultaneous MDT->OST connections at startup can cause the
2937 sync to abort, leaving the OSC in a bad state.
2939 Severity : enhancement
2941 Description: add optional extra BUILD_VERSION info
2942 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2943 to override the lustre version.
2946 Frequency : 2.6.18 servers only
2948 Description: ll_kern_mount() doesn't release the module reference
2949 Details : The ldiskfs module reference count never drops down to 0
2950 because ll_kern_mount() doesn't release the module reference.
2955 Description: server LBUG when using old ost_num_threads parameter
2956 Details : Accept the old ost_num_threads parameter but warn that it
2957 is deprecated, and fix an off-by-one error that caused an LBUG.
2962 Description: Transient SCSI error results in persistent IO issue
2963 Details : iobuf->dr_error is not reinitialized to 0 between two
2967 Frequency : sometimes when underlying device returns I/O errors
2969 Description: OSTs not going read-only during write failures
2970 Details : OSTs are not remounted read-only when the journal commit threads
2971 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2972 instead of the ext3 wrappers.
2976 Description: poor connect scaling with increasing client count
2977 Details : Don't run filter_grant_sanity_check for more than 100 exports
2978 to improve scaling for large numbers of clients.
2981 Frequency : SLES10 only
2983 Description: sanity-quota.sh quotacheck failed: rc = -22
2984 Details : Quotas cannot be enabled on SLES10.
2987 Frequency : liblustre clients only
2989 Description: getdirentries does not give error when run on compute nodes
2990 Details : getdirentries does not fail when the size specified as an argument
2991 is too small to contain at least one entry
2993 Severity : enhancement
2995 Description: Add LNET router traceability for debug purposes
2996 Details : If a checksum failure occurs with a router as part of the
2997 IO path, the NID of the last router that forwarded the bulk data
2998 is printed so it can be identified.
3003 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3004 Details : Due to a race condition, liblustre clients were occasionally
3005 evicted incorrectly.
3007 Severity : enhancement
3009 Description: lfs setstripe use optional parameters instead of postional
3012 Severity : enhancement
3014 Description: Nanosecond timestamp support for ldiskfs
3015 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3016 resolution timestamps. There is not yet support for this at
3017 the Lustre filesystem level.
3020 Frequency : during server recovery
3022 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3023 Details : request with rq_no_resend flag not awake l_wait_event if they get
3027 Frequency : nfs export on patchless client
3029 Description: connectathon hang when test nfs export over patchless client
3030 Details : Disconnected dentry cannot be found with lookup, so we do not need
3031 to unhash it or make it invalid
3034 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3035 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3036 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3041 Description: open req refcounting wrong on reconnect
3042 Details : If reconnect happened between getting open reply from server and
3043 call to mdc_set_replay_data in ll_file_open, we will schedule
3044 replay for unreferenced request that we are about to free.
3045 Subsequent close will crash in variety of ways.
3046 Check that request is still eligible for replay in
3047 mdc_set_replay_data().
3052 Description: disable writes to filesystem when reading health_check file
3053 Details : the default for reading the health_check proc file has changed
3054 to NOT do a journal transaction and write to disk, because this
3055 can cause reads of the /proc file to hang and block HA state
3056 checking on a healthy but otherwise heavily loaded system. It
3057 is possible to return to the previous behaviour during configure
3058 with --enable-health-write.
3060 Severity : enhancement
3062 Description: 64-bit inode version
3063 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3064 to the inode. This will be required for version-based recovery.
3069 Description: MDS fails to start if a duplicate client export is detected
3070 Details : in some rare cases it was possible for a client to connect to
3071 an MDS multiple times. Upon recovery the MDS would detect this
3072 and fail during startup. Handle this more gracefully.
3074 Severity : enhancement
3076 Description: Add -o localflock option to simulate old noflock
3078 Details : This will achieve local-only flock/fcntl locks
3084 Description: log_commit_thread vs filter_destroy race leads to crash
3085 Details : Take import reference before releasing llog record semaphore
3090 Description: Wrong request locking in request set processing
3091 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3092 imp_delayed_list, in this place should be used imp_lock.
3095 Frequency : when reconnection
3097 Description: Grant Leak when osc reconnect to OST
3098 Details : When osc reconnect ost, OST(filter) should check whether it
3099 should grant more space to client by comparing fed_grant and
3100 cl_avail_grant, and return the granted space to client instead
3101 of "new granted" space, because client will call osc_init_grant
3102 to update the client grant space info.
3105 Frequency : when client reconnect to OST
3107 Description: Grant Leak when osc do resend and replay bulk write
3108 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3109 bulk write request, because the grant info while be sync between
3110 OSC and OST when reconnect, and we should ignore the grant info
3111 these of resend/replay write req.
3116 Description: Grant space more than available left space sometimes.
3117 Details : When then OST is about to be full, if two bulk writing from
3118 different clients came to OST. Accord the avaliable space of the
3119 OST, the first req should be permitted, and the second one
3120 should be denied by ENOSPC. But if the seconde arrived before
3121 the first one is committed. The OST might wrongly permit second
3122 writing, which will cause grant space > available space.
3125 Frequency : when client is evicted
3127 Description: Grant might be wrongly erased when osc is evicted by OST
3128 Details : when the import is evicted by server, it will fork another
3129 thread ptlrpc_invalidate_import_thread to invalidate the
3130 import, where the grant will be set to 0. While the original
3131 thread will update the grant it got when connecting. So if
3132 the former happened latter, the grant will be wrongly errased
3133 because of this race.
3138 Description: Checking Stale with correct fid
3139 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3140 to check whether it is stale, because sometimes, we want the
3141 enqueue happened anyway, and op_data.fid2 will not be initialized.
3143 Severity : enhancement
3145 Description: update patchless client
3146 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3149 Frequency : only with 2.4 kernel
3151 Description: random memory corruption
3152 Details : size of struct ll_inode_info is to big for union inode.u and this
3153 can be cause of random memory corruption.
3158 Description: Memory leak in recovery
3159 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3160 It should also check obd_fail before fsfilt_start, otherwise if
3161 fsfilt_start return -EROFS,(failover mds during mds recovery).
3162 then the req will return with repmsg->transno = 0 and rc = EROFS.
3163 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3164 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3165 be freed no matter whether fsfilt_commit success or not.
3168 Frequency : only with huge count clients
3170 Description: Prevents from taking the superblock lock in llap_from_page for
3172 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3173 for llap_from_page call in ll_removepage prevents from taking the
3174 superblock lock for a soon died page.
3179 Description: Not check open intent error before release open handle
3180 Details : in some rare cases, the open intent error is not checked before
3181 release open handle, which may cause
3182 ASSERTION(open_req->rq_transno != 0), because it tries to release
3183 the failed open handle.
3188 Description: Set cat log bitmap only after create log success.
3189 Details : in some rare cases, the cat log bitmap is set too early. and it
3190 should be set only after create log success.
3194 Description: Accessing a block bevice can re-enable I/O when Lustre is
3195 tearing down a device.
3196 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3201 Description: service threads may hog cpus when there are a lot of requests
3202 Details : Insert cond_resched to give other threads a chance to use some CPU
3207 Description: the cat log was not initialized in recovery
3208 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3209 unlink log on mds will not be initialized until mds post
3210 recovery. And also in mds post recovery, the unlink log will
3211 initialization will be done asynchronausly, so there will be race
3212 between add unlink log and unlink log initialization.
3216 Description: brw_stats were being printed incorrectly
3217 Details : brw_stats were being printed as log2 but all of them were not
3218 recorded as log2. Also remove some code duplication arising from
3219 filter_tally_{read,write}.
3223 Frequency : rare, only in recovery.
3224 Description: ASSERTION(req->rq_type != LI_POISON) failed
3225 Details : imp_lock should be held while iterating over imp_sending_list for
3226 prevent destroy request after get timeout in ptlrpc_queue_wait.
3230 Description: replay-single.sh test 52 fails
3231 Details : A lock's skiplist need to be cleanup when it being unlinked
3232 from its resource list.
3236 Description: Short directio read returns full requested size rather than
3238 Details : Direct I/O operations should return actual amount of bytes
3239 transferred rather than requested size.
3241 Severity : enhancement
3243 Description: metadata RPC reduction (e.g. for rm performance)
3244 Details : decrease the amount of synchronous RPC between clients and servers
3245 by canceling conflicing lock before the operation on the client
3246 and packing thier handles into the main operation RPC to server.
3248 Severity : enhancement
3250 Description: Async OSC create to avoid the blocking unnecessarily.
3251 Details : If an OST has no remaining object, system will block on the
3252 creation when it needs to create a new object on this OST. Now,
3253 ways use pre-created objects when available, instead of blocking on
3254 an empty osc while others are not empty. If we must block, we
3255 block for the shortest possible period of time.
3259 Description: improve handling recoverable errors
3260 Details : if request processig with error which can be recoverable on server
3261 request should be resend, otherwise page released from cache and
3264 Severity : enhancement
3266 Description: refine locking for avoid write wrong info into lov_objid file
3267 Details : fix possible races with add new target and write/update data in
3270 --------------------------------------------------------------------------------
3272 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3277 Frequency : on some architectures
3279 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3280 Details : Uninitialized flags sometimes cause configuration commands to
3284 Frequency : patchless clients only
3286 Description: missing __iget() symbol export
3287 Details : The __iget() symbol export is missing. To avoid the need for
3288 this on patchless clients the deathrow inode reaper is turned
3289 off, and we depend on the VM to clean up old inodes. This
3290 dependency was during via the fix for bug 12181.
3292 --------------------------------------------------------------------------------
3294 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3296 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3297 INTEROPERATE with older versions automatically. In many cases a
3298 special upgrade step is needed. Please read the
3299 user documentation before upgrading any part of a live system.
3300 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3301 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3302 * WARNING: Lustre configuration and startup changes are required with
3303 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3307 Severity : enhancement
3309 Description: MountConf
3310 Details : Lustre configuration is now managed via mkfs and mount
3311 commands instead of lmc and lconf. New obd types (MGS, MGC)
3312 are added for dynamic configuration management. See
3313 https://mail.clusterfs.com/wikis/lustre/MountConf for
3316 Severity : enhancement
3318 Description: dynamic OST addition
3319 Details : OSTs can now be added to a live filesystem
3321 Severity : enhancement
3323 Description: startup order invariance
3324 Details : MDTs and OSTs can be started in any order. Clients only
3325 require the MDT to complete startup.
3327 Severity : enhancement
3329 Description: parallel, asynchronous orphan cleanup
3330 Details : orphan cleanup is now performed in separate threads for each
3331 OST, allowing parallel non-blocking operation.
3333 Severity : enhancement
3335 Description: optimized stripe assignment
3336 Details : stripe assignments are now made based on ost space available,
3337 ost previous usage, and OSS previous usage, in order to try
3338 to optimize storage space and networking resources.
3340 Severity : enhancement
3342 Description: Permanently set tunables
3343 Details : All writable /proc/fs/lustre tunables can now be permanently
3344 set on a per-server basis, at mkfs time or on a live system.
3346 Severity : enhancement
3348 Description: Lustre message v2
3349 Details : Add lustre message format v2.
3351 Severity : enhancement
3353 Description: client OST exclusion list
3354 Details : Clients can be started with a list of OSTs that should be
3355 declared "inactive" for known non-responsive OSTs.
3359 Description: SPEC SFS validation failure on NFS v2 over lustre.
3360 Details : Changes the blocksize for regular files to be 2x RPC size,
3361 and not depend on stripe size.
3363 Severity : enhancement
3365 Description: Multiple MD RPCs in flight.
3366 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3367 To avoid overly-overloading MDS, introduce a limit on number of
3368 MDS RPCs in flight for a single client and add /proc controls
3369 to adjust this limit.
3371 Severity : enhancement
3373 Description: client read/write statistics
3374 Details : Add client read/write call usage stats for performance
3375 analysis of user processes.
3376 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3377 file access. extents_stats shows chunk size distribution.
3378 extents_stats_per_process show chunk size distribution per
3381 Severity : enhancement
3383 Description: mds statistics
3384 Details : Add detailed mds operations statistics in
3385 /proc/fs/lustre/mds/*/stats.
3389 Description: Failure of copying files with lustre special EAs.
3390 Details : Client side always return success for setxattr call for lustre
3391 special xattr (currently only "trusted.lov").
3396 Description: Refcount LNET uuids
3397 Details : The global LNET uuid list grew linearly with every startup;
3398 refcount repeated list entries instead of always adding to
3401 Severity : enhancement
3403 Description: Dynamic service threads
3404 Details : Within a small range, start extra service threads
3405 automatically when the request queue builds up.
3408 Frequency : mixed-endian client/server environments
3410 Description: mixed-endian crashes
3411 Details : The new msg_v2 system had some failures in mixed-endian
3414 Severity : enhancement
3416 Description: Easy OST removal
3417 Details : OSTs can be permanently deactivated with e.g. 'lctl
3418 conf_param lustre-OST0001.osc.active=0'
3420 Severity : enhancement
3422 Description: MGS proc entries
3423 Details : Added basic proc entries for the MGS showing what filesystems
3426 Severity : enhancement
3428 Description: provide MGS failover
3429 Details : Added config lock reacquisition after MGS server failover.
3431 Severity : enhancement
3433 Description: add Linux 2.4 support
3434 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3438 Description: a large application tries to do I/O to the same resource and dies
3439 in the middle of it.
3440 Details : Check the req->rq_arrival time after the call to
3441 ost_brw_lock_get(), but before we do anything about
3442 processing it & sending the BULK transfer request. This
3443 should help move old stale pending locks off the queue as
3444 quickly as obd_timeout.
3447 Frequency : when an incorrect nid is specified during startup
3449 Description: ptlrpc connect to non-existant node causes kernel crash
3450 Details : LNET can't be re-entered from an event callback, which
3451 happened when we expire a message after the export has been
3452 cleaned up. Instead, hand the zombie cleanup off to another
3455 Severity : enhancement
3457 Description: plain/inodebits lock performance improvement
3458 Details : Grouping plain/inodebits in granted list by their request modes
3459 and bits policy, thus improving the performance of search through
3463 Frequency : only if OST filesystem is corrupted
3465 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3466 Details : for a short time RPCs with bulk IO are in the replay list,
3467 but replay of bulk IOs is unimplemented. If the OST filesystem
3468 is corrupted due to disk cache incoherency and then replay is
3469 started it is possible to trip an assertion. Avoid putting
3470 committed RPCs into the replay list at all to avoid this issue.
3473 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3475 Description: System hang on startup
3476 Details : This bug allowed the liblustre (e.g. catamount) client to
3477 return to the app before handling all startup RPCs. This
3478 could leave the node unresponsive to lustre network traffic
3479 and manifested as a server ptllnd timeout.
3481 Severity : enhancement
3483 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3484 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3485 boolean to control whether to print peer debug info when a
3486 client's RPC times out.
3489 Frequency : only for kernels with patches from Lustre below 1.4.3
3491 Description: Remove old rdonly API
3492 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3495 Frequency : only for devices with external journals
3497 Description: Set external device read-only also
3498 Details : During a commanded failover stop, we set the disk device
3499 read-only while the server shuts down. We now also set any
3500 external journal device read-only at the same time.
3503 Frequency : when upgrading from 1.4 while trying to change parameters
3505 Description: The wrong (new) MDC name was used when setting parameters for
3506 upgraded MDT's. Also allows changing of OSC (and MDC)
3507 parameters if --writeconf is specified at tunefs upgrade time.
3510 Frequency : when setting specific ost indicies
3512 Description: QOS code breaks on skipped indicies
3513 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3514 created with --index need not be sequential.
3516 Severity : enhancement
3518 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3519 Details : The uninit_groups feature works in conjunction with the kernel
3520 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3521 up the pass1 processing of e2fsck. This is a read-only feature
3522 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3523 mount filesystems that have had this feature enabled.
3525 Severity : enhancement
3527 Description: Improve multi-block allocation algorithm to avoid fragmentation
3528 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3529 allocation locality and avoid filesystem fragmentation.
3531 ------------------------------------------------------------------------------
3533 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3535 * Support for kernels:
3536 2.4.21-47.0.1.EL (RHEL 3)
3537 2.6.5-7.283 (SLES 9)
3538 2.6.9-42.0.10.EL (RHEL 4)
3539 2.6.12.6 vanilla (kernel.org)
3540 2.6.16.27-0.9 (SLES 10)
3541 * Recommended e2fsprogs version: 1.39.cfs5
3543 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3547 Frequency : occasional, depends on client load and configuration
3548 Bugzilla : 12181, 12203
3549 Description: data loss for recently-modified files
3551 Details : In some cases it is possible that recently written or created
3552 files may not be written to disk in a timely manner (this should
3553 normally be within 30s unless client IO load is very high).
3554 The problem appears as zero-length files or files that are a
3555 multiple of 1MB in size after a client crash or client eviction
3556 that are missing data at the end of the file.
3558 This problem is more likely to be hit on clients where files are
3559 repeatedly created and unlinked in the same directory, clients
3560 have a large amount of RAM, have many CPUs, the filesystem has
3561 many OSTs, the clients are rebooted frequently, and/or the files
3562 are not accessed by other nodes after being written.
3564 The presence of the problem can be detected by looking at
3565 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3566 smaller than the second (nr_unused) then dirty files will not
3567 be flushed automatically to disk. "sync; sleep 10" should be
3568 run several times on the node before unmounting it to update
3569 Lustre (this is also safe to run on nodes without this problem).
3571 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3572 that can cause this same problem, so customers using that kernel
3573 also need to update the kernel in addition to Lustre. In order
3574 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3576 It is normal that files written just before a client crash (less
3577 than 30s) may not yet have been flushed to disk, even for local
3581 Frequency : frequent on thin XT3 nodes
3583 Description: UUID collision on thin XT3 Linux nodes
3584 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3585 randomly, since we relied on an insufficiently-seeded PRNG.
3590 Description: OSS hangs after "All ost request buffers busy"
3591 Details : A deadlock between quota and journal operations caused OSS
3592 hangs after printing "All ost request buffers busy."
3595 Frequency : always on liblustre builds
3597 Description: Cleanup compiler warnings on liblustre
3600 Frequency : always on liblustre builds on XT3
3602 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3606 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3607 > 32000 subdirectories
3608 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3609 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3610 subdirectory count crosses 32000. This will aid e2fsck to
3611 correctly handle more than 32000 subdirectories.
3614 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3616 Description: System hang on startup
3617 Details : This bug allowed the liblustre (e.g. catamount) client to
3618 return to the app before handling all startup RPCs. This
3619 could leave the node unresponsive to lustre network traffic
3620 and manifested as a server ptllnd timeout.
3622 Severity : enhancement
3624 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3625 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3626 boolean to control whether to print peer debug info when a
3627 client's RPC times out.
3632 Description: make O_SYNC working on 2.6 kernels
3633 Details : 2.6 kernels use different method for mark pages for write,
3634 so need add a code to lustre for O_SYNC work.
3639 Description: Failure to close file and release space on NFS
3640 Details : Put inode details into lock acquired in ll_intent_file_open.
3641 Use mdc_intent_lock in ll_intent_open to properly
3642 detect all kind of errors unhandled by mdc_enqueue.
3647 Description: proc file read during shutdown sometimes raced obd removal,
3649 Details : Add lock to prevent obd access after proc file removal.
3652 Frequency : Only for files larger than 4GB on 32-bit clients.
3654 Description: improperly doing page alignment of locks
3655 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3656 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3661 Description: RPCs being resent when they shouldn't be
3662 Details : Some RPCs that should not be resent are being resent. This
3663 can cause inconsistencies in the RPC state machine. Do not
3664 resend such requests.
3667 Frequency : rare, only with NFS export
3669 Description: Crash on NFS re-export node
3670 Details : under very unusual load conditions an assertion is hit in
3671 ll_intent_file_open()
3674 Frequency : only if OST filesystem is corrupted
3676 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3677 Details : for a short time RPCs with bulk IO are in the replay list,
3678 but replay of bulk IOs is unimplemented. If the OST filesystem
3679 is corrupted due to disk cache incoherency and then replay is
3680 started it is possible to trip an assertion. Avoid putting
3681 committed RPCs into the replay list at all to avoid this issue.
3686 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3687 Details : Large single O_DIRECT read and write calls can fail to allocate
3688 a sufficiently large buffer to process the request. In case of
3689 allocation failure the allocation is retried with a smaller
3690 buffer and broken into smaller requests.
3692 Severity : enhancement
3694 Description: Add -o localflock option to simulate old noflock behaviour.
3695 Details : This will achieve local-only flock/fcntl locks coherentness.
3700 Description: versioning check is incomplete
3701 Details : Checking the version difference of client vs. server, report
3702 error if the gap is too big.
3707 Description: add support PG_writeback bit
3708 Details : add support for PG_writeback bit for Lustre, for more carefull
3709 work with page cache in 2.6 kernel. This also fix some deadlocks
3710 and remove hack for work O_SYNC with 2.6 kernel.
3712 Severity : enhancement
3714 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3715 Details : The uninit_groups feature works in conjunction with the kernel
3716 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3717 up the pass1 processing of e2fsck. This is a read-only feature
3718 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3719 mount filesystems that have had this feature enabled.
3721 Severity : enhancement
3723 Description: Improve multi-block allocation algorithm to avoid fragmentation
3724 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3725 allocation locality and avoid filesystem fragmentation.
3727 ------------------------------------------------------------------------------
3729 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3731 * Support for kernels:
3732 2.6.9-42.0.3.EL (RHEL 4)
3733 2.6.5-7.276 (SLES 9)
3734 2.4.21-47.0.1.EL (RHEL 3)
3735 2.6.12.6 vanilla (kernel.org)
3736 2.6.16.21-0.8 (SLES10)
3737 * Recommended e2fsprogs version: 1.39.cfs2-0
3739 * The backwards-compatible /proc/sys/portals symlink has been removed
3740 in this release. Before upgrading, please ensure that you change
3741 any configuration scripts or /etc/sysctl.conf files that access
3742 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3743 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3744 in advance of the upgrade on any system running Lustre 1.4.6 or
3745 newer, since /proc/sys/lnet was added in that version.
3746 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3750 Frequency : only when quota is used
3752 Description: avoid scanning export list for quota master
3753 Details : Change the algorithms to avoid scanning export list in order
3754 to improve the efficiency.
3757 Frequency : MDS failover only, very rarely
3759 Description: "went back in time" messages on mds failover
3760 Details : The greatest transno may be lost when the current operation
3761 finishes with an error (transno==0) and the client's last_rcvd
3762 record is over-written. Save the greatest transno in the
3763 mds_last_transno for this case.
3766 Frequency : always for specific kernels and striping counts
3768 Description: client may get "Matching packet too big" without ACL support
3769 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3770 when trying to access files in certain configurations. The
3771 clients should in fact be denied when mounting because they do
3772 not understand ACLs.
3775 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3777 Description: many clients connecting with IO in progress causes connect timeouts
3778 Details : Avoid synchronous journal commits to avoid delays caused by many
3779 clients connecting/disconnecting when bulk IO is in progress.
3780 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3781 OST_IO_PORTAL to avoid delays behind potentially many pending
3785 Frequency : occasionally with multiple writers to a single file
3787 Description: shared writes to file may result in wrong size reported by stat()
3788 Details : Allow growing of kms when extent lock is cancelled
3791 Frequency : always with random mmap IO to multi-striped file
3793 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3794 Details : Only if the hole is at the end of OST object so that kms is too
3795 small. Fix is to increase kms accordingly in ll_nopage.
3798 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3800 Description: writes to a missing object would leak memory on the OST
3801 Details : If there is an inconsistency between the MDS and OST filesystems,
3802 such that the MDS references an object that doesn't exist, writes
3803 to that object will leak memory due to incorrect cleanup in the
3804 error handling path, eventually running out of memory on the OST.
3809 Description: Creating too long symlink causes lustre errors
3810 Details : Check symlink and name lengths before sending requests to MDS.
3813 Frequency : only if flock is enabled (not on by default)
3815 Description: posix locks not released on fd closure on 2.6.9+
3816 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3817 kernels, this caused such locks not to be released on fd close and
3818 then assertions on fs unmount about still used locks.
3821 Frequency : MDS failover only, very rarely
3823 Description: clients may get ASSERTION(granted_lock != NULL)
3824 Details : When request was taking a long time, and a client was resending
3825 a getattr by name lock request. The were multiple lock requests
3826 with the same client lock handle and
3827 mds_getattr_name->fixup_handle_for_resent_request found one of the
3828 lock handles but later failed with ASSERTION(granted_lock != NULL).
3833 Description: handle->h_buffer_credits > 0, assertion failure
3834 Details : h_buffer_credits is zero after truncate, causing assertion
3835 failure. This patch extends the transaction or creates a new
3839 Frequency : NFS re-export or patchless client
3840 Bugzilla : 11179, 10796
3841 Description: Crash on NFS re-export node (__d_move)
3842 Details : We do not want to hash the dentry if we don't have a lock.
3843 But if this dentry is later used in d_move, we'd hit uninitialised
3844 list head d_hash, so we just do this to init d_hash field but
3845 leave dentry unhashed.
3848 Frequency : NFS re-export or patchless client
3850 Description: NFS exports has problem with symbolic link
3851 Details : lustre client didn't properly install dentry when re-exported
3852 to NFS or running patchless client.
3855 Frequency : NFS re-export or patchless client
3857 Description: Various nfs/patchless fixes.
3858 Details : fixes reuse disconected alias for lookup process - this fixes
3859 warning "find_exported_dentry: npd != pd",
3860 fix permission error with open files at nfs.
3861 fix apply umask when do revalidate.
3864 Frequency : occasional
3866 Description: Crash on NFS re-export node
3867 Details : calling clear_page() on the wrong pointer triggered oops in
3868 generic_mapping_read().
3871 Frequency : rarely, using O_DIRECT IO
3873 Description: unaligned directio crashes client with LASSERT
3874 Details : check for unaligned buffers before trying any requests.
3877 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3879 Description: stale data returned from RAID cache
3880 Details : If only a small amount of IO is done to the RAID device before
3881 reading it again it is possible to get stale data from the RAID
3882 cache instead of reading it from disk.
3885 Frequency : always for sles10 kernel
3887 Description: sles10 support
3888 Details : ll_follow_link: compile fixes and using of nd_set_link
3889 under newer kernels.
3892 Frequency : depends on arch, kernel and compiler version, always on sles10
3895 Description: recursive or deep enough symlinks cause stack overflow
3896 Details : getting rid of large stack-allocated variable in
3900 Frequency : depends on hardware
3902 Description: lustre write performance loss in the SLES10 kernel
3903 Details : the performance loss is caused by using of write barriers in the
3904 ext3 code. The SLES10 kernel turns barrier support on by
3905 default. The fix is to undo that change for ldiskfs.
3907 ------------------------------------------------------------------------------
3909 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3911 * Support for kernels:
3912 2.6.9-42.0.3EL (RHEL 4)
3913 2.6.5-7.276 (SLES 9)
3914 2.4.21-47.0.1.EL (RHEL 3)
3915 2.6.12.6 vanilla (kernel.org)
3919 Frequency : quota enabled and large files being deleted
3921 Description: releasing more than 4GB of quota at once hangs OST
3922 Details : If a user deletes more than 4GB of files on a single OST it
3923 will cause the OST to spin in an infinite loop. Release
3924 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3929 Description: statfs data retrieved from /proc may be stale or zero
3930 Details : When reading per-device statfs data from /proc, in the
3931 {kbytes,files}_{total,free,avail} files, it may appear
3932 as zero or be out of date.
3935 Frequency : systems with MD RAID1 external journal devices
3937 Description: lconf's call to blkid is confused by RAID1 journal devices
3938 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3939 of returning all block devices that match the journal UUID.
3942 Frequency : always, for aggregate stripe size over 4GB
3944 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3945 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3946 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3949 Frequency : always if "lfs find" used on a local file/directory
3951 Description: "lfs find" segfaults if used on a local file/directory
3952 Details : The case where a directory component was not specified wasn't
3953 handled correctly. Handle this properly.
3956 Frequency : always on ppc64
3958 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3959 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3960 perform endianness conversion.
3963 Frequency : rarely (truncate to non-zero file size after write under load)
3964 Bugzilla : 10730, 10687
3965 Description: Files padded with zeros to next 4K multiple
3966 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3967 it is possible that files that are truncated to a non-zero size
3968 immediately after being written are filled with zero bytes beyond
3969 the truncated size. No file data is lost.
3971 Severity : enhancement
3973 Description: Allow recovery/failover for liblustre clients.
3974 Details : liblustre clients were unaware of failover configurations until
3977 Severity : enhancement
3979 Description: user file locks should fail when not mounting with flock option
3980 Details : Set up an error-returning stub in ll_file_operations.lock field
3981 to prevent incorrect behaviour when client is mounted without
3982 flock option. Also, set up properly f_op->flock field for
3986 Frequency : always on ia64
3988 Description: "lfs df" loops on printing out MDS statfs information
3989 Details : The obd_ioctl_data was not initialized and in some systems
3990 this caused a failure during the ioctl that did not return
3991 an error. Initialize the struct and return an error on failure.
3994 Frequency : SLES 9 only
3996 Description: Error of copying files with lustre special EAs as root
3997 Details : Client side always return success for setxattr call for lustre
3998 special xattr (currently only "trusted.lov").
4001 Frequency : rarely on clusters with both ia64+i386 clients
4003 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4004 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4005 clients are doing concurrent writes to the same file it is
4006 possible that the ia64 clients may LASSERT because the OST
4007 extent locks are not PAGE_SIZE aligned. Ensure that grown
4008 locks are always aligned on the request boundary.
4011 Frequency : specific use, occasional
4013 Description: Overwriting in use executable truncates on-disk binary image
4014 Details : If one node attempts to overwrite an executable in use by
4015 another node, we now correctly return ETXTBSY instead of
4016 truncating the file.
4018 Severity : enhancement
4020 Description: Async OSC create to avoid the blocking unnecessarily.
4021 Details : If an OST has no remaining object, system will block on the creating
4022 when need to create a new object on this OST. Now, ways use
4023 pre-created objects when available, instead of blocking on an
4024 empty osc while others are not empty. If we must block, we block
4025 for the shortest possible period of time.
4030 Description: chmod on Lustre root is propagated to other clients
4031 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4032 invalid by the follow_mount time.
4035 Frequency : liblustre clients only
4037 Description: Race in 'instant cancel' lock handling could lead to such locks
4038 never to be granted in case of SMP MDS
4039 Details : Do not destroy not yet granted but cbpending locks in
4043 Frequency : replay/resend of open
4045 Description: non null lock assetion failure in mds_intent_policy
4046 Details : Trying to replay/resend lockless open requests resulted in
4047 mds_open() returning 0 with no lock. Now it sets a flag if
4048 a lock is going to be returned.
4050 Severity : enhancement
4052 Description: Checksum enhancements
4053 Details : New checksum enhancements allow for resending RPCs that failed
4056 Severity : enhancement
4058 Description: Tunables on number of dirty pages in cacche
4059 Details : Allow to set limit on number of dirty pages cached.
4064 Description: client crash on unmount - lock still has references
4065 Details : In some error handling cases it was possible to leak a lock
4066 reference on a client while accessing a file. This was not
4067 harmful to the client during operation, but would cause the
4068 client to crash when the filesystem is unmounted.
4071 Frequency : specific case, rare
4073 Description: ETXTBSY on mds though file not in use by client
4074 Details : ETXTBSY is no longer incorrectly returned when attempting to
4075 chmod or chown a directory that the user previously tried to
4076 execute or a currently-executing binary.
4079 Frequency : extremely rare except on liblustre-based clients
4081 Description: Lustre space not freed when files are deleted
4082 Details : Clean up open-unlinked files after client eviction. Previously
4083 the unlink was skipped and the files remained as orphans.
4088 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4089 Details : In some cases it was possible to send a blocking callback to
4090 a client doing a glimpse, even though that client didn't get
4091 a lock granted. When the glimpse lock is cancelled on the OST
4092 the freed lock is left on the waiting list and corrupted the list.
4095 Frequency : all core dumps
4097 Description: Broke core dumps to lustre
4098 Details : Negative dentry may be unhashed if parent does not have UPDATE
4099 lock, but some callers, e.g. do_coredump, expect dentry to be
4100 hashed after successful create, hash it in ll_create_it.
4102 ------------------------------------------------------------------------------
4104 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4106 * Support for kernels:
4107 2.6.9-42.0.2.EL (RHEL 4)
4108 2.6.5-7.276 (SLES 9)
4109 2.4.21-40.EL (RHEL 3)
4110 2.6.12.6 vanilla (kernel.org)
4114 Frequency : always on RHEL 3
4116 Description: Number of open files grows over time
4117 Details : The number of open files grows over time, whether or not
4118 Lustre is started. This was due to a filp leak introduced
4119 by one of our kernel patches.
4121 ------------------------------------------------------------------------------
4123 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4125 * Support for kernels:
4126 2.6.9-42.EL (RHEL 4)
4127 2.6.5-7.276 (SLES 9)
4128 2.4.21-40.EL (RHEL 3)
4129 2.6.12.6 vanilla (kernel.org)
4134 Bugzilla : 5719, 9635, 9792, 9684
4135 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4136 Details : If a server is under heavy load and cannot reply to new
4137 connection requests before the client resends the (re)connect,
4138 the connection handling code can behave badly if two service
4139 threads are concurrently handing separate (re)connections from
4140 the same client. Add better locking to the connection handling
4141 code, and ensure that only a single connection will be processed
4142 for a given client UUID, even if the lock is dropped.
4144 Severity : enhancement
4146 Description: add TCP zero-copy support to kernel
4147 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4148 sends if the hardware supports scatter-gather and checksumming.
4149 This allows socklnd to do client-write and server-read more
4150 efficiently and reduce CPU utilization from skbuf copying.
4153 Frequency : only if NFS exporting from client
4155 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4156 Details : If setting attributes on a file created under NFS that had
4157 never been opened it would be possible to oops the client
4158 if the file had no objects.
4161 Frequency : always for liblustre
4163 Description: liblustre client does MDS+OSTs setattr RPC for each write
4164 Details : When doing a write from a liblustre client, the client
4165 incorrectly issued an RPC to the MDS and each OST the file was
4166 striped over in order to update the timestamps. When writing
4167 with small chunks and many clients this could overwhelm the MDS
4168 with RPCs. In all cases it would slow down the write because
4169 these RPCs are unnecessary.
4171 Severity : enhancement
4173 Description: allow number of MDS service threads to be changed at module load
4174 Details : It is now possible to change the number of MDS service threads
4175 running. Adding "options mds mds_num_threads={N}" to the MDS's
4176 /etc/modprobe.conf will set the number of threads for the next
4177 time Lustre is restarted (assuming the "mds" module is also
4178 reloaded at that time). The default number of threads will
4179 stay the same, 32 for most systems.
4184 Description: OST crash if filesystem is unformatted or corrupt
4185 Details : If an OST is started on a device that has never been formatted
4186 or if the filesystem is corrupt and cannot even mount then the
4187 error handling cleanup routines would dereference a NULL pointer.
4192 Description: NULL pointer deref in llap_from_page.
4193 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4194 incorrect) mapping if the page was truncated/reclaimed while it was
4195 searched for. Check for this condition and skip such pages when
4196 doing readahead. Introduce extra check to llap_from_page() to
4197 verify page->mapping->host is non-NULL (so page is not anonymous).
4200 Frequency : Sometimes when using sys_sendfile
4202 Description: "page not covered by a lock" warnings from ll_readpage
4203 Details : sendfile called ll_readpage without right page locks present.
4204 Now we introduced ll_file_sendfile that does necessary locking
4205 around call to generic_file_sendfile() much like we do in
4209 Frequency : with certain MDS communication failures at client mount time
4211 Description: NULL pointer deref after failed client mount
4212 Details : a client connection request may delayed by the network layer
4213 and not be sent until after the PTLRPC layer has timed out the
4214 request. If the client fails the mount immediately it will try
4215 to clean up before the network times out the request. Add a
4216 reference from the request import to the obd device and delay
4217 the cleanup until the network drops the request.
4220 Frequency : occasionally during client (re)connect
4222 Description: assertion failure during client (re)connect
4223 Details : processing a client connection request may be delayed by the
4224 client or server longer than the client connect timeout. This
4225 causes the client to resend the connection request. If the
4226 original connection request is replied in this interval, the
4227 client may trip an assertion failure in ptlrpc_connect_interpret()
4228 which thought it would be the only running connect process.
4231 Frequency : only with obd_echo servers and clients that are rebooted
4233 Description: kernel BUG accessing uninitialized data structure
4234 Details : When running an obd_echo server it did not start the ping_evictor
4235 thread, and when a client was evicted an uninitialized data
4236 structure was accessed. Start the ping_evictor in the RPC
4237 service startup instead of the OBD startup.
4239 Severity : enhancement
4240 Bugzilla : 10193 (patchless)
4241 Description: Remove dependency on various unexported kernel interfaces.
4242 Details : No longer need reparent_to_init, exit_mm, exit_files,
4243 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4246 Frequency : rare (only users of deprecated and unsupported LDAP config)
4248 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4249 Details : LDAP apparently contains 'lustreName' attributes instead of
4250 'name'. A simple remapping of the name is sufficient.
4253 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4255 Description: waiting_locks_callback trips kernel BUG if client is evicted
4256 Details : Running with the dump_on_timeout debug flag turned on makes
4257 it possible that the waiting_locks_callback() can try to dump
4258 the Lustre kernel debug logs from an interrupt handler. Defer
4259 this log dumping to the expired_lock_main() thread.
4261 Severity : enhancement
4263 Description: Support NFS exporting on 2.6 kernels.
4264 Details : Implement non-rawops metadata methods for NFS server to use without
4265 changing NFS server code.
4268 Frequency : very rare (synthetic metadata workload only)
4270 Description: two racing renames might cause an MDS thread to deadlock
4271 Details : Running the "racer" program may cause one MDS thread to rename
4272 a file from being the source of a rename to being the target of
4273 a rename at exactly the same time that another thread is doing
4274 so, and the second thread has already enqueued these locks after
4275 doing a lookup of the target and is trying to relock them in
4276 order. Ensure that we don't try to re-lock the same resource.
4279 Frequency : only very large systems with liblustre clients
4281 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4282 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4283 all OSTs in parallel. This allows the eviction of stale liblustre
4284 clients to proceed much faster than if they were done in series,
4285 and also offers similar improvements for other set_info RPCs.
4290 Description: excessive CPU usage during initial read phase on client
4291 Details : During the initial read phase on a client, it would agressively
4292 retry readahead on the file, consuming too much CPU and impacting
4293 performance (since 1.4.5.8). Improve the readahead algorithm
4294 to avoid this, and also improve some other common cases (read
4295 of small files in particular, where "small" is files smaller than
4296 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4301 Description: MDS crash when receiving packet with unknown intent.
4302 Details : Do not LBUG in unknown intent case, just return -EFAULT
4304 Severity : enhancement
4305 Bugzilla : 9293, 9385
4306 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4307 Details : Do not serialize getattr (non-intent version) and statfs.
4310 Frequency : occasional, when OST network is overloaded/intermittent
4312 Description: client evicted by OST after bulk IO timeout
4313 Details : If a client sends a bulk IO request (read or write) the OST
4314 may evict the client if it is unresposive to its data GET/PUT
4315 request. This is incorrect if the network is overloaded (takes
4316 too long to transfer the RPC data) or dropped the OST GET/PUT
4317 request. There is no need to evict the client at all, since
4318 the pinger and/or lock callbacks will handle this, and the
4319 client can restart the bulk request.
4322 Frequency : Always when mmapping file with no objects
4324 Description: client crashes when mmapping file with no objects
4325 Details : Check that we actually have objects in a file before doing any
4326 operations on objects in ll_vm_open, ll_vm_close and
4332 Description: Request leak when working with deleted CWD
4333 Details : Introduce advanced request refcount tracking for requests
4334 referenced from lustre intent.
4336 Severity : Enhancement
4338 Description: Cache open file handles on client.
4339 Details : MDS now will return special lock along with openhandle, if
4340 requested and client is allowed to hold openhandle, even if unused,
4341 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4342 closing files for every read/write openration.
4344 Severity : Enhancement
4346 Description: Cache open negative dentries on client when possible.
4347 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4348 negative dentries on lock revocation.
4353 Description: Remounting a client read-only wasn't possible with a zconf mount
4354 Details : It wasn't possible to remount a client read-only with llmount.
4356 Severity : enhancement
4357 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4358 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4359 MPICH is included to add Lustre-specific ADIO interfaces.
4360 This is based closely on the UFS ADIO layer and only differs
4361 in file creation, in order to allow the OST striping to be set.
4362 This is user-contributed code and not supported by CFS.
4367 Description: extended inode attributes (immutable, append-only) work improperly
4368 when 2.4 and 2.6 kernels are used on client/server or vice versa
4369 Details : Introduce kernel-independent values for these flags.
4371 Severity : enhancement
4374 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4375 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4376 granularity. Now, Fractional MB granularity is be supported,
4377 this is very useful for low memory system.
4379 Severity : enhancement
4381 Description: Getattr by fid
4382 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4383 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4384 have locks and avoids getattr rpc requests for it on every stat.
4387 Frequency : Always, for filesystems larger than 2TB
4389 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4390 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4391 filesystems larger than 2TB to be mounted with mballoc enabled.
4394 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4396 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4397 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4398 filesystems larger than 2TB is mounted then the kernel will
4399 silently corrupt the start of the filesystem. CONFIG_LBD is
4400 enabled for all CFS-supported kernels, but the possibility of
4401 this happening with a modified kernel config exists.
4403 Severity : enhancement
4405 Description: add client O_DIRECT support for 2.6 kernels
4406 Details : It is now possible to do O_DIRECT reads and writes to files
4407 in the Lustre client mountpoint on 2.6 kernel clients.
4409 Severity : enhancement
4411 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4412 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4413 parallel, not waiting for response from every OST before sending
4414 a rpc to the next OST.
4419 Description: setattr vs write race when updating file timestamps
4420 Details : Client processes that update a file timestamp into the past
4421 right after writing to the file (e.g. tar) it is possible that
4422 the updated file modification time can be reset to the current
4423 time due to a race between processing the setattr and write RPC.
4425 Severity : enhancement
4427 Description: Bring 'lfs find' closer in line with regular Linux find.
4428 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4429 -print0 options and obtains all the needed info through the lustre
4432 Severity : enhancement
4434 Description: support up to 1024 configured devices on one node
4435 Details : change obd_dev array from statically allocated to dynamically
4436 allocated structs as they are first used to reduce memory usage
4441 Description: Flush dirty partially truncated pages during truncate
4442 Details : Immediatelly flush partially truncated pages in filter_setattr,
4443 this way we completely avoid having any pages in page cache on OST
4444 and can retire ugly workarounds during writes to flush such pages.
4449 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4450 Details : protect inode from truncation within vfs_unlink() context
4451 just take a reference before calling vfs_unlink() and release it
4452 when parent's i_sem is free.
4457 Description: last_id value checked outside lock on OST caused LASSERT failure
4458 Details : If there were multiple MDS->OST object precreate requests in
4459 flight, it was possible that the OST's last object id was checked
4460 outside a lock and incorrectly tripped an assertion. Move checks
4461 inside locks, and discard old precreate requests.
4464 Frequency : always, if extents are used on OSTs
4466 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4467 Details : OSTs using the extents format would not zero the high 16 bits of
4468 the index physical block number. This is not a problem for any
4469 OST filesystems smaller than 16TB, and no kernels support ext3
4470 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4471 new/modified files) and can be fixed for existing filesystems
4472 with e2fsprogs-1.39-cfs1.
4477 Description: import connection selection may be incorrect if timer wraps
4478 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4479 import connections to be ignored if the 32-bit jiffies counter
4480 wraps. Use a 64-bit jiffies counter.
4483 Frequency : very large clusters immediately after boot
4485 Description: LNET request buffers exhausted under heavy short-term load
4486 Details : If a large number of client requests are generated on a service
4487 that has previously never seen so many requests it is possible
4488 that the request buffer growth cannot keep up with the spike in
4489 demand. Instead of dropping incoming requests, they are held in
4490 the LND until the RPC service can accept more requests.
4493 Frequency : Sometimes during replay
4495 Description: Assertion failure in ll_local_open after replay.
4496 Details : If replay happened on an open request reply before we were able
4497 to set replay handler, reply will become not swabbed tripping the
4498 assertion in ll_local_open. Now we set the handler right after
4499 recognising of open request
4502 Frequency : very rare
4504 Description: kernel reports "badness in vsnprintf"
4505 Details : Reading from the "recovery_status" /proc file in small chunks
4506 may cause a negative length in lprocfs_obd_rd_recovery_status()
4507 call to vsnprintf() (which is otherwise harmless). Exit early
4508 if there is no more space in the output buffer.
4510 Severity : enhancement
4512 Description: clear OBD RPC statistics by writing to them
4513 Details : It is now possible to clear the OBD RPC statistics by writing
4514 to the "stats" file.
4519 Description: Client mtime is not the same on different clients after utimes
4520 Details : In some cases, the client was using the utimes() syscall on
4521 a file cached on another node. The clients now validate the
4522 ctime from the MDS + OSTs to determine which one is right.
4527 Description: Inability to activate failout mode
4528 Details : lconf script incorrectly assumed that in pythong string's numeric
4529 value is used in comparisons.
4532 Frequency : always with multiple stripes per file
4534 Description: Inefficient object allocation for mutli-stripe files
4535 Details : When selecting which OSTs to stripe files over, for files with
4536 a stripe count that divides evenly into the number of OSTs,
4537 the MDS is always picking the same starting OST for each file.
4538 Return the OST selection heuristic to the original design.
4543 Description: mount failures may take full timeout to return an error
4544 Details : Under some heavy load conditions it is possible that a
4545 failed mount can wait for the full obd_timeout interval,
4546 possibly several minutes, before reporting an error.
4547 Instead return an error as soon as the status is known.
4549 Frequency : quota enabled and large files being deleted
4551 Description: releasing more than 4GB of quota at once hangs OST
4552 Details : If a user deletes more than 4GB of files on a single OST it
4553 will cause the OST to spin in an infinite loop. Release
4554 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4559 Description: statfs data retrieved from /proc may be stale or zero
4560 Details : When reading per-device statfs data from /proc, in the
4561 {kbytes,files}_{total,free,avail} files, it may appear
4562 as zero or be out of date.
4565 Frequency : systems with MD RAID1 external journal devices
4567 Description: lconf's call to blkid is confused by RAID1 journal devices
4568 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4569 of returning all block devices that match the journal UUID.
4572 Frequency : always, for aggregate stripe size over 4GB
4574 Description: assertion fails when trying to use 4GB stripe size
4575 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4576 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4579 Frequency : always on ppc64
4581 Description: the first write on an ext3 filesystem with mballoc got stuck
4582 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4583 perform endianness conversion.
4585 ------------------------------------------------------------------------------
4587 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4589 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4590 INTEROPERATE with older versions automatically. Please read the
4591 user documentation before upgrading any part of a live system.
4592 * WARNING: Lustre networking configuration changes are required with
4593 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4596 * Support for kernels:
4597 2.6.9-22.0.2.EL (RHEL 4)
4598 2.6.5-7.244 (SLES 9)
4599 2.6.12.6 vanilla (kernel.org)
4602 Severity : enhancement
4603 Bugzilla : 7981/8208
4604 Description: Introduced Lustre Networking (LNET)
4605 Details : LNET is new networking infrastructure for Lustre, it includes
4606 a reorganized network configuration mode (see the user
4607 documentation for full details) as well as support for routing
4608 between different network fabrics. Lustre Networking Devices
4609 (LNDs) for the supported network fabrics have also been
4610 created for this new infrastructure.
4612 Severity : enhancement
4613 Description: Introduced Access control lists
4614 Details : clients can set ACLs on files and directories in order to have
4615 more fine-grained permissions than the standard Unix UGO+RWX.
4616 The MDS must be started with the "-o acl" mount option.
4618 Severity : enhancement
4619 Description: Introduced filesystem quotas
4620 Details : Administrators may now establish per-user quotas on the
4623 Severity : enhancement
4625 Description: Configuration change for the XT3
4626 The PTLLND is now used to run Lustre over Portals on the XT3
4627 The configure option(s) --with-cray-portals are no longer used.
4628 Rather --with-portals=<path-to-portals-includes> is used to
4629 enable building on the XT3. In addition to enable XT3 specific
4630 features the option --enable-cray-xt3 must be used.
4635 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4636 Details : A race between allocating a new llcd and re-getting the llcd_lock
4637 allowed another thread to grab newly-allocated llcd.
4639 Severity : enhancement
4641 Description: 2.6 OST async journal commit and locking fix to improve performance
4642 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4643 2.6 kernels are now async as they already were in 2.4 kernels so
4644 that they can commit concurrently with the network bulk transfer.
4645 For block-allocated files the filter allocation semaphore is held
4646 to avoid filesystem fragmentation during allocation. BKL lock
4647 removed for 2.6 xattr operations where it is no longer needed.
4652 Description: lconf incorrectly determined whether two IP networks could talk
4653 Details : In some more complicated routing and multiple-network
4654 configurations, lconf will avoid trying to make a network
4655 connection to a disjoint part of the IP space. It was doing the
4656 math incorrectly for one set of cases.
4661 Description: Fix for potential infinite loop processing records in an llog.
4662 Details : If an llog record is corrupted/zeroed, it is possible to loop
4663 forever in llog_process(). Validate the llog record length
4664 and skip the remainder of the block on error.
4667 Frequency : occasional (liblustre only)
4669 Description: liblustre could not open files whose last component is a symlink
4670 Details : sysio_path_walk() would incorrectly pass the open intent to
4671 intermediate path components.
4674 Frequency : rare (liblustre only with non-standard tuning)
4675 Bugzilla : 7201 (7350)
4676 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4677 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4678 after it has been released, only do this for referenced locks
4680 Severity : enhancement
4682 Description: specifying an (invalid) directory default stripe_size of -1
4683 would reset the directory default striping
4684 Details : stripe_size -1 was used internally to signal directory stripe
4685 removal, now use "all default" to signal dir stripe removal
4686 as a directory striping of "all default" is not useful
4689 Frequency : common for large clusters running liblustre clients
4691 Description: doing an ls when liblustre clients are running is slow
4692 Details : sending a glimpse AST to a liblustre client waits for every AST
4693 to time out, as liblustre clients will not respond. Since they
4694 cannot cache data we refresh the OST lock LVB from disk instead.
4696 Severity : enhancement
4698 Description: doing an ls at the same time as file IO can be slow
4699 Details : enqueue and other "small" requests can be blocked behind many
4700 large IO requests. Create a new OST IO portal for non-IO
4701 requests so they can be processed faster.
4704 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4706 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4707 created file with mode 000
4708 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4709 allow an ftruncate() as a non-root user to a file with mode 000.
4710 Lustre now respects this flag to disable mode checks when
4711 truncating a file owned by the user
4714 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4717 Description: Revoking locks from clients that went dead or catatonic might take
4719 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4720 cancellation of such locks instant on servers without waiting for
4721 any reply from clients. Clients drops these locks when cancel
4722 notification from server is received without replying.
4725 Frequency : liblustre-only, when liblustre client dies or becomes busy
4727 Description: Doing ls on Linux clients can take a long time with active
4729 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4730 granting such locks to it in the first place if possible. Locks
4731 are taken by proxy on the OST during the read or write and
4732 dropped immediately afterward. Add connect flags handling, do
4733 not grant locks to liblustre clients for glimpse ASTs.
4735 Severity : enhancement
4737 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4738 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4739 sequential reads and behaves badly with applications doing random
4740 reads. Improve it by reading ahead at least read region, and
4741 avoiding excessive large RPC for small reads.
4743 Severity : enhancement
4745 Description: Creating more than 1000 files for a single job may cause a load
4746 imbalance on the OSTs if there are also a large number of OSTs.
4747 Details : qos_prep_create() uses an OST index reseed value that is an
4748 even multiple of the number of available OSTs so that if the
4749 reseed happens in the middle of the object allocation it will
4750 still utilize the OSTs as uniformly as possible.
4755 Description: OST or MDS may oops in ping_evictor_main()
4756 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4757 but doesn't restart at beginning of obd_exports_timed list
4760 Severity : enhancement
4762 Description: improve by-nid export eviction on the MDS and OST
4763 Details : allow multiple exports with the same NID to be evicted at one
4764 time without re-searching the exports list.
4767 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4769 Description: MDS may oops in groups_free()
4770 Details : in rare race conditions a newly allocated group_info struct is
4771 freed again, and this can be NULL. The 2.4 compatibility code
4772 for groups_free() checked for a NULL pointer, but 2.6 did not.
4775 Frequency : common for liblustre clients doing little filesystem IO
4776 Bugzilla : 9352, 7313
4777 Description: server may evict liblustre clients accessing contended locks
4778 Details : if a client is granted a lock or receives a completion AST
4779 with a blocking AST already set it would not reply to the AST
4780 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4783 Frequency : lfs setstripe, only systems with more than 160 OSTs
4785 Description: unable to set striping with a starting offset beyond OST 160
4786 Details : llapi_create_file() incorrectly limited the starting stripe
4787 index to the maximum single-file stripe count.
4790 Frequency : LDAP users only
4792 Description: lconf did not handle in-kernel recovery with LDAP properly
4793 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4795 Severity : enhancement
4797 Description: bind OST threads to NUMA nodes to improve performance
4798 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4799 node and do their allocations there to localize memory access
4801 Severity : enhancement
4803 Description: llmount can determine client NID directly from Myrinet (GM)
4804 Details : the client NID code from gmnalnid was moved directly into
4805 llmount, removing the need to use this or specifying the
4806 client NID explicitly when mounting GM clients with zeroconf
4809 Frequency : if client is started with down MDS
4811 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4812 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4813 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4814 ptlrpc_import_delayed_req() because we didn't check if the
4815 request was interrupted, and we also didn't break out of the
4816 event loop if there was no timeout
4821 Description: data loss during non-page-aligned writes to a single file from
4822 both multiple nodes and multiple threads on one node at same time
4823 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4824 inconsistency led to false short-reads, that were cached and later
4825 used by ->prepare_write() to fill in partially written page,
4826 leading to data loss.
4829 Frequency : always, if lconf --abort_recovery used
4831 Description: lconf --abort_recovery fails with 'Operation not supported'
4832 Details : lconf was attempting to abort recovery on the MDT device and not
4835 Severity : enhancement
4837 Description: remove cleanup logs
4838 Details : replace lconf-generated cleanup logs with lustre internal
4839 cleanup routines. Eliminates the need for client-cleanup and
4842 Severity : enhancement
4844 Description: add support for EAs (user and system) on lustre filesystems
4845 Details : it is now possible to store extended attributes in the Lustre
4846 client filesystem, and with the user_xattr mount option it
4847 is possible to allow users to store EAs on their files also
4849 Severity : enhancement
4851 Description: Add possibility (config option) to show minimal available OST free
4853 Details : When compiled with --enable-mindf configure option, statfs(2)
4854 (and so, df) will return least minimal free space available from
4855 all OSTs as amount of free space on FS, instead of summary of
4856 free spaces of all OSTs.
4858 Severity : enhancement
4860 Description: do not expand extent locks acquired on OST-side
4861 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4862 by server: they are not cached anyway.
4865 Frequency : when mmap is used/binaries executed from Lustre
4867 Description: Unmmap pages before throwing them away from read cache.
4868 Details : llap_shrink cache now attempts to unmap pages before discarding
4869 them (if unmapping failed - do not discard). SLES9 kernel has
4870 extra checks that trigger if this unmapping is not done first.
4875 Description: lconf didn't resolve symlinks before checking to see whether a
4876 given mountpoint was already in use
4879 Frequency : when migrating failover services
4880 Bugzilla : 6395, 9514
4881 Description: When migrating a subset of services from a node (e.g. failback
4882 from a failover service node) the remaining services would
4883 time out and evict clients.
4884 Details : lconf --force (implied by --failover) sets the global obd_timeout
4885 to 5 seconds in order to quickly disconnect, but this caused
4886 other RPCs to time out too quickly. Do not change the global
4887 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4889 Severity : enhancement
4890 Frequency : if MDS is started with down OST
4891 Bugzilla : 9439,5706
4892 Description: Allow startup/shutdown of an MDS without depending on the
4893 availability of the OSTs.
4894 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4895 Add appropriate locking and lov-osc refcounts for safe
4896 cleaning. Add osc abort_inflight calls in case the
4897 synchronize never started.
4900 Frequency : occasional (Cray XT3 only)
4902 Description: root not authorized to access files in CRAY_PORTALS environment
4903 Details : The client process capabilities were not honoured on the MDS in
4904 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4905 been accessed by an authorized user then root was able to access
4906 the file on the local client also. The root user capabilities
4907 are now allowed on the MDS, as this environment has secure UID.
4910 Frequency : occasional
4912 Description: ldiskfs "too long searching" message happens too often
4913 Details : A debugging message (otherwise harmless) prints too often on
4914 the OST console. This has been reduced to only happen when
4915 there are fragmentation problems on the filesystem.
4920 Description: Division by zero in statfs when all OSCs are inactive
4921 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4922 lov_statfs divides by value returned by lov_get_stripecnt().
4926 Bugzilla : 9489, 3273
4927 Description: First write from each client to each OST was only 4kB in size,
4928 to initialize client writeback cache, which caused sub-optimal
4929 RPCs and poor layout on disk for the first written file.
4930 Details : Clients now request an initial cache grant at (re)connect time
4931 and so that they can start streaming writes to the cache right
4932 away and always do full-sized RPCs if there is enough data.
4933 If the OST is rebooted the client also re-establishes its grant
4934 so that client cached writes will be honoured under the grant.
4939 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4940 Details : Now I/O RPCs go to different portal number and (presumably) fast
4941 lock requests (and glimses) and other RPCs get their own service
4942 threads pool that should be able to service those RPCs
4945 Severity : enhancement
4947 Description: Ability to exchange lustre version between client and servers and
4948 issue warnings at client side if client is too old. Also for
4949 liblustre clients there is ability to refuse connection of too old
4951 Details : New 'version' field is added to connect data structure that is
4952 filled with version info. That info is later checked by server and
4956 Frequency : rare, liblustre only.
4957 Bugzilla : 9296, 9581
4958 Description: Two simultaneous writes from liblustre at offset within same page
4959 might proceed at the same time overwriting eachother with stale
4961 Details : I/O lock withing llu_file_prwv was released too early, before data
4962 actually was hitting the wire. Extended lock-holding time until
4963 server acknowledges receiving data.
4966 Frequency : extremely rare. Never observed in practice.
4968 Description: avoid generating lustre_handle cookie of 0.
4969 Details : class_handle_hash() generates handle cookies by incrementing
4970 global counter, and can hit 0 occasionaly (this is unlikely, but
4971 not impossible, because initial value of cookie counter is
4972 selected randonly). Value of 0 is used as a sentinel meaning
4973 "unassigned handle" --- avoid it. Also coalesce two critical
4974 sections in this function into one.
4976 Severity : enhancement
4978 Description: allow liblustre clients to delegate truncate locking to OST
4979 Details : To avoid overhead of locking, liblustre client instructs OST to
4980 take extent lock in ost_punch() on client's behalf. New connection
4981 flag is added to handle backward compatibility.
4983 Severity : enhancement
4984 Bugzilla : 4928, 7341, 9758
4985 Description: allow number of OST service threads to be specified
4986 Details : a module parameter allows the number of OST service threads
4987 to be specified via "options ost ost_num_threads={N}" in the
4988 OSS's /etc/modules.conf or /etc/modprobe.conf.
4992 Bugzilla : 6146, 9635, 9895
4993 Description: servers crash with bad pointer in target_handle_connect()
4994 Details : In rare cases when a client is reconnecting it was possible that
4995 the connection request was the last reference for that export.
4996 We would temporarily drop the export reference and get a new
4997 one, but this may have been the last reference and the export
4998 was just destroyed. Get new reference before dropping old one.
5000 Severity : enhancement
5001 Frequency : if client is started with failover MDS
5003 Description: Allow multiple MDS hostnames in the mount command
5004 Details : Try to read the configuration from all specified MDS
5005 hostnames during a client mount in case the "primary"
5008 Severity : enhancement
5010 Description: Stop sending data to evicted clients as soon as possible.
5011 Details : Check if the client we are about to send or are sending data to
5012 was evicted already. (Check is done every second of waiting,
5013 for which l_wait_event interface was extended to allow checking
5014 of exit condition at specified intervals).
5017 Frequency : rare, normally only when NFS exporting is done from client
5019 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5021 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5024 Frequency : always, if srand() is called before liblustre initialization
5026 Description: Liblustre uses system PRNG disturbing its usage by user application
5027 Details : Introduce internal to lustre fast and high-quality PRNG for
5028 lustre usage and make liblustre and some other places in generic
5029 lustre code to use it.
5031 Severity : enhancement
5032 Bugzilla : 9477, 9557, 9870
5033 Description: Verify that the MDS configuration logs are updated when xml is
5034 Details : Check if the .xml configuration logs are newer than the config
5035 logs stored on the MDS and report an error if this is the case.
5036 Request --write-conf, or allow starting with --old_conf.
5038 Severity : enhancement
5040 Description: Handle symlinks in the path when checking if Lustre is mounted.
5041 Details : Resolve intermediate symlinks when checking if a client has
5042 mounted a filesystem to avoid duplicate client mounts.
5047 Description: lconf can hit an error exception but still return success.
5048 Details : The lconf command catches the Command error exception at the top
5049 level script context and will exit with the associated exit
5050 status, but doesn't ensure that this exit status is non-zero.
5055 Description: failure of ptlrpc thread startup can cause oops
5056 Details : Starting a ptlrpc service thread can fail if there are a large
5057 number of threads or the server memory is very fragmented.
5058 Handle this without oopsing.
5061 Frequency : always, only if liblustre and non-default acceptor port was used
5063 Description: liblustre cannot connect to servers with non-default acceptor port
5064 Details : tcpnal_set_default_params() was not called and was therefore
5065 ignoring the environment varaible TCPNAL_PORT, as well as other
5066 TCPNAL_ environment variables
5071 Description: two objects could be created on the same OST for a single file
5072 Details : If an OST is down, in some cases it was possible to create two
5073 objects on a single OST for a single file. No problems other
5074 than potential performance impact and spurious error messages.
5078 Bugzilla : 5681, 9562
5079 Description: Client may oops in ll_unhash_aliases
5080 Details : Client dcache may become inconsistent in race condition.
5081 In some cases "getcwd" can fail if the current directory is
5087 Description: Inode refcounting problems in NFS export code
5088 Details : link_raw functions used to call d_instantiate without obtaining
5089 extra inode reference first.
5093 Bugzilla : 9942, 9903
5094 Description: Referencing freed requests leading to crash, memleaks with NFS.
5095 Details : We used to require that call to ll_revalidate_it was always
5096 followed by ll_lookup_it. Also with revalidate_special() it is
5097 possible to call ll_revalidate_it() twice for the same dentry
5098 even if first occurence returned success. This fix changes semantic
5099 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5100 reference on a request referred from the intent.
5101 ll_intent_release() then releases such a request.
5104 Frequency : rare, normally benchmark loads only
5106 Description: unlinked inodes were kept in memory on the client
5107 Details : If a client is repeatedly creating and unlinking files it
5108 can accumulate a lot of stale inodes in the inode slab cache.
5109 If there is no other client load running this can cause the
5110 client node to run out of memory. Instead flush old inodes
5111 from client cache that have the same inode number as a new inode.
5114 Frequency : SLES9 2.6.5 kernel and long filenames only
5115 Bugzilla : 9969, 10379
5116 Description: utime reports stale NFS file handle
5117 Details : SLES9 uses out-of-dentry names in some cases, which confused
5118 the lustre dentry revalidation. Change it to always use the
5122 Frequency : rare, unless heavy write-truncate concurrency is continuous
5123 Bugzilla : 4180, 6984, 7171, 9963, 9331
5124 Description: OST becomes very slow and/or deadlocked during object unlink
5125 Details : filter_destroy() was holding onto the parent directory lock
5126 while truncating+unlinking objects. For very large objects this
5127 may block other threads for a long time and slow overall OST
5128 responsiveness. It may also be possible to get a lock ordering
5129 deadlock in this case, or run out of journal credits because of
5130 the combined truncate+unlink. Solution is to do object truncate
5131 first in one transaction without parent lock, and then do the
5132 final unlink in a new transaction with the parent lock. This
5133 reduces the lock hold time dramatically.
5136 Frequency : rare, 2.4 kernels only
5138 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5139 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5140 before unmounting their filesystem in order to avoid deadlock.
5141 The kernel_locked() function in 2.4 kernels only checks whether
5142 the kernel lock is held, not whether it is this process that is
5143 holding it as 2.6 kernels do.
5148 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5149 Details : The client ptlrpc code may be trying to reconnect to a down
5150 server before a previous connection attempt has timed out.
5151 Increase the reconnect interval to be longer than the connection
5152 timeout interval to avoid sending duplicate connections to
5156 Frequency : echo_client brw_test command
5158 Description: fix echo_client to work with OST preallocated code
5159 Details : OST preallocation code (5137) didn't take echo_client IO path
5160 into account: echo_client calls filter methods outside of any
5161 OST thread and, hence, there is no per-thread preallocated
5162 pages and buffers to use. Solution: hijack pga pages for IO. As
5163 a byproduct, this avoids unnecessary data copying.
5167 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5168 Description: Client can oops in mdc_commit_close() after open replay
5169 Details : It was possible for the MDS to return an open request with no
5170 transaction number in mds_finish_transno() if the client was
5171 evicted, but without actually returning an error. Clients
5172 would later try to replay that open and may trip an assertion
5173 Simplify the client close codepath, and always return an error
5174 from the MDS in case the open is not successful.
5177 Frequency : rare, 2.6 OSTs only
5179 Description: OST may deadlock under high load on fragmented files
5180 Details : If there was a heavy load and highly-fragmented OST filesystems
5181 it was possible to have all the OST threads deadlock waiting on
5182 allocation of biovecs, because the biovecs were not released
5183 until the entire RPC IO was completed. Instead, release biovecs
5184 as soon as they are complete to ensure forward IO progress.
5186 Severity : enhancement
5188 Description: Support for specifying external journal device at mount
5189 Details : If an OST or MDS device is formatted with an external journal
5190 device, this device major/minor is stored in the ext3 superblock
5191 and may not be valid for failover. Allow detecting and
5192 specifying the external journal at mount time.
5197 Description: Mounting an MDS with pending unlinked files may cause oops
5198 Details : target_finish_recovery() calls mds_postrecov() which returned
5199 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5200 considers this an error and immediately begins cleaning up the
5201 lov, just after starting the mds_lov process
5203 Severity : enhancement
5205 Description: Implement 'lfs df' to report actual free space on per-OST basis
5206 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5207 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5208 usage of objects; '-h' to report in human readable format.
5210 ------------------------------------------------------------------------------
5212 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5219 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5220 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5221 at startup during orphan unlinking. Instead, skip buddy update
5222 before setup, as it will be regenerated after recovery is complete.
5227 Description: If an OST is inactive, its locks might reference stale inodes.
5228 Details : lov_change_cbdata() must iterate over all namespaces, even if
5229 they are inactive to clear inode references from the lock.
5231 Severity : enhancement
5232 Frequency : occasional, if non-standard max_dirty_mb used
5234 Description: Client will block write RPCs if not enough grant
5235 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5236 then the client will block writes while waiting for another RPC
5237 to complete instead of consuming its dirty limit. With change
5238 we get improved performance when max_dirty_mb is small.
5240 Severity : enhancement
5241 Bugzilla : 3389, 6253
5242 Description: Add support for supplementary groups on the MDS.
5243 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5244 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5245 lookups for user supplementary groups into a cache.
5249 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5250 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5251 for liblustre O_EXCL is enforced by MDS.
5255 Description: readdir never returns NULL in liblustre.
5256 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5257 dentry in struct dirent.
5261 Frequency : liblustre only, depends on application IO pattern
5262 Description: liblustre clients evicted if not contacting servers
5263 Details : Don't put liblustre clients into the ping_evictor list, so
5264 they will not be evicted by the pinger ever.
5266 Severity : enhancement
5268 Description: Add ability to evict clients by NID from MDS.
5269 Details : By echoing "nid:$NID" string into
5270 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5271 $NID would be instantly evicted from this MDS and from all active
5272 OSTs connected to it.
5276 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5277 Details : lookup_it_finish() used to query file size from OSTs that was not
5282 Description: service threads change working directory to that of init
5283 Details : Starting lustre service threads may pin the working directory
5284 of the parent thread, making that filesystem busy. Threads
5285 now change to the working directory of init to avoid this.
5289 Frequency : during shutdown only
5290 Description: shutdown with a failed MDS or OST can cause unmount to hang
5291 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5296 Frequency : 2.6 only
5297 Description: chmod/chown may include an extra supplementary group
5298 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5299 supplementary group and if none is specified this is used.
5302 Bugzilla : 5479 (6816)
5304 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5305 Details : If lookup is in progress on a file that is unlinked we might try
5306 to revalidate the inode and fail in revalidate after lookup is
5307 complete and ll_file_open() enqueues the open again but
5308 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5311 Frequency : always, if lconf --abort_recovery used
5313 Description: lconf --abort_recovery fails with 'Operation not supported'
5314 Details : lconf was attempting to abort recovery on the MDT device and not
5317 ------------------------------------------------------------------------------
5319 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5324 Frequency : rare (only unsupported configurations with a node running as an
5326 Bugzilla : 6514, 5137
5327 Description: Mounting a Lustre file system on a node running as an OST could
5329 Details : OSTs now preallocates memory needed to write out data at
5330 startup, instead of when needed, to avoid having to
5331 allocate memory in possibly low memory situations.
5332 Specifically, if the file system is mounted on on OST,
5333 memory pressure could force it to try to write out data,
5334 which it needed to allocate memory to do. Due to the low
5335 memory, it would be unable to do so and the node would
5336 become unresponsive.
5338 Severity : enhancement
5340 Description: Addition of lconf --service command line option
5341 Details : lconf now accepts a '--service <arg>' option, which is
5342 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5344 Severity : enhancement
5346 Description: Failover mode is now the default for OSTs.
5347 Details : By default, OSTs will now run in failover mode. To return to
5348 the old behaviour, add '--failout' to the lmc line for OSTs.
5350 Severity : enhancement
5352 Description: Health checks are now provided for MDS and OSTs
5353 Details : Additional detailed health check information on MSD and OSTs
5354 is now provided through the procfs health_check value.
5357 Frequency : occasional, depends on IO load
5359 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5360 after numerous create/delete cycles
5361 Details : The ext3 inode allocation policy would not allocate new inodes
5362 very well on the OSTs because there are no new directories
5363 being created. Instead we look for groups with free space if
5364 the parent directories are nearly full.
5369 Description: Network or server problems during mount may cause partially
5370 mounted clients instead of returning an error.
5371 Details : The config llog parsing code may overwrite the error return
5372 code during mount error handling, returning success instead
5378 Description: MDS can fail to allocate large reply buffers
5379 Details : After long uptimes the MDS can fail to allocate large reply
5380 buffers (e.g. zconf client mount config records) due to memory
5381 fragmentation or consumption by the buffer cache. Preallocate
5382 some large reply buffers so that these replies can be sent even
5383 under memory pressure.
5387 Frequency : rare (liblustre)
5388 Description: fsx running with liblustre complained that using truncate() to
5389 extend the file doesn't work. This patch corrects that issue.
5390 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5391 ATTR_SIZE and lsm use in llu_setattr_raw.
5395 Frequency : rare, only 2.6 kernels
5396 Description: Unusual file access patterns on the MDS may result in inode
5397 data being lost in very rare circumstances.
5398 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5399 code in the 2.6 kernel caused the inode and/or EA data not to
5400 be read from disk, causing single-file corruption.
5404 Frequency : rare, only 2.6 filesystems using extents
5405 Description: Heavy concurrent write and delete load may cause data corruption.
5406 Details : It was possible under high-load situations to have an extent
5407 metadata block in the block device cache from a just-unlinked
5408 file overwrite a newly-allocated data block. We now unmap any
5409 metadata buffers that alias just-allocated data blocks.
5413 Frequency : filesystems with default stripe_count larger than 77
5414 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5415 Details : lconf specifies an inode size of 4096 bytes when the default
5416 stripe_count is larger than 77. This conflicts with the default
5417 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5418 this case to avoid pinning too much memory for large EAs.
5420 ------------------------------------------------------------------------------
5422 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5427 Frequency : rare (extremely heavy IO load with hundreds of clients)
5429 Description: Client is evicted, gets IO error writing to file
5430 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5431 caused clients to be evicted for AST timeouts. The fixes in
5432 bug 5192 mean we no longer need to have such short AST timeouts
5433 so ldlm_timeout has been increased.
5436 Frequency : occasional during --force or --failover shutdown under load
5437 Bugzilla : 5949, 4834
5438 Description: Server oops/LBUG if stopped with --force or --failover under load
5439 Details : a collection of import/export refcount and cleanup ordering
5440 issues fixed for safer force cleanup
5443 Frequency : only filesystems larger than 120 OSTs
5444 Bugzilla : 5990, 6223
5445 Description: lfs getstripe would oops on a very large filesystem
5446 Details : lov_getconfig used kfree on vmalloc'd memory
5449 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5450 Bugzilla : 6242, 6243
5451 Description: reading from files that had been truncated to a non-zero size
5452 but never opened returned no data
5453 Details : ll_file_read() reads zeros from no-object files to EOF
5458 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5459 Details : There's little harm in aborting MDS/OSS recovery and letting it
5460 try again, so I removed the LASSERT and return an error instead.
5462 Severity : enhancement
5464 Description: New debugging infrastructure for tracking down data corruption
5465 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5466 (b) cover more of the client-side code path, and (c) try to narrow
5467 down where problems occurred
5471 Bugzilla : 3819, 4364, 4397, 6313
5472 Description: Racing close and eviction MDS could cause assertion in mds_close
5473 Details : It was possible to get multiple mfd references during close and
5474 client eviction, leading to one thread referencing a freed mfd.
5476 Severity: : enhancement
5477 Bugzilla : 3262, 6359
5478 Description: Attempts to reconnect to servers are now more aggressive.
5479 Details : This builds on the enhanced upcall-less recovery that was added
5480 in 1.4.2. When trying to reconnect to servers, clients will
5481 now try each server in the failover group every 10 seconds. By
5482 default, clients would previously try one server every 25 seconds.
5487 Description: After recovery, certain operations trigger a failed
5488 assertion on a client.
5489 Details : Failing over an mds, using lconf -d --failover, while a
5490 client was doing a readdir() call would cause the client to
5491 LBUG after recovery completed and the readdir() was resent.
5493 Severity : enhancement
5495 Description: Default groups are now added by lconf
5496 Details : You can now run lconf --group <servicename> without having to
5497 manually add groups with lmc.
5500 Frequency : occasional
5502 Description: Nodes with an elan id of 0 trigger a failed assertion
5505 Frequency : always when accessing e.g. tty/console device nodes
5507 Description: tty and some other devices nodes cannot be used on lustre
5508 Details : file's private_data field is used by device data and lustre
5509 values in there got lost. New field was added to struct file to
5510 store fs-specific private data.
5513 Frequency : when exporting Lustre via NFS
5515 Description: NFSD failed occasionally when looking up a path component
5516 Details : NFSD is looking up ".." which was broken in ext3 directories
5517 that had grown large enough to become hashed.
5520 Frequency : Clusters with multiple interfaces not on the same subnet
5522 Description: Nodes will repeatedly try to reconnect to an interface which it
5523 cannot reach and report an error to the log.
5524 Details : Extra peer list entries will be created by lconf with some peers
5525 unreachable. lconf now validates the peer before adding it.
5528 Frequency : Only if a default stripe is set on the filesystem root.
5530 Description: Setting a default stripe on the filesystem root prevented the
5531 filesystem from being remounted.
5532 Details : The client was sending extra request flags in the root getattr
5533 request and did not allocate a reply buffer for the dir EA.
5536 Frequency : occasional, higher if lots of files are accessed by one client
5537 Bugzilla : 6159, 6097
5538 Description: Client trips assertion regarding lsm mismatch/magic
5539 Details : While revalidating inodes the VFS looks up inodes with ifind()
5540 and in rare cases can find an inode that is being freed.
5541 The ll_test_inode() code will free the lsm during ifind()
5542 when it finds an existing inode and then the VFS later attaches
5543 this free lsm to a new inode.
5547 Bugzilla : 6422, 7030
5548 Description: MDS deadlock between mkdir and client eviction
5549 Details : Creating a new file via mkdir or mknod (starting a transaction
5550 and getting the ns lock) can deadlock with client eviction
5551 (gets ns lock and trying to finish a synchronous transaction).
5554 Frequency : occasional
5555 Description: While starting a server, the fsfilt_ext3 module could not be
5557 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5558 kernels. Previously, lconf would still use the ext3 name
5559 when trying to load modules. Now, it will correctly use
5560 ext3 on 2.4 and ldiskfs on 2.6.
5562 Severity : enhancement
5563 Description: The default stripe count has been changed to 1
5564 Details : The interpretation of the default stripe count (0, to lfs
5565 or lmc) has been changed to mean striping across a single
5566 OST, rather than all available. For general usage we have
5567 found a stripe count of 1 or 2 works best.
5569 Severity : enhancement
5570 Description: Add support for compiling against Cray portals.
5571 Details : Conditional compiling for some areas that are different
5575 Frequency : occasional
5576 Bugzilla : 6409, 6834
5577 Description: Creating files with an explicit stripe count may lead to
5578 a failed assertion on the MDS
5579 Details : If some OSTs are full or unavailable, creating files may
5580 trigger a failed assertion on the MDS. Now, Lustre will
5581 try to use other servers or return an error to the
5585 Frequency : occasional
5587 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5588 caused lock timeout during readahead (since 1.4.2).
5589 Details : Processes doing readahead might match a lock that hasn't been
5590 granted yet if there are overlapping and conflicting lock
5591 requests. The readahead process waits on ungranted lock
5592 (original lock is CBPENDING), while OST waits for that process
5593 to cancel CBPENDING read lock and eventually evicts client.
5595 Severity : enhancement
5597 Description: Initial enabling of flock support for clients
5598 Details : Implements fcntl advisory locking and file status functions.
5599 This feature is provided as an optional mount flag (default
5600 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5601 locking are implemented yet, and those that are are not guaranteed
5602 to be completely correct in production environments.
5603 mount -t lustre -o [flock|noflock] ...
5606 Frequency : occasional
5608 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5609 assertion in the ext3 JBD (journaling) layer.
5610 Details : The b_committed_data struct is protected by the big kernel lock
5611 in 2.4 kernels, serializing journal_commit_transaction() and
5612 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5613 access to this struct. In 2.6 kernels there is finer grained
5614 locking to improve SMP performance of the JBD layer.
5618 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5623 Description: merge and cleanup kernel patches.
5626 -----------------------------------------------------------------------------
5628 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5630 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5631 versions of Lustre. Please update all servers and clients to
5632 version 1.4.2 or later at the same time. You must also run
5633 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5634 to update the configuration logs.
5636 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5637 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5638 - swab llog records of type '0' so we get proper header size/idx (5861)
5639 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5640 - fix rename of one directory over another leaking an inode (5953)
5641 - avoid SetPageDirty on 2.6 (5981)
5642 - don't re-add just-being-destroyed locks to the waiting list (5653)
5643 - when creating new directories, inherit the parent's custom
5644 striping settings if present parent (3048)
5645 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5646 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5647 - don't hold client locks on temporary worklist from l_lru (5666)
5648 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5649 - automatically evict dead clients (5921)
5650 - Update file size properly in create+truncate+fstat case (6196)
5651 - Do not unhash mountpoint dentries, do not allow removal of
5653 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5654 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5655 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5656 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5657 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5658 - Correctly handle reads of files with no objects (6243)
5659 - lctl recover will also mark a device active if deactivate used (5933)
5661 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5662 - allow --write-conf on an MDS with different nettype than client (5619)
5663 - don't write config llogs to MDS for mounts not from that MDS (5617)
5664 - lconf should create multiple TCP connections from a client (5201)
5665 - init scripts are now turned off by default; run chkconfig --on
5666 lustre and chkconfig --on lustrefs to use them
5667 - upcalls are no longer needed for clients to recover to failover
5669 - add --abort-recovery option to lconf to abort recovery on device
5671 - add support for an arbitrary number of OSTs (3026)
5672 - Quota support protocol changes.
5673 - forward compatibility changes to wire structs (6007)
5674 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5675 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5676 - improved "lustre status" script
5677 - initialize blocksize for non-regular files (6062)
5678 - added --disable-server and --disable-client configure options (5782)
5679 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5680 - Vanilla 2.4.29 support
5681 - increase maximum number of obd devices to 520 (6242)
5682 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5683 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5684 SLES 9 2.6.5 kernels
5685 - limit stripes per file to 160 (the maximum EA size) (6093)
5687 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5690 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5691 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5692 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5693 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5694 - small ext3 extents cleanups and fixes (5733)
5695 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5696 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5697 - avoid needless client->OST connect, fix handle mismatch (5317)
5698 - fix DLM error path that led to out-of-sync client, long delays (5779)
5699 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5700 - fix several locking issues related to i_size (5492,5624,5654,5672)
5701 - don't move pending lock onto export if it is already evicted (5683)
5702 - fix kernel oops when creating .foo in unlinked directory (5548)
5703 - fix deadlock in obdfilter statistics vs. object create (5811)
5704 - use time_{before,after} to avoid timer jiffies wrap (5882)
5705 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5706 - Do not leak request if server was not able to process it (5154)
5707 - If mds_open unable to find parent dir, make that negative lookup(5154)
5708 - don't create new directories with extent-mapping (5909, 5936)
5710 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5711 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5712 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5713 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5714 - support for RHEL4 kernel on i686 (5773)
5715 - provide error messages when incompatible logs are encountered (5898)
5717 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5718 * version 1.4.0.10 (1.4.1 release candidate 1)
5720 - don't keep a lock reference when lock is not granted (4238)
5721 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5722 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5723 - drop import inflight refcount on signal_completed_replay error (5255)
5724 - unlock page after async write error during send (3677)
5725 - handle missing objects in filter_preprw_read properly (5265)
5726 - no transno return for symlink open, don't save no-trasno open (3440)
5727 - don't try to complete elan receive that already failed (4012)
5728 - free RPC server reply state on error (5406)
5729 - clean up thread from ptlrpc_start_thread() on error (5160)
5730 - readahead could read extra page into cache that wasn't ejected (5388)
5731 - prevent races in class_attach/setup/cleanup/detach (5260)
5732 - don't dereference de->d_inode after l_dput of de (5458)
5733 - use "int" for stripe value returned from lock_to_stripe (5544)
5734 - mballoc allocation and error-checking fixes in 2.6 (5504)
5735 - block device patches to fix I/O request sizes in 2.6 (5482)
5736 - look up hostnames for IB nals (5602)
5737 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5738 - don't start multiple acceptors for the same port (5277)
5739 - fix incorrect LASSERT in mds_getattr_name (5635)
5740 - export a proc file for general "ping" checking (5628)
5741 - fix "lfs check" to not block when the MDS is down (5628)
5743 - service request history (4965)
5744 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5745 - create an "evict_client" /proc entry on OSTs, like the MDS has
5746 - fix mount usage message, return errors per mount(8) (5168)
5747 - change grep [] to grep "[]" in tests so they work in more UMLs
5748 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5749 - remove ancient LOV_MAGIC_V0 EA support (5047)
5750 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5751 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5752 - allow readahead allocations to fail when low on memory (5383)
5753 - mmap locking landed again, after considerable improvement (2828)
5754 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5756 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5759 - send OST transaction number in read/write reply to free req (4966)
5760 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5761 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5762 - the watchdog thread now runs as interruptible (5246)
5763 - flock/lockf fixes (but it's still disabled, pending 5135)
5764 - don't use EXT3 constants in llite code (5094)
5765 - memory shortage at startup could cause assertion (5176)
5767 - reorganization of lov code
5768 - single portals codebase
5770 - add extents/mballoc support (5025)
5771 - direct I/O reads in the obdfilter (4048)
5772 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5774 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5777 - send OST transaction number in read/write reply to free req (4966)
5778 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5779 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5780 - the watchdog thread now runs as interruptible (5246)
5781 - handle missing objects in filter_preprw_read properly (5265)
5782 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5783 - drop import inflight refcount on signal_completed_replay error (5255)
5784 - unlock page after async write error during send (3677)
5785 - return original error code on reconstructed replies (3761)
5786 - no transno return for symlink open, don't save no-trasno open (3440)
5788 - add pid to ldlm debugging output (4922)
5789 - bump the watchdog timeouts -- we can't handle 30sec yet
5790 - extra debugging for orphan dentry/inode bug (5259)
5792 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5795 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5796 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5797 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5798 - fix NULL dereference in /proc/sys/portals/routes (4827)
5799 - allow failed mdc_close() operations to be interrupted (4561)
5800 - stop precreate on OST before MDS would time out on it (4778)
5801 - don't send partial-page writes before EOF from client (4410)
5802 - discard client grant for sub-page writes on large-page clients (4520)
5803 - don't free dentries not owned by NFS code, check generation (4806)
5804 - fix lsm leak if mds_create_objects() fails (4801)
5805 - limit debug_daemon file size, always print CERROR messages (4789)
5806 - use transno after validating reply (3892)
5807 - process timed out requests if import state changes (3754)
5808 - update mtime on OST during writes, return in glimpse (4829)
5809 - add mkfsoptions to LDAP (4679)
5810 - use ->max_readahead method instead of zapping global ra (5039)
5811 - don't interrupt __l_wait_event() during strace
5813 - add software watchdogs to catch hung threads quickly (4941)
5814 - make lustrefs init script start after nfs is mounted
5815 - fix CWARN/ERROR duplication (4930)
5816 - return async write errors to application if possible (2248)
5817 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5818 - print NAL number in %x format (4645)
5819 - update barely-supported suse-2.4.21-171 series (4842)
5820 - support for sles 9 %post scripts
5821 - support for building 2.6 kernel-source packages
5822 - support for sles km_* packages
5824 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5827 - ignore -ENOENT errors in osc_destroy (3639)
5828 - notify osc create thread that OSC is being cleaned up (4600)
5829 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5830 - reconstruct ost_handle() like mds_handle() (4657)
5831 - create a new thread to do import eviction to avoid deadlock (3969)
5832 - let lconf resolve symlinked-to devices (4629)
5833 - don't unlink "objects" from directory with default EA (4554)
5834 - hold socknal file ref over connect in case target is down (4394)
5835 - allow more than 32000 subdirectories in a single directory (3244)
5836 - fix blocks count for O_DIRECT writes (3751)
5837 - OST returns ENOSPC from object create when no space left (4539)
5838 - don't send truncate RPC if file size isn't changing (4410)
5839 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5840 - bind to privileged port in socknal and tcpnal (3689)
5842 - rate limit CERROR/CWARN console message to avoid overload (4519)
5843 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5844 - basic mmap support (3918)
5845 - kernel patch series update from b1_4 (4711)
5847 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5850 - avoid crash during MDS cleanup with OST shut down (2775)
5851 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5852 - don't use bad inodes on the MDS (3744)
5853 - dynamic object preallocation to improve recovery speed (4236)
5854 - don't hold spinlock over lock dumping or change debug flags (4401)
5855 - don't zero obd_dev when it is force cleaned (3651)
5856 - print grants to console if they go negative (4431)
5857 - "lctl deactivate" will stop automatic recovery attempts (3406)
5858 - look for existing locks in ldlm_handle_enqueue() (3764)
5859 - don't resolve lock handle twice in recovery avoiding race (4401)
5860 - revalidate should check working dir is a directory (4134)
5862 - don't always mark "slow" obdfilter messages as errors (4418)
5864 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5867 - don't close LustreDB during write_conf until it is done (3860)
5868 - fix typo in lconf for_each_profile (3821)
5869 - allow dumping logs from multiple threads at one time (3820)
5870 - don't allow multiple threads in OSC recovery (3812)
5871 - fix debug_size parameters (3864)
5872 - fix mds_postrecov to initialize import for llog ctxt (3121)
5873 - replace config semaphore with spinlock (3306)
5874 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5875 - don't allow enqueue to complete on a destroyed export (3822)
5876 - down write_lock before checking llog header bitmap (3825)
5877 - recover from lock replay timeout (3764)
5878 - up llog sem before sending rpc (3652)
5879 - reduce ns lock hold times when setting kms (3267)
5880 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5881 - fix NULL deref and obd_dev leak on setup error (3312)
5882 - replace some LBUG about llog ops with error handling (3841)
5883 - don't match INVALID dentries from d_lookup and spin (3784)
5884 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5885 - fix invalid assertion in ptlrpc_set_wait (3880)
5887 - add libwrap support for the TCP acceptor (3996)
5888 - add /proc/sys/portals/routes for non-root route listing (3994)
5889 - allow setting MDS UUID in .xml (2580)
5890 - print the stack of a process that LBUGs (4228)
5892 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5895 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5896 - make sure to unset replay flag from failed open requests (3440)
5897 - if default stripe count is 0, use OST count for inode size (3636)
5898 - update parent mtime/ctime on client for create/unlink (2611)
5899 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5900 - free recovery state on server during a forced cleanup (3571)
5901 - unregister_reply for resent reqs (3063)
5902 - loop back devices mounting and status check on 2.6 (3563)
5903 - fix resource-creation race that can provoke i_size == 0 (3513)
5904 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5905 - more debugging for page-accounting assertion (3746)
5906 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5907 - avoid deadlock after precreation failure (3758)
5908 - fix race and lock order deadlock in orphan handling (3450, 3750)
5909 - add validity checks when grabbing inodes from l_ast_data (3599)
5911 - add /proc/.../recovery_status to obdfilter (3428)
5912 - lightweight CDEBUG infrastructure, debug daemon (3668)
5913 - change default OSC RPC parameters to be better on small clusters
5914 - turn off OST read cache for files smaller than 32MB
5915 - install man pages and include them in rpms (3100)
5916 - add new init script for (un)mounting lustre filesystems (2593)
5917 - run chkconfig in %post for init scripts (3701)
5918 - drop scimac NAL (unmaintained)
5920 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - clean kiobufs before and after use (3485)
5924 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5925 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5926 - change init script to fail more clearly if not run as root (1528)
5927 - allow clients to reconnect during replay (1742)
5928 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5929 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5930 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5931 - fix cond_resched (3554)
5932 - only evict unfinished clients after recovery (3515)
5933 - allow bulk resend, prevent data loss (3570)
5934 - dynamic ptlrpc request buffer allocation (2102)
5935 - don't allow unlinking open directory if it isn't empty (2904)
5936 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5937 - remove extraneous obd dereference causing LASSERT failure (3334)
5938 - don't use get_cycles() when creating temp. files on the mds (3156)
5939 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5940 - handle EEXIST for set-stripe, set proper directory name (3336)
5942 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5943 - fix ksocknal_fmb_callback() error messages (2918)
5945 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5948 - don't copy lvb into (possibly NULL) reply on error (2983)
5949 - don't deref dentry after dput, don't free lvb on error (2922)
5950 - use the kms to determine writeback rpc length (2947)
5951 - increment oti_logcookies when osc is inactive (2948)
5952 - update client's i_blocks count via lvb messages (2543)
5953 - handle intent open/close of special files properly (1557)
5954 - mount MDS with errors=remount-ro, like obdfilter (2009)
5955 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5956 - don't use cancelling-locks' kms values (2947)
5957 - use highest lock extent for kms, not last one (2925)
5958 - don't dereference ERR_PTR() dentry in error handling path (3107)
5959 - fix thread race in portals_debug_dumplog() (3122)
5960 - create lprocfs device entries at setup instead of at attach (1519)
5961 - common AST error handler, don't evict client on completion race (3145)
5962 - zero nameidata in detach_mnt in 2.6 (3118)
5963 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5964 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5965 - initialize RPC timeout timer earlier for 2.6 (3219)
5966 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5967 - print nal/nid for unknown nid (3258)
5968 - additional checks for oscc recovery before doing precreate (3284)
5969 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5970 - don't crash in mdc_close for bad permissions on open (3285)
5971 - zero i_rdev for non-device files (3147)
5972 - clear page->private before handing to FS, better assertion (3119)
5973 - tune the read pipeline (3236)
5974 - fix incorrect decref of invalidated dentry (2350)
5975 - provide read-ahead stats and refine rpc in flight stats (3328)
5976 - don't hold journal transaction open across create RPC (3313)
5977 - update atime on MDS at close time (3265)
5978 - close LDAP connection when recovering to avoid server load (3315)
5979 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5980 - don't leak open file on MDS after open resend (3325)
5981 - serialize filter_precreate and filter_destroy_precreated (3329)
5982 - loop device shouldn't call sync_dev() for nul device (3092)
5983 - clear page cache after eviction (2766)
5984 - resynchronize MDS->OST in background (2824)
5985 - refuse to mount the same filesystem twice on same mountpoint (3394)
5986 - allow llmount to create routes for mounting behind routers (3320)
5987 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5988 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5989 - fix rare problem with rename on htree directories (3417)
5991 - allow default OST striping configuration per directory (1414)
5992 - fix compilation for qswnal for 2.6 kernels (3125)
5993 - increase maximum number of MDS request buffers for large systems
5994 - change liblustreapi to be useful for external progs like lfsck (3098)
5995 - increase local configuration timeout for slow disks (3353)
5996 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5998 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6001 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6002 - don't overwrite extent policy data in reply if lock was blocked (2901)
6003 - drop filter export grants atomically with removal from device (2663)
6004 - del obd_self_export from work_list in class_disconnect_exports (2908)
6005 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6006 - swab reply message in mdc_close, other PPC fixes (2464)
6007 - fix destroying of named logs (2325)
6008 - overwrite old logs when running lconf --write_conf (2264)
6009 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6010 - fix race in target_handle_connect (2898)
6011 - mds_reint_create() should take same inode create lock (2926)
6012 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6013 - don't close files for self_export to avoid uninitialized obd (2936)
6014 - allow MDS with the same name as client node (2939)
6015 - hold dentry reference for closed log files for unlink (2325)
6016 - reserve space for all logs during transactions (2059)
6017 - don't evict page beyond end of stripe extent (2925)
6018 - don't oops on a deleted current working directory (2399)
6019 - handle hard links to targets without a parent properly (2517)
6020 - don't dereference NULL lock when racing during eviction (2867)
6021 - don't grow lock extents when lots of conflicting locks (2919)
6023 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6026 - account for cache space usage on clients to avoid data loss (974)
6027 - lfsck support in lustre kernel code (2349)
6028 - reduce journal credits needed for BRW writes (2370)
6029 - orphan handling to avoid losing space on client/server crashes
6030 - ptlrpcd can be blocked, stopping ALL progress (2477)
6031 - use lock value blocks to assist in proper KMS, faster stat (1021)
6032 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6033 - recovery for initial connections (2355)
6034 - fixes for mds_cleanup_orphans (1934)
6035 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6036 - block all file creations until orphan recovery completes (1901)
6037 - client remove rq_connection from request struct (2423)
6038 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6039 - recovery timer race (2670)
6040 - mdc_close recovey bug (2532)
6041 - ptlrpc cleanup bug (2710)
6042 - mds timeout on local locks (2588)
6043 - namespace lock held during RPCs (2431)
6044 - handle interrupted sync write properly (2503)
6045 - don't try to handle a message that hasn't been replied to (2699)
6046 - client assert failure during cleanup after abort recovery (2701)
6047 - leak mdc device after failed mount (2712)
6048 - ptlrpc_check_set allows timedout requests to complete (2714)
6049 - wait for inflight reqs when ptlrpcd finishes (2710)
6050 - make sure unregistered services are removed from the srv_list
6051 - reset bulk XID's when resending them (caught by 1138 test)
6052 - unregister_bulk after timeout
6053 - fix lconf error (2694)
6054 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6055 - readahead locks pages, leaves pending causing memory pressure (2673)
6056 - increase OST request buffers to 4096 on large machines (2729)
6057 - fix up permission of existing directories in simple_mkdir (2661)
6058 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6059 - don't assign transno to errored transactions (2742)
6060 - don't delete objects on OST if given a bogus objid from MDS (2751)
6061 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6062 - if rq_no_resend, then timeout request after recovery (2432)
6063 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6064 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6065 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6066 - MDS recovery shouldn't delete objects that it has given out (2730)
6067 - if enqueue arrives after completion, don't clobber LVB (2819)
6068 - don't unlock pages twice when trigger_group_io returns error (2814)
6069 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6070 - don't write pages to disk if there was an error (1450)
6071 - don't ping imports that have recovery disabled (2676)
6072 - take buffered bytes into account when balancing socknal conn (2817)
6073 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6074 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6075 - remove little-endian swabbing of llog records (1987)
6076 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6077 - retry reposting request buffers if they fail (1191)
6078 - grow extent at grant time to avoid granting a revoked lock (2809)
6079 - lock revoke doesn't evict page if covered by a second lock (2765)
6080 - disable VM readahead to avoid reading outside lock extents (2805)
6082 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6083 - updated kernel patches for hp-2.4.20 kernel (2681)
6085 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6088 - fix truncated write corruption (2366)
6089 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6091 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6092 - ensure len(uuid) < 37 in lmc (1171)
6093 - fix ia64 OOPS in llog_test (2255)
6094 - zero end of page at obdfilter for partial page writes (2648)
6095 - don't leave stale dentries around after renames (bug 2428)
6096 - fix timeouts when evicting a client with a single lock held (2642)
6097 - set deadline for the initial HELLO message to drain (2634)
6098 - print out dotted-quad IP addresses in the socknal (2302)
6100 - additional debugging for MDS client eviction problem (2443)
6101 - fix mkfsoptions support for osts (2603, 2604)
6103 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6106 - add series for the vanilla 2.6.0 kernel
6107 - add series for the vanilla 2.4.24 kernel
6108 - add series for a cray x86/64 UL kernel drop
6109 - fix xattr patches for the vanilla 2.4.19 series
6111 - generate true UUIDs in lmc (1171)
6112 - have portals stack dumping break in UML (2466)
6113 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6114 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6115 - wake callback waiting threads on client eviction (2460)
6116 - Add --ptldebug and --subsystem to lmc (1719)
6117 - update assertion to allow safe interrupt allocation
6118 - set rq_no_resend for cancel requests (2432)
6119 - recalculate ptlrpcd timeout after resend (2494)
6120 - call vfs_rmdir when removing pending directories (2368)
6121 - fix renaming a file to itself (2429)
6122 - lmc creates a default one-stripe lov (2454)
6123 - expand procfs space to handle large clusters (2326)
6124 - increase UML stack to avoid overflow
6125 - update lconf's list of debug and subsystem masks
6126 - fix lfs find --obd (2510)
6127 - /proc tunable for disabling filter read caching (2591)
6128 - stop rpm packages from altering slapd.conf (2301)
6129 - disable nagle in the socknal under 0conf (2578)
6130 - choose mds inode size based on stripe count (2572)
6131 - fix kernel-source rpm problems (2516)
6133 - add --disable-doc to avoid pdf generation (2421)
6134 - update documentation, tests, type-os, comments
6135 - avoid format warnings on ia64
6136 - remove the TOE NAL
6137 - tiny code cleanups by removing unused fields
6139 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6142 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6143 - remove the most blatant lies from BUILDING file (2371)
6144 - change default debug level to reasonable production setting
6145 - reduce client side cache size to reduce cache flush time
6146 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6147 - make TCP zerocopy and pinger support enabled by default (2476)
6148 - sync writes completed after process exits caused crashes (2319)
6149 - maintain correct mount count on the MDS (2356)
6150 - backout 1557, because 2316 wasn't really fixed
6151 - better file I/O statistics gathering in /proc
6152 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6153 - another kernel patch to fix zero-copy TCP function export
6154 - don't take duplicate lock when processing re-sent getattr (2420)
6155 - lctl uses obd_self_export instead of creating new conn (2353)
6156 - MDS/OST recovery case which requires object creation asserted (2425)
6157 - move lfs from /usr/sbin to /usr/bin in packages
6158 - fix race between mds_client_add and mds_client_free (2417)
6159 - use kmalloc instead of slabs in portals (2430)
6160 - don't create duplicate records when a failover MDS is present (2442)
6161 - remove unnecessary mount age check (2332)
6162 - don't remove directory inodes from locks prematurely (2451)
6163 - don't break if MDS service name is the same as hostname (2103)
6164 - fix races in client write RPC generation when cache full (2482)
6166 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6169 - remove now-unused request->rq_obd (278)
6170 - if an allocation fails, print out how much memory we've used (1933)
6171 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6172 - add the "configurable stack size" patch to most series files (1256)
6173 - ability to write large log records, for 100+ OST configs (2306)
6174 - fix NULL deref when filter_prep fails (2314)
6175 - fix operator precedence error in filter_sync
6176 - dynamic allocation of socknal TX descriptors (2315)
6177 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6178 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6179 - ensure that CWARN messages in Portals always get to the syslog
6180 - __init/__exit are not for prototype decls (ldlm_init/exit)
6181 - x86-64 compile warning fixes
6182 - fix gateway LMC keyword conflict (2318)
6183 - fix MDS lock inversions in getattr/reint paths (1844)
6184 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6185 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6186 - DLM race condition prevented some lock evictions (2328)
6187 - ENOMEM detection and retry on socknal sends (2230)
6188 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6189 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6190 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6191 - better error messages when a client is rejected during recovery (1505)
6192 - avoid cancelling locks which were never granted, after failure (2330)
6193 - fix i_sem/journal inversion in mds_client_add (2333)
6194 - fix truncate/getattr lock cycle deadlock (2334)
6195 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6196 - fix two rare exit paths which could leak an l_lock() ref (2321)
6197 - fencepost error in MDS/OST orphan recovery (2226)
6198 - make log record alignment 8 bytes (1988)
6199 - lstripe now fails when requested offset > ost_count (2237)
6200 - ensure that all kernel series have a complete list.h (1607)
6201 - fix crashes in special-file operations (2316)
6202 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6204 - allow configurable automake binary, for testing new versions
6205 - small update to the lfs documentation
6207 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6209 * fix negative export reference count in fsfilt_sync (2312)
6211 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6212 * release candidate 0.9.1
6214 - orphans are moved into the PENDING directory for possible recovery
6215 - replayed opens now open by fid for orphan/rename safety (1042)
6216 - last close of an orphan inode generates a transno (683)
6217 - chdir() and mount() now pin the directory entry (1020)
6218 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6219 - discard very old requests without processing them (1502)
6220 - remove some common, well-understood CERRORs (1505)
6221 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6222 - clear "grant" flags in OST replies until OST grant code lands (1644)
6223 - fix read performance by not clobbering i_blksize on client (1598)
6224 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6225 - make LRU size a /proc tunable, clears locks when reduced (707)
6226 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6227 - server locks take references on exports now (1558)
6228 - build fixes for 2.4.20-rh trees (1663)
6229 - return an error from lov_create if all OSCs are inactive (1751)
6230 - fix import levels when a reconnect happens without a timeout (1597)
6231 - exit early from mds_open if we get a lookup error (1749)
6232 - partial page read at EOF wouldn't wait for disk before sending (1642)
6233 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6234 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6235 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6236 - write/truncate lock inversion (1639)
6237 - Don't auto-load obdclass, portals modules during cleanup (1495)
6238 - fix timestamps from jumping to "now" (1763)
6239 - extra journal assertions (1648)
6240 - add an extra multiunlink test (1771)
6241 - fix read_record/write_record API (1776)
6242 - fix leak of offset_extent, possible incorrect i_size later (1772)
6243 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6244 - Debugging for the kqswnal_get_idle_tx problems (1820)
6245 - Allow recovery to be attempted multiple times (1536)
6246 - Write out MDS last_rcvd file after it is first created (1600)
6247 - Fix tx_descriptor leak in failed transmit situations (1827)
6248 - ext3 journaling fixes for assertion failure after IO error (1871)
6249 - class_export_put() on freed export after completion AST error (1896)
6250 - Fix revalidate looping in VFS (1322)
6251 - Don't access a freed export during MDS_REINT timeout (1521)
6252 - Add open-unlink recovery support on the MDS (1673,1764)
6253 - Return an error if no MDS data was read from last_rcvd (1946)
6254 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6255 - Don't setup a disk device that doesn't match exported UUID (317)
6256 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6257 - avoid committing NULL handle in force close
6258 - local.sh is now a one-stripe LOV configuration
6259 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6260 - fix ext3 htree duplicate directory entry corruption (1516)
6261 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6262 - update to new LOV EA format (2097)
6263 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6265 2003-06-15 Phil Schwan <phil@clusterfs.com>
6268 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6269 - per-import/export recovery handling (958, 931, 959)
6270 - multiple last-rcvd slots, for serving multiple FSes (949)
6271 - connections are again shared between multiple imp/exports (963, 964)
6272 - "umount -f" would hang if any requests needed to be sent (393, 978)
6273 - avoid pinning large req buffer by copying for queued messages (989)
6274 - add "uuid" to "lctl device" command to help upcalls (991)
6275 - "open" RPCs with transnos would confuse recovery counters (1037)
6276 - do proper endian conversion of all wire messages (288, 340, 891)
6277 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6278 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6279 - fix ptlrpc_connection leak in target_handle_connect (1174)
6280 - fix import refcounting bug in OST and MDS cleanup (1134)
6281 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6282 - fix very unlikely obd_types race condition (501)
6283 - remove osc_open hack for echo_client (1187)
6284 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6285 - a failure in read_inode2 leads to deadlock (1139)
6286 - cancel ack-locks as soon as transaction is committed (1072)
6287 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6288 - make sure to commitrw after any preprw to avoid deadlock (1162)
6289 - failing to execute a file in a lustre FS would lock inode (1203)
6290 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6291 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6292 - obd_uuid2tgt would walk off the end of the list (1255)
6293 - on IA64 the osc would give portals incorrect bulk size (1258)
6294 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6295 - fix lock inversion caused by new llite matching code (1282)
6296 - limit the number of dirty pages on a client to 10MB (1286)
6297 - timed out locks were not being corrected cancelled (1289)
6298 - fix O_DIRECT above 4GB on IA-32 (1292)
6299 * major user-visible changes
6300 - fail out/fail over policy now controlled by the upcall (993)
6302 - add OBD_PING to check server availability and failure (954)
6303 - lustre messages are now sent in sending host order (288, 340, 891)
6304 - add eadatalen to MDS getattr reply (340)
6305 - OST read replies may contain second buffer, with per-page status (593)
6307 2003-03-11 Phil Schwan <phil@clusterfs.com>
6310 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6311 - failed open()s could cause deadlock; fixed (867, 869)
6312 - stop cancelling OST locks when files are closed (481)
6313 - overlapping XID spaces caused network corruption (851, 853)
6314 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6315 - setattr_raw added, to do single-RPC, server-side setattrs
6316 - lmc/lconf syntax change for OST UUIDs
6317 - fix crashy race condition between ptlrpc_free_req and osc_close
6318 - don't use request in mdc_enqueue if we hit a timeout (889)
6319 - don't set the inode i_size for regular files from the MDS (896)
6320 - handle out of order completion AST (842)
6321 - don't LBUG if a lock request times out after receiving AST (913)
6322 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6323 - if a bad lock AST arrives, send an error instead of dropping entirely
6324 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6325 - fix leak in bulk IO when only partially completed (899, 900, 926)
6326 - fix O_DIRECT for ia64 (55)
6327 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6328 - C-z after timeout could hang a process forever; fixed (977)
6330 - client-side I/O cache (678, 924, 929, 941, 970)
6332 - READPAGE and SETATTRs which don't take server-side locks get
6335 2003-02-11 Phil Schwan <phil@clusterfs.com>
6338 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6339 - Fix fsfilt_extN_readpage() to read a full page of directory
6340 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6341 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6342 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6343 - Most intent operations were removed, in favour of a new RPC mode
6344 that does a single RPC to the server and bypasses most of the VFS
6345 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6346 - Aggressively cancel local locks on DLM servers
6347 - mds_reint_unlink sends EA to the client if it's the last nlink.
6348 client uses that EA to unlink OST objects.
6349 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6350 - recursive symlinks were fixed (439)
6351 - fixed NULL deref in DEBUG_REQ
6352 - filter_update_lastobjid no longer calls sync, which annoyed extN
6353 - fixed multi-client small-writes to a single file problem (445)
6354 - fixed mtime updates during file writes (607)
6355 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6356 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6357 - fixed Linux OST size reporting problem (444, 656)
6358 - OST now updates object mtime with writes or setattr (607, 619)
6359 - client verifies file size before zeroing page past EOF (445)
6360 - OST now writes last allocated objid to disk with allocation (108)
6361 - LOV on echo now works (409)
6363 - mds_reint_unlink sends a new buffer, with the EA included. this
6364 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6365 set if a regular file was being unlinked, and it was the last link
6366 - use PtlGet from the target for bulk writes (315)
6367 - OST now updates object mtime with writes or setattr (607, 619)
6368 - LDLM now has a grant-time callback to revalidate locked items, if
6370 - Many MDS operations were reorganized to combat race conditions
6372 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6373 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6375 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6378 - Fully reactivate OST imports after reconnection (512, others)
6379 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6380 - More graceful error handling for truncating on dead OST (515)
6381 - Don't error out unless we're actually accessing dead stripes (474)
6382 - Fix garbage sizes when stripes are missing (410)
6383 - LRU counters were broken, causing constant lock purge (433, 432)
6384 - garbage on read from stripes with failed OSTs (441)
6385 - mark OSCs as active before reconnecting during recovery (438)
6386 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6387 - lfind did not preserve OST order in output (443)
6388 - symlinks cause hung clients, incorrect data (439)
6389 - stop dereferencing request after dropping refcount (457)
6390 - don't LASSERT(spin_is_locked) on non-SMP (455)
6391 - fixes for many rename() bugs
6392 - fstat didn't correctly synchronize attributes (399)
6393 - server must handle lock cancellation during blocking AST prep (487)
6394 - bulk descriptors were free()d too soon (511)
6395 - fix paths in lconf, which would load incorrect modules (451, 507)
6396 - fix confusing lconf 'host not found' error message (386)
6397 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6398 - fix race condition in mdc_blocking_ast() for inode access (526)
6399 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6400 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6401 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6402 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6403 - fix lstripe in several ways (kernel side) (527)
6404 - fix request leak in ldlm_cli_enqueue (262)
6405 - incorrect OSC was marked inactive after OST failure
6406 - call mds_fs_cleanup before unmounting filesystem (524)
6407 - fix races between taking ns_lock and ldlm_lock_change_resource
6408 - fix races updating LOV export open file list
6409 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6410 - fix recovery NULL deref in ldlm_cli_cancel_unused
6411 - fix some DLM races by using new hash table for lock handles (419)
6412 - permit the client to specify desired inodes, at replay
6413 - duplicate requests when we queue them for replay reintegration
6414 - fix last_rcvd offset calculation
6415 - sync after each recovered transaction, so we always make progress
6416 - never, not always, ERESTART requests without transnos
6417 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6419 - skip replay if the MDS says that the client is already connected
6420 - don't check for a recovery-enabled export to match lctl's UUID
6421 - don't INC_USE_COUNT for phantom exports
6422 - don't crash when cleaning up phantom exports (567)
6423 - don't double-finish or set replay data for errored mdc_open requests
6424 - abort requests when they time out, so we don't get old replies
6425 - send/receive replies for AST messages again
6426 - if the client says that it doesn't have the lock, cancel it on the
6428 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6429 mark it as destroyed, it will all work out when decref is called
6430 - fix module use counts (22, 581)
6432 - ASTs now expect a reply (server cancels lock on error reply)
6434 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6437 - fix many simultaneous client startup (392)
6438 - fix dentry->d_it clobbering
6439 - credentials weren't being shipped for readdir/getattr operations
6440 - remove invalid assertions triggered during some concurrent MD
6442 - proper Lustre versions added (336, 389)
6443 - fix memory leak for create error case (398)
6444 - fix LOV locking bug that would get cli/srv out of sync
6445 - fix echo client over LOV (409)
6446 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6447 - fix double-O_EXCL intent crash (424)
6448 - avoid sending multiple lock CANCELs (352)
6450 - MDS can do multi-client recovery (modulo bugs in new code)
6452 - many updates, edits, cleanups
6454 2002-11-18 Phil Schwan <phil@clusterfs.com>
6457 - fix null d_it dereference (346)
6458 - fix full OST/dbench hang (333)
6459 - fix permission problem with file removal (286)
6460 - fix removal of OSCs from LOV when they fail
6461 - fix NULL deref during bulk timeout (214)
6462 - fix problems related to multiple filesystems on one MDS (241)
6463 - fixed serious subtle metadata locking bugs
6464 - free locks on clients when inodes are removed due to memory
6466 - fix inode pointer in lock data (285)
6467 - partial support for multiple MDS on a single host (241)
6468 - data locks weren't cancelled at clear_inode time (290, 311)
6469 - intent locks could lead to unbounded lock growth (205)
6470 - added a maximum lock count, an LRU list, and a flusher
6471 - fix multiple rename (365)
6472 - properly abstracted the echo client
6473 - OSC locked 1 byte too many; fixed
6474 - rewrote brw callback code:
6475 - fixed recovery bugs related to LOVs (306)
6476 - fixed too-many-pages-in-one-write crash (191)
6477 - fixed (again) crash in sync_io_timeout (214)
6478 - probably fixed callback-related race (385)
6480 - Add capability to MDS protocol
6481 - LDLM cancellations and callbacks on different portals
6483 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6486 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6487 - defer open object destruction to close time (601981, 138)
6488 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6489 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6490 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6491 - don't call obd_setattr in ll_file_release for destroyed objects
6492 * protocol change to lustre_msg: move |version| and add |flags|
6493 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6494 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6495 * added replay of create, unlink, link and rename operations during
6496 MDS failover; recovery should be much more robust now
6497 * remove failed OSCs from LOVs (only lov_create uses this so far)
6498 * the lustre-HOWTO was brought (more) up to date (582544)
6500 2002-10-23 Phil Schwan <phil@clusterfs.com>
6503 - in-use dentries weren't being reused properly (617851)
6504 - prevent multiple LDLM setup (599178)
6505 - fix LOV size calculations for truncate (617853)
6506 - fix client handling of MDS intent errors (POSIX)
6507 - fix permission bug in lovstripe.c test (624321)
6508 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6509 - truncate past end of file could corrupt data
6510 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6511 - a race in recovery could return ETIMEDOUT to apps (623947)
6512 - building outside the source directory was fixed
6513 * the lustre-HOWTO was brought (more) up to date (582544)
6514 * major progress was made on recovery functionality
6516 2002-10-10 Phil Schwan <phil@clusterfs.com>
6519 - recovery deadlock fix
6520 - rm -rf causes LBUG fix (617817)
6521 - file open by multiple tasks fix (618962)
6522 - directory permissions bugs (602707 and 620007)
6523 - journal_stop fixed with locking (611313)
6524 - O_APPEND failures resolved (618273, perhaps 614459)
6525 - lconf PATH fix (619770)
6526 - IA64 build fix (621450)
6527 - RPC buffer sizes scale with amount of memory
6529 2002-10-01 Phil Schwan <phil@clusterfs.com>
6532 - locks would be cancelled without throwing away data pages,
6533 resulting in inconsistent data (605627)
6534 - inode attributes were not always being refreshed (605627, 612449)
6535 - lconf now continues to cleanup after lctl reports an error
6536 - MDS now enforces user permissions (602707)
6537 - lprocfs cleanup fixed, but not yet enabled (614157)
6538 - fixed infinite server hang, should a client not respond to an AST
6539 - avoid going into recovery if user calls readlink() with a buffer
6540 that's too small (613941)
6541 - AST RPCs no longer require replies (614867) -- this may be changed
6542 - don't crash server if client sends an IOV that's too big (611336)
6543 - fixed lock conversion deadlock (611892)
6544 - fixed the following of symlinks (614622)
6545 * recovery: the server can remove locks from a client that dies, other
6546 clients can make progress
6547 * more extN patch fixes
6548 * compile-time configurable ptlrpc buffer allocations
6550 - collaborative read cache document
6551 - Lustre Lite Performance CDR document-in-progress
6553 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6556 - fix typo in patch-2.4.18
6558 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6561 - clear ptlrpc request each time in handle_incoming_request()
6562 - unlink of files now destroys the object on the OST
6564 2002-09-19 Peter Braam <braam@clusterfs.com>
6566 * add hard link support
6567 * change obdfile creation method
6568 * kernel patch changed
6570 2002-09-19 Peter Braam <braam@clusterfs.com>
6573 - stack overflow bug in extN fixed
6575 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6577 * documentation updates
6578 - add man pages for config tools
6579 - update tests/README to describe testing with new config tools
6580 - finish metadata API descriptions
6581 * bug fixes and cleanups
6582 - statfs workaround for 16TB limit
6583 - LOV stripe allocation improved, can stripe on subset of OSTs
6584 - LOV file size/IO offset was wrong for files > 4GB in size
6585 - object EA data was being dropped, caused files to be unreadable
6586 - memory overflow with non-LOV OST caused memory corruption
6587 - fixed regression tests to work with new config tools, obdfilter
6588 - fixed bug when directory size became larger than 1 block
6589 - fixed bug (for single client case) when PWD was deleted
6590 - invalidate local directory pages when doing intent-based ops
6591 - avoid LDLM oops when lock callback contained bad data
6593 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6595 * documentation updates
6596 * bug fixes and cleanups
6597 - configuration tools
6600 - 64-bit compile warnings
6601 - 64-bit internal statfs data
6603 * test_brw on persistent OST devices
6605 * lprocfs (disabled)
6607 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6609 * documentation updates
6610 * bug fixes and cleanups
6611 * configuration tools
6613 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6616 * many small fixes to 0_5_4
6617 * io/network handling
6618 * thinkos in MDS operations
6620 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6623 * crucial basic fixes to 0.5.3
6624 * IOR, Iozone work over Elan
6627 2002-08-07 Phil Schwan <phil@clusterfs.com>
6628 * version 0_5_3, our first alpha
6629 * we use the new Portals iovs
6630 * documentation updates
6631 * bug fixes and cleanups
6632 * small changes in the DLM wire protocol
6634 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6635 * version 0_5_1 with some initial stability,
6636 * locking on MD and file I/O.
6637 * documentation updates
6638 * several bug fixes since 0.5.0
6639 * small changes in wire protocol
6641 2002-07-18 Phil Schwan <phil@clusterfs.com>
6643 * delivered as Lustre Light Alpha
6644 * fixed a crash after handling invalid MDS requests
6645 * fixed directory pages for architectures with non-4k pages sizes
6647 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6648 * release version v0_4_4
6649 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6651 * Fixes a number of interruption problems with OST operations.
6652 * Update documentation for portals header changes
6653 * Move all wire protocol structs/defines to lustre_idl.h
6654 * Fixes symlink length bug.
6655 * Add tcpdump to repository.
6657 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6658 * release version v0_4_3
6659 * Fixes statfs for inodes on extN.
6660 * Fixes bug in runtests which would delete /etc/hosts.
6661 * Use 64-bit object IDs wherever possible (not into VFS though)
6662 Remove ost_get_info, which is unused by lustre, and out of date.
6664 2002-07-03 Peter Braam <braam@clusterfs.com>
6665 * release version v0_4_2 Fixes a lookup error (type not passed)
6666 * move forward to head of Portals
6667 * move forward to latest Lustre kernel
6669 2002-06-25 Peter Braam <braam@clusterfs.com>
6670 * release version v0_4_1. Hopefully stable on single node use.