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 4.4.120-92.70 (SLES12 SP2)
22 4.4.162-94.69 (SLES12 SP3)
23 4.4.0-131 (Ubuntu 16.04)
24 4.15.0-32 (Ubuntu 18.04)
25 vanilla linux 4.15.0 (ZFS only)
26 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
27 * Recommended ZFS / SPL version: 0.7.9
28 * Tested with ZFS / SPL version: 0.7.9
29 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
30 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
31 stack overflow. For more information, please refer to bugzilla 17630.
32 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
33 the re-exporting nodes to be mounted with "32bitapi" mount option
34 * Note for Ubuntu 14.04.5:
35 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
36 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
38 --------------------------------------------------------------------------------
41 2018-04-03 Intel Corporation
43 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
44 for currently supported client and server kernel versions.
45 * Server known to build on patched kernels:
46 2.6.32-431.29.2.el6 (RHEL6.5)
47 2.6.32-504.30.3.el6 (RHEL6.6)
48 2.6.32-573.26.1.el6 (RHEL6.7)
49 2.6.32-642.15.1.el6 (RHEL6.8)
50 2.6.32-696.18.7.el6 (RHEL6.9)
51 3.10.0-693.21.1.el7 (RHEL7.4)
52 3.0.101-0.47.71 (SLES11 SP3)
53 3.0.101-107 (SLES11 SP4)
54 3.12.74-60.64.40 (SLES12 SP1)
55 4.4.114-92.64 (SLES12 SP2)
56 4.4.114-94.11 (SLES12 SP3)
57 3.13.0-101 (Ubuntu 14.04, ZFS only)
58 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
59 4.4.0-116 (Ubuntu 16.04)
60 vanilla linux 4.6.7 (ZFS only)
61 * Client known to build on unpatched kernels:
62 2.6.32-431.29.2.el6 (RHEL6.5)
63 2.6.32-504.30.3.el6 (RHEL6.6)
64 2.6.32-573.26.1.el6 (RHEL6.7)
65 2.6.32-642.15.1.el6 (RHEL6.8)
66 2.6.32-696.18.7.el6 (RHEL6.9)
67 3.10.0-693.21.1.el7 (RHEL7.4)
68 3.0.101-0.47.71 (SLES11 SP3)
69 3.0.101-107 (SLES11 SP4)
70 3.12.74-60.64.40 (SLES12 SP1)
71 4.4.114-92.64 (SLES12 SP2)
72 4.4.114-94.11 (SLES12 SP3)
73 3.13.0-101 (Ubuntu 14.04)
74 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
75 4.4.0-116 (Ubuntu 16.04)
77 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
78 * Recommended ZFS / SPL version: 0.7.6
79 * Tested with ZFS / SPL version: 0.7.6
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
85 * Note for Ubuntu 14.04.5:
86 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
87 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
89 --------------------------------------------------------------------------------
91 2017-07-05 Intel Corporation
93 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
94 for currently supported client and server kernel versions.
95 * Server known to build on patched kernels:
96 2.6.32-431.29.2.el6 (RHEL6.5)
97 2.6.32-504.30.3.el6 (RHEL6.6)
98 2.6.32-573.26.1.el6 (RHEL6.7)
99 2.6.32-642.15.1.el6 (RHEL6.8)
100 2.6.32-696.3.1.el6 (RHEL6.9)
101 3.10.0-514.21.1.el7 (RHEL7.3)
102 3.0.101-0.47.71 (SLES11 SP3)
103 3.0.101-100 (SLES11 SP4)
104 3.12.74-60.64.40 (SLES12 SP1)
105 4.4.59-92.17 (SLES12 SP2)
106 vanilla linux 4.6.7 (ZFS only)
107 * Client known to build on unpatched kernels:
108 2.6.32-431.29.2.el6 (RHEL6.5)
109 2.6.32-504.30.3.el6 (RHEL6.6)
110 2.6.32-573.26.1.el6 (RHEL6.7)
111 2.6.32-642.15.1.el6 (RHEL6.8)
112 2.6.32-696.3.1.el6 (RHEL6.9)
113 3.10.0-514.21.1.el7 (RHEL7.3)
114 3.0.101-0.47.71 (SLES11 SP3)
115 3.0.101-100 (SLES11 SP4)
116 3.12.74-60.64.40 (SLES12 SP1)
117 4.4.59-92.17 (SLES12 SP2)
119 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
120 * Recommended ZFS / SPL version: 0.6.5.9
121 * Tested with ZFS / SPL version: 0.6.5.9
122 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
123 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
124 stack overflow. For more information, please refer to bugzilla 17630.
125 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
126 the re-exporting nodes to be mounted with "32bitapi" mount option
128 --------------------------------------------------------------------------------
130 12-07-2016 Intel Corporation
132 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
133 for currently supported client and server kernel versions.
134 * Server known to build on patched kernels:
135 2.6.32-431.29.2.el6 (RHEL6.5)
136 2.6.32-504.30.3.el6 (RHEL6.6)
137 2.6.32-573.26.1.el6 (RHEL6.7)
138 2.6.32-642.11.1.el6 (RHEL6.8)
139 3.10.0-514.el7 (RHEL7.3)
140 3.0.101-0.47.71 (SLES11 SP3)
141 3.0.101-80 (SLES11 SP4)
142 3.12.62-60.64.8 (SLES12 SP1)
143 vanilla linux 4.5.7 (ZFS only)
144 * Client known to build on unpatched kernels:
145 2.6.32-431.29.2.el6 (RHEL6.5)
146 2.6.32-504.30.3.el6 (RHEL6.6)
147 2.6.32-573.26.1.el6 (RHEL6.7)
148 2.6.32-642.11.1.el6 (RHEL6.8)
149 3.10.0-514.el7 (RHEL7.3)
150 3.0.101-0.47.71 (SLES11 SP3)
151 3.0.101-80 (SLES11 SP4)
152 3.12.62-60.64.8 (SLES12 SP1)
154 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
155 * Recommended ZFS / SPL version: 0.6.5.8
156 * Tested with ZFS / SPL version: 0.6.5.8
157 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
158 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
159 stack overflow. For more information, please refer to bugzilla 17630.
160 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
161 the re-exporting nodes to be mounted with "32bitapi" mount option
163 --------------------------------------------------------------------------------
165 02-29-2016 Intel Corporation
167 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
168 for currently supported client and server kernel versions.
169 * Server known to build on patched kernels:
170 2.6.32-431.29.2.el6 (RHEL6.5)
171 2.6.32-504.30.3.el6 (RHEL6.6)
172 2.6.32-573.12.1.el6 (RHEL6.7)
173 3.10.0-327.3.1.el7 (RHEL7.2)
174 3.0.101-0.47.71 (SLES11 SP3)
175 3.0.101-68 (SLES11 SP4)
176 vanilla linux 4.2.1 (ZFS only)
177 * Client known to build on unpatched kernels:
178 2.6.32-431.29.2.el6 (RHEL6.5)
179 2.6.32-504.30.3.el6 (RHEL6.6)
180 2.6.32-573.12.1.el6 (RHEL6.7)
181 3.10.0-327.3.1.el7 (RHEL7.2)
182 3.0.101-0.47.71 (SLES11 SP3)
183 3.0.101-68 (SLES11 SP4)
186 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
187 * Recommended ZFS / SPL version: 0.6.4.2
188 * Tested with ZFS / SPL version: 0.6.4.2
189 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
190 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
191 stack overflow. For more information, please refer to bugzilla 17630.
192 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
193 the re-exporting nodes to be mounted with "32bitapi" mount option
195 --------------------------------------------------------------------------------
197 03-10-2015 Intel Corporation
199 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
200 for currently supported client and server kernel versions.
201 * Server known to build on patched kernels:
202 2.6.32-431.29.2.el6 (RHEL6.5)
203 2.6.32-504.8.1.el6 (RHEL6.6)
204 3.0.101-0.46 (SLES11 SP3)
205 * Client known to build on unpatched kernels:
206 2.6.32-431.29.2.el6 (RHEL6.5)
207 2.6.32-504.8.1.el6 (RHEL6.6)
208 3.10.0-123.20.1.el7 (RHEL7)
209 3.0.101-0.46 (SLES11 SP3)
210 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
211 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
212 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
213 stack overflow. For more information, please refer to bugzilla 17630.
214 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
215 the re-exporting nodes to be mounted with "32bitapi" mount option
217 Severity : enhancement
219 Description: control OST-index in IDIF via ROCOMPAT flag.
220 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
221 last_rcvd file. For new formatted OST device, it will be auto set;
222 for the case of upgrading from old OST device, you can enable it
223 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
224 enabled, for new created OST-object, its IDIF-in-LMA will contain
225 the OST-index; for the existing OST-object, the OSD will convert
226 old format IDIF as new format IDIF with OST-index stored in the
227 LMA EA when accessing such OST-object or via OI scrub. Once such
228 flag is enabled, it cannot be reverted back, so the system cannot
229 be downgraded to the orignal incompatible version.
231 --------------------------------------------------------------------------------
233 07-30-2014 Intel Corporation
235 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
236 for currently supported client and server kernel versions.
237 * Server known to build on patched kernels:
238 2.6.32-431.20.3.el6 (RHEL6.5)
239 3.0.101-0.31 (SLES11 SP3)
240 * Client known to build on unpatched kernels:
241 2.6.32-431.20.3.el6 (RHEL6.5)
242 3.0.101-0.31 (SLES11 SP3)
245 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
246 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
247 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
248 stack overflow. For more information, please refer to bugzilla 17630.
249 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
250 the re-exporting nodes to be mounted with "32bitapi" mount option
252 --------------------------------------------------------------------------------
254 10-31-2013 Intel Corporation
256 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
257 * Server builds for kernels:
258 2.6.32-358.18.1.el6 (RHEL6.4)
259 * Client builds for unpatched kernels:
260 2.6.32-358.18.1.el6 (RHEL6.4)
261 3.0.92-0.8 (SLES11 SP3)
263 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
264 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
265 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
266 stack overflow. For more information, please refer to bugzilla 17630.
267 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
268 the re-exporting nodes to be mounted with "32bitapi" mount option
270 --------------------------------------------------------------------------------
272 05-30-2013 Intel Corporation
274 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
275 * Server builds for kernels:
276 2.6.32-358.6.2.el6 (RHEL6.4)
277 * Client builds for unpatched kernels:
278 2.6.32-358.6.2.el6 (RHEL6.4)
279 3.0.74-0.6.6 (SLES11 SP2)
281 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
282 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
283 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
284 stack overflow. For more information, please refer to bugzilla 17630.
285 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
286 the re-exporting nodes to be mounted with "32bitapi" mount option
288 --------------------------------------------------------------------------------
290 9-30-2012 Whamcloud, Inc.
292 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
293 * Server support for kernels:
294 2.6.32-279.5.1.el6 (RHEL6)
295 * Client support for unpatched kernels:
296 2.6.32-279.5.1.el6 (RHEL6)
298 2.6.18-238.19.1.el5 (RHEL5)
299 2.6.16.60-0.69.1 (SLES10)
300 * Recommended e2fsprogs version: 1.41.90.wc4
301 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
302 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
303 stack overflow. For more information, please refer to bugzilla 17630.
304 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
305 the re-exporting nodes to be mounted with "32bitapi" mount option
307 Severity : enhancement
309 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
311 Severity : enhancement
313 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
315 Severity : enhancement
317 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
319 Severity : enhancement
321 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
323 Severity : enhancement
325 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
327 --------------------------------------------------------------------------------
329 04-30-2012 Whamcloud, Inc.
331 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
332 * Server support for kernels:
333 2.6.32-220.4.2.el6 (RHEL6)
334 * Client support for unpatched kernels:
335 2.6.32-220.4.2.el6 (RHEL6)
337 2.6.18-238.19.1.el5 (RHEL5)
338 2.6.16.60-0.69.1 (SLES10)
339 * Recommended e2fsprogs version: 1.41.90.wc4
340 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
341 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
342 stack overflow. For more information, please refer to bugzilla 17630.
343 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
344 the re-exporting nodes to be mounted with "32bitapi" mount option
346 Severity : enhancement
348 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
350 Severity : enhancement
352 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
354 Severity : enhancement
356 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
358 Severity : enhancement
360 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
362 --------------------------------------------------------------------------------
364 09-30-2011 Whamcloud, Inc.
366 * Server support for kernels:
367 2.6.18-238.19.1.el5 (RHEL5)
368 2.6.32-131.6.1.el6 (RHEL6)
369 * Client support for unpatched kernels:
370 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
371 2.6.18-238.19.1.el5 (RHEL5)
372 2.6.32-131.6.1.el6 (RHEL6)
373 2.6.29.4-167.fc11 (FC11)
374 2.6.16.60-0.69.1 (SLES10)
376 * Recommended e2fsprogs version: 1.41.90.wc3
377 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
378 Lustre filesystem with 4K stack may cause a stack overflow. For more
379 information, please refer to bugzilla 17630.
380 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
381 mounted with "32bitapi" mount option
383 Severity : enhancement
384 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
386 Severity : enhancement
387 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
389 Severity : enhancement
390 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
393 Frequency : only with SLES10
395 Description: Use OFED "KMP" provided by Novell
396 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
397 Lustre is now built against this package. That means you need to
398 install the ofed-kmp package from Novell for the patchless client.
399 Note that the ofed-kmp that Novell ships may not exactly match the
400 kernel version but should still be compatible.
402 Severity : enhancement
404 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
405 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
407 --------------------------------------------------------------------------------
409 2010-07-15 Oracle, Inc.
411 * Server support for kernels:
412 2.6.18-164.11.1.el5 (RHEL 5)
413 2.6.18-164.11.1.0.1.el5 (OEL 5)
414 * Client support for unpatched kernels:
415 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
416 2.6.18-164.11.1.el5 (RHEL 5),
417 2.6.18-164.11.1.0.1.el5 (OEL 5)
418 2.6.16.60-0.42.8 (SLES 10),
420 2.6.29.4-167.fc11 (FC11)
421 * Recommended e2fsprogs version: 1.41.10-sun2
423 Severity : enhancement
425 Description: Distribute lfs_migrate script with lustre
426 Details : The lfs_migrate mentioned in the manual is now included in the
427 lustre tree in lustre/scripts and will distributed in lustre RPMs.
429 Severity : enhancement
431 Description: Failover nid list added to import proc file.
433 Severity : enhancement
435 Description: "som_preview" mount option required to use SOM
436 Details : The clients will need to use som_preview mount option
440 Description: Lustre does not do 1MB IOs to HW RAID
441 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
443 Severity : enhancement
445 Description: lfs quota output is not very convenient for awk/sed-parsing
446 Details : Some positions in lfs quota output table could be empty or
447 non-empty which made it hard to parse it with scripts, now
448 a dash is put instead of space where there is not supposed
453 Description: Update to OFED 1.5.1
454 Details : OFED 1.5.1 will be used in environments that don't use
455 vendor supplied OFED.
459 Description: Remove lnet. prefix from lctl params display
460 Details : Also renamed lnet.memused to lnet_memused to
465 Description: Use conf_param -d to remove permanant settings.
469 Description: Add list_param -R
470 Details : lctl list_param now supports recursive mode.
473 Bugzilla : 21829, 22312
474 Description: Fix llobdstat.
475 Details : 1. Now updating the write_bytes counter
476 2. Port script fixes from 1.8.
480 Description: decrease the usage of memory on clients.
481 Details : 1. On clients, recycle dentries and inodes unused.
482 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
487 Description: llobdstat fix and enhancement
488 Details : add a counter to set a limit to how many samples will be returned
489 fix a wildcard in the path to limit to obdfilter stats only
493 Description: lfs check only allowed for root
494 Details : Code cleanups and added a regresion test of non-root use of
499 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
500 Details : Fix hr->hr_index overload.
504 Description: Fix FIEMAP functionality.
508 Description: FC11 patchless client support added
512 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
513 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
517 Description: New acc-sm script and various test-framework enhancments.
521 Description: Previously landed patch was reverted.
525 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
529 Description: Add ldiskfs maxdirsize mount option.
533 Description: Fix mkfs.lustre support for 16TB LUNs.
537 Description: Print mmp_check_interval
538 Details : print mmp_check_interval and make it possible to abort
539 mount operation in case it takes too long.
543 Description: Add cascading_rw to lustre mpi tests.
547 Description: Store SOM data in EA.
548 Details : Also several SOM related cleanups.
552 Description: Add refcount to md_open_data
556 Description: Use refcount on lov_request_set to prevent memory corruption.
560 Description: Statahead: use parent dir's i_mutex for synchronization.
564 Description: loadgen improvements.
568 Description: Enable peer health detection in ksocklnd and ko2iblnd.
572 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
576 Description: Fix race on obd_devcie:md_stats
580 Description: Fix SLES 11 client builds.
581 Details : Resolved uninitalized pointer gcc warnings.
585 Description: Allow non-root access for "lfs check".
586 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
590 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
592 Severity : enhancement
594 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
595 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
599 Description: Sanity test_27p fails.
600 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
601 Sanity tests 27 were cleaned up and improved.
605 Description: Unify req format on client/servers
606 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
607 buffers and swabbers approach..
612 Description: drop join file stuff.
614 Severity : enhancement
616 Description: Add 2.6.27 and SLES11 (patchless client) support
620 Description: Fix in ptlrpc_expire_one_request() to print the signed time
623 Severity : enhancement
625 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
628 Severity : enhancement
630 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
632 Severity : enhancement
634 Description: correctly handle big reply message.
635 Details : send LNet event if reply is bigger than buffer and adjust this buffer
640 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
642 Severity : enhancement
644 Description: Add support for OEL5.
648 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
650 Severity : enhancement
652 Description: "weak-modules" support
653 Details : Implement "weak-modules" support which enables kernel modules
654 to be used with any kernel that implements the same kABI. In
655 order to achieve this modules are now installed in
656 /lib/modules/$(uname -r)/updates/kernel on all distributions.
660 Description: client could not reconnect to OST because of an active request.
661 Details : The client now retries to reconnect to the same server, if a
662 connect request failed with EBUSY or -EAGAIN. Test case 35b
663 conf-sanity.sh has been created.
667 Description: Don't allow make backward step on assignin osc next id.
668 Details : race between allocation next id and ll_sync thread can be cause
669 of set wrong osc next id and can be kill valid ost objects.
673 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
677 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
681 Description: Error handling in osc_statfs_interpret() has been improved.
682 Details : Check in osc_statfs_interpret() for EBADR.
686 Description: Do not send statfs() requests to OSTs disabled by administrator.
687 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
689 Severity : enhancement
691 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
693 Severity : enhancement
695 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
697 Severity : enhancement
699 Description: Enhancement for lfs(1) command to use numeric uid/gid.
703 Description: Improved cleanup procedure for conf-sanity test_46a.
707 Description: Do not start bulk transfer, if client tried to reconnect.
708 Details : Do not start bulk transfer, if client tried to reconnect,
709 but failed due to active rpcs. Check for exp_abort_active_req
710 has been implemented.
714 Description: Prevent inconsistences between linux and lustre mount structures.
715 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
716 Make the sleep interruptible.
718 Severity : enhancement
720 Description: provide server to client comms path
721 Details : server to client communications path via new LDLM_SET_INFO rpc
725 Description: truncate starts GFP_FS allocation under transaction causing deadlock
726 Details : ldiskfs_truncate calls grab_cache_page which may start page
727 allocation under an open transaction. This may lead to
728 calling prune_icache with consequent lustre reentrance.
732 Description: send correctly lsm on open replay
733 Details : MDS is trust to LSM size on replay open, but client can set wrong size
738 Description: enable client interface failover
739 Details : When a child reconnects from another NID, properly update export
740 nid hash position and ldlm reverse import.
742 Severity : enhancement
744 Description: Communicate OST degraded/readonly state via statfs to MDS
745 Details : Flags in the statfs returned from OSTs indicate whether the
746 OST is in a degraded RAID state, or if the filesystem has
747 turned read-only after a filesystem error is detected.
752 Description: don't panic if EPROTO was hit when reading symlink
753 Details : correctly handling request reference in error cases.
755 Severity : enhancement
757 Description: Add LustreNetLink, a kernel-userspace communcation path.
759 Severity : enhancement
761 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
766 Description: access to llog context before init.
767 Details : move handling CATALOGS file at osc layer and forbid access to llog
771 Frequency : rare, only when client is writing to OST on local node
773 Description: Avoid deadlock for local client writes
774 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
775 memory freeing context. This allows OST threads to set the
776 PF_MEMALLOC flag on task structures in order to allocate memory
777 from reserved pools and complete IO.
778 Use GFP_HIGHUSER for OST allocations for non-local client writes,
779 so that the OST threads generate memory pressure and allow
780 inactive pages to be reclaimed.
782 Severity : enhancement
784 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
789 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
790 Details : move ldlm namespace creation in setup phase to avoid grab
791 _lprocfs_lock with cli_sem held.
795 Description: Temporarily disable grant shrink.
796 Details : Disable the feature for debugging.
799 Frequency : only when formatting test filesystems
801 Description: Unable to run several mkfs.lustre on loop devices at the same time
802 Details : mkfs.lustre returns error 256 on the concurrent loop devices
803 formatting. The solution is to proper handle the error.
805 Severity : enhancement
807 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
809 Severity : enhancement
811 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
816 Description: resolve race between obd_disconnect and class_disconnect_exports
817 Details : if obd_disconnect will be called to already disconnected export he
818 forget release one reference and osc module can't unloaded.
820 Severity : enhancement
822 Description: Allow tuning service thread via /proc
823 Details : For each service a new
824 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
825 created that can be used to set min/max thread counts, and get the
826 current number of running threads.
831 Descriptoin: don't return error if have partially created objects for file.
832 Details : lov_update_create_set uses set->set_success as index for created
833 objects, so if some requests will be failed, they will have hole
834 at end of array and qos_shrink_lsm for allocate correct lsm.
836 Severity : enhancement
838 Description: Update OFED support to 1.4.1
843 Description: Reduce small size read RPC
844 Details : Set read-ahead limit for every file and only do read-ahead when
845 available read-ahead pages are bigger than 1M to avoid small size
848 Severity : enhancement
850 Description: move AT tunable parameters for more consistent usage
851 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
853 Severity : enhancement
855 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
856 Details : allow skip disconnected ost for send statfs request and hide error
862 Description: fix racy locking of mballoc block bitmaps causing BUG
863 Details : The locking of the mballoc buddy bitmap and the in-memory
864 block bitmap was using two different spin locks in some
865 cases. This made it possible to incorrectly access the
866 mballoc bitmap while another process was modifying it,
867 causing a sanity assertion to fail. While no on-disk corruption
868 was reported, there was some risk of this happening.
871 Frequency : rare, on llog test 6
873 Descriptoin: don't allow connect to already connected import
874 Details : allowing connect to already connected import is hide connecting problem.
879 Descriptoin: allow kill process which wait statahead result
880 Details : for some reasons 'ls' can stick in waiting result from statahead,
881 in this case need way for kill this process.
883 Severity : enhancement
885 Description: Add state history info file, enhance import info file
886 Details : Track import connection state changes in a new osc/mdc proc file;
887 add overview-type data to the osc/mdc import proc file.
889 Severity : enhancement
891 Description: MDS create should not wait for statfs RPC while holding DLM lock.
893 Severity : enhancement
895 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
900 Descriptoin: OST grant too much space to client even there are not enough space.
901 Details : Client will shrink its grant cache to OST if there are no write
902 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
903 this grant cache if there are already not enough avaible space
904 (left_space < total_clients * 32M).
907 Frequency : start MDS on uncleanly shutdowned MDS device
909 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
910 Details : stay in waiting mds<>ost recovery finished produce random bugs
911 due race between two ll_sync thread for one lov target. send
912 ACTIVATE event only if connect realy finished and import have
916 Frequency : rare, connect and disconnect target at same time
918 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
919 Details : don't call obd_disconnect under lov_lock. this long time
920 operation and can block ptlrpcd which answer to connect request.
925 Descriptoin: don't lose wakeup for imp_recovery_waitq
926 Details : recover_import_no_retry or invalidate_import and import_close can
927 both sleep on imp_recovery_waitq, but we was send only one wakeup
931 Frequency : always with long access acl
933 Descriptoin: mds can't pack reply with long acl.
934 Details : mds don't control size of acl but they limited by reint/getattr
937 Severity : enhancement
939 Description: Update to SLES10 kernel-2.6.16.60-0.33.
941 Severity : enhancement
943 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
946 Frequency : start MDS on uncleanly shutdowned MDS device
948 Descriptoin: aborting recovery hang on MDS
949 Details : don't throttle destroy RPCs for the MDT.
952 Frequency : on remount
954 Description: external journal device not working after the remount
955 Details : clear dev_rdonly flag for external journal devices in
961 Description: shutdown vs evict race
962 Details : client_disconnect_export vs connect request race.
963 if client will evicted at this time - we start invalidate
964 thread without referece to import and import can be freed
968 Frequency : rare, need acl's on inode.
970 Description: client can't handle ost additional correctly
971 Details : if ost was added after client connected to mds client can have
972 hit lnet_try_match_md ... to big messages to wide striped files.
973 in this case need teach client to handle config events about add
974 lov target and update client max ea size at that event.
976 Severity : enhancement
978 Description: Changelogs
979 Details : Changelogs are a lightweight mechanism to track filesystem
980 metadata and namespace changes. The changelog is recorded
981 permanently on the MDTs, and is periodically "consumed" / purged
982 when records are no longer needed.
984 Severity : enhancement
986 Description: compact fld format with extents
987 Details : Store range of seq rather than every seq in FLD. Seq
988 controller update FLD rather than clients. In Case of CMD, mdt0
989 has FLD, all other metadata server act as non persistent proxy
990 for FLD queries and cache fld entries in fld cache.
995 Description: don't skip ost target if they assigned to file
996 Details : Drop slow OSCs if we can, but not for requested start idx.
997 This means "if OSC is slow and it is not the requested
998 start OST, then it can be skipped, otherwise skip it only
999 if it is inactive/recovering/out-of-space.
1003 Description: more cleanup in mds_lov
1004 Details : not send LOV EA under replay, we can't know about they size at this
1005 time. Don't allow client connect to mds before any ost connected,
1006 for avoid problems with LOV EA size and returning EIO to client.
1008 Severity : enhancement
1010 Description: Interoperability at server side (Disk interoperability)
1012 Severity : enhancement
1014 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1016 Severity : enhancement
1018 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1020 Severity : enhancement
1022 Description: New client IO stack (CLIO).
1024 Severity : enhancement
1026 Description: Commit on sharing. Eliminate inter-client dependencies between
1027 uncommitted transactions by doing transaction commits.
1028 Thereby clients may recovery independently.
1031 Frequency : Create a symlink file with a very long name
1033 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1034 Details : If there is no extra space in the request for early cancels,
1035 ldlm_req_handles_avail() returns 0 instead of a negative value.
1037 Severity : enhancement
1039 Description: Add /proc entry for import status
1040 Details : The mdc, osc, and mgc import directories now have
1041 an import directory that contains useful import data for debugging
1042 connection problems.
1044 Severity : enhancement
1046 Description: Re-disable certain /proc logging
1047 Details : Enable and disable client's offset_stats, extents_stats and
1048 extents_stats_per_process stats logging on the fly.
1051 Frequency : Only on FC kernels 2.6.22+
1053 Description: oops in statahead
1054 Details : Do not drop reference count for the dentry from VFS when lookup,
1055 VFS will do that by itself.
1057 Severity : enhancement
1059 Description: Generic /proc file permissions
1060 Details : Set /Proc file permissions in a more generic way to enable non-
1061 root users operate on some /proc files.
1065 Description: Hitting mdc_commit_close() ASSERTION
1066 Details : Properly handle request reference release in
1067 ll_release_openhandle().
1071 Description: quota recovery deadlock during mds failover
1072 Details : This patch includes att18982, att18236, att18237 in bz14840.
1074 1. fix osts hang when mds does failover with quotaon
1075 2. prevent watchdog storm when osts threads wait for the
1080 Frequency : only patchless client
1081 Description: add workaround for race between add/remove dentry from hash
1083 Severity : enhancement
1085 Description: Allow OST glimpses to return PW locks
1089 Description: LBUG when llog conf file is full
1090 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1094 Description: Prevent import from entering FULL state when server in recovery
1098 Description: service mount cannot take device name with ":"
1099 Details : Only when device name contains ":/" will mount treat it as
1105 Description: replace ptlrpcd with the statahead thread to interpret the async
1106 statahead RPC callback
1110 Frequency : on recovery
1111 Description: I/O failures after umount during fail back
1112 Details : if client reconnected to restarted server we need join to recovery
1113 instead of find server handler is changed and process self
1114 eviction with cancel all locks.
1116 Severity : enhancement
1118 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1120 Severity : enhancement
1122 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1124 Severity : enhancement
1126 Description: Upcall on Lustre log has been dumped
1127 Details : Allow for a user mode script to be called once a Lustre log has
1128 been dumped. It passes the filename of the dumped log to the
1129 script, the location of the script can be specified via
1130 /proc/sys/lnet/debug_log_upcall.
1135 Description: avoid idr_remove called for id which is not allocated.
1136 Details : Move assigment s_dev for clustered nfs to end of initialization,
1137 to avoid problem with error handling.
1142 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1143 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1144 into CONN_USED_HASH and this prodice warning when put connection
1145 again in unused hash.
1148 Severity : enhancement
1150 Description: File striping can now be set to use an arbitrary pool of OSTs.
1152 Severity : enhancement
1154 Description: Export bytes_read/bytes_write count on OSC/OST.
1158 Description: Early reply size mismatch, MGC loses connection
1159 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1160 the connect flags are properly negotiated.
1165 Description: quotas are not honored with O_DIRECT
1166 Details : all writes with the flag O_DIRECT will use grants which leads to
1167 this problem. Now using OBD_BRW_SYNC to guard this.
1171 Description: add quota statistics
1172 Details : 1. sort out quota proc entries and proc code.
1173 2. add quota statistics
1175 Severity : enhancement
1177 Description: enable quota support for HEAD.
1181 Description: Properly propagate oinfo flags from lov to osc for statfs
1182 Details : restore missing copy oi_flags to lov requests.
1184 Severity : enhancement
1186 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1187 : l_getgroups(8), lst(8), routerstat(8)
1188 Details : included man pages for llobdstat(8), llstat(8),
1189 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1191 Severity : enhancement
1193 Description: configure's --enable-quota should check the
1194 : kernel .config for CONFIG_QUOTA
1195 Details : configure is terminated if --enable-quota is passed but
1196 : no quota support is in kernel
1200 Description: Remove portals compatibility
1201 Details : Remove portals compatibility, not interoperable with releases
1206 Description: Resolve device initialization race
1207 Details : Prevent proc handler from accessing devices added to the
1208 obd_devs array but yet be intialized.
1210 Severity : enhancement
1212 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1214 Severity : enhancement
1216 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1221 Description: Using wrong pointer in osc_brw_prep_request
1222 Details : Access to array[-1] can produce panic if kernel compiled with
1223 CONFIG_PAGE_ALLOC enabled
1227 Description: Client runs out of low memory
1228 Details : Consider only lowmem when counting initial number of llap pages
1232 Description: *optional* service tags registration
1233 Details : if the "service tags" package is installed on a Lustre node
1234 When the filesystem is mounted, a local-node service tag will
1235 be created. See http://inventory.sun.com/ for more information
1236 about the Service Tags asset management system.
1240 Description: Kernel BUG tries to release flock
1241 Details : Lustre does not destroy flock lock before last reference goes
1242 away. So always drop flock locks when client is evicted and
1243 perform unlock regardless of successfulness of speaking to MDS.
1247 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1251 Description: sanity test 65a fails if stripecount of -1 is set
1252 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1257 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1258 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1261 Severity : enhancement
1263 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1267 Description: do not process already freed flock
1268 Details : flock can possibly be freed by another thread before it reaches
1269 to ldlm_flock_completion_ast.
1273 Description: LBUG during stress test
1274 Details : Need properly lock accesses the flock deadlock detection list.
1278 Description: oops in page fault handler
1279 Details : kernel page fault handler can return two special 'pages' in error
1280 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1284 Description: timeout with invalidate import.
1285 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1286 should be handled by ptlrpcd. This produce long age waiting and
1287 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1289 Severity : enhancement
1291 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1295 Description: enable MGS and MDT services start separately
1296 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1297 a co-located MGS without starting the MGS, which is a complement
1298 to 'nosvc' mount option.
1302 Frequency : after recovery
1303 Description: precreate to many object's after del orphan.
1304 Details : del orphan st in oscc last_id == next_id and this triger growing
1305 count of precreated objects. Set flag LOW to skip increase count
1306 of precreated objects.
1310 Frequency : rare, on clear nid stats
1311 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1312 Details : when clean nid stats sometimes try destroy live entry,
1313 and this produce panic in free.
1317 Description: Stack overflow during MDS log replay
1318 ease stack pressure by using a thread dealing llog_process.
1322 Description: wait until IO finished before start new when do lock cancel.
1323 Details : VM protocol want old IO finished before start new, in this case
1324 need wait until PG_writeback is cleared until check dirty flag and
1325 call writepages in lock cancel callback.
1327 Severity : enhancement
1329 Description: using special macro for print time and cleanup in includes.
1333 Description: mds_mfd_close() ASSERTION(rc == 0)
1334 Details : In mds_mfd_close(), we need protect inode's writecount change
1335 within its orphan write semaphore to prevent possible races.
1339 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1343 Frequency : rare, on shutdown ost
1344 Description: don't hit live lock with umount ost.
1345 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1346 use shrink_dcache_sb instead.
1350 Description: don't panic with use echo client
1351 Details : echo client pass NULL as client nid pointer and this produce null
1352 pointer dereference.
1356 Description: fix build on ppc32
1357 Details : compile code with -m64 flag produce wrong object file for ppc32.
1361 Description: add message levels for liblustreapi
1365 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1366 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1367 will result in an -EAGAIN and in the exit call path we call
1368 alloc_rr() which will with increasing aggressiveness attempt to
1369 aquire precreated objects on the minimum number of required OSCs.
1373 Description: Use old size assignment to avoid deadlock
1374 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1375 the scheduling while holding a spinlock. We do not need locking
1376 for size in ll_update_inode() because size is only updated from
1377 the MDS for directories or files without objects, so there is no
1378 other place to do the update, and concurrent access to such inodes
1379 are protected by the inode lock.
1383 Description: resolve "_IOWR redefined" build error on SLES10
1387 Description: dump the memory debugging after all modules are unloaded to
1388 suppress false negative in conf_sanity test 39
1390 Severity : enhancement
1392 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1397 Description: fix SLES kernel versioning
1398 Details : the kernel version for our SLES 10 kernel did not include a "-"
1399 before the "smp" at the end. while this was not a problem in
1400 general, it did mean that software trying to use the kernel
1401 version to try to detect a vendor specific kernel would fail.
1402 this was most evident by the OFED build scripts.
1406 Description: Don't update lov_desc members until making sure they are valid
1407 Details : When updating lov_desc members via proc fs, need fix their
1408 validities before doing the real update.
1412 Description: don't put request into delay list while invalidate in flight.
1413 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1414 import in flight. this produce timeout for invalidate and sometimes
1415 can cause stale data.
1419 Frequency : on ppc only
1420 Description: not convert ost objects for directory because it's not exist.
1421 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1425 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1427 Severity : enhancement
1429 Description: Update to support 2.6.22.14 vanilla kernel.
1433 Frequency : rare, on recovery
1434 Description: read procfs can produce deadlock in some situation
1435 Details : Holding lprocfs lock which send rpc can produce block for destroy
1436 obd objects and this also block reconnect with -EALREADY. This
1437 isn't fix all lprocfs bugs - but make it rare.
1439 Severity : enhancement
1441 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1444 Frequency : frequent on X2 node
1446 Description: mdc_set_open_replay_data LBUG
1447 Details : Set replay data for requests that are eligible for replay.
1451 Description: lustre_mgs: operation 101 on unconnected MGS
1452 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1453 MGC, and late on MGC cannot successfully connect to MGS and a lot
1454 of the error messages complaining that MGS is not connected.
1458 Frequency : on network error
1459 Description: panic with double free request if network error
1460 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1461 but it's true only for synchronus enqueue, for async enqueue
1462 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1465 Severity : enhancement
1467 Description: client-side metadata stat-ahead during readdir(directory readahead)
1468 Details : perform client-side metadata stat-ahead when the client detects
1469 readdir and sequential stat of dir entries therein
1472 Frequency : on start mds
1474 Description: Implement get_info(last_id) in obdfilter.
1477 Frequency : occasional
1479 Description: Correctly check stale fid, not start epoch if ost not support SOM
1480 Details : open with flag O_CREATE need set old fid in op_fid3 because
1481 op_fid2 was overwritten with new generated fid, but mds can answer
1482 with one of these two fids and both is not stale. Setattr
1483 incorrectly started an epoch and assume will be called
1484 done_writing, but without SOM done_writing ever being called.
1487 Frequency : rare, depends on device drivers and load
1489 Description: MDS or OSS nodes crash due to stack overflow
1490 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1491 In some cases, in conjunction with device drivers that use a lot
1492 of stack the MDS (or possibly OSS) service threads could overflow
1493 the stack. One change which was identified to consume additional
1494 stack has been reworked to avoid the extra stack usage.
1497 Frequency : occasional
1499 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1500 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1501 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1502 oscc_has_objects and see OSC has no objests and call
1503 oscc_internal_create to resend create request.
1505 Severity : enhancement
1507 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1509 Severity : enhancement
1511 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1514 Frequency : very rare
1516 Description: Fix replay if there is an un-replied request and open
1517 Details : In some cases, older replay request will revert the
1518 mcd->mcd_last_xid on MDS which is used to record the client's
1519 latest sent request.
1521 Severity : enhancement
1523 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1525 Severity : enhancement
1527 Description: Add rhel5 support to HEAD.
1529 Serverity : enhancement
1531 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1536 Description: Don't allow skipping OSTs if index has been specified.
1537 Details : Don't allow skipping OSTs if index has been specified, make
1538 locking in internal create lots better.
1542 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1543 Details : access to bitfield in structure is always rounded to long
1544 and this produce problem with not atomic change any bit.
1548 Description: Lustre make rpms failed.
1549 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1550 builds Lustre rpms from tarball.
1553 Frequency : rare on shutdown ost
1555 Description: If llog cancel was not send before clean_exports phase, this can
1556 produce deadlock in llog code.
1557 Details : If llog thread has last reference to obd and call class_import_put
1558 this produce deadlock because llog_cleanup_commit_master wait when
1559 last llog_commit_thread exited, but this never success because was
1560 called from llog_commit_thread.
1564 Description: allow userland application know is lost one of stripes.
1565 Details : fill lvb_blocks with error code on ost and return it to
1566 application if error flag found.
1570 Description: NULL lov_tgts causing MDS oops
1571 Details : more safe checks for NULL lov_tgts for avoid oops.
1573 Severity : enhancement
1575 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1579 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1580 Details : If an OST is down the MDS will hang indefinitely in
1581 obd_statfs_rqset() waiting for the statfs data. While for
1582 MDS QOS usage of statfs, it should not stuck in waiting.
1584 Severity : enhancement
1586 Description: remote_acl support
1587 Details : Support ACL-based permission check for remote user.
1588 Support setfacl/getfacl for remote user with the utils
1589 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1590 the system "{s,g}etfacl" utils.
1592 Severity : enhancement
1594 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1596 Severity : enhancement
1598 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1602 Description: make mgs_setparam() handle fsname containing dash
1603 Details : fsname containing a dash does not work with lctl conf_param
1605 Severity : enhancement
1607 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1609 Severity : enhancement
1611 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1615 Frequency : rare, at shutdown
1616 Description: access already free / zero obd_namespace.
1617 Details : if client_disconnect_export was called without force flag set,
1618 and exist connect request in flight, this can produce access to
1619 NULL pointer (or already free pointer) when connect_interpret
1620 store ocd flags in obd_namespace.
1624 Frequency : only at startup
1625 Description: not alloc memory with spinlock held.
1626 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1627 if any spinlock held.
1629 Severity : enhancement
1631 Description: make lustre randomly fail allocating memory
1632 Details : Make lustre randomly failed allocating memory for testing purpose.
1634 Severity : enhancement
1636 Description: lost problems with lov objid file
1637 Details : Fixes some scability and access to not inited memory problems
1638 in work with lov objdid file.
1643 Description: lfs find does not continue on file error
1644 Details : Continue other files processing when a file/dir is absent.
1648 Description: Inconsistent usage of lustre_pack_reply()
1649 Details : Standardize the usage of lustre_pack_reply() such that it
1650 always generate a CERROR on failure.
1653 Frequency : occasional
1655 Description: MDS hang or stay in waiting lock
1656 Details : If client receive lock with CBPENDING flag ldlm need send lock
1657 cancel as separate rpc, to avoid situation when cancel request
1658 can't processed due all i/o threads stay in wait lock.
1662 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1666 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1670 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1674 Description: Update to RHEL4 latest kernel.
1676 Severity : enhancement
1678 Description: Build SLES10 patchless client fails
1679 Details : The configure was broken by run ./configure with
1680 --with-linux-obj=.... argument for patchless client. When the
1681 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1682 can't search header adequately. Use absolute path such as
1683 -I($LINUX)/include instead.
1687 Description: interrupt oig_wait produce painc on resend.
1688 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1689 private set, and this produce situation when rq_ptlrpcd_data not
1690 copyed to new allocated request and triggered LBUG on assert
1691 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1692 ptlrpcd set if request is changed and can be safety changed to use
1697 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1698 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1699 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1700 these macros and replace them with their difinitions instead.
1704 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1705 Details : Update which_patch & target file for SLES10 latest kernel.
1708 Description: Remove client patches from SLES 10 kernel.
1709 Details : This causes SLES 10 clients to behave as patchless clients
1710 even on a Lustre-patched (server) kernel.
1712 Severity : enhancement
1714 Description: self-adjustable client's lru lists
1715 Details : use adaptive algorithm for managing client cached locks lru
1716 lists according to current server load, other client's work
1717 pattern, memory activities, etc. Both, server and client
1718 side namespaces provide number of proc tunables for controlling
1721 Severity : enhancement
1723 Description: light-weight GSS support
1724 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1725 and reduce performance overhead.
1727 Severity : enhancement
1729 Description: Linux keyring support
1730 Details : Support using service of Linux keyring for Lustre GSS internal
1731 context refresh/cache mechanism.
1735 Description: Fix errors in lfs documentation
1736 Details : Fixes man pages
1740 Description: don't use GFP_* in generic Lustre code.
1741 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1745 Description: obdclass is limited by single OBD_ALLOC(idarray)
1746 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1750 Description: warnings with build patchless client with vanila 2.6.19 and up
1751 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1752 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1756 Frequency : Only for SLES
1757 Description: sanity_quota fail test_1
1758 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1759 file, which makes TSTID[2] inunique.
1763 Frequency : Only for SLES9
1764 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1765 Details : Update target/ChangeLog/which_patch .
1769 Frequency : Only for test_10 in sanity-quota.sh
1770 Description: a bug in quota test code
1771 Details : checking a test flag in a wrong place
1775 Frequency : Only for test_13 in sanity-quota.sh
1776 Description: a bug in quota test script
1777 Details : shouldn't check overfull allocated quota
1781 Description: jbd statistics
1782 Details : Port older jbd statistics patch for sles10
1784 Severity : enhancement
1786 Description: Add printing inode info into message about error in writepage.
1791 Description: reply_lock_interpret crash due to race with it and lock cancel.
1792 Details : Do not replay locks that are being cancelled. Do not reference
1793 locks by their address during replay, just by their handle.
1797 Frequency : When flocks are used.
1798 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1799 Details : Flock locks might destroy just granted lock if it could be merged
1800 with another existing flock, this is done in completion handler,
1801 so teach ldlm_cli_enquque_fini that this is a valid case for
1807 Description: Oops in read and write path when failing to allocate lock.
1808 Details : Check if lock allocation failed and return error back.
1812 Description: lstripe command fails for valid OST index
1813 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1814 instead of lov->desc.ld_active_tgt_count.
1818 Description: sanity.sh failed test 103
1819 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1824 Description: df doesn't work properly if diskfs blocksize != 4K
1825 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1829 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1830 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1832 Severity : enhancement
1834 Description: Adaptive timeouts
1835 Details : RPC timeouts adapt to changing server load and network
1836 conditions to reduce resend attempts and improve recovery time.
1840 Description: llapi_file_create() does not allow some changes
1841 Details : add llapi_file_open() that allows specifying the mode and
1842 open flags, and also returns an open file handle.
1846 Description: merge and cleanup kernel patches.
1847 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1851 Description: Add journal checksum support.(Kernel part)
1852 Details : The journal checksum feature adds two new flags i.e
1853 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1854 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1855 indicates that the commit block contains the checksum for
1856 the blocks described by the descriptor blocks. Now commit
1857 record can be sent to disk without waiting for descriptor
1858 blocks to be written to disk. This behavior is controlled
1859 using JBD2_FEATURE_ASYNC_COMMIT flag.
1863 Description: OSS needs mutliple precreate threads
1864 Details : Add ability to start more than one create thread per OSS.
1868 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1869 Details : execute lfs setstripe on client
1873 Description: mds_obd_create error creating tmp object
1874 Details : When the user sets quota on root, llog will be affected and can't
1875 create files and write files.
1878 Frequency : Always on ia64 patchless client, and possibly others.
1880 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1881 Details : This allows the patchless client to be loaded on architectures
1882 without this export.
1886 Description: RedHat Update kernel for RHEL5
1887 Details : Add the kernel config file for RHEL5.
1891 Description: RedHat Update kernel for RHEL5
1892 Details : Modify the kernel config file more closer RHEL5.
1896 Description: Build failure against Centos5 (RHEL5)
1897 Details : Define PAGE_SIZE when it isn't present.
1901 Description: test_7 on sanity-quota.sh can't be used on separate machines
1902 Details : add facet to handle it
1906 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1907 Details : with NFS, the anon dentry's parent was set to itself in
1908 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1909 corresponding dentry other than use rec->ur_name.
1911 Severity : enhancement
1913 Description: lfs setstripe enhancement
1914 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1918 Description: enable data checksumming by default
1919 Details : enable checksum by default, allow --disable-checksum
1920 configure option and "-o nochecksum" mount option. Checksums
1921 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1925 Description: lustre support for RHEL5
1926 Details : Add support for RHEL5.
1930 Description: Client eviction due to failover config
1931 Details : after a connection loss, the lustre client should attempt to
1932 reconnect to the last active server first before trying the
1933 other potential connections.
1937 Description: when mds and osts use different quota unit(32bit and 64bit),
1938 quota will be released repeatly.
1939 Details : void sending multiple quota reqs to mds, which will keep the
1940 status between the reqs.
1944 Description: osts not allocated evenly to files
1945 Details : change the condition to increase offset_idx
1948 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1950 Description: Data corruption for OSTs that are formatted larger than 2TB
1952 Details : When generating the bio request for lustre file writes the
1953 sector number would overflow a temporary variable before being
1954 used for the IO. The data reads correctly from Lustre (which
1955 will overflow in a similar manner) but other file data or
1956 filesystem metadata may be corrupted in some cases.
1960 Description: Tune the kernel for good SCSI performance.
1961 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1962 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1967 Description: rewrite ext2-derived code in obdclass/uuid.c
1968 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1969 from scratch preserving functionality.
1973 Description: "lfs find -obd UUID" prints directories
1974 Details : "lfs find -obd UUID" will return all directory names instead
1975 of just file names. It is incorrect because the directories
1976 do not reside on the OSTs.
1980 Description: use i_size_read and i_size_write in 2.6 port
1981 Details : replace inode->i_size access with i_size_read/write()
1984 Frequency : when removing large files
1986 Description: scheduling issue during removal of large Lustre files
1987 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1988 It causes scheduling issues when removing large files (17TB in the
1992 Frequency : only with liblustre clients on XT3
1994 Description: evictions taking too long
1995 Details : allow llrd to evict clients directly on OSTs
1998 Frequency : only on ppc
2000 Description: /proc/fs/lustre/devices broken on ppc
2001 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2002 We should make sure the type of 'index' is loff_t and then cast
2003 explicitly as needed below. Do not assign an explicitly cast
2007 Frequency : only for rhel5
2009 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2010 Details : Modify the target file & which_kernel.
2012 Severity : enhancement
2014 Description: omit set fsid for export NFS
2015 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2016 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2020 Frequency : Always, for kernels after 2.6.16
2021 Description: Fix warning idr_remove called for id=.. which is not allocated.
2022 Details : Last kernels save old s_dev before kill super and not allow
2023 to restore from callback - restore it before call kill_anon_super.
2027 Description: buffer overruns could theoretically occur
2028 Details : llapi_semantic_traverse() modifies the "path" argument by
2029 appending values to the end of the origin string, and a
2030 overrun may occur. Adding buffer overrun check in liblustreapi.
2034 Description: Fix error on 'ls .' at the top of the Lustre mount.
2035 Details : Don't revalidate dentry if it is a root dentry.
2039 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2040 Details : Modify vm-tunables-rhel4.patch.
2044 Description: Kernel config for 2.6.18-vanilla.
2045 Details : Modify targets/2.6-vanilla.target.in.
2046 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2047 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2048 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2049 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2053 Description: improve handling recoverable errors
2054 Details : if request processig with error which can be recoverable on server
2055 request should be resend, otherwise page released from cache and
2060 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2061 Details : Modify target file & which_patch.
2062 A flaw was found in the IA32 system call emulation provided
2063 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2064 value could be stored in the %RAX register, which could trigger an
2065 out-of-bounds system call table access. An untrusted local user
2066 could exploit this flaw to run code in the kernel
2067 (ie a root privilege escalation). (CVE-2007-4573).
2071 Description: O_DIRECT bypasses client statistics.
2072 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2073 read_ahead_stats were updated. copied the stats section from
2074 osc_send_oap_rpc() into async_internal().
2078 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2082 Description: change order of libsysio includes
2083 Details : '#include sysio.h' should always come before '#include xtio.h'
2087 Description: handle "serious error: objid * already exists" more gracefully
2088 Details : If LAST_ID value on disk is smaller than the objects existing in
2089 the O/0/d* directories, it indicates disk corruption and causes an
2090 LBUG(). If the object is 0-length, then we should use the existing
2091 object. This will help to avoid a full fsck in most cases.
2093 Severity : enhancement
2095 Description: adapt the lustre_config script to support the upgrade case
2096 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2097 server targets to 1.6 in parallel.
2101 Description: To avoid grant space > avaible space when the disk is almost
2102 full. Without this patch you might see the error "grant XXXX >
2103 available" or some LBUG about grant, when the disk is almost
2105 Details : In filter_check_grant, for non_grant cache write, we should
2106 check the left space by if (*left > ungranted + bytes), instead
2107 of (*left > ungranted), because only we are sure the left space
2108 is enough for another "bytes", then the ungrant space should be
2109 increase. In client, we should update cl_avail_grant only there
2110 is OBD_MD_FLGRANT in the reply.
2114 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2115 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2116 described in CVE-2007-4573. This problem could allow untrusted
2117 local users to gain root access.
2120 Frequency : when using O_DIRECT and quotas
2122 Description: Incorrect file ownership on O_DIRECT output files
2123 Details : block usage reported by 'lfs quota' does not take into account
2124 files that have been written with O_DIRECT.
2129 Description: touch file failed when fs is not full
2130 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2131 otherwise we can get ENOSP while fs is not full.
2135 Description: parallel lock callbacks
2136 Details : Instead of sending blocking and completion callbacks as separated
2137 requests, adding them to a set and sending in parallel.
2140 Frequency : only for Cray XT3
2141 Bugzilla : 12829/13455
2142 Description: Changing primary group doesn't change the group lustre assigns to
2144 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2145 overridden with the primary group provided by the group upcall,
2146 whereas the supplied fsgid can be trusted if it is in the list of
2147 supplementary groups returned by the group upcall.
2149 Severity : enhancement
2151 Description: Allow masking D_WARNING, D_ERROR messages from console
2152 Details : Console messages can now be disabled via lnet.printk.
2156 Description: User code with malformed file open parameter crashes client node
2157 Details : Before packing join_file req, all the related reference should be
2158 checked carefully in case some malformed flags cause fake
2159 join_file req on client.
2163 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2164 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2165 mode is equal to the granted mode. However, it should also check
2166 that LDLM_FL_CP_REQD is not set.
2170 Description: Heavy nfs access might result in deadlocks
2171 Details : After ELC code landed, it is now improper to enqueue any mds
2172 locks under och_sem, because enqueue might want to decide to
2173 cancel open locks for same inode we are holding och_sem for.
2177 Description: Client eviction while running blogbench
2178 Details : A lot of unlink operations with concurrent I/O can lead to a
2179 deadlock causing evictions. To address the problem, the number of
2180 oustanding OST_DESTROY requests is now throttled to
2181 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2182 callbacks are processed in priority.
2186 Description: enable ACLs on MDS by default
2187 Details : ACLs must be enabled on MDS by default.
2190 Frequency : PPC/PPC64 only
2192 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2193 Details : fix duplicated definitions between asm-ppc64/types.h and
2194 lustre_types.h on PPC.
2197 Frequency : PPC/PPC64 only
2199 Description: asm-ppc/segment.h does not exist
2200 Details : fix compile issue on PPC.
2204 Description: better handle error messages in extents code
2207 Frequency : RHEL4 only
2209 Description: mkfs is very slow on IA64/RHEL4
2210 Details : A performance regression has been discovered in the MPT Fusion
2211 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2212 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2213 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2215 Severity : enhancement
2217 Description: SNMP support enhancement
2218 Details : Adding total number of sampled request for an MDS node in snmp
2221 Severity : enhancement
2223 Description: Optimize ldlm waiting list processing for PR extent locks
2224 Details : When processing waiting list for read extent lock and meeting read
2225 lock that is same or wider to it that is not contended, skip
2226 processing rest of the list and immediatelly return current
2227 status of conflictness, since we are guaranteed there are no
2228 conflicting locks in the rest of the list.
2232 Description: Time out and refuse to reconnect
2233 Details : When the failover node is the primary node, it is possible
2234 to have two identical connections in imp_conn_list. We must
2235 compare not conn's pointers but NIDs, otherwise we can defeat
2236 connection throttling.
2240 Description: port llog fixes from b1_6 into HEAD
2241 Details : Port llog reference couting and some llog cleanups from b1_6
2242 (bug 10800) into HEAD, for protect from panic and access to already
2243 free llog structures.
2247 Description: Detect stride IO mode in read-ahead
2248 Details : When a client does stride read, read-ahead should detect that and
2249 read-ahead pages according to the detected stride pattern.
2253 Description: data checksumming impacts single node performance
2254 Details : add support for several checksum algorithm. Currently, only CRC32
2255 and Adler-32 are supported. The checksum type can be changed on
2256 the fly via /proc/fs/lustre/osc/*/checksum_type.
2260 Description: use adler32 for page checksums
2261 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2266 Description: build for x2 fails
2267 Details : fix compile issue on Cray systems.
2271 Description: Properly match for duplicate locks
2272 Details : Due to different lock order from skiplists code, we need to
2273 traverse entire list for now
2276 Frequency : only on PPC/SLES10
2278 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2279 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2280 be defined. Add a hack in mkfs_lustre.c to work around this compile
2285 Description: LASSERT on MDS when client holding flock lock dies
2286 Details : ldlm pool logic depends on number of granted locks equal to
2287 number of released locks which is not true for flock locks, so
2288 just exclude such locks from consideration.
2292 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2293 Details : Use fsfilt_sync() for both the whole filesystem sync and
2294 individual file sync to eliminate dangerous inode locking
2295 with I_LOCK that can lead to a deadlock.
2299 Description: performance in 1.6.3
2300 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2301 sure that 1MB requests are not fragmented by the block layer.
2303 Severity : enhancement
2305 Description: organize the server-side client stats on per-nid basis
2306 Details : Change the structure of stats under obdfilter and mds to
2314 The "uuid"s file would list the uuids of _active_ exports.
2315 And the clear entry is to clear all stats and stale nids.
2317 Severity : enhancement
2319 Description: eliminate client locks in face of contention
2320 Details : file contention detection and lockless i/o implementation
2321 for contended files.
2325 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2328 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2330 Description: MDS may lose file striping (and hence file data) in some cases
2331 Details : If there are additional extended attributes stored on the MDS,
2332 in particular ACLs, SELinux, or user attributes (if user_xattr
2333 is specified for the client mount options) then there is a risk
2334 of attribute loss. Additionally, the Lustre file striping
2335 needs to be larger than default (e.g. striped over all OSTs),
2336 and an additional attribute must be stored initially in the
2337 inode and then increase in size enough to be moved to the
2338 external attribute block (e.g. ACL growing in size) for file
2343 Description: skiplist implementation simplification
2344 Details : skiplists are used to group compatible locks on granted list
2345 that was implemented as tracking first and last lock of each
2346 lock group the patch changes that to using doubly linked lists
2350 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2351 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2352 because LASSERT on both the data supplied by a client, and the
2353 data on disk is dangerous and incorrect.
2355 Severity : enhancement
2357 Description: Slow truncate/writes to huge files at high offsets.
2358 Details : Directly associate cached pages to lock that protect those pages,
2359 this allows us to quickly find what pages to write and remove
2360 once lock callback is received.
2364 Description: more ldlm soft lockups
2365 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2366 starve other threads from the resource lock for a long time in
2367 case of long waiting queue, so change the debug level from
2368 D_OTHER to the less frequently used D_INFO.
2370 Severity : enhancement
2372 Description: add -gid, -group, -uid, -user options to lfs find
2376 Description: Hung threads in invalidate_inode_pages2_range
2377 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2378 once one is completed. As a result, some RPCs are stuck in the
2379 queue and are never sent.
2383 Description: filter threads hungs on waiting journal commit
2384 Details : Cleanup filter group llog code, then only filter group llog will
2385 be only created in the MDS/OST syncing process.
2389 Description: Procfs and llog threads access destoryed import sometimes.
2390 Details : Sync the import destoryed process with procfs and llog threads by
2391 the import refcount and semaphore.
2393 Severity : enhancement
2395 Description: openlock cache of b1_6 port to HEAD
2400 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2401 Details : If insertion of an extent fails, then discard the inode
2402 preallocation and free data blocks else it can lead to duplicate
2407 Description: don't always update ctime in ext3_xattr_set_handle()
2408 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2409 In some cases the ctime should not be updated, for example for
2410 2.0->1.8 compatibility it is necessary to delete an xattr and it
2411 should not update the ctime.
2415 Bugzilla : 15713/16362
2416 Description: Assertion in iopen_connect_dentry in 1.6.3
2417 Details : looking up an inode via iopen with the wrong generation number can
2418 populate the dcache with a disconneced dentry while the inode
2419 number is in the process of being reallocated. This causes an
2420 assertion failure in iopen since the inode's dentry list contains
2421 both a connected and disconnected dentry.
2425 Description: assertion failure in ldlm_handle2lock()
2426 Details : fix a race between class_handle_unhash() and class_handle2object()
2427 introduced in lustre 1.6.5 by bug 13622.
2432 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2433 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2438 Description: oops in sd_iostats_seq_show()
2439 Details : unloading/reloading the scsi low level driver triggers a kernel
2440 bug when trying to access the sd iostat file.
2445 Description: Kernel panics during QLogic driver reload
2446 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2447 patch, causing memory corruption.
2452 Description: journal_dev option does not work in b1_6
2453 Details : pass mount option during pre-mount.
2455 Severity : enhancement
2457 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2458 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2459 extent information of a file. It can be used to map logical blocks
2460 in a file to physical blocks in the block device.
2464 Description: LDLM soft lockups - improvement
2465 Details : It is be possible to send the lock handle along with each read
2466 or write request because the client is already doing a lock match
2467 itself so there isn't any reason the OST should have to re-do that
2471 Frequency : only on Cray X2
2473 Description: X2 build failures
2474 Details : fix build failures on Cray X2.
2478 Description: xid & resent requests
2479 Details : Initialize RPC XID from clock at startup (randomly if clock is
2482 Severity : enhancement
2484 Description: Add lustre_start utility to start or stop multiple Lustre servers
2489 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2490 Details : In case of memory pressure, list_del() can be called twice on
2491 req->rq_history_list, causing a kernel oops.
2495 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2496 Details : fix stack overflow in the distributed lock manager by defering
2497 export eviction after a failed AST to the elt thread instead of
2498 handling it in the dlm interpret routine.
2502 Description: Convert some comments to new format.
2503 Details : Update documenting comments to match doxygen conventions.
2507 Description: Grammar fixes.
2508 Details : A couple of trivial spelling fixes.
2512 Description: OSD_COUNTERS-mandatory
2513 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2514 to simplify the code.
2518 Description: simplify cmm_device freeing logic.
2519 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2523 Description: Add lockdep support to dt_object_operations locking interface.
2524 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2525 parameter indicating lock ordering. Update mdd code to use new
2530 Description: Introduce failloc constants for lockless IO tests.
2531 Details : Add two new failloc constants to test lockless IO. Only one of
2532 them in implemented---another is checked in yet to be landed
2537 Description: Add lockdep support for inode mutex.
2538 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2539 in the situations where Lustre has to lock more than one inode
2544 Description: Add optional invariants checking support.
2545 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2546 configure switch. This macro is to be used for consistency and
2547 sanity checks that are too expensive to be left in `production'
2552 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2553 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2558 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2559 Details : Use spin_lock_nested() in (the only) situation where more than
2560 one ldlm_lock is locked simultaneously. Also, fix possible
2561 dead-lock in ldlm_lock_change_resource() by enforcing particular
2566 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2567 Details : Instead of specifying each ldlm_lock call-back through separate
2568 parameter, wrap them into struct ldlm_callback_suite.
2572 Description: Kill join_lru obd method and its callers.
2573 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2574 in memory---a requirement implemented through ->o_join_lru() obd
2575 method in HEAD. Remove this method and its users.
2579 Description: Add asynchronous ldlm ENQUEUE completion handler.
2580 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2581 case that stock ldlm_completion_ast() cannot handle as it waits
2582 until lock is granted. Introduce new ldlm_completion_ast_async()
2583 for this. Also comment ldlm_completion_ast().
2587 Description: ldlm_error <-> errno conversion.
2588 Details : Add functions to map (rather arbitrary) between LDLM error codes
2589 and standard errno values. CLIO needs this to prevent LDLM specific
2590 constants from escaping ldlm and osc.
2594 Description: Kill unused ldlm_handle2lock_ns() function.
2595 Details : Kill unused ldlm_handle2lock_ns() function.
2599 Description: Add lu_ref support to ldlm_lock
2600 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2601 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2602 and ldlm_resource. LDLM interface has to be changed, because code that
2603 releases a reference on a lock, has to "know" what reference this is.
2604 In the most frequent case
2606 lock = ldlm_handle2lock(handle);
2608 LDLM_LOCK_PUT(lock);
2610 no changes are required. When any other reference (received _not_ from
2611 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2612 instead of LDLM_LOCK_PUT().
2614 Arguably, changes are pervasive, and interface requires some discipline
2615 for proper use. On the other hand, it was very instrumental in finding
2616 a few leaked lock references.
2620 Description: Add ldlm_lock_addref_try().
2621 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2622 attempts to addref a lock that might be being canceled
2627 Description: Add ldlm_weigh_callback().
2628 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2629 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2630 hard-coded `number of pages' logic.
2634 Description: Add lockdep annotations to llog code.
2635 Details : Use appropriately tagged _nested() locking calls in the places
2636 where llog takes more than one ->lgh_lock lock.
2640 Description: Add loi_kms_set().
2641 Details : Wrap kms updates into a helper function.
2645 Description: Constify instances of struct lsm_operations.
2646 Details : Constify instances of struct lsm_operations.
2650 Description: lu_conf support.
2651 Details : On a server, a file system object is uniquely identified
2652 by a fid, which is sufficient to locate and load all object
2653 state (inode). On a client, on the other hand, more data are
2654 necessary instantiate an object. Change lu_object_find() and
2655 friends to take additional `lu_conf' argument describing object.
2656 Typically this includes layout information.
2660 Description: lu_context fixes.
2661 Details : Introduce new lu_context functions that are needed on the client
2662 side, where some system threads (ptlrpcd) are shared by multiple
2663 modules, and so cannot be stopped during module shutdown.
2667 Description: Add start and stop methods to lu_device_type_operations.
2668 Details : Introduce two new methods in lu_device_type_operations, that are
2669 invoked when first instance of a given type is created and last one
2670 is destroyed respectively. This is need by CLIO.
2674 Description: Add lu_ref support to struct lu_device.
2675 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2676 track leaked references.
2680 Description: Introduce lu_kmem_descr.
2681 Details : lu_kmem_descr and its companion interface allow to create
2682 and destroy a number of kmem caches at once.
2686 Description: Fix lu_object finalization race.
2687 Details : Fix a race between lu_object_find() finding an object and its
2688 concurrent finalization. This race is (most likely) not possible
2689 on the server, but might happen on the client.
2693 Description: Introduce lu_ref interface.
2694 Details : lu_ref is a debugging module allowing to track references to
2695 a given object. It is quite cpu expensive, and has to be
2696 explicitly enabled with --enable-lu_ref. See usage description
2701 Description: Factor lu_site procfs stats into a separate function.
2702 Details : Separate lu_site stats printing code into a separate function
2703 that can be reused on a client.
2707 Description: Constify instances of struct {lu,dt,md}_device_operations.
2708 Details : Constify instances of struct {lu,dt,md}_device_operations.
2712 Description: Introduce struct md_site and move meta-data specific parts of
2713 struct lu_site here.
2714 Details : Move md-specific fields out of struct lu_site into special struct
2715 md_site, so that lu_site can be used on a client.
2719 Description: Kill mdd_lov_destroy().
2720 Details : Remove unused mdd code.
2724 Description: Add st_block checking to multistat.c.
2725 Details : Add st_block checking to multistat.c.
2729 Description: Add lu_ref support to struct obd_device.
2730 Details : Add lu_ref tracking to obd_device.
2734 Description: Kill obd_set_fail_loc().
2735 Details : Remove unused code.
2739 Description: Add special type for ptlrpc_request interpret functions.
2740 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2741 there. Actual usage will be in CLIO.
2745 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2746 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2747 doesn't work with lockdep.
2751 Description: Add rwv.c test program.
2752 Details : New testing program exercising readv(2) and writev(2) (Qian).
2756 Description: Add sendfile.c test program.
2757 Details : New testing program exercising sendfile(2) (Jay).
2761 Description: Ratelimit a message that can be very frequent.
2762 Details : Ratelimit a memory allocation failure message that can
2767 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2768 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2772 Description: Liblustre build fixes.
2773 Details : Liblustre build fixes.
2777 Description: libcfs: add cfs_{need,cond}_resched() interface.
2778 Details : libcfs: add cfs_{need,cond}_resched() definition and
2779 implementations for Linux, NT, and liblustre.
2781 Severity : enhancement
2783 Description: More exported tunables for mballoc
2784 Details : Add support for tunable preallocation window and new tunables for
2785 large/small requests
2789 Description: Detect corruption of block bitmap and checking for preallocations
2790 Details : Checks validity of on-disk block bitmap. Also it does better
2791 checking of number of applied preallocations. When corruption is
2792 found, it turns filesystem readonly to prevent further corruptions.
2796 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2797 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2798 window to 0. If it does hit cache miss, and read-pattern is still
2799 stride-io mode, does not reset the stride window, but also does not
2800 increase the stride window length in this case.
2804 Frequency : only for big-endian servers
2805 Description: Check if system is big-endian while mounting fs with extents feature
2806 Details : Mounting a filesystem with extents feature will fail on big-endian
2807 systems since ext3-based ldiskfs is not supported on big-endian
2808 systems. This can be over-riden with "bigendian_extents" mount option.
2810 Severity : enhancement
2812 Description: The root squash functionality
2813 Details : A security feature, which is to prevent users from being able
2814 to mount lustre on their desktop, run as root, and delete
2815 all of the files in the filesystem. The goal is accomplished by
2816 remapping user id (UID) and group id (GID) of the root user to
2817 a UID and GID specified by the system administartor via Lustre
2818 configuration management server (MGS). The functionality also
2819 allows to specify sets of clients for which the remapping does
2824 Description: Excessive recovery window
2825 Details : With AT enabled, the recovery window can be excessively long (6000+
2826 seconds). To address this problem, we no longer use
2827 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2828 timeout no longer depends on the service time, it is set to
2829 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2830 time via pb_service_time.
2834 Description: Watchdog triggered on MDS failover
2835 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2836 the OSTs know that the MDS "UUID" can be reused for the same export
2837 from a different NID, so we do not need to wait for the export to be
2841 Frequency : rare, only if using MMP with Linux RAID
2843 Description: MMP doesn't work with Linux RAID
2844 Details : While using HA for Lustre servers with Linux RAID, it is possible
2845 that MMP will not detect multiple mounts. To make this work we
2846 need to unplug the device queue in RAID when the MMP block is being
2847 written. Also while reading the MMP block, we should read it from
2848 disk and not the cached one.
2850 Severity : enhancement
2852 Description: open file using fid
2853 Details : A file can be opened using just its fid, like
2854 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2857 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2860 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2861 Details : add proper locking for creation of jbd_4k slab cache
2865 Description: MMP check in ext3_remount() fails without displaying any error
2866 Details : When multiple mount protection fails during remount, proper error
2869 Severity : enhancement
2871 Description: Allow stripe size to be up to 4G-64k
2872 Details : Fix math logic to allow large stripe sizes.
2876 Description: add check for >8TB ldiskfs filesystems
2877 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2878 Don't allow >8TB ldiskfs filesystems to be mounted without
2879 force_over_8tb mount option
2884 Descriptoin: OSS complains "too many blocks" and crashes
2885 Details : In filter_commitrw_write():
2886 1. lquota_chkquota();
2887 2. obdo_from_inode();
2888 3. lquota_pending_commit();
2889 "2." has changed uid and gid in this case, then caused
2890 lquota_chkquota() and lquota_pending_commit() to use
2891 different uid and gid. That is the root of the bug.
2895 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2896 Details : Detect and print out the 'stripe count = -1' case properly.
2900 Description: Out or order replies might be lost on replay
2901 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2902 request with tid smaller than one currently being added, add it
2903 to the start, not end of the list.
2905 --------------------------------------------------------------------------------
2907 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2909 * Support for kernels:
2910 2.6.5-7.283 (SLES 9),
2911 2.6.9-55.EL (RHEL 4),
2912 2.6.16.46-0.14 (SLES 10),
2913 2.6.18.8 vanilla (kernel.org)
2914 * Client support for unpatched kernels:
2915 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2916 2.6.16 - 2.6.22 vanilla (kernel.org)
2917 * Due to recently discovered recovery problems, we do not recommend
2918 using patchless RHEL 4 clients with this or any earlier release.
2919 * Recommended e2fsprogs version: 1.39.cfs8
2920 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2921 * Starting with this release, the ldiskfs backing filesystem required
2922 by Lustre is now in its own package, lustre-ldiskfs. This package
2923 should be installed. It is versioned separately from Lustre and
2924 may be released separately in future.
2928 Description: block reactivating mgc import until all deactivates complete
2929 Details : Fix race when failing back MDT/MGS to itself (testing)
2932 Frequency : at statup only
2934 Description: mds_lov_synchronize race leads to various problems
2935 Details : simultaneous MDT->OST connections at startup can cause the
2936 sync to abort, leaving the OSC in a bad state.
2938 Severity : enhancement
2940 Description: add optional extra BUILD_VERSION info
2941 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2942 to override the lustre version.
2945 Frequency : 2.6.18 servers only
2947 Description: ll_kern_mount() doesn't release the module reference
2948 Details : The ldiskfs module reference count never drops down to 0
2949 because ll_kern_mount() doesn't release the module reference.
2954 Description: server LBUG when using old ost_num_threads parameter
2955 Details : Accept the old ost_num_threads parameter but warn that it
2956 is deprecated, and fix an off-by-one error that caused an LBUG.
2961 Description: Transient SCSI error results in persistent IO issue
2962 Details : iobuf->dr_error is not reinitialized to 0 between two
2966 Frequency : sometimes when underlying device returns I/O errors
2968 Description: OSTs not going read-only during write failures
2969 Details : OSTs are not remounted read-only when the journal commit threads
2970 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2971 instead of the ext3 wrappers.
2975 Description: poor connect scaling with increasing client count
2976 Details : Don't run filter_grant_sanity_check for more than 100 exports
2977 to improve scaling for large numbers of clients.
2980 Frequency : SLES10 only
2982 Description: sanity-quota.sh quotacheck failed: rc = -22
2983 Details : Quotas cannot be enabled on SLES10.
2986 Frequency : liblustre clients only
2988 Description: getdirentries does not give error when run on compute nodes
2989 Details : getdirentries does not fail when the size specified as an argument
2990 is too small to contain at least one entry
2992 Severity : enhancement
2994 Description: Add LNET router traceability for debug purposes
2995 Details : If a checksum failure occurs with a router as part of the
2996 IO path, the NID of the last router that forwarded the bulk data
2997 is printed so it can be identified.
3002 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3003 Details : Due to a race condition, liblustre clients were occasionally
3004 evicted incorrectly.
3006 Severity : enhancement
3008 Description: lfs setstripe use optional parameters instead of postional
3011 Severity : enhancement
3013 Description: Nanosecond timestamp support for ldiskfs
3014 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3015 resolution timestamps. There is not yet support for this at
3016 the Lustre filesystem level.
3019 Frequency : during server recovery
3021 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3022 Details : request with rq_no_resend flag not awake l_wait_event if they get
3026 Frequency : nfs export on patchless client
3028 Description: connectathon hang when test nfs export over patchless client
3029 Details : Disconnected dentry cannot be found with lookup, so we do not need
3030 to unhash it or make it invalid
3033 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3034 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3035 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3040 Description: open req refcounting wrong on reconnect
3041 Details : If reconnect happened between getting open reply from server and
3042 call to mdc_set_replay_data in ll_file_open, we will schedule
3043 replay for unreferenced request that we are about to free.
3044 Subsequent close will crash in variety of ways.
3045 Check that request is still eligible for replay in
3046 mdc_set_replay_data().
3051 Description: disable writes to filesystem when reading health_check file
3052 Details : the default for reading the health_check proc file has changed
3053 to NOT do a journal transaction and write to disk, because this
3054 can cause reads of the /proc file to hang and block HA state
3055 checking on a healthy but otherwise heavily loaded system. It
3056 is possible to return to the previous behaviour during configure
3057 with --enable-health-write.
3059 Severity : enhancement
3061 Description: 64-bit inode version
3062 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3063 to the inode. This will be required for version-based recovery.
3068 Description: MDS fails to start if a duplicate client export is detected
3069 Details : in some rare cases it was possible for a client to connect to
3070 an MDS multiple times. Upon recovery the MDS would detect this
3071 and fail during startup. Handle this more gracefully.
3073 Severity : enhancement
3075 Description: Add -o localflock option to simulate old noflock
3077 Details : This will achieve local-only flock/fcntl locks
3083 Description: log_commit_thread vs filter_destroy race leads to crash
3084 Details : Take import reference before releasing llog record semaphore
3089 Description: Wrong request locking in request set processing
3090 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3091 imp_delayed_list, in this place should be used imp_lock.
3094 Frequency : when reconnection
3096 Description: Grant Leak when osc reconnect to OST
3097 Details : When osc reconnect ost, OST(filter) should check whether it
3098 should grant more space to client by comparing fed_grant and
3099 cl_avail_grant, and return the granted space to client instead
3100 of "new granted" space, because client will call osc_init_grant
3101 to update the client grant space info.
3104 Frequency : when client reconnect to OST
3106 Description: Grant Leak when osc do resend and replay bulk write
3107 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3108 bulk write request, because the grant info while be sync between
3109 OSC and OST when reconnect, and we should ignore the grant info
3110 these of resend/replay write req.
3115 Description: Grant space more than available left space sometimes.
3116 Details : When then OST is about to be full, if two bulk writing from
3117 different clients came to OST. Accord the avaliable space of the
3118 OST, the first req should be permitted, and the second one
3119 should be denied by ENOSPC. But if the seconde arrived before
3120 the first one is committed. The OST might wrongly permit second
3121 writing, which will cause grant space > available space.
3124 Frequency : when client is evicted
3126 Description: Grant might be wrongly erased when osc is evicted by OST
3127 Details : when the import is evicted by server, it will fork another
3128 thread ptlrpc_invalidate_import_thread to invalidate the
3129 import, where the grant will be set to 0. While the original
3130 thread will update the grant it got when connecting. So if
3131 the former happened latter, the grant will be wrongly errased
3132 because of this race.
3137 Description: Checking Stale with correct fid
3138 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3139 to check whether it is stale, because sometimes, we want the
3140 enqueue happened anyway, and op_data.fid2 will not be initialized.
3142 Severity : enhancement
3144 Description: update patchless client
3145 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3148 Frequency : only with 2.4 kernel
3150 Description: random memory corruption
3151 Details : size of struct ll_inode_info is to big for union inode.u and this
3152 can be cause of random memory corruption.
3157 Description: Memory leak in recovery
3158 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3159 It should also check obd_fail before fsfilt_start, otherwise if
3160 fsfilt_start return -EROFS,(failover mds during mds recovery).
3161 then the req will return with repmsg->transno = 0 and rc = EROFS.
3162 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3163 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3164 be freed no matter whether fsfilt_commit success or not.
3167 Frequency : only with huge count clients
3169 Description: Prevents from taking the superblock lock in llap_from_page for
3171 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3172 for llap_from_page call in ll_removepage prevents from taking the
3173 superblock lock for a soon died page.
3178 Description: Not check open intent error before release open handle
3179 Details : in some rare cases, the open intent error is not checked before
3180 release open handle, which may cause
3181 ASSERTION(open_req->rq_transno != 0), because it tries to release
3182 the failed open handle.
3187 Description: Set cat log bitmap only after create log success.
3188 Details : in some rare cases, the cat log bitmap is set too early. and it
3189 should be set only after create log success.
3193 Description: Accessing a block bevice can re-enable I/O when Lustre is
3194 tearing down a device.
3195 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3200 Description: service threads may hog cpus when there are a lot of requests
3201 Details : Insert cond_resched to give other threads a chance to use some CPU
3206 Description: the cat log was not initialized in recovery
3207 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3208 unlink log on mds will not be initialized until mds post
3209 recovery. And also in mds post recovery, the unlink log will
3210 initialization will be done asynchronausly, so there will be race
3211 between add unlink log and unlink log initialization.
3215 Description: brw_stats were being printed incorrectly
3216 Details : brw_stats were being printed as log2 but all of them were not
3217 recorded as log2. Also remove some code duplication arising from
3218 filter_tally_{read,write}.
3222 Frequency : rare, only in recovery.
3223 Description: ASSERTION(req->rq_type != LI_POISON) failed
3224 Details : imp_lock should be held while iterating over imp_sending_list for
3225 prevent destroy request after get timeout in ptlrpc_queue_wait.
3229 Description: replay-single.sh test 52 fails
3230 Details : A lock's skiplist need to be cleanup when it being unlinked
3231 from its resource list.
3235 Description: Short directio read returns full requested size rather than
3237 Details : Direct I/O operations should return actual amount of bytes
3238 transferred rather than requested size.
3240 Severity : enhancement
3242 Description: metadata RPC reduction (e.g. for rm performance)
3243 Details : decrease the amount of synchronous RPC between clients and servers
3244 by canceling conflicing lock before the operation on the client
3245 and packing thier handles into the main operation RPC to server.
3247 Severity : enhancement
3249 Description: Async OSC create to avoid the blocking unnecessarily.
3250 Details : If an OST has no remaining object, system will block on the
3251 creation when it needs to create a new object on this OST. Now,
3252 ways use pre-created objects when available, instead of blocking on
3253 an empty osc while others are not empty. If we must block, we
3254 block for the shortest possible period of time.
3258 Description: improve handling recoverable errors
3259 Details : if request processig with error which can be recoverable on server
3260 request should be resend, otherwise page released from cache and
3263 Severity : enhancement
3265 Description: refine locking for avoid write wrong info into lov_objid file
3266 Details : fix possible races with add new target and write/update data in
3269 --------------------------------------------------------------------------------
3271 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3276 Frequency : on some architectures
3278 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3279 Details : Uninitialized flags sometimes cause configuration commands to
3283 Frequency : patchless clients only
3285 Description: missing __iget() symbol export
3286 Details : The __iget() symbol export is missing. To avoid the need for
3287 this on patchless clients the deathrow inode reaper is turned
3288 off, and we depend on the VM to clean up old inodes. This
3289 dependency was during via the fix for bug 12181.
3291 --------------------------------------------------------------------------------
3293 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3295 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3296 INTEROPERATE with older versions automatically. In many cases a
3297 special upgrade step is needed. Please read the
3298 user documentation before upgrading any part of a live system.
3299 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3300 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3301 * WARNING: Lustre configuration and startup changes are required with
3302 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3306 Severity : enhancement
3308 Description: MountConf
3309 Details : Lustre configuration is now managed via mkfs and mount
3310 commands instead of lmc and lconf. New obd types (MGS, MGC)
3311 are added for dynamic configuration management. See
3312 https://mail.clusterfs.com/wikis/lustre/MountConf for
3315 Severity : enhancement
3317 Description: dynamic OST addition
3318 Details : OSTs can now be added to a live filesystem
3320 Severity : enhancement
3322 Description: startup order invariance
3323 Details : MDTs and OSTs can be started in any order. Clients only
3324 require the MDT to complete startup.
3326 Severity : enhancement
3328 Description: parallel, asynchronous orphan cleanup
3329 Details : orphan cleanup is now performed in separate threads for each
3330 OST, allowing parallel non-blocking operation.
3332 Severity : enhancement
3334 Description: optimized stripe assignment
3335 Details : stripe assignments are now made based on ost space available,
3336 ost previous usage, and OSS previous usage, in order to try
3337 to optimize storage space and networking resources.
3339 Severity : enhancement
3341 Description: Permanently set tunables
3342 Details : All writable /proc/fs/lustre tunables can now be permanently
3343 set on a per-server basis, at mkfs time or on a live system.
3345 Severity : enhancement
3347 Description: Lustre message v2
3348 Details : Add lustre message format v2.
3350 Severity : enhancement
3352 Description: client OST exclusion list
3353 Details : Clients can be started with a list of OSTs that should be
3354 declared "inactive" for known non-responsive OSTs.
3358 Description: SPEC SFS validation failure on NFS v2 over lustre.
3359 Details : Changes the blocksize for regular files to be 2x RPC size,
3360 and not depend on stripe size.
3362 Severity : enhancement
3364 Description: Multiple MD RPCs in flight.
3365 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3366 To avoid overly-overloading MDS, introduce a limit on number of
3367 MDS RPCs in flight for a single client and add /proc controls
3368 to adjust this limit.
3370 Severity : enhancement
3372 Description: client read/write statistics
3373 Details : Add client read/write call usage stats for performance
3374 analysis of user processes.
3375 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3376 file access. extents_stats shows chunk size distribution.
3377 extents_stats_per_process show chunk size distribution per
3380 Severity : enhancement
3382 Description: mds statistics
3383 Details : Add detailed mds operations statistics in
3384 /proc/fs/lustre/mds/*/stats.
3388 Description: Failure of copying files with lustre special EAs.
3389 Details : Client side always return success for setxattr call for lustre
3390 special xattr (currently only "trusted.lov").
3395 Description: Refcount LNET uuids
3396 Details : The global LNET uuid list grew linearly with every startup;
3397 refcount repeated list entries instead of always adding to
3400 Severity : enhancement
3402 Description: Dynamic service threads
3403 Details : Within a small range, start extra service threads
3404 automatically when the request queue builds up.
3407 Frequency : mixed-endian client/server environments
3409 Description: mixed-endian crashes
3410 Details : The new msg_v2 system had some failures in mixed-endian
3413 Severity : enhancement
3415 Description: Easy OST removal
3416 Details : OSTs can be permanently deactivated with e.g. 'lctl
3417 conf_param lustre-OST0001.osc.active=0'
3419 Severity : enhancement
3421 Description: MGS proc entries
3422 Details : Added basic proc entries for the MGS showing what filesystems
3425 Severity : enhancement
3427 Description: provide MGS failover
3428 Details : Added config lock reacquisition after MGS server failover.
3430 Severity : enhancement
3432 Description: add Linux 2.4 support
3433 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3437 Description: a large application tries to do I/O to the same resource and dies
3438 in the middle of it.
3439 Details : Check the req->rq_arrival time after the call to
3440 ost_brw_lock_get(), but before we do anything about
3441 processing it & sending the BULK transfer request. This
3442 should help move old stale pending locks off the queue as
3443 quickly as obd_timeout.
3446 Frequency : when an incorrect nid is specified during startup
3448 Description: ptlrpc connect to non-existant node causes kernel crash
3449 Details : LNET can't be re-entered from an event callback, which
3450 happened when we expire a message after the export has been
3451 cleaned up. Instead, hand the zombie cleanup off to another
3454 Severity : enhancement
3456 Description: plain/inodebits lock performance improvement
3457 Details : Grouping plain/inodebits in granted list by their request modes
3458 and bits policy, thus improving the performance of search through
3462 Frequency : only if OST filesystem is corrupted
3464 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3465 Details : for a short time RPCs with bulk IO are in the replay list,
3466 but replay of bulk IOs is unimplemented. If the OST filesystem
3467 is corrupted due to disk cache incoherency and then replay is
3468 started it is possible to trip an assertion. Avoid putting
3469 committed RPCs into the replay list at all to avoid this issue.
3472 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3474 Description: System hang on startup
3475 Details : This bug allowed the liblustre (e.g. catamount) client to
3476 return to the app before handling all startup RPCs. This
3477 could leave the node unresponsive to lustre network traffic
3478 and manifested as a server ptllnd timeout.
3480 Severity : enhancement
3482 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3483 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3484 boolean to control whether to print peer debug info when a
3485 client's RPC times out.
3488 Frequency : only for kernels with patches from Lustre below 1.4.3
3490 Description: Remove old rdonly API
3491 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3494 Frequency : only for devices with external journals
3496 Description: Set external device read-only also
3497 Details : During a commanded failover stop, we set the disk device
3498 read-only while the server shuts down. We now also set any
3499 external journal device read-only at the same time.
3502 Frequency : when upgrading from 1.4 while trying to change parameters
3504 Description: The wrong (new) MDC name was used when setting parameters for
3505 upgraded MDT's. Also allows changing of OSC (and MDC)
3506 parameters if --writeconf is specified at tunefs upgrade time.
3509 Frequency : when setting specific ost indicies
3511 Description: QOS code breaks on skipped indicies
3512 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3513 created with --index need not be sequential.
3515 Severity : enhancement
3517 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3518 Details : The uninit_groups feature works in conjunction with the kernel
3519 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3520 up the pass1 processing of e2fsck. This is a read-only feature
3521 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3522 mount filesystems that have had this feature enabled.
3524 Severity : enhancement
3526 Description: Improve multi-block allocation algorithm to avoid fragmentation
3527 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3528 allocation locality and avoid filesystem fragmentation.
3530 ------------------------------------------------------------------------------
3532 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3534 * Support for kernels:
3535 2.4.21-47.0.1.EL (RHEL 3)
3536 2.6.5-7.283 (SLES 9)
3537 2.6.9-42.0.10.EL (RHEL 4)
3538 2.6.12.6 vanilla (kernel.org)
3539 2.6.16.27-0.9 (SLES 10)
3540 * Recommended e2fsprogs version: 1.39.cfs5
3542 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3546 Frequency : occasional, depends on client load and configuration
3547 Bugzilla : 12181, 12203
3548 Description: data loss for recently-modified files
3550 Details : In some cases it is possible that recently written or created
3551 files may not be written to disk in a timely manner (this should
3552 normally be within 30s unless client IO load is very high).
3553 The problem appears as zero-length files or files that are a
3554 multiple of 1MB in size after a client crash or client eviction
3555 that are missing data at the end of the file.
3557 This problem is more likely to be hit on clients where files are
3558 repeatedly created and unlinked in the same directory, clients
3559 have a large amount of RAM, have many CPUs, the filesystem has
3560 many OSTs, the clients are rebooted frequently, and/or the files
3561 are not accessed by other nodes after being written.
3563 The presence of the problem can be detected by looking at
3564 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3565 smaller than the second (nr_unused) then dirty files will not
3566 be flushed automatically to disk. "sync; sleep 10" should be
3567 run several times on the node before unmounting it to update
3568 Lustre (this is also safe to run on nodes without this problem).
3570 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3571 that can cause this same problem, so customers using that kernel
3572 also need to update the kernel in addition to Lustre. In order
3573 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3575 It is normal that files written just before a client crash (less
3576 than 30s) may not yet have been flushed to disk, even for local
3580 Frequency : frequent on thin XT3 nodes
3582 Description: UUID collision on thin XT3 Linux nodes
3583 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3584 randomly, since we relied on an insufficiently-seeded PRNG.
3589 Description: OSS hangs after "All ost request buffers busy"
3590 Details : A deadlock between quota and journal operations caused OSS
3591 hangs after printing "All ost request buffers busy."
3594 Frequency : always on liblustre builds
3596 Description: Cleanup compiler warnings on liblustre
3599 Frequency : always on liblustre builds on XT3
3601 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3605 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3606 > 32000 subdirectories
3607 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3608 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3609 subdirectory count crosses 32000. This will aid e2fsck to
3610 correctly handle more than 32000 subdirectories.
3613 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3615 Description: System hang on startup
3616 Details : This bug allowed the liblustre (e.g. catamount) client to
3617 return to the app before handling all startup RPCs. This
3618 could leave the node unresponsive to lustre network traffic
3619 and manifested as a server ptllnd timeout.
3621 Severity : enhancement
3623 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3624 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3625 boolean to control whether to print peer debug info when a
3626 client's RPC times out.
3631 Description: make O_SYNC working on 2.6 kernels
3632 Details : 2.6 kernels use different method for mark pages for write,
3633 so need add a code to lustre for O_SYNC work.
3638 Description: Failure to close file and release space on NFS
3639 Details : Put inode details into lock acquired in ll_intent_file_open.
3640 Use mdc_intent_lock in ll_intent_open to properly
3641 detect all kind of errors unhandled by mdc_enqueue.
3646 Description: proc file read during shutdown sometimes raced obd removal,
3648 Details : Add lock to prevent obd access after proc file removal.
3651 Frequency : Only for files larger than 4GB on 32-bit clients.
3653 Description: improperly doing page alignment of locks
3654 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3655 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3660 Description: RPCs being resent when they shouldn't be
3661 Details : Some RPCs that should not be resent are being resent. This
3662 can cause inconsistencies in the RPC state machine. Do not
3663 resend such requests.
3666 Frequency : rare, only with NFS export
3668 Description: Crash on NFS re-export node
3669 Details : under very unusual load conditions an assertion is hit in
3670 ll_intent_file_open()
3673 Frequency : only if OST filesystem is corrupted
3675 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3676 Details : for a short time RPCs with bulk IO are in the replay list,
3677 but replay of bulk IOs is unimplemented. If the OST filesystem
3678 is corrupted due to disk cache incoherency and then replay is
3679 started it is possible to trip an assertion. Avoid putting
3680 committed RPCs into the replay list at all to avoid this issue.
3685 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3686 Details : Large single O_DIRECT read and write calls can fail to allocate
3687 a sufficiently large buffer to process the request. In case of
3688 allocation failure the allocation is retried with a smaller
3689 buffer and broken into smaller requests.
3691 Severity : enhancement
3693 Description: Add -o localflock option to simulate old noflock behaviour.
3694 Details : This will achieve local-only flock/fcntl locks coherentness.
3699 Description: versioning check is incomplete
3700 Details : Checking the version difference of client vs. server, report
3701 error if the gap is too big.
3706 Description: add support PG_writeback bit
3707 Details : add support for PG_writeback bit for Lustre, for more carefull
3708 work with page cache in 2.6 kernel. This also fix some deadlocks
3709 and remove hack for work O_SYNC with 2.6 kernel.
3711 Severity : enhancement
3713 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3714 Details : The uninit_groups feature works in conjunction with the kernel
3715 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3716 up the pass1 processing of e2fsck. This is a read-only feature
3717 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3718 mount filesystems that have had this feature enabled.
3720 Severity : enhancement
3722 Description: Improve multi-block allocation algorithm to avoid fragmentation
3723 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3724 allocation locality and avoid filesystem fragmentation.
3726 ------------------------------------------------------------------------------
3728 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3730 * Support for kernels:
3731 2.6.9-42.0.3.EL (RHEL 4)
3732 2.6.5-7.276 (SLES 9)
3733 2.4.21-47.0.1.EL (RHEL 3)
3734 2.6.12.6 vanilla (kernel.org)
3735 2.6.16.21-0.8 (SLES10)
3736 * Recommended e2fsprogs version: 1.39.cfs2-0
3738 * The backwards-compatible /proc/sys/portals symlink has been removed
3739 in this release. Before upgrading, please ensure that you change
3740 any configuration scripts or /etc/sysctl.conf files that access
3741 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3742 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3743 in advance of the upgrade on any system running Lustre 1.4.6 or
3744 newer, since /proc/sys/lnet was added in that version.
3745 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3749 Frequency : only when quota is used
3751 Description: avoid scanning export list for quota master
3752 Details : Change the algorithms to avoid scanning export list in order
3753 to improve the efficiency.
3756 Frequency : MDS failover only, very rarely
3758 Description: "went back in time" messages on mds failover
3759 Details : The greatest transno may be lost when the current operation
3760 finishes with an error (transno==0) and the client's last_rcvd
3761 record is over-written. Save the greatest transno in the
3762 mds_last_transno for this case.
3765 Frequency : always for specific kernels and striping counts
3767 Description: client may get "Matching packet too big" without ACL support
3768 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3769 when trying to access files in certain configurations. The
3770 clients should in fact be denied when mounting because they do
3771 not understand ACLs.
3774 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3776 Description: many clients connecting with IO in progress causes connect timeouts
3777 Details : Avoid synchronous journal commits to avoid delays caused by many
3778 clients connecting/disconnecting when bulk IO is in progress.
3779 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3780 OST_IO_PORTAL to avoid delays behind potentially many pending
3784 Frequency : occasionally with multiple writers to a single file
3786 Description: shared writes to file may result in wrong size reported by stat()
3787 Details : Allow growing of kms when extent lock is cancelled
3790 Frequency : always with random mmap IO to multi-striped file
3792 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3793 Details : Only if the hole is at the end of OST object so that kms is too
3794 small. Fix is to increase kms accordingly in ll_nopage.
3797 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3799 Description: writes to a missing object would leak memory on the OST
3800 Details : If there is an inconsistency between the MDS and OST filesystems,
3801 such that the MDS references an object that doesn't exist, writes
3802 to that object will leak memory due to incorrect cleanup in the
3803 error handling path, eventually running out of memory on the OST.
3808 Description: Creating too long symlink causes lustre errors
3809 Details : Check symlink and name lengths before sending requests to MDS.
3812 Frequency : only if flock is enabled (not on by default)
3814 Description: posix locks not released on fd closure on 2.6.9+
3815 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3816 kernels, this caused such locks not to be released on fd close and
3817 then assertions on fs unmount about still used locks.
3820 Frequency : MDS failover only, very rarely
3822 Description: clients may get ASSERTION(granted_lock != NULL)
3823 Details : When request was taking a long time, and a client was resending
3824 a getattr by name lock request. The were multiple lock requests
3825 with the same client lock handle and
3826 mds_getattr_name->fixup_handle_for_resent_request found one of the
3827 lock handles but later failed with ASSERTION(granted_lock != NULL).
3832 Description: handle->h_buffer_credits > 0, assertion failure
3833 Details : h_buffer_credits is zero after truncate, causing assertion
3834 failure. This patch extends the transaction or creates a new
3838 Frequency : NFS re-export or patchless client
3839 Bugzilla : 11179, 10796
3840 Description: Crash on NFS re-export node (__d_move)
3841 Details : We do not want to hash the dentry if we don't have a lock.
3842 But if this dentry is later used in d_move, we'd hit uninitialised
3843 list head d_hash, so we just do this to init d_hash field but
3844 leave dentry unhashed.
3847 Frequency : NFS re-export or patchless client
3849 Description: NFS exports has problem with symbolic link
3850 Details : lustre client didn't properly install dentry when re-exported
3851 to NFS or running patchless client.
3854 Frequency : NFS re-export or patchless client
3856 Description: Various nfs/patchless fixes.
3857 Details : fixes reuse disconected alias for lookup process - this fixes
3858 warning "find_exported_dentry: npd != pd",
3859 fix permission error with open files at nfs.
3860 fix apply umask when do revalidate.
3863 Frequency : occasional
3865 Description: Crash on NFS re-export node
3866 Details : calling clear_page() on the wrong pointer triggered oops in
3867 generic_mapping_read().
3870 Frequency : rarely, using O_DIRECT IO
3872 Description: unaligned directio crashes client with LASSERT
3873 Details : check for unaligned buffers before trying any requests.
3876 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3878 Description: stale data returned from RAID cache
3879 Details : If only a small amount of IO is done to the RAID device before
3880 reading it again it is possible to get stale data from the RAID
3881 cache instead of reading it from disk.
3884 Frequency : always for sles10 kernel
3886 Description: sles10 support
3887 Details : ll_follow_link: compile fixes and using of nd_set_link
3888 under newer kernels.
3891 Frequency : depends on arch, kernel and compiler version, always on sles10
3894 Description: recursive or deep enough symlinks cause stack overflow
3895 Details : getting rid of large stack-allocated variable in
3899 Frequency : depends on hardware
3901 Description: lustre write performance loss in the SLES10 kernel
3902 Details : the performance loss is caused by using of write barriers in the
3903 ext3 code. The SLES10 kernel turns barrier support on by
3904 default. The fix is to undo that change for ldiskfs.
3906 ------------------------------------------------------------------------------
3908 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3910 * Support for kernels:
3911 2.6.9-42.0.3EL (RHEL 4)
3912 2.6.5-7.276 (SLES 9)
3913 2.4.21-47.0.1.EL (RHEL 3)
3914 2.6.12.6 vanilla (kernel.org)
3918 Frequency : quota enabled and large files being deleted
3920 Description: releasing more than 4GB of quota at once hangs OST
3921 Details : If a user deletes more than 4GB of files on a single OST it
3922 will cause the OST to spin in an infinite loop. Release
3923 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3928 Description: statfs data retrieved from /proc may be stale or zero
3929 Details : When reading per-device statfs data from /proc, in the
3930 {kbytes,files}_{total,free,avail} files, it may appear
3931 as zero or be out of date.
3934 Frequency : systems with MD RAID1 external journal devices
3936 Description: lconf's call to blkid is confused by RAID1 journal devices
3937 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3938 of returning all block devices that match the journal UUID.
3941 Frequency : always, for aggregate stripe size over 4GB
3943 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3944 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3945 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3948 Frequency : always if "lfs find" used on a local file/directory
3950 Description: "lfs find" segfaults if used on a local file/directory
3951 Details : The case where a directory component was not specified wasn't
3952 handled correctly. Handle this properly.
3955 Frequency : always on ppc64
3957 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3958 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3959 perform endianness conversion.
3962 Frequency : rarely (truncate to non-zero file size after write under load)
3963 Bugzilla : 10730, 10687
3964 Description: Files padded with zeros to next 4K multiple
3965 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3966 it is possible that files that are truncated to a non-zero size
3967 immediately after being written are filled with zero bytes beyond
3968 the truncated size. No file data is lost.
3970 Severity : enhancement
3972 Description: Allow recovery/failover for liblustre clients.
3973 Details : liblustre clients were unaware of failover configurations until
3976 Severity : enhancement
3978 Description: user file locks should fail when not mounting with flock option
3979 Details : Set up an error-returning stub in ll_file_operations.lock field
3980 to prevent incorrect behaviour when client is mounted without
3981 flock option. Also, set up properly f_op->flock field for
3985 Frequency : always on ia64
3987 Description: "lfs df" loops on printing out MDS statfs information
3988 Details : The obd_ioctl_data was not initialized and in some systems
3989 this caused a failure during the ioctl that did not return
3990 an error. Initialize the struct and return an error on failure.
3993 Frequency : SLES 9 only
3995 Description: Error of copying files with lustre special EAs as root
3996 Details : Client side always return success for setxattr call for lustre
3997 special xattr (currently only "trusted.lov").
4000 Frequency : rarely on clusters with both ia64+i386 clients
4002 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4003 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4004 clients are doing concurrent writes to the same file it is
4005 possible that the ia64 clients may LASSERT because the OST
4006 extent locks are not PAGE_SIZE aligned. Ensure that grown
4007 locks are always aligned on the request boundary.
4010 Frequency : specific use, occasional
4012 Description: Overwriting in use executable truncates on-disk binary image
4013 Details : If one node attempts to overwrite an executable in use by
4014 another node, we now correctly return ETXTBSY instead of
4015 truncating the file.
4017 Severity : enhancement
4019 Description: Async OSC create to avoid the blocking unnecessarily.
4020 Details : If an OST has no remaining object, system will block on the creating
4021 when need to create a new object on this OST. Now, ways use
4022 pre-created objects when available, instead of blocking on an
4023 empty osc while others are not empty. If we must block, we block
4024 for the shortest possible period of time.
4029 Description: chmod on Lustre root is propagated to other clients
4030 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4031 invalid by the follow_mount time.
4034 Frequency : liblustre clients only
4036 Description: Race in 'instant cancel' lock handling could lead to such locks
4037 never to be granted in case of SMP MDS
4038 Details : Do not destroy not yet granted but cbpending locks in
4042 Frequency : replay/resend of open
4044 Description: non null lock assetion failure in mds_intent_policy
4045 Details : Trying to replay/resend lockless open requests resulted in
4046 mds_open() returning 0 with no lock. Now it sets a flag if
4047 a lock is going to be returned.
4049 Severity : enhancement
4051 Description: Checksum enhancements
4052 Details : New checksum enhancements allow for resending RPCs that failed
4055 Severity : enhancement
4057 Description: Tunables on number of dirty pages in cacche
4058 Details : Allow to set limit on number of dirty pages cached.
4063 Description: client crash on unmount - lock still has references
4064 Details : In some error handling cases it was possible to leak a lock
4065 reference on a client while accessing a file. This was not
4066 harmful to the client during operation, but would cause the
4067 client to crash when the filesystem is unmounted.
4070 Frequency : specific case, rare
4072 Description: ETXTBSY on mds though file not in use by client
4073 Details : ETXTBSY is no longer incorrectly returned when attempting to
4074 chmod or chown a directory that the user previously tried to
4075 execute or a currently-executing binary.
4078 Frequency : extremely rare except on liblustre-based clients
4080 Description: Lustre space not freed when files are deleted
4081 Details : Clean up open-unlinked files after client eviction. Previously
4082 the unlink was skipped and the files remained as orphans.
4087 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4088 Details : In some cases it was possible to send a blocking callback to
4089 a client doing a glimpse, even though that client didn't get
4090 a lock granted. When the glimpse lock is cancelled on the OST
4091 the freed lock is left on the waiting list and corrupted the list.
4094 Frequency : all core dumps
4096 Description: Broke core dumps to lustre
4097 Details : Negative dentry may be unhashed if parent does not have UPDATE
4098 lock, but some callers, e.g. do_coredump, expect dentry to be
4099 hashed after successful create, hash it in ll_create_it.
4101 ------------------------------------------------------------------------------
4103 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4105 * Support for kernels:
4106 2.6.9-42.0.2.EL (RHEL 4)
4107 2.6.5-7.276 (SLES 9)
4108 2.4.21-40.EL (RHEL 3)
4109 2.6.12.6 vanilla (kernel.org)
4113 Frequency : always on RHEL 3
4115 Description: Number of open files grows over time
4116 Details : The number of open files grows over time, whether or not
4117 Lustre is started. This was due to a filp leak introduced
4118 by one of our kernel patches.
4120 ------------------------------------------------------------------------------
4122 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4124 * Support for kernels:
4125 2.6.9-42.EL (RHEL 4)
4126 2.6.5-7.276 (SLES 9)
4127 2.4.21-40.EL (RHEL 3)
4128 2.6.12.6 vanilla (kernel.org)
4133 Bugzilla : 5719, 9635, 9792, 9684
4134 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4135 Details : If a server is under heavy load and cannot reply to new
4136 connection requests before the client resends the (re)connect,
4137 the connection handling code can behave badly if two service
4138 threads are concurrently handing separate (re)connections from
4139 the same client. Add better locking to the connection handling
4140 code, and ensure that only a single connection will be processed
4141 for a given client UUID, even if the lock is dropped.
4143 Severity : enhancement
4145 Description: add TCP zero-copy support to kernel
4146 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4147 sends if the hardware supports scatter-gather and checksumming.
4148 This allows socklnd to do client-write and server-read more
4149 efficiently and reduce CPU utilization from skbuf copying.
4152 Frequency : only if NFS exporting from client
4154 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4155 Details : If setting attributes on a file created under NFS that had
4156 never been opened it would be possible to oops the client
4157 if the file had no objects.
4160 Frequency : always for liblustre
4162 Description: liblustre client does MDS+OSTs setattr RPC for each write
4163 Details : When doing a write from a liblustre client, the client
4164 incorrectly issued an RPC to the MDS and each OST the file was
4165 striped over in order to update the timestamps. When writing
4166 with small chunks and many clients this could overwhelm the MDS
4167 with RPCs. In all cases it would slow down the write because
4168 these RPCs are unnecessary.
4170 Severity : enhancement
4172 Description: allow number of MDS service threads to be changed at module load
4173 Details : It is now possible to change the number of MDS service threads
4174 running. Adding "options mds mds_num_threads={N}" to the MDS's
4175 /etc/modprobe.conf will set the number of threads for the next
4176 time Lustre is restarted (assuming the "mds" module is also
4177 reloaded at that time). The default number of threads will
4178 stay the same, 32 for most systems.
4183 Description: OST crash if filesystem is unformatted or corrupt
4184 Details : If an OST is started on a device that has never been formatted
4185 or if the filesystem is corrupt and cannot even mount then the
4186 error handling cleanup routines would dereference a NULL pointer.
4191 Description: NULL pointer deref in llap_from_page.
4192 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4193 incorrect) mapping if the page was truncated/reclaimed while it was
4194 searched for. Check for this condition and skip such pages when
4195 doing readahead. Introduce extra check to llap_from_page() to
4196 verify page->mapping->host is non-NULL (so page is not anonymous).
4199 Frequency : Sometimes when using sys_sendfile
4201 Description: "page not covered by a lock" warnings from ll_readpage
4202 Details : sendfile called ll_readpage without right page locks present.
4203 Now we introduced ll_file_sendfile that does necessary locking
4204 around call to generic_file_sendfile() much like we do in
4208 Frequency : with certain MDS communication failures at client mount time
4210 Description: NULL pointer deref after failed client mount
4211 Details : a client connection request may delayed by the network layer
4212 and not be sent until after the PTLRPC layer has timed out the
4213 request. If the client fails the mount immediately it will try
4214 to clean up before the network times out the request. Add a
4215 reference from the request import to the obd device and delay
4216 the cleanup until the network drops the request.
4219 Frequency : occasionally during client (re)connect
4221 Description: assertion failure during client (re)connect
4222 Details : processing a client connection request may be delayed by the
4223 client or server longer than the client connect timeout. This
4224 causes the client to resend the connection request. If the
4225 original connection request is replied in this interval, the
4226 client may trip an assertion failure in ptlrpc_connect_interpret()
4227 which thought it would be the only running connect process.
4230 Frequency : only with obd_echo servers and clients that are rebooted
4232 Description: kernel BUG accessing uninitialized data structure
4233 Details : When running an obd_echo server it did not start the ping_evictor
4234 thread, and when a client was evicted an uninitialized data
4235 structure was accessed. Start the ping_evictor in the RPC
4236 service startup instead of the OBD startup.
4238 Severity : enhancement
4239 Bugzilla : 10193 (patchless)
4240 Description: Remove dependency on various unexported kernel interfaces.
4241 Details : No longer need reparent_to_init, exit_mm, exit_files,
4242 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4245 Frequency : rare (only users of deprecated and unsupported LDAP config)
4247 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4248 Details : LDAP apparently contains 'lustreName' attributes instead of
4249 'name'. A simple remapping of the name is sufficient.
4252 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4254 Description: waiting_locks_callback trips kernel BUG if client is evicted
4255 Details : Running with the dump_on_timeout debug flag turned on makes
4256 it possible that the waiting_locks_callback() can try to dump
4257 the Lustre kernel debug logs from an interrupt handler. Defer
4258 this log dumping to the expired_lock_main() thread.
4260 Severity : enhancement
4262 Description: Support NFS exporting on 2.6 kernels.
4263 Details : Implement non-rawops metadata methods for NFS server to use without
4264 changing NFS server code.
4267 Frequency : very rare (synthetic metadata workload only)
4269 Description: two racing renames might cause an MDS thread to deadlock
4270 Details : Running the "racer" program may cause one MDS thread to rename
4271 a file from being the source of a rename to being the target of
4272 a rename at exactly the same time that another thread is doing
4273 so, and the second thread has already enqueued these locks after
4274 doing a lookup of the target and is trying to relock them in
4275 order. Ensure that we don't try to re-lock the same resource.
4278 Frequency : only very large systems with liblustre clients
4280 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4281 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4282 all OSTs in parallel. This allows the eviction of stale liblustre
4283 clients to proceed much faster than if they were done in series,
4284 and also offers similar improvements for other set_info RPCs.
4289 Description: excessive CPU usage during initial read phase on client
4290 Details : During the initial read phase on a client, it would agressively
4291 retry readahead on the file, consuming too much CPU and impacting
4292 performance (since 1.4.5.8). Improve the readahead algorithm
4293 to avoid this, and also improve some other common cases (read
4294 of small files in particular, where "small" is files smaller than
4295 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4300 Description: MDS crash when receiving packet with unknown intent.
4301 Details : Do not LBUG in unknown intent case, just return -EFAULT
4303 Severity : enhancement
4304 Bugzilla : 9293, 9385
4305 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4306 Details : Do not serialize getattr (non-intent version) and statfs.
4309 Frequency : occasional, when OST network is overloaded/intermittent
4311 Description: client evicted by OST after bulk IO timeout
4312 Details : If a client sends a bulk IO request (read or write) the OST
4313 may evict the client if it is unresposive to its data GET/PUT
4314 request. This is incorrect if the network is overloaded (takes
4315 too long to transfer the RPC data) or dropped the OST GET/PUT
4316 request. There is no need to evict the client at all, since
4317 the pinger and/or lock callbacks will handle this, and the
4318 client can restart the bulk request.
4321 Frequency : Always when mmapping file with no objects
4323 Description: client crashes when mmapping file with no objects
4324 Details : Check that we actually have objects in a file before doing any
4325 operations on objects in ll_vm_open, ll_vm_close and
4331 Description: Request leak when working with deleted CWD
4332 Details : Introduce advanced request refcount tracking for requests
4333 referenced from lustre intent.
4335 Severity : Enhancement
4337 Description: Cache open file handles on client.
4338 Details : MDS now will return special lock along with openhandle, if
4339 requested and client is allowed to hold openhandle, even if unused,
4340 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4341 closing files for every read/write openration.
4343 Severity : Enhancement
4345 Description: Cache open negative dentries on client when possible.
4346 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4347 negative dentries on lock revocation.
4352 Description: Remounting a client read-only wasn't possible with a zconf mount
4353 Details : It wasn't possible to remount a client read-only with llmount.
4355 Severity : enhancement
4356 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4357 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4358 MPICH is included to add Lustre-specific ADIO interfaces.
4359 This is based closely on the UFS ADIO layer and only differs
4360 in file creation, in order to allow the OST striping to be set.
4361 This is user-contributed code and not supported by CFS.
4366 Description: extended inode attributes (immutable, append-only) work improperly
4367 when 2.4 and 2.6 kernels are used on client/server or vice versa
4368 Details : Introduce kernel-independent values for these flags.
4370 Severity : enhancement
4373 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4374 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4375 granularity. Now, Fractional MB granularity is be supported,
4376 this is very useful for low memory system.
4378 Severity : enhancement
4380 Description: Getattr by fid
4381 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4382 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4383 have locks and avoids getattr rpc requests for it on every stat.
4386 Frequency : Always, for filesystems larger than 2TB
4388 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4389 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4390 filesystems larger than 2TB to be mounted with mballoc enabled.
4393 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4395 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4396 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4397 filesystems larger than 2TB is mounted then the kernel will
4398 silently corrupt the start of the filesystem. CONFIG_LBD is
4399 enabled for all CFS-supported kernels, but the possibility of
4400 this happening with a modified kernel config exists.
4402 Severity : enhancement
4404 Description: add client O_DIRECT support for 2.6 kernels
4405 Details : It is now possible to do O_DIRECT reads and writes to files
4406 in the Lustre client mountpoint on 2.6 kernel clients.
4408 Severity : enhancement
4410 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4411 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4412 parallel, not waiting for response from every OST before sending
4413 a rpc to the next OST.
4418 Description: setattr vs write race when updating file timestamps
4419 Details : Client processes that update a file timestamp into the past
4420 right after writing to the file (e.g. tar) it is possible that
4421 the updated file modification time can be reset to the current
4422 time due to a race between processing the setattr and write RPC.
4424 Severity : enhancement
4426 Description: Bring 'lfs find' closer in line with regular Linux find.
4427 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4428 -print0 options and obtains all the needed info through the lustre
4431 Severity : enhancement
4433 Description: support up to 1024 configured devices on one node
4434 Details : change obd_dev array from statically allocated to dynamically
4435 allocated structs as they are first used to reduce memory usage
4440 Description: Flush dirty partially truncated pages during truncate
4441 Details : Immediatelly flush partially truncated pages in filter_setattr,
4442 this way we completely avoid having any pages in page cache on OST
4443 and can retire ugly workarounds during writes to flush such pages.
4448 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4449 Details : protect inode from truncation within vfs_unlink() context
4450 just take a reference before calling vfs_unlink() and release it
4451 when parent's i_sem is free.
4456 Description: last_id value checked outside lock on OST caused LASSERT failure
4457 Details : If there were multiple MDS->OST object precreate requests in
4458 flight, it was possible that the OST's last object id was checked
4459 outside a lock and incorrectly tripped an assertion. Move checks
4460 inside locks, and discard old precreate requests.
4463 Frequency : always, if extents are used on OSTs
4465 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4466 Details : OSTs using the extents format would not zero the high 16 bits of
4467 the index physical block number. This is not a problem for any
4468 OST filesystems smaller than 16TB, and no kernels support ext3
4469 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4470 new/modified files) and can be fixed for existing filesystems
4471 with e2fsprogs-1.39-cfs1.
4476 Description: import connection selection may be incorrect if timer wraps
4477 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4478 import connections to be ignored if the 32-bit jiffies counter
4479 wraps. Use a 64-bit jiffies counter.
4482 Frequency : very large clusters immediately after boot
4484 Description: LNET request buffers exhausted under heavy short-term load
4485 Details : If a large number of client requests are generated on a service
4486 that has previously never seen so many requests it is possible
4487 that the request buffer growth cannot keep up with the spike in
4488 demand. Instead of dropping incoming requests, they are held in
4489 the LND until the RPC service can accept more requests.
4492 Frequency : Sometimes during replay
4494 Description: Assertion failure in ll_local_open after replay.
4495 Details : If replay happened on an open request reply before we were able
4496 to set replay handler, reply will become not swabbed tripping the
4497 assertion in ll_local_open. Now we set the handler right after
4498 recognising of open request
4501 Frequency : very rare
4503 Description: kernel reports "badness in vsnprintf"
4504 Details : Reading from the "recovery_status" /proc file in small chunks
4505 may cause a negative length in lprocfs_obd_rd_recovery_status()
4506 call to vsnprintf() (which is otherwise harmless). Exit early
4507 if there is no more space in the output buffer.
4509 Severity : enhancement
4511 Description: clear OBD RPC statistics by writing to them
4512 Details : It is now possible to clear the OBD RPC statistics by writing
4513 to the "stats" file.
4518 Description: Client mtime is not the same on different clients after utimes
4519 Details : In some cases, the client was using the utimes() syscall on
4520 a file cached on another node. The clients now validate the
4521 ctime from the MDS + OSTs to determine which one is right.
4526 Description: Inability to activate failout mode
4527 Details : lconf script incorrectly assumed that in pythong string's numeric
4528 value is used in comparisons.
4531 Frequency : always with multiple stripes per file
4533 Description: Inefficient object allocation for mutli-stripe files
4534 Details : When selecting which OSTs to stripe files over, for files with
4535 a stripe count that divides evenly into the number of OSTs,
4536 the MDS is always picking the same starting OST for each file.
4537 Return the OST selection heuristic to the original design.
4542 Description: mount failures may take full timeout to return an error
4543 Details : Under some heavy load conditions it is possible that a
4544 failed mount can wait for the full obd_timeout interval,
4545 possibly several minutes, before reporting an error.
4546 Instead return an error as soon as the status is known.
4548 Frequency : quota enabled and large files being deleted
4550 Description: releasing more than 4GB of quota at once hangs OST
4551 Details : If a user deletes more than 4GB of files on a single OST it
4552 will cause the OST to spin in an infinite loop. Release
4553 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4558 Description: statfs data retrieved from /proc may be stale or zero
4559 Details : When reading per-device statfs data from /proc, in the
4560 {kbytes,files}_{total,free,avail} files, it may appear
4561 as zero or be out of date.
4564 Frequency : systems with MD RAID1 external journal devices
4566 Description: lconf's call to blkid is confused by RAID1 journal devices
4567 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4568 of returning all block devices that match the journal UUID.
4571 Frequency : always, for aggregate stripe size over 4GB
4573 Description: assertion fails when trying to use 4GB stripe size
4574 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4575 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4578 Frequency : always on ppc64
4580 Description: the first write on an ext3 filesystem with mballoc got stuck
4581 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4582 perform endianness conversion.
4584 ------------------------------------------------------------------------------
4586 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4588 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4589 INTEROPERATE with older versions automatically. Please read the
4590 user documentation before upgrading any part of a live system.
4591 * WARNING: Lustre networking configuration changes are required with
4592 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4595 * Support for kernels:
4596 2.6.9-22.0.2.EL (RHEL 4)
4597 2.6.5-7.244 (SLES 9)
4598 2.6.12.6 vanilla (kernel.org)
4601 Severity : enhancement
4602 Bugzilla : 7981/8208
4603 Description: Introduced Lustre Networking (LNET)
4604 Details : LNET is new networking infrastructure for Lustre, it includes
4605 a reorganized network configuration mode (see the user
4606 documentation for full details) as well as support for routing
4607 between different network fabrics. Lustre Networking Devices
4608 (LNDs) for the supported network fabrics have also been
4609 created for this new infrastructure.
4611 Severity : enhancement
4612 Description: Introduced Access control lists
4613 Details : clients can set ACLs on files and directories in order to have
4614 more fine-grained permissions than the standard Unix UGO+RWX.
4615 The MDS must be started with the "-o acl" mount option.
4617 Severity : enhancement
4618 Description: Introduced filesystem quotas
4619 Details : Administrators may now establish per-user quotas on the
4622 Severity : enhancement
4624 Description: Configuration change for the XT3
4625 The PTLLND is now used to run Lustre over Portals on the XT3
4626 The configure option(s) --with-cray-portals are no longer used.
4627 Rather --with-portals=<path-to-portals-includes> is used to
4628 enable building on the XT3. In addition to enable XT3 specific
4629 features the option --enable-cray-xt3 must be used.
4634 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4635 Details : A race between allocating a new llcd and re-getting the llcd_lock
4636 allowed another thread to grab newly-allocated llcd.
4638 Severity : enhancement
4640 Description: 2.6 OST async journal commit and locking fix to improve performance
4641 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4642 2.6 kernels are now async as they already were in 2.4 kernels so
4643 that they can commit concurrently with the network bulk transfer.
4644 For block-allocated files the filter allocation semaphore is held
4645 to avoid filesystem fragmentation during allocation. BKL lock
4646 removed for 2.6 xattr operations where it is no longer needed.
4651 Description: lconf incorrectly determined whether two IP networks could talk
4652 Details : In some more complicated routing and multiple-network
4653 configurations, lconf will avoid trying to make a network
4654 connection to a disjoint part of the IP space. It was doing the
4655 math incorrectly for one set of cases.
4660 Description: Fix for potential infinite loop processing records in an llog.
4661 Details : If an llog record is corrupted/zeroed, it is possible to loop
4662 forever in llog_process(). Validate the llog record length
4663 and skip the remainder of the block on error.
4666 Frequency : occasional (liblustre only)
4668 Description: liblustre could not open files whose last component is a symlink
4669 Details : sysio_path_walk() would incorrectly pass the open intent to
4670 intermediate path components.
4673 Frequency : rare (liblustre only with non-standard tuning)
4674 Bugzilla : 7201 (7350)
4675 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4676 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4677 after it has been released, only do this for referenced locks
4679 Severity : enhancement
4681 Description: specifying an (invalid) directory default stripe_size of -1
4682 would reset the directory default striping
4683 Details : stripe_size -1 was used internally to signal directory stripe
4684 removal, now use "all default" to signal dir stripe removal
4685 as a directory striping of "all default" is not useful
4688 Frequency : common for large clusters running liblustre clients
4690 Description: doing an ls when liblustre clients are running is slow
4691 Details : sending a glimpse AST to a liblustre client waits for every AST
4692 to time out, as liblustre clients will not respond. Since they
4693 cannot cache data we refresh the OST lock LVB from disk instead.
4695 Severity : enhancement
4697 Description: doing an ls at the same time as file IO can be slow
4698 Details : enqueue and other "small" requests can be blocked behind many
4699 large IO requests. Create a new OST IO portal for non-IO
4700 requests so they can be processed faster.
4703 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4705 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4706 created file with mode 000
4707 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4708 allow an ftruncate() as a non-root user to a file with mode 000.
4709 Lustre now respects this flag to disable mode checks when
4710 truncating a file owned by the user
4713 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4716 Description: Revoking locks from clients that went dead or catatonic might take
4718 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4719 cancellation of such locks instant on servers without waiting for
4720 any reply from clients. Clients drops these locks when cancel
4721 notification from server is received without replying.
4724 Frequency : liblustre-only, when liblustre client dies or becomes busy
4726 Description: Doing ls on Linux clients can take a long time with active
4728 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4729 granting such locks to it in the first place if possible. Locks
4730 are taken by proxy on the OST during the read or write and
4731 dropped immediately afterward. Add connect flags handling, do
4732 not grant locks to liblustre clients for glimpse ASTs.
4734 Severity : enhancement
4736 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4737 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4738 sequential reads and behaves badly with applications doing random
4739 reads. Improve it by reading ahead at least read region, and
4740 avoiding excessive large RPC for small reads.
4742 Severity : enhancement
4744 Description: Creating more than 1000 files for a single job may cause a load
4745 imbalance on the OSTs if there are also a large number of OSTs.
4746 Details : qos_prep_create() uses an OST index reseed value that is an
4747 even multiple of the number of available OSTs so that if the
4748 reseed happens in the middle of the object allocation it will
4749 still utilize the OSTs as uniformly as possible.
4754 Description: OST or MDS may oops in ping_evictor_main()
4755 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4756 but doesn't restart at beginning of obd_exports_timed list
4759 Severity : enhancement
4761 Description: improve by-nid export eviction on the MDS and OST
4762 Details : allow multiple exports with the same NID to be evicted at one
4763 time without re-searching the exports list.
4766 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4768 Description: MDS may oops in groups_free()
4769 Details : in rare race conditions a newly allocated group_info struct is
4770 freed again, and this can be NULL. The 2.4 compatibility code
4771 for groups_free() checked for a NULL pointer, but 2.6 did not.
4774 Frequency : common for liblustre clients doing little filesystem IO
4775 Bugzilla : 9352, 7313
4776 Description: server may evict liblustre clients accessing contended locks
4777 Details : if a client is granted a lock or receives a completion AST
4778 with a blocking AST already set it would not reply to the AST
4779 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4782 Frequency : lfs setstripe, only systems with more than 160 OSTs
4784 Description: unable to set striping with a starting offset beyond OST 160
4785 Details : llapi_create_file() incorrectly limited the starting stripe
4786 index to the maximum single-file stripe count.
4789 Frequency : LDAP users only
4791 Description: lconf did not handle in-kernel recovery with LDAP properly
4792 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4794 Severity : enhancement
4796 Description: bind OST threads to NUMA nodes to improve performance
4797 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4798 node and do their allocations there to localize memory access
4800 Severity : enhancement
4802 Description: llmount can determine client NID directly from Myrinet (GM)
4803 Details : the client NID code from gmnalnid was moved directly into
4804 llmount, removing the need to use this or specifying the
4805 client NID explicitly when mounting GM clients with zeroconf
4808 Frequency : if client is started with down MDS
4810 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4811 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4812 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4813 ptlrpc_import_delayed_req() because we didn't check if the
4814 request was interrupted, and we also didn't break out of the
4815 event loop if there was no timeout
4820 Description: data loss during non-page-aligned writes to a single file from
4821 both multiple nodes and multiple threads on one node at same time
4822 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4823 inconsistency led to false short-reads, that were cached and later
4824 used by ->prepare_write() to fill in partially written page,
4825 leading to data loss.
4828 Frequency : always, if lconf --abort_recovery used
4830 Description: lconf --abort_recovery fails with 'Operation not supported'
4831 Details : lconf was attempting to abort recovery on the MDT device and not
4834 Severity : enhancement
4836 Description: remove cleanup logs
4837 Details : replace lconf-generated cleanup logs with lustre internal
4838 cleanup routines. Eliminates the need for client-cleanup and
4841 Severity : enhancement
4843 Description: add support for EAs (user and system) on lustre filesystems
4844 Details : it is now possible to store extended attributes in the Lustre
4845 client filesystem, and with the user_xattr mount option it
4846 is possible to allow users to store EAs on their files also
4848 Severity : enhancement
4850 Description: Add possibility (config option) to show minimal available OST free
4852 Details : When compiled with --enable-mindf configure option, statfs(2)
4853 (and so, df) will return least minimal free space available from
4854 all OSTs as amount of free space on FS, instead of summary of
4855 free spaces of all OSTs.
4857 Severity : enhancement
4859 Description: do not expand extent locks acquired on OST-side
4860 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4861 by server: they are not cached anyway.
4864 Frequency : when mmap is used/binaries executed from Lustre
4866 Description: Unmmap pages before throwing them away from read cache.
4867 Details : llap_shrink cache now attempts to unmap pages before discarding
4868 them (if unmapping failed - do not discard). SLES9 kernel has
4869 extra checks that trigger if this unmapping is not done first.
4874 Description: lconf didn't resolve symlinks before checking to see whether a
4875 given mountpoint was already in use
4878 Frequency : when migrating failover services
4879 Bugzilla : 6395, 9514
4880 Description: When migrating a subset of services from a node (e.g. failback
4881 from a failover service node) the remaining services would
4882 time out and evict clients.
4883 Details : lconf --force (implied by --failover) sets the global obd_timeout
4884 to 5 seconds in order to quickly disconnect, but this caused
4885 other RPCs to time out too quickly. Do not change the global
4886 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4888 Severity : enhancement
4889 Frequency : if MDS is started with down OST
4890 Bugzilla : 9439,5706
4891 Description: Allow startup/shutdown of an MDS without depending on the
4892 availability of the OSTs.
4893 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4894 Add appropriate locking and lov-osc refcounts for safe
4895 cleaning. Add osc abort_inflight calls in case the
4896 synchronize never started.
4899 Frequency : occasional (Cray XT3 only)
4901 Description: root not authorized to access files in CRAY_PORTALS environment
4902 Details : The client process capabilities were not honoured on the MDS in
4903 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4904 been accessed by an authorized user then root was able to access
4905 the file on the local client also. The root user capabilities
4906 are now allowed on the MDS, as this environment has secure UID.
4909 Frequency : occasional
4911 Description: ldiskfs "too long searching" message happens too often
4912 Details : A debugging message (otherwise harmless) prints too often on
4913 the OST console. This has been reduced to only happen when
4914 there are fragmentation problems on the filesystem.
4919 Description: Division by zero in statfs when all OSCs are inactive
4920 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4921 lov_statfs divides by value returned by lov_get_stripecnt().
4925 Bugzilla : 9489, 3273
4926 Description: First write from each client to each OST was only 4kB in size,
4927 to initialize client writeback cache, which caused sub-optimal
4928 RPCs and poor layout on disk for the first written file.
4929 Details : Clients now request an initial cache grant at (re)connect time
4930 and so that they can start streaming writes to the cache right
4931 away and always do full-sized RPCs if there is enough data.
4932 If the OST is rebooted the client also re-establishes its grant
4933 so that client cached writes will be honoured under the grant.
4938 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4939 Details : Now I/O RPCs go to different portal number and (presumably) fast
4940 lock requests (and glimses) and other RPCs get their own service
4941 threads pool that should be able to service those RPCs
4944 Severity : enhancement
4946 Description: Ability to exchange lustre version between client and servers and
4947 issue warnings at client side if client is too old. Also for
4948 liblustre clients there is ability to refuse connection of too old
4950 Details : New 'version' field is added to connect data structure that is
4951 filled with version info. That info is later checked by server and
4955 Frequency : rare, liblustre only.
4956 Bugzilla : 9296, 9581
4957 Description: Two simultaneous writes from liblustre at offset within same page
4958 might proceed at the same time overwriting eachother with stale
4960 Details : I/O lock withing llu_file_prwv was released too early, before data
4961 actually was hitting the wire. Extended lock-holding time until
4962 server acknowledges receiving data.
4965 Frequency : extremely rare. Never observed in practice.
4967 Description: avoid generating lustre_handle cookie of 0.
4968 Details : class_handle_hash() generates handle cookies by incrementing
4969 global counter, and can hit 0 occasionaly (this is unlikely, but
4970 not impossible, because initial value of cookie counter is
4971 selected randonly). Value of 0 is used as a sentinel meaning
4972 "unassigned handle" --- avoid it. Also coalesce two critical
4973 sections in this function into one.
4975 Severity : enhancement
4977 Description: allow liblustre clients to delegate truncate locking to OST
4978 Details : To avoid overhead of locking, liblustre client instructs OST to
4979 take extent lock in ost_punch() on client's behalf. New connection
4980 flag is added to handle backward compatibility.
4982 Severity : enhancement
4983 Bugzilla : 4928, 7341, 9758
4984 Description: allow number of OST service threads to be specified
4985 Details : a module parameter allows the number of OST service threads
4986 to be specified via "options ost ost_num_threads={N}" in the
4987 OSS's /etc/modules.conf or /etc/modprobe.conf.
4991 Bugzilla : 6146, 9635, 9895
4992 Description: servers crash with bad pointer in target_handle_connect()
4993 Details : In rare cases when a client is reconnecting it was possible that
4994 the connection request was the last reference for that export.
4995 We would temporarily drop the export reference and get a new
4996 one, but this may have been the last reference and the export
4997 was just destroyed. Get new reference before dropping old one.
4999 Severity : enhancement
5000 Frequency : if client is started with failover MDS
5002 Description: Allow multiple MDS hostnames in the mount command
5003 Details : Try to read the configuration from all specified MDS
5004 hostnames during a client mount in case the "primary"
5007 Severity : enhancement
5009 Description: Stop sending data to evicted clients as soon as possible.
5010 Details : Check if the client we are about to send or are sending data to
5011 was evicted already. (Check is done every second of waiting,
5012 for which l_wait_event interface was extended to allow checking
5013 of exit condition at specified intervals).
5016 Frequency : rare, normally only when NFS exporting is done from client
5018 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5020 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5023 Frequency : always, if srand() is called before liblustre initialization
5025 Description: Liblustre uses system PRNG disturbing its usage by user application
5026 Details : Introduce internal to lustre fast and high-quality PRNG for
5027 lustre usage and make liblustre and some other places in generic
5028 lustre code to use it.
5030 Severity : enhancement
5031 Bugzilla : 9477, 9557, 9870
5032 Description: Verify that the MDS configuration logs are updated when xml is
5033 Details : Check if the .xml configuration logs are newer than the config
5034 logs stored on the MDS and report an error if this is the case.
5035 Request --write-conf, or allow starting with --old_conf.
5037 Severity : enhancement
5039 Description: Handle symlinks in the path when checking if Lustre is mounted.
5040 Details : Resolve intermediate symlinks when checking if a client has
5041 mounted a filesystem to avoid duplicate client mounts.
5046 Description: lconf can hit an error exception but still return success.
5047 Details : The lconf command catches the Command error exception at the top
5048 level script context and will exit with the associated exit
5049 status, but doesn't ensure that this exit status is non-zero.
5054 Description: failure of ptlrpc thread startup can cause oops
5055 Details : Starting a ptlrpc service thread can fail if there are a large
5056 number of threads or the server memory is very fragmented.
5057 Handle this without oopsing.
5060 Frequency : always, only if liblustre and non-default acceptor port was used
5062 Description: liblustre cannot connect to servers with non-default acceptor port
5063 Details : tcpnal_set_default_params() was not called and was therefore
5064 ignoring the environment varaible TCPNAL_PORT, as well as other
5065 TCPNAL_ environment variables
5070 Description: two objects could be created on the same OST for a single file
5071 Details : If an OST is down, in some cases it was possible to create two
5072 objects on a single OST for a single file. No problems other
5073 than potential performance impact and spurious error messages.
5077 Bugzilla : 5681, 9562
5078 Description: Client may oops in ll_unhash_aliases
5079 Details : Client dcache may become inconsistent in race condition.
5080 In some cases "getcwd" can fail if the current directory is
5086 Description: Inode refcounting problems in NFS export code
5087 Details : link_raw functions used to call d_instantiate without obtaining
5088 extra inode reference first.
5092 Bugzilla : 9942, 9903
5093 Description: Referencing freed requests leading to crash, memleaks with NFS.
5094 Details : We used to require that call to ll_revalidate_it was always
5095 followed by ll_lookup_it. Also with revalidate_special() it is
5096 possible to call ll_revalidate_it() twice for the same dentry
5097 even if first occurence returned success. This fix changes semantic
5098 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5099 reference on a request referred from the intent.
5100 ll_intent_release() then releases such a request.
5103 Frequency : rare, normally benchmark loads only
5105 Description: unlinked inodes were kept in memory on the client
5106 Details : If a client is repeatedly creating and unlinking files it
5107 can accumulate a lot of stale inodes in the inode slab cache.
5108 If there is no other client load running this can cause the
5109 client node to run out of memory. Instead flush old inodes
5110 from client cache that have the same inode number as a new inode.
5113 Frequency : SLES9 2.6.5 kernel and long filenames only
5114 Bugzilla : 9969, 10379
5115 Description: utime reports stale NFS file handle
5116 Details : SLES9 uses out-of-dentry names in some cases, which confused
5117 the lustre dentry revalidation. Change it to always use the
5121 Frequency : rare, unless heavy write-truncate concurrency is continuous
5122 Bugzilla : 4180, 6984, 7171, 9963, 9331
5123 Description: OST becomes very slow and/or deadlocked during object unlink
5124 Details : filter_destroy() was holding onto the parent directory lock
5125 while truncating+unlinking objects. For very large objects this
5126 may block other threads for a long time and slow overall OST
5127 responsiveness. It may also be possible to get a lock ordering
5128 deadlock in this case, or run out of journal credits because of
5129 the combined truncate+unlink. Solution is to do object truncate
5130 first in one transaction without parent lock, and then do the
5131 final unlink in a new transaction with the parent lock. This
5132 reduces the lock hold time dramatically.
5135 Frequency : rare, 2.4 kernels only
5137 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5138 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5139 before unmounting their filesystem in order to avoid deadlock.
5140 The kernel_locked() function in 2.4 kernels only checks whether
5141 the kernel lock is held, not whether it is this process that is
5142 holding it as 2.6 kernels do.
5147 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5148 Details : The client ptlrpc code may be trying to reconnect to a down
5149 server before a previous connection attempt has timed out.
5150 Increase the reconnect interval to be longer than the connection
5151 timeout interval to avoid sending duplicate connections to
5155 Frequency : echo_client brw_test command
5157 Description: fix echo_client to work with OST preallocated code
5158 Details : OST preallocation code (5137) didn't take echo_client IO path
5159 into account: echo_client calls filter methods outside of any
5160 OST thread and, hence, there is no per-thread preallocated
5161 pages and buffers to use. Solution: hijack pga pages for IO. As
5162 a byproduct, this avoids unnecessary data copying.
5166 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5167 Description: Client can oops in mdc_commit_close() after open replay
5168 Details : It was possible for the MDS to return an open request with no
5169 transaction number in mds_finish_transno() if the client was
5170 evicted, but without actually returning an error. Clients
5171 would later try to replay that open and may trip an assertion
5172 Simplify the client close codepath, and always return an error
5173 from the MDS in case the open is not successful.
5176 Frequency : rare, 2.6 OSTs only
5178 Description: OST may deadlock under high load on fragmented files
5179 Details : If there was a heavy load and highly-fragmented OST filesystems
5180 it was possible to have all the OST threads deadlock waiting on
5181 allocation of biovecs, because the biovecs were not released
5182 until the entire RPC IO was completed. Instead, release biovecs
5183 as soon as they are complete to ensure forward IO progress.
5185 Severity : enhancement
5187 Description: Support for specifying external journal device at mount
5188 Details : If an OST or MDS device is formatted with an external journal
5189 device, this device major/minor is stored in the ext3 superblock
5190 and may not be valid for failover. Allow detecting and
5191 specifying the external journal at mount time.
5196 Description: Mounting an MDS with pending unlinked files may cause oops
5197 Details : target_finish_recovery() calls mds_postrecov() which returned
5198 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5199 considers this an error and immediately begins cleaning up the
5200 lov, just after starting the mds_lov process
5202 Severity : enhancement
5204 Description: Implement 'lfs df' to report actual free space on per-OST basis
5205 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5206 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5207 usage of objects; '-h' to report in human readable format.
5209 ------------------------------------------------------------------------------
5211 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5218 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5219 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5220 at startup during orphan unlinking. Instead, skip buddy update
5221 before setup, as it will be regenerated after recovery is complete.
5226 Description: If an OST is inactive, its locks might reference stale inodes.
5227 Details : lov_change_cbdata() must iterate over all namespaces, even if
5228 they are inactive to clear inode references from the lock.
5230 Severity : enhancement
5231 Frequency : occasional, if non-standard max_dirty_mb used
5233 Description: Client will block write RPCs if not enough grant
5234 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5235 then the client will block writes while waiting for another RPC
5236 to complete instead of consuming its dirty limit. With change
5237 we get improved performance when max_dirty_mb is small.
5239 Severity : enhancement
5240 Bugzilla : 3389, 6253
5241 Description: Add support for supplementary groups on the MDS.
5242 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5243 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5244 lookups for user supplementary groups into a cache.
5248 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5249 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5250 for liblustre O_EXCL is enforced by MDS.
5254 Description: readdir never returns NULL in liblustre.
5255 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5256 dentry in struct dirent.
5260 Frequency : liblustre only, depends on application IO pattern
5261 Description: liblustre clients evicted if not contacting servers
5262 Details : Don't put liblustre clients into the ping_evictor list, so
5263 they will not be evicted by the pinger ever.
5265 Severity : enhancement
5267 Description: Add ability to evict clients by NID from MDS.
5268 Details : By echoing "nid:$NID" string into
5269 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5270 $NID would be instantly evicted from this MDS and from all active
5271 OSTs connected to it.
5275 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5276 Details : lookup_it_finish() used to query file size from OSTs that was not
5281 Description: service threads change working directory to that of init
5282 Details : Starting lustre service threads may pin the working directory
5283 of the parent thread, making that filesystem busy. Threads
5284 now change to the working directory of init to avoid this.
5288 Frequency : during shutdown only
5289 Description: shutdown with a failed MDS or OST can cause unmount to hang
5290 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5295 Frequency : 2.6 only
5296 Description: chmod/chown may include an extra supplementary group
5297 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5298 supplementary group and if none is specified this is used.
5301 Bugzilla : 5479 (6816)
5303 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5304 Details : If lookup is in progress on a file that is unlinked we might try
5305 to revalidate the inode and fail in revalidate after lookup is
5306 complete and ll_file_open() enqueues the open again but
5307 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5310 Frequency : always, if lconf --abort_recovery used
5312 Description: lconf --abort_recovery fails with 'Operation not supported'
5313 Details : lconf was attempting to abort recovery on the MDT device and not
5316 ------------------------------------------------------------------------------
5318 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5323 Frequency : rare (only unsupported configurations with a node running as an
5325 Bugzilla : 6514, 5137
5326 Description: Mounting a Lustre file system on a node running as an OST could
5328 Details : OSTs now preallocates memory needed to write out data at
5329 startup, instead of when needed, to avoid having to
5330 allocate memory in possibly low memory situations.
5331 Specifically, if the file system is mounted on on OST,
5332 memory pressure could force it to try to write out data,
5333 which it needed to allocate memory to do. Due to the low
5334 memory, it would be unable to do so and the node would
5335 become unresponsive.
5337 Severity : enhancement
5339 Description: Addition of lconf --service command line option
5340 Details : lconf now accepts a '--service <arg>' option, which is
5341 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5343 Severity : enhancement
5345 Description: Failover mode is now the default for OSTs.
5346 Details : By default, OSTs will now run in failover mode. To return to
5347 the old behaviour, add '--failout' to the lmc line for OSTs.
5349 Severity : enhancement
5351 Description: Health checks are now provided for MDS and OSTs
5352 Details : Additional detailed health check information on MSD and OSTs
5353 is now provided through the procfs health_check value.
5356 Frequency : occasional, depends on IO load
5358 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5359 after numerous create/delete cycles
5360 Details : The ext3 inode allocation policy would not allocate new inodes
5361 very well on the OSTs because there are no new directories
5362 being created. Instead we look for groups with free space if
5363 the parent directories are nearly full.
5368 Description: Network or server problems during mount may cause partially
5369 mounted clients instead of returning an error.
5370 Details : The config llog parsing code may overwrite the error return
5371 code during mount error handling, returning success instead
5377 Description: MDS can fail to allocate large reply buffers
5378 Details : After long uptimes the MDS can fail to allocate large reply
5379 buffers (e.g. zconf client mount config records) due to memory
5380 fragmentation or consumption by the buffer cache. Preallocate
5381 some large reply buffers so that these replies can be sent even
5382 under memory pressure.
5386 Frequency : rare (liblustre)
5387 Description: fsx running with liblustre complained that using truncate() to
5388 extend the file doesn't work. This patch corrects that issue.
5389 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5390 ATTR_SIZE and lsm use in llu_setattr_raw.
5394 Frequency : rare, only 2.6 kernels
5395 Description: Unusual file access patterns on the MDS may result in inode
5396 data being lost in very rare circumstances.
5397 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5398 code in the 2.6 kernel caused the inode and/or EA data not to
5399 be read from disk, causing single-file corruption.
5403 Frequency : rare, only 2.6 filesystems using extents
5404 Description: Heavy concurrent write and delete load may cause data corruption.
5405 Details : It was possible under high-load situations to have an extent
5406 metadata block in the block device cache from a just-unlinked
5407 file overwrite a newly-allocated data block. We now unmap any
5408 metadata buffers that alias just-allocated data blocks.
5412 Frequency : filesystems with default stripe_count larger than 77
5413 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5414 Details : lconf specifies an inode size of 4096 bytes when the default
5415 stripe_count is larger than 77. This conflicts with the default
5416 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5417 this case to avoid pinning too much memory for large EAs.
5419 ------------------------------------------------------------------------------
5421 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5426 Frequency : rare (extremely heavy IO load with hundreds of clients)
5428 Description: Client is evicted, gets IO error writing to file
5429 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5430 caused clients to be evicted for AST timeouts. The fixes in
5431 bug 5192 mean we no longer need to have such short AST timeouts
5432 so ldlm_timeout has been increased.
5435 Frequency : occasional during --force or --failover shutdown under load
5436 Bugzilla : 5949, 4834
5437 Description: Server oops/LBUG if stopped with --force or --failover under load
5438 Details : a collection of import/export refcount and cleanup ordering
5439 issues fixed for safer force cleanup
5442 Frequency : only filesystems larger than 120 OSTs
5443 Bugzilla : 5990, 6223
5444 Description: lfs getstripe would oops on a very large filesystem
5445 Details : lov_getconfig used kfree on vmalloc'd memory
5448 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5449 Bugzilla : 6242, 6243
5450 Description: reading from files that had been truncated to a non-zero size
5451 but never opened returned no data
5452 Details : ll_file_read() reads zeros from no-object files to EOF
5457 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5458 Details : There's little harm in aborting MDS/OSS recovery and letting it
5459 try again, so I removed the LASSERT and return an error instead.
5461 Severity : enhancement
5463 Description: New debugging infrastructure for tracking down data corruption
5464 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5465 (b) cover more of the client-side code path, and (c) try to narrow
5466 down where problems occurred
5470 Bugzilla : 3819, 4364, 4397, 6313
5471 Description: Racing close and eviction MDS could cause assertion in mds_close
5472 Details : It was possible to get multiple mfd references during close and
5473 client eviction, leading to one thread referencing a freed mfd.
5475 Severity: : enhancement
5476 Bugzilla : 3262, 6359
5477 Description: Attempts to reconnect to servers are now more aggressive.
5478 Details : This builds on the enhanced upcall-less recovery that was added
5479 in 1.4.2. When trying to reconnect to servers, clients will
5480 now try each server in the failover group every 10 seconds. By
5481 default, clients would previously try one server every 25 seconds.
5486 Description: After recovery, certain operations trigger a failed
5487 assertion on a client.
5488 Details : Failing over an mds, using lconf -d --failover, while a
5489 client was doing a readdir() call would cause the client to
5490 LBUG after recovery completed and the readdir() was resent.
5492 Severity : enhancement
5494 Description: Default groups are now added by lconf
5495 Details : You can now run lconf --group <servicename> without having to
5496 manually add groups with lmc.
5499 Frequency : occasional
5501 Description: Nodes with an elan id of 0 trigger a failed assertion
5504 Frequency : always when accessing e.g. tty/console device nodes
5506 Description: tty and some other devices nodes cannot be used on lustre
5507 Details : file's private_data field is used by device data and lustre
5508 values in there got lost. New field was added to struct file to
5509 store fs-specific private data.
5512 Frequency : when exporting Lustre via NFS
5514 Description: NFSD failed occasionally when looking up a path component
5515 Details : NFSD is looking up ".." which was broken in ext3 directories
5516 that had grown large enough to become hashed.
5519 Frequency : Clusters with multiple interfaces not on the same subnet
5521 Description: Nodes will repeatedly try to reconnect to an interface which it
5522 cannot reach and report an error to the log.
5523 Details : Extra peer list entries will be created by lconf with some peers
5524 unreachable. lconf now validates the peer before adding it.
5527 Frequency : Only if a default stripe is set on the filesystem root.
5529 Description: Setting a default stripe on the filesystem root prevented the
5530 filesystem from being remounted.
5531 Details : The client was sending extra request flags in the root getattr
5532 request and did not allocate a reply buffer for the dir EA.
5535 Frequency : occasional, higher if lots of files are accessed by one client
5536 Bugzilla : 6159, 6097
5537 Description: Client trips assertion regarding lsm mismatch/magic
5538 Details : While revalidating inodes the VFS looks up inodes with ifind()
5539 and in rare cases can find an inode that is being freed.
5540 The ll_test_inode() code will free the lsm during ifind()
5541 when it finds an existing inode and then the VFS later attaches
5542 this free lsm to a new inode.
5546 Bugzilla : 6422, 7030
5547 Description: MDS deadlock between mkdir and client eviction
5548 Details : Creating a new file via mkdir or mknod (starting a transaction
5549 and getting the ns lock) can deadlock with client eviction
5550 (gets ns lock and trying to finish a synchronous transaction).
5553 Frequency : occasional
5554 Description: While starting a server, the fsfilt_ext3 module could not be
5556 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5557 kernels. Previously, lconf would still use the ext3 name
5558 when trying to load modules. Now, it will correctly use
5559 ext3 on 2.4 and ldiskfs on 2.6.
5561 Severity : enhancement
5562 Description: The default stripe count has been changed to 1
5563 Details : The interpretation of the default stripe count (0, to lfs
5564 or lmc) has been changed to mean striping across a single
5565 OST, rather than all available. For general usage we have
5566 found a stripe count of 1 or 2 works best.
5568 Severity : enhancement
5569 Description: Add support for compiling against Cray portals.
5570 Details : Conditional compiling for some areas that are different
5574 Frequency : occasional
5575 Bugzilla : 6409, 6834
5576 Description: Creating files with an explicit stripe count may lead to
5577 a failed assertion on the MDS
5578 Details : If some OSTs are full or unavailable, creating files may
5579 trigger a failed assertion on the MDS. Now, Lustre will
5580 try to use other servers or return an error to the
5584 Frequency : occasional
5586 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5587 caused lock timeout during readahead (since 1.4.2).
5588 Details : Processes doing readahead might match a lock that hasn't been
5589 granted yet if there are overlapping and conflicting lock
5590 requests. The readahead process waits on ungranted lock
5591 (original lock is CBPENDING), while OST waits for that process
5592 to cancel CBPENDING read lock and eventually evicts client.
5594 Severity : enhancement
5596 Description: Initial enabling of flock support for clients
5597 Details : Implements fcntl advisory locking and file status functions.
5598 This feature is provided as an optional mount flag (default
5599 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5600 locking are implemented yet, and those that are are not guaranteed
5601 to be completely correct in production environments.
5602 mount -t lustre -o [flock|noflock] ...
5605 Frequency : occasional
5607 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5608 assertion in the ext3 JBD (journaling) layer.
5609 Details : The b_committed_data struct is protected by the big kernel lock
5610 in 2.4 kernels, serializing journal_commit_transaction() and
5611 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5612 access to this struct. In 2.6 kernels there is finer grained
5613 locking to improve SMP performance of the JBD layer.
5617 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5622 Description: merge and cleanup kernel patches.
5625 -----------------------------------------------------------------------------
5627 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5629 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5630 versions of Lustre. Please update all servers and clients to
5631 version 1.4.2 or later at the same time. You must also run
5632 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5633 to update the configuration logs.
5635 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5636 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5637 - swab llog records of type '0' so we get proper header size/idx (5861)
5638 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5639 - fix rename of one directory over another leaking an inode (5953)
5640 - avoid SetPageDirty on 2.6 (5981)
5641 - don't re-add just-being-destroyed locks to the waiting list (5653)
5642 - when creating new directories, inherit the parent's custom
5643 striping settings if present parent (3048)
5644 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5645 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5646 - don't hold client locks on temporary worklist from l_lru (5666)
5647 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5648 - automatically evict dead clients (5921)
5649 - Update file size properly in create+truncate+fstat case (6196)
5650 - Do not unhash mountpoint dentries, do not allow removal of
5652 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5653 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5654 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5655 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5656 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5657 - Correctly handle reads of files with no objects (6243)
5658 - lctl recover will also mark a device active if deactivate used (5933)
5660 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5661 - allow --write-conf on an MDS with different nettype than client (5619)
5662 - don't write config llogs to MDS for mounts not from that MDS (5617)
5663 - lconf should create multiple TCP connections from a client (5201)
5664 - init scripts are now turned off by default; run chkconfig --on
5665 lustre and chkconfig --on lustrefs to use them
5666 - upcalls are no longer needed for clients to recover to failover
5668 - add --abort-recovery option to lconf to abort recovery on device
5670 - add support for an arbitrary number of OSTs (3026)
5671 - Quota support protocol changes.
5672 - forward compatibility changes to wire structs (6007)
5673 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5674 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5675 - improved "lustre status" script
5676 - initialize blocksize for non-regular files (6062)
5677 - added --disable-server and --disable-client configure options (5782)
5678 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5679 - Vanilla 2.4.29 support
5680 - increase maximum number of obd devices to 520 (6242)
5681 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5682 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5683 SLES 9 2.6.5 kernels
5684 - limit stripes per file to 160 (the maximum EA size) (6093)
5686 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5689 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5690 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5691 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5692 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5693 - small ext3 extents cleanups and fixes (5733)
5694 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5695 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5696 - avoid needless client->OST connect, fix handle mismatch (5317)
5697 - fix DLM error path that led to out-of-sync client, long delays (5779)
5698 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5699 - fix several locking issues related to i_size (5492,5624,5654,5672)
5700 - don't move pending lock onto export if it is already evicted (5683)
5701 - fix kernel oops when creating .foo in unlinked directory (5548)
5702 - fix deadlock in obdfilter statistics vs. object create (5811)
5703 - use time_{before,after} to avoid timer jiffies wrap (5882)
5704 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5705 - Do not leak request if server was not able to process it (5154)
5706 - If mds_open unable to find parent dir, make that negative lookup(5154)
5707 - don't create new directories with extent-mapping (5909, 5936)
5709 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5710 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5711 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5712 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5713 - support for RHEL4 kernel on i686 (5773)
5714 - provide error messages when incompatible logs are encountered (5898)
5716 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5717 * version 1.4.0.10 (1.4.1 release candidate 1)
5719 - don't keep a lock reference when lock is not granted (4238)
5720 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5721 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5722 - drop import inflight refcount on signal_completed_replay error (5255)
5723 - unlock page after async write error during send (3677)
5724 - handle missing objects in filter_preprw_read properly (5265)
5725 - no transno return for symlink open, don't save no-trasno open (3440)
5726 - don't try to complete elan receive that already failed (4012)
5727 - free RPC server reply state on error (5406)
5728 - clean up thread from ptlrpc_start_thread() on error (5160)
5729 - readahead could read extra page into cache that wasn't ejected (5388)
5730 - prevent races in class_attach/setup/cleanup/detach (5260)
5731 - don't dereference de->d_inode after l_dput of de (5458)
5732 - use "int" for stripe value returned from lock_to_stripe (5544)
5733 - mballoc allocation and error-checking fixes in 2.6 (5504)
5734 - block device patches to fix I/O request sizes in 2.6 (5482)
5735 - look up hostnames for IB nals (5602)
5736 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5737 - don't start multiple acceptors for the same port (5277)
5738 - fix incorrect LASSERT in mds_getattr_name (5635)
5739 - export a proc file for general "ping" checking (5628)
5740 - fix "lfs check" to not block when the MDS is down (5628)
5742 - service request history (4965)
5743 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5744 - create an "evict_client" /proc entry on OSTs, like the MDS has
5745 - fix mount usage message, return errors per mount(8) (5168)
5746 - change grep [] to grep "[]" in tests so they work in more UMLs
5747 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5748 - remove ancient LOV_MAGIC_V0 EA support (5047)
5749 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5750 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5751 - allow readahead allocations to fail when low on memory (5383)
5752 - mmap locking landed again, after considerable improvement (2828)
5753 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5755 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5758 - send OST transaction number in read/write reply to free req (4966)
5759 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5760 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5761 - the watchdog thread now runs as interruptible (5246)
5762 - flock/lockf fixes (but it's still disabled, pending 5135)
5763 - don't use EXT3 constants in llite code (5094)
5764 - memory shortage at startup could cause assertion (5176)
5766 - reorganization of lov code
5767 - single portals codebase
5769 - add extents/mballoc support (5025)
5770 - direct I/O reads in the obdfilter (4048)
5771 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5773 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5776 - send OST transaction number in read/write reply to free req (4966)
5777 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5778 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5779 - the watchdog thread now runs as interruptible (5246)
5780 - handle missing objects in filter_preprw_read properly (5265)
5781 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5782 - drop import inflight refcount on signal_completed_replay error (5255)
5783 - unlock page after async write error during send (3677)
5784 - return original error code on reconstructed replies (3761)
5785 - no transno return for symlink open, don't save no-trasno open (3440)
5787 - add pid to ldlm debugging output (4922)
5788 - bump the watchdog timeouts -- we can't handle 30sec yet
5789 - extra debugging for orphan dentry/inode bug (5259)
5791 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5794 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5795 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5796 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5797 - fix NULL dereference in /proc/sys/portals/routes (4827)
5798 - allow failed mdc_close() operations to be interrupted (4561)
5799 - stop precreate on OST before MDS would time out on it (4778)
5800 - don't send partial-page writes before EOF from client (4410)
5801 - discard client grant for sub-page writes on large-page clients (4520)
5802 - don't free dentries not owned by NFS code, check generation (4806)
5803 - fix lsm leak if mds_create_objects() fails (4801)
5804 - limit debug_daemon file size, always print CERROR messages (4789)
5805 - use transno after validating reply (3892)
5806 - process timed out requests if import state changes (3754)
5807 - update mtime on OST during writes, return in glimpse (4829)
5808 - add mkfsoptions to LDAP (4679)
5809 - use ->max_readahead method instead of zapping global ra (5039)
5810 - don't interrupt __l_wait_event() during strace
5812 - add software watchdogs to catch hung threads quickly (4941)
5813 - make lustrefs init script start after nfs is mounted
5814 - fix CWARN/ERROR duplication (4930)
5815 - return async write errors to application if possible (2248)
5816 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5817 - print NAL number in %x format (4645)
5818 - update barely-supported suse-2.4.21-171 series (4842)
5819 - support for sles 9 %post scripts
5820 - support for building 2.6 kernel-source packages
5821 - support for sles km_* packages
5823 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5826 - ignore -ENOENT errors in osc_destroy (3639)
5827 - notify osc create thread that OSC is being cleaned up (4600)
5828 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5829 - reconstruct ost_handle() like mds_handle() (4657)
5830 - create a new thread to do import eviction to avoid deadlock (3969)
5831 - let lconf resolve symlinked-to devices (4629)
5832 - don't unlink "objects" from directory with default EA (4554)
5833 - hold socknal file ref over connect in case target is down (4394)
5834 - allow more than 32000 subdirectories in a single directory (3244)
5835 - fix blocks count for O_DIRECT writes (3751)
5836 - OST returns ENOSPC from object create when no space left (4539)
5837 - don't send truncate RPC if file size isn't changing (4410)
5838 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5839 - bind to privileged port in socknal and tcpnal (3689)
5841 - rate limit CERROR/CWARN console message to avoid overload (4519)
5842 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5843 - basic mmap support (3918)
5844 - kernel patch series update from b1_4 (4711)
5846 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5849 - avoid crash during MDS cleanup with OST shut down (2775)
5850 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5851 - don't use bad inodes on the MDS (3744)
5852 - dynamic object preallocation to improve recovery speed (4236)
5853 - don't hold spinlock over lock dumping or change debug flags (4401)
5854 - don't zero obd_dev when it is force cleaned (3651)
5855 - print grants to console if they go negative (4431)
5856 - "lctl deactivate" will stop automatic recovery attempts (3406)
5857 - look for existing locks in ldlm_handle_enqueue() (3764)
5858 - don't resolve lock handle twice in recovery avoiding race (4401)
5859 - revalidate should check working dir is a directory (4134)
5861 - don't always mark "slow" obdfilter messages as errors (4418)
5863 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5866 - don't close LustreDB during write_conf until it is done (3860)
5867 - fix typo in lconf for_each_profile (3821)
5868 - allow dumping logs from multiple threads at one time (3820)
5869 - don't allow multiple threads in OSC recovery (3812)
5870 - fix debug_size parameters (3864)
5871 - fix mds_postrecov to initialize import for llog ctxt (3121)
5872 - replace config semaphore with spinlock (3306)
5873 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5874 - don't allow enqueue to complete on a destroyed export (3822)
5875 - down write_lock before checking llog header bitmap (3825)
5876 - recover from lock replay timeout (3764)
5877 - up llog sem before sending rpc (3652)
5878 - reduce ns lock hold times when setting kms (3267)
5879 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5880 - fix NULL deref and obd_dev leak on setup error (3312)
5881 - replace some LBUG about llog ops with error handling (3841)
5882 - don't match INVALID dentries from d_lookup and spin (3784)
5883 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5884 - fix invalid assertion in ptlrpc_set_wait (3880)
5886 - add libwrap support for the TCP acceptor (3996)
5887 - add /proc/sys/portals/routes for non-root route listing (3994)
5888 - allow setting MDS UUID in .xml (2580)
5889 - print the stack of a process that LBUGs (4228)
5891 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5894 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5895 - make sure to unset replay flag from failed open requests (3440)
5896 - if default stripe count is 0, use OST count for inode size (3636)
5897 - update parent mtime/ctime on client for create/unlink (2611)
5898 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5899 - free recovery state on server during a forced cleanup (3571)
5900 - unregister_reply for resent reqs (3063)
5901 - loop back devices mounting and status check on 2.6 (3563)
5902 - fix resource-creation race that can provoke i_size == 0 (3513)
5903 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5904 - more debugging for page-accounting assertion (3746)
5905 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5906 - avoid deadlock after precreation failure (3758)
5907 - fix race and lock order deadlock in orphan handling (3450, 3750)
5908 - add validity checks when grabbing inodes from l_ast_data (3599)
5910 - add /proc/.../recovery_status to obdfilter (3428)
5911 - lightweight CDEBUG infrastructure, debug daemon (3668)
5912 - change default OSC RPC parameters to be better on small clusters
5913 - turn off OST read cache for files smaller than 32MB
5914 - install man pages and include them in rpms (3100)
5915 - add new init script for (un)mounting lustre filesystems (2593)
5916 - run chkconfig in %post for init scripts (3701)
5917 - drop scimac NAL (unmaintained)
5919 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5922 - clean kiobufs before and after use (3485)
5923 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5924 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5925 - change init script to fail more clearly if not run as root (1528)
5926 - allow clients to reconnect during replay (1742)
5927 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5928 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5929 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5930 - fix cond_resched (3554)
5931 - only evict unfinished clients after recovery (3515)
5932 - allow bulk resend, prevent data loss (3570)
5933 - dynamic ptlrpc request buffer allocation (2102)
5934 - don't allow unlinking open directory if it isn't empty (2904)
5935 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5936 - remove extraneous obd dereference causing LASSERT failure (3334)
5937 - don't use get_cycles() when creating temp. files on the mds (3156)
5938 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5939 - handle EEXIST for set-stripe, set proper directory name (3336)
5941 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5942 - fix ksocknal_fmb_callback() error messages (2918)
5944 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5947 - don't copy lvb into (possibly NULL) reply on error (2983)
5948 - don't deref dentry after dput, don't free lvb on error (2922)
5949 - use the kms to determine writeback rpc length (2947)
5950 - increment oti_logcookies when osc is inactive (2948)
5951 - update client's i_blocks count via lvb messages (2543)
5952 - handle intent open/close of special files properly (1557)
5953 - mount MDS with errors=remount-ro, like obdfilter (2009)
5954 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5955 - don't use cancelling-locks' kms values (2947)
5956 - use highest lock extent for kms, not last one (2925)
5957 - don't dereference ERR_PTR() dentry in error handling path (3107)
5958 - fix thread race in portals_debug_dumplog() (3122)
5959 - create lprocfs device entries at setup instead of at attach (1519)
5960 - common AST error handler, don't evict client on completion race (3145)
5961 - zero nameidata in detach_mnt in 2.6 (3118)
5962 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5963 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5964 - initialize RPC timeout timer earlier for 2.6 (3219)
5965 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5966 - print nal/nid for unknown nid (3258)
5967 - additional checks for oscc recovery before doing precreate (3284)
5968 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5969 - don't crash in mdc_close for bad permissions on open (3285)
5970 - zero i_rdev for non-device files (3147)
5971 - clear page->private before handing to FS, better assertion (3119)
5972 - tune the read pipeline (3236)
5973 - fix incorrect decref of invalidated dentry (2350)
5974 - provide read-ahead stats and refine rpc in flight stats (3328)
5975 - don't hold journal transaction open across create RPC (3313)
5976 - update atime on MDS at close time (3265)
5977 - close LDAP connection when recovering to avoid server load (3315)
5978 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5979 - don't leak open file on MDS after open resend (3325)
5980 - serialize filter_precreate and filter_destroy_precreated (3329)
5981 - loop device shouldn't call sync_dev() for nul device (3092)
5982 - clear page cache after eviction (2766)
5983 - resynchronize MDS->OST in background (2824)
5984 - refuse to mount the same filesystem twice on same mountpoint (3394)
5985 - allow llmount to create routes for mounting behind routers (3320)
5986 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5987 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5988 - fix rare problem with rename on htree directories (3417)
5990 - allow default OST striping configuration per directory (1414)
5991 - fix compilation for qswnal for 2.6 kernels (3125)
5992 - increase maximum number of MDS request buffers for large systems
5993 - change liblustreapi to be useful for external progs like lfsck (3098)
5994 - increase local configuration timeout for slow disks (3353)
5995 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5997 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6000 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6001 - don't overwrite extent policy data in reply if lock was blocked (2901)
6002 - drop filter export grants atomically with removal from device (2663)
6003 - del obd_self_export from work_list in class_disconnect_exports (2908)
6004 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6005 - swab reply message in mdc_close, other PPC fixes (2464)
6006 - fix destroying of named logs (2325)
6007 - overwrite old logs when running lconf --write_conf (2264)
6008 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6009 - fix race in target_handle_connect (2898)
6010 - mds_reint_create() should take same inode create lock (2926)
6011 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6012 - don't close files for self_export to avoid uninitialized obd (2936)
6013 - allow MDS with the same name as client node (2939)
6014 - hold dentry reference for closed log files for unlink (2325)
6015 - reserve space for all logs during transactions (2059)
6016 - don't evict page beyond end of stripe extent (2925)
6017 - don't oops on a deleted current working directory (2399)
6018 - handle hard links to targets without a parent properly (2517)
6019 - don't dereference NULL lock when racing during eviction (2867)
6020 - don't grow lock extents when lots of conflicting locks (2919)
6022 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6025 - account for cache space usage on clients to avoid data loss (974)
6026 - lfsck support in lustre kernel code (2349)
6027 - reduce journal credits needed for BRW writes (2370)
6028 - orphan handling to avoid losing space on client/server crashes
6029 - ptlrpcd can be blocked, stopping ALL progress (2477)
6030 - use lock value blocks to assist in proper KMS, faster stat (1021)
6031 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6032 - recovery for initial connections (2355)
6033 - fixes for mds_cleanup_orphans (1934)
6034 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6035 - block all file creations until orphan recovery completes (1901)
6036 - client remove rq_connection from request struct (2423)
6037 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6038 - recovery timer race (2670)
6039 - mdc_close recovey bug (2532)
6040 - ptlrpc cleanup bug (2710)
6041 - mds timeout on local locks (2588)
6042 - namespace lock held during RPCs (2431)
6043 - handle interrupted sync write properly (2503)
6044 - don't try to handle a message that hasn't been replied to (2699)
6045 - client assert failure during cleanup after abort recovery (2701)
6046 - leak mdc device after failed mount (2712)
6047 - ptlrpc_check_set allows timedout requests to complete (2714)
6048 - wait for inflight reqs when ptlrpcd finishes (2710)
6049 - make sure unregistered services are removed from the srv_list
6050 - reset bulk XID's when resending them (caught by 1138 test)
6051 - unregister_bulk after timeout
6052 - fix lconf error (2694)
6053 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6054 - readahead locks pages, leaves pending causing memory pressure (2673)
6055 - increase OST request buffers to 4096 on large machines (2729)
6056 - fix up permission of existing directories in simple_mkdir (2661)
6057 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6058 - don't assign transno to errored transactions (2742)
6059 - don't delete objects on OST if given a bogus objid from MDS (2751)
6060 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6061 - if rq_no_resend, then timeout request after recovery (2432)
6062 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6063 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6064 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6065 - MDS recovery shouldn't delete objects that it has given out (2730)
6066 - if enqueue arrives after completion, don't clobber LVB (2819)
6067 - don't unlock pages twice when trigger_group_io returns error (2814)
6068 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6069 - don't write pages to disk if there was an error (1450)
6070 - don't ping imports that have recovery disabled (2676)
6071 - take buffered bytes into account when balancing socknal conn (2817)
6072 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6073 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6074 - remove little-endian swabbing of llog records (1987)
6075 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6076 - retry reposting request buffers if they fail (1191)
6077 - grow extent at grant time to avoid granting a revoked lock (2809)
6078 - lock revoke doesn't evict page if covered by a second lock (2765)
6079 - disable VM readahead to avoid reading outside lock extents (2805)
6081 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6082 - updated kernel patches for hp-2.4.20 kernel (2681)
6084 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6087 - fix truncated write corruption (2366)
6088 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6090 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6091 - ensure len(uuid) < 37 in lmc (1171)
6092 - fix ia64 OOPS in llog_test (2255)
6093 - zero end of page at obdfilter for partial page writes (2648)
6094 - don't leave stale dentries around after renames (bug 2428)
6095 - fix timeouts when evicting a client with a single lock held (2642)
6096 - set deadline for the initial HELLO message to drain (2634)
6097 - print out dotted-quad IP addresses in the socknal (2302)
6099 - additional debugging for MDS client eviction problem (2443)
6100 - fix mkfsoptions support for osts (2603, 2604)
6102 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6105 - add series for the vanilla 2.6.0 kernel
6106 - add series for the vanilla 2.4.24 kernel
6107 - add series for a cray x86/64 UL kernel drop
6108 - fix xattr patches for the vanilla 2.4.19 series
6110 - generate true UUIDs in lmc (1171)
6111 - have portals stack dumping break in UML (2466)
6112 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6113 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6114 - wake callback waiting threads on client eviction (2460)
6115 - Add --ptldebug and --subsystem to lmc (1719)
6116 - update assertion to allow safe interrupt allocation
6117 - set rq_no_resend for cancel requests (2432)
6118 - recalculate ptlrpcd timeout after resend (2494)
6119 - call vfs_rmdir when removing pending directories (2368)
6120 - fix renaming a file to itself (2429)
6121 - lmc creates a default one-stripe lov (2454)
6122 - expand procfs space to handle large clusters (2326)
6123 - increase UML stack to avoid overflow
6124 - update lconf's list of debug and subsystem masks
6125 - fix lfs find --obd (2510)
6126 - /proc tunable for disabling filter read caching (2591)
6127 - stop rpm packages from altering slapd.conf (2301)
6128 - disable nagle in the socknal under 0conf (2578)
6129 - choose mds inode size based on stripe count (2572)
6130 - fix kernel-source rpm problems (2516)
6132 - add --disable-doc to avoid pdf generation (2421)
6133 - update documentation, tests, type-os, comments
6134 - avoid format warnings on ia64
6135 - remove the TOE NAL
6136 - tiny code cleanups by removing unused fields
6138 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6141 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6142 - remove the most blatant lies from BUILDING file (2371)
6143 - change default debug level to reasonable production setting
6144 - reduce client side cache size to reduce cache flush time
6145 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6146 - make TCP zerocopy and pinger support enabled by default (2476)
6147 - sync writes completed after process exits caused crashes (2319)
6148 - maintain correct mount count on the MDS (2356)
6149 - backout 1557, because 2316 wasn't really fixed
6150 - better file I/O statistics gathering in /proc
6151 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6152 - another kernel patch to fix zero-copy TCP function export
6153 - don't take duplicate lock when processing re-sent getattr (2420)
6154 - lctl uses obd_self_export instead of creating new conn (2353)
6155 - MDS/OST recovery case which requires object creation asserted (2425)
6156 - move lfs from /usr/sbin to /usr/bin in packages
6157 - fix race between mds_client_add and mds_client_free (2417)
6158 - use kmalloc instead of slabs in portals (2430)
6159 - don't create duplicate records when a failover MDS is present (2442)
6160 - remove unnecessary mount age check (2332)
6161 - don't remove directory inodes from locks prematurely (2451)
6162 - don't break if MDS service name is the same as hostname (2103)
6163 - fix races in client write RPC generation when cache full (2482)
6165 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6168 - remove now-unused request->rq_obd (278)
6169 - if an allocation fails, print out how much memory we've used (1933)
6170 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6171 - add the "configurable stack size" patch to most series files (1256)
6172 - ability to write large log records, for 100+ OST configs (2306)
6173 - fix NULL deref when filter_prep fails (2314)
6174 - fix operator precedence error in filter_sync
6175 - dynamic allocation of socknal TX descriptors (2315)
6176 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6177 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6178 - ensure that CWARN messages in Portals always get to the syslog
6179 - __init/__exit are not for prototype decls (ldlm_init/exit)
6180 - x86-64 compile warning fixes
6181 - fix gateway LMC keyword conflict (2318)
6182 - fix MDS lock inversions in getattr/reint paths (1844)
6183 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6184 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6185 - DLM race condition prevented some lock evictions (2328)
6186 - ENOMEM detection and retry on socknal sends (2230)
6187 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6188 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6189 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6190 - better error messages when a client is rejected during recovery (1505)
6191 - avoid cancelling locks which were never granted, after failure (2330)
6192 - fix i_sem/journal inversion in mds_client_add (2333)
6193 - fix truncate/getattr lock cycle deadlock (2334)
6194 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6195 - fix two rare exit paths which could leak an l_lock() ref (2321)
6196 - fencepost error in MDS/OST orphan recovery (2226)
6197 - make log record alignment 8 bytes (1988)
6198 - lstripe now fails when requested offset > ost_count (2237)
6199 - ensure that all kernel series have a complete list.h (1607)
6200 - fix crashes in special-file operations (2316)
6201 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6203 - allow configurable automake binary, for testing new versions
6204 - small update to the lfs documentation
6206 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6208 * fix negative export reference count in fsfilt_sync (2312)
6210 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6211 * release candidate 0.9.1
6213 - orphans are moved into the PENDING directory for possible recovery
6214 - replayed opens now open by fid for orphan/rename safety (1042)
6215 - last close of an orphan inode generates a transno (683)
6216 - chdir() and mount() now pin the directory entry (1020)
6217 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6218 - discard very old requests without processing them (1502)
6219 - remove some common, well-understood CERRORs (1505)
6220 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6221 - clear "grant" flags in OST replies until OST grant code lands (1644)
6222 - fix read performance by not clobbering i_blksize on client (1598)
6223 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6224 - make LRU size a /proc tunable, clears locks when reduced (707)
6225 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6226 - server locks take references on exports now (1558)
6227 - build fixes for 2.4.20-rh trees (1663)
6228 - return an error from lov_create if all OSCs are inactive (1751)
6229 - fix import levels when a reconnect happens without a timeout (1597)
6230 - exit early from mds_open if we get a lookup error (1749)
6231 - partial page read at EOF wouldn't wait for disk before sending (1642)
6232 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6233 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6234 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6235 - write/truncate lock inversion (1639)
6236 - Don't auto-load obdclass, portals modules during cleanup (1495)
6237 - fix timestamps from jumping to "now" (1763)
6238 - extra journal assertions (1648)
6239 - add an extra multiunlink test (1771)
6240 - fix read_record/write_record API (1776)
6241 - fix leak of offset_extent, possible incorrect i_size later (1772)
6242 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6243 - Debugging for the kqswnal_get_idle_tx problems (1820)
6244 - Allow recovery to be attempted multiple times (1536)
6245 - Write out MDS last_rcvd file after it is first created (1600)
6246 - Fix tx_descriptor leak in failed transmit situations (1827)
6247 - ext3 journaling fixes for assertion failure after IO error (1871)
6248 - class_export_put() on freed export after completion AST error (1896)
6249 - Fix revalidate looping in VFS (1322)
6250 - Don't access a freed export during MDS_REINT timeout (1521)
6251 - Add open-unlink recovery support on the MDS (1673,1764)
6252 - Return an error if no MDS data was read from last_rcvd (1946)
6253 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6254 - Don't setup a disk device that doesn't match exported UUID (317)
6255 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6256 - avoid committing NULL handle in force close
6257 - local.sh is now a one-stripe LOV configuration
6258 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6259 - fix ext3 htree duplicate directory entry corruption (1516)
6260 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6261 - update to new LOV EA format (2097)
6262 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6264 2003-06-15 Phil Schwan <phil@clusterfs.com>
6267 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6268 - per-import/export recovery handling (958, 931, 959)
6269 - multiple last-rcvd slots, for serving multiple FSes (949)
6270 - connections are again shared between multiple imp/exports (963, 964)
6271 - "umount -f" would hang if any requests needed to be sent (393, 978)
6272 - avoid pinning large req buffer by copying for queued messages (989)
6273 - add "uuid" to "lctl device" command to help upcalls (991)
6274 - "open" RPCs with transnos would confuse recovery counters (1037)
6275 - do proper endian conversion of all wire messages (288, 340, 891)
6276 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6277 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6278 - fix ptlrpc_connection leak in target_handle_connect (1174)
6279 - fix import refcounting bug in OST and MDS cleanup (1134)
6280 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6281 - fix very unlikely obd_types race condition (501)
6282 - remove osc_open hack for echo_client (1187)
6283 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6284 - a failure in read_inode2 leads to deadlock (1139)
6285 - cancel ack-locks as soon as transaction is committed (1072)
6286 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6287 - make sure to commitrw after any preprw to avoid deadlock (1162)
6288 - failing to execute a file in a lustre FS would lock inode (1203)
6289 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6290 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6291 - obd_uuid2tgt would walk off the end of the list (1255)
6292 - on IA64 the osc would give portals incorrect bulk size (1258)
6293 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6294 - fix lock inversion caused by new llite matching code (1282)
6295 - limit the number of dirty pages on a client to 10MB (1286)
6296 - timed out locks were not being corrected cancelled (1289)
6297 - fix O_DIRECT above 4GB on IA-32 (1292)
6298 * major user-visible changes
6299 - fail out/fail over policy now controlled by the upcall (993)
6301 - add OBD_PING to check server availability and failure (954)
6302 - lustre messages are now sent in sending host order (288, 340, 891)
6303 - add eadatalen to MDS getattr reply (340)
6304 - OST read replies may contain second buffer, with per-page status (593)
6306 2003-03-11 Phil Schwan <phil@clusterfs.com>
6309 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6310 - failed open()s could cause deadlock; fixed (867, 869)
6311 - stop cancelling OST locks when files are closed (481)
6312 - overlapping XID spaces caused network corruption (851, 853)
6313 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6314 - setattr_raw added, to do single-RPC, server-side setattrs
6315 - lmc/lconf syntax change for OST UUIDs
6316 - fix crashy race condition between ptlrpc_free_req and osc_close
6317 - don't use request in mdc_enqueue if we hit a timeout (889)
6318 - don't set the inode i_size for regular files from the MDS (896)
6319 - handle out of order completion AST (842)
6320 - don't LBUG if a lock request times out after receiving AST (913)
6321 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6322 - if a bad lock AST arrives, send an error instead of dropping entirely
6323 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6324 - fix leak in bulk IO when only partially completed (899, 900, 926)
6325 - fix O_DIRECT for ia64 (55)
6326 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6327 - C-z after timeout could hang a process forever; fixed (977)
6329 - client-side I/O cache (678, 924, 929, 941, 970)
6331 - READPAGE and SETATTRs which don't take server-side locks get
6334 2003-02-11 Phil Schwan <phil@clusterfs.com>
6337 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6338 - Fix fsfilt_extN_readpage() to read a full page of directory
6339 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6340 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6341 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6342 - Most intent operations were removed, in favour of a new RPC mode
6343 that does a single RPC to the server and bypasses most of the VFS
6344 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6345 - Aggressively cancel local locks on DLM servers
6346 - mds_reint_unlink sends EA to the client if it's the last nlink.
6347 client uses that EA to unlink OST objects.
6348 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6349 - recursive symlinks were fixed (439)
6350 - fixed NULL deref in DEBUG_REQ
6351 - filter_update_lastobjid no longer calls sync, which annoyed extN
6352 - fixed multi-client small-writes to a single file problem (445)
6353 - fixed mtime updates during file writes (607)
6354 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6355 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6356 - fixed Linux OST size reporting problem (444, 656)
6357 - OST now updates object mtime with writes or setattr (607, 619)
6358 - client verifies file size before zeroing page past EOF (445)
6359 - OST now writes last allocated objid to disk with allocation (108)
6360 - LOV on echo now works (409)
6362 - mds_reint_unlink sends a new buffer, with the EA included. this
6363 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6364 set if a regular file was being unlinked, and it was the last link
6365 - use PtlGet from the target for bulk writes (315)
6366 - OST now updates object mtime with writes or setattr (607, 619)
6367 - LDLM now has a grant-time callback to revalidate locked items, if
6369 - Many MDS operations were reorganized to combat race conditions
6371 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6372 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6374 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6377 - Fully reactivate OST imports after reconnection (512, others)
6378 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6379 - More graceful error handling for truncating on dead OST (515)
6380 - Don't error out unless we're actually accessing dead stripes (474)
6381 - Fix garbage sizes when stripes are missing (410)
6382 - LRU counters were broken, causing constant lock purge (433, 432)
6383 - garbage on read from stripes with failed OSTs (441)
6384 - mark OSCs as active before reconnecting during recovery (438)
6385 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6386 - lfind did not preserve OST order in output (443)
6387 - symlinks cause hung clients, incorrect data (439)
6388 - stop dereferencing request after dropping refcount (457)
6389 - don't LASSERT(spin_is_locked) on non-SMP (455)
6390 - fixes for many rename() bugs
6391 - fstat didn't correctly synchronize attributes (399)
6392 - server must handle lock cancellation during blocking AST prep (487)
6393 - bulk descriptors were free()d too soon (511)
6394 - fix paths in lconf, which would load incorrect modules (451, 507)
6395 - fix confusing lconf 'host not found' error message (386)
6396 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6397 - fix race condition in mdc_blocking_ast() for inode access (526)
6398 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6399 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6400 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6401 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6402 - fix lstripe in several ways (kernel side) (527)
6403 - fix request leak in ldlm_cli_enqueue (262)
6404 - incorrect OSC was marked inactive after OST failure
6405 - call mds_fs_cleanup before unmounting filesystem (524)
6406 - fix races between taking ns_lock and ldlm_lock_change_resource
6407 - fix races updating LOV export open file list
6408 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6409 - fix recovery NULL deref in ldlm_cli_cancel_unused
6410 - fix some DLM races by using new hash table for lock handles (419)
6411 - permit the client to specify desired inodes, at replay
6412 - duplicate requests when we queue them for replay reintegration
6413 - fix last_rcvd offset calculation
6414 - sync after each recovered transaction, so we always make progress
6415 - never, not always, ERESTART requests without transnos
6416 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6418 - skip replay if the MDS says that the client is already connected
6419 - don't check for a recovery-enabled export to match lctl's UUID
6420 - don't INC_USE_COUNT for phantom exports
6421 - don't crash when cleaning up phantom exports (567)
6422 - don't double-finish or set replay data for errored mdc_open requests
6423 - abort requests when they time out, so we don't get old replies
6424 - send/receive replies for AST messages again
6425 - if the client says that it doesn't have the lock, cancel it on the
6427 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6428 mark it as destroyed, it will all work out when decref is called
6429 - fix module use counts (22, 581)
6431 - ASTs now expect a reply (server cancels lock on error reply)
6433 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6436 - fix many simultaneous client startup (392)
6437 - fix dentry->d_it clobbering
6438 - credentials weren't being shipped for readdir/getattr operations
6439 - remove invalid assertions triggered during some concurrent MD
6441 - proper Lustre versions added (336, 389)
6442 - fix memory leak for create error case (398)
6443 - fix LOV locking bug that would get cli/srv out of sync
6444 - fix echo client over LOV (409)
6445 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6446 - fix double-O_EXCL intent crash (424)
6447 - avoid sending multiple lock CANCELs (352)
6449 - MDS can do multi-client recovery (modulo bugs in new code)
6451 - many updates, edits, cleanups
6453 2002-11-18 Phil Schwan <phil@clusterfs.com>
6456 - fix null d_it dereference (346)
6457 - fix full OST/dbench hang (333)
6458 - fix permission problem with file removal (286)
6459 - fix removal of OSCs from LOV when they fail
6460 - fix NULL deref during bulk timeout (214)
6461 - fix problems related to multiple filesystems on one MDS (241)
6462 - fixed serious subtle metadata locking bugs
6463 - free locks on clients when inodes are removed due to memory
6465 - fix inode pointer in lock data (285)
6466 - partial support for multiple MDS on a single host (241)
6467 - data locks weren't cancelled at clear_inode time (290, 311)
6468 - intent locks could lead to unbounded lock growth (205)
6469 - added a maximum lock count, an LRU list, and a flusher
6470 - fix multiple rename (365)
6471 - properly abstracted the echo client
6472 - OSC locked 1 byte too many; fixed
6473 - rewrote brw callback code:
6474 - fixed recovery bugs related to LOVs (306)
6475 - fixed too-many-pages-in-one-write crash (191)
6476 - fixed (again) crash in sync_io_timeout (214)
6477 - probably fixed callback-related race (385)
6479 - Add capability to MDS protocol
6480 - LDLM cancellations and callbacks on different portals
6482 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6485 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6486 - defer open object destruction to close time (601981, 138)
6487 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6488 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6489 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6490 - don't call obd_setattr in ll_file_release for destroyed objects
6491 * protocol change to lustre_msg: move |version| and add |flags|
6492 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6493 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6494 * added replay of create, unlink, link and rename operations during
6495 MDS failover; recovery should be much more robust now
6496 * remove failed OSCs from LOVs (only lov_create uses this so far)
6497 * the lustre-HOWTO was brought (more) up to date (582544)
6499 2002-10-23 Phil Schwan <phil@clusterfs.com>
6502 - in-use dentries weren't being reused properly (617851)
6503 - prevent multiple LDLM setup (599178)
6504 - fix LOV size calculations for truncate (617853)
6505 - fix client handling of MDS intent errors (POSIX)
6506 - fix permission bug in lovstripe.c test (624321)
6507 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6508 - truncate past end of file could corrupt data
6509 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6510 - a race in recovery could return ETIMEDOUT to apps (623947)
6511 - building outside the source directory was fixed
6512 * the lustre-HOWTO was brought (more) up to date (582544)
6513 * major progress was made on recovery functionality
6515 2002-10-10 Phil Schwan <phil@clusterfs.com>
6518 - recovery deadlock fix
6519 - rm -rf causes LBUG fix (617817)
6520 - file open by multiple tasks fix (618962)
6521 - directory permissions bugs (602707 and 620007)
6522 - journal_stop fixed with locking (611313)
6523 - O_APPEND failures resolved (618273, perhaps 614459)
6524 - lconf PATH fix (619770)
6525 - IA64 build fix (621450)
6526 - RPC buffer sizes scale with amount of memory
6528 2002-10-01 Phil Schwan <phil@clusterfs.com>
6531 - locks would be cancelled without throwing away data pages,
6532 resulting in inconsistent data (605627)
6533 - inode attributes were not always being refreshed (605627, 612449)
6534 - lconf now continues to cleanup after lctl reports an error
6535 - MDS now enforces user permissions (602707)
6536 - lprocfs cleanup fixed, but not yet enabled (614157)
6537 - fixed infinite server hang, should a client not respond to an AST
6538 - avoid going into recovery if user calls readlink() with a buffer
6539 that's too small (613941)
6540 - AST RPCs no longer require replies (614867) -- this may be changed
6541 - don't crash server if client sends an IOV that's too big (611336)
6542 - fixed lock conversion deadlock (611892)
6543 - fixed the following of symlinks (614622)
6544 * recovery: the server can remove locks from a client that dies, other
6545 clients can make progress
6546 * more extN patch fixes
6547 * compile-time configurable ptlrpc buffer allocations
6549 - collaborative read cache document
6550 - Lustre Lite Performance CDR document-in-progress
6552 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6555 - fix typo in patch-2.4.18
6557 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6560 - clear ptlrpc request each time in handle_incoming_request()
6561 - unlink of files now destroys the object on the OST
6563 2002-09-19 Peter Braam <braam@clusterfs.com>
6565 * add hard link support
6566 * change obdfile creation method
6567 * kernel patch changed
6569 2002-09-19 Peter Braam <braam@clusterfs.com>
6572 - stack overflow bug in extN fixed
6574 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6576 * documentation updates
6577 - add man pages for config tools
6578 - update tests/README to describe testing with new config tools
6579 - finish metadata API descriptions
6580 * bug fixes and cleanups
6581 - statfs workaround for 16TB limit
6582 - LOV stripe allocation improved, can stripe on subset of OSTs
6583 - LOV file size/IO offset was wrong for files > 4GB in size
6584 - object EA data was being dropped, caused files to be unreadable
6585 - memory overflow with non-LOV OST caused memory corruption
6586 - fixed regression tests to work with new config tools, obdfilter
6587 - fixed bug when directory size became larger than 1 block
6588 - fixed bug (for single client case) when PWD was deleted
6589 - invalidate local directory pages when doing intent-based ops
6590 - avoid LDLM oops when lock callback contained bad data
6592 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6594 * documentation updates
6595 * bug fixes and cleanups
6596 - configuration tools
6599 - 64-bit compile warnings
6600 - 64-bit internal statfs data
6602 * test_brw on persistent OST devices
6604 * lprocfs (disabled)
6606 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6608 * documentation updates
6609 * bug fixes and cleanups
6610 * configuration tools
6612 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6615 * many small fixes to 0_5_4
6616 * io/network handling
6617 * thinkos in MDS operations
6619 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6622 * crucial basic fixes to 0.5.3
6623 * IOR, Iozone work over Elan
6626 2002-08-07 Phil Schwan <phil@clusterfs.com>
6627 * version 0_5_3, our first alpha
6628 * we use the new Portals iovs
6629 * documentation updates
6630 * bug fixes and cleanups
6631 * small changes in the DLM wire protocol
6633 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6634 * version 0_5_1 with some initial stability,
6635 * locking on MD and file I/O.
6636 * documentation updates
6637 * several bug fixes since 0.5.0
6638 * small changes in wire protocol
6640 2002-07-18 Phil Schwan <phil@clusterfs.com>
6642 * delivered as Lustre Light Alpha
6643 * fixed a crash after handling invalid MDS requests
6644 * fixed directory pages for architectures with non-4k pages sizes
6646 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6647 * release version v0_4_4
6648 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6650 * Fixes a number of interruption problems with OST operations.
6651 * Update documentation for portals header changes
6652 * Move all wire protocol structs/defines to lustre_idl.h
6653 * Fixes symlink length bug.
6654 * Add tcpdump to repository.
6656 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6657 * release version v0_4_3
6658 * Fixes statfs for inodes on extN.
6659 * Fixes bug in runtests which would delete /etc/hosts.
6660 * Use 64-bit object IDs wherever possible (not into VFS though)
6661 Remove ost_get_info, which is unused by lustre, and out of date.
6663 2002-07-03 Peter Braam <braam@clusterfs.com>
6664 * release version v0_4_2 Fixes a lookup error (type not passed)
6665 * move forward to head of Portals
6666 * move forward to latest Lustre kernel
6668 2002-06-25 Peter Braam <braam@clusterfs.com>
6669 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob