1 12-11-2017 Intel Corporation
3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.16.1.el6 (RHEL6.9)
11 3.10.0-693.5.2.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.90-92.45 (SLES12 SP2)
16 4.4.92-6.18 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.16.1.el6 (RHEL6.9)
24 3.10.0-693.5.2.el7 (RHEL7.4)
25 3.0.101-0.47.71 (SLES11 SP3)
26 3.0.101-107 (SLES11 SP4)
27 3.12.74-60.64.40 (SLES12 SP1)
28 4.4.90-92.45 (SLES12 SP2)
29 4.4.92-6.18 (SLES12 SP3)
31 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
32 * Recommended ZFS / SPL version: 0.7.3
33 * Tested with ZFS / SPL version: 0.7.3
34 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
35 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
36 stack overflow. For more information, please refer to bugzilla 17630.
37 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
38 the re-exporting nodes to be mounted with "32bitapi" mount option
40 --------------------------------------------------------------------------------
42 10-02-2017 Intel Corporation
44 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on patched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.26.1.el6 (RHEL6.7)
50 2.6.32-642.15.1.el6 (RHEL6.8)
51 2.6.32-696.10.3.el6 (RHEL6.9)
52 3.10.0-693.2.2.el7 (RHEL7.4)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-107 (SLES11 SP4)
55 3.12.74-60.64.40 (SLES12 SP1)
56 4.4.74-92.35 (SLES12 SP2)
57 vanilla linux 4.6.7 (ZFS only)
58 * Client known to build on unpatched kernels:
59 2.6.32-431.29.2.el6 (RHEL6.5)
60 2.6.32-504.30.3.el6 (RHEL6.6)
61 2.6.32-573.26.1.el6 (RHEL6.7)
62 2.6.32-642.15.1.el6 (RHEL6.8)
63 2.6.32-696.10.3.el6 (RHEL6.9)
64 3.10.0-693.2.2.el7 (RHEL7.4)
65 3.0.101-0.47.71 (SLES11 SP3)
66 3.0.101-107 (SLES11 SP4)
67 3.12.74-60.64.40 (SLES12 SP1)
68 4.4.74-92.35 (SLES12 SP2)
70 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
71 * Recommended ZFS / SPL version: 0.7.1
72 * Tested with ZFS / SPL version: 0.7.1
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
79 --------------------------------------------------------------------------------
81 2017-07-05 Intel Corporation
83 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
84 for currently supported client and server kernel versions.
85 * Server known to build on patched kernels:
86 2.6.32-431.29.2.el6 (RHEL6.5)
87 2.6.32-504.30.3.el6 (RHEL6.6)
88 2.6.32-573.26.1.el6 (RHEL6.7)
89 2.6.32-642.15.1.el6 (RHEL6.8)
90 2.6.32-696.3.1.el6 (RHEL6.9)
91 3.10.0-514.21.1.el7 (RHEL7.3)
92 3.0.101-0.47.71 (SLES11 SP3)
93 3.0.101-100 (SLES11 SP4)
94 3.12.74-60.64.40 (SLES12 SP1)
95 4.4.59-92.17 (SLES12 SP2)
96 vanilla linux 4.6.7 (ZFS only)
97 * Client known to build on unpatched kernels:
98 2.6.32-431.29.2.el6 (RHEL6.5)
99 2.6.32-504.30.3.el6 (RHEL6.6)
100 2.6.32-573.26.1.el6 (RHEL6.7)
101 2.6.32-642.15.1.el6 (RHEL6.8)
102 2.6.32-696.3.1.el6 (RHEL6.9)
103 3.10.0-514.21.1.el7 (RHEL7.3)
104 3.0.101-0.47.71 (SLES11 SP3)
105 3.0.101-100 (SLES11 SP4)
106 3.12.74-60.64.40 (SLES12 SP1)
107 4.4.59-92.17 (SLES12 SP2)
109 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
110 * Recommended ZFS / SPL version: 0.6.5.9
111 * Tested with ZFS / SPL version: 0.6.5.9
112 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
113 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
114 stack overflow. For more information, please refer to bugzilla 17630.
115 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
116 the re-exporting nodes to be mounted with "32bitapi" mount option
118 --------------------------------------------------------------------------------
120 12-07-2016 Intel Corporation
122 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
123 for currently supported client and server kernel versions.
124 * Server known to build on patched kernels:
125 2.6.32-431.29.2.el6 (RHEL6.5)
126 2.6.32-504.30.3.el6 (RHEL6.6)
127 2.6.32-573.26.1.el6 (RHEL6.7)
128 2.6.32-642.11.1.el6 (RHEL6.8)
129 3.10.0-514.el7 (RHEL7.3)
130 3.0.101-0.47.71 (SLES11 SP3)
131 3.0.101-80 (SLES11 SP4)
132 3.12.62-60.64.8 (SLES12 SP1)
133 vanilla linux 4.5.7 (ZFS only)
134 * Client known to build on unpatched 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)
144 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
145 * Recommended ZFS / SPL version: 0.6.5.8
146 * Tested with ZFS / SPL version: 0.6.5.8
147 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
148 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
149 stack overflow. For more information, please refer to bugzilla 17630.
150 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
151 the re-exporting nodes to be mounted with "32bitapi" mount option
153 --------------------------------------------------------------------------------
155 02-29-2016 Intel Corporation
157 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
158 for currently supported client and server kernel versions.
159 * Server known to build on patched kernels:
160 2.6.32-431.29.2.el6 (RHEL6.5)
161 2.6.32-504.30.3.el6 (RHEL6.6)
162 2.6.32-573.12.1.el6 (RHEL6.7)
163 3.10.0-327.3.1.el7 (RHEL7.2)
164 3.0.101-0.47.71 (SLES11 SP3)
165 3.0.101-68 (SLES11 SP4)
166 vanilla linux 4.2.1 (ZFS only)
167 * Client known to build on unpatched kernels:
168 2.6.32-431.29.2.el6 (RHEL6.5)
169 2.6.32-504.30.3.el6 (RHEL6.6)
170 2.6.32-573.12.1.el6 (RHEL6.7)
171 3.10.0-327.3.1.el7 (RHEL7.2)
172 3.0.101-0.47.71 (SLES11 SP3)
173 3.0.101-68 (SLES11 SP4)
176 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
177 * Recommended ZFS / SPL version: 0.6.4.2
178 * Tested with ZFS / SPL version: 0.6.4.2
179 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
180 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
181 stack overflow. For more information, please refer to bugzilla 17630.
182 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
183 the re-exporting nodes to be mounted with "32bitapi" mount option
185 --------------------------------------------------------------------------------
187 03-10-2015 Intel Corporation
189 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
190 for currently supported client and server kernel versions.
191 * Server known to build on patched kernels:
192 2.6.32-431.29.2.el6 (RHEL6.5)
193 2.6.32-504.8.1.el6 (RHEL6.6)
194 3.0.101-0.46 (SLES11 SP3)
195 * Client known to build on unpatched kernels:
196 2.6.32-431.29.2.el6 (RHEL6.5)
197 2.6.32-504.8.1.el6 (RHEL6.6)
198 3.10.0-123.20.1.el7 (RHEL7)
199 3.0.101-0.46 (SLES11 SP3)
200 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
201 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
202 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
203 stack overflow. For more information, please refer to bugzilla 17630.
204 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
205 the re-exporting nodes to be mounted with "32bitapi" mount option
207 Severity : enhancement
209 Description: control OST-index in IDIF via ROCOMPAT flag.
210 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
211 last_rcvd file. For new formatted OST device, it will be auto set;
212 for the case of upgrading from old OST device, you can enable it
213 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
214 enabled, for new created OST-object, its IDIF-in-LMA will contain
215 the OST-index; for the existing OST-object, the OSD will convert
216 old format IDIF as new format IDIF with OST-index stored in the
217 LMA EA when accessing such OST-object or via OI scrub. Once such
218 flag is enabled, it cannot be reverted back, so the system cannot
219 be downgraded to the orignal incompatible version.
221 --------------------------------------------------------------------------------
223 07-30-2014 Intel Corporation
225 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
226 for currently supported client and server kernel versions.
227 * Server known to build on patched kernels:
228 2.6.32-431.20.3.el6 (RHEL6.5)
229 3.0.101-0.31 (SLES11 SP3)
230 * Client known to build on unpatched kernels:
231 2.6.32-431.20.3.el6 (RHEL6.5)
232 3.0.101-0.31 (SLES11 SP3)
235 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
236 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
237 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
238 stack overflow. For more information, please refer to bugzilla 17630.
239 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
240 the re-exporting nodes to be mounted with "32bitapi" mount option
242 --------------------------------------------------------------------------------
244 10-31-2013 Intel Corporation
246 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
247 * Server builds for kernels:
248 2.6.32-358.18.1.el6 (RHEL6.4)
249 * Client builds for unpatched kernels:
250 2.6.32-358.18.1.el6 (RHEL6.4)
251 3.0.92-0.8 (SLES11 SP3)
253 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
254 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
255 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
256 stack overflow. For more information, please refer to bugzilla 17630.
257 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
258 the re-exporting nodes to be mounted with "32bitapi" mount option
260 --------------------------------------------------------------------------------
262 05-30-2013 Intel Corporation
264 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
265 * Server builds for kernels:
266 2.6.32-358.6.2.el6 (RHEL6.4)
267 * Client builds for unpatched kernels:
268 2.6.32-358.6.2.el6 (RHEL6.4)
269 3.0.74-0.6.6 (SLES11 SP2)
271 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
272 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
273 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
274 stack overflow. For more information, please refer to bugzilla 17630.
275 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
276 the re-exporting nodes to be mounted with "32bitapi" mount option
278 --------------------------------------------------------------------------------
280 9-30-2012 Whamcloud, Inc.
282 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
283 * Server support for kernels:
284 2.6.32-279.5.1.el6 (RHEL6)
285 * Client support for unpatched kernels:
286 2.6.32-279.5.1.el6 (RHEL6)
288 2.6.18-238.19.1.el5 (RHEL5)
289 2.6.16.60-0.69.1 (SLES10)
290 * Recommended e2fsprogs version: 1.41.90.wc4
291 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
292 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
293 stack overflow. For more information, please refer to bugzilla 17630.
294 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
295 the re-exporting nodes to be mounted with "32bitapi" mount option
297 Severity : enhancement
299 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
301 Severity : enhancement
303 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
305 Severity : enhancement
307 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
309 Severity : enhancement
311 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
313 Severity : enhancement
315 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
317 --------------------------------------------------------------------------------
319 04-30-2012 Whamcloud, Inc.
321 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
322 * Server support for kernels:
323 2.6.32-220.4.2.el6 (RHEL6)
324 * Client support for unpatched kernels:
325 2.6.32-220.4.2.el6 (RHEL6)
327 2.6.18-238.19.1.el5 (RHEL5)
328 2.6.16.60-0.69.1 (SLES10)
329 * Recommended e2fsprogs version: 1.41.90.wc4
330 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
331 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
332 stack overflow. For more information, please refer to bugzilla 17630.
333 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
334 the re-exporting nodes to be mounted with "32bitapi" mount option
336 Severity : enhancement
338 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
340 Severity : enhancement
342 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
344 Severity : enhancement
346 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
348 Severity : enhancement
350 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
352 --------------------------------------------------------------------------------
354 09-30-2011 Whamcloud, Inc.
356 * Server support for kernels:
357 2.6.18-238.19.1.el5 (RHEL5)
358 2.6.32-131.6.1.el6 (RHEL6)
359 * Client support for unpatched kernels:
360 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
361 2.6.18-238.19.1.el5 (RHEL5)
362 2.6.32-131.6.1.el6 (RHEL6)
363 2.6.29.4-167.fc11 (FC11)
364 2.6.16.60-0.69.1 (SLES10)
366 * Recommended e2fsprogs version: 1.41.90.wc3
367 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
368 Lustre filesystem with 4K stack may cause a stack overflow. For more
369 information, please refer to bugzilla 17630.
370 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
371 mounted with "32bitapi" mount option
373 Severity : enhancement
374 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
376 Severity : enhancement
377 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
379 Severity : enhancement
380 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
383 Frequency : only with SLES10
385 Description: Use OFED "KMP" provided by Novell
386 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
387 Lustre is now built against this package. That means you need to
388 install the ofed-kmp package from Novell for the patchless client.
389 Note that the ofed-kmp that Novell ships may not exactly match the
390 kernel version but should still be compatible.
392 Severity : enhancement
394 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
395 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
397 --------------------------------------------------------------------------------
399 2010-07-15 Oracle, Inc.
401 * Server support for kernels:
402 2.6.18-164.11.1.el5 (RHEL 5)
403 2.6.18-164.11.1.0.1.el5 (OEL 5)
404 * Client support for unpatched kernels:
405 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
406 2.6.18-164.11.1.el5 (RHEL 5),
407 2.6.18-164.11.1.0.1.el5 (OEL 5)
408 2.6.16.60-0.42.8 (SLES 10),
410 2.6.29.4-167.fc11 (FC11)
411 * Recommended e2fsprogs version: 1.41.10-sun2
413 Severity : enhancement
415 Description: Distribute lfs_migrate script with lustre
416 Details : The lfs_migrate mentioned in the manual is now included in the
417 lustre tree in lustre/scripts and will distributed in lustre RPMs.
419 Severity : enhancement
421 Description: Failover nid list added to import proc file.
423 Severity : enhancement
425 Description: "som_preview" mount option required to use SOM
426 Details : The clients will need to use som_preview mount option
430 Description: Lustre does not do 1MB IOs to HW RAID
431 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
433 Severity : enhancement
435 Description: lfs quota output is not very convenient for awk/sed-parsing
436 Details : Some positions in lfs quota output table could be empty or
437 non-empty which made it hard to parse it with scripts, now
438 a dash is put instead of space where there is not supposed
443 Description: Update to OFED 1.5.1
444 Details : OFED 1.5.1 will be used in environments that don't use
445 vendor supplied OFED.
449 Description: Remove lnet. prefix from lctl params display
450 Details : Also renamed lnet.memused to lnet_memused to
455 Description: Use conf_param -d to remove permanant settings.
459 Description: Add list_param -R
460 Details : lctl list_param now supports recursive mode.
463 Bugzilla : 21829, 22312
464 Description: Fix llobdstat.
465 Details : 1. Now updating the write_bytes counter
466 2. Port script fixes from 1.8.
470 Description: decrease the usage of memory on clients.
471 Details : 1. On clients, recycle dentries and inodes unused.
472 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
477 Description: llobdstat fix and enhancement
478 Details : add a counter to set a limit to how many samples will be returned
479 fix a wildcard in the path to limit to obdfilter stats only
483 Description: lfs check only allowed for root
484 Details : Code cleanups and added a regresion test of non-root use of
489 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
490 Details : Fix hr->hr_index overload.
494 Description: Fix FIEMAP functionality.
498 Description: FC11 patchless client support added
502 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
503 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
507 Description: New acc-sm script and various test-framework enhancments.
511 Description: Previously landed patch was reverted.
515 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
519 Description: Add ldiskfs maxdirsize mount option.
523 Description: Fix mkfs.lustre support for 16TB LUNs.
527 Description: Print mmp_check_interval
528 Details : print mmp_check_interval and make it possible to abort
529 mount operation in case it takes too long.
533 Description: Add cascading_rw to lustre mpi tests.
537 Description: Store SOM data in EA.
538 Details : Also several SOM related cleanups.
542 Description: Add refcount to md_open_data
546 Description: Use refcount on lov_request_set to prevent memory corruption.
550 Description: Statahead: use parent dir's i_mutex for synchronization.
554 Description: loadgen improvements.
558 Description: Enable peer health detection in ksocklnd and ko2iblnd.
562 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
566 Description: Fix race on obd_devcie:md_stats
570 Description: Fix SLES 11 client builds.
571 Details : Resolved uninitalized pointer gcc warnings.
575 Description: Allow non-root access for "lfs check".
576 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
580 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
582 Severity : enhancement
584 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
585 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
589 Description: Sanity test_27p fails.
590 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
591 Sanity tests 27 were cleaned up and improved.
595 Description: Unify req format on client/servers
596 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
597 buffers and swabbers approach..
602 Description: drop join file stuff.
604 Severity : enhancement
606 Description: Add 2.6.27 and SLES11 (patchless client) support
610 Description: Fix in ptlrpc_expire_one_request() to print the signed time
613 Severity : enhancement
615 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
618 Severity : enhancement
620 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
622 Severity : enhancement
624 Description: correctly handle big reply message.
625 Details : send LNet event if reply is bigger than buffer and adjust this buffer
630 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
632 Severity : enhancement
634 Description: Add support for OEL5.
638 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
640 Severity : enhancement
642 Description: "weak-modules" support
643 Details : Implement "weak-modules" support which enables kernel modules
644 to be used with any kernel that implements the same kABI. In
645 order to achieve this modules are now installed in
646 /lib/modules/$(uname -r)/updates/kernel on all distributions.
650 Description: client could not reconnect to OST because of an active request.
651 Details : The client now retries to reconnect to the same server, if a
652 connect request failed with EBUSY or -EAGAIN. Test case 35b
653 conf-sanity.sh has been created.
657 Description: Don't allow make backward step on assignin osc next id.
658 Details : race between allocation next id and ll_sync thread can be cause
659 of set wrong osc next id and can be kill valid ost objects.
663 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
667 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
671 Description: Error handling in osc_statfs_interpret() has been improved.
672 Details : Check in osc_statfs_interpret() for EBADR.
676 Description: Do not send statfs() requests to OSTs disabled by administrator.
677 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
679 Severity : enhancement
681 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
683 Severity : enhancement
685 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
687 Severity : enhancement
689 Description: Enhancement for lfs(1) command to use numeric uid/gid.
693 Description: Improved cleanup procedure for conf-sanity test_46a.
697 Description: Do not start bulk transfer, if client tried to reconnect.
698 Details : Do not start bulk transfer, if client tried to reconnect,
699 but failed due to active rpcs. Check for exp_abort_active_req
700 has been implemented.
704 Description: Prevent inconsistences between linux and lustre mount structures.
705 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
706 Make the sleep interruptible.
708 Severity : enhancement
710 Description: provide server to client comms path
711 Details : server to client communications path via new LDLM_SET_INFO rpc
715 Description: truncate starts GFP_FS allocation under transaction causing deadlock
716 Details : ldiskfs_truncate calls grab_cache_page which may start page
717 allocation under an open transaction. This may lead to
718 calling prune_icache with consequent lustre reentrance.
722 Description: send correctly lsm on open replay
723 Details : MDS is trust to LSM size on replay open, but client can set wrong size
728 Description: enable client interface failover
729 Details : When a child reconnects from another NID, properly update export
730 nid hash position and ldlm reverse import.
732 Severity : enhancement
734 Description: Communicate OST degraded/readonly state via statfs to MDS
735 Details : Flags in the statfs returned from OSTs indicate whether the
736 OST is in a degraded RAID state, or if the filesystem has
737 turned read-only after a filesystem error is detected.
742 Description: don't panic if EPROTO was hit when reading symlink
743 Details : correctly handling request reference in error cases.
745 Severity : enhancement
747 Description: Add LustreNetLink, a kernel-userspace communcation path.
749 Severity : enhancement
751 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
756 Description: access to llog context before init.
757 Details : move handling CATALOGS file at osc layer and forbid access to llog
761 Frequency : rare, only when client is writing to OST on local node
763 Description: Avoid deadlock for local client writes
764 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
765 memory freeing context. This allows OST threads to set the
766 PF_MEMALLOC flag on task structures in order to allocate memory
767 from reserved pools and complete IO.
768 Use GFP_HIGHUSER for OST allocations for non-local client writes,
769 so that the OST threads generate memory pressure and allow
770 inactive pages to be reclaimed.
772 Severity : enhancement
774 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
779 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
780 Details : move ldlm namespace creation in setup phase to avoid grab
781 _lprocfs_lock with cli_sem held.
785 Description: Temporarily disable grant shrink.
786 Details : Disable the feature for debugging.
789 Frequency : only when formatting test filesystems
791 Description: Unable to run several mkfs.lustre on loop devices at the same time
792 Details : mkfs.lustre returns error 256 on the concurrent loop devices
793 formatting. The solution is to proper handle the error.
795 Severity : enhancement
797 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
799 Severity : enhancement
801 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
806 Description: resolve race between obd_disconnect and class_disconnect_exports
807 Details : if obd_disconnect will be called to already disconnected export he
808 forget release one reference and osc module can't unloaded.
810 Severity : enhancement
812 Description: Allow tuning service thread via /proc
813 Details : For each service a new
814 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
815 created that can be used to set min/max thread counts, and get the
816 current number of running threads.
821 Descriptoin: don't return error if have partially created objects for file.
822 Details : lov_update_create_set uses set->set_success as index for created
823 objects, so if some requests will be failed, they will have hole
824 at end of array and qos_shrink_lsm for allocate correct lsm.
826 Severity : enhancement
828 Description: Update OFED support to 1.4.1
833 Description: Reduce small size read RPC
834 Details : Set read-ahead limit for every file and only do read-ahead when
835 available read-ahead pages are bigger than 1M to avoid small size
838 Severity : enhancement
840 Description: move AT tunable parameters for more consistent usage
841 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
843 Severity : enhancement
845 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
846 Details : allow skip disconnected ost for send statfs request and hide error
852 Description: fix racy locking of mballoc block bitmaps causing BUG
853 Details : The locking of the mballoc buddy bitmap and the in-memory
854 block bitmap was using two different spin locks in some
855 cases. This made it possible to incorrectly access the
856 mballoc bitmap while another process was modifying it,
857 causing a sanity assertion to fail. While no on-disk corruption
858 was reported, there was some risk of this happening.
861 Frequency : rare, on llog test 6
863 Descriptoin: don't allow connect to already connected import
864 Details : allowing connect to already connected import is hide connecting problem.
869 Descriptoin: allow kill process which wait statahead result
870 Details : for some reasons 'ls' can stick in waiting result from statahead,
871 in this case need way for kill this process.
873 Severity : enhancement
875 Description: Add state history info file, enhance import info file
876 Details : Track import connection state changes in a new osc/mdc proc file;
877 add overview-type data to the osc/mdc import proc file.
879 Severity : enhancement
881 Description: MDS create should not wait for statfs RPC while holding DLM lock.
883 Severity : enhancement
885 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
890 Descriptoin: OST grant too much space to client even there are not enough space.
891 Details : Client will shrink its grant cache to OST if there are no write
892 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
893 this grant cache if there are already not enough avaible space
894 (left_space < total_clients * 32M).
897 Frequency : start MDS on uncleanly shutdowned MDS device
899 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
900 Details : stay in waiting mds<>ost recovery finished produce random bugs
901 due race between two ll_sync thread for one lov target. send
902 ACTIVATE event only if connect realy finished and import have
906 Frequency : rare, connect and disconnect target at same time
908 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
909 Details : don't call obd_disconnect under lov_lock. this long time
910 operation and can block ptlrpcd which answer to connect request.
915 Descriptoin: don't lose wakeup for imp_recovery_waitq
916 Details : recover_import_no_retry or invalidate_import and import_close can
917 both sleep on imp_recovery_waitq, but we was send only one wakeup
921 Frequency : always with long access acl
923 Descriptoin: mds can't pack reply with long acl.
924 Details : mds don't control size of acl but they limited by reint/getattr
927 Severity : enhancement
929 Description: Update to SLES10 kernel-2.6.16.60-0.33.
931 Severity : enhancement
933 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
936 Frequency : start MDS on uncleanly shutdowned MDS device
938 Descriptoin: aborting recovery hang on MDS
939 Details : don't throttle destroy RPCs for the MDT.
942 Frequency : on remount
944 Description: external journal device not working after the remount
945 Details : clear dev_rdonly flag for external journal devices in
951 Description: shutdown vs evict race
952 Details : client_disconnect_export vs connect request race.
953 if client will evicted at this time - we start invalidate
954 thread without referece to import and import can be freed
958 Frequency : rare, need acl's on inode.
960 Description: client can't handle ost additional correctly
961 Details : if ost was added after client connected to mds client can have
962 hit lnet_try_match_md ... to big messages to wide striped files.
963 in this case need teach client to handle config events about add
964 lov target and update client max ea size at that event.
966 Severity : enhancement
968 Description: Changelogs
969 Details : Changelogs are a lightweight mechanism to track filesystem
970 metadata and namespace changes. The changelog is recorded
971 permanently on the MDTs, and is periodically "consumed" / purged
972 when records are no longer needed.
974 Severity : enhancement
976 Description: compact fld format with extents
977 Details : Store range of seq rather than every seq in FLD. Seq
978 controller update FLD rather than clients. In Case of CMD, mdt0
979 has FLD, all other metadata server act as non persistent proxy
980 for FLD queries and cache fld entries in fld cache.
985 Description: don't skip ost target if they assigned to file
986 Details : Drop slow OSCs if we can, but not for requested start idx.
987 This means "if OSC is slow and it is not the requested
988 start OST, then it can be skipped, otherwise skip it only
989 if it is inactive/recovering/out-of-space.
993 Description: more cleanup in mds_lov
994 Details : not send LOV EA under replay, we can't know about they size at this
995 time. Don't allow client connect to mds before any ost connected,
996 for avoid problems with LOV EA size and returning EIO to client.
998 Severity : enhancement
1000 Description: Interoperability at server side (Disk interoperability)
1002 Severity : enhancement
1004 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1006 Severity : enhancement
1008 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1010 Severity : enhancement
1012 Description: New client IO stack (CLIO).
1014 Severity : enhancement
1016 Description: Commit on sharing. Eliminate inter-client dependencies between
1017 uncommitted transactions by doing transaction commits.
1018 Thereby clients may recovery independently.
1021 Frequency : Create a symlink file with a very long name
1023 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1024 Details : If there is no extra space in the request for early cancels,
1025 ldlm_req_handles_avail() returns 0 instead of a negative value.
1027 Severity : enhancement
1029 Description: Add /proc entry for import status
1030 Details : The mdc, osc, and mgc import directories now have
1031 an import directory that contains useful import data for debugging
1032 connection problems.
1034 Severity : enhancement
1036 Description: Re-disable certain /proc logging
1037 Details : Enable and disable client's offset_stats, extents_stats and
1038 extents_stats_per_process stats logging on the fly.
1041 Frequency : Only on FC kernels 2.6.22+
1043 Description: oops in statahead
1044 Details : Do not drop reference count for the dentry from VFS when lookup,
1045 VFS will do that by itself.
1047 Severity : enhancement
1049 Description: Generic /proc file permissions
1050 Details : Set /Proc file permissions in a more generic way to enable non-
1051 root users operate on some /proc files.
1055 Description: Hitting mdc_commit_close() ASSERTION
1056 Details : Properly handle request reference release in
1057 ll_release_openhandle().
1061 Description: quota recovery deadlock during mds failover
1062 Details : This patch includes att18982, att18236, att18237 in bz14840.
1064 1. fix osts hang when mds does failover with quotaon
1065 2. prevent watchdog storm when osts threads wait for the
1070 Frequency : only patchless client
1071 Description: add workaround for race between add/remove dentry from hash
1073 Severity : enhancement
1075 Description: Allow OST glimpses to return PW locks
1079 Description: LBUG when llog conf file is full
1080 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1084 Description: Prevent import from entering FULL state when server in recovery
1088 Description: service mount cannot take device name with ":"
1089 Details : Only when device name contains ":/" will mount treat it as
1095 Description: replace ptlrpcd with the statahead thread to interpret the async
1096 statahead RPC callback
1100 Frequency : on recovery
1101 Description: I/O failures after umount during fail back
1102 Details : if client reconnected to restarted server we need join to recovery
1103 instead of find server handler is changed and process self
1104 eviction with cancel all locks.
1106 Severity : enhancement
1108 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1110 Severity : enhancement
1112 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1114 Severity : enhancement
1116 Description: Upcall on Lustre log has been dumped
1117 Details : Allow for a user mode script to be called once a Lustre log has
1118 been dumped. It passes the filename of the dumped log to the
1119 script, the location of the script can be specified via
1120 /proc/sys/lnet/debug_log_upcall.
1125 Description: avoid idr_remove called for id which is not allocated.
1126 Details : Move assigment s_dev for clustered nfs to end of initialization,
1127 to avoid problem with error handling.
1132 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1133 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1134 into CONN_USED_HASH and this prodice warning when put connection
1135 again in unused hash.
1138 Severity : enhancement
1140 Description: File striping can now be set to use an arbitrary pool of OSTs.
1142 Severity : enhancement
1144 Description: Export bytes_read/bytes_write count on OSC/OST.
1148 Description: Early reply size mismatch, MGC loses connection
1149 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1150 the connect flags are properly negotiated.
1155 Description: quotas are not honored with O_DIRECT
1156 Details : all writes with the flag O_DIRECT will use grants which leads to
1157 this problem. Now using OBD_BRW_SYNC to guard this.
1161 Description: add quota statistics
1162 Details : 1. sort out quota proc entries and proc code.
1163 2. add quota statistics
1165 Severity : enhancement
1167 Description: enable quota support for HEAD.
1171 Description: Properly propagate oinfo flags from lov to osc for statfs
1172 Details : restore missing copy oi_flags to lov requests.
1174 Severity : enhancement
1176 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1177 : l_getgroups(8), lst(8), routerstat(8)
1178 Details : included man pages for llobdstat(8), llstat(8),
1179 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1181 Severity : enhancement
1183 Description: configure's --enable-quota should check the
1184 : kernel .config for CONFIG_QUOTA
1185 Details : configure is terminated if --enable-quota is passed but
1186 : no quota support is in kernel
1190 Description: Remove portals compatibility
1191 Details : Remove portals compatibility, not interoperable with releases
1196 Description: Resolve device initialization race
1197 Details : Prevent proc handler from accessing devices added to the
1198 obd_devs array but yet be intialized.
1200 Severity : enhancement
1202 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1204 Severity : enhancement
1206 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1211 Description: Using wrong pointer in osc_brw_prep_request
1212 Details : Access to array[-1] can produce panic if kernel compiled with
1213 CONFIG_PAGE_ALLOC enabled
1217 Description: Client runs out of low memory
1218 Details : Consider only lowmem when counting initial number of llap pages
1222 Description: *optional* service tags registration
1223 Details : if the "service tags" package is installed on a Lustre node
1224 When the filesystem is mounted, a local-node service tag will
1225 be created. See http://inventory.sun.com/ for more information
1226 about the Service Tags asset management system.
1230 Description: Kernel BUG tries to release flock
1231 Details : Lustre does not destroy flock lock before last reference goes
1232 away. So always drop flock locks when client is evicted and
1233 perform unlock regardless of successfulness of speaking to MDS.
1237 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1241 Description: sanity test 65a fails if stripecount of -1 is set
1242 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1247 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1248 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1251 Severity : enhancement
1253 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1257 Description: do not process already freed flock
1258 Details : flock can possibly be freed by another thread before it reaches
1259 to ldlm_flock_completion_ast.
1263 Description: LBUG during stress test
1264 Details : Need properly lock accesses the flock deadlock detection list.
1268 Description: oops in page fault handler
1269 Details : kernel page fault handler can return two special 'pages' in error
1270 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1274 Description: timeout with invalidate import.
1275 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1276 should be handled by ptlrpcd. This produce long age waiting and
1277 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1279 Severity : enhancement
1281 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1285 Description: enable MGS and MDT services start separately
1286 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1287 a co-located MGS without starting the MGS, which is a complement
1288 to 'nosvc' mount option.
1292 Frequency : after recovery
1293 Description: precreate to many object's after del orphan.
1294 Details : del orphan st in oscc last_id == next_id and this triger growing
1295 count of precreated objects. Set flag LOW to skip increase count
1296 of precreated objects.
1300 Frequency : rare, on clear nid stats
1301 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1302 Details : when clean nid stats sometimes try destroy live entry,
1303 and this produce panic in free.
1307 Description: Stack overflow during MDS log replay
1308 ease stack pressure by using a thread dealing llog_process.
1312 Description: wait until IO finished before start new when do lock cancel.
1313 Details : VM protocol want old IO finished before start new, in this case
1314 need wait until PG_writeback is cleared until check dirty flag and
1315 call writepages in lock cancel callback.
1317 Severity : enhancement
1319 Description: using special macro for print time and cleanup in includes.
1323 Description: mds_mfd_close() ASSERTION(rc == 0)
1324 Details : In mds_mfd_close(), we need protect inode's writecount change
1325 within its orphan write semaphore to prevent possible races.
1329 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1333 Frequency : rare, on shutdown ost
1334 Description: don't hit live lock with umount ost.
1335 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1336 use shrink_dcache_sb instead.
1340 Description: don't panic with use echo client
1341 Details : echo client pass NULL as client nid pointer and this produce null
1342 pointer dereference.
1346 Description: fix build on ppc32
1347 Details : compile code with -m64 flag produce wrong object file for ppc32.
1351 Description: add message levels for liblustreapi
1355 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1356 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1357 will result in an -EAGAIN and in the exit call path we call
1358 alloc_rr() which will with increasing aggressiveness attempt to
1359 aquire precreated objects on the minimum number of required OSCs.
1363 Description: Use old size assignment to avoid deadlock
1364 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1365 the scheduling while holding a spinlock. We do not need locking
1366 for size in ll_update_inode() because size is only updated from
1367 the MDS for directories or files without objects, so there is no
1368 other place to do the update, and concurrent access to such inodes
1369 are protected by the inode lock.
1373 Description: resolve "_IOWR redefined" build error on SLES10
1377 Description: dump the memory debugging after all modules are unloaded to
1378 suppress false negative in conf_sanity test 39
1380 Severity : enhancement
1382 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1387 Description: fix SLES kernel versioning
1388 Details : the kernel version for our SLES 10 kernel did not include a "-"
1389 before the "smp" at the end. while this was not a problem in
1390 general, it did mean that software trying to use the kernel
1391 version to try to detect a vendor specific kernel would fail.
1392 this was most evident by the OFED build scripts.
1396 Description: Don't update lov_desc members until making sure they are valid
1397 Details : When updating lov_desc members via proc fs, need fix their
1398 validities before doing the real update.
1402 Description: don't put request into delay list while invalidate in flight.
1403 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1404 import in flight. this produce timeout for invalidate and sometimes
1405 can cause stale data.
1409 Frequency : on ppc only
1410 Description: not convert ost objects for directory because it's not exist.
1411 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1415 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1417 Severity : enhancement
1419 Description: Update to support 2.6.22.14 vanilla kernel.
1423 Frequency : rare, on recovery
1424 Description: read procfs can produce deadlock in some situation
1425 Details : Holding lprocfs lock which send rpc can produce block for destroy
1426 obd objects and this also block reconnect with -EALREADY. This
1427 isn't fix all lprocfs bugs - but make it rare.
1429 Severity : enhancement
1431 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1434 Frequency : frequent on X2 node
1436 Description: mdc_set_open_replay_data LBUG
1437 Details : Set replay data for requests that are eligible for replay.
1441 Description: lustre_mgs: operation 101 on unconnected MGS
1442 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1443 MGC, and late on MGC cannot successfully connect to MGS and a lot
1444 of the error messages complaining that MGS is not connected.
1448 Frequency : on network error
1449 Description: panic with double free request if network error
1450 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1451 but it's true only for synchronus enqueue, for async enqueue
1452 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1455 Severity : enhancement
1457 Description: client-side metadata stat-ahead during readdir(directory readahead)
1458 Details : perform client-side metadata stat-ahead when the client detects
1459 readdir and sequential stat of dir entries therein
1462 Frequency : on start mds
1464 Description: Implement get_info(last_id) in obdfilter.
1467 Frequency : occasional
1469 Description: Correctly check stale fid, not start epoch if ost not support SOM
1470 Details : open with flag O_CREATE need set old fid in op_fid3 because
1471 op_fid2 was overwritten with new generated fid, but mds can answer
1472 with one of these two fids and both is not stale. Setattr
1473 incorrectly started an epoch and assume will be called
1474 done_writing, but without SOM done_writing ever being called.
1477 Frequency : rare, depends on device drivers and load
1479 Description: MDS or OSS nodes crash due to stack overflow
1480 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1481 In some cases, in conjunction with device drivers that use a lot
1482 of stack the MDS (or possibly OSS) service threads could overflow
1483 the stack. One change which was identified to consume additional
1484 stack has been reworked to avoid the extra stack usage.
1487 Frequency : occasional
1489 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1490 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1491 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1492 oscc_has_objects and see OSC has no objests and call
1493 oscc_internal_create to resend create request.
1495 Severity : enhancement
1497 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1499 Severity : enhancement
1501 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1504 Frequency : very rare
1506 Description: Fix replay if there is an un-replied request and open
1507 Details : In some cases, older replay request will revert the
1508 mcd->mcd_last_xid on MDS which is used to record the client's
1509 latest sent request.
1511 Severity : enhancement
1513 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1515 Severity : enhancement
1517 Description: Add rhel5 support to HEAD.
1519 Serverity : enhancement
1521 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1526 Description: Don't allow skipping OSTs if index has been specified.
1527 Details : Don't allow skipping OSTs if index has been specified, make
1528 locking in internal create lots better.
1532 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1533 Details : access to bitfield in structure is always rounded to long
1534 and this produce problem with not atomic change any bit.
1538 Description: Lustre make rpms failed.
1539 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1540 builds Lustre rpms from tarball.
1543 Frequency : rare on shutdown ost
1545 Description: If llog cancel was not send before clean_exports phase, this can
1546 produce deadlock in llog code.
1547 Details : If llog thread has last reference to obd and call class_import_put
1548 this produce deadlock because llog_cleanup_commit_master wait when
1549 last llog_commit_thread exited, but this never success because was
1550 called from llog_commit_thread.
1554 Description: allow userland application know is lost one of stripes.
1555 Details : fill lvb_blocks with error code on ost and return it to
1556 application if error flag found.
1560 Description: NULL lov_tgts causing MDS oops
1561 Details : more safe checks for NULL lov_tgts for avoid oops.
1563 Severity : enhancement
1565 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1569 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1570 Details : If an OST is down the MDS will hang indefinitely in
1571 obd_statfs_rqset() waiting for the statfs data. While for
1572 MDS QOS usage of statfs, it should not stuck in waiting.
1574 Severity : enhancement
1576 Description: remote_acl support
1577 Details : Support ACL-based permission check for remote user.
1578 Support setfacl/getfacl for remote user with the utils
1579 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1580 the system "{s,g}etfacl" utils.
1582 Severity : enhancement
1584 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1586 Severity : enhancement
1588 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1592 Description: make mgs_setparam() handle fsname containing dash
1593 Details : fsname containing a dash does not work with lctl conf_param
1595 Severity : enhancement
1597 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1599 Severity : enhancement
1601 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1605 Frequency : rare, at shutdown
1606 Description: access already free / zero obd_namespace.
1607 Details : if client_disconnect_export was called without force flag set,
1608 and exist connect request in flight, this can produce access to
1609 NULL pointer (or already free pointer) when connect_interpret
1610 store ocd flags in obd_namespace.
1614 Frequency : only at startup
1615 Description: not alloc memory with spinlock held.
1616 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1617 if any spinlock held.
1619 Severity : enhancement
1621 Description: make lustre randomly fail allocating memory
1622 Details : Make lustre randomly failed allocating memory for testing purpose.
1624 Severity : enhancement
1626 Description: lost problems with lov objid file
1627 Details : Fixes some scability and access to not inited memory problems
1628 in work with lov objdid file.
1633 Description: lfs find does not continue on file error
1634 Details : Continue other files processing when a file/dir is absent.
1638 Description: Inconsistent usage of lustre_pack_reply()
1639 Details : Standardize the usage of lustre_pack_reply() such that it
1640 always generate a CERROR on failure.
1643 Frequency : occasional
1645 Description: MDS hang or stay in waiting lock
1646 Details : If client receive lock with CBPENDING flag ldlm need send lock
1647 cancel as separate rpc, to avoid situation when cancel request
1648 can't processed due all i/o threads stay in wait lock.
1652 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1656 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1660 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1664 Description: Update to RHEL4 latest kernel.
1666 Severity : enhancement
1668 Description: Build SLES10 patchless client fails
1669 Details : The configure was broken by run ./configure with
1670 --with-linux-obj=.... argument for patchless client. When the
1671 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1672 can't search header adequately. Use absolute path such as
1673 -I($LINUX)/include instead.
1677 Description: interrupt oig_wait produce painc on resend.
1678 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1679 private set, and this produce situation when rq_ptlrpcd_data not
1680 copyed to new allocated request and triggered LBUG on assert
1681 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1682 ptlrpcd set if request is changed and can be safety changed to use
1687 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1688 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1689 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1690 these macros and replace them with their difinitions instead.
1694 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1695 Details : Update which_patch & target file for SLES10 latest kernel.
1698 Description: Remove client patches from SLES 10 kernel.
1699 Details : This causes SLES 10 clients to behave as patchless clients
1700 even on a Lustre-patched (server) kernel.
1702 Severity : enhancement
1704 Description: self-adjustable client's lru lists
1705 Details : use adaptive algorithm for managing client cached locks lru
1706 lists according to current server load, other client's work
1707 pattern, memory activities, etc. Both, server and client
1708 side namespaces provide number of proc tunables for controlling
1711 Severity : enhancement
1713 Description: light-weight GSS support
1714 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1715 and reduce performance overhead.
1717 Severity : enhancement
1719 Description: Linux keyring support
1720 Details : Support using service of Linux keyring for Lustre GSS internal
1721 context refresh/cache mechanism.
1725 Description: Fix errors in lfs documentation
1726 Details : Fixes man pages
1730 Description: don't use GFP_* in generic Lustre code.
1731 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1735 Description: obdclass is limited by single OBD_ALLOC(idarray)
1736 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1740 Description: warnings with build patchless client with vanila 2.6.19 and up
1741 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1742 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1746 Frequency : Only for SLES
1747 Description: sanity_quota fail test_1
1748 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1749 file, which makes TSTID[2] inunique.
1753 Frequency : Only for SLES9
1754 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1755 Details : Update target/ChangeLog/which_patch .
1759 Frequency : Only for test_10 in sanity-quota.sh
1760 Description: a bug in quota test code
1761 Details : checking a test flag in a wrong place
1765 Frequency : Only for test_13 in sanity-quota.sh
1766 Description: a bug in quota test script
1767 Details : shouldn't check overfull allocated quota
1771 Description: jbd statistics
1772 Details : Port older jbd statistics patch for sles10
1774 Severity : enhancement
1776 Description: Add printing inode info into message about error in writepage.
1781 Description: reply_lock_interpret crash due to race with it and lock cancel.
1782 Details : Do not replay locks that are being cancelled. Do not reference
1783 locks by their address during replay, just by their handle.
1787 Frequency : When flocks are used.
1788 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1789 Details : Flock locks might destroy just granted lock if it could be merged
1790 with another existing flock, this is done in completion handler,
1791 so teach ldlm_cli_enquque_fini that this is a valid case for
1797 Description: Oops in read and write path when failing to allocate lock.
1798 Details : Check if lock allocation failed and return error back.
1802 Description: lstripe command fails for valid OST index
1803 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1804 instead of lov->desc.ld_active_tgt_count.
1808 Description: sanity.sh failed test 103
1809 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1814 Description: df doesn't work properly if diskfs blocksize != 4K
1815 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1819 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1820 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1822 Severity : enhancement
1824 Description: Adaptive timeouts
1825 Details : RPC timeouts adapt to changing server load and network
1826 conditions to reduce resend attempts and improve recovery time.
1830 Description: llapi_file_create() does not allow some changes
1831 Details : add llapi_file_open() that allows specifying the mode and
1832 open flags, and also returns an open file handle.
1836 Description: merge and cleanup kernel patches.
1837 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1841 Description: Add journal checksum support.(Kernel part)
1842 Details : The journal checksum feature adds two new flags i.e
1843 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1844 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1845 indicates that the commit block contains the checksum for
1846 the blocks described by the descriptor blocks. Now commit
1847 record can be sent to disk without waiting for descriptor
1848 blocks to be written to disk. This behavior is controlled
1849 using JBD2_FEATURE_ASYNC_COMMIT flag.
1853 Description: OSS needs mutliple precreate threads
1854 Details : Add ability to start more than one create thread per OSS.
1858 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1859 Details : execute lfs setstripe on client
1863 Description: mds_obd_create error creating tmp object
1864 Details : When the user sets quota on root, llog will be affected and can't
1865 create files and write files.
1868 Frequency : Always on ia64 patchless client, and possibly others.
1870 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1871 Details : This allows the patchless client to be loaded on architectures
1872 without this export.
1876 Description: RedHat Update kernel for RHEL5
1877 Details : Add the kernel config file for RHEL5.
1881 Description: RedHat Update kernel for RHEL5
1882 Details : Modify the kernel config file more closer RHEL5.
1886 Description: Build failure against Centos5 (RHEL5)
1887 Details : Define PAGE_SIZE when it isn't present.
1891 Description: test_7 on sanity-quota.sh can't be used on separate machines
1892 Details : add facet to handle it
1896 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1897 Details : with NFS, the anon dentry's parent was set to itself in
1898 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1899 corresponding dentry other than use rec->ur_name.
1901 Severity : enhancement
1903 Description: lfs setstripe enhancement
1904 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1908 Description: enable data checksumming by default
1909 Details : enable checksum by default, allow --disable-checksum
1910 configure option and "-o nochecksum" mount option. Checksums
1911 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1915 Description: lustre support for RHEL5
1916 Details : Add support for RHEL5.
1920 Description: Client eviction due to failover config
1921 Details : after a connection loss, the lustre client should attempt to
1922 reconnect to the last active server first before trying the
1923 other potential connections.
1927 Description: when mds and osts use different quota unit(32bit and 64bit),
1928 quota will be released repeatly.
1929 Details : void sending multiple quota reqs to mds, which will keep the
1930 status between the reqs.
1934 Description: osts not allocated evenly to files
1935 Details : change the condition to increase offset_idx
1938 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1940 Description: Data corruption for OSTs that are formatted larger than 2TB
1942 Details : When generating the bio request for lustre file writes the
1943 sector number would overflow a temporary variable before being
1944 used for the IO. The data reads correctly from Lustre (which
1945 will overflow in a similar manner) but other file data or
1946 filesystem metadata may be corrupted in some cases.
1950 Description: Tune the kernel for good SCSI performance.
1951 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1952 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1957 Description: rewrite ext2-derived code in obdclass/uuid.c
1958 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1959 from scratch preserving functionality.
1963 Description: "lfs find -obd UUID" prints directories
1964 Details : "lfs find -obd UUID" will return all directory names instead
1965 of just file names. It is incorrect because the directories
1966 do not reside on the OSTs.
1970 Description: use i_size_read and i_size_write in 2.6 port
1971 Details : replace inode->i_size access with i_size_read/write()
1974 Frequency : when removing large files
1976 Description: scheduling issue during removal of large Lustre files
1977 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1978 It causes scheduling issues when removing large files (17TB in the
1982 Frequency : only with liblustre clients on XT3
1984 Description: evictions taking too long
1985 Details : allow llrd to evict clients directly on OSTs
1988 Frequency : only on ppc
1990 Description: /proc/fs/lustre/devices broken on ppc
1991 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1992 We should make sure the type of 'index' is loff_t and then cast
1993 explicitly as needed below. Do not assign an explicitly cast
1997 Frequency : only for rhel5
1999 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2000 Details : Modify the target file & which_kernel.
2002 Severity : enhancement
2004 Description: omit set fsid for export NFS
2005 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2006 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2010 Frequency : Always, for kernels after 2.6.16
2011 Description: Fix warning idr_remove called for id=.. which is not allocated.
2012 Details : Last kernels save old s_dev before kill super and not allow
2013 to restore from callback - restore it before call kill_anon_super.
2017 Description: buffer overruns could theoretically occur
2018 Details : llapi_semantic_traverse() modifies the "path" argument by
2019 appending values to the end of the origin string, and a
2020 overrun may occur. Adding buffer overrun check in liblustreapi.
2024 Description: Fix error on 'ls .' at the top of the Lustre mount.
2025 Details : Don't revalidate dentry if it is a root dentry.
2029 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2030 Details : Modify vm-tunables-rhel4.patch.
2034 Description: Kernel config for 2.6.18-vanilla.
2035 Details : Modify targets/2.6-vanilla.target.in.
2036 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2037 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2038 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2039 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2043 Description: improve handling recoverable errors
2044 Details : if request processig with error which can be recoverable on server
2045 request should be resend, otherwise page released from cache and
2050 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2051 Details : Modify target file & which_patch.
2052 A flaw was found in the IA32 system call emulation provided
2053 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2054 value could be stored in the %RAX register, which could trigger an
2055 out-of-bounds system call table access. An untrusted local user
2056 could exploit this flaw to run code in the kernel
2057 (ie a root privilege escalation). (CVE-2007-4573).
2061 Description: O_DIRECT bypasses client statistics.
2062 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2063 read_ahead_stats were updated. copied the stats section from
2064 osc_send_oap_rpc() into async_internal().
2068 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2072 Description: change order of libsysio includes
2073 Details : '#include sysio.h' should always come before '#include xtio.h'
2077 Description: handle "serious error: objid * already exists" more gracefully
2078 Details : If LAST_ID value on disk is smaller than the objects existing in
2079 the O/0/d* directories, it indicates disk corruption and causes an
2080 LBUG(). If the object is 0-length, then we should use the existing
2081 object. This will help to avoid a full fsck in most cases.
2083 Severity : enhancement
2085 Description: adapt the lustre_config script to support the upgrade case
2086 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2087 server targets to 1.6 in parallel.
2091 Description: To avoid grant space > avaible space when the disk is almost
2092 full. Without this patch you might see the error "grant XXXX >
2093 available" or some LBUG about grant, when the disk is almost
2095 Details : In filter_check_grant, for non_grant cache write, we should
2096 check the left space by if (*left > ungranted + bytes), instead
2097 of (*left > ungranted), because only we are sure the left space
2098 is enough for another "bytes", then the ungrant space should be
2099 increase. In client, we should update cl_avail_grant only there
2100 is OBD_MD_FLGRANT in the reply.
2104 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2105 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2106 described in CVE-2007-4573. This problem could allow untrusted
2107 local users to gain root access.
2110 Frequency : when using O_DIRECT and quotas
2112 Description: Incorrect file ownership on O_DIRECT output files
2113 Details : block usage reported by 'lfs quota' does not take into account
2114 files that have been written with O_DIRECT.
2119 Description: touch file failed when fs is not full
2120 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2121 otherwise we can get ENOSP while fs is not full.
2125 Description: parallel lock callbacks
2126 Details : Instead of sending blocking and completion callbacks as separated
2127 requests, adding them to a set and sending in parallel.
2130 Frequency : only for Cray XT3
2131 Bugzilla : 12829/13455
2132 Description: Changing primary group doesn't change the group lustre assigns to
2134 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2135 overridden with the primary group provided by the group upcall,
2136 whereas the supplied fsgid can be trusted if it is in the list of
2137 supplementary groups returned by the group upcall.
2139 Severity : enhancement
2141 Description: Allow masking D_WARNING, D_ERROR messages from console
2142 Details : Console messages can now be disabled via lnet.printk.
2146 Description: User code with malformed file open parameter crashes client node
2147 Details : Before packing join_file req, all the related reference should be
2148 checked carefully in case some malformed flags cause fake
2149 join_file req on client.
2153 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2154 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2155 mode is equal to the granted mode. However, it should also check
2156 that LDLM_FL_CP_REQD is not set.
2160 Description: Heavy nfs access might result in deadlocks
2161 Details : After ELC code landed, it is now improper to enqueue any mds
2162 locks under och_sem, because enqueue might want to decide to
2163 cancel open locks for same inode we are holding och_sem for.
2167 Description: Client eviction while running blogbench
2168 Details : A lot of unlink operations with concurrent I/O can lead to a
2169 deadlock causing evictions. To address the problem, the number of
2170 oustanding OST_DESTROY requests is now throttled to
2171 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2172 callbacks are processed in priority.
2176 Description: enable ACLs on MDS by default
2177 Details : ACLs must be enabled on MDS by default.
2180 Frequency : PPC/PPC64 only
2182 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2183 Details : fix duplicated definitions between asm-ppc64/types.h and
2184 lustre_types.h on PPC.
2187 Frequency : PPC/PPC64 only
2189 Description: asm-ppc/segment.h does not exist
2190 Details : fix compile issue on PPC.
2194 Description: better handle error messages in extents code
2197 Frequency : RHEL4 only
2199 Description: mkfs is very slow on IA64/RHEL4
2200 Details : A performance regression has been discovered in the MPT Fusion
2201 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2202 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2203 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2205 Severity : enhancement
2207 Description: SNMP support enhancement
2208 Details : Adding total number of sampled request for an MDS node in snmp
2211 Severity : enhancement
2213 Description: Optimize ldlm waiting list processing for PR extent locks
2214 Details : When processing waiting list for read extent lock and meeting read
2215 lock that is same or wider to it that is not contended, skip
2216 processing rest of the list and immediatelly return current
2217 status of conflictness, since we are guaranteed there are no
2218 conflicting locks in the rest of the list.
2222 Description: Time out and refuse to reconnect
2223 Details : When the failover node is the primary node, it is possible
2224 to have two identical connections in imp_conn_list. We must
2225 compare not conn's pointers but NIDs, otherwise we can defeat
2226 connection throttling.
2230 Description: port llog fixes from b1_6 into HEAD
2231 Details : Port llog reference couting and some llog cleanups from b1_6
2232 (bug 10800) into HEAD, for protect from panic and access to already
2233 free llog structures.
2237 Description: Detect stride IO mode in read-ahead
2238 Details : When a client does stride read, read-ahead should detect that and
2239 read-ahead pages according to the detected stride pattern.
2243 Description: data checksumming impacts single node performance
2244 Details : add support for several checksum algorithm. Currently, only CRC32
2245 and Adler-32 are supported. The checksum type can be changed on
2246 the fly via /proc/fs/lustre/osc/*/checksum_type.
2250 Description: use adler32 for page checksums
2251 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2256 Description: build for x2 fails
2257 Details : fix compile issue on Cray systems.
2261 Description: Properly match for duplicate locks
2262 Details : Due to different lock order from skiplists code, we need to
2263 traverse entire list for now
2266 Frequency : only on PPC/SLES10
2268 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2269 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2270 be defined. Add a hack in mkfs_lustre.c to work around this compile
2275 Description: LASSERT on MDS when client holding flock lock dies
2276 Details : ldlm pool logic depends on number of granted locks equal to
2277 number of released locks which is not true for flock locks, so
2278 just exclude such locks from consideration.
2282 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2283 Details : Use fsfilt_sync() for both the whole filesystem sync and
2284 individual file sync to eliminate dangerous inode locking
2285 with I_LOCK that can lead to a deadlock.
2289 Description: performance in 1.6.3
2290 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2291 sure that 1MB requests are not fragmented by the block layer.
2293 Severity : enhancement
2295 Description: organize the server-side client stats on per-nid basis
2296 Details : Change the structure of stats under obdfilter and mds to
2304 The "uuid"s file would list the uuids of _active_ exports.
2305 And the clear entry is to clear all stats and stale nids.
2307 Severity : enhancement
2309 Description: eliminate client locks in face of contention
2310 Details : file contention detection and lockless i/o implementation
2311 for contended files.
2315 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2318 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2320 Description: MDS may lose file striping (and hence file data) in some cases
2321 Details : If there are additional extended attributes stored on the MDS,
2322 in particular ACLs, SELinux, or user attributes (if user_xattr
2323 is specified for the client mount options) then there is a risk
2324 of attribute loss. Additionally, the Lustre file striping
2325 needs to be larger than default (e.g. striped over all OSTs),
2326 and an additional attribute must be stored initially in the
2327 inode and then increase in size enough to be moved to the
2328 external attribute block (e.g. ACL growing in size) for file
2333 Description: skiplist implementation simplification
2334 Details : skiplists are used to group compatible locks on granted list
2335 that was implemented as tracking first and last lock of each
2336 lock group the patch changes that to using doubly linked lists
2340 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2341 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2342 because LASSERT on both the data supplied by a client, and the
2343 data on disk is dangerous and incorrect.
2345 Severity : enhancement
2347 Description: Slow truncate/writes to huge files at high offsets.
2348 Details : Directly associate cached pages to lock that protect those pages,
2349 this allows us to quickly find what pages to write and remove
2350 once lock callback is received.
2354 Description: more ldlm soft lockups
2355 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2356 starve other threads from the resource lock for a long time in
2357 case of long waiting queue, so change the debug level from
2358 D_OTHER to the less frequently used D_INFO.
2360 Severity : enhancement
2362 Description: add -gid, -group, -uid, -user options to lfs find
2366 Description: Hung threads in invalidate_inode_pages2_range
2367 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2368 once one is completed. As a result, some RPCs are stuck in the
2369 queue and are never sent.
2373 Description: filter threads hungs on waiting journal commit
2374 Details : Cleanup filter group llog code, then only filter group llog will
2375 be only created in the MDS/OST syncing process.
2379 Description: Procfs and llog threads access destoryed import sometimes.
2380 Details : Sync the import destoryed process with procfs and llog threads by
2381 the import refcount and semaphore.
2383 Severity : enhancement
2385 Description: openlock cache of b1_6 port to HEAD
2390 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2391 Details : If insertion of an extent fails, then discard the inode
2392 preallocation and free data blocks else it can lead to duplicate
2397 Description: don't always update ctime in ext3_xattr_set_handle()
2398 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2399 In some cases the ctime should not be updated, for example for
2400 2.0->1.8 compatibility it is necessary to delete an xattr and it
2401 should not update the ctime.
2405 Bugzilla : 15713/16362
2406 Description: Assertion in iopen_connect_dentry in 1.6.3
2407 Details : looking up an inode via iopen with the wrong generation number can
2408 populate the dcache with a disconneced dentry while the inode
2409 number is in the process of being reallocated. This causes an
2410 assertion failure in iopen since the inode's dentry list contains
2411 both a connected and disconnected dentry.
2415 Description: assertion failure in ldlm_handle2lock()
2416 Details : fix a race between class_handle_unhash() and class_handle2object()
2417 introduced in lustre 1.6.5 by bug 13622.
2422 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2423 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2428 Description: oops in sd_iostats_seq_show()
2429 Details : unloading/reloading the scsi low level driver triggers a kernel
2430 bug when trying to access the sd iostat file.
2435 Description: Kernel panics during QLogic driver reload
2436 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2437 patch, causing memory corruption.
2442 Description: journal_dev option does not work in b1_6
2443 Details : pass mount option during pre-mount.
2445 Severity : enhancement
2447 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2448 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2449 extent information of a file. It can be used to map logical blocks
2450 in a file to physical blocks in the block device.
2454 Description: LDLM soft lockups - improvement
2455 Details : It is be possible to send the lock handle along with each read
2456 or write request because the client is already doing a lock match
2457 itself so there isn't any reason the OST should have to re-do that
2461 Frequency : only on Cray X2
2463 Description: X2 build failures
2464 Details : fix build failures on Cray X2.
2468 Description: xid & resent requests
2469 Details : Initialize RPC XID from clock at startup (randomly if clock is
2472 Severity : enhancement
2474 Description: Add lustre_start utility to start or stop multiple Lustre servers
2479 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2480 Details : In case of memory pressure, list_del() can be called twice on
2481 req->rq_history_list, causing a kernel oops.
2485 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2486 Details : fix stack overflow in the distributed lock manager by defering
2487 export eviction after a failed AST to the elt thread instead of
2488 handling it in the dlm interpret routine.
2492 Description: Convert some comments to new format.
2493 Details : Update documenting comments to match doxygen conventions.
2497 Description: Grammar fixes.
2498 Details : A couple of trivial spelling fixes.
2502 Description: OSD_COUNTERS-mandatory
2503 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2504 to simplify the code.
2508 Description: simplify cmm_device freeing logic.
2509 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2513 Description: Add lockdep support to dt_object_operations locking interface.
2514 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2515 parameter indicating lock ordering. Update mdd code to use new
2520 Description: Introduce failloc constants for lockless IO tests.
2521 Details : Add two new failloc constants to test lockless IO. Only one of
2522 them in implemented---another is checked in yet to be landed
2527 Description: Add lockdep support for inode mutex.
2528 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2529 in the situations where Lustre has to lock more than one inode
2534 Description: Add optional invariants checking support.
2535 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2536 configure switch. This macro is to be used for consistency and
2537 sanity checks that are too expensive to be left in `production'
2542 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2543 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2548 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2549 Details : Use spin_lock_nested() in (the only) situation where more than
2550 one ldlm_lock is locked simultaneously. Also, fix possible
2551 dead-lock in ldlm_lock_change_resource() by enforcing particular
2556 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2557 Details : Instead of specifying each ldlm_lock call-back through separate
2558 parameter, wrap them into struct ldlm_callback_suite.
2562 Description: Kill join_lru obd method and its callers.
2563 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2564 in memory---a requirement implemented through ->o_join_lru() obd
2565 method in HEAD. Remove this method and its users.
2569 Description: Add asynchronous ldlm ENQUEUE completion handler.
2570 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2571 case that stock ldlm_completion_ast() cannot handle as it waits
2572 until lock is granted. Introduce new ldlm_completion_ast_async()
2573 for this. Also comment ldlm_completion_ast().
2577 Description: ldlm_error <-> errno conversion.
2578 Details : Add functions to map (rather arbitrary) between LDLM error codes
2579 and standard errno values. CLIO needs this to prevent LDLM specific
2580 constants from escaping ldlm and osc.
2584 Description: Kill unused ldlm_handle2lock_ns() function.
2585 Details : Kill unused ldlm_handle2lock_ns() function.
2589 Description: Add lu_ref support to ldlm_lock
2590 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2591 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2592 and ldlm_resource. LDLM interface has to be changed, because code that
2593 releases a reference on a lock, has to "know" what reference this is.
2594 In the most frequent case
2596 lock = ldlm_handle2lock(handle);
2598 LDLM_LOCK_PUT(lock);
2600 no changes are required. When any other reference (received _not_ from
2601 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2602 instead of LDLM_LOCK_PUT().
2604 Arguably, changes are pervasive, and interface requires some discipline
2605 for proper use. On the other hand, it was very instrumental in finding
2606 a few leaked lock references.
2610 Description: Add ldlm_lock_addref_try().
2611 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2612 attempts to addref a lock that might be being canceled
2617 Description: Add ldlm_weigh_callback().
2618 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2619 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2620 hard-coded `number of pages' logic.
2624 Description: Add lockdep annotations to llog code.
2625 Details : Use appropriately tagged _nested() locking calls in the places
2626 where llog takes more than one ->lgh_lock lock.
2630 Description: Add loi_kms_set().
2631 Details : Wrap kms updates into a helper function.
2635 Description: Constify instances of struct lsm_operations.
2636 Details : Constify instances of struct lsm_operations.
2640 Description: lu_conf support.
2641 Details : On a server, a file system object is uniquely identified
2642 by a fid, which is sufficient to locate and load all object
2643 state (inode). On a client, on the other hand, more data are
2644 necessary instantiate an object. Change lu_object_find() and
2645 friends to take additional `lu_conf' argument describing object.
2646 Typically this includes layout information.
2650 Description: lu_context fixes.
2651 Details : Introduce new lu_context functions that are needed on the client
2652 side, where some system threads (ptlrpcd) are shared by multiple
2653 modules, and so cannot be stopped during module shutdown.
2657 Description: Add start and stop methods to lu_device_type_operations.
2658 Details : Introduce two new methods in lu_device_type_operations, that are
2659 invoked when first instance of a given type is created and last one
2660 is destroyed respectively. This is need by CLIO.
2664 Description: Add lu_ref support to struct lu_device.
2665 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2666 track leaked references.
2670 Description: Introduce lu_kmem_descr.
2671 Details : lu_kmem_descr and its companion interface allow to create
2672 and destroy a number of kmem caches at once.
2676 Description: Fix lu_object finalization race.
2677 Details : Fix a race between lu_object_find() finding an object and its
2678 concurrent finalization. This race is (most likely) not possible
2679 on the server, but might happen on the client.
2683 Description: Introduce lu_ref interface.
2684 Details : lu_ref is a debugging module allowing to track references to
2685 a given object. It is quite cpu expensive, and has to be
2686 explicitly enabled with --enable-lu_ref. See usage description
2691 Description: Factor lu_site procfs stats into a separate function.
2692 Details : Separate lu_site stats printing code into a separate function
2693 that can be reused on a client.
2697 Description: Constify instances of struct {lu,dt,md}_device_operations.
2698 Details : Constify instances of struct {lu,dt,md}_device_operations.
2702 Description: Introduce struct md_site and move meta-data specific parts of
2703 struct lu_site here.
2704 Details : Move md-specific fields out of struct lu_site into special struct
2705 md_site, so that lu_site can be used on a client.
2709 Description: Kill mdd_lov_destroy().
2710 Details : Remove unused mdd code.
2714 Description: Add st_block checking to multistat.c.
2715 Details : Add st_block checking to multistat.c.
2719 Description: Add lu_ref support to struct obd_device.
2720 Details : Add lu_ref tracking to obd_device.
2724 Description: Kill obd_set_fail_loc().
2725 Details : Remove unused code.
2729 Description: Add special type for ptlrpc_request interpret functions.
2730 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2731 there. Actual usage will be in CLIO.
2735 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2736 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2737 doesn't work with lockdep.
2741 Description: Add rwv.c test program.
2742 Details : New testing program exercising readv(2) and writev(2) (Qian).
2746 Description: Add sendfile.c test program.
2747 Details : New testing program exercising sendfile(2) (Jay).
2751 Description: Ratelimit a message that can be very frequent.
2752 Details : Ratelimit a memory allocation failure message that can
2757 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2758 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2762 Description: Liblustre build fixes.
2763 Details : Liblustre build fixes.
2767 Description: libcfs: add cfs_{need,cond}_resched() interface.
2768 Details : libcfs: add cfs_{need,cond}_resched() definition and
2769 implementations for Linux, NT, and liblustre.
2771 Severity : enhancement
2773 Description: More exported tunables for mballoc
2774 Details : Add support for tunable preallocation window and new tunables for
2775 large/small requests
2779 Description: Detect corruption of block bitmap and checking for preallocations
2780 Details : Checks validity of on-disk block bitmap. Also it does better
2781 checking of number of applied preallocations. When corruption is
2782 found, it turns filesystem readonly to prevent further corruptions.
2786 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2787 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2788 window to 0. If it does hit cache miss, and read-pattern is still
2789 stride-io mode, does not reset the stride window, but also does not
2790 increase the stride window length in this case.
2794 Frequency : only for big-endian servers
2795 Description: Check if system is big-endian while mounting fs with extents feature
2796 Details : Mounting a filesystem with extents feature will fail on big-endian
2797 systems since ext3-based ldiskfs is not supported on big-endian
2798 systems. This can be over-riden with "bigendian_extents" mount option.
2800 Severity : enhancement
2802 Description: The root squash functionality
2803 Details : A security feature, which is to prevent users from being able
2804 to mount lustre on their desktop, run as root, and delete
2805 all of the files in the filesystem. The goal is accomplished by
2806 remapping user id (UID) and group id (GID) of the root user to
2807 a UID and GID specified by the system administartor via Lustre
2808 configuration management server (MGS). The functionality also
2809 allows to specify sets of clients for which the remapping does
2814 Description: Excessive recovery window
2815 Details : With AT enabled, the recovery window can be excessively long (6000+
2816 seconds). To address this problem, we no longer use
2817 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2818 timeout no longer depends on the service time, it is set to
2819 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2820 time via pb_service_time.
2824 Description: Watchdog triggered on MDS failover
2825 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2826 the OSTs know that the MDS "UUID" can be reused for the same export
2827 from a different NID, so we do not need to wait for the export to be
2831 Frequency : rare, only if using MMP with Linux RAID
2833 Description: MMP doesn't work with Linux RAID
2834 Details : While using HA for Lustre servers with Linux RAID, it is possible
2835 that MMP will not detect multiple mounts. To make this work we
2836 need to unplug the device queue in RAID when the MMP block is being
2837 written. Also while reading the MMP block, we should read it from
2838 disk and not the cached one.
2840 Severity : enhancement
2842 Description: open file using fid
2843 Details : A file can be opened using just its fid, like
2844 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2847 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2850 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2851 Details : add proper locking for creation of jbd_4k slab cache
2855 Description: MMP check in ext3_remount() fails without displaying any error
2856 Details : When multiple mount protection fails during remount, proper error
2859 Severity : enhancement
2861 Description: Allow stripe size to be up to 4G-64k
2862 Details : Fix math logic to allow large stripe sizes.
2866 Description: add check for >8TB ldiskfs filesystems
2867 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2868 Don't allow >8TB ldiskfs filesystems to be mounted without
2869 force_over_8tb mount option
2874 Descriptoin: OSS complains "too many blocks" and crashes
2875 Details : In filter_commitrw_write():
2876 1. lquota_chkquota();
2877 2. obdo_from_inode();
2878 3. lquota_pending_commit();
2879 "2." has changed uid and gid in this case, then caused
2880 lquota_chkquota() and lquota_pending_commit() to use
2881 different uid and gid. That is the root of the bug.
2885 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2886 Details : Detect and print out the 'stripe count = -1' case properly.
2890 Description: Out or order replies might be lost on replay
2891 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2892 request with tid smaller than one currently being added, add it
2893 to the start, not end of the list.
2895 --------------------------------------------------------------------------------
2897 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2899 * Support for kernels:
2900 2.6.5-7.283 (SLES 9),
2901 2.6.9-55.EL (RHEL 4),
2902 2.6.16.46-0.14 (SLES 10),
2903 2.6.18.8 vanilla (kernel.org)
2904 * Client support for unpatched kernels:
2905 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2906 2.6.16 - 2.6.22 vanilla (kernel.org)
2907 * Due to recently discovered recovery problems, we do not recommend
2908 using patchless RHEL 4 clients with this or any earlier release.
2909 * Recommended e2fsprogs version: 1.39.cfs8
2910 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2911 * Starting with this release, the ldiskfs backing filesystem required
2912 by Lustre is now in its own package, lustre-ldiskfs. This package
2913 should be installed. It is versioned separately from Lustre and
2914 may be released separately in future.
2918 Description: block reactivating mgc import until all deactivates complete
2919 Details : Fix race when failing back MDT/MGS to itself (testing)
2922 Frequency : at statup only
2924 Description: mds_lov_synchronize race leads to various problems
2925 Details : simultaneous MDT->OST connections at startup can cause the
2926 sync to abort, leaving the OSC in a bad state.
2928 Severity : enhancement
2930 Description: add optional extra BUILD_VERSION info
2931 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2932 to override the lustre version.
2935 Frequency : 2.6.18 servers only
2937 Description: ll_kern_mount() doesn't release the module reference
2938 Details : The ldiskfs module reference count never drops down to 0
2939 because ll_kern_mount() doesn't release the module reference.
2944 Description: server LBUG when using old ost_num_threads parameter
2945 Details : Accept the old ost_num_threads parameter but warn that it
2946 is deprecated, and fix an off-by-one error that caused an LBUG.
2951 Description: Transient SCSI error results in persistent IO issue
2952 Details : iobuf->dr_error is not reinitialized to 0 between two
2956 Frequency : sometimes when underlying device returns I/O errors
2958 Description: OSTs not going read-only during write failures
2959 Details : OSTs are not remounted read-only when the journal commit threads
2960 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2961 instead of the ext3 wrappers.
2965 Description: poor connect scaling with increasing client count
2966 Details : Don't run filter_grant_sanity_check for more than 100 exports
2967 to improve scaling for large numbers of clients.
2970 Frequency : SLES10 only
2972 Description: sanity-quota.sh quotacheck failed: rc = -22
2973 Details : Quotas cannot be enabled on SLES10.
2976 Frequency : liblustre clients only
2978 Description: getdirentries does not give error when run on compute nodes
2979 Details : getdirentries does not fail when the size specified as an argument
2980 is too small to contain at least one entry
2982 Severity : enhancement
2984 Description: Add LNET router traceability for debug purposes
2985 Details : If a checksum failure occurs with a router as part of the
2986 IO path, the NID of the last router that forwarded the bulk data
2987 is printed so it can be identified.
2992 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2993 Details : Due to a race condition, liblustre clients were occasionally
2994 evicted incorrectly.
2996 Severity : enhancement
2998 Description: lfs setstripe use optional parameters instead of postional
3001 Severity : enhancement
3003 Description: Nanosecond timestamp support for ldiskfs
3004 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3005 resolution timestamps. There is not yet support for this at
3006 the Lustre filesystem level.
3009 Frequency : during server recovery
3011 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3012 Details : request with rq_no_resend flag not awake l_wait_event if they get
3016 Frequency : nfs export on patchless client
3018 Description: connectathon hang when test nfs export over patchless client
3019 Details : Disconnected dentry cannot be found with lookup, so we do not need
3020 to unhash it or make it invalid
3023 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3024 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3025 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3030 Description: open req refcounting wrong on reconnect
3031 Details : If reconnect happened between getting open reply from server and
3032 call to mdc_set_replay_data in ll_file_open, we will schedule
3033 replay for unreferenced request that we are about to free.
3034 Subsequent close will crash in variety of ways.
3035 Check that request is still eligible for replay in
3036 mdc_set_replay_data().
3041 Description: disable writes to filesystem when reading health_check file
3042 Details : the default for reading the health_check proc file has changed
3043 to NOT do a journal transaction and write to disk, because this
3044 can cause reads of the /proc file to hang and block HA state
3045 checking on a healthy but otherwise heavily loaded system. It
3046 is possible to return to the previous behaviour during configure
3047 with --enable-health-write.
3049 Severity : enhancement
3051 Description: 64-bit inode version
3052 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3053 to the inode. This will be required for version-based recovery.
3058 Description: MDS fails to start if a duplicate client export is detected
3059 Details : in some rare cases it was possible for a client to connect to
3060 an MDS multiple times. Upon recovery the MDS would detect this
3061 and fail during startup. Handle this more gracefully.
3063 Severity : enhancement
3065 Description: Add -o localflock option to simulate old noflock
3067 Details : This will achieve local-only flock/fcntl locks
3073 Description: log_commit_thread vs filter_destroy race leads to crash
3074 Details : Take import reference before releasing llog record semaphore
3079 Description: Wrong request locking in request set processing
3080 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3081 imp_delayed_list, in this place should be used imp_lock.
3084 Frequency : when reconnection
3086 Description: Grant Leak when osc reconnect to OST
3087 Details : When osc reconnect ost, OST(filter) should check whether it
3088 should grant more space to client by comparing fed_grant and
3089 cl_avail_grant, and return the granted space to client instead
3090 of "new granted" space, because client will call osc_init_grant
3091 to update the client grant space info.
3094 Frequency : when client reconnect to OST
3096 Description: Grant Leak when osc do resend and replay bulk write
3097 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3098 bulk write request, because the grant info while be sync between
3099 OSC and OST when reconnect, and we should ignore the grant info
3100 these of resend/replay write req.
3105 Description: Grant space more than available left space sometimes.
3106 Details : When then OST is about to be full, if two bulk writing from
3107 different clients came to OST. Accord the avaliable space of the
3108 OST, the first req should be permitted, and the second one
3109 should be denied by ENOSPC. But if the seconde arrived before
3110 the first one is committed. The OST might wrongly permit second
3111 writing, which will cause grant space > available space.
3114 Frequency : when client is evicted
3116 Description: Grant might be wrongly erased when osc is evicted by OST
3117 Details : when the import is evicted by server, it will fork another
3118 thread ptlrpc_invalidate_import_thread to invalidate the
3119 import, where the grant will be set to 0. While the original
3120 thread will update the grant it got when connecting. So if
3121 the former happened latter, the grant will be wrongly errased
3122 because of this race.
3127 Description: Checking Stale with correct fid
3128 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3129 to check whether it is stale, because sometimes, we want the
3130 enqueue happened anyway, and op_data.fid2 will not be initialized.
3132 Severity : enhancement
3134 Description: update patchless client
3135 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3138 Frequency : only with 2.4 kernel
3140 Description: random memory corruption
3141 Details : size of struct ll_inode_info is to big for union inode.u and this
3142 can be cause of random memory corruption.
3147 Description: Memory leak in recovery
3148 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3149 It should also check obd_fail before fsfilt_start, otherwise if
3150 fsfilt_start return -EROFS,(failover mds during mds recovery).
3151 then the req will return with repmsg->transno = 0 and rc = EROFS.
3152 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3153 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3154 be freed no matter whether fsfilt_commit success or not.
3157 Frequency : only with huge count clients
3159 Description: Prevents from taking the superblock lock in llap_from_page for
3161 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3162 for llap_from_page call in ll_removepage prevents from taking the
3163 superblock lock for a soon died page.
3168 Description: Not check open intent error before release open handle
3169 Details : in some rare cases, the open intent error is not checked before
3170 release open handle, which may cause
3171 ASSERTION(open_req->rq_transno != 0), because it tries to release
3172 the failed open handle.
3177 Description: Set cat log bitmap only after create log success.
3178 Details : in some rare cases, the cat log bitmap is set too early. and it
3179 should be set only after create log success.
3183 Description: Accessing a block bevice can re-enable I/O when Lustre is
3184 tearing down a device.
3185 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3190 Description: service threads may hog cpus when there are a lot of requests
3191 Details : Insert cond_resched to give other threads a chance to use some CPU
3196 Description: the cat log was not initialized in recovery
3197 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3198 unlink log on mds will not be initialized until mds post
3199 recovery. And also in mds post recovery, the unlink log will
3200 initialization will be done asynchronausly, so there will be race
3201 between add unlink log and unlink log initialization.
3205 Description: brw_stats were being printed incorrectly
3206 Details : brw_stats were being printed as log2 but all of them were not
3207 recorded as log2. Also remove some code duplication arising from
3208 filter_tally_{read,write}.
3212 Frequency : rare, only in recovery.
3213 Description: ASSERTION(req->rq_type != LI_POISON) failed
3214 Details : imp_lock should be held while iterating over imp_sending_list for
3215 prevent destroy request after get timeout in ptlrpc_queue_wait.
3219 Description: replay-single.sh test 52 fails
3220 Details : A lock's skiplist need to be cleanup when it being unlinked
3221 from its resource list.
3225 Description: Short directio read returns full requested size rather than
3227 Details : Direct I/O operations should return actual amount of bytes
3228 transferred rather than requested size.
3230 Severity : enhancement
3232 Description: metadata RPC reduction (e.g. for rm performance)
3233 Details : decrease the amount of synchronous RPC between clients and servers
3234 by canceling conflicing lock before the operation on the client
3235 and packing thier handles into the main operation RPC to server.
3237 Severity : enhancement
3239 Description: Async OSC create to avoid the blocking unnecessarily.
3240 Details : If an OST has no remaining object, system will block on the
3241 creation when it needs to create a new object on this OST. Now,
3242 ways use pre-created objects when available, instead of blocking on
3243 an empty osc while others are not empty. If we must block, we
3244 block for the shortest possible period of time.
3248 Description: improve handling recoverable errors
3249 Details : if request processig with error which can be recoverable on server
3250 request should be resend, otherwise page released from cache and
3253 Severity : enhancement
3255 Description: refine locking for avoid write wrong info into lov_objid file
3256 Details : fix possible races with add new target and write/update data in
3259 --------------------------------------------------------------------------------
3261 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3266 Frequency : on some architectures
3268 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3269 Details : Uninitialized flags sometimes cause configuration commands to
3273 Frequency : patchless clients only
3275 Description: missing __iget() symbol export
3276 Details : The __iget() symbol export is missing. To avoid the need for
3277 this on patchless clients the deathrow inode reaper is turned
3278 off, and we depend on the VM to clean up old inodes. This
3279 dependency was during via the fix for bug 12181.
3281 --------------------------------------------------------------------------------
3283 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3285 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3286 INTEROPERATE with older versions automatically. In many cases a
3287 special upgrade step is needed. Please read the
3288 user documentation before upgrading any part of a live system.
3289 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3290 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3291 * WARNING: Lustre configuration and startup changes are required with
3292 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3296 Severity : enhancement
3298 Description: MountConf
3299 Details : Lustre configuration is now managed via mkfs and mount
3300 commands instead of lmc and lconf. New obd types (MGS, MGC)
3301 are added for dynamic configuration management. See
3302 https://mail.clusterfs.com/wikis/lustre/MountConf for
3305 Severity : enhancement
3307 Description: dynamic OST addition
3308 Details : OSTs can now be added to a live filesystem
3310 Severity : enhancement
3312 Description: startup order invariance
3313 Details : MDTs and OSTs can be started in any order. Clients only
3314 require the MDT to complete startup.
3316 Severity : enhancement
3318 Description: parallel, asynchronous orphan cleanup
3319 Details : orphan cleanup is now performed in separate threads for each
3320 OST, allowing parallel non-blocking operation.
3322 Severity : enhancement
3324 Description: optimized stripe assignment
3325 Details : stripe assignments are now made based on ost space available,
3326 ost previous usage, and OSS previous usage, in order to try
3327 to optimize storage space and networking resources.
3329 Severity : enhancement
3331 Description: Permanently set tunables
3332 Details : All writable /proc/fs/lustre tunables can now be permanently
3333 set on a per-server basis, at mkfs time or on a live system.
3335 Severity : enhancement
3337 Description: Lustre message v2
3338 Details : Add lustre message format v2.
3340 Severity : enhancement
3342 Description: client OST exclusion list
3343 Details : Clients can be started with a list of OSTs that should be
3344 declared "inactive" for known non-responsive OSTs.
3348 Description: SPEC SFS validation failure on NFS v2 over lustre.
3349 Details : Changes the blocksize for regular files to be 2x RPC size,
3350 and not depend on stripe size.
3352 Severity : enhancement
3354 Description: Multiple MD RPCs in flight.
3355 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3356 To avoid overly-overloading MDS, introduce a limit on number of
3357 MDS RPCs in flight for a single client and add /proc controls
3358 to adjust this limit.
3360 Severity : enhancement
3362 Description: client read/write statistics
3363 Details : Add client read/write call usage stats for performance
3364 analysis of user processes.
3365 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3366 file access. extents_stats shows chunk size distribution.
3367 extents_stats_per_process show chunk size distribution per
3370 Severity : enhancement
3372 Description: mds statistics
3373 Details : Add detailed mds operations statistics in
3374 /proc/fs/lustre/mds/*/stats.
3378 Description: Failure of copying files with lustre special EAs.
3379 Details : Client side always return success for setxattr call for lustre
3380 special xattr (currently only "trusted.lov").
3385 Description: Refcount LNET uuids
3386 Details : The global LNET uuid list grew linearly with every startup;
3387 refcount repeated list entries instead of always adding to
3390 Severity : enhancement
3392 Description: Dynamic service threads
3393 Details : Within a small range, start extra service threads
3394 automatically when the request queue builds up.
3397 Frequency : mixed-endian client/server environments
3399 Description: mixed-endian crashes
3400 Details : The new msg_v2 system had some failures in mixed-endian
3403 Severity : enhancement
3405 Description: Easy OST removal
3406 Details : OSTs can be permanently deactivated with e.g. 'lctl
3407 conf_param lustre-OST0001.osc.active=0'
3409 Severity : enhancement
3411 Description: MGS proc entries
3412 Details : Added basic proc entries for the MGS showing what filesystems
3415 Severity : enhancement
3417 Description: provide MGS failover
3418 Details : Added config lock reacquisition after MGS server failover.
3420 Severity : enhancement
3422 Description: add Linux 2.4 support
3423 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3427 Description: a large application tries to do I/O to the same resource and dies
3428 in the middle of it.
3429 Details : Check the req->rq_arrival time after the call to
3430 ost_brw_lock_get(), but before we do anything about
3431 processing it & sending the BULK transfer request. This
3432 should help move old stale pending locks off the queue as
3433 quickly as obd_timeout.
3436 Frequency : when an incorrect nid is specified during startup
3438 Description: ptlrpc connect to non-existant node causes kernel crash
3439 Details : LNET can't be re-entered from an event callback, which
3440 happened when we expire a message after the export has been
3441 cleaned up. Instead, hand the zombie cleanup off to another
3444 Severity : enhancement
3446 Description: plain/inodebits lock performance improvement
3447 Details : Grouping plain/inodebits in granted list by their request modes
3448 and bits policy, thus improving the performance of search through
3452 Frequency : only if OST filesystem is corrupted
3454 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3455 Details : for a short time RPCs with bulk IO are in the replay list,
3456 but replay of bulk IOs is unimplemented. If the OST filesystem
3457 is corrupted due to disk cache incoherency and then replay is
3458 started it is possible to trip an assertion. Avoid putting
3459 committed RPCs into the replay list at all to avoid this issue.
3462 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3464 Description: System hang on startup
3465 Details : This bug allowed the liblustre (e.g. catamount) client to
3466 return to the app before handling all startup RPCs. This
3467 could leave the node unresponsive to lustre network traffic
3468 and manifested as a server ptllnd timeout.
3470 Severity : enhancement
3472 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3473 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3474 boolean to control whether to print peer debug info when a
3475 client's RPC times out.
3478 Frequency : only for kernels with patches from Lustre below 1.4.3
3480 Description: Remove old rdonly API
3481 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3484 Frequency : only for devices with external journals
3486 Description: Set external device read-only also
3487 Details : During a commanded failover stop, we set the disk device
3488 read-only while the server shuts down. We now also set any
3489 external journal device read-only at the same time.
3492 Frequency : when upgrading from 1.4 while trying to change parameters
3494 Description: The wrong (new) MDC name was used when setting parameters for
3495 upgraded MDT's. Also allows changing of OSC (and MDC)
3496 parameters if --writeconf is specified at tunefs upgrade time.
3499 Frequency : when setting specific ost indicies
3501 Description: QOS code breaks on skipped indicies
3502 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3503 created with --index need not be sequential.
3505 Severity : enhancement
3507 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3508 Details : The uninit_groups feature works in conjunction with the kernel
3509 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3510 up the pass1 processing of e2fsck. This is a read-only feature
3511 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3512 mount filesystems that have had this feature enabled.
3514 Severity : enhancement
3516 Description: Improve multi-block allocation algorithm to avoid fragmentation
3517 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3518 allocation locality and avoid filesystem fragmentation.
3520 ------------------------------------------------------------------------------
3522 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3524 * Support for kernels:
3525 2.4.21-47.0.1.EL (RHEL 3)
3526 2.6.5-7.283 (SLES 9)
3527 2.6.9-42.0.10.EL (RHEL 4)
3528 2.6.12.6 vanilla (kernel.org)
3529 2.6.16.27-0.9 (SLES 10)
3530 * Recommended e2fsprogs version: 1.39.cfs5
3532 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3536 Frequency : occasional, depends on client load and configuration
3537 Bugzilla : 12181, 12203
3538 Description: data loss for recently-modified files
3540 Details : In some cases it is possible that recently written or created
3541 files may not be written to disk in a timely manner (this should
3542 normally be within 30s unless client IO load is very high).
3543 The problem appears as zero-length files or files that are a
3544 multiple of 1MB in size after a client crash or client eviction
3545 that are missing data at the end of the file.
3547 This problem is more likely to be hit on clients where files are
3548 repeatedly created and unlinked in the same directory, clients
3549 have a large amount of RAM, have many CPUs, the filesystem has
3550 many OSTs, the clients are rebooted frequently, and/or the files
3551 are not accessed by other nodes after being written.
3553 The presence of the problem can be detected by looking at
3554 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3555 smaller than the second (nr_unused) then dirty files will not
3556 be flushed automatically to disk. "sync; sleep 10" should be
3557 run several times on the node before unmounting it to update
3558 Lustre (this is also safe to run on nodes without this problem).
3560 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3561 that can cause this same problem, so customers using that kernel
3562 also need to update the kernel in addition to Lustre. In order
3563 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3565 It is normal that files written just before a client crash (less
3566 than 30s) may not yet have been flushed to disk, even for local
3570 Frequency : frequent on thin XT3 nodes
3572 Description: UUID collision on thin XT3 Linux nodes
3573 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3574 randomly, since we relied on an insufficiently-seeded PRNG.
3579 Description: OSS hangs after "All ost request buffers busy"
3580 Details : A deadlock between quota and journal operations caused OSS
3581 hangs after printing "All ost request buffers busy."
3584 Frequency : always on liblustre builds
3586 Description: Cleanup compiler warnings on liblustre
3589 Frequency : always on liblustre builds on XT3
3591 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3595 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3596 > 32000 subdirectories
3597 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3598 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3599 subdirectory count crosses 32000. This will aid e2fsck to
3600 correctly handle more than 32000 subdirectories.
3603 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3605 Description: System hang on startup
3606 Details : This bug allowed the liblustre (e.g. catamount) client to
3607 return to the app before handling all startup RPCs. This
3608 could leave the node unresponsive to lustre network traffic
3609 and manifested as a server ptllnd timeout.
3611 Severity : enhancement
3613 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3614 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3615 boolean to control whether to print peer debug info when a
3616 client's RPC times out.
3621 Description: make O_SYNC working on 2.6 kernels
3622 Details : 2.6 kernels use different method for mark pages for write,
3623 so need add a code to lustre for O_SYNC work.
3628 Description: Failure to close file and release space on NFS
3629 Details : Put inode details into lock acquired in ll_intent_file_open.
3630 Use mdc_intent_lock in ll_intent_open to properly
3631 detect all kind of errors unhandled by mdc_enqueue.
3636 Description: proc file read during shutdown sometimes raced obd removal,
3638 Details : Add lock to prevent obd access after proc file removal.
3641 Frequency : Only for files larger than 4GB on 32-bit clients.
3643 Description: improperly doing page alignment of locks
3644 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3645 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3650 Description: RPCs being resent when they shouldn't be
3651 Details : Some RPCs that should not be resent are being resent. This
3652 can cause inconsistencies in the RPC state machine. Do not
3653 resend such requests.
3656 Frequency : rare, only with NFS export
3658 Description: Crash on NFS re-export node
3659 Details : under very unusual load conditions an assertion is hit in
3660 ll_intent_file_open()
3663 Frequency : only if OST filesystem is corrupted
3665 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3666 Details : for a short time RPCs with bulk IO are in the replay list,
3667 but replay of bulk IOs is unimplemented. If the OST filesystem
3668 is corrupted due to disk cache incoherency and then replay is
3669 started it is possible to trip an assertion. Avoid putting
3670 committed RPCs into the replay list at all to avoid this issue.
3675 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3676 Details : Large single O_DIRECT read and write calls can fail to allocate
3677 a sufficiently large buffer to process the request. In case of
3678 allocation failure the allocation is retried with a smaller
3679 buffer and broken into smaller requests.
3681 Severity : enhancement
3683 Description: Add -o localflock option to simulate old noflock behaviour.
3684 Details : This will achieve local-only flock/fcntl locks coherentness.
3689 Description: versioning check is incomplete
3690 Details : Checking the version difference of client vs. server, report
3691 error if the gap is too big.
3696 Description: add support PG_writeback bit
3697 Details : add support for PG_writeback bit for Lustre, for more carefull
3698 work with page cache in 2.6 kernel. This also fix some deadlocks
3699 and remove hack for work O_SYNC with 2.6 kernel.
3701 Severity : enhancement
3703 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3704 Details : The uninit_groups feature works in conjunction with the kernel
3705 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3706 up the pass1 processing of e2fsck. This is a read-only feature
3707 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3708 mount filesystems that have had this feature enabled.
3710 Severity : enhancement
3712 Description: Improve multi-block allocation algorithm to avoid fragmentation
3713 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3714 allocation locality and avoid filesystem fragmentation.
3716 ------------------------------------------------------------------------------
3718 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3720 * Support for kernels:
3721 2.6.9-42.0.3.EL (RHEL 4)
3722 2.6.5-7.276 (SLES 9)
3723 2.4.21-47.0.1.EL (RHEL 3)
3724 2.6.12.6 vanilla (kernel.org)
3725 2.6.16.21-0.8 (SLES10)
3726 * Recommended e2fsprogs version: 1.39.cfs2-0
3728 * The backwards-compatible /proc/sys/portals symlink has been removed
3729 in this release. Before upgrading, please ensure that you change
3730 any configuration scripts or /etc/sysctl.conf files that access
3731 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3732 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3733 in advance of the upgrade on any system running Lustre 1.4.6 or
3734 newer, since /proc/sys/lnet was added in that version.
3735 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3739 Frequency : only when quota is used
3741 Description: avoid scanning export list for quota master
3742 Details : Change the algorithms to avoid scanning export list in order
3743 to improve the efficiency.
3746 Frequency : MDS failover only, very rarely
3748 Description: "went back in time" messages on mds failover
3749 Details : The greatest transno may be lost when the current operation
3750 finishes with an error (transno==0) and the client's last_rcvd
3751 record is over-written. Save the greatest transno in the
3752 mds_last_transno for this case.
3755 Frequency : always for specific kernels and striping counts
3757 Description: client may get "Matching packet too big" without ACL support
3758 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3759 when trying to access files in certain configurations. The
3760 clients should in fact be denied when mounting because they do
3761 not understand ACLs.
3764 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3766 Description: many clients connecting with IO in progress causes connect timeouts
3767 Details : Avoid synchronous journal commits to avoid delays caused by many
3768 clients connecting/disconnecting when bulk IO is in progress.
3769 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3770 OST_IO_PORTAL to avoid delays behind potentially many pending
3774 Frequency : occasionally with multiple writers to a single file
3776 Description: shared writes to file may result in wrong size reported by stat()
3777 Details : Allow growing of kms when extent lock is cancelled
3780 Frequency : always with random mmap IO to multi-striped file
3782 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3783 Details : Only if the hole is at the end of OST object so that kms is too
3784 small. Fix is to increase kms accordingly in ll_nopage.
3787 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3789 Description: writes to a missing object would leak memory on the OST
3790 Details : If there is an inconsistency between the MDS and OST filesystems,
3791 such that the MDS references an object that doesn't exist, writes
3792 to that object will leak memory due to incorrect cleanup in the
3793 error handling path, eventually running out of memory on the OST.
3798 Description: Creating too long symlink causes lustre errors
3799 Details : Check symlink and name lengths before sending requests to MDS.
3802 Frequency : only if flock is enabled (not on by default)
3804 Description: posix locks not released on fd closure on 2.6.9+
3805 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3806 kernels, this caused such locks not to be released on fd close and
3807 then assertions on fs unmount about still used locks.
3810 Frequency : MDS failover only, very rarely
3812 Description: clients may get ASSERTION(granted_lock != NULL)
3813 Details : When request was taking a long time, and a client was resending
3814 a getattr by name lock request. The were multiple lock requests
3815 with the same client lock handle and
3816 mds_getattr_name->fixup_handle_for_resent_request found one of the
3817 lock handles but later failed with ASSERTION(granted_lock != NULL).
3822 Description: handle->h_buffer_credits > 0, assertion failure
3823 Details : h_buffer_credits is zero after truncate, causing assertion
3824 failure. This patch extends the transaction or creates a new
3828 Frequency : NFS re-export or patchless client
3829 Bugzilla : 11179, 10796
3830 Description: Crash on NFS re-export node (__d_move)
3831 Details : We do not want to hash the dentry if we don't have a lock.
3832 But if this dentry is later used in d_move, we'd hit uninitialised
3833 list head d_hash, so we just do this to init d_hash field but
3834 leave dentry unhashed.
3837 Frequency : NFS re-export or patchless client
3839 Description: NFS exports has problem with symbolic link
3840 Details : lustre client didn't properly install dentry when re-exported
3841 to NFS or running patchless client.
3844 Frequency : NFS re-export or patchless client
3846 Description: Various nfs/patchless fixes.
3847 Details : fixes reuse disconected alias for lookup process - this fixes
3848 warning "find_exported_dentry: npd != pd",
3849 fix permission error with open files at nfs.
3850 fix apply umask when do revalidate.
3853 Frequency : occasional
3855 Description: Crash on NFS re-export node
3856 Details : calling clear_page() on the wrong pointer triggered oops in
3857 generic_mapping_read().
3860 Frequency : rarely, using O_DIRECT IO
3862 Description: unaligned directio crashes client with LASSERT
3863 Details : check for unaligned buffers before trying any requests.
3866 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3868 Description: stale data returned from RAID cache
3869 Details : If only a small amount of IO is done to the RAID device before
3870 reading it again it is possible to get stale data from the RAID
3871 cache instead of reading it from disk.
3874 Frequency : always for sles10 kernel
3876 Description: sles10 support
3877 Details : ll_follow_link: compile fixes and using of nd_set_link
3878 under newer kernels.
3881 Frequency : depends on arch, kernel and compiler version, always on sles10
3884 Description: recursive or deep enough symlinks cause stack overflow
3885 Details : getting rid of large stack-allocated variable in
3889 Frequency : depends on hardware
3891 Description: lustre write performance loss in the SLES10 kernel
3892 Details : the performance loss is caused by using of write barriers in the
3893 ext3 code. The SLES10 kernel turns barrier support on by
3894 default. The fix is to undo that change for ldiskfs.
3896 ------------------------------------------------------------------------------
3898 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3900 * Support for kernels:
3901 2.6.9-42.0.3EL (RHEL 4)
3902 2.6.5-7.276 (SLES 9)
3903 2.4.21-47.0.1.EL (RHEL 3)
3904 2.6.12.6 vanilla (kernel.org)
3908 Frequency : quota enabled and large files being deleted
3910 Description: releasing more than 4GB of quota at once hangs OST
3911 Details : If a user deletes more than 4GB of files on a single OST it
3912 will cause the OST to spin in an infinite loop. Release
3913 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3918 Description: statfs data retrieved from /proc may be stale or zero
3919 Details : When reading per-device statfs data from /proc, in the
3920 {kbytes,files}_{total,free,avail} files, it may appear
3921 as zero or be out of date.
3924 Frequency : systems with MD RAID1 external journal devices
3926 Description: lconf's call to blkid is confused by RAID1 journal devices
3927 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3928 of returning all block devices that match the journal UUID.
3931 Frequency : always, for aggregate stripe size over 4GB
3933 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3934 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3935 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3938 Frequency : always if "lfs find" used on a local file/directory
3940 Description: "lfs find" segfaults if used on a local file/directory
3941 Details : The case where a directory component was not specified wasn't
3942 handled correctly. Handle this properly.
3945 Frequency : always on ppc64
3947 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3948 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3949 perform endianness conversion.
3952 Frequency : rarely (truncate to non-zero file size after write under load)
3953 Bugzilla : 10730, 10687
3954 Description: Files padded with zeros to next 4K multiple
3955 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3956 it is possible that files that are truncated to a non-zero size
3957 immediately after being written are filled with zero bytes beyond
3958 the truncated size. No file data is lost.
3960 Severity : enhancement
3962 Description: Allow recovery/failover for liblustre clients.
3963 Details : liblustre clients were unaware of failover configurations until
3966 Severity : enhancement
3968 Description: user file locks should fail when not mounting with flock option
3969 Details : Set up an error-returning stub in ll_file_operations.lock field
3970 to prevent incorrect behaviour when client is mounted without
3971 flock option. Also, set up properly f_op->flock field for
3975 Frequency : always on ia64
3977 Description: "lfs df" loops on printing out MDS statfs information
3978 Details : The obd_ioctl_data was not initialized and in some systems
3979 this caused a failure during the ioctl that did not return
3980 an error. Initialize the struct and return an error on failure.
3983 Frequency : SLES 9 only
3985 Description: Error of copying files with lustre special EAs as root
3986 Details : Client side always return success for setxattr call for lustre
3987 special xattr (currently only "trusted.lov").
3990 Frequency : rarely on clusters with both ia64+i386 clients
3992 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3993 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3994 clients are doing concurrent writes to the same file it is
3995 possible that the ia64 clients may LASSERT because the OST
3996 extent locks are not PAGE_SIZE aligned. Ensure that grown
3997 locks are always aligned on the request boundary.
4000 Frequency : specific use, occasional
4002 Description: Overwriting in use executable truncates on-disk binary image
4003 Details : If one node attempts to overwrite an executable in use by
4004 another node, we now correctly return ETXTBSY instead of
4005 truncating the file.
4007 Severity : enhancement
4009 Description: Async OSC create to avoid the blocking unnecessarily.
4010 Details : If an OST has no remaining object, system will block on the creating
4011 when need to create a new object on this OST. Now, ways use
4012 pre-created objects when available, instead of blocking on an
4013 empty osc while others are not empty. If we must block, we block
4014 for the shortest possible period of time.
4019 Description: chmod on Lustre root is propagated to other clients
4020 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4021 invalid by the follow_mount time.
4024 Frequency : liblustre clients only
4026 Description: Race in 'instant cancel' lock handling could lead to such locks
4027 never to be granted in case of SMP MDS
4028 Details : Do not destroy not yet granted but cbpending locks in
4032 Frequency : replay/resend of open
4034 Description: non null lock assetion failure in mds_intent_policy
4035 Details : Trying to replay/resend lockless open requests resulted in
4036 mds_open() returning 0 with no lock. Now it sets a flag if
4037 a lock is going to be returned.
4039 Severity : enhancement
4041 Description: Checksum enhancements
4042 Details : New checksum enhancements allow for resending RPCs that failed
4045 Severity : enhancement
4047 Description: Tunables on number of dirty pages in cacche
4048 Details : Allow to set limit on number of dirty pages cached.
4053 Description: client crash on unmount - lock still has references
4054 Details : In some error handling cases it was possible to leak a lock
4055 reference on a client while accessing a file. This was not
4056 harmful to the client during operation, but would cause the
4057 client to crash when the filesystem is unmounted.
4060 Frequency : specific case, rare
4062 Description: ETXTBSY on mds though file not in use by client
4063 Details : ETXTBSY is no longer incorrectly returned when attempting to
4064 chmod or chown a directory that the user previously tried to
4065 execute or a currently-executing binary.
4068 Frequency : extremely rare except on liblustre-based clients
4070 Description: Lustre space not freed when files are deleted
4071 Details : Clean up open-unlinked files after client eviction. Previously
4072 the unlink was skipped and the files remained as orphans.
4077 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4078 Details : In some cases it was possible to send a blocking callback to
4079 a client doing a glimpse, even though that client didn't get
4080 a lock granted. When the glimpse lock is cancelled on the OST
4081 the freed lock is left on the waiting list and corrupted the list.
4084 Frequency : all core dumps
4086 Description: Broke core dumps to lustre
4087 Details : Negative dentry may be unhashed if parent does not have UPDATE
4088 lock, but some callers, e.g. do_coredump, expect dentry to be
4089 hashed after successful create, hash it in ll_create_it.
4091 ------------------------------------------------------------------------------
4093 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4095 * Support for kernels:
4096 2.6.9-42.0.2.EL (RHEL 4)
4097 2.6.5-7.276 (SLES 9)
4098 2.4.21-40.EL (RHEL 3)
4099 2.6.12.6 vanilla (kernel.org)
4103 Frequency : always on RHEL 3
4105 Description: Number of open files grows over time
4106 Details : The number of open files grows over time, whether or not
4107 Lustre is started. This was due to a filp leak introduced
4108 by one of our kernel patches.
4110 ------------------------------------------------------------------------------
4112 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4114 * Support for kernels:
4115 2.6.9-42.EL (RHEL 4)
4116 2.6.5-7.276 (SLES 9)
4117 2.4.21-40.EL (RHEL 3)
4118 2.6.12.6 vanilla (kernel.org)
4123 Bugzilla : 5719, 9635, 9792, 9684
4124 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4125 Details : If a server is under heavy load and cannot reply to new
4126 connection requests before the client resends the (re)connect,
4127 the connection handling code can behave badly if two service
4128 threads are concurrently handing separate (re)connections from
4129 the same client. Add better locking to the connection handling
4130 code, and ensure that only a single connection will be processed
4131 for a given client UUID, even if the lock is dropped.
4133 Severity : enhancement
4135 Description: add TCP zero-copy support to kernel
4136 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4137 sends if the hardware supports scatter-gather and checksumming.
4138 This allows socklnd to do client-write and server-read more
4139 efficiently and reduce CPU utilization from skbuf copying.
4142 Frequency : only if NFS exporting from client
4144 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4145 Details : If setting attributes on a file created under NFS that had
4146 never been opened it would be possible to oops the client
4147 if the file had no objects.
4150 Frequency : always for liblustre
4152 Description: liblustre client does MDS+OSTs setattr RPC for each write
4153 Details : When doing a write from a liblustre client, the client
4154 incorrectly issued an RPC to the MDS and each OST the file was
4155 striped over in order to update the timestamps. When writing
4156 with small chunks and many clients this could overwhelm the MDS
4157 with RPCs. In all cases it would slow down the write because
4158 these RPCs are unnecessary.
4160 Severity : enhancement
4162 Description: allow number of MDS service threads to be changed at module load
4163 Details : It is now possible to change the number of MDS service threads
4164 running. Adding "options mds mds_num_threads={N}" to the MDS's
4165 /etc/modprobe.conf will set the number of threads for the next
4166 time Lustre is restarted (assuming the "mds" module is also
4167 reloaded at that time). The default number of threads will
4168 stay the same, 32 for most systems.
4173 Description: OST crash if filesystem is unformatted or corrupt
4174 Details : If an OST is started on a device that has never been formatted
4175 or if the filesystem is corrupt and cannot even mount then the
4176 error handling cleanup routines would dereference a NULL pointer.
4181 Description: NULL pointer deref in llap_from_page.
4182 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4183 incorrect) mapping if the page was truncated/reclaimed while it was
4184 searched for. Check for this condition and skip such pages when
4185 doing readahead. Introduce extra check to llap_from_page() to
4186 verify page->mapping->host is non-NULL (so page is not anonymous).
4189 Frequency : Sometimes when using sys_sendfile
4191 Description: "page not covered by a lock" warnings from ll_readpage
4192 Details : sendfile called ll_readpage without right page locks present.
4193 Now we introduced ll_file_sendfile that does necessary locking
4194 around call to generic_file_sendfile() much like we do in
4198 Frequency : with certain MDS communication failures at client mount time
4200 Description: NULL pointer deref after failed client mount
4201 Details : a client connection request may delayed by the network layer
4202 and not be sent until after the PTLRPC layer has timed out the
4203 request. If the client fails the mount immediately it will try
4204 to clean up before the network times out the request. Add a
4205 reference from the request import to the obd device and delay
4206 the cleanup until the network drops the request.
4209 Frequency : occasionally during client (re)connect
4211 Description: assertion failure during client (re)connect
4212 Details : processing a client connection request may be delayed by the
4213 client or server longer than the client connect timeout. This
4214 causes the client to resend the connection request. If the
4215 original connection request is replied in this interval, the
4216 client may trip an assertion failure in ptlrpc_connect_interpret()
4217 which thought it would be the only running connect process.
4220 Frequency : only with obd_echo servers and clients that are rebooted
4222 Description: kernel BUG accessing uninitialized data structure
4223 Details : When running an obd_echo server it did not start the ping_evictor
4224 thread, and when a client was evicted an uninitialized data
4225 structure was accessed. Start the ping_evictor in the RPC
4226 service startup instead of the OBD startup.
4228 Severity : enhancement
4229 Bugzilla : 10193 (patchless)
4230 Description: Remove dependency on various unexported kernel interfaces.
4231 Details : No longer need reparent_to_init, exit_mm, exit_files,
4232 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4235 Frequency : rare (only users of deprecated and unsupported LDAP config)
4237 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4238 Details : LDAP apparently contains 'lustreName' attributes instead of
4239 'name'. A simple remapping of the name is sufficient.
4242 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4244 Description: waiting_locks_callback trips kernel BUG if client is evicted
4245 Details : Running with the dump_on_timeout debug flag turned on makes
4246 it possible that the waiting_locks_callback() can try to dump
4247 the Lustre kernel debug logs from an interrupt handler. Defer
4248 this log dumping to the expired_lock_main() thread.
4250 Severity : enhancement
4252 Description: Support NFS exporting on 2.6 kernels.
4253 Details : Implement non-rawops metadata methods for NFS server to use without
4254 changing NFS server code.
4257 Frequency : very rare (synthetic metadata workload only)
4259 Description: two racing renames might cause an MDS thread to deadlock
4260 Details : Running the "racer" program may cause one MDS thread to rename
4261 a file from being the source of a rename to being the target of
4262 a rename at exactly the same time that another thread is doing
4263 so, and the second thread has already enqueued these locks after
4264 doing a lookup of the target and is trying to relock them in
4265 order. Ensure that we don't try to re-lock the same resource.
4268 Frequency : only very large systems with liblustre clients
4270 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4271 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4272 all OSTs in parallel. This allows the eviction of stale liblustre
4273 clients to proceed much faster than if they were done in series,
4274 and also offers similar improvements for other set_info RPCs.
4279 Description: excessive CPU usage during initial read phase on client
4280 Details : During the initial read phase on a client, it would agressively
4281 retry readahead on the file, consuming too much CPU and impacting
4282 performance (since 1.4.5.8). Improve the readahead algorithm
4283 to avoid this, and also improve some other common cases (read
4284 of small files in particular, where "small" is files smaller than
4285 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4290 Description: MDS crash when receiving packet with unknown intent.
4291 Details : Do not LBUG in unknown intent case, just return -EFAULT
4293 Severity : enhancement
4294 Bugzilla : 9293, 9385
4295 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4296 Details : Do not serialize getattr (non-intent version) and statfs.
4299 Frequency : occasional, when OST network is overloaded/intermittent
4301 Description: client evicted by OST after bulk IO timeout
4302 Details : If a client sends a bulk IO request (read or write) the OST
4303 may evict the client if it is unresposive to its data GET/PUT
4304 request. This is incorrect if the network is overloaded (takes
4305 too long to transfer the RPC data) or dropped the OST GET/PUT
4306 request. There is no need to evict the client at all, since
4307 the pinger and/or lock callbacks will handle this, and the
4308 client can restart the bulk request.
4311 Frequency : Always when mmapping file with no objects
4313 Description: client crashes when mmapping file with no objects
4314 Details : Check that we actually have objects in a file before doing any
4315 operations on objects in ll_vm_open, ll_vm_close and
4321 Description: Request leak when working with deleted CWD
4322 Details : Introduce advanced request refcount tracking for requests
4323 referenced from lustre intent.
4325 Severity : Enhancement
4327 Description: Cache open file handles on client.
4328 Details : MDS now will return special lock along with openhandle, if
4329 requested and client is allowed to hold openhandle, even if unused,
4330 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4331 closing files for every read/write openration.
4333 Severity : Enhancement
4335 Description: Cache open negative dentries on client when possible.
4336 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4337 negative dentries on lock revocation.
4342 Description: Remounting a client read-only wasn't possible with a zconf mount
4343 Details : It wasn't possible to remount a client read-only with llmount.
4345 Severity : enhancement
4346 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4347 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4348 MPICH is included to add Lustre-specific ADIO interfaces.
4349 This is based closely on the UFS ADIO layer and only differs
4350 in file creation, in order to allow the OST striping to be set.
4351 This is user-contributed code and not supported by CFS.
4356 Description: extended inode attributes (immutable, append-only) work improperly
4357 when 2.4 and 2.6 kernels are used on client/server or vice versa
4358 Details : Introduce kernel-independent values for these flags.
4360 Severity : enhancement
4363 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4364 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4365 granularity. Now, Fractional MB granularity is be supported,
4366 this is very useful for low memory system.
4368 Severity : enhancement
4370 Description: Getattr by fid
4371 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4372 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4373 have locks and avoids getattr rpc requests for it on every stat.
4376 Frequency : Always, for filesystems larger than 2TB
4378 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4379 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4380 filesystems larger than 2TB to be mounted with mballoc enabled.
4383 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4385 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4386 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4387 filesystems larger than 2TB is mounted then the kernel will
4388 silently corrupt the start of the filesystem. CONFIG_LBD is
4389 enabled for all CFS-supported kernels, but the possibility of
4390 this happening with a modified kernel config exists.
4392 Severity : enhancement
4394 Description: add client O_DIRECT support for 2.6 kernels
4395 Details : It is now possible to do O_DIRECT reads and writes to files
4396 in the Lustre client mountpoint on 2.6 kernel clients.
4398 Severity : enhancement
4400 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4401 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4402 parallel, not waiting for response from every OST before sending
4403 a rpc to the next OST.
4408 Description: setattr vs write race when updating file timestamps
4409 Details : Client processes that update a file timestamp into the past
4410 right after writing to the file (e.g. tar) it is possible that
4411 the updated file modification time can be reset to the current
4412 time due to a race between processing the setattr and write RPC.
4414 Severity : enhancement
4416 Description: Bring 'lfs find' closer in line with regular Linux find.
4417 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4418 -print0 options and obtains all the needed info through the lustre
4421 Severity : enhancement
4423 Description: support up to 1024 configured devices on one node
4424 Details : change obd_dev array from statically allocated to dynamically
4425 allocated structs as they are first used to reduce memory usage
4430 Description: Flush dirty partially truncated pages during truncate
4431 Details : Immediatelly flush partially truncated pages in filter_setattr,
4432 this way we completely avoid having any pages in page cache on OST
4433 and can retire ugly workarounds during writes to flush such pages.
4438 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4439 Details : protect inode from truncation within vfs_unlink() context
4440 just take a reference before calling vfs_unlink() and release it
4441 when parent's i_sem is free.
4446 Description: last_id value checked outside lock on OST caused LASSERT failure
4447 Details : If there were multiple MDS->OST object precreate requests in
4448 flight, it was possible that the OST's last object id was checked
4449 outside a lock and incorrectly tripped an assertion. Move checks
4450 inside locks, and discard old precreate requests.
4453 Frequency : always, if extents are used on OSTs
4455 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4456 Details : OSTs using the extents format would not zero the high 16 bits of
4457 the index physical block number. This is not a problem for any
4458 OST filesystems smaller than 16TB, and no kernels support ext3
4459 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4460 new/modified files) and can be fixed for existing filesystems
4461 with e2fsprogs-1.39-cfs1.
4466 Description: import connection selection may be incorrect if timer wraps
4467 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4468 import connections to be ignored if the 32-bit jiffies counter
4469 wraps. Use a 64-bit jiffies counter.
4472 Frequency : very large clusters immediately after boot
4474 Description: LNET request buffers exhausted under heavy short-term load
4475 Details : If a large number of client requests are generated on a service
4476 that has previously never seen so many requests it is possible
4477 that the request buffer growth cannot keep up with the spike in
4478 demand. Instead of dropping incoming requests, they are held in
4479 the LND until the RPC service can accept more requests.
4482 Frequency : Sometimes during replay
4484 Description: Assertion failure in ll_local_open after replay.
4485 Details : If replay happened on an open request reply before we were able
4486 to set replay handler, reply will become not swabbed tripping the
4487 assertion in ll_local_open. Now we set the handler right after
4488 recognising of open request
4491 Frequency : very rare
4493 Description: kernel reports "badness in vsnprintf"
4494 Details : Reading from the "recovery_status" /proc file in small chunks
4495 may cause a negative length in lprocfs_obd_rd_recovery_status()
4496 call to vsnprintf() (which is otherwise harmless). Exit early
4497 if there is no more space in the output buffer.
4499 Severity : enhancement
4501 Description: clear OBD RPC statistics by writing to them
4502 Details : It is now possible to clear the OBD RPC statistics by writing
4503 to the "stats" file.
4508 Description: Client mtime is not the same on different clients after utimes
4509 Details : In some cases, the client was using the utimes() syscall on
4510 a file cached on another node. The clients now validate the
4511 ctime from the MDS + OSTs to determine which one is right.
4516 Description: Inability to activate failout mode
4517 Details : lconf script incorrectly assumed that in pythong string's numeric
4518 value is used in comparisons.
4521 Frequency : always with multiple stripes per file
4523 Description: Inefficient object allocation for mutli-stripe files
4524 Details : When selecting which OSTs to stripe files over, for files with
4525 a stripe count that divides evenly into the number of OSTs,
4526 the MDS is always picking the same starting OST for each file.
4527 Return the OST selection heuristic to the original design.
4532 Description: mount failures may take full timeout to return an error
4533 Details : Under some heavy load conditions it is possible that a
4534 failed mount can wait for the full obd_timeout interval,
4535 possibly several minutes, before reporting an error.
4536 Instead return an error as soon as the status is known.
4538 Frequency : quota enabled and large files being deleted
4540 Description: releasing more than 4GB of quota at once hangs OST
4541 Details : If a user deletes more than 4GB of files on a single OST it
4542 will cause the OST to spin in an infinite loop. Release
4543 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4548 Description: statfs data retrieved from /proc may be stale or zero
4549 Details : When reading per-device statfs data from /proc, in the
4550 {kbytes,files}_{total,free,avail} files, it may appear
4551 as zero or be out of date.
4554 Frequency : systems with MD RAID1 external journal devices
4556 Description: lconf's call to blkid is confused by RAID1 journal devices
4557 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4558 of returning all block devices that match the journal UUID.
4561 Frequency : always, for aggregate stripe size over 4GB
4563 Description: assertion fails when trying to use 4GB stripe size
4564 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4565 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4568 Frequency : always on ppc64
4570 Description: the first write on an ext3 filesystem with mballoc got stuck
4571 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4572 perform endianness conversion.
4574 ------------------------------------------------------------------------------
4576 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4578 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4579 INTEROPERATE with older versions automatically. Please read the
4580 user documentation before upgrading any part of a live system.
4581 * WARNING: Lustre networking configuration changes are required with
4582 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4585 * Support for kernels:
4586 2.6.9-22.0.2.EL (RHEL 4)
4587 2.6.5-7.244 (SLES 9)
4588 2.6.12.6 vanilla (kernel.org)
4591 Severity : enhancement
4592 Bugzilla : 7981/8208
4593 Description: Introduced Lustre Networking (LNET)
4594 Details : LNET is new networking infrastructure for Lustre, it includes
4595 a reorganized network configuration mode (see the user
4596 documentation for full details) as well as support for routing
4597 between different network fabrics. Lustre Networking Devices
4598 (LNDs) for the supported network fabrics have also been
4599 created for this new infrastructure.
4601 Severity : enhancement
4602 Description: Introduced Access control lists
4603 Details : clients can set ACLs on files and directories in order to have
4604 more fine-grained permissions than the standard Unix UGO+RWX.
4605 The MDS must be started with the "-o acl" mount option.
4607 Severity : enhancement
4608 Description: Introduced filesystem quotas
4609 Details : Administrators may now establish per-user quotas on the
4612 Severity : enhancement
4614 Description: Configuration change for the XT3
4615 The PTLLND is now used to run Lustre over Portals on the XT3
4616 The configure option(s) --with-cray-portals are no longer used.
4617 Rather --with-portals=<path-to-portals-includes> is used to
4618 enable building on the XT3. In addition to enable XT3 specific
4619 features the option --enable-cray-xt3 must be used.
4624 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4625 Details : A race between allocating a new llcd and re-getting the llcd_lock
4626 allowed another thread to grab newly-allocated llcd.
4628 Severity : enhancement
4630 Description: 2.6 OST async journal commit and locking fix to improve performance
4631 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4632 2.6 kernels are now async as they already were in 2.4 kernels so
4633 that they can commit concurrently with the network bulk transfer.
4634 For block-allocated files the filter allocation semaphore is held
4635 to avoid filesystem fragmentation during allocation. BKL lock
4636 removed for 2.6 xattr operations where it is no longer needed.
4641 Description: lconf incorrectly determined whether two IP networks could talk
4642 Details : In some more complicated routing and multiple-network
4643 configurations, lconf will avoid trying to make a network
4644 connection to a disjoint part of the IP space. It was doing the
4645 math incorrectly for one set of cases.
4650 Description: Fix for potential infinite loop processing records in an llog.
4651 Details : If an llog record is corrupted/zeroed, it is possible to loop
4652 forever in llog_process(). Validate the llog record length
4653 and skip the remainder of the block on error.
4656 Frequency : occasional (liblustre only)
4658 Description: liblustre could not open files whose last component is a symlink
4659 Details : sysio_path_walk() would incorrectly pass the open intent to
4660 intermediate path components.
4663 Frequency : rare (liblustre only with non-standard tuning)
4664 Bugzilla : 7201 (7350)
4665 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4666 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4667 after it has been released, only do this for referenced locks
4669 Severity : enhancement
4671 Description: specifying an (invalid) directory default stripe_size of -1
4672 would reset the directory default striping
4673 Details : stripe_size -1 was used internally to signal directory stripe
4674 removal, now use "all default" to signal dir stripe removal
4675 as a directory striping of "all default" is not useful
4678 Frequency : common for large clusters running liblustre clients
4680 Description: doing an ls when liblustre clients are running is slow
4681 Details : sending a glimpse AST to a liblustre client waits for every AST
4682 to time out, as liblustre clients will not respond. Since they
4683 cannot cache data we refresh the OST lock LVB from disk instead.
4685 Severity : enhancement
4687 Description: doing an ls at the same time as file IO can be slow
4688 Details : enqueue and other "small" requests can be blocked behind many
4689 large IO requests. Create a new OST IO portal for non-IO
4690 requests so they can be processed faster.
4693 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4695 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4696 created file with mode 000
4697 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4698 allow an ftruncate() as a non-root user to a file with mode 000.
4699 Lustre now respects this flag to disable mode checks when
4700 truncating a file owned by the user
4703 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4706 Description: Revoking locks from clients that went dead or catatonic might take
4708 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4709 cancellation of such locks instant on servers without waiting for
4710 any reply from clients. Clients drops these locks when cancel
4711 notification from server is received without replying.
4714 Frequency : liblustre-only, when liblustre client dies or becomes busy
4716 Description: Doing ls on Linux clients can take a long time with active
4718 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4719 granting such locks to it in the first place if possible. Locks
4720 are taken by proxy on the OST during the read or write and
4721 dropped immediately afterward. Add connect flags handling, do
4722 not grant locks to liblustre clients for glimpse ASTs.
4724 Severity : enhancement
4726 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4727 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4728 sequential reads and behaves badly with applications doing random
4729 reads. Improve it by reading ahead at least read region, and
4730 avoiding excessive large RPC for small reads.
4732 Severity : enhancement
4734 Description: Creating more than 1000 files for a single job may cause a load
4735 imbalance on the OSTs if there are also a large number of OSTs.
4736 Details : qos_prep_create() uses an OST index reseed value that is an
4737 even multiple of the number of available OSTs so that if the
4738 reseed happens in the middle of the object allocation it will
4739 still utilize the OSTs as uniformly as possible.
4744 Description: OST or MDS may oops in ping_evictor_main()
4745 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4746 but doesn't restart at beginning of obd_exports_timed list
4749 Severity : enhancement
4751 Description: improve by-nid export eviction on the MDS and OST
4752 Details : allow multiple exports with the same NID to be evicted at one
4753 time without re-searching the exports list.
4756 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4758 Description: MDS may oops in groups_free()
4759 Details : in rare race conditions a newly allocated group_info struct is
4760 freed again, and this can be NULL. The 2.4 compatibility code
4761 for groups_free() checked for a NULL pointer, but 2.6 did not.
4764 Frequency : common for liblustre clients doing little filesystem IO
4765 Bugzilla : 9352, 7313
4766 Description: server may evict liblustre clients accessing contended locks
4767 Details : if a client is granted a lock or receives a completion AST
4768 with a blocking AST already set it would not reply to the AST
4769 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4772 Frequency : lfs setstripe, only systems with more than 160 OSTs
4774 Description: unable to set striping with a starting offset beyond OST 160
4775 Details : llapi_create_file() incorrectly limited the starting stripe
4776 index to the maximum single-file stripe count.
4779 Frequency : LDAP users only
4781 Description: lconf did not handle in-kernel recovery with LDAP properly
4782 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4784 Severity : enhancement
4786 Description: bind OST threads to NUMA nodes to improve performance
4787 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4788 node and do their allocations there to localize memory access
4790 Severity : enhancement
4792 Description: llmount can determine client NID directly from Myrinet (GM)
4793 Details : the client NID code from gmnalnid was moved directly into
4794 llmount, removing the need to use this or specifying the
4795 client NID explicitly when mounting GM clients with zeroconf
4798 Frequency : if client is started with down MDS
4800 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4801 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4802 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4803 ptlrpc_import_delayed_req() because we didn't check if the
4804 request was interrupted, and we also didn't break out of the
4805 event loop if there was no timeout
4810 Description: data loss during non-page-aligned writes to a single file from
4811 both multiple nodes and multiple threads on one node at same time
4812 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4813 inconsistency led to false short-reads, that were cached and later
4814 used by ->prepare_write() to fill in partially written page,
4815 leading to data loss.
4818 Frequency : always, if lconf --abort_recovery used
4820 Description: lconf --abort_recovery fails with 'Operation not supported'
4821 Details : lconf was attempting to abort recovery on the MDT device and not
4824 Severity : enhancement
4826 Description: remove cleanup logs
4827 Details : replace lconf-generated cleanup logs with lustre internal
4828 cleanup routines. Eliminates the need for client-cleanup and
4831 Severity : enhancement
4833 Description: add support for EAs (user and system) on lustre filesystems
4834 Details : it is now possible to store extended attributes in the Lustre
4835 client filesystem, and with the user_xattr mount option it
4836 is possible to allow users to store EAs on their files also
4838 Severity : enhancement
4840 Description: Add possibility (config option) to show minimal available OST free
4842 Details : When compiled with --enable-mindf configure option, statfs(2)
4843 (and so, df) will return least minimal free space available from
4844 all OSTs as amount of free space on FS, instead of summary of
4845 free spaces of all OSTs.
4847 Severity : enhancement
4849 Description: do not expand extent locks acquired on OST-side
4850 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4851 by server: they are not cached anyway.
4854 Frequency : when mmap is used/binaries executed from Lustre
4856 Description: Unmmap pages before throwing them away from read cache.
4857 Details : llap_shrink cache now attempts to unmap pages before discarding
4858 them (if unmapping failed - do not discard). SLES9 kernel has
4859 extra checks that trigger if this unmapping is not done first.
4864 Description: lconf didn't resolve symlinks before checking to see whether a
4865 given mountpoint was already in use
4868 Frequency : when migrating failover services
4869 Bugzilla : 6395, 9514
4870 Description: When migrating a subset of services from a node (e.g. failback
4871 from a failover service node) the remaining services would
4872 time out and evict clients.
4873 Details : lconf --force (implied by --failover) sets the global obd_timeout
4874 to 5 seconds in order to quickly disconnect, but this caused
4875 other RPCs to time out too quickly. Do not change the global
4876 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4878 Severity : enhancement
4879 Frequency : if MDS is started with down OST
4880 Bugzilla : 9439,5706
4881 Description: Allow startup/shutdown of an MDS without depending on the
4882 availability of the OSTs.
4883 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4884 Add appropriate locking and lov-osc refcounts for safe
4885 cleaning. Add osc abort_inflight calls in case the
4886 synchronize never started.
4889 Frequency : occasional (Cray XT3 only)
4891 Description: root not authorized to access files in CRAY_PORTALS environment
4892 Details : The client process capabilities were not honoured on the MDS in
4893 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4894 been accessed by an authorized user then root was able to access
4895 the file on the local client also. The root user capabilities
4896 are now allowed on the MDS, as this environment has secure UID.
4899 Frequency : occasional
4901 Description: ldiskfs "too long searching" message happens too often
4902 Details : A debugging message (otherwise harmless) prints too often on
4903 the OST console. This has been reduced to only happen when
4904 there are fragmentation problems on the filesystem.
4909 Description: Division by zero in statfs when all OSCs are inactive
4910 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4911 lov_statfs divides by value returned by lov_get_stripecnt().
4915 Bugzilla : 9489, 3273
4916 Description: First write from each client to each OST was only 4kB in size,
4917 to initialize client writeback cache, which caused sub-optimal
4918 RPCs and poor layout on disk for the first written file.
4919 Details : Clients now request an initial cache grant at (re)connect time
4920 and so that they can start streaming writes to the cache right
4921 away and always do full-sized RPCs if there is enough data.
4922 If the OST is rebooted the client also re-establishes its grant
4923 so that client cached writes will be honoured under the grant.
4928 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4929 Details : Now I/O RPCs go to different portal number and (presumably) fast
4930 lock requests (and glimses) and other RPCs get their own service
4931 threads pool that should be able to service those RPCs
4934 Severity : enhancement
4936 Description: Ability to exchange lustre version between client and servers and
4937 issue warnings at client side if client is too old. Also for
4938 liblustre clients there is ability to refuse connection of too old
4940 Details : New 'version' field is added to connect data structure that is
4941 filled with version info. That info is later checked by server and
4945 Frequency : rare, liblustre only.
4946 Bugzilla : 9296, 9581
4947 Description: Two simultaneous writes from liblustre at offset within same page
4948 might proceed at the same time overwriting eachother with stale
4950 Details : I/O lock withing llu_file_prwv was released too early, before data
4951 actually was hitting the wire. Extended lock-holding time until
4952 server acknowledges receiving data.
4955 Frequency : extremely rare. Never observed in practice.
4957 Description: avoid generating lustre_handle cookie of 0.
4958 Details : class_handle_hash() generates handle cookies by incrementing
4959 global counter, and can hit 0 occasionaly (this is unlikely, but
4960 not impossible, because initial value of cookie counter is
4961 selected randonly). Value of 0 is used as a sentinel meaning
4962 "unassigned handle" --- avoid it. Also coalesce two critical
4963 sections in this function into one.
4965 Severity : enhancement
4967 Description: allow liblustre clients to delegate truncate locking to OST
4968 Details : To avoid overhead of locking, liblustre client instructs OST to
4969 take extent lock in ost_punch() on client's behalf. New connection
4970 flag is added to handle backward compatibility.
4972 Severity : enhancement
4973 Bugzilla : 4928, 7341, 9758
4974 Description: allow number of OST service threads to be specified
4975 Details : a module parameter allows the number of OST service threads
4976 to be specified via "options ost ost_num_threads={N}" in the
4977 OSS's /etc/modules.conf or /etc/modprobe.conf.
4981 Bugzilla : 6146, 9635, 9895
4982 Description: servers crash with bad pointer in target_handle_connect()
4983 Details : In rare cases when a client is reconnecting it was possible that
4984 the connection request was the last reference for that export.
4985 We would temporarily drop the export reference and get a new
4986 one, but this may have been the last reference and the export
4987 was just destroyed. Get new reference before dropping old one.
4989 Severity : enhancement
4990 Frequency : if client is started with failover MDS
4992 Description: Allow multiple MDS hostnames in the mount command
4993 Details : Try to read the configuration from all specified MDS
4994 hostnames during a client mount in case the "primary"
4997 Severity : enhancement
4999 Description: Stop sending data to evicted clients as soon as possible.
5000 Details : Check if the client we are about to send or are sending data to
5001 was evicted already. (Check is done every second of waiting,
5002 for which l_wait_event interface was extended to allow checking
5003 of exit condition at specified intervals).
5006 Frequency : rare, normally only when NFS exporting is done from client
5008 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5010 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5013 Frequency : always, if srand() is called before liblustre initialization
5015 Description: Liblustre uses system PRNG disturbing its usage by user application
5016 Details : Introduce internal to lustre fast and high-quality PRNG for
5017 lustre usage and make liblustre and some other places in generic
5018 lustre code to use it.
5020 Severity : enhancement
5021 Bugzilla : 9477, 9557, 9870
5022 Description: Verify that the MDS configuration logs are updated when xml is
5023 Details : Check if the .xml configuration logs are newer than the config
5024 logs stored on the MDS and report an error if this is the case.
5025 Request --write-conf, or allow starting with --old_conf.
5027 Severity : enhancement
5029 Description: Handle symlinks in the path when checking if Lustre is mounted.
5030 Details : Resolve intermediate symlinks when checking if a client has
5031 mounted a filesystem to avoid duplicate client mounts.
5036 Description: lconf can hit an error exception but still return success.
5037 Details : The lconf command catches the Command error exception at the top
5038 level script context and will exit with the associated exit
5039 status, but doesn't ensure that this exit status is non-zero.
5044 Description: failure of ptlrpc thread startup can cause oops
5045 Details : Starting a ptlrpc service thread can fail if there are a large
5046 number of threads or the server memory is very fragmented.
5047 Handle this without oopsing.
5050 Frequency : always, only if liblustre and non-default acceptor port was used
5052 Description: liblustre cannot connect to servers with non-default acceptor port
5053 Details : tcpnal_set_default_params() was not called and was therefore
5054 ignoring the environment varaible TCPNAL_PORT, as well as other
5055 TCPNAL_ environment variables
5060 Description: two objects could be created on the same OST for a single file
5061 Details : If an OST is down, in some cases it was possible to create two
5062 objects on a single OST for a single file. No problems other
5063 than potential performance impact and spurious error messages.
5067 Bugzilla : 5681, 9562
5068 Description: Client may oops in ll_unhash_aliases
5069 Details : Client dcache may become inconsistent in race condition.
5070 In some cases "getcwd" can fail if the current directory is
5076 Description: Inode refcounting problems in NFS export code
5077 Details : link_raw functions used to call d_instantiate without obtaining
5078 extra inode reference first.
5082 Bugzilla : 9942, 9903
5083 Description: Referencing freed requests leading to crash, memleaks with NFS.
5084 Details : We used to require that call to ll_revalidate_it was always
5085 followed by ll_lookup_it. Also with revalidate_special() it is
5086 possible to call ll_revalidate_it() twice for the same dentry
5087 even if first occurence returned success. This fix changes semantic
5088 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5089 reference on a request referred from the intent.
5090 ll_intent_release() then releases such a request.
5093 Frequency : rare, normally benchmark loads only
5095 Description: unlinked inodes were kept in memory on the client
5096 Details : If a client is repeatedly creating and unlinking files it
5097 can accumulate a lot of stale inodes in the inode slab cache.
5098 If there is no other client load running this can cause the
5099 client node to run out of memory. Instead flush old inodes
5100 from client cache that have the same inode number as a new inode.
5103 Frequency : SLES9 2.6.5 kernel and long filenames only
5104 Bugzilla : 9969, 10379
5105 Description: utime reports stale NFS file handle
5106 Details : SLES9 uses out-of-dentry names in some cases, which confused
5107 the lustre dentry revalidation. Change it to always use the
5111 Frequency : rare, unless heavy write-truncate concurrency is continuous
5112 Bugzilla : 4180, 6984, 7171, 9963, 9331
5113 Description: OST becomes very slow and/or deadlocked during object unlink
5114 Details : filter_destroy() was holding onto the parent directory lock
5115 while truncating+unlinking objects. For very large objects this
5116 may block other threads for a long time and slow overall OST
5117 responsiveness. It may also be possible to get a lock ordering
5118 deadlock in this case, or run out of journal credits because of
5119 the combined truncate+unlink. Solution is to do object truncate
5120 first in one transaction without parent lock, and then do the
5121 final unlink in a new transaction with the parent lock. This
5122 reduces the lock hold time dramatically.
5125 Frequency : rare, 2.4 kernels only
5127 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5128 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5129 before unmounting their filesystem in order to avoid deadlock.
5130 The kernel_locked() function in 2.4 kernels only checks whether
5131 the kernel lock is held, not whether it is this process that is
5132 holding it as 2.6 kernels do.
5137 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5138 Details : The client ptlrpc code may be trying to reconnect to a down
5139 server before a previous connection attempt has timed out.
5140 Increase the reconnect interval to be longer than the connection
5141 timeout interval to avoid sending duplicate connections to
5145 Frequency : echo_client brw_test command
5147 Description: fix echo_client to work with OST preallocated code
5148 Details : OST preallocation code (5137) didn't take echo_client IO path
5149 into account: echo_client calls filter methods outside of any
5150 OST thread and, hence, there is no per-thread preallocated
5151 pages and buffers to use. Solution: hijack pga pages for IO. As
5152 a byproduct, this avoids unnecessary data copying.
5156 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5157 Description: Client can oops in mdc_commit_close() after open replay
5158 Details : It was possible for the MDS to return an open request with no
5159 transaction number in mds_finish_transno() if the client was
5160 evicted, but without actually returning an error. Clients
5161 would later try to replay that open and may trip an assertion
5162 Simplify the client close codepath, and always return an error
5163 from the MDS in case the open is not successful.
5166 Frequency : rare, 2.6 OSTs only
5168 Description: OST may deadlock under high load on fragmented files
5169 Details : If there was a heavy load and highly-fragmented OST filesystems
5170 it was possible to have all the OST threads deadlock waiting on
5171 allocation of biovecs, because the biovecs were not released
5172 until the entire RPC IO was completed. Instead, release biovecs
5173 as soon as they are complete to ensure forward IO progress.
5175 Severity : enhancement
5177 Description: Support for specifying external journal device at mount
5178 Details : If an OST or MDS device is formatted with an external journal
5179 device, this device major/minor is stored in the ext3 superblock
5180 and may not be valid for failover. Allow detecting and
5181 specifying the external journal at mount time.
5186 Description: Mounting an MDS with pending unlinked files may cause oops
5187 Details : target_finish_recovery() calls mds_postrecov() which returned
5188 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5189 considers this an error and immediately begins cleaning up the
5190 lov, just after starting the mds_lov process
5192 Severity : enhancement
5194 Description: Implement 'lfs df' to report actual free space on per-OST basis
5195 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5196 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5197 usage of objects; '-h' to report in human readable format.
5199 ------------------------------------------------------------------------------
5201 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5208 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5209 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5210 at startup during orphan unlinking. Instead, skip buddy update
5211 before setup, as it will be regenerated after recovery is complete.
5216 Description: If an OST is inactive, its locks might reference stale inodes.
5217 Details : lov_change_cbdata() must iterate over all namespaces, even if
5218 they are inactive to clear inode references from the lock.
5220 Severity : enhancement
5221 Frequency : occasional, if non-standard max_dirty_mb used
5223 Description: Client will block write RPCs if not enough grant
5224 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5225 then the client will block writes while waiting for another RPC
5226 to complete instead of consuming its dirty limit. With change
5227 we get improved performance when max_dirty_mb is small.
5229 Severity : enhancement
5230 Bugzilla : 3389, 6253
5231 Description: Add support for supplementary groups on the MDS.
5232 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5233 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5234 lookups for user supplementary groups into a cache.
5238 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5239 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5240 for liblustre O_EXCL is enforced by MDS.
5244 Description: readdir never returns NULL in liblustre.
5245 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5246 dentry in struct dirent.
5250 Frequency : liblustre only, depends on application IO pattern
5251 Description: liblustre clients evicted if not contacting servers
5252 Details : Don't put liblustre clients into the ping_evictor list, so
5253 they will not be evicted by the pinger ever.
5255 Severity : enhancement
5257 Description: Add ability to evict clients by NID from MDS.
5258 Details : By echoing "nid:$NID" string into
5259 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5260 $NID would be instantly evicted from this MDS and from all active
5261 OSTs connected to it.
5265 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5266 Details : lookup_it_finish() used to query file size from OSTs that was not
5271 Description: service threads change working directory to that of init
5272 Details : Starting lustre service threads may pin the working directory
5273 of the parent thread, making that filesystem busy. Threads
5274 now change to the working directory of init to avoid this.
5278 Frequency : during shutdown only
5279 Description: shutdown with a failed MDS or OST can cause unmount to hang
5280 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5285 Frequency : 2.6 only
5286 Description: chmod/chown may include an extra supplementary group
5287 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5288 supplementary group and if none is specified this is used.
5291 Bugzilla : 5479 (6816)
5293 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5294 Details : If lookup is in progress on a file that is unlinked we might try
5295 to revalidate the inode and fail in revalidate after lookup is
5296 complete and ll_file_open() enqueues the open again but
5297 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5300 Frequency : always, if lconf --abort_recovery used
5302 Description: lconf --abort_recovery fails with 'Operation not supported'
5303 Details : lconf was attempting to abort recovery on the MDT device and not
5306 ------------------------------------------------------------------------------
5308 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5313 Frequency : rare (only unsupported configurations with a node running as an
5315 Bugzilla : 6514, 5137
5316 Description: Mounting a Lustre file system on a node running as an OST could
5318 Details : OSTs now preallocates memory needed to write out data at
5319 startup, instead of when needed, to avoid having to
5320 allocate memory in possibly low memory situations.
5321 Specifically, if the file system is mounted on on OST,
5322 memory pressure could force it to try to write out data,
5323 which it needed to allocate memory to do. Due to the low
5324 memory, it would be unable to do so and the node would
5325 become unresponsive.
5327 Severity : enhancement
5329 Description: Addition of lconf --service command line option
5330 Details : lconf now accepts a '--service <arg>' option, which is
5331 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5333 Severity : enhancement
5335 Description: Failover mode is now the default for OSTs.
5336 Details : By default, OSTs will now run in failover mode. To return to
5337 the old behaviour, add '--failout' to the lmc line for OSTs.
5339 Severity : enhancement
5341 Description: Health checks are now provided for MDS and OSTs
5342 Details : Additional detailed health check information on MSD and OSTs
5343 is now provided through the procfs health_check value.
5346 Frequency : occasional, depends on IO load
5348 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5349 after numerous create/delete cycles
5350 Details : The ext3 inode allocation policy would not allocate new inodes
5351 very well on the OSTs because there are no new directories
5352 being created. Instead we look for groups with free space if
5353 the parent directories are nearly full.
5358 Description: Network or server problems during mount may cause partially
5359 mounted clients instead of returning an error.
5360 Details : The config llog parsing code may overwrite the error return
5361 code during mount error handling, returning success instead
5367 Description: MDS can fail to allocate large reply buffers
5368 Details : After long uptimes the MDS can fail to allocate large reply
5369 buffers (e.g. zconf client mount config records) due to memory
5370 fragmentation or consumption by the buffer cache. Preallocate
5371 some large reply buffers so that these replies can be sent even
5372 under memory pressure.
5376 Frequency : rare (liblustre)
5377 Description: fsx running with liblustre complained that using truncate() to
5378 extend the file doesn't work. This patch corrects that issue.
5379 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5380 ATTR_SIZE and lsm use in llu_setattr_raw.
5384 Frequency : rare, only 2.6 kernels
5385 Description: Unusual file access patterns on the MDS may result in inode
5386 data being lost in very rare circumstances.
5387 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5388 code in the 2.6 kernel caused the inode and/or EA data not to
5389 be read from disk, causing single-file corruption.
5393 Frequency : rare, only 2.6 filesystems using extents
5394 Description: Heavy concurrent write and delete load may cause data corruption.
5395 Details : It was possible under high-load situations to have an extent
5396 metadata block in the block device cache from a just-unlinked
5397 file overwrite a newly-allocated data block. We now unmap any
5398 metadata buffers that alias just-allocated data blocks.
5402 Frequency : filesystems with default stripe_count larger than 77
5403 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5404 Details : lconf specifies an inode size of 4096 bytes when the default
5405 stripe_count is larger than 77. This conflicts with the default
5406 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5407 this case to avoid pinning too much memory for large EAs.
5409 ------------------------------------------------------------------------------
5411 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5416 Frequency : rare (extremely heavy IO load with hundreds of clients)
5418 Description: Client is evicted, gets IO error writing to file
5419 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5420 caused clients to be evicted for AST timeouts. The fixes in
5421 bug 5192 mean we no longer need to have such short AST timeouts
5422 so ldlm_timeout has been increased.
5425 Frequency : occasional during --force or --failover shutdown under load
5426 Bugzilla : 5949, 4834
5427 Description: Server oops/LBUG if stopped with --force or --failover under load
5428 Details : a collection of import/export refcount and cleanup ordering
5429 issues fixed for safer force cleanup
5432 Frequency : only filesystems larger than 120 OSTs
5433 Bugzilla : 5990, 6223
5434 Description: lfs getstripe would oops on a very large filesystem
5435 Details : lov_getconfig used kfree on vmalloc'd memory
5438 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5439 Bugzilla : 6242, 6243
5440 Description: reading from files that had been truncated to a non-zero size
5441 but never opened returned no data
5442 Details : ll_file_read() reads zeros from no-object files to EOF
5447 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5448 Details : There's little harm in aborting MDS/OSS recovery and letting it
5449 try again, so I removed the LASSERT and return an error instead.
5451 Severity : enhancement
5453 Description: New debugging infrastructure for tracking down data corruption
5454 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5455 (b) cover more of the client-side code path, and (c) try to narrow
5456 down where problems occurred
5460 Bugzilla : 3819, 4364, 4397, 6313
5461 Description: Racing close and eviction MDS could cause assertion in mds_close
5462 Details : It was possible to get multiple mfd references during close and
5463 client eviction, leading to one thread referencing a freed mfd.
5465 Severity: : enhancement
5466 Bugzilla : 3262, 6359
5467 Description: Attempts to reconnect to servers are now more aggressive.
5468 Details : This builds on the enhanced upcall-less recovery that was added
5469 in 1.4.2. When trying to reconnect to servers, clients will
5470 now try each server in the failover group every 10 seconds. By
5471 default, clients would previously try one server every 25 seconds.
5476 Description: After recovery, certain operations trigger a failed
5477 assertion on a client.
5478 Details : Failing over an mds, using lconf -d --failover, while a
5479 client was doing a readdir() call would cause the client to
5480 LBUG after recovery completed and the readdir() was resent.
5482 Severity : enhancement
5484 Description: Default groups are now added by lconf
5485 Details : You can now run lconf --group <servicename> without having to
5486 manually add groups with lmc.
5489 Frequency : occasional
5491 Description: Nodes with an elan id of 0 trigger a failed assertion
5494 Frequency : always when accessing e.g. tty/console device nodes
5496 Description: tty and some other devices nodes cannot be used on lustre
5497 Details : file's private_data field is used by device data and lustre
5498 values in there got lost. New field was added to struct file to
5499 store fs-specific private data.
5502 Frequency : when exporting Lustre via NFS
5504 Description: NFSD failed occasionally when looking up a path component
5505 Details : NFSD is looking up ".." which was broken in ext3 directories
5506 that had grown large enough to become hashed.
5509 Frequency : Clusters with multiple interfaces not on the same subnet
5511 Description: Nodes will repeatedly try to reconnect to an interface which it
5512 cannot reach and report an error to the log.
5513 Details : Extra peer list entries will be created by lconf with some peers
5514 unreachable. lconf now validates the peer before adding it.
5517 Frequency : Only if a default stripe is set on the filesystem root.
5519 Description: Setting a default stripe on the filesystem root prevented the
5520 filesystem from being remounted.
5521 Details : The client was sending extra request flags in the root getattr
5522 request and did not allocate a reply buffer for the dir EA.
5525 Frequency : occasional, higher if lots of files are accessed by one client
5526 Bugzilla : 6159, 6097
5527 Description: Client trips assertion regarding lsm mismatch/magic
5528 Details : While revalidating inodes the VFS looks up inodes with ifind()
5529 and in rare cases can find an inode that is being freed.
5530 The ll_test_inode() code will free the lsm during ifind()
5531 when it finds an existing inode and then the VFS later attaches
5532 this free lsm to a new inode.
5536 Bugzilla : 6422, 7030
5537 Description: MDS deadlock between mkdir and client eviction
5538 Details : Creating a new file via mkdir or mknod (starting a transaction
5539 and getting the ns lock) can deadlock with client eviction
5540 (gets ns lock and trying to finish a synchronous transaction).
5543 Frequency : occasional
5544 Description: While starting a server, the fsfilt_ext3 module could not be
5546 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5547 kernels. Previously, lconf would still use the ext3 name
5548 when trying to load modules. Now, it will correctly use
5549 ext3 on 2.4 and ldiskfs on 2.6.
5551 Severity : enhancement
5552 Description: The default stripe count has been changed to 1
5553 Details : The interpretation of the default stripe count (0, to lfs
5554 or lmc) has been changed to mean striping across a single
5555 OST, rather than all available. For general usage we have
5556 found a stripe count of 1 or 2 works best.
5558 Severity : enhancement
5559 Description: Add support for compiling against Cray portals.
5560 Details : Conditional compiling for some areas that are different
5564 Frequency : occasional
5565 Bugzilla : 6409, 6834
5566 Description: Creating files with an explicit stripe count may lead to
5567 a failed assertion on the MDS
5568 Details : If some OSTs are full or unavailable, creating files may
5569 trigger a failed assertion on the MDS. Now, Lustre will
5570 try to use other servers or return an error to the
5574 Frequency : occasional
5576 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5577 caused lock timeout during readahead (since 1.4.2).
5578 Details : Processes doing readahead might match a lock that hasn't been
5579 granted yet if there are overlapping and conflicting lock
5580 requests. The readahead process waits on ungranted lock
5581 (original lock is CBPENDING), while OST waits for that process
5582 to cancel CBPENDING read lock and eventually evicts client.
5584 Severity : enhancement
5586 Description: Initial enabling of flock support for clients
5587 Details : Implements fcntl advisory locking and file status functions.
5588 This feature is provided as an optional mount flag (default
5589 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5590 locking are implemented yet, and those that are are not guaranteed
5591 to be completely correct in production environments.
5592 mount -t lustre -o [flock|noflock] ...
5595 Frequency : occasional
5597 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5598 assertion in the ext3 JBD (journaling) layer.
5599 Details : The b_committed_data struct is protected by the big kernel lock
5600 in 2.4 kernels, serializing journal_commit_transaction() and
5601 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5602 access to this struct. In 2.6 kernels there is finer grained
5603 locking to improve SMP performance of the JBD layer.
5607 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5612 Description: merge and cleanup kernel patches.
5615 -----------------------------------------------------------------------------
5617 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5619 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5620 versions of Lustre. Please update all servers and clients to
5621 version 1.4.2 or later at the same time. You must also run
5622 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5623 to update the configuration logs.
5625 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5626 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5627 - swab llog records of type '0' so we get proper header size/idx (5861)
5628 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5629 - fix rename of one directory over another leaking an inode (5953)
5630 - avoid SetPageDirty on 2.6 (5981)
5631 - don't re-add just-being-destroyed locks to the waiting list (5653)
5632 - when creating new directories, inherit the parent's custom
5633 striping settings if present parent (3048)
5634 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5635 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5636 - don't hold client locks on temporary worklist from l_lru (5666)
5637 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5638 - automatically evict dead clients (5921)
5639 - Update file size properly in create+truncate+fstat case (6196)
5640 - Do not unhash mountpoint dentries, do not allow removal of
5642 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5643 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5644 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5645 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5646 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5647 - Correctly handle reads of files with no objects (6243)
5648 - lctl recover will also mark a device active if deactivate used (5933)
5650 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5651 - allow --write-conf on an MDS with different nettype than client (5619)
5652 - don't write config llogs to MDS for mounts not from that MDS (5617)
5653 - lconf should create multiple TCP connections from a client (5201)
5654 - init scripts are now turned off by default; run chkconfig --on
5655 lustre and chkconfig --on lustrefs to use them
5656 - upcalls are no longer needed for clients to recover to failover
5658 - add --abort-recovery option to lconf to abort recovery on device
5660 - add support for an arbitrary number of OSTs (3026)
5661 - Quota support protocol changes.
5662 - forward compatibility changes to wire structs (6007)
5663 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5664 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5665 - improved "lustre status" script
5666 - initialize blocksize for non-regular files (6062)
5667 - added --disable-server and --disable-client configure options (5782)
5668 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5669 - Vanilla 2.4.29 support
5670 - increase maximum number of obd devices to 520 (6242)
5671 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5672 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5673 SLES 9 2.6.5 kernels
5674 - limit stripes per file to 160 (the maximum EA size) (6093)
5676 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5679 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5680 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5681 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5682 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5683 - small ext3 extents cleanups and fixes (5733)
5684 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5685 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5686 - avoid needless client->OST connect, fix handle mismatch (5317)
5687 - fix DLM error path that led to out-of-sync client, long delays (5779)
5688 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5689 - fix several locking issues related to i_size (5492,5624,5654,5672)
5690 - don't move pending lock onto export if it is already evicted (5683)
5691 - fix kernel oops when creating .foo in unlinked directory (5548)
5692 - fix deadlock in obdfilter statistics vs. object create (5811)
5693 - use time_{before,after} to avoid timer jiffies wrap (5882)
5694 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5695 - Do not leak request if server was not able to process it (5154)
5696 - If mds_open unable to find parent dir, make that negative lookup(5154)
5697 - don't create new directories with extent-mapping (5909, 5936)
5699 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5700 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5701 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5702 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5703 - support for RHEL4 kernel on i686 (5773)
5704 - provide error messages when incompatible logs are encountered (5898)
5706 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5707 * version 1.4.0.10 (1.4.1 release candidate 1)
5709 - don't keep a lock reference when lock is not granted (4238)
5710 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5711 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5712 - drop import inflight refcount on signal_completed_replay error (5255)
5713 - unlock page after async write error during send (3677)
5714 - handle missing objects in filter_preprw_read properly (5265)
5715 - no transno return for symlink open, don't save no-trasno open (3440)
5716 - don't try to complete elan receive that already failed (4012)
5717 - free RPC server reply state on error (5406)
5718 - clean up thread from ptlrpc_start_thread() on error (5160)
5719 - readahead could read extra page into cache that wasn't ejected (5388)
5720 - prevent races in class_attach/setup/cleanup/detach (5260)
5721 - don't dereference de->d_inode after l_dput of de (5458)
5722 - use "int" for stripe value returned from lock_to_stripe (5544)
5723 - mballoc allocation and error-checking fixes in 2.6 (5504)
5724 - block device patches to fix I/O request sizes in 2.6 (5482)
5725 - look up hostnames for IB nals (5602)
5726 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5727 - don't start multiple acceptors for the same port (5277)
5728 - fix incorrect LASSERT in mds_getattr_name (5635)
5729 - export a proc file for general "ping" checking (5628)
5730 - fix "lfs check" to not block when the MDS is down (5628)
5732 - service request history (4965)
5733 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5734 - create an "evict_client" /proc entry on OSTs, like the MDS has
5735 - fix mount usage message, return errors per mount(8) (5168)
5736 - change grep [] to grep "[]" in tests so they work in more UMLs
5737 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5738 - remove ancient LOV_MAGIC_V0 EA support (5047)
5739 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5740 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5741 - allow readahead allocations to fail when low on memory (5383)
5742 - mmap locking landed again, after considerable improvement (2828)
5743 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5745 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5748 - send OST transaction number in read/write reply to free req (4966)
5749 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5750 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5751 - the watchdog thread now runs as interruptible (5246)
5752 - flock/lockf fixes (but it's still disabled, pending 5135)
5753 - don't use EXT3 constants in llite code (5094)
5754 - memory shortage at startup could cause assertion (5176)
5756 - reorganization of lov code
5757 - single portals codebase
5759 - add extents/mballoc support (5025)
5760 - direct I/O reads in the obdfilter (4048)
5761 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5763 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5766 - send OST transaction number in read/write reply to free req (4966)
5767 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5768 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5769 - the watchdog thread now runs as interruptible (5246)
5770 - handle missing objects in filter_preprw_read properly (5265)
5771 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5772 - drop import inflight refcount on signal_completed_replay error (5255)
5773 - unlock page after async write error during send (3677)
5774 - return original error code on reconstructed replies (3761)
5775 - no transno return for symlink open, don't save no-trasno open (3440)
5777 - add pid to ldlm debugging output (4922)
5778 - bump the watchdog timeouts -- we can't handle 30sec yet
5779 - extra debugging for orphan dentry/inode bug (5259)
5781 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5784 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5785 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5786 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5787 - fix NULL dereference in /proc/sys/portals/routes (4827)
5788 - allow failed mdc_close() operations to be interrupted (4561)
5789 - stop precreate on OST before MDS would time out on it (4778)
5790 - don't send partial-page writes before EOF from client (4410)
5791 - discard client grant for sub-page writes on large-page clients (4520)
5792 - don't free dentries not owned by NFS code, check generation (4806)
5793 - fix lsm leak if mds_create_objects() fails (4801)
5794 - limit debug_daemon file size, always print CERROR messages (4789)
5795 - use transno after validating reply (3892)
5796 - process timed out requests if import state changes (3754)
5797 - update mtime on OST during writes, return in glimpse (4829)
5798 - add mkfsoptions to LDAP (4679)
5799 - use ->max_readahead method instead of zapping global ra (5039)
5800 - don't interrupt __l_wait_event() during strace
5802 - add software watchdogs to catch hung threads quickly (4941)
5803 - make lustrefs init script start after nfs is mounted
5804 - fix CWARN/ERROR duplication (4930)
5805 - return async write errors to application if possible (2248)
5806 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5807 - print NAL number in %x format (4645)
5808 - update barely-supported suse-2.4.21-171 series (4842)
5809 - support for sles 9 %post scripts
5810 - support for building 2.6 kernel-source packages
5811 - support for sles km_* packages
5813 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5816 - ignore -ENOENT errors in osc_destroy (3639)
5817 - notify osc create thread that OSC is being cleaned up (4600)
5818 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5819 - reconstruct ost_handle() like mds_handle() (4657)
5820 - create a new thread to do import eviction to avoid deadlock (3969)
5821 - let lconf resolve symlinked-to devices (4629)
5822 - don't unlink "objects" from directory with default EA (4554)
5823 - hold socknal file ref over connect in case target is down (4394)
5824 - allow more than 32000 subdirectories in a single directory (3244)
5825 - fix blocks count for O_DIRECT writes (3751)
5826 - OST returns ENOSPC from object create when no space left (4539)
5827 - don't send truncate RPC if file size isn't changing (4410)
5828 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5829 - bind to privileged port in socknal and tcpnal (3689)
5831 - rate limit CERROR/CWARN console message to avoid overload (4519)
5832 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5833 - basic mmap support (3918)
5834 - kernel patch series update from b1_4 (4711)
5836 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5839 - avoid crash during MDS cleanup with OST shut down (2775)
5840 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5841 - don't use bad inodes on the MDS (3744)
5842 - dynamic object preallocation to improve recovery speed (4236)
5843 - don't hold spinlock over lock dumping or change debug flags (4401)
5844 - don't zero obd_dev when it is force cleaned (3651)
5845 - print grants to console if they go negative (4431)
5846 - "lctl deactivate" will stop automatic recovery attempts (3406)
5847 - look for existing locks in ldlm_handle_enqueue() (3764)
5848 - don't resolve lock handle twice in recovery avoiding race (4401)
5849 - revalidate should check working dir is a directory (4134)
5851 - don't always mark "slow" obdfilter messages as errors (4418)
5853 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5856 - don't close LustreDB during write_conf until it is done (3860)
5857 - fix typo in lconf for_each_profile (3821)
5858 - allow dumping logs from multiple threads at one time (3820)
5859 - don't allow multiple threads in OSC recovery (3812)
5860 - fix debug_size parameters (3864)
5861 - fix mds_postrecov to initialize import for llog ctxt (3121)
5862 - replace config semaphore with spinlock (3306)
5863 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5864 - don't allow enqueue to complete on a destroyed export (3822)
5865 - down write_lock before checking llog header bitmap (3825)
5866 - recover from lock replay timeout (3764)
5867 - up llog sem before sending rpc (3652)
5868 - reduce ns lock hold times when setting kms (3267)
5869 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5870 - fix NULL deref and obd_dev leak on setup error (3312)
5871 - replace some LBUG about llog ops with error handling (3841)
5872 - don't match INVALID dentries from d_lookup and spin (3784)
5873 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5874 - fix invalid assertion in ptlrpc_set_wait (3880)
5876 - add libwrap support for the TCP acceptor (3996)
5877 - add /proc/sys/portals/routes for non-root route listing (3994)
5878 - allow setting MDS UUID in .xml (2580)
5879 - print the stack of a process that LBUGs (4228)
5881 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5884 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5885 - make sure to unset replay flag from failed open requests (3440)
5886 - if default stripe count is 0, use OST count for inode size (3636)
5887 - update parent mtime/ctime on client for create/unlink (2611)
5888 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5889 - free recovery state on server during a forced cleanup (3571)
5890 - unregister_reply for resent reqs (3063)
5891 - loop back devices mounting and status check on 2.6 (3563)
5892 - fix resource-creation race that can provoke i_size == 0 (3513)
5893 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5894 - more debugging for page-accounting assertion (3746)
5895 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5896 - avoid deadlock after precreation failure (3758)
5897 - fix race and lock order deadlock in orphan handling (3450, 3750)
5898 - add validity checks when grabbing inodes from l_ast_data (3599)
5900 - add /proc/.../recovery_status to obdfilter (3428)
5901 - lightweight CDEBUG infrastructure, debug daemon (3668)
5902 - change default OSC RPC parameters to be better on small clusters
5903 - turn off OST read cache for files smaller than 32MB
5904 - install man pages and include them in rpms (3100)
5905 - add new init script for (un)mounting lustre filesystems (2593)
5906 - run chkconfig in %post for init scripts (3701)
5907 - drop scimac NAL (unmaintained)
5909 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - clean kiobufs before and after use (3485)
5913 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5914 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5915 - change init script to fail more clearly if not run as root (1528)
5916 - allow clients to reconnect during replay (1742)
5917 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5918 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5919 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5920 - fix cond_resched (3554)
5921 - only evict unfinished clients after recovery (3515)
5922 - allow bulk resend, prevent data loss (3570)
5923 - dynamic ptlrpc request buffer allocation (2102)
5924 - don't allow unlinking open directory if it isn't empty (2904)
5925 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5926 - remove extraneous obd dereference causing LASSERT failure (3334)
5927 - don't use get_cycles() when creating temp. files on the mds (3156)
5928 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5929 - handle EEXIST for set-stripe, set proper directory name (3336)
5931 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5932 - fix ksocknal_fmb_callback() error messages (2918)
5934 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - don't copy lvb into (possibly NULL) reply on error (2983)
5938 - don't deref dentry after dput, don't free lvb on error (2922)
5939 - use the kms to determine writeback rpc length (2947)
5940 - increment oti_logcookies when osc is inactive (2948)
5941 - update client's i_blocks count via lvb messages (2543)
5942 - handle intent open/close of special files properly (1557)
5943 - mount MDS with errors=remount-ro, like obdfilter (2009)
5944 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5945 - don't use cancelling-locks' kms values (2947)
5946 - use highest lock extent for kms, not last one (2925)
5947 - don't dereference ERR_PTR() dentry in error handling path (3107)
5948 - fix thread race in portals_debug_dumplog() (3122)
5949 - create lprocfs device entries at setup instead of at attach (1519)
5950 - common AST error handler, don't evict client on completion race (3145)
5951 - zero nameidata in detach_mnt in 2.6 (3118)
5952 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5953 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5954 - initialize RPC timeout timer earlier for 2.6 (3219)
5955 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5956 - print nal/nid for unknown nid (3258)
5957 - additional checks for oscc recovery before doing precreate (3284)
5958 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5959 - don't crash in mdc_close for bad permissions on open (3285)
5960 - zero i_rdev for non-device files (3147)
5961 - clear page->private before handing to FS, better assertion (3119)
5962 - tune the read pipeline (3236)
5963 - fix incorrect decref of invalidated dentry (2350)
5964 - provide read-ahead stats and refine rpc in flight stats (3328)
5965 - don't hold journal transaction open across create RPC (3313)
5966 - update atime on MDS at close time (3265)
5967 - close LDAP connection when recovering to avoid server load (3315)
5968 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5969 - don't leak open file on MDS after open resend (3325)
5970 - serialize filter_precreate and filter_destroy_precreated (3329)
5971 - loop device shouldn't call sync_dev() for nul device (3092)
5972 - clear page cache after eviction (2766)
5973 - resynchronize MDS->OST in background (2824)
5974 - refuse to mount the same filesystem twice on same mountpoint (3394)
5975 - allow llmount to create routes for mounting behind routers (3320)
5976 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5977 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5978 - fix rare problem with rename on htree directories (3417)
5980 - allow default OST striping configuration per directory (1414)
5981 - fix compilation for qswnal for 2.6 kernels (3125)
5982 - increase maximum number of MDS request buffers for large systems
5983 - change liblustreapi to be useful for external progs like lfsck (3098)
5984 - increase local configuration timeout for slow disks (3353)
5985 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5987 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5990 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5991 - don't overwrite extent policy data in reply if lock was blocked (2901)
5992 - drop filter export grants atomically with removal from device (2663)
5993 - del obd_self_export from work_list in class_disconnect_exports (2908)
5994 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5995 - swab reply message in mdc_close, other PPC fixes (2464)
5996 - fix destroying of named logs (2325)
5997 - overwrite old logs when running lconf --write_conf (2264)
5998 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5999 - fix race in target_handle_connect (2898)
6000 - mds_reint_create() should take same inode create lock (2926)
6001 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6002 - don't close files for self_export to avoid uninitialized obd (2936)
6003 - allow MDS with the same name as client node (2939)
6004 - hold dentry reference for closed log files for unlink (2325)
6005 - reserve space for all logs during transactions (2059)
6006 - don't evict page beyond end of stripe extent (2925)
6007 - don't oops on a deleted current working directory (2399)
6008 - handle hard links to targets without a parent properly (2517)
6009 - don't dereference NULL lock when racing during eviction (2867)
6010 - don't grow lock extents when lots of conflicting locks (2919)
6012 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6015 - account for cache space usage on clients to avoid data loss (974)
6016 - lfsck support in lustre kernel code (2349)
6017 - reduce journal credits needed for BRW writes (2370)
6018 - orphan handling to avoid losing space on client/server crashes
6019 - ptlrpcd can be blocked, stopping ALL progress (2477)
6020 - use lock value blocks to assist in proper KMS, faster stat (1021)
6021 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6022 - recovery for initial connections (2355)
6023 - fixes for mds_cleanup_orphans (1934)
6024 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6025 - block all file creations until orphan recovery completes (1901)
6026 - client remove rq_connection from request struct (2423)
6027 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6028 - recovery timer race (2670)
6029 - mdc_close recovey bug (2532)
6030 - ptlrpc cleanup bug (2710)
6031 - mds timeout on local locks (2588)
6032 - namespace lock held during RPCs (2431)
6033 - handle interrupted sync write properly (2503)
6034 - don't try to handle a message that hasn't been replied to (2699)
6035 - client assert failure during cleanup after abort recovery (2701)
6036 - leak mdc device after failed mount (2712)
6037 - ptlrpc_check_set allows timedout requests to complete (2714)
6038 - wait for inflight reqs when ptlrpcd finishes (2710)
6039 - make sure unregistered services are removed from the srv_list
6040 - reset bulk XID's when resending them (caught by 1138 test)
6041 - unregister_bulk after timeout
6042 - fix lconf error (2694)
6043 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6044 - readahead locks pages, leaves pending causing memory pressure (2673)
6045 - increase OST request buffers to 4096 on large machines (2729)
6046 - fix up permission of existing directories in simple_mkdir (2661)
6047 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6048 - don't assign transno to errored transactions (2742)
6049 - don't delete objects on OST if given a bogus objid from MDS (2751)
6050 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6051 - if rq_no_resend, then timeout request after recovery (2432)
6052 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6053 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6054 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6055 - MDS recovery shouldn't delete objects that it has given out (2730)
6056 - if enqueue arrives after completion, don't clobber LVB (2819)
6057 - don't unlock pages twice when trigger_group_io returns error (2814)
6058 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6059 - don't write pages to disk if there was an error (1450)
6060 - don't ping imports that have recovery disabled (2676)
6061 - take buffered bytes into account when balancing socknal conn (2817)
6062 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6063 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6064 - remove little-endian swabbing of llog records (1987)
6065 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6066 - retry reposting request buffers if they fail (1191)
6067 - grow extent at grant time to avoid granting a revoked lock (2809)
6068 - lock revoke doesn't evict page if covered by a second lock (2765)
6069 - disable VM readahead to avoid reading outside lock extents (2805)
6071 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6072 - updated kernel patches for hp-2.4.20 kernel (2681)
6074 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6077 - fix truncated write corruption (2366)
6078 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6080 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6081 - ensure len(uuid) < 37 in lmc (1171)
6082 - fix ia64 OOPS in llog_test (2255)
6083 - zero end of page at obdfilter for partial page writes (2648)
6084 - don't leave stale dentries around after renames (bug 2428)
6085 - fix timeouts when evicting a client with a single lock held (2642)
6086 - set deadline for the initial HELLO message to drain (2634)
6087 - print out dotted-quad IP addresses in the socknal (2302)
6089 - additional debugging for MDS client eviction problem (2443)
6090 - fix mkfsoptions support for osts (2603, 2604)
6092 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6095 - add series for the vanilla 2.6.0 kernel
6096 - add series for the vanilla 2.4.24 kernel
6097 - add series for a cray x86/64 UL kernel drop
6098 - fix xattr patches for the vanilla 2.4.19 series
6100 - generate true UUIDs in lmc (1171)
6101 - have portals stack dumping break in UML (2466)
6102 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6103 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6104 - wake callback waiting threads on client eviction (2460)
6105 - Add --ptldebug and --subsystem to lmc (1719)
6106 - update assertion to allow safe interrupt allocation
6107 - set rq_no_resend for cancel requests (2432)
6108 - recalculate ptlrpcd timeout after resend (2494)
6109 - call vfs_rmdir when removing pending directories (2368)
6110 - fix renaming a file to itself (2429)
6111 - lmc creates a default one-stripe lov (2454)
6112 - expand procfs space to handle large clusters (2326)
6113 - increase UML stack to avoid overflow
6114 - update lconf's list of debug and subsystem masks
6115 - fix lfs find --obd (2510)
6116 - /proc tunable for disabling filter read caching (2591)
6117 - stop rpm packages from altering slapd.conf (2301)
6118 - disable nagle in the socknal under 0conf (2578)
6119 - choose mds inode size based on stripe count (2572)
6120 - fix kernel-source rpm problems (2516)
6122 - add --disable-doc to avoid pdf generation (2421)
6123 - update documentation, tests, type-os, comments
6124 - avoid format warnings on ia64
6125 - remove the TOE NAL
6126 - tiny code cleanups by removing unused fields
6128 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6131 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6132 - remove the most blatant lies from BUILDING file (2371)
6133 - change default debug level to reasonable production setting
6134 - reduce client side cache size to reduce cache flush time
6135 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6136 - make TCP zerocopy and pinger support enabled by default (2476)
6137 - sync writes completed after process exits caused crashes (2319)
6138 - maintain correct mount count on the MDS (2356)
6139 - backout 1557, because 2316 wasn't really fixed
6140 - better file I/O statistics gathering in /proc
6141 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6142 - another kernel patch to fix zero-copy TCP function export
6143 - don't take duplicate lock when processing re-sent getattr (2420)
6144 - lctl uses obd_self_export instead of creating new conn (2353)
6145 - MDS/OST recovery case which requires object creation asserted (2425)
6146 - move lfs from /usr/sbin to /usr/bin in packages
6147 - fix race between mds_client_add and mds_client_free (2417)
6148 - use kmalloc instead of slabs in portals (2430)
6149 - don't create duplicate records when a failover MDS is present (2442)
6150 - remove unnecessary mount age check (2332)
6151 - don't remove directory inodes from locks prematurely (2451)
6152 - don't break if MDS service name is the same as hostname (2103)
6153 - fix races in client write RPC generation when cache full (2482)
6155 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6158 - remove now-unused request->rq_obd (278)
6159 - if an allocation fails, print out how much memory we've used (1933)
6160 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6161 - add the "configurable stack size" patch to most series files (1256)
6162 - ability to write large log records, for 100+ OST configs (2306)
6163 - fix NULL deref when filter_prep fails (2314)
6164 - fix operator precedence error in filter_sync
6165 - dynamic allocation of socknal TX descriptors (2315)
6166 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6167 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6168 - ensure that CWARN messages in Portals always get to the syslog
6169 - __init/__exit are not for prototype decls (ldlm_init/exit)
6170 - x86-64 compile warning fixes
6171 - fix gateway LMC keyword conflict (2318)
6172 - fix MDS lock inversions in getattr/reint paths (1844)
6173 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6174 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6175 - DLM race condition prevented some lock evictions (2328)
6176 - ENOMEM detection and retry on socknal sends (2230)
6177 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6178 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6179 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6180 - better error messages when a client is rejected during recovery (1505)
6181 - avoid cancelling locks which were never granted, after failure (2330)
6182 - fix i_sem/journal inversion in mds_client_add (2333)
6183 - fix truncate/getattr lock cycle deadlock (2334)
6184 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6185 - fix two rare exit paths which could leak an l_lock() ref (2321)
6186 - fencepost error in MDS/OST orphan recovery (2226)
6187 - make log record alignment 8 bytes (1988)
6188 - lstripe now fails when requested offset > ost_count (2237)
6189 - ensure that all kernel series have a complete list.h (1607)
6190 - fix crashes in special-file operations (2316)
6191 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6193 - allow configurable automake binary, for testing new versions
6194 - small update to the lfs documentation
6196 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6198 * fix negative export reference count in fsfilt_sync (2312)
6200 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6201 * release candidate 0.9.1
6203 - orphans are moved into the PENDING directory for possible recovery
6204 - replayed opens now open by fid for orphan/rename safety (1042)
6205 - last close of an orphan inode generates a transno (683)
6206 - chdir() and mount() now pin the directory entry (1020)
6207 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6208 - discard very old requests without processing them (1502)
6209 - remove some common, well-understood CERRORs (1505)
6210 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6211 - clear "grant" flags in OST replies until OST grant code lands (1644)
6212 - fix read performance by not clobbering i_blksize on client (1598)
6213 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6214 - make LRU size a /proc tunable, clears locks when reduced (707)
6215 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6216 - server locks take references on exports now (1558)
6217 - build fixes for 2.4.20-rh trees (1663)
6218 - return an error from lov_create if all OSCs are inactive (1751)
6219 - fix import levels when a reconnect happens without a timeout (1597)
6220 - exit early from mds_open if we get a lookup error (1749)
6221 - partial page read at EOF wouldn't wait for disk before sending (1642)
6222 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6223 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6224 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6225 - write/truncate lock inversion (1639)
6226 - Don't auto-load obdclass, portals modules during cleanup (1495)
6227 - fix timestamps from jumping to "now" (1763)
6228 - extra journal assertions (1648)
6229 - add an extra multiunlink test (1771)
6230 - fix read_record/write_record API (1776)
6231 - fix leak of offset_extent, possible incorrect i_size later (1772)
6232 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6233 - Debugging for the kqswnal_get_idle_tx problems (1820)
6234 - Allow recovery to be attempted multiple times (1536)
6235 - Write out MDS last_rcvd file after it is first created (1600)
6236 - Fix tx_descriptor leak in failed transmit situations (1827)
6237 - ext3 journaling fixes for assertion failure after IO error (1871)
6238 - class_export_put() on freed export after completion AST error (1896)
6239 - Fix revalidate looping in VFS (1322)
6240 - Don't access a freed export during MDS_REINT timeout (1521)
6241 - Add open-unlink recovery support on the MDS (1673,1764)
6242 - Return an error if no MDS data was read from last_rcvd (1946)
6243 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6244 - Don't setup a disk device that doesn't match exported UUID (317)
6245 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6246 - avoid committing NULL handle in force close
6247 - local.sh is now a one-stripe LOV configuration
6248 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6249 - fix ext3 htree duplicate directory entry corruption (1516)
6250 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6251 - update to new LOV EA format (2097)
6252 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6254 2003-06-15 Phil Schwan <phil@clusterfs.com>
6257 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6258 - per-import/export recovery handling (958, 931, 959)
6259 - multiple last-rcvd slots, for serving multiple FSes (949)
6260 - connections are again shared between multiple imp/exports (963, 964)
6261 - "umount -f" would hang if any requests needed to be sent (393, 978)
6262 - avoid pinning large req buffer by copying for queued messages (989)
6263 - add "uuid" to "lctl device" command to help upcalls (991)
6264 - "open" RPCs with transnos would confuse recovery counters (1037)
6265 - do proper endian conversion of all wire messages (288, 340, 891)
6266 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6267 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6268 - fix ptlrpc_connection leak in target_handle_connect (1174)
6269 - fix import refcounting bug in OST and MDS cleanup (1134)
6270 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6271 - fix very unlikely obd_types race condition (501)
6272 - remove osc_open hack for echo_client (1187)
6273 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6274 - a failure in read_inode2 leads to deadlock (1139)
6275 - cancel ack-locks as soon as transaction is committed (1072)
6276 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6277 - make sure to commitrw after any preprw to avoid deadlock (1162)
6278 - failing to execute a file in a lustre FS would lock inode (1203)
6279 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6280 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6281 - obd_uuid2tgt would walk off the end of the list (1255)
6282 - on IA64 the osc would give portals incorrect bulk size (1258)
6283 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6284 - fix lock inversion caused by new llite matching code (1282)
6285 - limit the number of dirty pages on a client to 10MB (1286)
6286 - timed out locks were not being corrected cancelled (1289)
6287 - fix O_DIRECT above 4GB on IA-32 (1292)
6288 * major user-visible changes
6289 - fail out/fail over policy now controlled by the upcall (993)
6291 - add OBD_PING to check server availability and failure (954)
6292 - lustre messages are now sent in sending host order (288, 340, 891)
6293 - add eadatalen to MDS getattr reply (340)
6294 - OST read replies may contain second buffer, with per-page status (593)
6296 2003-03-11 Phil Schwan <phil@clusterfs.com>
6299 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6300 - failed open()s could cause deadlock; fixed (867, 869)
6301 - stop cancelling OST locks when files are closed (481)
6302 - overlapping XID spaces caused network corruption (851, 853)
6303 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6304 - setattr_raw added, to do single-RPC, server-side setattrs
6305 - lmc/lconf syntax change for OST UUIDs
6306 - fix crashy race condition between ptlrpc_free_req and osc_close
6307 - don't use request in mdc_enqueue if we hit a timeout (889)
6308 - don't set the inode i_size for regular files from the MDS (896)
6309 - handle out of order completion AST (842)
6310 - don't LBUG if a lock request times out after receiving AST (913)
6311 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6312 - if a bad lock AST arrives, send an error instead of dropping entirely
6313 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6314 - fix leak in bulk IO when only partially completed (899, 900, 926)
6315 - fix O_DIRECT for ia64 (55)
6316 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6317 - C-z after timeout could hang a process forever; fixed (977)
6319 - client-side I/O cache (678, 924, 929, 941, 970)
6321 - READPAGE and SETATTRs which don't take server-side locks get
6324 2003-02-11 Phil Schwan <phil@clusterfs.com>
6327 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6328 - Fix fsfilt_extN_readpage() to read a full page of directory
6329 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6330 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6331 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6332 - Most intent operations were removed, in favour of a new RPC mode
6333 that does a single RPC to the server and bypasses most of the VFS
6334 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6335 - Aggressively cancel local locks on DLM servers
6336 - mds_reint_unlink sends EA to the client if it's the last nlink.
6337 client uses that EA to unlink OST objects.
6338 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6339 - recursive symlinks were fixed (439)
6340 - fixed NULL deref in DEBUG_REQ
6341 - filter_update_lastobjid no longer calls sync, which annoyed extN
6342 - fixed multi-client small-writes to a single file problem (445)
6343 - fixed mtime updates during file writes (607)
6344 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6345 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6346 - fixed Linux OST size reporting problem (444, 656)
6347 - OST now updates object mtime with writes or setattr (607, 619)
6348 - client verifies file size before zeroing page past EOF (445)
6349 - OST now writes last allocated objid to disk with allocation (108)
6350 - LOV on echo now works (409)
6352 - mds_reint_unlink sends a new buffer, with the EA included. this
6353 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6354 set if a regular file was being unlinked, and it was the last link
6355 - use PtlGet from the target for bulk writes (315)
6356 - OST now updates object mtime with writes or setattr (607, 619)
6357 - LDLM now has a grant-time callback to revalidate locked items, if
6359 - Many MDS operations were reorganized to combat race conditions
6361 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6362 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6364 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6367 - Fully reactivate OST imports after reconnection (512, others)
6368 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6369 - More graceful error handling for truncating on dead OST (515)
6370 - Don't error out unless we're actually accessing dead stripes (474)
6371 - Fix garbage sizes when stripes are missing (410)
6372 - LRU counters were broken, causing constant lock purge (433, 432)
6373 - garbage on read from stripes with failed OSTs (441)
6374 - mark OSCs as active before reconnecting during recovery (438)
6375 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6376 - lfind did not preserve OST order in output (443)
6377 - symlinks cause hung clients, incorrect data (439)
6378 - stop dereferencing request after dropping refcount (457)
6379 - don't LASSERT(spin_is_locked) on non-SMP (455)
6380 - fixes for many rename() bugs
6381 - fstat didn't correctly synchronize attributes (399)
6382 - server must handle lock cancellation during blocking AST prep (487)
6383 - bulk descriptors were free()d too soon (511)
6384 - fix paths in lconf, which would load incorrect modules (451, 507)
6385 - fix confusing lconf 'host not found' error message (386)
6386 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6387 - fix race condition in mdc_blocking_ast() for inode access (526)
6388 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6389 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6390 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6391 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6392 - fix lstripe in several ways (kernel side) (527)
6393 - fix request leak in ldlm_cli_enqueue (262)
6394 - incorrect OSC was marked inactive after OST failure
6395 - call mds_fs_cleanup before unmounting filesystem (524)
6396 - fix races between taking ns_lock and ldlm_lock_change_resource
6397 - fix races updating LOV export open file list
6398 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6399 - fix recovery NULL deref in ldlm_cli_cancel_unused
6400 - fix some DLM races by using new hash table for lock handles (419)
6401 - permit the client to specify desired inodes, at replay
6402 - duplicate requests when we queue them for replay reintegration
6403 - fix last_rcvd offset calculation
6404 - sync after each recovered transaction, so we always make progress
6405 - never, not always, ERESTART requests without transnos
6406 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6408 - skip replay if the MDS says that the client is already connected
6409 - don't check for a recovery-enabled export to match lctl's UUID
6410 - don't INC_USE_COUNT for phantom exports
6411 - don't crash when cleaning up phantom exports (567)
6412 - don't double-finish or set replay data for errored mdc_open requests
6413 - abort requests when they time out, so we don't get old replies
6414 - send/receive replies for AST messages again
6415 - if the client says that it doesn't have the lock, cancel it on the
6417 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6418 mark it as destroyed, it will all work out when decref is called
6419 - fix module use counts (22, 581)
6421 - ASTs now expect a reply (server cancels lock on error reply)
6423 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6426 - fix many simultaneous client startup (392)
6427 - fix dentry->d_it clobbering
6428 - credentials weren't being shipped for readdir/getattr operations
6429 - remove invalid assertions triggered during some concurrent MD
6431 - proper Lustre versions added (336, 389)
6432 - fix memory leak for create error case (398)
6433 - fix LOV locking bug that would get cli/srv out of sync
6434 - fix echo client over LOV (409)
6435 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6436 - fix double-O_EXCL intent crash (424)
6437 - avoid sending multiple lock CANCELs (352)
6439 - MDS can do multi-client recovery (modulo bugs in new code)
6441 - many updates, edits, cleanups
6443 2002-11-18 Phil Schwan <phil@clusterfs.com>
6446 - fix null d_it dereference (346)
6447 - fix full OST/dbench hang (333)
6448 - fix permission problem with file removal (286)
6449 - fix removal of OSCs from LOV when they fail
6450 - fix NULL deref during bulk timeout (214)
6451 - fix problems related to multiple filesystems on one MDS (241)
6452 - fixed serious subtle metadata locking bugs
6453 - free locks on clients when inodes are removed due to memory
6455 - fix inode pointer in lock data (285)
6456 - partial support for multiple MDS on a single host (241)
6457 - data locks weren't cancelled at clear_inode time (290, 311)
6458 - intent locks could lead to unbounded lock growth (205)
6459 - added a maximum lock count, an LRU list, and a flusher
6460 - fix multiple rename (365)
6461 - properly abstracted the echo client
6462 - OSC locked 1 byte too many; fixed
6463 - rewrote brw callback code:
6464 - fixed recovery bugs related to LOVs (306)
6465 - fixed too-many-pages-in-one-write crash (191)
6466 - fixed (again) crash in sync_io_timeout (214)
6467 - probably fixed callback-related race (385)
6469 - Add capability to MDS protocol
6470 - LDLM cancellations and callbacks on different portals
6472 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6475 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6476 - defer open object destruction to close time (601981, 138)
6477 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6478 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6479 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6480 - don't call obd_setattr in ll_file_release for destroyed objects
6481 * protocol change to lustre_msg: move |version| and add |flags|
6482 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6483 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6484 * added replay of create, unlink, link and rename operations during
6485 MDS failover; recovery should be much more robust now
6486 * remove failed OSCs from LOVs (only lov_create uses this so far)
6487 * the lustre-HOWTO was brought (more) up to date (582544)
6489 2002-10-23 Phil Schwan <phil@clusterfs.com>
6492 - in-use dentries weren't being reused properly (617851)
6493 - prevent multiple LDLM setup (599178)
6494 - fix LOV size calculations for truncate (617853)
6495 - fix client handling of MDS intent errors (POSIX)
6496 - fix permission bug in lovstripe.c test (624321)
6497 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6498 - truncate past end of file could corrupt data
6499 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6500 - a race in recovery could return ETIMEDOUT to apps (623947)
6501 - building outside the source directory was fixed
6502 * the lustre-HOWTO was brought (more) up to date (582544)
6503 * major progress was made on recovery functionality
6505 2002-10-10 Phil Schwan <phil@clusterfs.com>
6508 - recovery deadlock fix
6509 - rm -rf causes LBUG fix (617817)
6510 - file open by multiple tasks fix (618962)
6511 - directory permissions bugs (602707 and 620007)
6512 - journal_stop fixed with locking (611313)
6513 - O_APPEND failures resolved (618273, perhaps 614459)
6514 - lconf PATH fix (619770)
6515 - IA64 build fix (621450)
6516 - RPC buffer sizes scale with amount of memory
6518 2002-10-01 Phil Schwan <phil@clusterfs.com>
6521 - locks would be cancelled without throwing away data pages,
6522 resulting in inconsistent data (605627)
6523 - inode attributes were not always being refreshed (605627, 612449)
6524 - lconf now continues to cleanup after lctl reports an error
6525 - MDS now enforces user permissions (602707)
6526 - lprocfs cleanup fixed, but not yet enabled (614157)
6527 - fixed infinite server hang, should a client not respond to an AST
6528 - avoid going into recovery if user calls readlink() with a buffer
6529 that's too small (613941)
6530 - AST RPCs no longer require replies (614867) -- this may be changed
6531 - don't crash server if client sends an IOV that's too big (611336)
6532 - fixed lock conversion deadlock (611892)
6533 - fixed the following of symlinks (614622)
6534 * recovery: the server can remove locks from a client that dies, other
6535 clients can make progress
6536 * more extN patch fixes
6537 * compile-time configurable ptlrpc buffer allocations
6539 - collaborative read cache document
6540 - Lustre Lite Performance CDR document-in-progress
6542 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6545 - fix typo in patch-2.4.18
6547 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6550 - clear ptlrpc request each time in handle_incoming_request()
6551 - unlink of files now destroys the object on the OST
6553 2002-09-19 Peter Braam <braam@clusterfs.com>
6555 * add hard link support
6556 * change obdfile creation method
6557 * kernel patch changed
6559 2002-09-19 Peter Braam <braam@clusterfs.com>
6562 - stack overflow bug in extN fixed
6564 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6566 * documentation updates
6567 - add man pages for config tools
6568 - update tests/README to describe testing with new config tools
6569 - finish metadata API descriptions
6570 * bug fixes and cleanups
6571 - statfs workaround for 16TB limit
6572 - LOV stripe allocation improved, can stripe on subset of OSTs
6573 - LOV file size/IO offset was wrong for files > 4GB in size
6574 - object EA data was being dropped, caused files to be unreadable
6575 - memory overflow with non-LOV OST caused memory corruption
6576 - fixed regression tests to work with new config tools, obdfilter
6577 - fixed bug when directory size became larger than 1 block
6578 - fixed bug (for single client case) when PWD was deleted
6579 - invalidate local directory pages when doing intent-based ops
6580 - avoid LDLM oops when lock callback contained bad data
6582 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6584 * documentation updates
6585 * bug fixes and cleanups
6586 - configuration tools
6589 - 64-bit compile warnings
6590 - 64-bit internal statfs data
6592 * test_brw on persistent OST devices
6594 * lprocfs (disabled)
6596 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6598 * documentation updates
6599 * bug fixes and cleanups
6600 * configuration tools
6602 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6605 * many small fixes to 0_5_4
6606 * io/network handling
6607 * thinkos in MDS operations
6609 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6612 * crucial basic fixes to 0.5.3
6613 * IOR, Iozone work over Elan
6616 2002-08-07 Phil Schwan <phil@clusterfs.com>
6617 * version 0_5_3, our first alpha
6618 * we use the new Portals iovs
6619 * documentation updates
6620 * bug fixes and cleanups
6621 * small changes in the DLM wire protocol
6623 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6624 * version 0_5_1 with some initial stability,
6625 * locking on MD and file I/O.
6626 * documentation updates
6627 * several bug fixes since 0.5.0
6628 * small changes in wire protocol
6630 2002-07-18 Phil Schwan <phil@clusterfs.com>
6632 * delivered as Lustre Light Alpha
6633 * fixed a crash after handling invalid MDS requests
6634 * fixed directory pages for architectures with non-4k pages sizes
6636 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6637 * release version v0_4_4
6638 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6640 * Fixes a number of interruption problems with OST operations.
6641 * Update documentation for portals header changes
6642 * Move all wire protocol structs/defines to lustre_idl.h
6643 * Fixes symlink length bug.
6644 * Add tcpdump to repository.
6646 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6647 * release version v0_4_3
6648 * Fixes statfs for inodes on extN.
6649 * Fixes bug in runtests which would delete /etc/hosts.
6650 * Use 64-bit object IDs wherever possible (not into VFS though)
6651 Remove ost_get_info, which is unused by lustre, and out of date.
6653 2002-07-03 Peter Braam <braam@clusterfs.com>
6654 * release version v0_4_2 Fixes a lookup error (type not passed)
6655 * move forward to head of Portals
6656 * move forward to latest Lustre kernel
6658 2002-06-25 Peter Braam <braam@clusterfs.com>
6659 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob