1 01-31-2018 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 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-693.11.6.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 4.4.114-92.67 (SLES12 SP2)
16 4.4.114-94.14 (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 2.6.32-696.18.7.el6 (RHEL6.9)
24 3.10.0-693.11.6.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 4.4.114-92.67 (SLES12 SP2)
29 4.4.114-94.14 (SLES12 SP3)
31 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
32 * Recommended ZFS / SPL version: 0.7.5
33 * Tested with ZFS / SPL version: 0.7.5
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 12-11-2017 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 2.6.32-696.20.1.el6 (RHEL6.9)
52 3.10.0-693.17.1.el7 (RHEL7.4)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-107 (SLES11 SP4)
55 3.12.74-60.64.40 (SLES12 SP1)
56 4.4.103-92.56 (SLES12 SP2)
57 4.4.103-6.38 (SLES12 SP3)
58 vanilla linux 4.6.7 (ZFS only)
59 * Client known to build on unpatched kernels:
60 2.6.32-431.29.2.el6 (RHEL6.5)
61 2.6.32-504.30.3.el6 (RHEL6.6)
62 2.6.32-573.26.1.el6 (RHEL6.7)
63 2.6.32-642.15.1.el6 (RHEL6.8)
64 2.6.32-696.20.1.el6 (RHEL6.9)
65 3.10.0-693.17.1.el7 (RHEL7.4)
66 3.0.101-0.47.71 (SLES11 SP3)
67 3.0.101-107 (SLES11 SP4)
68 3.12.74-60.64.40 (SLES12 SP1)
69 4.4.103-92.56 (SLES12 SP2)
70 4.4.103-6.38 (SLES12 SP3)
72 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
73 * Recommended ZFS / SPL version: 0.7.3
74 * Tested with ZFS / SPL version: 0.7.3
75 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
76 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
77 stack overflow. For more information, please refer to bugzilla 17630.
78 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
79 the re-exporting nodes to be mounted with "32bitapi" mount option
81 --------------------------------------------------------------------------------
83 10-02-2017 Intel Corporation
85 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
86 for currently supported client and server kernel versions.
87 * Server known to build on patched kernels:
88 2.6.32-431.29.2.el6 (RHEL6.5)
89 2.6.32-504.30.3.el6 (RHEL6.6)
90 2.6.32-573.26.1.el6 (RHEL6.7)
91 2.6.32-642.15.1.el6 (RHEL6.8)
92 2.6.32-696.10.3.el6 (RHEL6.9)
93 3.10.0-693.2.2.el7 (RHEL7.4)
94 3.0.101-0.47.71 (SLES11 SP3)
95 3.0.101-107 (SLES11 SP4)
96 3.12.74-60.64.40 (SLES12 SP1)
97 4.4.74-92.35 (SLES12 SP2)
98 vanilla linux 4.6.7 (ZFS only)
99 * Client known to build on unpatched kernels:
100 2.6.32-431.29.2.el6 (RHEL6.5)
101 2.6.32-504.30.3.el6 (RHEL6.6)
102 2.6.32-573.26.1.el6 (RHEL6.7)
103 2.6.32-642.15.1.el6 (RHEL6.8)
104 2.6.32-696.10.3.el6 (RHEL6.9)
105 3.10.0-693.2.2.el7 (RHEL7.4)
106 3.0.101-0.47.71 (SLES11 SP3)
107 3.0.101-107 (SLES11 SP4)
108 3.12.74-60.64.40 (SLES12 SP1)
109 4.4.74-92.35 (SLES12 SP2)
111 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
112 * Recommended ZFS / SPL version: 0.7.1
113 * Tested with ZFS / SPL version: 0.7.1
114 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
115 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
116 stack overflow. For more information, please refer to bugzilla 17630.
117 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
118 the re-exporting nodes to be mounted with "32bitapi" mount option
120 --------------------------------------------------------------------------------
122 2017-07-05 Intel Corporation
124 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
125 for currently supported client and server kernel versions.
126 * Server known to build on patched kernels:
127 2.6.32-431.29.2.el6 (RHEL6.5)
128 2.6.32-504.30.3.el6 (RHEL6.6)
129 2.6.32-573.26.1.el6 (RHEL6.7)
130 2.6.32-642.15.1.el6 (RHEL6.8)
131 2.6.32-696.3.1.el6 (RHEL6.9)
132 3.10.0-514.21.1.el7 (RHEL7.3)
133 3.0.101-0.47.71 (SLES11 SP3)
134 3.0.101-100 (SLES11 SP4)
135 3.12.74-60.64.40 (SLES12 SP1)
136 4.4.59-92.17 (SLES12 SP2)
137 vanilla linux 4.6.7 (ZFS only)
138 * Client known to build on unpatched kernels:
139 2.6.32-431.29.2.el6 (RHEL6.5)
140 2.6.32-504.30.3.el6 (RHEL6.6)
141 2.6.32-573.26.1.el6 (RHEL6.7)
142 2.6.32-642.15.1.el6 (RHEL6.8)
143 2.6.32-696.3.1.el6 (RHEL6.9)
144 3.10.0-514.21.1.el7 (RHEL7.3)
145 3.0.101-0.47.71 (SLES11 SP3)
146 3.0.101-100 (SLES11 SP4)
147 3.12.74-60.64.40 (SLES12 SP1)
148 4.4.59-92.17 (SLES12 SP2)
150 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
151 * Recommended ZFS / SPL version: 0.6.5.9
152 * Tested with ZFS / SPL version: 0.6.5.9
153 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
154 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
155 stack overflow. For more information, please refer to bugzilla 17630.
156 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
157 the re-exporting nodes to be mounted with "32bitapi" mount option
159 --------------------------------------------------------------------------------
161 12-07-2016 Intel Corporation
163 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
164 for currently supported client and server kernel versions.
165 * Server known to build on patched kernels:
166 2.6.32-431.29.2.el6 (RHEL6.5)
167 2.6.32-504.30.3.el6 (RHEL6.6)
168 2.6.32-573.26.1.el6 (RHEL6.7)
169 2.6.32-642.11.1.el6 (RHEL6.8)
170 3.10.0-514.el7 (RHEL7.3)
171 3.0.101-0.47.71 (SLES11 SP3)
172 3.0.101-80 (SLES11 SP4)
173 3.12.62-60.64.8 (SLES12 SP1)
174 vanilla linux 4.5.7 (ZFS only)
175 * Client known to build on unpatched kernels:
176 2.6.32-431.29.2.el6 (RHEL6.5)
177 2.6.32-504.30.3.el6 (RHEL6.6)
178 2.6.32-573.26.1.el6 (RHEL6.7)
179 2.6.32-642.11.1.el6 (RHEL6.8)
180 3.10.0-514.el7 (RHEL7.3)
181 3.0.101-0.47.71 (SLES11 SP3)
182 3.0.101-80 (SLES11 SP4)
183 3.12.62-60.64.8 (SLES12 SP1)
185 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
186 * Recommended ZFS / SPL version: 0.6.5.8
187 * Tested with ZFS / SPL version: 0.6.5.8
188 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
189 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
190 stack overflow. For more information, please refer to bugzilla 17630.
191 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
192 the re-exporting nodes to be mounted with "32bitapi" mount option
194 --------------------------------------------------------------------------------
196 02-29-2016 Intel Corporation
198 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
199 for currently supported client and server kernel versions.
200 * Server known to build on patched kernels:
201 2.6.32-431.29.2.el6 (RHEL6.5)
202 2.6.32-504.30.3.el6 (RHEL6.6)
203 2.6.32-573.12.1.el6 (RHEL6.7)
204 3.10.0-327.3.1.el7 (RHEL7.2)
205 3.0.101-0.47.71 (SLES11 SP3)
206 3.0.101-68 (SLES11 SP4)
207 vanilla linux 4.2.1 (ZFS only)
208 * Client known to build on unpatched kernels:
209 2.6.32-431.29.2.el6 (RHEL6.5)
210 2.6.32-504.30.3.el6 (RHEL6.6)
211 2.6.32-573.12.1.el6 (RHEL6.7)
212 3.10.0-327.3.1.el7 (RHEL7.2)
213 3.0.101-0.47.71 (SLES11 SP3)
214 3.0.101-68 (SLES11 SP4)
217 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
218 * Recommended ZFS / SPL version: 0.6.4.2
219 * Tested with ZFS / SPL version: 0.6.4.2
220 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
221 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
222 stack overflow. For more information, please refer to bugzilla 17630.
223 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
224 the re-exporting nodes to be mounted with "32bitapi" mount option
226 --------------------------------------------------------------------------------
228 03-10-2015 Intel Corporation
230 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
231 for currently supported client and server kernel versions.
232 * Server known to build on patched kernels:
233 2.6.32-431.29.2.el6 (RHEL6.5)
234 2.6.32-504.8.1.el6 (RHEL6.6)
235 3.0.101-0.46 (SLES11 SP3)
236 * Client known to build on unpatched kernels:
237 2.6.32-431.29.2.el6 (RHEL6.5)
238 2.6.32-504.8.1.el6 (RHEL6.6)
239 3.10.0-123.20.1.el7 (RHEL7)
240 3.0.101-0.46 (SLES11 SP3)
241 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
242 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
243 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
244 stack overflow. For more information, please refer to bugzilla 17630.
245 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
246 the re-exporting nodes to be mounted with "32bitapi" mount option
248 Severity : enhancement
250 Description: control OST-index in IDIF via ROCOMPAT flag.
251 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
252 last_rcvd file. For new formatted OST device, it will be auto set;
253 for the case of upgrading from old OST device, you can enable it
254 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
255 enabled, for new created OST-object, its IDIF-in-LMA will contain
256 the OST-index; for the existing OST-object, the OSD will convert
257 old format IDIF as new format IDIF with OST-index stored in the
258 LMA EA when accessing such OST-object or via OI scrub. Once such
259 flag is enabled, it cannot be reverted back, so the system cannot
260 be downgraded to the orignal incompatible version.
262 --------------------------------------------------------------------------------
264 07-30-2014 Intel Corporation
266 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
267 for currently supported client and server kernel versions.
268 * Server known to build on patched kernels:
269 2.6.32-431.20.3.el6 (RHEL6.5)
270 3.0.101-0.31 (SLES11 SP3)
271 * Client known to build on unpatched kernels:
272 2.6.32-431.20.3.el6 (RHEL6.5)
273 3.0.101-0.31 (SLES11 SP3)
276 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
277 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
278 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
279 stack overflow. For more information, please refer to bugzilla 17630.
280 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
281 the re-exporting nodes to be mounted with "32bitapi" mount option
283 --------------------------------------------------------------------------------
285 10-31-2013 Intel Corporation
287 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
288 * Server builds for kernels:
289 2.6.32-358.18.1.el6 (RHEL6.4)
290 * Client builds for unpatched kernels:
291 2.6.32-358.18.1.el6 (RHEL6.4)
292 3.0.92-0.8 (SLES11 SP3)
294 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
295 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
296 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
297 stack overflow. For more information, please refer to bugzilla 17630.
298 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
299 the re-exporting nodes to be mounted with "32bitapi" mount option
301 --------------------------------------------------------------------------------
303 05-30-2013 Intel Corporation
305 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
306 * Server builds for kernels:
307 2.6.32-358.6.2.el6 (RHEL6.4)
308 * Client builds for unpatched kernels:
309 2.6.32-358.6.2.el6 (RHEL6.4)
310 3.0.74-0.6.6 (SLES11 SP2)
312 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
313 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
314 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
315 stack overflow. For more information, please refer to bugzilla 17630.
316 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
317 the re-exporting nodes to be mounted with "32bitapi" mount option
319 --------------------------------------------------------------------------------
321 9-30-2012 Whamcloud, Inc.
323 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
324 * Server support for kernels:
325 2.6.32-279.5.1.el6 (RHEL6)
326 * Client support for unpatched kernels:
327 2.6.32-279.5.1.el6 (RHEL6)
329 2.6.18-238.19.1.el5 (RHEL5)
330 2.6.16.60-0.69.1 (SLES10)
331 * Recommended e2fsprogs version: 1.41.90.wc4
332 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
333 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
334 stack overflow. For more information, please refer to bugzilla 17630.
335 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
336 the re-exporting nodes to be mounted with "32bitapi" mount option
338 Severity : enhancement
340 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
342 Severity : enhancement
344 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
346 Severity : enhancement
348 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
350 Severity : enhancement
352 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
354 Severity : enhancement
356 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
358 --------------------------------------------------------------------------------
360 04-30-2012 Whamcloud, Inc.
362 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
363 * Server support for kernels:
364 2.6.32-220.4.2.el6 (RHEL6)
365 * Client support for unpatched kernels:
366 2.6.32-220.4.2.el6 (RHEL6)
368 2.6.18-238.19.1.el5 (RHEL5)
369 2.6.16.60-0.69.1 (SLES10)
370 * Recommended e2fsprogs version: 1.41.90.wc4
371 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
372 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
373 stack overflow. For more information, please refer to bugzilla 17630.
374 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
375 the re-exporting nodes to be mounted with "32bitapi" mount option
377 Severity : enhancement
379 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
381 Severity : enhancement
383 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
385 Severity : enhancement
387 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
389 Severity : enhancement
391 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
393 --------------------------------------------------------------------------------
395 09-30-2011 Whamcloud, Inc.
397 * Server support for kernels:
398 2.6.18-238.19.1.el5 (RHEL5)
399 2.6.32-131.6.1.el6 (RHEL6)
400 * Client support for unpatched kernels:
401 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
402 2.6.18-238.19.1.el5 (RHEL5)
403 2.6.32-131.6.1.el6 (RHEL6)
404 2.6.29.4-167.fc11 (FC11)
405 2.6.16.60-0.69.1 (SLES10)
407 * Recommended e2fsprogs version: 1.41.90.wc3
408 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
409 Lustre filesystem with 4K stack may cause a stack overflow. For more
410 information, please refer to bugzilla 17630.
411 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
412 mounted with "32bitapi" mount option
414 Severity : enhancement
415 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
417 Severity : enhancement
418 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
420 Severity : enhancement
421 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
424 Frequency : only with SLES10
426 Description: Use OFED "KMP" provided by Novell
427 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
428 Lustre is now built against this package. That means you need to
429 install the ofed-kmp package from Novell for the patchless client.
430 Note that the ofed-kmp that Novell ships may not exactly match the
431 kernel version but should still be compatible.
433 Severity : enhancement
435 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
436 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
438 --------------------------------------------------------------------------------
440 2010-07-15 Oracle, Inc.
442 * Server support for kernels:
443 2.6.18-164.11.1.el5 (RHEL 5)
444 2.6.18-164.11.1.0.1.el5 (OEL 5)
445 * Client support for unpatched kernels:
446 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
447 2.6.18-164.11.1.el5 (RHEL 5),
448 2.6.18-164.11.1.0.1.el5 (OEL 5)
449 2.6.16.60-0.42.8 (SLES 10),
451 2.6.29.4-167.fc11 (FC11)
452 * Recommended e2fsprogs version: 1.41.10-sun2
454 Severity : enhancement
456 Description: Distribute lfs_migrate script with lustre
457 Details : The lfs_migrate mentioned in the manual is now included in the
458 lustre tree in lustre/scripts and will distributed in lustre RPMs.
460 Severity : enhancement
462 Description: Failover nid list added to import proc file.
464 Severity : enhancement
466 Description: "som_preview" mount option required to use SOM
467 Details : The clients will need to use som_preview mount option
471 Description: Lustre does not do 1MB IOs to HW RAID
472 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
474 Severity : enhancement
476 Description: lfs quota output is not very convenient for awk/sed-parsing
477 Details : Some positions in lfs quota output table could be empty or
478 non-empty which made it hard to parse it with scripts, now
479 a dash is put instead of space where there is not supposed
484 Description: Update to OFED 1.5.1
485 Details : OFED 1.5.1 will be used in environments that don't use
486 vendor supplied OFED.
490 Description: Remove lnet. prefix from lctl params display
491 Details : Also renamed lnet.memused to lnet_memused to
496 Description: Use conf_param -d to remove permanant settings.
500 Description: Add list_param -R
501 Details : lctl list_param now supports recursive mode.
504 Bugzilla : 21829, 22312
505 Description: Fix llobdstat.
506 Details : 1. Now updating the write_bytes counter
507 2. Port script fixes from 1.8.
511 Description: decrease the usage of memory on clients.
512 Details : 1. On clients, recycle dentries and inodes unused.
513 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
518 Description: llobdstat fix and enhancement
519 Details : add a counter to set a limit to how many samples will be returned
520 fix a wildcard in the path to limit to obdfilter stats only
524 Description: lfs check only allowed for root
525 Details : Code cleanups and added a regresion test of non-root use of
530 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
531 Details : Fix hr->hr_index overload.
535 Description: Fix FIEMAP functionality.
539 Description: FC11 patchless client support added
543 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
544 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
548 Description: New acc-sm script and various test-framework enhancments.
552 Description: Previously landed patch was reverted.
556 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
560 Description: Add ldiskfs maxdirsize mount option.
564 Description: Fix mkfs.lustre support for 16TB LUNs.
568 Description: Print mmp_check_interval
569 Details : print mmp_check_interval and make it possible to abort
570 mount operation in case it takes too long.
574 Description: Add cascading_rw to lustre mpi tests.
578 Description: Store SOM data in EA.
579 Details : Also several SOM related cleanups.
583 Description: Add refcount to md_open_data
587 Description: Use refcount on lov_request_set to prevent memory corruption.
591 Description: Statahead: use parent dir's i_mutex for synchronization.
595 Description: loadgen improvements.
599 Description: Enable peer health detection in ksocklnd and ko2iblnd.
603 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
607 Description: Fix race on obd_devcie:md_stats
611 Description: Fix SLES 11 client builds.
612 Details : Resolved uninitalized pointer gcc warnings.
616 Description: Allow non-root access for "lfs check".
617 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
621 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
623 Severity : enhancement
625 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
626 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
630 Description: Sanity test_27p fails.
631 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
632 Sanity tests 27 were cleaned up and improved.
636 Description: Unify req format on client/servers
637 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
638 buffers and swabbers approach..
643 Description: drop join file stuff.
645 Severity : enhancement
647 Description: Add 2.6.27 and SLES11 (patchless client) support
651 Description: Fix in ptlrpc_expire_one_request() to print the signed time
654 Severity : enhancement
656 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
659 Severity : enhancement
661 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
663 Severity : enhancement
665 Description: correctly handle big reply message.
666 Details : send LNet event if reply is bigger than buffer and adjust this buffer
671 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
673 Severity : enhancement
675 Description: Add support for OEL5.
679 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
681 Severity : enhancement
683 Description: "weak-modules" support
684 Details : Implement "weak-modules" support which enables kernel modules
685 to be used with any kernel that implements the same kABI. In
686 order to achieve this modules are now installed in
687 /lib/modules/$(uname -r)/updates/kernel on all distributions.
691 Description: client could not reconnect to OST because of an active request.
692 Details : The client now retries to reconnect to the same server, if a
693 connect request failed with EBUSY or -EAGAIN. Test case 35b
694 conf-sanity.sh has been created.
698 Description: Don't allow make backward step on assignin osc next id.
699 Details : race between allocation next id and ll_sync thread can be cause
700 of set wrong osc next id and can be kill valid ost objects.
704 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
708 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
712 Description: Error handling in osc_statfs_interpret() has been improved.
713 Details : Check in osc_statfs_interpret() for EBADR.
717 Description: Do not send statfs() requests to OSTs disabled by administrator.
718 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
720 Severity : enhancement
722 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
724 Severity : enhancement
726 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
728 Severity : enhancement
730 Description: Enhancement for lfs(1) command to use numeric uid/gid.
734 Description: Improved cleanup procedure for conf-sanity test_46a.
738 Description: Do not start bulk transfer, if client tried to reconnect.
739 Details : Do not start bulk transfer, if client tried to reconnect,
740 but failed due to active rpcs. Check for exp_abort_active_req
741 has been implemented.
745 Description: Prevent inconsistences between linux and lustre mount structures.
746 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
747 Make the sleep interruptible.
749 Severity : enhancement
751 Description: provide server to client comms path
752 Details : server to client communications path via new LDLM_SET_INFO rpc
756 Description: truncate starts GFP_FS allocation under transaction causing deadlock
757 Details : ldiskfs_truncate calls grab_cache_page which may start page
758 allocation under an open transaction. This may lead to
759 calling prune_icache with consequent lustre reentrance.
763 Description: send correctly lsm on open replay
764 Details : MDS is trust to LSM size on replay open, but client can set wrong size
769 Description: enable client interface failover
770 Details : When a child reconnects from another NID, properly update export
771 nid hash position and ldlm reverse import.
773 Severity : enhancement
775 Description: Communicate OST degraded/readonly state via statfs to MDS
776 Details : Flags in the statfs returned from OSTs indicate whether the
777 OST is in a degraded RAID state, or if the filesystem has
778 turned read-only after a filesystem error is detected.
783 Description: don't panic if EPROTO was hit when reading symlink
784 Details : correctly handling request reference in error cases.
786 Severity : enhancement
788 Description: Add LustreNetLink, a kernel-userspace communcation path.
790 Severity : enhancement
792 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
797 Description: access to llog context before init.
798 Details : move handling CATALOGS file at osc layer and forbid access to llog
802 Frequency : rare, only when client is writing to OST on local node
804 Description: Avoid deadlock for local client writes
805 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
806 memory freeing context. This allows OST threads to set the
807 PF_MEMALLOC flag on task structures in order to allocate memory
808 from reserved pools and complete IO.
809 Use GFP_HIGHUSER for OST allocations for non-local client writes,
810 so that the OST threads generate memory pressure and allow
811 inactive pages to be reclaimed.
813 Severity : enhancement
815 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
820 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
821 Details : move ldlm namespace creation in setup phase to avoid grab
822 _lprocfs_lock with cli_sem held.
826 Description: Temporarily disable grant shrink.
827 Details : Disable the feature for debugging.
830 Frequency : only when formatting test filesystems
832 Description: Unable to run several mkfs.lustre on loop devices at the same time
833 Details : mkfs.lustre returns error 256 on the concurrent loop devices
834 formatting. The solution is to proper handle the error.
836 Severity : enhancement
838 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
840 Severity : enhancement
842 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
847 Description: resolve race between obd_disconnect and class_disconnect_exports
848 Details : if obd_disconnect will be called to already disconnected export he
849 forget release one reference and osc module can't unloaded.
851 Severity : enhancement
853 Description: Allow tuning service thread via /proc
854 Details : For each service a new
855 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
856 created that can be used to set min/max thread counts, and get the
857 current number of running threads.
862 Descriptoin: don't return error if have partially created objects for file.
863 Details : lov_update_create_set uses set->set_success as index for created
864 objects, so if some requests will be failed, they will have hole
865 at end of array and qos_shrink_lsm for allocate correct lsm.
867 Severity : enhancement
869 Description: Update OFED support to 1.4.1
874 Description: Reduce small size read RPC
875 Details : Set read-ahead limit for every file and only do read-ahead when
876 available read-ahead pages are bigger than 1M to avoid small size
879 Severity : enhancement
881 Description: move AT tunable parameters for more consistent usage
882 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
884 Severity : enhancement
886 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
887 Details : allow skip disconnected ost for send statfs request and hide error
893 Description: fix racy locking of mballoc block bitmaps causing BUG
894 Details : The locking of the mballoc buddy bitmap and the in-memory
895 block bitmap was using two different spin locks in some
896 cases. This made it possible to incorrectly access the
897 mballoc bitmap while another process was modifying it,
898 causing a sanity assertion to fail. While no on-disk corruption
899 was reported, there was some risk of this happening.
902 Frequency : rare, on llog test 6
904 Descriptoin: don't allow connect to already connected import
905 Details : allowing connect to already connected import is hide connecting problem.
910 Descriptoin: allow kill process which wait statahead result
911 Details : for some reasons 'ls' can stick in waiting result from statahead,
912 in this case need way for kill this process.
914 Severity : enhancement
916 Description: Add state history info file, enhance import info file
917 Details : Track import connection state changes in a new osc/mdc proc file;
918 add overview-type data to the osc/mdc import proc file.
920 Severity : enhancement
922 Description: MDS create should not wait for statfs RPC while holding DLM lock.
924 Severity : enhancement
926 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
931 Descriptoin: OST grant too much space to client even there are not enough space.
932 Details : Client will shrink its grant cache to OST if there are no write
933 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
934 this grant cache if there are already not enough avaible space
935 (left_space < total_clients * 32M).
938 Frequency : start MDS on uncleanly shutdowned MDS device
940 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
941 Details : stay in waiting mds<>ost recovery finished produce random bugs
942 due race between two ll_sync thread for one lov target. send
943 ACTIVATE event only if connect realy finished and import have
947 Frequency : rare, connect and disconnect target at same time
949 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
950 Details : don't call obd_disconnect under lov_lock. this long time
951 operation and can block ptlrpcd which answer to connect request.
956 Descriptoin: don't lose wakeup for imp_recovery_waitq
957 Details : recover_import_no_retry or invalidate_import and import_close can
958 both sleep on imp_recovery_waitq, but we was send only one wakeup
962 Frequency : always with long access acl
964 Descriptoin: mds can't pack reply with long acl.
965 Details : mds don't control size of acl but they limited by reint/getattr
968 Severity : enhancement
970 Description: Update to SLES10 kernel-2.6.16.60-0.33.
972 Severity : enhancement
974 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
977 Frequency : start MDS on uncleanly shutdowned MDS device
979 Descriptoin: aborting recovery hang on MDS
980 Details : don't throttle destroy RPCs for the MDT.
983 Frequency : on remount
985 Description: external journal device not working after the remount
986 Details : clear dev_rdonly flag for external journal devices in
992 Description: shutdown vs evict race
993 Details : client_disconnect_export vs connect request race.
994 if client will evicted at this time - we start invalidate
995 thread without referece to import and import can be freed
999 Frequency : rare, need acl's on inode.
1001 Description: client can't handle ost additional correctly
1002 Details : if ost was added after client connected to mds client can have
1003 hit lnet_try_match_md ... to big messages to wide striped files.
1004 in this case need teach client to handle config events about add
1005 lov target and update client max ea size at that event.
1007 Severity : enhancement
1009 Description: Changelogs
1010 Details : Changelogs are a lightweight mechanism to track filesystem
1011 metadata and namespace changes. The changelog is recorded
1012 permanently on the MDTs, and is periodically "consumed" / purged
1013 when records are no longer needed.
1015 Severity : enhancement
1017 Description: compact fld format with extents
1018 Details : Store range of seq rather than every seq in FLD. Seq
1019 controller update FLD rather than clients. In Case of CMD, mdt0
1020 has FLD, all other metadata server act as non persistent proxy
1021 for FLD queries and cache fld entries in fld cache.
1026 Description: don't skip ost target if they assigned to file
1027 Details : Drop slow OSCs if we can, but not for requested start idx.
1028 This means "if OSC is slow and it is not the requested
1029 start OST, then it can be skipped, otherwise skip it only
1030 if it is inactive/recovering/out-of-space.
1034 Description: more cleanup in mds_lov
1035 Details : not send LOV EA under replay, we can't know about they size at this
1036 time. Don't allow client connect to mds before any ost connected,
1037 for avoid problems with LOV EA size and returning EIO to client.
1039 Severity : enhancement
1041 Description: Interoperability at server side (Disk interoperability)
1043 Severity : enhancement
1045 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1047 Severity : enhancement
1049 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1051 Severity : enhancement
1053 Description: New client IO stack (CLIO).
1055 Severity : enhancement
1057 Description: Commit on sharing. Eliminate inter-client dependencies between
1058 uncommitted transactions by doing transaction commits.
1059 Thereby clients may recovery independently.
1062 Frequency : Create a symlink file with a very long name
1064 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1065 Details : If there is no extra space in the request for early cancels,
1066 ldlm_req_handles_avail() returns 0 instead of a negative value.
1068 Severity : enhancement
1070 Description: Add /proc entry for import status
1071 Details : The mdc, osc, and mgc import directories now have
1072 an import directory that contains useful import data for debugging
1073 connection problems.
1075 Severity : enhancement
1077 Description: Re-disable certain /proc logging
1078 Details : Enable and disable client's offset_stats, extents_stats and
1079 extents_stats_per_process stats logging on the fly.
1082 Frequency : Only on FC kernels 2.6.22+
1084 Description: oops in statahead
1085 Details : Do not drop reference count for the dentry from VFS when lookup,
1086 VFS will do that by itself.
1088 Severity : enhancement
1090 Description: Generic /proc file permissions
1091 Details : Set /Proc file permissions in a more generic way to enable non-
1092 root users operate on some /proc files.
1096 Description: Hitting mdc_commit_close() ASSERTION
1097 Details : Properly handle request reference release in
1098 ll_release_openhandle().
1102 Description: quota recovery deadlock during mds failover
1103 Details : This patch includes att18982, att18236, att18237 in bz14840.
1105 1. fix osts hang when mds does failover with quotaon
1106 2. prevent watchdog storm when osts threads wait for the
1111 Frequency : only patchless client
1112 Description: add workaround for race between add/remove dentry from hash
1114 Severity : enhancement
1116 Description: Allow OST glimpses to return PW locks
1120 Description: LBUG when llog conf file is full
1121 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1125 Description: Prevent import from entering FULL state when server in recovery
1129 Description: service mount cannot take device name with ":"
1130 Details : Only when device name contains ":/" will mount treat it as
1136 Description: replace ptlrpcd with the statahead thread to interpret the async
1137 statahead RPC callback
1141 Frequency : on recovery
1142 Description: I/O failures after umount during fail back
1143 Details : if client reconnected to restarted server we need join to recovery
1144 instead of find server handler is changed and process self
1145 eviction with cancel all locks.
1147 Severity : enhancement
1149 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1151 Severity : enhancement
1153 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1155 Severity : enhancement
1157 Description: Upcall on Lustre log has been dumped
1158 Details : Allow for a user mode script to be called once a Lustre log has
1159 been dumped. It passes the filename of the dumped log to the
1160 script, the location of the script can be specified via
1161 /proc/sys/lnet/debug_log_upcall.
1166 Description: avoid idr_remove called for id which is not allocated.
1167 Details : Move assigment s_dev for clustered nfs to end of initialization,
1168 to avoid problem with error handling.
1173 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1174 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1175 into CONN_USED_HASH and this prodice warning when put connection
1176 again in unused hash.
1179 Severity : enhancement
1181 Description: File striping can now be set to use an arbitrary pool of OSTs.
1183 Severity : enhancement
1185 Description: Export bytes_read/bytes_write count on OSC/OST.
1189 Description: Early reply size mismatch, MGC loses connection
1190 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1191 the connect flags are properly negotiated.
1196 Description: quotas are not honored with O_DIRECT
1197 Details : all writes with the flag O_DIRECT will use grants which leads to
1198 this problem. Now using OBD_BRW_SYNC to guard this.
1202 Description: add quota statistics
1203 Details : 1. sort out quota proc entries and proc code.
1204 2. add quota statistics
1206 Severity : enhancement
1208 Description: enable quota support for HEAD.
1212 Description: Properly propagate oinfo flags from lov to osc for statfs
1213 Details : restore missing copy oi_flags to lov requests.
1215 Severity : enhancement
1217 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1218 : l_getgroups(8), lst(8), routerstat(8)
1219 Details : included man pages for llobdstat(8), llstat(8),
1220 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1222 Severity : enhancement
1224 Description: configure's --enable-quota should check the
1225 : kernel .config for CONFIG_QUOTA
1226 Details : configure is terminated if --enable-quota is passed but
1227 : no quota support is in kernel
1231 Description: Remove portals compatibility
1232 Details : Remove portals compatibility, not interoperable with releases
1237 Description: Resolve device initialization race
1238 Details : Prevent proc handler from accessing devices added to the
1239 obd_devs array but yet be intialized.
1241 Severity : enhancement
1243 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1245 Severity : enhancement
1247 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1252 Description: Using wrong pointer in osc_brw_prep_request
1253 Details : Access to array[-1] can produce panic if kernel compiled with
1254 CONFIG_PAGE_ALLOC enabled
1258 Description: Client runs out of low memory
1259 Details : Consider only lowmem when counting initial number of llap pages
1263 Description: *optional* service tags registration
1264 Details : if the "service tags" package is installed on a Lustre node
1265 When the filesystem is mounted, a local-node service tag will
1266 be created. See http://inventory.sun.com/ for more information
1267 about the Service Tags asset management system.
1271 Description: Kernel BUG tries to release flock
1272 Details : Lustre does not destroy flock lock before last reference goes
1273 away. So always drop flock locks when client is evicted and
1274 perform unlock regardless of successfulness of speaking to MDS.
1278 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1282 Description: sanity test 65a fails if stripecount of -1 is set
1283 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1288 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1289 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1292 Severity : enhancement
1294 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1298 Description: do not process already freed flock
1299 Details : flock can possibly be freed by another thread before it reaches
1300 to ldlm_flock_completion_ast.
1304 Description: LBUG during stress test
1305 Details : Need properly lock accesses the flock deadlock detection list.
1309 Description: oops in page fault handler
1310 Details : kernel page fault handler can return two special 'pages' in error
1311 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1315 Description: timeout with invalidate import.
1316 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1317 should be handled by ptlrpcd. This produce long age waiting and
1318 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1320 Severity : enhancement
1322 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1326 Description: enable MGS and MDT services start separately
1327 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1328 a co-located MGS without starting the MGS, which is a complement
1329 to 'nosvc' mount option.
1333 Frequency : after recovery
1334 Description: precreate to many object's after del orphan.
1335 Details : del orphan st in oscc last_id == next_id and this triger growing
1336 count of precreated objects. Set flag LOW to skip increase count
1337 of precreated objects.
1341 Frequency : rare, on clear nid stats
1342 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1343 Details : when clean nid stats sometimes try destroy live entry,
1344 and this produce panic in free.
1348 Description: Stack overflow during MDS log replay
1349 ease stack pressure by using a thread dealing llog_process.
1353 Description: wait until IO finished before start new when do lock cancel.
1354 Details : VM protocol want old IO finished before start new, in this case
1355 need wait until PG_writeback is cleared until check dirty flag and
1356 call writepages in lock cancel callback.
1358 Severity : enhancement
1360 Description: using special macro for print time and cleanup in includes.
1364 Description: mds_mfd_close() ASSERTION(rc == 0)
1365 Details : In mds_mfd_close(), we need protect inode's writecount change
1366 within its orphan write semaphore to prevent possible races.
1370 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1374 Frequency : rare, on shutdown ost
1375 Description: don't hit live lock with umount ost.
1376 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1377 use shrink_dcache_sb instead.
1381 Description: don't panic with use echo client
1382 Details : echo client pass NULL as client nid pointer and this produce null
1383 pointer dereference.
1387 Description: fix build on ppc32
1388 Details : compile code with -m64 flag produce wrong object file for ppc32.
1392 Description: add message levels for liblustreapi
1396 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1397 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1398 will result in an -EAGAIN and in the exit call path we call
1399 alloc_rr() which will with increasing aggressiveness attempt to
1400 aquire precreated objects on the minimum number of required OSCs.
1404 Description: Use old size assignment to avoid deadlock
1405 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1406 the scheduling while holding a spinlock. We do not need locking
1407 for size in ll_update_inode() because size is only updated from
1408 the MDS for directories or files without objects, so there is no
1409 other place to do the update, and concurrent access to such inodes
1410 are protected by the inode lock.
1414 Description: resolve "_IOWR redefined" build error on SLES10
1418 Description: dump the memory debugging after all modules are unloaded to
1419 suppress false negative in conf_sanity test 39
1421 Severity : enhancement
1423 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1428 Description: fix SLES kernel versioning
1429 Details : the kernel version for our SLES 10 kernel did not include a "-"
1430 before the "smp" at the end. while this was not a problem in
1431 general, it did mean that software trying to use the kernel
1432 version to try to detect a vendor specific kernel would fail.
1433 this was most evident by the OFED build scripts.
1437 Description: Don't update lov_desc members until making sure they are valid
1438 Details : When updating lov_desc members via proc fs, need fix their
1439 validities before doing the real update.
1443 Description: don't put request into delay list while invalidate in flight.
1444 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1445 import in flight. this produce timeout for invalidate and sometimes
1446 can cause stale data.
1450 Frequency : on ppc only
1451 Description: not convert ost objects for directory because it's not exist.
1452 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1456 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1458 Severity : enhancement
1460 Description: Update to support 2.6.22.14 vanilla kernel.
1464 Frequency : rare, on recovery
1465 Description: read procfs can produce deadlock in some situation
1466 Details : Holding lprocfs lock which send rpc can produce block for destroy
1467 obd objects and this also block reconnect with -EALREADY. This
1468 isn't fix all lprocfs bugs - but make it rare.
1470 Severity : enhancement
1472 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1475 Frequency : frequent on X2 node
1477 Description: mdc_set_open_replay_data LBUG
1478 Details : Set replay data for requests that are eligible for replay.
1482 Description: lustre_mgs: operation 101 on unconnected MGS
1483 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1484 MGC, and late on MGC cannot successfully connect to MGS and a lot
1485 of the error messages complaining that MGS is not connected.
1489 Frequency : on network error
1490 Description: panic with double free request if network error
1491 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1492 but it's true only for synchronus enqueue, for async enqueue
1493 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1496 Severity : enhancement
1498 Description: client-side metadata stat-ahead during readdir(directory readahead)
1499 Details : perform client-side metadata stat-ahead when the client detects
1500 readdir and sequential stat of dir entries therein
1503 Frequency : on start mds
1505 Description: Implement get_info(last_id) in obdfilter.
1508 Frequency : occasional
1510 Description: Correctly check stale fid, not start epoch if ost not support SOM
1511 Details : open with flag O_CREATE need set old fid in op_fid3 because
1512 op_fid2 was overwritten with new generated fid, but mds can answer
1513 with one of these two fids and both is not stale. Setattr
1514 incorrectly started an epoch and assume will be called
1515 done_writing, but without SOM done_writing ever being called.
1518 Frequency : rare, depends on device drivers and load
1520 Description: MDS or OSS nodes crash due to stack overflow
1521 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1522 In some cases, in conjunction with device drivers that use a lot
1523 of stack the MDS (or possibly OSS) service threads could overflow
1524 the stack. One change which was identified to consume additional
1525 stack has been reworked to avoid the extra stack usage.
1528 Frequency : occasional
1530 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1531 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1532 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1533 oscc_has_objects and see OSC has no objests and call
1534 oscc_internal_create to resend create request.
1536 Severity : enhancement
1538 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1540 Severity : enhancement
1542 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1545 Frequency : very rare
1547 Description: Fix replay if there is an un-replied request and open
1548 Details : In some cases, older replay request will revert the
1549 mcd->mcd_last_xid on MDS which is used to record the client's
1550 latest sent request.
1552 Severity : enhancement
1554 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1556 Severity : enhancement
1558 Description: Add rhel5 support to HEAD.
1560 Serverity : enhancement
1562 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1567 Description: Don't allow skipping OSTs if index has been specified.
1568 Details : Don't allow skipping OSTs if index has been specified, make
1569 locking in internal create lots better.
1573 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1574 Details : access to bitfield in structure is always rounded to long
1575 and this produce problem with not atomic change any bit.
1579 Description: Lustre make rpms failed.
1580 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1581 builds Lustre rpms from tarball.
1584 Frequency : rare on shutdown ost
1586 Description: If llog cancel was not send before clean_exports phase, this can
1587 produce deadlock in llog code.
1588 Details : If llog thread has last reference to obd and call class_import_put
1589 this produce deadlock because llog_cleanup_commit_master wait when
1590 last llog_commit_thread exited, but this never success because was
1591 called from llog_commit_thread.
1595 Description: allow userland application know is lost one of stripes.
1596 Details : fill lvb_blocks with error code on ost and return it to
1597 application if error flag found.
1601 Description: NULL lov_tgts causing MDS oops
1602 Details : more safe checks for NULL lov_tgts for avoid oops.
1604 Severity : enhancement
1606 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1610 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1611 Details : If an OST is down the MDS will hang indefinitely in
1612 obd_statfs_rqset() waiting for the statfs data. While for
1613 MDS QOS usage of statfs, it should not stuck in waiting.
1615 Severity : enhancement
1617 Description: remote_acl support
1618 Details : Support ACL-based permission check for remote user.
1619 Support setfacl/getfacl for remote user with the utils
1620 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1621 the system "{s,g}etfacl" utils.
1623 Severity : enhancement
1625 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1627 Severity : enhancement
1629 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1633 Description: make mgs_setparam() handle fsname containing dash
1634 Details : fsname containing a dash does not work with lctl conf_param
1636 Severity : enhancement
1638 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1640 Severity : enhancement
1642 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1646 Frequency : rare, at shutdown
1647 Description: access already free / zero obd_namespace.
1648 Details : if client_disconnect_export was called without force flag set,
1649 and exist connect request in flight, this can produce access to
1650 NULL pointer (or already free pointer) when connect_interpret
1651 store ocd flags in obd_namespace.
1655 Frequency : only at startup
1656 Description: not alloc memory with spinlock held.
1657 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1658 if any spinlock held.
1660 Severity : enhancement
1662 Description: make lustre randomly fail allocating memory
1663 Details : Make lustre randomly failed allocating memory for testing purpose.
1665 Severity : enhancement
1667 Description: lost problems with lov objid file
1668 Details : Fixes some scability and access to not inited memory problems
1669 in work with lov objdid file.
1674 Description: lfs find does not continue on file error
1675 Details : Continue other files processing when a file/dir is absent.
1679 Description: Inconsistent usage of lustre_pack_reply()
1680 Details : Standardize the usage of lustre_pack_reply() such that it
1681 always generate a CERROR on failure.
1684 Frequency : occasional
1686 Description: MDS hang or stay in waiting lock
1687 Details : If client receive lock with CBPENDING flag ldlm need send lock
1688 cancel as separate rpc, to avoid situation when cancel request
1689 can't processed due all i/o threads stay in wait lock.
1693 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1697 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1701 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1705 Description: Update to RHEL4 latest kernel.
1707 Severity : enhancement
1709 Description: Build SLES10 patchless client fails
1710 Details : The configure was broken by run ./configure with
1711 --with-linux-obj=.... argument for patchless client. When the
1712 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1713 can't search header adequately. Use absolute path such as
1714 -I($LINUX)/include instead.
1718 Description: interrupt oig_wait produce painc on resend.
1719 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1720 private set, and this produce situation when rq_ptlrpcd_data not
1721 copyed to new allocated request and triggered LBUG on assert
1722 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1723 ptlrpcd set if request is changed and can be safety changed to use
1728 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1729 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1730 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1731 these macros and replace them with their difinitions instead.
1735 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1736 Details : Update which_patch & target file for SLES10 latest kernel.
1739 Description: Remove client patches from SLES 10 kernel.
1740 Details : This causes SLES 10 clients to behave as patchless clients
1741 even on a Lustre-patched (server) kernel.
1743 Severity : enhancement
1745 Description: self-adjustable client's lru lists
1746 Details : use adaptive algorithm for managing client cached locks lru
1747 lists according to current server load, other client's work
1748 pattern, memory activities, etc. Both, server and client
1749 side namespaces provide number of proc tunables for controlling
1752 Severity : enhancement
1754 Description: light-weight GSS support
1755 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1756 and reduce performance overhead.
1758 Severity : enhancement
1760 Description: Linux keyring support
1761 Details : Support using service of Linux keyring for Lustre GSS internal
1762 context refresh/cache mechanism.
1766 Description: Fix errors in lfs documentation
1767 Details : Fixes man pages
1771 Description: don't use GFP_* in generic Lustre code.
1772 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1776 Description: obdclass is limited by single OBD_ALLOC(idarray)
1777 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1781 Description: warnings with build patchless client with vanila 2.6.19 and up
1782 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1783 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1787 Frequency : Only for SLES
1788 Description: sanity_quota fail test_1
1789 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1790 file, which makes TSTID[2] inunique.
1794 Frequency : Only for SLES9
1795 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1796 Details : Update target/ChangeLog/which_patch .
1800 Frequency : Only for test_10 in sanity-quota.sh
1801 Description: a bug in quota test code
1802 Details : checking a test flag in a wrong place
1806 Frequency : Only for test_13 in sanity-quota.sh
1807 Description: a bug in quota test script
1808 Details : shouldn't check overfull allocated quota
1812 Description: jbd statistics
1813 Details : Port older jbd statistics patch for sles10
1815 Severity : enhancement
1817 Description: Add printing inode info into message about error in writepage.
1822 Description: reply_lock_interpret crash due to race with it and lock cancel.
1823 Details : Do not replay locks that are being cancelled. Do not reference
1824 locks by their address during replay, just by their handle.
1828 Frequency : When flocks are used.
1829 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1830 Details : Flock locks might destroy just granted lock if it could be merged
1831 with another existing flock, this is done in completion handler,
1832 so teach ldlm_cli_enquque_fini that this is a valid case for
1838 Description: Oops in read and write path when failing to allocate lock.
1839 Details : Check if lock allocation failed and return error back.
1843 Description: lstripe command fails for valid OST index
1844 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1845 instead of lov->desc.ld_active_tgt_count.
1849 Description: sanity.sh failed test 103
1850 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1855 Description: df doesn't work properly if diskfs blocksize != 4K
1856 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1860 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1861 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1863 Severity : enhancement
1865 Description: Adaptive timeouts
1866 Details : RPC timeouts adapt to changing server load and network
1867 conditions to reduce resend attempts and improve recovery time.
1871 Description: llapi_file_create() does not allow some changes
1872 Details : add llapi_file_open() that allows specifying the mode and
1873 open flags, and also returns an open file handle.
1877 Description: merge and cleanup kernel patches.
1878 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1882 Description: Add journal checksum support.(Kernel part)
1883 Details : The journal checksum feature adds two new flags i.e
1884 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1885 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1886 indicates that the commit block contains the checksum for
1887 the blocks described by the descriptor blocks. Now commit
1888 record can be sent to disk without waiting for descriptor
1889 blocks to be written to disk. This behavior is controlled
1890 using JBD2_FEATURE_ASYNC_COMMIT flag.
1894 Description: OSS needs mutliple precreate threads
1895 Details : Add ability to start more than one create thread per OSS.
1899 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1900 Details : execute lfs setstripe on client
1904 Description: mds_obd_create error creating tmp object
1905 Details : When the user sets quota on root, llog will be affected and can't
1906 create files and write files.
1909 Frequency : Always on ia64 patchless client, and possibly others.
1911 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1912 Details : This allows the patchless client to be loaded on architectures
1913 without this export.
1917 Description: RedHat Update kernel for RHEL5
1918 Details : Add the kernel config file for RHEL5.
1922 Description: RedHat Update kernel for RHEL5
1923 Details : Modify the kernel config file more closer RHEL5.
1927 Description: Build failure against Centos5 (RHEL5)
1928 Details : Define PAGE_SIZE when it isn't present.
1932 Description: test_7 on sanity-quota.sh can't be used on separate machines
1933 Details : add facet to handle it
1937 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1938 Details : with NFS, the anon dentry's parent was set to itself in
1939 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1940 corresponding dentry other than use rec->ur_name.
1942 Severity : enhancement
1944 Description: lfs setstripe enhancement
1945 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1949 Description: enable data checksumming by default
1950 Details : enable checksum by default, allow --disable-checksum
1951 configure option and "-o nochecksum" mount option. Checksums
1952 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1956 Description: lustre support for RHEL5
1957 Details : Add support for RHEL5.
1961 Description: Client eviction due to failover config
1962 Details : after a connection loss, the lustre client should attempt to
1963 reconnect to the last active server first before trying the
1964 other potential connections.
1968 Description: when mds and osts use different quota unit(32bit and 64bit),
1969 quota will be released repeatly.
1970 Details : void sending multiple quota reqs to mds, which will keep the
1971 status between the reqs.
1975 Description: osts not allocated evenly to files
1976 Details : change the condition to increase offset_idx
1979 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1981 Description: Data corruption for OSTs that are formatted larger than 2TB
1983 Details : When generating the bio request for lustre file writes the
1984 sector number would overflow a temporary variable before being
1985 used for the IO. The data reads correctly from Lustre (which
1986 will overflow in a similar manner) but other file data or
1987 filesystem metadata may be corrupted in some cases.
1991 Description: Tune the kernel for good SCSI performance.
1992 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1993 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1998 Description: rewrite ext2-derived code in obdclass/uuid.c
1999 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2000 from scratch preserving functionality.
2004 Description: "lfs find -obd UUID" prints directories
2005 Details : "lfs find -obd UUID" will return all directory names instead
2006 of just file names. It is incorrect because the directories
2007 do not reside on the OSTs.
2011 Description: use i_size_read and i_size_write in 2.6 port
2012 Details : replace inode->i_size access with i_size_read/write()
2015 Frequency : when removing large files
2017 Description: scheduling issue during removal of large Lustre files
2018 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2019 It causes scheduling issues when removing large files (17TB in the
2023 Frequency : only with liblustre clients on XT3
2025 Description: evictions taking too long
2026 Details : allow llrd to evict clients directly on OSTs
2029 Frequency : only on ppc
2031 Description: /proc/fs/lustre/devices broken on ppc
2032 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2033 We should make sure the type of 'index' is loff_t and then cast
2034 explicitly as needed below. Do not assign an explicitly cast
2038 Frequency : only for rhel5
2040 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2041 Details : Modify the target file & which_kernel.
2043 Severity : enhancement
2045 Description: omit set fsid for export NFS
2046 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2047 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2051 Frequency : Always, for kernels after 2.6.16
2052 Description: Fix warning idr_remove called for id=.. which is not allocated.
2053 Details : Last kernels save old s_dev before kill super and not allow
2054 to restore from callback - restore it before call kill_anon_super.
2058 Description: buffer overruns could theoretically occur
2059 Details : llapi_semantic_traverse() modifies the "path" argument by
2060 appending values to the end of the origin string, and a
2061 overrun may occur. Adding buffer overrun check in liblustreapi.
2065 Description: Fix error on 'ls .' at the top of the Lustre mount.
2066 Details : Don't revalidate dentry if it is a root dentry.
2070 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2071 Details : Modify vm-tunables-rhel4.patch.
2075 Description: Kernel config for 2.6.18-vanilla.
2076 Details : Modify targets/2.6-vanilla.target.in.
2077 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2078 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2079 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2080 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2084 Description: improve handling recoverable errors
2085 Details : if request processig with error which can be recoverable on server
2086 request should be resend, otherwise page released from cache and
2091 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2092 Details : Modify target file & which_patch.
2093 A flaw was found in the IA32 system call emulation provided
2094 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2095 value could be stored in the %RAX register, which could trigger an
2096 out-of-bounds system call table access. An untrusted local user
2097 could exploit this flaw to run code in the kernel
2098 (ie a root privilege escalation). (CVE-2007-4573).
2102 Description: O_DIRECT bypasses client statistics.
2103 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2104 read_ahead_stats were updated. copied the stats section from
2105 osc_send_oap_rpc() into async_internal().
2109 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2113 Description: change order of libsysio includes
2114 Details : '#include sysio.h' should always come before '#include xtio.h'
2118 Description: handle "serious error: objid * already exists" more gracefully
2119 Details : If LAST_ID value on disk is smaller than the objects existing in
2120 the O/0/d* directories, it indicates disk corruption and causes an
2121 LBUG(). If the object is 0-length, then we should use the existing
2122 object. This will help to avoid a full fsck in most cases.
2124 Severity : enhancement
2126 Description: adapt the lustre_config script to support the upgrade case
2127 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2128 server targets to 1.6 in parallel.
2132 Description: To avoid grant space > avaible space when the disk is almost
2133 full. Without this patch you might see the error "grant XXXX >
2134 available" or some LBUG about grant, when the disk is almost
2136 Details : In filter_check_grant, for non_grant cache write, we should
2137 check the left space by if (*left > ungranted + bytes), instead
2138 of (*left > ungranted), because only we are sure the left space
2139 is enough for another "bytes", then the ungrant space should be
2140 increase. In client, we should update cl_avail_grant only there
2141 is OBD_MD_FLGRANT in the reply.
2145 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2146 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2147 described in CVE-2007-4573. This problem could allow untrusted
2148 local users to gain root access.
2151 Frequency : when using O_DIRECT and quotas
2153 Description: Incorrect file ownership on O_DIRECT output files
2154 Details : block usage reported by 'lfs quota' does not take into account
2155 files that have been written with O_DIRECT.
2160 Description: touch file failed when fs is not full
2161 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2162 otherwise we can get ENOSP while fs is not full.
2166 Description: parallel lock callbacks
2167 Details : Instead of sending blocking and completion callbacks as separated
2168 requests, adding them to a set and sending in parallel.
2171 Frequency : only for Cray XT3
2172 Bugzilla : 12829/13455
2173 Description: Changing primary group doesn't change the group lustre assigns to
2175 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2176 overridden with the primary group provided by the group upcall,
2177 whereas the supplied fsgid can be trusted if it is in the list of
2178 supplementary groups returned by the group upcall.
2180 Severity : enhancement
2182 Description: Allow masking D_WARNING, D_ERROR messages from console
2183 Details : Console messages can now be disabled via lnet.printk.
2187 Description: User code with malformed file open parameter crashes client node
2188 Details : Before packing join_file req, all the related reference should be
2189 checked carefully in case some malformed flags cause fake
2190 join_file req on client.
2194 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2195 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2196 mode is equal to the granted mode. However, it should also check
2197 that LDLM_FL_CP_REQD is not set.
2201 Description: Heavy nfs access might result in deadlocks
2202 Details : After ELC code landed, it is now improper to enqueue any mds
2203 locks under och_sem, because enqueue might want to decide to
2204 cancel open locks for same inode we are holding och_sem for.
2208 Description: Client eviction while running blogbench
2209 Details : A lot of unlink operations with concurrent I/O can lead to a
2210 deadlock causing evictions. To address the problem, the number of
2211 oustanding OST_DESTROY requests is now throttled to
2212 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2213 callbacks are processed in priority.
2217 Description: enable ACLs on MDS by default
2218 Details : ACLs must be enabled on MDS by default.
2221 Frequency : PPC/PPC64 only
2223 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2224 Details : fix duplicated definitions between asm-ppc64/types.h and
2225 lustre_types.h on PPC.
2228 Frequency : PPC/PPC64 only
2230 Description: asm-ppc/segment.h does not exist
2231 Details : fix compile issue on PPC.
2235 Description: better handle error messages in extents code
2238 Frequency : RHEL4 only
2240 Description: mkfs is very slow on IA64/RHEL4
2241 Details : A performance regression has been discovered in the MPT Fusion
2242 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2243 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2244 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2246 Severity : enhancement
2248 Description: SNMP support enhancement
2249 Details : Adding total number of sampled request for an MDS node in snmp
2252 Severity : enhancement
2254 Description: Optimize ldlm waiting list processing for PR extent locks
2255 Details : When processing waiting list for read extent lock and meeting read
2256 lock that is same or wider to it that is not contended, skip
2257 processing rest of the list and immediatelly return current
2258 status of conflictness, since we are guaranteed there are no
2259 conflicting locks in the rest of the list.
2263 Description: Time out and refuse to reconnect
2264 Details : When the failover node is the primary node, it is possible
2265 to have two identical connections in imp_conn_list. We must
2266 compare not conn's pointers but NIDs, otherwise we can defeat
2267 connection throttling.
2271 Description: port llog fixes from b1_6 into HEAD
2272 Details : Port llog reference couting and some llog cleanups from b1_6
2273 (bug 10800) into HEAD, for protect from panic and access to already
2274 free llog structures.
2278 Description: Detect stride IO mode in read-ahead
2279 Details : When a client does stride read, read-ahead should detect that and
2280 read-ahead pages according to the detected stride pattern.
2284 Description: data checksumming impacts single node performance
2285 Details : add support for several checksum algorithm. Currently, only CRC32
2286 and Adler-32 are supported. The checksum type can be changed on
2287 the fly via /proc/fs/lustre/osc/*/checksum_type.
2291 Description: use adler32 for page checksums
2292 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2297 Description: build for x2 fails
2298 Details : fix compile issue on Cray systems.
2302 Description: Properly match for duplicate locks
2303 Details : Due to different lock order from skiplists code, we need to
2304 traverse entire list for now
2307 Frequency : only on PPC/SLES10
2309 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2310 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2311 be defined. Add a hack in mkfs_lustre.c to work around this compile
2316 Description: LASSERT on MDS when client holding flock lock dies
2317 Details : ldlm pool logic depends on number of granted locks equal to
2318 number of released locks which is not true for flock locks, so
2319 just exclude such locks from consideration.
2323 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2324 Details : Use fsfilt_sync() for both the whole filesystem sync and
2325 individual file sync to eliminate dangerous inode locking
2326 with I_LOCK that can lead to a deadlock.
2330 Description: performance in 1.6.3
2331 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2332 sure that 1MB requests are not fragmented by the block layer.
2334 Severity : enhancement
2336 Description: organize the server-side client stats on per-nid basis
2337 Details : Change the structure of stats under obdfilter and mds to
2345 The "uuid"s file would list the uuids of _active_ exports.
2346 And the clear entry is to clear all stats and stale nids.
2348 Severity : enhancement
2350 Description: eliminate client locks in face of contention
2351 Details : file contention detection and lockless i/o implementation
2352 for contended files.
2356 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2359 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2361 Description: MDS may lose file striping (and hence file data) in some cases
2362 Details : If there are additional extended attributes stored on the MDS,
2363 in particular ACLs, SELinux, or user attributes (if user_xattr
2364 is specified for the client mount options) then there is a risk
2365 of attribute loss. Additionally, the Lustre file striping
2366 needs to be larger than default (e.g. striped over all OSTs),
2367 and an additional attribute must be stored initially in the
2368 inode and then increase in size enough to be moved to the
2369 external attribute block (e.g. ACL growing in size) for file
2374 Description: skiplist implementation simplification
2375 Details : skiplists are used to group compatible locks on granted list
2376 that was implemented as tracking first and last lock of each
2377 lock group the patch changes that to using doubly linked lists
2381 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2382 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2383 because LASSERT on both the data supplied by a client, and the
2384 data on disk is dangerous and incorrect.
2386 Severity : enhancement
2388 Description: Slow truncate/writes to huge files at high offsets.
2389 Details : Directly associate cached pages to lock that protect those pages,
2390 this allows us to quickly find what pages to write and remove
2391 once lock callback is received.
2395 Description: more ldlm soft lockups
2396 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2397 starve other threads from the resource lock for a long time in
2398 case of long waiting queue, so change the debug level from
2399 D_OTHER to the less frequently used D_INFO.
2401 Severity : enhancement
2403 Description: add -gid, -group, -uid, -user options to lfs find
2407 Description: Hung threads in invalidate_inode_pages2_range
2408 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2409 once one is completed. As a result, some RPCs are stuck in the
2410 queue and are never sent.
2414 Description: filter threads hungs on waiting journal commit
2415 Details : Cleanup filter group llog code, then only filter group llog will
2416 be only created in the MDS/OST syncing process.
2420 Description: Procfs and llog threads access destoryed import sometimes.
2421 Details : Sync the import destoryed process with procfs and llog threads by
2422 the import refcount and semaphore.
2424 Severity : enhancement
2426 Description: openlock cache of b1_6 port to HEAD
2431 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2432 Details : If insertion of an extent fails, then discard the inode
2433 preallocation and free data blocks else it can lead to duplicate
2438 Description: don't always update ctime in ext3_xattr_set_handle()
2439 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2440 In some cases the ctime should not be updated, for example for
2441 2.0->1.8 compatibility it is necessary to delete an xattr and it
2442 should not update the ctime.
2446 Bugzilla : 15713/16362
2447 Description: Assertion in iopen_connect_dentry in 1.6.3
2448 Details : looking up an inode via iopen with the wrong generation number can
2449 populate the dcache with a disconneced dentry while the inode
2450 number is in the process of being reallocated. This causes an
2451 assertion failure in iopen since the inode's dentry list contains
2452 both a connected and disconnected dentry.
2456 Description: assertion failure in ldlm_handle2lock()
2457 Details : fix a race between class_handle_unhash() and class_handle2object()
2458 introduced in lustre 1.6.5 by bug 13622.
2463 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2464 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2469 Description: oops in sd_iostats_seq_show()
2470 Details : unloading/reloading the scsi low level driver triggers a kernel
2471 bug when trying to access the sd iostat file.
2476 Description: Kernel panics during QLogic driver reload
2477 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2478 patch, causing memory corruption.
2483 Description: journal_dev option does not work in b1_6
2484 Details : pass mount option during pre-mount.
2486 Severity : enhancement
2488 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2489 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2490 extent information of a file. It can be used to map logical blocks
2491 in a file to physical blocks in the block device.
2495 Description: LDLM soft lockups - improvement
2496 Details : It is be possible to send the lock handle along with each read
2497 or write request because the client is already doing a lock match
2498 itself so there isn't any reason the OST should have to re-do that
2502 Frequency : only on Cray X2
2504 Description: X2 build failures
2505 Details : fix build failures on Cray X2.
2509 Description: xid & resent requests
2510 Details : Initialize RPC XID from clock at startup (randomly if clock is
2513 Severity : enhancement
2515 Description: Add lustre_start utility to start or stop multiple Lustre servers
2520 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2521 Details : In case of memory pressure, list_del() can be called twice on
2522 req->rq_history_list, causing a kernel oops.
2526 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2527 Details : fix stack overflow in the distributed lock manager by defering
2528 export eviction after a failed AST to the elt thread instead of
2529 handling it in the dlm interpret routine.
2533 Description: Convert some comments to new format.
2534 Details : Update documenting comments to match doxygen conventions.
2538 Description: Grammar fixes.
2539 Details : A couple of trivial spelling fixes.
2543 Description: OSD_COUNTERS-mandatory
2544 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2545 to simplify the code.
2549 Description: simplify cmm_device freeing logic.
2550 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2554 Description: Add lockdep support to dt_object_operations locking interface.
2555 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2556 parameter indicating lock ordering. Update mdd code to use new
2561 Description: Introduce failloc constants for lockless IO tests.
2562 Details : Add two new failloc constants to test lockless IO. Only one of
2563 them in implemented---another is checked in yet to be landed
2568 Description: Add lockdep support for inode mutex.
2569 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2570 in the situations where Lustre has to lock more than one inode
2575 Description: Add optional invariants checking support.
2576 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2577 configure switch. This macro is to be used for consistency and
2578 sanity checks that are too expensive to be left in `production'
2583 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2584 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2589 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2590 Details : Use spin_lock_nested() in (the only) situation where more than
2591 one ldlm_lock is locked simultaneously. Also, fix possible
2592 dead-lock in ldlm_lock_change_resource() by enforcing particular
2597 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2598 Details : Instead of specifying each ldlm_lock call-back through separate
2599 parameter, wrap them into struct ldlm_callback_suite.
2603 Description: Kill join_lru obd method and its callers.
2604 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2605 in memory---a requirement implemented through ->o_join_lru() obd
2606 method in HEAD. Remove this method and its users.
2610 Description: Add asynchronous ldlm ENQUEUE completion handler.
2611 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2612 case that stock ldlm_completion_ast() cannot handle as it waits
2613 until lock is granted. Introduce new ldlm_completion_ast_async()
2614 for this. Also comment ldlm_completion_ast().
2618 Description: ldlm_error <-> errno conversion.
2619 Details : Add functions to map (rather arbitrary) between LDLM error codes
2620 and standard errno values. CLIO needs this to prevent LDLM specific
2621 constants from escaping ldlm and osc.
2625 Description: Kill unused ldlm_handle2lock_ns() function.
2626 Details : Kill unused ldlm_handle2lock_ns() function.
2630 Description: Add lu_ref support to ldlm_lock
2631 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2632 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2633 and ldlm_resource. LDLM interface has to be changed, because code that
2634 releases a reference on a lock, has to "know" what reference this is.
2635 In the most frequent case
2637 lock = ldlm_handle2lock(handle);
2639 LDLM_LOCK_PUT(lock);
2641 no changes are required. When any other reference (received _not_ from
2642 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2643 instead of LDLM_LOCK_PUT().
2645 Arguably, changes are pervasive, and interface requires some discipline
2646 for proper use. On the other hand, it was very instrumental in finding
2647 a few leaked lock references.
2651 Description: Add ldlm_lock_addref_try().
2652 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2653 attempts to addref a lock that might be being canceled
2658 Description: Add ldlm_weigh_callback().
2659 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2660 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2661 hard-coded `number of pages' logic.
2665 Description: Add lockdep annotations to llog code.
2666 Details : Use appropriately tagged _nested() locking calls in the places
2667 where llog takes more than one ->lgh_lock lock.
2671 Description: Add loi_kms_set().
2672 Details : Wrap kms updates into a helper function.
2676 Description: Constify instances of struct lsm_operations.
2677 Details : Constify instances of struct lsm_operations.
2681 Description: lu_conf support.
2682 Details : On a server, a file system object is uniquely identified
2683 by a fid, which is sufficient to locate and load all object
2684 state (inode). On a client, on the other hand, more data are
2685 necessary instantiate an object. Change lu_object_find() and
2686 friends to take additional `lu_conf' argument describing object.
2687 Typically this includes layout information.
2691 Description: lu_context fixes.
2692 Details : Introduce new lu_context functions that are needed on the client
2693 side, where some system threads (ptlrpcd) are shared by multiple
2694 modules, and so cannot be stopped during module shutdown.
2698 Description: Add start and stop methods to lu_device_type_operations.
2699 Details : Introduce two new methods in lu_device_type_operations, that are
2700 invoked when first instance of a given type is created and last one
2701 is destroyed respectively. This is need by CLIO.
2705 Description: Add lu_ref support to struct lu_device.
2706 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2707 track leaked references.
2711 Description: Introduce lu_kmem_descr.
2712 Details : lu_kmem_descr and its companion interface allow to create
2713 and destroy a number of kmem caches at once.
2717 Description: Fix lu_object finalization race.
2718 Details : Fix a race between lu_object_find() finding an object and its
2719 concurrent finalization. This race is (most likely) not possible
2720 on the server, but might happen on the client.
2724 Description: Introduce lu_ref interface.
2725 Details : lu_ref is a debugging module allowing to track references to
2726 a given object. It is quite cpu expensive, and has to be
2727 explicitly enabled with --enable-lu_ref. See usage description
2732 Description: Factor lu_site procfs stats into a separate function.
2733 Details : Separate lu_site stats printing code into a separate function
2734 that can be reused on a client.
2738 Description: Constify instances of struct {lu,dt,md}_device_operations.
2739 Details : Constify instances of struct {lu,dt,md}_device_operations.
2743 Description: Introduce struct md_site and move meta-data specific parts of
2744 struct lu_site here.
2745 Details : Move md-specific fields out of struct lu_site into special struct
2746 md_site, so that lu_site can be used on a client.
2750 Description: Kill mdd_lov_destroy().
2751 Details : Remove unused mdd code.
2755 Description: Add st_block checking to multistat.c.
2756 Details : Add st_block checking to multistat.c.
2760 Description: Add lu_ref support to struct obd_device.
2761 Details : Add lu_ref tracking to obd_device.
2765 Description: Kill obd_set_fail_loc().
2766 Details : Remove unused code.
2770 Description: Add special type for ptlrpc_request interpret functions.
2771 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2772 there. Actual usage will be in CLIO.
2776 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2777 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2778 doesn't work with lockdep.
2782 Description: Add rwv.c test program.
2783 Details : New testing program exercising readv(2) and writev(2) (Qian).
2787 Description: Add sendfile.c test program.
2788 Details : New testing program exercising sendfile(2) (Jay).
2792 Description: Ratelimit a message that can be very frequent.
2793 Details : Ratelimit a memory allocation failure message that can
2798 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2799 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2803 Description: Liblustre build fixes.
2804 Details : Liblustre build fixes.
2808 Description: libcfs: add cfs_{need,cond}_resched() interface.
2809 Details : libcfs: add cfs_{need,cond}_resched() definition and
2810 implementations for Linux, NT, and liblustre.
2812 Severity : enhancement
2814 Description: More exported tunables for mballoc
2815 Details : Add support for tunable preallocation window and new tunables for
2816 large/small requests
2820 Description: Detect corruption of block bitmap and checking for preallocations
2821 Details : Checks validity of on-disk block bitmap. Also it does better
2822 checking of number of applied preallocations. When corruption is
2823 found, it turns filesystem readonly to prevent further corruptions.
2827 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2828 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2829 window to 0. If it does hit cache miss, and read-pattern is still
2830 stride-io mode, does not reset the stride window, but also does not
2831 increase the stride window length in this case.
2835 Frequency : only for big-endian servers
2836 Description: Check if system is big-endian while mounting fs with extents feature
2837 Details : Mounting a filesystem with extents feature will fail on big-endian
2838 systems since ext3-based ldiskfs is not supported on big-endian
2839 systems. This can be over-riden with "bigendian_extents" mount option.
2841 Severity : enhancement
2843 Description: The root squash functionality
2844 Details : A security feature, which is to prevent users from being able
2845 to mount lustre on their desktop, run as root, and delete
2846 all of the files in the filesystem. The goal is accomplished by
2847 remapping user id (UID) and group id (GID) of the root user to
2848 a UID and GID specified by the system administartor via Lustre
2849 configuration management server (MGS). The functionality also
2850 allows to specify sets of clients for which the remapping does
2855 Description: Excessive recovery window
2856 Details : With AT enabled, the recovery window can be excessively long (6000+
2857 seconds). To address this problem, we no longer use
2858 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2859 timeout no longer depends on the service time, it is set to
2860 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2861 time via pb_service_time.
2865 Description: Watchdog triggered on MDS failover
2866 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2867 the OSTs know that the MDS "UUID" can be reused for the same export
2868 from a different NID, so we do not need to wait for the export to be
2872 Frequency : rare, only if using MMP with Linux RAID
2874 Description: MMP doesn't work with Linux RAID
2875 Details : While using HA for Lustre servers with Linux RAID, it is possible
2876 that MMP will not detect multiple mounts. To make this work we
2877 need to unplug the device queue in RAID when the MMP block is being
2878 written. Also while reading the MMP block, we should read it from
2879 disk and not the cached one.
2881 Severity : enhancement
2883 Description: open file using fid
2884 Details : A file can be opened using just its fid, like
2885 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2888 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2891 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2892 Details : add proper locking for creation of jbd_4k slab cache
2896 Description: MMP check in ext3_remount() fails without displaying any error
2897 Details : When multiple mount protection fails during remount, proper error
2900 Severity : enhancement
2902 Description: Allow stripe size to be up to 4G-64k
2903 Details : Fix math logic to allow large stripe sizes.
2907 Description: add check for >8TB ldiskfs filesystems
2908 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2909 Don't allow >8TB ldiskfs filesystems to be mounted without
2910 force_over_8tb mount option
2915 Descriptoin: OSS complains "too many blocks" and crashes
2916 Details : In filter_commitrw_write():
2917 1. lquota_chkquota();
2918 2. obdo_from_inode();
2919 3. lquota_pending_commit();
2920 "2." has changed uid and gid in this case, then caused
2921 lquota_chkquota() and lquota_pending_commit() to use
2922 different uid and gid. That is the root of the bug.
2926 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2927 Details : Detect and print out the 'stripe count = -1' case properly.
2931 Description: Out or order replies might be lost on replay
2932 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2933 request with tid smaller than one currently being added, add it
2934 to the start, not end of the list.
2936 --------------------------------------------------------------------------------
2938 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2940 * Support for kernels:
2941 2.6.5-7.283 (SLES 9),
2942 2.6.9-55.EL (RHEL 4),
2943 2.6.16.46-0.14 (SLES 10),
2944 2.6.18.8 vanilla (kernel.org)
2945 * Client support for unpatched kernels:
2946 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2947 2.6.16 - 2.6.22 vanilla (kernel.org)
2948 * Due to recently discovered recovery problems, we do not recommend
2949 using patchless RHEL 4 clients with this or any earlier release.
2950 * Recommended e2fsprogs version: 1.39.cfs8
2951 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2952 * Starting with this release, the ldiskfs backing filesystem required
2953 by Lustre is now in its own package, lustre-ldiskfs. This package
2954 should be installed. It is versioned separately from Lustre and
2955 may be released separately in future.
2959 Description: block reactivating mgc import until all deactivates complete
2960 Details : Fix race when failing back MDT/MGS to itself (testing)
2963 Frequency : at statup only
2965 Description: mds_lov_synchronize race leads to various problems
2966 Details : simultaneous MDT->OST connections at startup can cause the
2967 sync to abort, leaving the OSC in a bad state.
2969 Severity : enhancement
2971 Description: add optional extra BUILD_VERSION info
2972 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2973 to override the lustre version.
2976 Frequency : 2.6.18 servers only
2978 Description: ll_kern_mount() doesn't release the module reference
2979 Details : The ldiskfs module reference count never drops down to 0
2980 because ll_kern_mount() doesn't release the module reference.
2985 Description: server LBUG when using old ost_num_threads parameter
2986 Details : Accept the old ost_num_threads parameter but warn that it
2987 is deprecated, and fix an off-by-one error that caused an LBUG.
2992 Description: Transient SCSI error results in persistent IO issue
2993 Details : iobuf->dr_error is not reinitialized to 0 between two
2997 Frequency : sometimes when underlying device returns I/O errors
2999 Description: OSTs not going read-only during write failures
3000 Details : OSTs are not remounted read-only when the journal commit threads
3001 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3002 instead of the ext3 wrappers.
3006 Description: poor connect scaling with increasing client count
3007 Details : Don't run filter_grant_sanity_check for more than 100 exports
3008 to improve scaling for large numbers of clients.
3011 Frequency : SLES10 only
3013 Description: sanity-quota.sh quotacheck failed: rc = -22
3014 Details : Quotas cannot be enabled on SLES10.
3017 Frequency : liblustre clients only
3019 Description: getdirentries does not give error when run on compute nodes
3020 Details : getdirentries does not fail when the size specified as an argument
3021 is too small to contain at least one entry
3023 Severity : enhancement
3025 Description: Add LNET router traceability for debug purposes
3026 Details : If a checksum failure occurs with a router as part of the
3027 IO path, the NID of the last router that forwarded the bulk data
3028 is printed so it can be identified.
3033 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3034 Details : Due to a race condition, liblustre clients were occasionally
3035 evicted incorrectly.
3037 Severity : enhancement
3039 Description: lfs setstripe use optional parameters instead of postional
3042 Severity : enhancement
3044 Description: Nanosecond timestamp support for ldiskfs
3045 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3046 resolution timestamps. There is not yet support for this at
3047 the Lustre filesystem level.
3050 Frequency : during server recovery
3052 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3053 Details : request with rq_no_resend flag not awake l_wait_event if they get
3057 Frequency : nfs export on patchless client
3059 Description: connectathon hang when test nfs export over patchless client
3060 Details : Disconnected dentry cannot be found with lookup, so we do not need
3061 to unhash it or make it invalid
3064 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3065 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3066 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3071 Description: open req refcounting wrong on reconnect
3072 Details : If reconnect happened between getting open reply from server and
3073 call to mdc_set_replay_data in ll_file_open, we will schedule
3074 replay for unreferenced request that we are about to free.
3075 Subsequent close will crash in variety of ways.
3076 Check that request is still eligible for replay in
3077 mdc_set_replay_data().
3082 Description: disable writes to filesystem when reading health_check file
3083 Details : the default for reading the health_check proc file has changed
3084 to NOT do a journal transaction and write to disk, because this
3085 can cause reads of the /proc file to hang and block HA state
3086 checking on a healthy but otherwise heavily loaded system. It
3087 is possible to return to the previous behaviour during configure
3088 with --enable-health-write.
3090 Severity : enhancement
3092 Description: 64-bit inode version
3093 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3094 to the inode. This will be required for version-based recovery.
3099 Description: MDS fails to start if a duplicate client export is detected
3100 Details : in some rare cases it was possible for a client to connect to
3101 an MDS multiple times. Upon recovery the MDS would detect this
3102 and fail during startup. Handle this more gracefully.
3104 Severity : enhancement
3106 Description: Add -o localflock option to simulate old noflock
3108 Details : This will achieve local-only flock/fcntl locks
3114 Description: log_commit_thread vs filter_destroy race leads to crash
3115 Details : Take import reference before releasing llog record semaphore
3120 Description: Wrong request locking in request set processing
3121 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3122 imp_delayed_list, in this place should be used imp_lock.
3125 Frequency : when reconnection
3127 Description: Grant Leak when osc reconnect to OST
3128 Details : When osc reconnect ost, OST(filter) should check whether it
3129 should grant more space to client by comparing fed_grant and
3130 cl_avail_grant, and return the granted space to client instead
3131 of "new granted" space, because client will call osc_init_grant
3132 to update the client grant space info.
3135 Frequency : when client reconnect to OST
3137 Description: Grant Leak when osc do resend and replay bulk write
3138 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3139 bulk write request, because the grant info while be sync between
3140 OSC and OST when reconnect, and we should ignore the grant info
3141 these of resend/replay write req.
3146 Description: Grant space more than available left space sometimes.
3147 Details : When then OST is about to be full, if two bulk writing from
3148 different clients came to OST. Accord the avaliable space of the
3149 OST, the first req should be permitted, and the second one
3150 should be denied by ENOSPC. But if the seconde arrived before
3151 the first one is committed. The OST might wrongly permit second
3152 writing, which will cause grant space > available space.
3155 Frequency : when client is evicted
3157 Description: Grant might be wrongly erased when osc is evicted by OST
3158 Details : when the import is evicted by server, it will fork another
3159 thread ptlrpc_invalidate_import_thread to invalidate the
3160 import, where the grant will be set to 0. While the original
3161 thread will update the grant it got when connecting. So if
3162 the former happened latter, the grant will be wrongly errased
3163 because of this race.
3168 Description: Checking Stale with correct fid
3169 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3170 to check whether it is stale, because sometimes, we want the
3171 enqueue happened anyway, and op_data.fid2 will not be initialized.
3173 Severity : enhancement
3175 Description: update patchless client
3176 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3179 Frequency : only with 2.4 kernel
3181 Description: random memory corruption
3182 Details : size of struct ll_inode_info is to big for union inode.u and this
3183 can be cause of random memory corruption.
3188 Description: Memory leak in recovery
3189 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3190 It should also check obd_fail before fsfilt_start, otherwise if
3191 fsfilt_start return -EROFS,(failover mds during mds recovery).
3192 then the req will return with repmsg->transno = 0 and rc = EROFS.
3193 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3194 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3195 be freed no matter whether fsfilt_commit success or not.
3198 Frequency : only with huge count clients
3200 Description: Prevents from taking the superblock lock in llap_from_page for
3202 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3203 for llap_from_page call in ll_removepage prevents from taking the
3204 superblock lock for a soon died page.
3209 Description: Not check open intent error before release open handle
3210 Details : in some rare cases, the open intent error is not checked before
3211 release open handle, which may cause
3212 ASSERTION(open_req->rq_transno != 0), because it tries to release
3213 the failed open handle.
3218 Description: Set cat log bitmap only after create log success.
3219 Details : in some rare cases, the cat log bitmap is set too early. and it
3220 should be set only after create log success.
3224 Description: Accessing a block bevice can re-enable I/O when Lustre is
3225 tearing down a device.
3226 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3231 Description: service threads may hog cpus when there are a lot of requests
3232 Details : Insert cond_resched to give other threads a chance to use some CPU
3237 Description: the cat log was not initialized in recovery
3238 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3239 unlink log on mds will not be initialized until mds post
3240 recovery. And also in mds post recovery, the unlink log will
3241 initialization will be done asynchronausly, so there will be race
3242 between add unlink log and unlink log initialization.
3246 Description: brw_stats were being printed incorrectly
3247 Details : brw_stats were being printed as log2 but all of them were not
3248 recorded as log2. Also remove some code duplication arising from
3249 filter_tally_{read,write}.
3253 Frequency : rare, only in recovery.
3254 Description: ASSERTION(req->rq_type != LI_POISON) failed
3255 Details : imp_lock should be held while iterating over imp_sending_list for
3256 prevent destroy request after get timeout in ptlrpc_queue_wait.
3260 Description: replay-single.sh test 52 fails
3261 Details : A lock's skiplist need to be cleanup when it being unlinked
3262 from its resource list.
3266 Description: Short directio read returns full requested size rather than
3268 Details : Direct I/O operations should return actual amount of bytes
3269 transferred rather than requested size.
3271 Severity : enhancement
3273 Description: metadata RPC reduction (e.g. for rm performance)
3274 Details : decrease the amount of synchronous RPC between clients and servers
3275 by canceling conflicing lock before the operation on the client
3276 and packing thier handles into the main operation RPC to server.
3278 Severity : enhancement
3280 Description: Async OSC create to avoid the blocking unnecessarily.
3281 Details : If an OST has no remaining object, system will block on the
3282 creation when it needs to create a new object on this OST. Now,
3283 ways use pre-created objects when available, instead of blocking on
3284 an empty osc while others are not empty. If we must block, we
3285 block for the shortest possible period of time.
3289 Description: improve handling recoverable errors
3290 Details : if request processig with error which can be recoverable on server
3291 request should be resend, otherwise page released from cache and
3294 Severity : enhancement
3296 Description: refine locking for avoid write wrong info into lov_objid file
3297 Details : fix possible races with add new target and write/update data in
3300 --------------------------------------------------------------------------------
3302 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3307 Frequency : on some architectures
3309 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3310 Details : Uninitialized flags sometimes cause configuration commands to
3314 Frequency : patchless clients only
3316 Description: missing __iget() symbol export
3317 Details : The __iget() symbol export is missing. To avoid the need for
3318 this on patchless clients the deathrow inode reaper is turned
3319 off, and we depend on the VM to clean up old inodes. This
3320 dependency was during via the fix for bug 12181.
3322 --------------------------------------------------------------------------------
3324 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3326 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3327 INTEROPERATE with older versions automatically. In many cases a
3328 special upgrade step is needed. Please read the
3329 user documentation before upgrading any part of a live system.
3330 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3331 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3332 * WARNING: Lustre configuration and startup changes are required with
3333 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3337 Severity : enhancement
3339 Description: MountConf
3340 Details : Lustre configuration is now managed via mkfs and mount
3341 commands instead of lmc and lconf. New obd types (MGS, MGC)
3342 are added for dynamic configuration management. See
3343 https://mail.clusterfs.com/wikis/lustre/MountConf for
3346 Severity : enhancement
3348 Description: dynamic OST addition
3349 Details : OSTs can now be added to a live filesystem
3351 Severity : enhancement
3353 Description: startup order invariance
3354 Details : MDTs and OSTs can be started in any order. Clients only
3355 require the MDT to complete startup.
3357 Severity : enhancement
3359 Description: parallel, asynchronous orphan cleanup
3360 Details : orphan cleanup is now performed in separate threads for each
3361 OST, allowing parallel non-blocking operation.
3363 Severity : enhancement
3365 Description: optimized stripe assignment
3366 Details : stripe assignments are now made based on ost space available,
3367 ost previous usage, and OSS previous usage, in order to try
3368 to optimize storage space and networking resources.
3370 Severity : enhancement
3372 Description: Permanently set tunables
3373 Details : All writable /proc/fs/lustre tunables can now be permanently
3374 set on a per-server basis, at mkfs time or on a live system.
3376 Severity : enhancement
3378 Description: Lustre message v2
3379 Details : Add lustre message format v2.
3381 Severity : enhancement
3383 Description: client OST exclusion list
3384 Details : Clients can be started with a list of OSTs that should be
3385 declared "inactive" for known non-responsive OSTs.
3389 Description: SPEC SFS validation failure on NFS v2 over lustre.
3390 Details : Changes the blocksize for regular files to be 2x RPC size,
3391 and not depend on stripe size.
3393 Severity : enhancement
3395 Description: Multiple MD RPCs in flight.
3396 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3397 To avoid overly-overloading MDS, introduce a limit on number of
3398 MDS RPCs in flight for a single client and add /proc controls
3399 to adjust this limit.
3401 Severity : enhancement
3403 Description: client read/write statistics
3404 Details : Add client read/write call usage stats for performance
3405 analysis of user processes.
3406 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3407 file access. extents_stats shows chunk size distribution.
3408 extents_stats_per_process show chunk size distribution per
3411 Severity : enhancement
3413 Description: mds statistics
3414 Details : Add detailed mds operations statistics in
3415 /proc/fs/lustre/mds/*/stats.
3419 Description: Failure of copying files with lustre special EAs.
3420 Details : Client side always return success for setxattr call for lustre
3421 special xattr (currently only "trusted.lov").
3426 Description: Refcount LNET uuids
3427 Details : The global LNET uuid list grew linearly with every startup;
3428 refcount repeated list entries instead of always adding to
3431 Severity : enhancement
3433 Description: Dynamic service threads
3434 Details : Within a small range, start extra service threads
3435 automatically when the request queue builds up.
3438 Frequency : mixed-endian client/server environments
3440 Description: mixed-endian crashes
3441 Details : The new msg_v2 system had some failures in mixed-endian
3444 Severity : enhancement
3446 Description: Easy OST removal
3447 Details : OSTs can be permanently deactivated with e.g. 'lctl
3448 conf_param lustre-OST0001.osc.active=0'
3450 Severity : enhancement
3452 Description: MGS proc entries
3453 Details : Added basic proc entries for the MGS showing what filesystems
3456 Severity : enhancement
3458 Description: provide MGS failover
3459 Details : Added config lock reacquisition after MGS server failover.
3461 Severity : enhancement
3463 Description: add Linux 2.4 support
3464 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3468 Description: a large application tries to do I/O to the same resource and dies
3469 in the middle of it.
3470 Details : Check the req->rq_arrival time after the call to
3471 ost_brw_lock_get(), but before we do anything about
3472 processing it & sending the BULK transfer request. This
3473 should help move old stale pending locks off the queue as
3474 quickly as obd_timeout.
3477 Frequency : when an incorrect nid is specified during startup
3479 Description: ptlrpc connect to non-existant node causes kernel crash
3480 Details : LNET can't be re-entered from an event callback, which
3481 happened when we expire a message after the export has been
3482 cleaned up. Instead, hand the zombie cleanup off to another
3485 Severity : enhancement
3487 Description: plain/inodebits lock performance improvement
3488 Details : Grouping plain/inodebits in granted list by their request modes
3489 and bits policy, thus improving the performance of search through
3493 Frequency : only if OST filesystem is corrupted
3495 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3496 Details : for a short time RPCs with bulk IO are in the replay list,
3497 but replay of bulk IOs is unimplemented. If the OST filesystem
3498 is corrupted due to disk cache incoherency and then replay is
3499 started it is possible to trip an assertion. Avoid putting
3500 committed RPCs into the replay list at all to avoid this issue.
3503 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3505 Description: System hang on startup
3506 Details : This bug allowed the liblustre (e.g. catamount) client to
3507 return to the app before handling all startup RPCs. This
3508 could leave the node unresponsive to lustre network traffic
3509 and manifested as a server ptllnd timeout.
3511 Severity : enhancement
3513 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3514 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3515 boolean to control whether to print peer debug info when a
3516 client's RPC times out.
3519 Frequency : only for kernels with patches from Lustre below 1.4.3
3521 Description: Remove old rdonly API
3522 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3525 Frequency : only for devices with external journals
3527 Description: Set external device read-only also
3528 Details : During a commanded failover stop, we set the disk device
3529 read-only while the server shuts down. We now also set any
3530 external journal device read-only at the same time.
3533 Frequency : when upgrading from 1.4 while trying to change parameters
3535 Description: The wrong (new) MDC name was used when setting parameters for
3536 upgraded MDT's. Also allows changing of OSC (and MDC)
3537 parameters if --writeconf is specified at tunefs upgrade time.
3540 Frequency : when setting specific ost indicies
3542 Description: QOS code breaks on skipped indicies
3543 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3544 created with --index need not be sequential.
3546 Severity : enhancement
3548 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3549 Details : The uninit_groups feature works in conjunction with the kernel
3550 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3551 up the pass1 processing of e2fsck. This is a read-only feature
3552 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3553 mount filesystems that have had this feature enabled.
3555 Severity : enhancement
3557 Description: Improve multi-block allocation algorithm to avoid fragmentation
3558 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3559 allocation locality and avoid filesystem fragmentation.
3561 ------------------------------------------------------------------------------
3563 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3565 * Support for kernels:
3566 2.4.21-47.0.1.EL (RHEL 3)
3567 2.6.5-7.283 (SLES 9)
3568 2.6.9-42.0.10.EL (RHEL 4)
3569 2.6.12.6 vanilla (kernel.org)
3570 2.6.16.27-0.9 (SLES 10)
3571 * Recommended e2fsprogs version: 1.39.cfs5
3573 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3577 Frequency : occasional, depends on client load and configuration
3578 Bugzilla : 12181, 12203
3579 Description: data loss for recently-modified files
3581 Details : In some cases it is possible that recently written or created
3582 files may not be written to disk in a timely manner (this should
3583 normally be within 30s unless client IO load is very high).
3584 The problem appears as zero-length files or files that are a
3585 multiple of 1MB in size after a client crash or client eviction
3586 that are missing data at the end of the file.
3588 This problem is more likely to be hit on clients where files are
3589 repeatedly created and unlinked in the same directory, clients
3590 have a large amount of RAM, have many CPUs, the filesystem has
3591 many OSTs, the clients are rebooted frequently, and/or the files
3592 are not accessed by other nodes after being written.
3594 The presence of the problem can be detected by looking at
3595 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3596 smaller than the second (nr_unused) then dirty files will not
3597 be flushed automatically to disk. "sync; sleep 10" should be
3598 run several times on the node before unmounting it to update
3599 Lustre (this is also safe to run on nodes without this problem).
3601 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3602 that can cause this same problem, so customers using that kernel
3603 also need to update the kernel in addition to Lustre. In order
3604 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3606 It is normal that files written just before a client crash (less
3607 than 30s) may not yet have been flushed to disk, even for local
3611 Frequency : frequent on thin XT3 nodes
3613 Description: UUID collision on thin XT3 Linux nodes
3614 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3615 randomly, since we relied on an insufficiently-seeded PRNG.
3620 Description: OSS hangs after "All ost request buffers busy"
3621 Details : A deadlock between quota and journal operations caused OSS
3622 hangs after printing "All ost request buffers busy."
3625 Frequency : always on liblustre builds
3627 Description: Cleanup compiler warnings on liblustre
3630 Frequency : always on liblustre builds on XT3
3632 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3636 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3637 > 32000 subdirectories
3638 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3639 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3640 subdirectory count crosses 32000. This will aid e2fsck to
3641 correctly handle more than 32000 subdirectories.
3644 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3646 Description: System hang on startup
3647 Details : This bug allowed the liblustre (e.g. catamount) client to
3648 return to the app before handling all startup RPCs. This
3649 could leave the node unresponsive to lustre network traffic
3650 and manifested as a server ptllnd timeout.
3652 Severity : enhancement
3654 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3655 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3656 boolean to control whether to print peer debug info when a
3657 client's RPC times out.
3662 Description: make O_SYNC working on 2.6 kernels
3663 Details : 2.6 kernels use different method for mark pages for write,
3664 so need add a code to lustre for O_SYNC work.
3669 Description: Failure to close file and release space on NFS
3670 Details : Put inode details into lock acquired in ll_intent_file_open.
3671 Use mdc_intent_lock in ll_intent_open to properly
3672 detect all kind of errors unhandled by mdc_enqueue.
3677 Description: proc file read during shutdown sometimes raced obd removal,
3679 Details : Add lock to prevent obd access after proc file removal.
3682 Frequency : Only for files larger than 4GB on 32-bit clients.
3684 Description: improperly doing page alignment of locks
3685 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3686 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3691 Description: RPCs being resent when they shouldn't be
3692 Details : Some RPCs that should not be resent are being resent. This
3693 can cause inconsistencies in the RPC state machine. Do not
3694 resend such requests.
3697 Frequency : rare, only with NFS export
3699 Description: Crash on NFS re-export node
3700 Details : under very unusual load conditions an assertion is hit in
3701 ll_intent_file_open()
3704 Frequency : only if OST filesystem is corrupted
3706 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3707 Details : for a short time RPCs with bulk IO are in the replay list,
3708 but replay of bulk IOs is unimplemented. If the OST filesystem
3709 is corrupted due to disk cache incoherency and then replay is
3710 started it is possible to trip an assertion. Avoid putting
3711 committed RPCs into the replay list at all to avoid this issue.
3716 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3717 Details : Large single O_DIRECT read and write calls can fail to allocate
3718 a sufficiently large buffer to process the request. In case of
3719 allocation failure the allocation is retried with a smaller
3720 buffer and broken into smaller requests.
3722 Severity : enhancement
3724 Description: Add -o localflock option to simulate old noflock behaviour.
3725 Details : This will achieve local-only flock/fcntl locks coherentness.
3730 Description: versioning check is incomplete
3731 Details : Checking the version difference of client vs. server, report
3732 error if the gap is too big.
3737 Description: add support PG_writeback bit
3738 Details : add support for PG_writeback bit for Lustre, for more carefull
3739 work with page cache in 2.6 kernel. This also fix some deadlocks
3740 and remove hack for work O_SYNC with 2.6 kernel.
3742 Severity : enhancement
3744 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3745 Details : The uninit_groups feature works in conjunction with the kernel
3746 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3747 up the pass1 processing of e2fsck. This is a read-only feature
3748 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3749 mount filesystems that have had this feature enabled.
3751 Severity : enhancement
3753 Description: Improve multi-block allocation algorithm to avoid fragmentation
3754 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3755 allocation locality and avoid filesystem fragmentation.
3757 ------------------------------------------------------------------------------
3759 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3761 * Support for kernels:
3762 2.6.9-42.0.3.EL (RHEL 4)
3763 2.6.5-7.276 (SLES 9)
3764 2.4.21-47.0.1.EL (RHEL 3)
3765 2.6.12.6 vanilla (kernel.org)
3766 2.6.16.21-0.8 (SLES10)
3767 * Recommended e2fsprogs version: 1.39.cfs2-0
3769 * The backwards-compatible /proc/sys/portals symlink has been removed
3770 in this release. Before upgrading, please ensure that you change
3771 any configuration scripts or /etc/sysctl.conf files that access
3772 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3773 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3774 in advance of the upgrade on any system running Lustre 1.4.6 or
3775 newer, since /proc/sys/lnet was added in that version.
3776 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3780 Frequency : only when quota is used
3782 Description: avoid scanning export list for quota master
3783 Details : Change the algorithms to avoid scanning export list in order
3784 to improve the efficiency.
3787 Frequency : MDS failover only, very rarely
3789 Description: "went back in time" messages on mds failover
3790 Details : The greatest transno may be lost when the current operation
3791 finishes with an error (transno==0) and the client's last_rcvd
3792 record is over-written. Save the greatest transno in the
3793 mds_last_transno for this case.
3796 Frequency : always for specific kernels and striping counts
3798 Description: client may get "Matching packet too big" without ACL support
3799 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3800 when trying to access files in certain configurations. The
3801 clients should in fact be denied when mounting because they do
3802 not understand ACLs.
3805 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3807 Description: many clients connecting with IO in progress causes connect timeouts
3808 Details : Avoid synchronous journal commits to avoid delays caused by many
3809 clients connecting/disconnecting when bulk IO is in progress.
3810 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3811 OST_IO_PORTAL to avoid delays behind potentially many pending
3815 Frequency : occasionally with multiple writers to a single file
3817 Description: shared writes to file may result in wrong size reported by stat()
3818 Details : Allow growing of kms when extent lock is cancelled
3821 Frequency : always with random mmap IO to multi-striped file
3823 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3824 Details : Only if the hole is at the end of OST object so that kms is too
3825 small. Fix is to increase kms accordingly in ll_nopage.
3828 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3830 Description: writes to a missing object would leak memory on the OST
3831 Details : If there is an inconsistency between the MDS and OST filesystems,
3832 such that the MDS references an object that doesn't exist, writes
3833 to that object will leak memory due to incorrect cleanup in the
3834 error handling path, eventually running out of memory on the OST.
3839 Description: Creating too long symlink causes lustre errors
3840 Details : Check symlink and name lengths before sending requests to MDS.
3843 Frequency : only if flock is enabled (not on by default)
3845 Description: posix locks not released on fd closure on 2.6.9+
3846 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3847 kernels, this caused such locks not to be released on fd close and
3848 then assertions on fs unmount about still used locks.
3851 Frequency : MDS failover only, very rarely
3853 Description: clients may get ASSERTION(granted_lock != NULL)
3854 Details : When request was taking a long time, and a client was resending
3855 a getattr by name lock request. The were multiple lock requests
3856 with the same client lock handle and
3857 mds_getattr_name->fixup_handle_for_resent_request found one of the
3858 lock handles but later failed with ASSERTION(granted_lock != NULL).
3863 Description: handle->h_buffer_credits > 0, assertion failure
3864 Details : h_buffer_credits is zero after truncate, causing assertion
3865 failure. This patch extends the transaction or creates a new
3869 Frequency : NFS re-export or patchless client
3870 Bugzilla : 11179, 10796
3871 Description: Crash on NFS re-export node (__d_move)
3872 Details : We do not want to hash the dentry if we don't have a lock.
3873 But if this dentry is later used in d_move, we'd hit uninitialised
3874 list head d_hash, so we just do this to init d_hash field but
3875 leave dentry unhashed.
3878 Frequency : NFS re-export or patchless client
3880 Description: NFS exports has problem with symbolic link
3881 Details : lustre client didn't properly install dentry when re-exported
3882 to NFS or running patchless client.
3885 Frequency : NFS re-export or patchless client
3887 Description: Various nfs/patchless fixes.
3888 Details : fixes reuse disconected alias for lookup process - this fixes
3889 warning "find_exported_dentry: npd != pd",
3890 fix permission error with open files at nfs.
3891 fix apply umask when do revalidate.
3894 Frequency : occasional
3896 Description: Crash on NFS re-export node
3897 Details : calling clear_page() on the wrong pointer triggered oops in
3898 generic_mapping_read().
3901 Frequency : rarely, using O_DIRECT IO
3903 Description: unaligned directio crashes client with LASSERT
3904 Details : check for unaligned buffers before trying any requests.
3907 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3909 Description: stale data returned from RAID cache
3910 Details : If only a small amount of IO is done to the RAID device before
3911 reading it again it is possible to get stale data from the RAID
3912 cache instead of reading it from disk.
3915 Frequency : always for sles10 kernel
3917 Description: sles10 support
3918 Details : ll_follow_link: compile fixes and using of nd_set_link
3919 under newer kernels.
3922 Frequency : depends on arch, kernel and compiler version, always on sles10
3925 Description: recursive or deep enough symlinks cause stack overflow
3926 Details : getting rid of large stack-allocated variable in
3930 Frequency : depends on hardware
3932 Description: lustre write performance loss in the SLES10 kernel
3933 Details : the performance loss is caused by using of write barriers in the
3934 ext3 code. The SLES10 kernel turns barrier support on by
3935 default. The fix is to undo that change for ldiskfs.
3937 ------------------------------------------------------------------------------
3939 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3941 * Support for kernels:
3942 2.6.9-42.0.3EL (RHEL 4)
3943 2.6.5-7.276 (SLES 9)
3944 2.4.21-47.0.1.EL (RHEL 3)
3945 2.6.12.6 vanilla (kernel.org)
3949 Frequency : quota enabled and large files being deleted
3951 Description: releasing more than 4GB of quota at once hangs OST
3952 Details : If a user deletes more than 4GB of files on a single OST it
3953 will cause the OST to spin in an infinite loop. Release
3954 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3959 Description: statfs data retrieved from /proc may be stale or zero
3960 Details : When reading per-device statfs data from /proc, in the
3961 {kbytes,files}_{total,free,avail} files, it may appear
3962 as zero or be out of date.
3965 Frequency : systems with MD RAID1 external journal devices
3967 Description: lconf's call to blkid is confused by RAID1 journal devices
3968 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3969 of returning all block devices that match the journal UUID.
3972 Frequency : always, for aggregate stripe size over 4GB
3974 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3975 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3976 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3979 Frequency : always if "lfs find" used on a local file/directory
3981 Description: "lfs find" segfaults if used on a local file/directory
3982 Details : The case where a directory component was not specified wasn't
3983 handled correctly. Handle this properly.
3986 Frequency : always on ppc64
3988 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3989 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3990 perform endianness conversion.
3993 Frequency : rarely (truncate to non-zero file size after write under load)
3994 Bugzilla : 10730, 10687
3995 Description: Files padded with zeros to next 4K multiple
3996 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3997 it is possible that files that are truncated to a non-zero size
3998 immediately after being written are filled with zero bytes beyond
3999 the truncated size. No file data is lost.
4001 Severity : enhancement
4003 Description: Allow recovery/failover for liblustre clients.
4004 Details : liblustre clients were unaware of failover configurations until
4007 Severity : enhancement
4009 Description: user file locks should fail when not mounting with flock option
4010 Details : Set up an error-returning stub in ll_file_operations.lock field
4011 to prevent incorrect behaviour when client is mounted without
4012 flock option. Also, set up properly f_op->flock field for
4016 Frequency : always on ia64
4018 Description: "lfs df" loops on printing out MDS statfs information
4019 Details : The obd_ioctl_data was not initialized and in some systems
4020 this caused a failure during the ioctl that did not return
4021 an error. Initialize the struct and return an error on failure.
4024 Frequency : SLES 9 only
4026 Description: Error of copying files with lustre special EAs as root
4027 Details : Client side always return success for setxattr call for lustre
4028 special xattr (currently only "trusted.lov").
4031 Frequency : rarely on clusters with both ia64+i386 clients
4033 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4034 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4035 clients are doing concurrent writes to the same file it is
4036 possible that the ia64 clients may LASSERT because the OST
4037 extent locks are not PAGE_SIZE aligned. Ensure that grown
4038 locks are always aligned on the request boundary.
4041 Frequency : specific use, occasional
4043 Description: Overwriting in use executable truncates on-disk binary image
4044 Details : If one node attempts to overwrite an executable in use by
4045 another node, we now correctly return ETXTBSY instead of
4046 truncating the file.
4048 Severity : enhancement
4050 Description: Async OSC create to avoid the blocking unnecessarily.
4051 Details : If an OST has no remaining object, system will block on the creating
4052 when need to create a new object on this OST. Now, ways use
4053 pre-created objects when available, instead of blocking on an
4054 empty osc while others are not empty. If we must block, we block
4055 for the shortest possible period of time.
4060 Description: chmod on Lustre root is propagated to other clients
4061 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4062 invalid by the follow_mount time.
4065 Frequency : liblustre clients only
4067 Description: Race in 'instant cancel' lock handling could lead to such locks
4068 never to be granted in case of SMP MDS
4069 Details : Do not destroy not yet granted but cbpending locks in
4073 Frequency : replay/resend of open
4075 Description: non null lock assetion failure in mds_intent_policy
4076 Details : Trying to replay/resend lockless open requests resulted in
4077 mds_open() returning 0 with no lock. Now it sets a flag if
4078 a lock is going to be returned.
4080 Severity : enhancement
4082 Description: Checksum enhancements
4083 Details : New checksum enhancements allow for resending RPCs that failed
4086 Severity : enhancement
4088 Description: Tunables on number of dirty pages in cacche
4089 Details : Allow to set limit on number of dirty pages cached.
4094 Description: client crash on unmount - lock still has references
4095 Details : In some error handling cases it was possible to leak a lock
4096 reference on a client while accessing a file. This was not
4097 harmful to the client during operation, but would cause the
4098 client to crash when the filesystem is unmounted.
4101 Frequency : specific case, rare
4103 Description: ETXTBSY on mds though file not in use by client
4104 Details : ETXTBSY is no longer incorrectly returned when attempting to
4105 chmod or chown a directory that the user previously tried to
4106 execute or a currently-executing binary.
4109 Frequency : extremely rare except on liblustre-based clients
4111 Description: Lustre space not freed when files are deleted
4112 Details : Clean up open-unlinked files after client eviction. Previously
4113 the unlink was skipped and the files remained as orphans.
4118 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4119 Details : In some cases it was possible to send a blocking callback to
4120 a client doing a glimpse, even though that client didn't get
4121 a lock granted. When the glimpse lock is cancelled on the OST
4122 the freed lock is left on the waiting list and corrupted the list.
4125 Frequency : all core dumps
4127 Description: Broke core dumps to lustre
4128 Details : Negative dentry may be unhashed if parent does not have UPDATE
4129 lock, but some callers, e.g. do_coredump, expect dentry to be
4130 hashed after successful create, hash it in ll_create_it.
4132 ------------------------------------------------------------------------------
4134 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4136 * Support for kernels:
4137 2.6.9-42.0.2.EL (RHEL 4)
4138 2.6.5-7.276 (SLES 9)
4139 2.4.21-40.EL (RHEL 3)
4140 2.6.12.6 vanilla (kernel.org)
4144 Frequency : always on RHEL 3
4146 Description: Number of open files grows over time
4147 Details : The number of open files grows over time, whether or not
4148 Lustre is started. This was due to a filp leak introduced
4149 by one of our kernel patches.
4151 ------------------------------------------------------------------------------
4153 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4155 * Support for kernels:
4156 2.6.9-42.EL (RHEL 4)
4157 2.6.5-7.276 (SLES 9)
4158 2.4.21-40.EL (RHEL 3)
4159 2.6.12.6 vanilla (kernel.org)
4164 Bugzilla : 5719, 9635, 9792, 9684
4165 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4166 Details : If a server is under heavy load and cannot reply to new
4167 connection requests before the client resends the (re)connect,
4168 the connection handling code can behave badly if two service
4169 threads are concurrently handing separate (re)connections from
4170 the same client. Add better locking to the connection handling
4171 code, and ensure that only a single connection will be processed
4172 for a given client UUID, even if the lock is dropped.
4174 Severity : enhancement
4176 Description: add TCP zero-copy support to kernel
4177 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4178 sends if the hardware supports scatter-gather and checksumming.
4179 This allows socklnd to do client-write and server-read more
4180 efficiently and reduce CPU utilization from skbuf copying.
4183 Frequency : only if NFS exporting from client
4185 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4186 Details : If setting attributes on a file created under NFS that had
4187 never been opened it would be possible to oops the client
4188 if the file had no objects.
4191 Frequency : always for liblustre
4193 Description: liblustre client does MDS+OSTs setattr RPC for each write
4194 Details : When doing a write from a liblustre client, the client
4195 incorrectly issued an RPC to the MDS and each OST the file was
4196 striped over in order to update the timestamps. When writing
4197 with small chunks and many clients this could overwhelm the MDS
4198 with RPCs. In all cases it would slow down the write because
4199 these RPCs are unnecessary.
4201 Severity : enhancement
4203 Description: allow number of MDS service threads to be changed at module load
4204 Details : It is now possible to change the number of MDS service threads
4205 running. Adding "options mds mds_num_threads={N}" to the MDS's
4206 /etc/modprobe.conf will set the number of threads for the next
4207 time Lustre is restarted (assuming the "mds" module is also
4208 reloaded at that time). The default number of threads will
4209 stay the same, 32 for most systems.
4214 Description: OST crash if filesystem is unformatted or corrupt
4215 Details : If an OST is started on a device that has never been formatted
4216 or if the filesystem is corrupt and cannot even mount then the
4217 error handling cleanup routines would dereference a NULL pointer.
4222 Description: NULL pointer deref in llap_from_page.
4223 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4224 incorrect) mapping if the page was truncated/reclaimed while it was
4225 searched for. Check for this condition and skip such pages when
4226 doing readahead. Introduce extra check to llap_from_page() to
4227 verify page->mapping->host is non-NULL (so page is not anonymous).
4230 Frequency : Sometimes when using sys_sendfile
4232 Description: "page not covered by a lock" warnings from ll_readpage
4233 Details : sendfile called ll_readpage without right page locks present.
4234 Now we introduced ll_file_sendfile that does necessary locking
4235 around call to generic_file_sendfile() much like we do in
4239 Frequency : with certain MDS communication failures at client mount time
4241 Description: NULL pointer deref after failed client mount
4242 Details : a client connection request may delayed by the network layer
4243 and not be sent until after the PTLRPC layer has timed out the
4244 request. If the client fails the mount immediately it will try
4245 to clean up before the network times out the request. Add a
4246 reference from the request import to the obd device and delay
4247 the cleanup until the network drops the request.
4250 Frequency : occasionally during client (re)connect
4252 Description: assertion failure during client (re)connect
4253 Details : processing a client connection request may be delayed by the
4254 client or server longer than the client connect timeout. This
4255 causes the client to resend the connection request. If the
4256 original connection request is replied in this interval, the
4257 client may trip an assertion failure in ptlrpc_connect_interpret()
4258 which thought it would be the only running connect process.
4261 Frequency : only with obd_echo servers and clients that are rebooted
4263 Description: kernel BUG accessing uninitialized data structure
4264 Details : When running an obd_echo server it did not start the ping_evictor
4265 thread, and when a client was evicted an uninitialized data
4266 structure was accessed. Start the ping_evictor in the RPC
4267 service startup instead of the OBD startup.
4269 Severity : enhancement
4270 Bugzilla : 10193 (patchless)
4271 Description: Remove dependency on various unexported kernel interfaces.
4272 Details : No longer need reparent_to_init, exit_mm, exit_files,
4273 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4276 Frequency : rare (only users of deprecated and unsupported LDAP config)
4278 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4279 Details : LDAP apparently contains 'lustreName' attributes instead of
4280 'name'. A simple remapping of the name is sufficient.
4283 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4285 Description: waiting_locks_callback trips kernel BUG if client is evicted
4286 Details : Running with the dump_on_timeout debug flag turned on makes
4287 it possible that the waiting_locks_callback() can try to dump
4288 the Lustre kernel debug logs from an interrupt handler. Defer
4289 this log dumping to the expired_lock_main() thread.
4291 Severity : enhancement
4293 Description: Support NFS exporting on 2.6 kernels.
4294 Details : Implement non-rawops metadata methods for NFS server to use without
4295 changing NFS server code.
4298 Frequency : very rare (synthetic metadata workload only)
4300 Description: two racing renames might cause an MDS thread to deadlock
4301 Details : Running the "racer" program may cause one MDS thread to rename
4302 a file from being the source of a rename to being the target of
4303 a rename at exactly the same time that another thread is doing
4304 so, and the second thread has already enqueued these locks after
4305 doing a lookup of the target and is trying to relock them in
4306 order. Ensure that we don't try to re-lock the same resource.
4309 Frequency : only very large systems with liblustre clients
4311 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4312 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4313 all OSTs in parallel. This allows the eviction of stale liblustre
4314 clients to proceed much faster than if they were done in series,
4315 and also offers similar improvements for other set_info RPCs.
4320 Description: excessive CPU usage during initial read phase on client
4321 Details : During the initial read phase on a client, it would agressively
4322 retry readahead on the file, consuming too much CPU and impacting
4323 performance (since 1.4.5.8). Improve the readahead algorithm
4324 to avoid this, and also improve some other common cases (read
4325 of small files in particular, where "small" is files smaller than
4326 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4331 Description: MDS crash when receiving packet with unknown intent.
4332 Details : Do not LBUG in unknown intent case, just return -EFAULT
4334 Severity : enhancement
4335 Bugzilla : 9293, 9385
4336 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4337 Details : Do not serialize getattr (non-intent version) and statfs.
4340 Frequency : occasional, when OST network is overloaded/intermittent
4342 Description: client evicted by OST after bulk IO timeout
4343 Details : If a client sends a bulk IO request (read or write) the OST
4344 may evict the client if it is unresposive to its data GET/PUT
4345 request. This is incorrect if the network is overloaded (takes
4346 too long to transfer the RPC data) or dropped the OST GET/PUT
4347 request. There is no need to evict the client at all, since
4348 the pinger and/or lock callbacks will handle this, and the
4349 client can restart the bulk request.
4352 Frequency : Always when mmapping file with no objects
4354 Description: client crashes when mmapping file with no objects
4355 Details : Check that we actually have objects in a file before doing any
4356 operations on objects in ll_vm_open, ll_vm_close and
4362 Description: Request leak when working with deleted CWD
4363 Details : Introduce advanced request refcount tracking for requests
4364 referenced from lustre intent.
4366 Severity : Enhancement
4368 Description: Cache open file handles on client.
4369 Details : MDS now will return special lock along with openhandle, if
4370 requested and client is allowed to hold openhandle, even if unused,
4371 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4372 closing files for every read/write openration.
4374 Severity : Enhancement
4376 Description: Cache open negative dentries on client when possible.
4377 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4378 negative dentries on lock revocation.
4383 Description: Remounting a client read-only wasn't possible with a zconf mount
4384 Details : It wasn't possible to remount a client read-only with llmount.
4386 Severity : enhancement
4387 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4388 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4389 MPICH is included to add Lustre-specific ADIO interfaces.
4390 This is based closely on the UFS ADIO layer and only differs
4391 in file creation, in order to allow the OST striping to be set.
4392 This is user-contributed code and not supported by CFS.
4397 Description: extended inode attributes (immutable, append-only) work improperly
4398 when 2.4 and 2.6 kernels are used on client/server or vice versa
4399 Details : Introduce kernel-independent values for these flags.
4401 Severity : enhancement
4404 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4405 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4406 granularity. Now, Fractional MB granularity is be supported,
4407 this is very useful for low memory system.
4409 Severity : enhancement
4411 Description: Getattr by fid
4412 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4413 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4414 have locks and avoids getattr rpc requests for it on every stat.
4417 Frequency : Always, for filesystems larger than 2TB
4419 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4420 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4421 filesystems larger than 2TB to be mounted with mballoc enabled.
4424 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4426 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4427 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4428 filesystems larger than 2TB is mounted then the kernel will
4429 silently corrupt the start of the filesystem. CONFIG_LBD is
4430 enabled for all CFS-supported kernels, but the possibility of
4431 this happening with a modified kernel config exists.
4433 Severity : enhancement
4435 Description: add client O_DIRECT support for 2.6 kernels
4436 Details : It is now possible to do O_DIRECT reads and writes to files
4437 in the Lustre client mountpoint on 2.6 kernel clients.
4439 Severity : enhancement
4441 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4442 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4443 parallel, not waiting for response from every OST before sending
4444 a rpc to the next OST.
4449 Description: setattr vs write race when updating file timestamps
4450 Details : Client processes that update a file timestamp into the past
4451 right after writing to the file (e.g. tar) it is possible that
4452 the updated file modification time can be reset to the current
4453 time due to a race between processing the setattr and write RPC.
4455 Severity : enhancement
4457 Description: Bring 'lfs find' closer in line with regular Linux find.
4458 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4459 -print0 options and obtains all the needed info through the lustre
4462 Severity : enhancement
4464 Description: support up to 1024 configured devices on one node
4465 Details : change obd_dev array from statically allocated to dynamically
4466 allocated structs as they are first used to reduce memory usage
4471 Description: Flush dirty partially truncated pages during truncate
4472 Details : Immediatelly flush partially truncated pages in filter_setattr,
4473 this way we completely avoid having any pages in page cache on OST
4474 and can retire ugly workarounds during writes to flush such pages.
4479 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4480 Details : protect inode from truncation within vfs_unlink() context
4481 just take a reference before calling vfs_unlink() and release it
4482 when parent's i_sem is free.
4487 Description: last_id value checked outside lock on OST caused LASSERT failure
4488 Details : If there were multiple MDS->OST object precreate requests in
4489 flight, it was possible that the OST's last object id was checked
4490 outside a lock and incorrectly tripped an assertion. Move checks
4491 inside locks, and discard old precreate requests.
4494 Frequency : always, if extents are used on OSTs
4496 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4497 Details : OSTs using the extents format would not zero the high 16 bits of
4498 the index physical block number. This is not a problem for any
4499 OST filesystems smaller than 16TB, and no kernels support ext3
4500 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4501 new/modified files) and can be fixed for existing filesystems
4502 with e2fsprogs-1.39-cfs1.
4507 Description: import connection selection may be incorrect if timer wraps
4508 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4509 import connections to be ignored if the 32-bit jiffies counter
4510 wraps. Use a 64-bit jiffies counter.
4513 Frequency : very large clusters immediately after boot
4515 Description: LNET request buffers exhausted under heavy short-term load
4516 Details : If a large number of client requests are generated on a service
4517 that has previously never seen so many requests it is possible
4518 that the request buffer growth cannot keep up with the spike in
4519 demand. Instead of dropping incoming requests, they are held in
4520 the LND until the RPC service can accept more requests.
4523 Frequency : Sometimes during replay
4525 Description: Assertion failure in ll_local_open after replay.
4526 Details : If replay happened on an open request reply before we were able
4527 to set replay handler, reply will become not swabbed tripping the
4528 assertion in ll_local_open. Now we set the handler right after
4529 recognising of open request
4532 Frequency : very rare
4534 Description: kernel reports "badness in vsnprintf"
4535 Details : Reading from the "recovery_status" /proc file in small chunks
4536 may cause a negative length in lprocfs_obd_rd_recovery_status()
4537 call to vsnprintf() (which is otherwise harmless). Exit early
4538 if there is no more space in the output buffer.
4540 Severity : enhancement
4542 Description: clear OBD RPC statistics by writing to them
4543 Details : It is now possible to clear the OBD RPC statistics by writing
4544 to the "stats" file.
4549 Description: Client mtime is not the same on different clients after utimes
4550 Details : In some cases, the client was using the utimes() syscall on
4551 a file cached on another node. The clients now validate the
4552 ctime from the MDS + OSTs to determine which one is right.
4557 Description: Inability to activate failout mode
4558 Details : lconf script incorrectly assumed that in pythong string's numeric
4559 value is used in comparisons.
4562 Frequency : always with multiple stripes per file
4564 Description: Inefficient object allocation for mutli-stripe files
4565 Details : When selecting which OSTs to stripe files over, for files with
4566 a stripe count that divides evenly into the number of OSTs,
4567 the MDS is always picking the same starting OST for each file.
4568 Return the OST selection heuristic to the original design.
4573 Description: mount failures may take full timeout to return an error
4574 Details : Under some heavy load conditions it is possible that a
4575 failed mount can wait for the full obd_timeout interval,
4576 possibly several minutes, before reporting an error.
4577 Instead return an error as soon as the status is known.
4579 Frequency : quota enabled and large files being deleted
4581 Description: releasing more than 4GB of quota at once hangs OST
4582 Details : If a user deletes more than 4GB of files on a single OST it
4583 will cause the OST to spin in an infinite loop. Release
4584 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4589 Description: statfs data retrieved from /proc may be stale or zero
4590 Details : When reading per-device statfs data from /proc, in the
4591 {kbytes,files}_{total,free,avail} files, it may appear
4592 as zero or be out of date.
4595 Frequency : systems with MD RAID1 external journal devices
4597 Description: lconf's call to blkid is confused by RAID1 journal devices
4598 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4599 of returning all block devices that match the journal UUID.
4602 Frequency : always, for aggregate stripe size over 4GB
4604 Description: assertion fails when trying to use 4GB stripe size
4605 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4606 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4609 Frequency : always on ppc64
4611 Description: the first write on an ext3 filesystem with mballoc got stuck
4612 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4613 perform endianness conversion.
4615 ------------------------------------------------------------------------------
4617 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4619 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4620 INTEROPERATE with older versions automatically. Please read the
4621 user documentation before upgrading any part of a live system.
4622 * WARNING: Lustre networking configuration changes are required with
4623 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4626 * Support for kernels:
4627 2.6.9-22.0.2.EL (RHEL 4)
4628 2.6.5-7.244 (SLES 9)
4629 2.6.12.6 vanilla (kernel.org)
4632 Severity : enhancement
4633 Bugzilla : 7981/8208
4634 Description: Introduced Lustre Networking (LNET)
4635 Details : LNET is new networking infrastructure for Lustre, it includes
4636 a reorganized network configuration mode (see the user
4637 documentation for full details) as well as support for routing
4638 between different network fabrics. Lustre Networking Devices
4639 (LNDs) for the supported network fabrics have also been
4640 created for this new infrastructure.
4642 Severity : enhancement
4643 Description: Introduced Access control lists
4644 Details : clients can set ACLs on files and directories in order to have
4645 more fine-grained permissions than the standard Unix UGO+RWX.
4646 The MDS must be started with the "-o acl" mount option.
4648 Severity : enhancement
4649 Description: Introduced filesystem quotas
4650 Details : Administrators may now establish per-user quotas on the
4653 Severity : enhancement
4655 Description: Configuration change for the XT3
4656 The PTLLND is now used to run Lustre over Portals on the XT3
4657 The configure option(s) --with-cray-portals are no longer used.
4658 Rather --with-portals=<path-to-portals-includes> is used to
4659 enable building on the XT3. In addition to enable XT3 specific
4660 features the option --enable-cray-xt3 must be used.
4665 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4666 Details : A race between allocating a new llcd and re-getting the llcd_lock
4667 allowed another thread to grab newly-allocated llcd.
4669 Severity : enhancement
4671 Description: 2.6 OST async journal commit and locking fix to improve performance
4672 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4673 2.6 kernels are now async as they already were in 2.4 kernels so
4674 that they can commit concurrently with the network bulk transfer.
4675 For block-allocated files the filter allocation semaphore is held
4676 to avoid filesystem fragmentation during allocation. BKL lock
4677 removed for 2.6 xattr operations where it is no longer needed.
4682 Description: lconf incorrectly determined whether two IP networks could talk
4683 Details : In some more complicated routing and multiple-network
4684 configurations, lconf will avoid trying to make a network
4685 connection to a disjoint part of the IP space. It was doing the
4686 math incorrectly for one set of cases.
4691 Description: Fix for potential infinite loop processing records in an llog.
4692 Details : If an llog record is corrupted/zeroed, it is possible to loop
4693 forever in llog_process(). Validate the llog record length
4694 and skip the remainder of the block on error.
4697 Frequency : occasional (liblustre only)
4699 Description: liblustre could not open files whose last component is a symlink
4700 Details : sysio_path_walk() would incorrectly pass the open intent to
4701 intermediate path components.
4704 Frequency : rare (liblustre only with non-standard tuning)
4705 Bugzilla : 7201 (7350)
4706 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4707 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4708 after it has been released, only do this for referenced locks
4710 Severity : enhancement
4712 Description: specifying an (invalid) directory default stripe_size of -1
4713 would reset the directory default striping
4714 Details : stripe_size -1 was used internally to signal directory stripe
4715 removal, now use "all default" to signal dir stripe removal
4716 as a directory striping of "all default" is not useful
4719 Frequency : common for large clusters running liblustre clients
4721 Description: doing an ls when liblustre clients are running is slow
4722 Details : sending a glimpse AST to a liblustre client waits for every AST
4723 to time out, as liblustre clients will not respond. Since they
4724 cannot cache data we refresh the OST lock LVB from disk instead.
4726 Severity : enhancement
4728 Description: doing an ls at the same time as file IO can be slow
4729 Details : enqueue and other "small" requests can be blocked behind many
4730 large IO requests. Create a new OST IO portal for non-IO
4731 requests so they can be processed faster.
4734 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4736 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4737 created file with mode 000
4738 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4739 allow an ftruncate() as a non-root user to a file with mode 000.
4740 Lustre now respects this flag to disable mode checks when
4741 truncating a file owned by the user
4744 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4747 Description: Revoking locks from clients that went dead or catatonic might take
4749 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4750 cancellation of such locks instant on servers without waiting for
4751 any reply from clients. Clients drops these locks when cancel
4752 notification from server is received without replying.
4755 Frequency : liblustre-only, when liblustre client dies or becomes busy
4757 Description: Doing ls on Linux clients can take a long time with active
4759 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4760 granting such locks to it in the first place if possible. Locks
4761 are taken by proxy on the OST during the read or write and
4762 dropped immediately afterward. Add connect flags handling, do
4763 not grant locks to liblustre clients for glimpse ASTs.
4765 Severity : enhancement
4767 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4768 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4769 sequential reads and behaves badly with applications doing random
4770 reads. Improve it by reading ahead at least read region, and
4771 avoiding excessive large RPC for small reads.
4773 Severity : enhancement
4775 Description: Creating more than 1000 files for a single job may cause a load
4776 imbalance on the OSTs if there are also a large number of OSTs.
4777 Details : qos_prep_create() uses an OST index reseed value that is an
4778 even multiple of the number of available OSTs so that if the
4779 reseed happens in the middle of the object allocation it will
4780 still utilize the OSTs as uniformly as possible.
4785 Description: OST or MDS may oops in ping_evictor_main()
4786 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4787 but doesn't restart at beginning of obd_exports_timed list
4790 Severity : enhancement
4792 Description: improve by-nid export eviction on the MDS and OST
4793 Details : allow multiple exports with the same NID to be evicted at one
4794 time without re-searching the exports list.
4797 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4799 Description: MDS may oops in groups_free()
4800 Details : in rare race conditions a newly allocated group_info struct is
4801 freed again, and this can be NULL. The 2.4 compatibility code
4802 for groups_free() checked for a NULL pointer, but 2.6 did not.
4805 Frequency : common for liblustre clients doing little filesystem IO
4806 Bugzilla : 9352, 7313
4807 Description: server may evict liblustre clients accessing contended locks
4808 Details : if a client is granted a lock or receives a completion AST
4809 with a blocking AST already set it would not reply to the AST
4810 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4813 Frequency : lfs setstripe, only systems with more than 160 OSTs
4815 Description: unable to set striping with a starting offset beyond OST 160
4816 Details : llapi_create_file() incorrectly limited the starting stripe
4817 index to the maximum single-file stripe count.
4820 Frequency : LDAP users only
4822 Description: lconf did not handle in-kernel recovery with LDAP properly
4823 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4825 Severity : enhancement
4827 Description: bind OST threads to NUMA nodes to improve performance
4828 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4829 node and do their allocations there to localize memory access
4831 Severity : enhancement
4833 Description: llmount can determine client NID directly from Myrinet (GM)
4834 Details : the client NID code from gmnalnid was moved directly into
4835 llmount, removing the need to use this or specifying the
4836 client NID explicitly when mounting GM clients with zeroconf
4839 Frequency : if client is started with down MDS
4841 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4842 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4843 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4844 ptlrpc_import_delayed_req() because we didn't check if the
4845 request was interrupted, and we also didn't break out of the
4846 event loop if there was no timeout
4851 Description: data loss during non-page-aligned writes to a single file from
4852 both multiple nodes and multiple threads on one node at same time
4853 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4854 inconsistency led to false short-reads, that were cached and later
4855 used by ->prepare_write() to fill in partially written page,
4856 leading to data loss.
4859 Frequency : always, if lconf --abort_recovery used
4861 Description: lconf --abort_recovery fails with 'Operation not supported'
4862 Details : lconf was attempting to abort recovery on the MDT device and not
4865 Severity : enhancement
4867 Description: remove cleanup logs
4868 Details : replace lconf-generated cleanup logs with lustre internal
4869 cleanup routines. Eliminates the need for client-cleanup and
4872 Severity : enhancement
4874 Description: add support for EAs (user and system) on lustre filesystems
4875 Details : it is now possible to store extended attributes in the Lustre
4876 client filesystem, and with the user_xattr mount option it
4877 is possible to allow users to store EAs on their files also
4879 Severity : enhancement
4881 Description: Add possibility (config option) to show minimal available OST free
4883 Details : When compiled with --enable-mindf configure option, statfs(2)
4884 (and so, df) will return least minimal free space available from
4885 all OSTs as amount of free space on FS, instead of summary of
4886 free spaces of all OSTs.
4888 Severity : enhancement
4890 Description: do not expand extent locks acquired on OST-side
4891 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4892 by server: they are not cached anyway.
4895 Frequency : when mmap is used/binaries executed from Lustre
4897 Description: Unmmap pages before throwing them away from read cache.
4898 Details : llap_shrink cache now attempts to unmap pages before discarding
4899 them (if unmapping failed - do not discard). SLES9 kernel has
4900 extra checks that trigger if this unmapping is not done first.
4905 Description: lconf didn't resolve symlinks before checking to see whether a
4906 given mountpoint was already in use
4909 Frequency : when migrating failover services
4910 Bugzilla : 6395, 9514
4911 Description: When migrating a subset of services from a node (e.g. failback
4912 from a failover service node) the remaining services would
4913 time out and evict clients.
4914 Details : lconf --force (implied by --failover) sets the global obd_timeout
4915 to 5 seconds in order to quickly disconnect, but this caused
4916 other RPCs to time out too quickly. Do not change the global
4917 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4919 Severity : enhancement
4920 Frequency : if MDS is started with down OST
4921 Bugzilla : 9439,5706
4922 Description: Allow startup/shutdown of an MDS without depending on the
4923 availability of the OSTs.
4924 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4925 Add appropriate locking and lov-osc refcounts for safe
4926 cleaning. Add osc abort_inflight calls in case the
4927 synchronize never started.
4930 Frequency : occasional (Cray XT3 only)
4932 Description: root not authorized to access files in CRAY_PORTALS environment
4933 Details : The client process capabilities were not honoured on the MDS in
4934 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4935 been accessed by an authorized user then root was able to access
4936 the file on the local client also. The root user capabilities
4937 are now allowed on the MDS, as this environment has secure UID.
4940 Frequency : occasional
4942 Description: ldiskfs "too long searching" message happens too often
4943 Details : A debugging message (otherwise harmless) prints too often on
4944 the OST console. This has been reduced to only happen when
4945 there are fragmentation problems on the filesystem.
4950 Description: Division by zero in statfs when all OSCs are inactive
4951 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4952 lov_statfs divides by value returned by lov_get_stripecnt().
4956 Bugzilla : 9489, 3273
4957 Description: First write from each client to each OST was only 4kB in size,
4958 to initialize client writeback cache, which caused sub-optimal
4959 RPCs and poor layout on disk for the first written file.
4960 Details : Clients now request an initial cache grant at (re)connect time
4961 and so that they can start streaming writes to the cache right
4962 away and always do full-sized RPCs if there is enough data.
4963 If the OST is rebooted the client also re-establishes its grant
4964 so that client cached writes will be honoured under the grant.
4969 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4970 Details : Now I/O RPCs go to different portal number and (presumably) fast
4971 lock requests (and glimses) and other RPCs get their own service
4972 threads pool that should be able to service those RPCs
4975 Severity : enhancement
4977 Description: Ability to exchange lustre version between client and servers and
4978 issue warnings at client side if client is too old. Also for
4979 liblustre clients there is ability to refuse connection of too old
4981 Details : New 'version' field is added to connect data structure that is
4982 filled with version info. That info is later checked by server and
4986 Frequency : rare, liblustre only.
4987 Bugzilla : 9296, 9581
4988 Description: Two simultaneous writes from liblustre at offset within same page
4989 might proceed at the same time overwriting eachother with stale
4991 Details : I/O lock withing llu_file_prwv was released too early, before data
4992 actually was hitting the wire. Extended lock-holding time until
4993 server acknowledges receiving data.
4996 Frequency : extremely rare. Never observed in practice.
4998 Description: avoid generating lustre_handle cookie of 0.
4999 Details : class_handle_hash() generates handle cookies by incrementing
5000 global counter, and can hit 0 occasionaly (this is unlikely, but
5001 not impossible, because initial value of cookie counter is
5002 selected randonly). Value of 0 is used as a sentinel meaning
5003 "unassigned handle" --- avoid it. Also coalesce two critical
5004 sections in this function into one.
5006 Severity : enhancement
5008 Description: allow liblustre clients to delegate truncate locking to OST
5009 Details : To avoid overhead of locking, liblustre client instructs OST to
5010 take extent lock in ost_punch() on client's behalf. New connection
5011 flag is added to handle backward compatibility.
5013 Severity : enhancement
5014 Bugzilla : 4928, 7341, 9758
5015 Description: allow number of OST service threads to be specified
5016 Details : a module parameter allows the number of OST service threads
5017 to be specified via "options ost ost_num_threads={N}" in the
5018 OSS's /etc/modules.conf or /etc/modprobe.conf.
5022 Bugzilla : 6146, 9635, 9895
5023 Description: servers crash with bad pointer in target_handle_connect()
5024 Details : In rare cases when a client is reconnecting it was possible that
5025 the connection request was the last reference for that export.
5026 We would temporarily drop the export reference and get a new
5027 one, but this may have been the last reference and the export
5028 was just destroyed. Get new reference before dropping old one.
5030 Severity : enhancement
5031 Frequency : if client is started with failover MDS
5033 Description: Allow multiple MDS hostnames in the mount command
5034 Details : Try to read the configuration from all specified MDS
5035 hostnames during a client mount in case the "primary"
5038 Severity : enhancement
5040 Description: Stop sending data to evicted clients as soon as possible.
5041 Details : Check if the client we are about to send or are sending data to
5042 was evicted already. (Check is done every second of waiting,
5043 for which l_wait_event interface was extended to allow checking
5044 of exit condition at specified intervals).
5047 Frequency : rare, normally only when NFS exporting is done from client
5049 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5051 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5054 Frequency : always, if srand() is called before liblustre initialization
5056 Description: Liblustre uses system PRNG disturbing its usage by user application
5057 Details : Introduce internal to lustre fast and high-quality PRNG for
5058 lustre usage and make liblustre and some other places in generic
5059 lustre code to use it.
5061 Severity : enhancement
5062 Bugzilla : 9477, 9557, 9870
5063 Description: Verify that the MDS configuration logs are updated when xml is
5064 Details : Check if the .xml configuration logs are newer than the config
5065 logs stored on the MDS and report an error if this is the case.
5066 Request --write-conf, or allow starting with --old_conf.
5068 Severity : enhancement
5070 Description: Handle symlinks in the path when checking if Lustre is mounted.
5071 Details : Resolve intermediate symlinks when checking if a client has
5072 mounted a filesystem to avoid duplicate client mounts.
5077 Description: lconf can hit an error exception but still return success.
5078 Details : The lconf command catches the Command error exception at the top
5079 level script context and will exit with the associated exit
5080 status, but doesn't ensure that this exit status is non-zero.
5085 Description: failure of ptlrpc thread startup can cause oops
5086 Details : Starting a ptlrpc service thread can fail if there are a large
5087 number of threads or the server memory is very fragmented.
5088 Handle this without oopsing.
5091 Frequency : always, only if liblustre and non-default acceptor port was used
5093 Description: liblustre cannot connect to servers with non-default acceptor port
5094 Details : tcpnal_set_default_params() was not called and was therefore
5095 ignoring the environment varaible TCPNAL_PORT, as well as other
5096 TCPNAL_ environment variables
5101 Description: two objects could be created on the same OST for a single file
5102 Details : If an OST is down, in some cases it was possible to create two
5103 objects on a single OST for a single file. No problems other
5104 than potential performance impact and spurious error messages.
5108 Bugzilla : 5681, 9562
5109 Description: Client may oops in ll_unhash_aliases
5110 Details : Client dcache may become inconsistent in race condition.
5111 In some cases "getcwd" can fail if the current directory is
5117 Description: Inode refcounting problems in NFS export code
5118 Details : link_raw functions used to call d_instantiate without obtaining
5119 extra inode reference first.
5123 Bugzilla : 9942, 9903
5124 Description: Referencing freed requests leading to crash, memleaks with NFS.
5125 Details : We used to require that call to ll_revalidate_it was always
5126 followed by ll_lookup_it. Also with revalidate_special() it is
5127 possible to call ll_revalidate_it() twice for the same dentry
5128 even if first occurence returned success. This fix changes semantic
5129 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5130 reference on a request referred from the intent.
5131 ll_intent_release() then releases such a request.
5134 Frequency : rare, normally benchmark loads only
5136 Description: unlinked inodes were kept in memory on the client
5137 Details : If a client is repeatedly creating and unlinking files it
5138 can accumulate a lot of stale inodes in the inode slab cache.
5139 If there is no other client load running this can cause the
5140 client node to run out of memory. Instead flush old inodes
5141 from client cache that have the same inode number as a new inode.
5144 Frequency : SLES9 2.6.5 kernel and long filenames only
5145 Bugzilla : 9969, 10379
5146 Description: utime reports stale NFS file handle
5147 Details : SLES9 uses out-of-dentry names in some cases, which confused
5148 the lustre dentry revalidation. Change it to always use the
5152 Frequency : rare, unless heavy write-truncate concurrency is continuous
5153 Bugzilla : 4180, 6984, 7171, 9963, 9331
5154 Description: OST becomes very slow and/or deadlocked during object unlink
5155 Details : filter_destroy() was holding onto the parent directory lock
5156 while truncating+unlinking objects. For very large objects this
5157 may block other threads for a long time and slow overall OST
5158 responsiveness. It may also be possible to get a lock ordering
5159 deadlock in this case, or run out of journal credits because of
5160 the combined truncate+unlink. Solution is to do object truncate
5161 first in one transaction without parent lock, and then do the
5162 final unlink in a new transaction with the parent lock. This
5163 reduces the lock hold time dramatically.
5166 Frequency : rare, 2.4 kernels only
5168 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5169 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5170 before unmounting their filesystem in order to avoid deadlock.
5171 The kernel_locked() function in 2.4 kernels only checks whether
5172 the kernel lock is held, not whether it is this process that is
5173 holding it as 2.6 kernels do.
5178 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5179 Details : The client ptlrpc code may be trying to reconnect to a down
5180 server before a previous connection attempt has timed out.
5181 Increase the reconnect interval to be longer than the connection
5182 timeout interval to avoid sending duplicate connections to
5186 Frequency : echo_client brw_test command
5188 Description: fix echo_client to work with OST preallocated code
5189 Details : OST preallocation code (5137) didn't take echo_client IO path
5190 into account: echo_client calls filter methods outside of any
5191 OST thread and, hence, there is no per-thread preallocated
5192 pages and buffers to use. Solution: hijack pga pages for IO. As
5193 a byproduct, this avoids unnecessary data copying.
5197 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5198 Description: Client can oops in mdc_commit_close() after open replay
5199 Details : It was possible for the MDS to return an open request with no
5200 transaction number in mds_finish_transno() if the client was
5201 evicted, but without actually returning an error. Clients
5202 would later try to replay that open and may trip an assertion
5203 Simplify the client close codepath, and always return an error
5204 from the MDS in case the open is not successful.
5207 Frequency : rare, 2.6 OSTs only
5209 Description: OST may deadlock under high load on fragmented files
5210 Details : If there was a heavy load and highly-fragmented OST filesystems
5211 it was possible to have all the OST threads deadlock waiting on
5212 allocation of biovecs, because the biovecs were not released
5213 until the entire RPC IO was completed. Instead, release biovecs
5214 as soon as they are complete to ensure forward IO progress.
5216 Severity : enhancement
5218 Description: Support for specifying external journal device at mount
5219 Details : If an OST or MDS device is formatted with an external journal
5220 device, this device major/minor is stored in the ext3 superblock
5221 and may not be valid for failover. Allow detecting and
5222 specifying the external journal at mount time.
5227 Description: Mounting an MDS with pending unlinked files may cause oops
5228 Details : target_finish_recovery() calls mds_postrecov() which returned
5229 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5230 considers this an error and immediately begins cleaning up the
5231 lov, just after starting the mds_lov process
5233 Severity : enhancement
5235 Description: Implement 'lfs df' to report actual free space on per-OST basis
5236 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5237 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5238 usage of objects; '-h' to report in human readable format.
5240 ------------------------------------------------------------------------------
5242 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5249 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5250 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5251 at startup during orphan unlinking. Instead, skip buddy update
5252 before setup, as it will be regenerated after recovery is complete.
5257 Description: If an OST is inactive, its locks might reference stale inodes.
5258 Details : lov_change_cbdata() must iterate over all namespaces, even if
5259 they are inactive to clear inode references from the lock.
5261 Severity : enhancement
5262 Frequency : occasional, if non-standard max_dirty_mb used
5264 Description: Client will block write RPCs if not enough grant
5265 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5266 then the client will block writes while waiting for another RPC
5267 to complete instead of consuming its dirty limit. With change
5268 we get improved performance when max_dirty_mb is small.
5270 Severity : enhancement
5271 Bugzilla : 3389, 6253
5272 Description: Add support for supplementary groups on the MDS.
5273 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5274 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5275 lookups for user supplementary groups into a cache.
5279 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5280 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5281 for liblustre O_EXCL is enforced by MDS.
5285 Description: readdir never returns NULL in liblustre.
5286 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5287 dentry in struct dirent.
5291 Frequency : liblustre only, depends on application IO pattern
5292 Description: liblustre clients evicted if not contacting servers
5293 Details : Don't put liblustre clients into the ping_evictor list, so
5294 they will not be evicted by the pinger ever.
5296 Severity : enhancement
5298 Description: Add ability to evict clients by NID from MDS.
5299 Details : By echoing "nid:$NID" string into
5300 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5301 $NID would be instantly evicted from this MDS and from all active
5302 OSTs connected to it.
5306 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5307 Details : lookup_it_finish() used to query file size from OSTs that was not
5312 Description: service threads change working directory to that of init
5313 Details : Starting lustre service threads may pin the working directory
5314 of the parent thread, making that filesystem busy. Threads
5315 now change to the working directory of init to avoid this.
5319 Frequency : during shutdown only
5320 Description: shutdown with a failed MDS or OST can cause unmount to hang
5321 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5326 Frequency : 2.6 only
5327 Description: chmod/chown may include an extra supplementary group
5328 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5329 supplementary group and if none is specified this is used.
5332 Bugzilla : 5479 (6816)
5334 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5335 Details : If lookup is in progress on a file that is unlinked we might try
5336 to revalidate the inode and fail in revalidate after lookup is
5337 complete and ll_file_open() enqueues the open again but
5338 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5341 Frequency : always, if lconf --abort_recovery used
5343 Description: lconf --abort_recovery fails with 'Operation not supported'
5344 Details : lconf was attempting to abort recovery on the MDT device and not
5347 ------------------------------------------------------------------------------
5349 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5354 Frequency : rare (only unsupported configurations with a node running as an
5356 Bugzilla : 6514, 5137
5357 Description: Mounting a Lustre file system on a node running as an OST could
5359 Details : OSTs now preallocates memory needed to write out data at
5360 startup, instead of when needed, to avoid having to
5361 allocate memory in possibly low memory situations.
5362 Specifically, if the file system is mounted on on OST,
5363 memory pressure could force it to try to write out data,
5364 which it needed to allocate memory to do. Due to the low
5365 memory, it would be unable to do so and the node would
5366 become unresponsive.
5368 Severity : enhancement
5370 Description: Addition of lconf --service command line option
5371 Details : lconf now accepts a '--service <arg>' option, which is
5372 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5374 Severity : enhancement
5376 Description: Failover mode is now the default for OSTs.
5377 Details : By default, OSTs will now run in failover mode. To return to
5378 the old behaviour, add '--failout' to the lmc line for OSTs.
5380 Severity : enhancement
5382 Description: Health checks are now provided for MDS and OSTs
5383 Details : Additional detailed health check information on MSD and OSTs
5384 is now provided through the procfs health_check value.
5387 Frequency : occasional, depends on IO load
5389 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5390 after numerous create/delete cycles
5391 Details : The ext3 inode allocation policy would not allocate new inodes
5392 very well on the OSTs because there are no new directories
5393 being created. Instead we look for groups with free space if
5394 the parent directories are nearly full.
5399 Description: Network or server problems during mount may cause partially
5400 mounted clients instead of returning an error.
5401 Details : The config llog parsing code may overwrite the error return
5402 code during mount error handling, returning success instead
5408 Description: MDS can fail to allocate large reply buffers
5409 Details : After long uptimes the MDS can fail to allocate large reply
5410 buffers (e.g. zconf client mount config records) due to memory
5411 fragmentation or consumption by the buffer cache. Preallocate
5412 some large reply buffers so that these replies can be sent even
5413 under memory pressure.
5417 Frequency : rare (liblustre)
5418 Description: fsx running with liblustre complained that using truncate() to
5419 extend the file doesn't work. This patch corrects that issue.
5420 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5421 ATTR_SIZE and lsm use in llu_setattr_raw.
5425 Frequency : rare, only 2.6 kernels
5426 Description: Unusual file access patterns on the MDS may result in inode
5427 data being lost in very rare circumstances.
5428 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5429 code in the 2.6 kernel caused the inode and/or EA data not to
5430 be read from disk, causing single-file corruption.
5434 Frequency : rare, only 2.6 filesystems using extents
5435 Description: Heavy concurrent write and delete load may cause data corruption.
5436 Details : It was possible under high-load situations to have an extent
5437 metadata block in the block device cache from a just-unlinked
5438 file overwrite a newly-allocated data block. We now unmap any
5439 metadata buffers that alias just-allocated data blocks.
5443 Frequency : filesystems with default stripe_count larger than 77
5444 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5445 Details : lconf specifies an inode size of 4096 bytes when the default
5446 stripe_count is larger than 77. This conflicts with the default
5447 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5448 this case to avoid pinning too much memory for large EAs.
5450 ------------------------------------------------------------------------------
5452 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5457 Frequency : rare (extremely heavy IO load with hundreds of clients)
5459 Description: Client is evicted, gets IO error writing to file
5460 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5461 caused clients to be evicted for AST timeouts. The fixes in
5462 bug 5192 mean we no longer need to have such short AST timeouts
5463 so ldlm_timeout has been increased.
5466 Frequency : occasional during --force or --failover shutdown under load
5467 Bugzilla : 5949, 4834
5468 Description: Server oops/LBUG if stopped with --force or --failover under load
5469 Details : a collection of import/export refcount and cleanup ordering
5470 issues fixed for safer force cleanup
5473 Frequency : only filesystems larger than 120 OSTs
5474 Bugzilla : 5990, 6223
5475 Description: lfs getstripe would oops on a very large filesystem
5476 Details : lov_getconfig used kfree on vmalloc'd memory
5479 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5480 Bugzilla : 6242, 6243
5481 Description: reading from files that had been truncated to a non-zero size
5482 but never opened returned no data
5483 Details : ll_file_read() reads zeros from no-object files to EOF
5488 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5489 Details : There's little harm in aborting MDS/OSS recovery and letting it
5490 try again, so I removed the LASSERT and return an error instead.
5492 Severity : enhancement
5494 Description: New debugging infrastructure for tracking down data corruption
5495 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5496 (b) cover more of the client-side code path, and (c) try to narrow
5497 down where problems occurred
5501 Bugzilla : 3819, 4364, 4397, 6313
5502 Description: Racing close and eviction MDS could cause assertion in mds_close
5503 Details : It was possible to get multiple mfd references during close and
5504 client eviction, leading to one thread referencing a freed mfd.
5506 Severity: : enhancement
5507 Bugzilla : 3262, 6359
5508 Description: Attempts to reconnect to servers are now more aggressive.
5509 Details : This builds on the enhanced upcall-less recovery that was added
5510 in 1.4.2. When trying to reconnect to servers, clients will
5511 now try each server in the failover group every 10 seconds. By
5512 default, clients would previously try one server every 25 seconds.
5517 Description: After recovery, certain operations trigger a failed
5518 assertion on a client.
5519 Details : Failing over an mds, using lconf -d --failover, while a
5520 client was doing a readdir() call would cause the client to
5521 LBUG after recovery completed and the readdir() was resent.
5523 Severity : enhancement
5525 Description: Default groups are now added by lconf
5526 Details : You can now run lconf --group <servicename> without having to
5527 manually add groups with lmc.
5530 Frequency : occasional
5532 Description: Nodes with an elan id of 0 trigger a failed assertion
5535 Frequency : always when accessing e.g. tty/console device nodes
5537 Description: tty and some other devices nodes cannot be used on lustre
5538 Details : file's private_data field is used by device data and lustre
5539 values in there got lost. New field was added to struct file to
5540 store fs-specific private data.
5543 Frequency : when exporting Lustre via NFS
5545 Description: NFSD failed occasionally when looking up a path component
5546 Details : NFSD is looking up ".." which was broken in ext3 directories
5547 that had grown large enough to become hashed.
5550 Frequency : Clusters with multiple interfaces not on the same subnet
5552 Description: Nodes will repeatedly try to reconnect to an interface which it
5553 cannot reach and report an error to the log.
5554 Details : Extra peer list entries will be created by lconf with some peers
5555 unreachable. lconf now validates the peer before adding it.
5558 Frequency : Only if a default stripe is set on the filesystem root.
5560 Description: Setting a default stripe on the filesystem root prevented the
5561 filesystem from being remounted.
5562 Details : The client was sending extra request flags in the root getattr
5563 request and did not allocate a reply buffer for the dir EA.
5566 Frequency : occasional, higher if lots of files are accessed by one client
5567 Bugzilla : 6159, 6097
5568 Description: Client trips assertion regarding lsm mismatch/magic
5569 Details : While revalidating inodes the VFS looks up inodes with ifind()
5570 and in rare cases can find an inode that is being freed.
5571 The ll_test_inode() code will free the lsm during ifind()
5572 when it finds an existing inode and then the VFS later attaches
5573 this free lsm to a new inode.
5577 Bugzilla : 6422, 7030
5578 Description: MDS deadlock between mkdir and client eviction
5579 Details : Creating a new file via mkdir or mknod (starting a transaction
5580 and getting the ns lock) can deadlock with client eviction
5581 (gets ns lock and trying to finish a synchronous transaction).
5584 Frequency : occasional
5585 Description: While starting a server, the fsfilt_ext3 module could not be
5587 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5588 kernels. Previously, lconf would still use the ext3 name
5589 when trying to load modules. Now, it will correctly use
5590 ext3 on 2.4 and ldiskfs on 2.6.
5592 Severity : enhancement
5593 Description: The default stripe count has been changed to 1
5594 Details : The interpretation of the default stripe count (0, to lfs
5595 or lmc) has been changed to mean striping across a single
5596 OST, rather than all available. For general usage we have
5597 found a stripe count of 1 or 2 works best.
5599 Severity : enhancement
5600 Description: Add support for compiling against Cray portals.
5601 Details : Conditional compiling for some areas that are different
5605 Frequency : occasional
5606 Bugzilla : 6409, 6834
5607 Description: Creating files with an explicit stripe count may lead to
5608 a failed assertion on the MDS
5609 Details : If some OSTs are full or unavailable, creating files may
5610 trigger a failed assertion on the MDS. Now, Lustre will
5611 try to use other servers or return an error to the
5615 Frequency : occasional
5617 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5618 caused lock timeout during readahead (since 1.4.2).
5619 Details : Processes doing readahead might match a lock that hasn't been
5620 granted yet if there are overlapping and conflicting lock
5621 requests. The readahead process waits on ungranted lock
5622 (original lock is CBPENDING), while OST waits for that process
5623 to cancel CBPENDING read lock and eventually evicts client.
5625 Severity : enhancement
5627 Description: Initial enabling of flock support for clients
5628 Details : Implements fcntl advisory locking and file status functions.
5629 This feature is provided as an optional mount flag (default
5630 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5631 locking are implemented yet, and those that are are not guaranteed
5632 to be completely correct in production environments.
5633 mount -t lustre -o [flock|noflock] ...
5636 Frequency : occasional
5638 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5639 assertion in the ext3 JBD (journaling) layer.
5640 Details : The b_committed_data struct is protected by the big kernel lock
5641 in 2.4 kernels, serializing journal_commit_transaction() and
5642 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5643 access to this struct. In 2.6 kernels there is finer grained
5644 locking to improve SMP performance of the JBD layer.
5648 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5653 Description: merge and cleanup kernel patches.
5656 -----------------------------------------------------------------------------
5658 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5660 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5661 versions of Lustre. Please update all servers and clients to
5662 version 1.4.2 or later at the same time. You must also run
5663 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5664 to update the configuration logs.
5666 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5667 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5668 - swab llog records of type '0' so we get proper header size/idx (5861)
5669 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5670 - fix rename of one directory over another leaking an inode (5953)
5671 - avoid SetPageDirty on 2.6 (5981)
5672 - don't re-add just-being-destroyed locks to the waiting list (5653)
5673 - when creating new directories, inherit the parent's custom
5674 striping settings if present parent (3048)
5675 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5676 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5677 - don't hold client locks on temporary worklist from l_lru (5666)
5678 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5679 - automatically evict dead clients (5921)
5680 - Update file size properly in create+truncate+fstat case (6196)
5681 - Do not unhash mountpoint dentries, do not allow removal of
5683 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5684 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5685 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5686 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5687 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5688 - Correctly handle reads of files with no objects (6243)
5689 - lctl recover will also mark a device active if deactivate used (5933)
5691 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5692 - allow --write-conf on an MDS with different nettype than client (5619)
5693 - don't write config llogs to MDS for mounts not from that MDS (5617)
5694 - lconf should create multiple TCP connections from a client (5201)
5695 - init scripts are now turned off by default; run chkconfig --on
5696 lustre and chkconfig --on lustrefs to use them
5697 - upcalls are no longer needed for clients to recover to failover
5699 - add --abort-recovery option to lconf to abort recovery on device
5701 - add support for an arbitrary number of OSTs (3026)
5702 - Quota support protocol changes.
5703 - forward compatibility changes to wire structs (6007)
5704 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5705 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5706 - improved "lustre status" script
5707 - initialize blocksize for non-regular files (6062)
5708 - added --disable-server and --disable-client configure options (5782)
5709 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5710 - Vanilla 2.4.29 support
5711 - increase maximum number of obd devices to 520 (6242)
5712 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5713 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5714 SLES 9 2.6.5 kernels
5715 - limit stripes per file to 160 (the maximum EA size) (6093)
5717 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5720 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5721 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5722 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5723 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5724 - small ext3 extents cleanups and fixes (5733)
5725 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5726 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5727 - avoid needless client->OST connect, fix handle mismatch (5317)
5728 - fix DLM error path that led to out-of-sync client, long delays (5779)
5729 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5730 - fix several locking issues related to i_size (5492,5624,5654,5672)
5731 - don't move pending lock onto export if it is already evicted (5683)
5732 - fix kernel oops when creating .foo in unlinked directory (5548)
5733 - fix deadlock in obdfilter statistics vs. object create (5811)
5734 - use time_{before,after} to avoid timer jiffies wrap (5882)
5735 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5736 - Do not leak request if server was not able to process it (5154)
5737 - If mds_open unable to find parent dir, make that negative lookup(5154)
5738 - don't create new directories with extent-mapping (5909, 5936)
5740 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5741 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5742 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5743 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5744 - support for RHEL4 kernel on i686 (5773)
5745 - provide error messages when incompatible logs are encountered (5898)
5747 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5748 * version 1.4.0.10 (1.4.1 release candidate 1)
5750 - don't keep a lock reference when lock is not granted (4238)
5751 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5752 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5753 - drop import inflight refcount on signal_completed_replay error (5255)
5754 - unlock page after async write error during send (3677)
5755 - handle missing objects in filter_preprw_read properly (5265)
5756 - no transno return for symlink open, don't save no-trasno open (3440)
5757 - don't try to complete elan receive that already failed (4012)
5758 - free RPC server reply state on error (5406)
5759 - clean up thread from ptlrpc_start_thread() on error (5160)
5760 - readahead could read extra page into cache that wasn't ejected (5388)
5761 - prevent races in class_attach/setup/cleanup/detach (5260)
5762 - don't dereference de->d_inode after l_dput of de (5458)
5763 - use "int" for stripe value returned from lock_to_stripe (5544)
5764 - mballoc allocation and error-checking fixes in 2.6 (5504)
5765 - block device patches to fix I/O request sizes in 2.6 (5482)
5766 - look up hostnames for IB nals (5602)
5767 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5768 - don't start multiple acceptors for the same port (5277)
5769 - fix incorrect LASSERT in mds_getattr_name (5635)
5770 - export a proc file for general "ping" checking (5628)
5771 - fix "lfs check" to not block when the MDS is down (5628)
5773 - service request history (4965)
5774 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5775 - create an "evict_client" /proc entry on OSTs, like the MDS has
5776 - fix mount usage message, return errors per mount(8) (5168)
5777 - change grep [] to grep "[]" in tests so they work in more UMLs
5778 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5779 - remove ancient LOV_MAGIC_V0 EA support (5047)
5780 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5781 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5782 - allow readahead allocations to fail when low on memory (5383)
5783 - mmap locking landed again, after considerable improvement (2828)
5784 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5786 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5789 - send OST transaction number in read/write reply to free req (4966)
5790 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5791 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5792 - the watchdog thread now runs as interruptible (5246)
5793 - flock/lockf fixes (but it's still disabled, pending 5135)
5794 - don't use EXT3 constants in llite code (5094)
5795 - memory shortage at startup could cause assertion (5176)
5797 - reorganization of lov code
5798 - single portals codebase
5800 - add extents/mballoc support (5025)
5801 - direct I/O reads in the obdfilter (4048)
5802 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5804 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5807 - send OST transaction number in read/write reply to free req (4966)
5808 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5809 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5810 - the watchdog thread now runs as interruptible (5246)
5811 - handle missing objects in filter_preprw_read properly (5265)
5812 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5813 - drop import inflight refcount on signal_completed_replay error (5255)
5814 - unlock page after async write error during send (3677)
5815 - return original error code on reconstructed replies (3761)
5816 - no transno return for symlink open, don't save no-trasno open (3440)
5818 - add pid to ldlm debugging output (4922)
5819 - bump the watchdog timeouts -- we can't handle 30sec yet
5820 - extra debugging for orphan dentry/inode bug (5259)
5822 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5825 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5826 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5827 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5828 - fix NULL dereference in /proc/sys/portals/routes (4827)
5829 - allow failed mdc_close() operations to be interrupted (4561)
5830 - stop precreate on OST before MDS would time out on it (4778)
5831 - don't send partial-page writes before EOF from client (4410)
5832 - discard client grant for sub-page writes on large-page clients (4520)
5833 - don't free dentries not owned by NFS code, check generation (4806)
5834 - fix lsm leak if mds_create_objects() fails (4801)
5835 - limit debug_daemon file size, always print CERROR messages (4789)
5836 - use transno after validating reply (3892)
5837 - process timed out requests if import state changes (3754)
5838 - update mtime on OST during writes, return in glimpse (4829)
5839 - add mkfsoptions to LDAP (4679)
5840 - use ->max_readahead method instead of zapping global ra (5039)
5841 - don't interrupt __l_wait_event() during strace
5843 - add software watchdogs to catch hung threads quickly (4941)
5844 - make lustrefs init script start after nfs is mounted
5845 - fix CWARN/ERROR duplication (4930)
5846 - return async write errors to application if possible (2248)
5847 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5848 - print NAL number in %x format (4645)
5849 - update barely-supported suse-2.4.21-171 series (4842)
5850 - support for sles 9 %post scripts
5851 - support for building 2.6 kernel-source packages
5852 - support for sles km_* packages
5854 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5857 - ignore -ENOENT errors in osc_destroy (3639)
5858 - notify osc create thread that OSC is being cleaned up (4600)
5859 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5860 - reconstruct ost_handle() like mds_handle() (4657)
5861 - create a new thread to do import eviction to avoid deadlock (3969)
5862 - let lconf resolve symlinked-to devices (4629)
5863 - don't unlink "objects" from directory with default EA (4554)
5864 - hold socknal file ref over connect in case target is down (4394)
5865 - allow more than 32000 subdirectories in a single directory (3244)
5866 - fix blocks count for O_DIRECT writes (3751)
5867 - OST returns ENOSPC from object create when no space left (4539)
5868 - don't send truncate RPC if file size isn't changing (4410)
5869 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5870 - bind to privileged port in socknal and tcpnal (3689)
5872 - rate limit CERROR/CWARN console message to avoid overload (4519)
5873 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5874 - basic mmap support (3918)
5875 - kernel patch series update from b1_4 (4711)
5877 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5880 - avoid crash during MDS cleanup with OST shut down (2775)
5881 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5882 - don't use bad inodes on the MDS (3744)
5883 - dynamic object preallocation to improve recovery speed (4236)
5884 - don't hold spinlock over lock dumping or change debug flags (4401)
5885 - don't zero obd_dev when it is force cleaned (3651)
5886 - print grants to console if they go negative (4431)
5887 - "lctl deactivate" will stop automatic recovery attempts (3406)
5888 - look for existing locks in ldlm_handle_enqueue() (3764)
5889 - don't resolve lock handle twice in recovery avoiding race (4401)
5890 - revalidate should check working dir is a directory (4134)
5892 - don't always mark "slow" obdfilter messages as errors (4418)
5894 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5897 - don't close LustreDB during write_conf until it is done (3860)
5898 - fix typo in lconf for_each_profile (3821)
5899 - allow dumping logs from multiple threads at one time (3820)
5900 - don't allow multiple threads in OSC recovery (3812)
5901 - fix debug_size parameters (3864)
5902 - fix mds_postrecov to initialize import for llog ctxt (3121)
5903 - replace config semaphore with spinlock (3306)
5904 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5905 - don't allow enqueue to complete on a destroyed export (3822)
5906 - down write_lock before checking llog header bitmap (3825)
5907 - recover from lock replay timeout (3764)
5908 - up llog sem before sending rpc (3652)
5909 - reduce ns lock hold times when setting kms (3267)
5910 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5911 - fix NULL deref and obd_dev leak on setup error (3312)
5912 - replace some LBUG about llog ops with error handling (3841)
5913 - don't match INVALID dentries from d_lookup and spin (3784)
5914 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5915 - fix invalid assertion in ptlrpc_set_wait (3880)
5917 - add libwrap support for the TCP acceptor (3996)
5918 - add /proc/sys/portals/routes for non-root route listing (3994)
5919 - allow setting MDS UUID in .xml (2580)
5920 - print the stack of a process that LBUGs (4228)
5922 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5925 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5926 - make sure to unset replay flag from failed open requests (3440)
5927 - if default stripe count is 0, use OST count for inode size (3636)
5928 - update parent mtime/ctime on client for create/unlink (2611)
5929 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5930 - free recovery state on server during a forced cleanup (3571)
5931 - unregister_reply for resent reqs (3063)
5932 - loop back devices mounting and status check on 2.6 (3563)
5933 - fix resource-creation race that can provoke i_size == 0 (3513)
5934 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5935 - more debugging for page-accounting assertion (3746)
5936 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5937 - avoid deadlock after precreation failure (3758)
5938 - fix race and lock order deadlock in orphan handling (3450, 3750)
5939 - add validity checks when grabbing inodes from l_ast_data (3599)
5941 - add /proc/.../recovery_status to obdfilter (3428)
5942 - lightweight CDEBUG infrastructure, debug daemon (3668)
5943 - change default OSC RPC parameters to be better on small clusters
5944 - turn off OST read cache for files smaller than 32MB
5945 - install man pages and include them in rpms (3100)
5946 - add new init script for (un)mounting lustre filesystems (2593)
5947 - run chkconfig in %post for init scripts (3701)
5948 - drop scimac NAL (unmaintained)
5950 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5953 - clean kiobufs before and after use (3485)
5954 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5955 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5956 - change init script to fail more clearly if not run as root (1528)
5957 - allow clients to reconnect during replay (1742)
5958 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5959 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5960 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5961 - fix cond_resched (3554)
5962 - only evict unfinished clients after recovery (3515)
5963 - allow bulk resend, prevent data loss (3570)
5964 - dynamic ptlrpc request buffer allocation (2102)
5965 - don't allow unlinking open directory if it isn't empty (2904)
5966 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5967 - remove extraneous obd dereference causing LASSERT failure (3334)
5968 - don't use get_cycles() when creating temp. files on the mds (3156)
5969 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5970 - handle EEXIST for set-stripe, set proper directory name (3336)
5972 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5973 - fix ksocknal_fmb_callback() error messages (2918)
5975 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5978 - don't copy lvb into (possibly NULL) reply on error (2983)
5979 - don't deref dentry after dput, don't free lvb on error (2922)
5980 - use the kms to determine writeback rpc length (2947)
5981 - increment oti_logcookies when osc is inactive (2948)
5982 - update client's i_blocks count via lvb messages (2543)
5983 - handle intent open/close of special files properly (1557)
5984 - mount MDS with errors=remount-ro, like obdfilter (2009)
5985 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5986 - don't use cancelling-locks' kms values (2947)
5987 - use highest lock extent for kms, not last one (2925)
5988 - don't dereference ERR_PTR() dentry in error handling path (3107)
5989 - fix thread race in portals_debug_dumplog() (3122)
5990 - create lprocfs device entries at setup instead of at attach (1519)
5991 - common AST error handler, don't evict client on completion race (3145)
5992 - zero nameidata in detach_mnt in 2.6 (3118)
5993 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5994 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5995 - initialize RPC timeout timer earlier for 2.6 (3219)
5996 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5997 - print nal/nid for unknown nid (3258)
5998 - additional checks for oscc recovery before doing precreate (3284)
5999 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6000 - don't crash in mdc_close for bad permissions on open (3285)
6001 - zero i_rdev for non-device files (3147)
6002 - clear page->private before handing to FS, better assertion (3119)
6003 - tune the read pipeline (3236)
6004 - fix incorrect decref of invalidated dentry (2350)
6005 - provide read-ahead stats and refine rpc in flight stats (3328)
6006 - don't hold journal transaction open across create RPC (3313)
6007 - update atime on MDS at close time (3265)
6008 - close LDAP connection when recovering to avoid server load (3315)
6009 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6010 - don't leak open file on MDS after open resend (3325)
6011 - serialize filter_precreate and filter_destroy_precreated (3329)
6012 - loop device shouldn't call sync_dev() for nul device (3092)
6013 - clear page cache after eviction (2766)
6014 - resynchronize MDS->OST in background (2824)
6015 - refuse to mount the same filesystem twice on same mountpoint (3394)
6016 - allow llmount to create routes for mounting behind routers (3320)
6017 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6018 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6019 - fix rare problem with rename on htree directories (3417)
6021 - allow default OST striping configuration per directory (1414)
6022 - fix compilation for qswnal for 2.6 kernels (3125)
6023 - increase maximum number of MDS request buffers for large systems
6024 - change liblustreapi to be useful for external progs like lfsck (3098)
6025 - increase local configuration timeout for slow disks (3353)
6026 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6028 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6031 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6032 - don't overwrite extent policy data in reply if lock was blocked (2901)
6033 - drop filter export grants atomically with removal from device (2663)
6034 - del obd_self_export from work_list in class_disconnect_exports (2908)
6035 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6036 - swab reply message in mdc_close, other PPC fixes (2464)
6037 - fix destroying of named logs (2325)
6038 - overwrite old logs when running lconf --write_conf (2264)
6039 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6040 - fix race in target_handle_connect (2898)
6041 - mds_reint_create() should take same inode create lock (2926)
6042 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6043 - don't close files for self_export to avoid uninitialized obd (2936)
6044 - allow MDS with the same name as client node (2939)
6045 - hold dentry reference for closed log files for unlink (2325)
6046 - reserve space for all logs during transactions (2059)
6047 - don't evict page beyond end of stripe extent (2925)
6048 - don't oops on a deleted current working directory (2399)
6049 - handle hard links to targets without a parent properly (2517)
6050 - don't dereference NULL lock when racing during eviction (2867)
6051 - don't grow lock extents when lots of conflicting locks (2919)
6053 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - account for cache space usage on clients to avoid data loss (974)
6057 - lfsck support in lustre kernel code (2349)
6058 - reduce journal credits needed for BRW writes (2370)
6059 - orphan handling to avoid losing space on client/server crashes
6060 - ptlrpcd can be blocked, stopping ALL progress (2477)
6061 - use lock value blocks to assist in proper KMS, faster stat (1021)
6062 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6063 - recovery for initial connections (2355)
6064 - fixes for mds_cleanup_orphans (1934)
6065 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6066 - block all file creations until orphan recovery completes (1901)
6067 - client remove rq_connection from request struct (2423)
6068 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6069 - recovery timer race (2670)
6070 - mdc_close recovey bug (2532)
6071 - ptlrpc cleanup bug (2710)
6072 - mds timeout on local locks (2588)
6073 - namespace lock held during RPCs (2431)
6074 - handle interrupted sync write properly (2503)
6075 - don't try to handle a message that hasn't been replied to (2699)
6076 - client assert failure during cleanup after abort recovery (2701)
6077 - leak mdc device after failed mount (2712)
6078 - ptlrpc_check_set allows timedout requests to complete (2714)
6079 - wait for inflight reqs when ptlrpcd finishes (2710)
6080 - make sure unregistered services are removed from the srv_list
6081 - reset bulk XID's when resending them (caught by 1138 test)
6082 - unregister_bulk after timeout
6083 - fix lconf error (2694)
6084 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6085 - readahead locks pages, leaves pending causing memory pressure (2673)
6086 - increase OST request buffers to 4096 on large machines (2729)
6087 - fix up permission of existing directories in simple_mkdir (2661)
6088 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6089 - don't assign transno to errored transactions (2742)
6090 - don't delete objects on OST if given a bogus objid from MDS (2751)
6091 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6092 - if rq_no_resend, then timeout request after recovery (2432)
6093 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6094 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6095 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6096 - MDS recovery shouldn't delete objects that it has given out (2730)
6097 - if enqueue arrives after completion, don't clobber LVB (2819)
6098 - don't unlock pages twice when trigger_group_io returns error (2814)
6099 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6100 - don't write pages to disk if there was an error (1450)
6101 - don't ping imports that have recovery disabled (2676)
6102 - take buffered bytes into account when balancing socknal conn (2817)
6103 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6104 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6105 - remove little-endian swabbing of llog records (1987)
6106 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6107 - retry reposting request buffers if they fail (1191)
6108 - grow extent at grant time to avoid granting a revoked lock (2809)
6109 - lock revoke doesn't evict page if covered by a second lock (2765)
6110 - disable VM readahead to avoid reading outside lock extents (2805)
6112 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6113 - updated kernel patches for hp-2.4.20 kernel (2681)
6115 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6118 - fix truncated write corruption (2366)
6119 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6121 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6122 - ensure len(uuid) < 37 in lmc (1171)
6123 - fix ia64 OOPS in llog_test (2255)
6124 - zero end of page at obdfilter for partial page writes (2648)
6125 - don't leave stale dentries around after renames (bug 2428)
6126 - fix timeouts when evicting a client with a single lock held (2642)
6127 - set deadline for the initial HELLO message to drain (2634)
6128 - print out dotted-quad IP addresses in the socknal (2302)
6130 - additional debugging for MDS client eviction problem (2443)
6131 - fix mkfsoptions support for osts (2603, 2604)
6133 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6136 - add series for the vanilla 2.6.0 kernel
6137 - add series for the vanilla 2.4.24 kernel
6138 - add series for a cray x86/64 UL kernel drop
6139 - fix xattr patches for the vanilla 2.4.19 series
6141 - generate true UUIDs in lmc (1171)
6142 - have portals stack dumping break in UML (2466)
6143 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6144 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6145 - wake callback waiting threads on client eviction (2460)
6146 - Add --ptldebug and --subsystem to lmc (1719)
6147 - update assertion to allow safe interrupt allocation
6148 - set rq_no_resend for cancel requests (2432)
6149 - recalculate ptlrpcd timeout after resend (2494)
6150 - call vfs_rmdir when removing pending directories (2368)
6151 - fix renaming a file to itself (2429)
6152 - lmc creates a default one-stripe lov (2454)
6153 - expand procfs space to handle large clusters (2326)
6154 - increase UML stack to avoid overflow
6155 - update lconf's list of debug and subsystem masks
6156 - fix lfs find --obd (2510)
6157 - /proc tunable for disabling filter read caching (2591)
6158 - stop rpm packages from altering slapd.conf (2301)
6159 - disable nagle in the socknal under 0conf (2578)
6160 - choose mds inode size based on stripe count (2572)
6161 - fix kernel-source rpm problems (2516)
6163 - add --disable-doc to avoid pdf generation (2421)
6164 - update documentation, tests, type-os, comments
6165 - avoid format warnings on ia64
6166 - remove the TOE NAL
6167 - tiny code cleanups by removing unused fields
6169 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6172 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6173 - remove the most blatant lies from BUILDING file (2371)
6174 - change default debug level to reasonable production setting
6175 - reduce client side cache size to reduce cache flush time
6176 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6177 - make TCP zerocopy and pinger support enabled by default (2476)
6178 - sync writes completed after process exits caused crashes (2319)
6179 - maintain correct mount count on the MDS (2356)
6180 - backout 1557, because 2316 wasn't really fixed
6181 - better file I/O statistics gathering in /proc
6182 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6183 - another kernel patch to fix zero-copy TCP function export
6184 - don't take duplicate lock when processing re-sent getattr (2420)
6185 - lctl uses obd_self_export instead of creating new conn (2353)
6186 - MDS/OST recovery case which requires object creation asserted (2425)
6187 - move lfs from /usr/sbin to /usr/bin in packages
6188 - fix race between mds_client_add and mds_client_free (2417)
6189 - use kmalloc instead of slabs in portals (2430)
6190 - don't create duplicate records when a failover MDS is present (2442)
6191 - remove unnecessary mount age check (2332)
6192 - don't remove directory inodes from locks prematurely (2451)
6193 - don't break if MDS service name is the same as hostname (2103)
6194 - fix races in client write RPC generation when cache full (2482)
6196 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6199 - remove now-unused request->rq_obd (278)
6200 - if an allocation fails, print out how much memory we've used (1933)
6201 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6202 - add the "configurable stack size" patch to most series files (1256)
6203 - ability to write large log records, for 100+ OST configs (2306)
6204 - fix NULL deref when filter_prep fails (2314)
6205 - fix operator precedence error in filter_sync
6206 - dynamic allocation of socknal TX descriptors (2315)
6207 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6208 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6209 - ensure that CWARN messages in Portals always get to the syslog
6210 - __init/__exit are not for prototype decls (ldlm_init/exit)
6211 - x86-64 compile warning fixes
6212 - fix gateway LMC keyword conflict (2318)
6213 - fix MDS lock inversions in getattr/reint paths (1844)
6214 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6215 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6216 - DLM race condition prevented some lock evictions (2328)
6217 - ENOMEM detection and retry on socknal sends (2230)
6218 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6219 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6220 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6221 - better error messages when a client is rejected during recovery (1505)
6222 - avoid cancelling locks which were never granted, after failure (2330)
6223 - fix i_sem/journal inversion in mds_client_add (2333)
6224 - fix truncate/getattr lock cycle deadlock (2334)
6225 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6226 - fix two rare exit paths which could leak an l_lock() ref (2321)
6227 - fencepost error in MDS/OST orphan recovery (2226)
6228 - make log record alignment 8 bytes (1988)
6229 - lstripe now fails when requested offset > ost_count (2237)
6230 - ensure that all kernel series have a complete list.h (1607)
6231 - fix crashes in special-file operations (2316)
6232 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6234 - allow configurable automake binary, for testing new versions
6235 - small update to the lfs documentation
6237 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6239 * fix negative export reference count in fsfilt_sync (2312)
6241 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6242 * release candidate 0.9.1
6244 - orphans are moved into the PENDING directory for possible recovery
6245 - replayed opens now open by fid for orphan/rename safety (1042)
6246 - last close of an orphan inode generates a transno (683)
6247 - chdir() and mount() now pin the directory entry (1020)
6248 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6249 - discard very old requests without processing them (1502)
6250 - remove some common, well-understood CERRORs (1505)
6251 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6252 - clear "grant" flags in OST replies until OST grant code lands (1644)
6253 - fix read performance by not clobbering i_blksize on client (1598)
6254 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6255 - make LRU size a /proc tunable, clears locks when reduced (707)
6256 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6257 - server locks take references on exports now (1558)
6258 - build fixes for 2.4.20-rh trees (1663)
6259 - return an error from lov_create if all OSCs are inactive (1751)
6260 - fix import levels when a reconnect happens without a timeout (1597)
6261 - exit early from mds_open if we get a lookup error (1749)
6262 - partial page read at EOF wouldn't wait for disk before sending (1642)
6263 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6264 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6265 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6266 - write/truncate lock inversion (1639)
6267 - Don't auto-load obdclass, portals modules during cleanup (1495)
6268 - fix timestamps from jumping to "now" (1763)
6269 - extra journal assertions (1648)
6270 - add an extra multiunlink test (1771)
6271 - fix read_record/write_record API (1776)
6272 - fix leak of offset_extent, possible incorrect i_size later (1772)
6273 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6274 - Debugging for the kqswnal_get_idle_tx problems (1820)
6275 - Allow recovery to be attempted multiple times (1536)
6276 - Write out MDS last_rcvd file after it is first created (1600)
6277 - Fix tx_descriptor leak in failed transmit situations (1827)
6278 - ext3 journaling fixes for assertion failure after IO error (1871)
6279 - class_export_put() on freed export after completion AST error (1896)
6280 - Fix revalidate looping in VFS (1322)
6281 - Don't access a freed export during MDS_REINT timeout (1521)
6282 - Add open-unlink recovery support on the MDS (1673,1764)
6283 - Return an error if no MDS data was read from last_rcvd (1946)
6284 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6285 - Don't setup a disk device that doesn't match exported UUID (317)
6286 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6287 - avoid committing NULL handle in force close
6288 - local.sh is now a one-stripe LOV configuration
6289 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6290 - fix ext3 htree duplicate directory entry corruption (1516)
6291 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6292 - update to new LOV EA format (2097)
6293 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6295 2003-06-15 Phil Schwan <phil@clusterfs.com>
6298 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6299 - per-import/export recovery handling (958, 931, 959)
6300 - multiple last-rcvd slots, for serving multiple FSes (949)
6301 - connections are again shared between multiple imp/exports (963, 964)
6302 - "umount -f" would hang if any requests needed to be sent (393, 978)
6303 - avoid pinning large req buffer by copying for queued messages (989)
6304 - add "uuid" to "lctl device" command to help upcalls (991)
6305 - "open" RPCs with transnos would confuse recovery counters (1037)
6306 - do proper endian conversion of all wire messages (288, 340, 891)
6307 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6308 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6309 - fix ptlrpc_connection leak in target_handle_connect (1174)
6310 - fix import refcounting bug in OST and MDS cleanup (1134)
6311 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6312 - fix very unlikely obd_types race condition (501)
6313 - remove osc_open hack for echo_client (1187)
6314 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6315 - a failure in read_inode2 leads to deadlock (1139)
6316 - cancel ack-locks as soon as transaction is committed (1072)
6317 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6318 - make sure to commitrw after any preprw to avoid deadlock (1162)
6319 - failing to execute a file in a lustre FS would lock inode (1203)
6320 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6321 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6322 - obd_uuid2tgt would walk off the end of the list (1255)
6323 - on IA64 the osc would give portals incorrect bulk size (1258)
6324 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6325 - fix lock inversion caused by new llite matching code (1282)
6326 - limit the number of dirty pages on a client to 10MB (1286)
6327 - timed out locks were not being corrected cancelled (1289)
6328 - fix O_DIRECT above 4GB on IA-32 (1292)
6329 * major user-visible changes
6330 - fail out/fail over policy now controlled by the upcall (993)
6332 - add OBD_PING to check server availability and failure (954)
6333 - lustre messages are now sent in sending host order (288, 340, 891)
6334 - add eadatalen to MDS getattr reply (340)
6335 - OST read replies may contain second buffer, with per-page status (593)
6337 2003-03-11 Phil Schwan <phil@clusterfs.com>
6340 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6341 - failed open()s could cause deadlock; fixed (867, 869)
6342 - stop cancelling OST locks when files are closed (481)
6343 - overlapping XID spaces caused network corruption (851, 853)
6344 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6345 - setattr_raw added, to do single-RPC, server-side setattrs
6346 - lmc/lconf syntax change for OST UUIDs
6347 - fix crashy race condition between ptlrpc_free_req and osc_close
6348 - don't use request in mdc_enqueue if we hit a timeout (889)
6349 - don't set the inode i_size for regular files from the MDS (896)
6350 - handle out of order completion AST (842)
6351 - don't LBUG if a lock request times out after receiving AST (913)
6352 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6353 - if a bad lock AST arrives, send an error instead of dropping entirely
6354 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6355 - fix leak in bulk IO when only partially completed (899, 900, 926)
6356 - fix O_DIRECT for ia64 (55)
6357 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6358 - C-z after timeout could hang a process forever; fixed (977)
6360 - client-side I/O cache (678, 924, 929, 941, 970)
6362 - READPAGE and SETATTRs which don't take server-side locks get
6365 2003-02-11 Phil Schwan <phil@clusterfs.com>
6368 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6369 - Fix fsfilt_extN_readpage() to read a full page of directory
6370 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6371 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6372 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6373 - Most intent operations were removed, in favour of a new RPC mode
6374 that does a single RPC to the server and bypasses most of the VFS
6375 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6376 - Aggressively cancel local locks on DLM servers
6377 - mds_reint_unlink sends EA to the client if it's the last nlink.
6378 client uses that EA to unlink OST objects.
6379 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6380 - recursive symlinks were fixed (439)
6381 - fixed NULL deref in DEBUG_REQ
6382 - filter_update_lastobjid no longer calls sync, which annoyed extN
6383 - fixed multi-client small-writes to a single file problem (445)
6384 - fixed mtime updates during file writes (607)
6385 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6386 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6387 - fixed Linux OST size reporting problem (444, 656)
6388 - OST now updates object mtime with writes or setattr (607, 619)
6389 - client verifies file size before zeroing page past EOF (445)
6390 - OST now writes last allocated objid to disk with allocation (108)
6391 - LOV on echo now works (409)
6393 - mds_reint_unlink sends a new buffer, with the EA included. this
6394 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6395 set if a regular file was being unlinked, and it was the last link
6396 - use PtlGet from the target for bulk writes (315)
6397 - OST now updates object mtime with writes or setattr (607, 619)
6398 - LDLM now has a grant-time callback to revalidate locked items, if
6400 - Many MDS operations were reorganized to combat race conditions
6402 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6403 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6405 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6408 - Fully reactivate OST imports after reconnection (512, others)
6409 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6410 - More graceful error handling for truncating on dead OST (515)
6411 - Don't error out unless we're actually accessing dead stripes (474)
6412 - Fix garbage sizes when stripes are missing (410)
6413 - LRU counters were broken, causing constant lock purge (433, 432)
6414 - garbage on read from stripes with failed OSTs (441)
6415 - mark OSCs as active before reconnecting during recovery (438)
6416 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6417 - lfind did not preserve OST order in output (443)
6418 - symlinks cause hung clients, incorrect data (439)
6419 - stop dereferencing request after dropping refcount (457)
6420 - don't LASSERT(spin_is_locked) on non-SMP (455)
6421 - fixes for many rename() bugs
6422 - fstat didn't correctly synchronize attributes (399)
6423 - server must handle lock cancellation during blocking AST prep (487)
6424 - bulk descriptors were free()d too soon (511)
6425 - fix paths in lconf, which would load incorrect modules (451, 507)
6426 - fix confusing lconf 'host not found' error message (386)
6427 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6428 - fix race condition in mdc_blocking_ast() for inode access (526)
6429 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6430 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6431 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6432 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6433 - fix lstripe in several ways (kernel side) (527)
6434 - fix request leak in ldlm_cli_enqueue (262)
6435 - incorrect OSC was marked inactive after OST failure
6436 - call mds_fs_cleanup before unmounting filesystem (524)
6437 - fix races between taking ns_lock and ldlm_lock_change_resource
6438 - fix races updating LOV export open file list
6439 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6440 - fix recovery NULL deref in ldlm_cli_cancel_unused
6441 - fix some DLM races by using new hash table for lock handles (419)
6442 - permit the client to specify desired inodes, at replay
6443 - duplicate requests when we queue them for replay reintegration
6444 - fix last_rcvd offset calculation
6445 - sync after each recovered transaction, so we always make progress
6446 - never, not always, ERESTART requests without transnos
6447 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6449 - skip replay if the MDS says that the client is already connected
6450 - don't check for a recovery-enabled export to match lctl's UUID
6451 - don't INC_USE_COUNT for phantom exports
6452 - don't crash when cleaning up phantom exports (567)
6453 - don't double-finish or set replay data for errored mdc_open requests
6454 - abort requests when they time out, so we don't get old replies
6455 - send/receive replies for AST messages again
6456 - if the client says that it doesn't have the lock, cancel it on the
6458 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6459 mark it as destroyed, it will all work out when decref is called
6460 - fix module use counts (22, 581)
6462 - ASTs now expect a reply (server cancels lock on error reply)
6464 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6467 - fix many simultaneous client startup (392)
6468 - fix dentry->d_it clobbering
6469 - credentials weren't being shipped for readdir/getattr operations
6470 - remove invalid assertions triggered during some concurrent MD
6472 - proper Lustre versions added (336, 389)
6473 - fix memory leak for create error case (398)
6474 - fix LOV locking bug that would get cli/srv out of sync
6475 - fix echo client over LOV (409)
6476 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6477 - fix double-O_EXCL intent crash (424)
6478 - avoid sending multiple lock CANCELs (352)
6480 - MDS can do multi-client recovery (modulo bugs in new code)
6482 - many updates, edits, cleanups
6484 2002-11-18 Phil Schwan <phil@clusterfs.com>
6487 - fix null d_it dereference (346)
6488 - fix full OST/dbench hang (333)
6489 - fix permission problem with file removal (286)
6490 - fix removal of OSCs from LOV when they fail
6491 - fix NULL deref during bulk timeout (214)
6492 - fix problems related to multiple filesystems on one MDS (241)
6493 - fixed serious subtle metadata locking bugs
6494 - free locks on clients when inodes are removed due to memory
6496 - fix inode pointer in lock data (285)
6497 - partial support for multiple MDS on a single host (241)
6498 - data locks weren't cancelled at clear_inode time (290, 311)
6499 - intent locks could lead to unbounded lock growth (205)
6500 - added a maximum lock count, an LRU list, and a flusher
6501 - fix multiple rename (365)
6502 - properly abstracted the echo client
6503 - OSC locked 1 byte too many; fixed
6504 - rewrote brw callback code:
6505 - fixed recovery bugs related to LOVs (306)
6506 - fixed too-many-pages-in-one-write crash (191)
6507 - fixed (again) crash in sync_io_timeout (214)
6508 - probably fixed callback-related race (385)
6510 - Add capability to MDS protocol
6511 - LDLM cancellations and callbacks on different portals
6513 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6516 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6517 - defer open object destruction to close time (601981, 138)
6518 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6519 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6520 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6521 - don't call obd_setattr in ll_file_release for destroyed objects
6522 * protocol change to lustre_msg: move |version| and add |flags|
6523 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6524 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6525 * added replay of create, unlink, link and rename operations during
6526 MDS failover; recovery should be much more robust now
6527 * remove failed OSCs from LOVs (only lov_create uses this so far)
6528 * the lustre-HOWTO was brought (more) up to date (582544)
6530 2002-10-23 Phil Schwan <phil@clusterfs.com>
6533 - in-use dentries weren't being reused properly (617851)
6534 - prevent multiple LDLM setup (599178)
6535 - fix LOV size calculations for truncate (617853)
6536 - fix client handling of MDS intent errors (POSIX)
6537 - fix permission bug in lovstripe.c test (624321)
6538 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6539 - truncate past end of file could corrupt data
6540 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6541 - a race in recovery could return ETIMEDOUT to apps (623947)
6542 - building outside the source directory was fixed
6543 * the lustre-HOWTO was brought (more) up to date (582544)
6544 * major progress was made on recovery functionality
6546 2002-10-10 Phil Schwan <phil@clusterfs.com>
6549 - recovery deadlock fix
6550 - rm -rf causes LBUG fix (617817)
6551 - file open by multiple tasks fix (618962)
6552 - directory permissions bugs (602707 and 620007)
6553 - journal_stop fixed with locking (611313)
6554 - O_APPEND failures resolved (618273, perhaps 614459)
6555 - lconf PATH fix (619770)
6556 - IA64 build fix (621450)
6557 - RPC buffer sizes scale with amount of memory
6559 2002-10-01 Phil Schwan <phil@clusterfs.com>
6562 - locks would be cancelled without throwing away data pages,
6563 resulting in inconsistent data (605627)
6564 - inode attributes were not always being refreshed (605627, 612449)
6565 - lconf now continues to cleanup after lctl reports an error
6566 - MDS now enforces user permissions (602707)
6567 - lprocfs cleanup fixed, but not yet enabled (614157)
6568 - fixed infinite server hang, should a client not respond to an AST
6569 - avoid going into recovery if user calls readlink() with a buffer
6570 that's too small (613941)
6571 - AST RPCs no longer require replies (614867) -- this may be changed
6572 - don't crash server if client sends an IOV that's too big (611336)
6573 - fixed lock conversion deadlock (611892)
6574 - fixed the following of symlinks (614622)
6575 * recovery: the server can remove locks from a client that dies, other
6576 clients can make progress
6577 * more extN patch fixes
6578 * compile-time configurable ptlrpc buffer allocations
6580 - collaborative read cache document
6581 - Lustre Lite Performance CDR document-in-progress
6583 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6586 - fix typo in patch-2.4.18
6588 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6591 - clear ptlrpc request each time in handle_incoming_request()
6592 - unlink of files now destroys the object on the OST
6594 2002-09-19 Peter Braam <braam@clusterfs.com>
6596 * add hard link support
6597 * change obdfile creation method
6598 * kernel patch changed
6600 2002-09-19 Peter Braam <braam@clusterfs.com>
6603 - stack overflow bug in extN fixed
6605 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6607 * documentation updates
6608 - add man pages for config tools
6609 - update tests/README to describe testing with new config tools
6610 - finish metadata API descriptions
6611 * bug fixes and cleanups
6612 - statfs workaround for 16TB limit
6613 - LOV stripe allocation improved, can stripe on subset of OSTs
6614 - LOV file size/IO offset was wrong for files > 4GB in size
6615 - object EA data was being dropped, caused files to be unreadable
6616 - memory overflow with non-LOV OST caused memory corruption
6617 - fixed regression tests to work with new config tools, obdfilter
6618 - fixed bug when directory size became larger than 1 block
6619 - fixed bug (for single client case) when PWD was deleted
6620 - invalidate local directory pages when doing intent-based ops
6621 - avoid LDLM oops when lock callback contained bad data
6623 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6625 * documentation updates
6626 * bug fixes and cleanups
6627 - configuration tools
6630 - 64-bit compile warnings
6631 - 64-bit internal statfs data
6633 * test_brw on persistent OST devices
6635 * lprocfs (disabled)
6637 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6639 * documentation updates
6640 * bug fixes and cleanups
6641 * configuration tools
6643 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6646 * many small fixes to 0_5_4
6647 * io/network handling
6648 * thinkos in MDS operations
6650 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6653 * crucial basic fixes to 0.5.3
6654 * IOR, Iozone work over Elan
6657 2002-08-07 Phil Schwan <phil@clusterfs.com>
6658 * version 0_5_3, our first alpha
6659 * we use the new Portals iovs
6660 * documentation updates
6661 * bug fixes and cleanups
6662 * small changes in the DLM wire protocol
6664 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6665 * version 0_5_1 with some initial stability,
6666 * locking on MD and file I/O.
6667 * documentation updates
6668 * several bug fixes since 0.5.0
6669 * small changes in wire protocol
6671 2002-07-18 Phil Schwan <phil@clusterfs.com>
6673 * delivered as Lustre Light Alpha
6674 * fixed a crash after handling invalid MDS requests
6675 * fixed directory pages for architectures with non-4k pages sizes
6677 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6678 * release version v0_4_4
6679 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6681 * Fixes a number of interruption problems with OST operations.
6682 * Update documentation for portals header changes
6683 * Move all wire protocol structs/defines to lustre_idl.h
6684 * Fixes symlink length bug.
6685 * Add tcpdump to repository.
6687 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6688 * release version v0_4_3
6689 * Fixes statfs for inodes on extN.
6690 * Fixes bug in runtests which would delete /etc/hosts.
6691 * Use 64-bit object IDs wherever possible (not into VFS though)
6692 Remove ost_get_info, which is unused by lustre, and out of date.
6694 2002-07-03 Peter Braam <braam@clusterfs.com>
6695 * release version v0_4_2 Fixes a lookup error (type not passed)
6696 * move forward to head of Portals
6697 * move forward to latest Lustre kernel
6699 2002-06-25 Peter Braam <braam@clusterfs.com>
6700 * release version v0_4_1. Hopefully stable on single node use.