3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-693.21.1.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.114-92.64 (SLES12 SP2)
16 4.4.114-94.11 (SLES12 SP3)
17 3.13.0-101 (Ubuntu 14.04, ZFS only)
18 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
19 4.4.0-116 (Ubuntu 16.04)
20 vanilla linux 4.6.7 (ZFS only)
21 * Client known to build on unpatched kernels:
22 2.6.32-431.29.2.el6 (RHEL6.5)
23 2.6.32-504.30.3.el6 (RHEL6.6)
24 2.6.32-573.26.1.el6 (RHEL6.7)
25 2.6.32-642.15.1.el6 (RHEL6.8)
26 2.6.32-696.18.7.el6 (RHEL6.9)
27 3.10.0-693.21.1.el7 (RHEL7.4)
28 3.0.101-0.47.71 (SLES11 SP3)
29 3.0.101-107 (SLES11 SP4)
30 3.12.74-60.64.40 (SLES12 SP1)
31 4.4.114-92.64 (SLES12 SP2)
32 4.4.114-94.11 (SLES12 SP3)
33 3.13.0-101 (Ubuntu 14.04)
34 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
35 4.4.0-116 (Ubuntu 16.04)
37 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
38 * Recommended ZFS / SPL version: 0.7.6
39 * Tested with ZFS / SPL version: 0.7.6
40 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
41 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
42 stack overflow. For more information, please refer to bugzilla 17630.
43 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
44 the re-exporting nodes to be mounted with "32bitapi" mount option
45 * Note for Ubuntu 14.04.5:
46 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
47 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
49 --------------------------------------------------------------------------------
51 2017-07-05 Intel Corporation
53 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
54 for currently supported client and server kernel versions.
55 * Server known to build on patched kernels:
56 2.6.32-431.29.2.el6 (RHEL6.5)
57 2.6.32-504.30.3.el6 (RHEL6.6)
58 2.6.32-573.26.1.el6 (RHEL6.7)
59 2.6.32-642.15.1.el6 (RHEL6.8)
60 2.6.32-696.3.1.el6 (RHEL6.9)
61 3.10.0-514.21.1.el7 (RHEL7.3)
62 3.0.101-0.47.71 (SLES11 SP3)
63 3.0.101-100 (SLES11 SP4)
64 3.12.74-60.64.40 (SLES12 SP1)
65 4.4.59-92.17 (SLES12 SP2)
66 vanilla linux 4.6.7 (ZFS only)
67 * Client known to build on unpatched kernels:
68 2.6.32-431.29.2.el6 (RHEL6.5)
69 2.6.32-504.30.3.el6 (RHEL6.6)
70 2.6.32-573.26.1.el6 (RHEL6.7)
71 2.6.32-642.15.1.el6 (RHEL6.8)
72 2.6.32-696.3.1.el6 (RHEL6.9)
73 3.10.0-514.21.1.el7 (RHEL7.3)
74 3.0.101-0.47.71 (SLES11 SP3)
75 3.0.101-100 (SLES11 SP4)
76 3.12.74-60.64.40 (SLES12 SP1)
77 4.4.59-92.17 (SLES12 SP2)
79 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
80 * Recommended ZFS / SPL version: 0.6.5.9
81 * Tested with ZFS / SPL version: 0.6.5.9
82 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
83 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
84 stack overflow. For more information, please refer to bugzilla 17630.
85 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
86 the re-exporting nodes to be mounted with "32bitapi" mount option
88 --------------------------------------------------------------------------------
90 12-07-2016 Intel Corporation
92 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
93 for currently supported client and server kernel versions.
94 * Server known to build on patched kernels:
95 2.6.32-431.29.2.el6 (RHEL6.5)
96 2.6.32-504.30.3.el6 (RHEL6.6)
97 2.6.32-573.26.1.el6 (RHEL6.7)
98 2.6.32-642.11.1.el6 (RHEL6.8)
99 3.10.0-514.el7 (RHEL7.3)
100 3.0.101-0.47.71 (SLES11 SP3)
101 3.0.101-80 (SLES11 SP4)
102 3.12.62-60.64.8 (SLES12 SP1)
103 vanilla linux 4.5.7 (ZFS only)
104 * Client known to build on unpatched kernels:
105 2.6.32-431.29.2.el6 (RHEL6.5)
106 2.6.32-504.30.3.el6 (RHEL6.6)
107 2.6.32-573.26.1.el6 (RHEL6.7)
108 2.6.32-642.11.1.el6 (RHEL6.8)
109 3.10.0-514.el7 (RHEL7.3)
110 3.0.101-0.47.71 (SLES11 SP3)
111 3.0.101-80 (SLES11 SP4)
112 3.12.62-60.64.8 (SLES12 SP1)
114 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
115 * Recommended ZFS / SPL version: 0.6.5.8
116 * Tested with ZFS / SPL version: 0.6.5.8
117 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
118 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
119 stack overflow. For more information, please refer to bugzilla 17630.
120 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
121 the re-exporting nodes to be mounted with "32bitapi" mount option
123 --------------------------------------------------------------------------------
125 02-29-2016 Intel Corporation
127 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
128 for currently supported client and server kernel versions.
129 * Server known to build on patched kernels:
130 2.6.32-431.29.2.el6 (RHEL6.5)
131 2.6.32-504.30.3.el6 (RHEL6.6)
132 2.6.32-573.12.1.el6 (RHEL6.7)
133 3.10.0-327.3.1.el7 (RHEL7.2)
134 3.0.101-0.47.71 (SLES11 SP3)
135 3.0.101-68 (SLES11 SP4)
136 vanilla linux 4.2.1 (ZFS only)
137 * Client known to build on unpatched kernels:
138 2.6.32-431.29.2.el6 (RHEL6.5)
139 2.6.32-504.30.3.el6 (RHEL6.6)
140 2.6.32-573.12.1.el6 (RHEL6.7)
141 3.10.0-327.3.1.el7 (RHEL7.2)
142 3.0.101-0.47.71 (SLES11 SP3)
143 3.0.101-68 (SLES11 SP4)
146 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
147 * Recommended ZFS / SPL version: 0.6.4.2
148 * Tested with ZFS / SPL version: 0.6.4.2
149 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
150 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
151 stack overflow. For more information, please refer to bugzilla 17630.
152 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
153 the re-exporting nodes to be mounted with "32bitapi" mount option
155 --------------------------------------------------------------------------------
157 03-10-2015 Intel Corporation
159 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
160 for currently supported client and server kernel versions.
161 * Server known to build on patched kernels:
162 2.6.32-431.29.2.el6 (RHEL6.5)
163 2.6.32-504.8.1.el6 (RHEL6.6)
164 3.0.101-0.46 (SLES11 SP3)
165 * Client known to build on unpatched kernels:
166 2.6.32-431.29.2.el6 (RHEL6.5)
167 2.6.32-504.8.1.el6 (RHEL6.6)
168 3.10.0-123.20.1.el7 (RHEL7)
169 3.0.101-0.46 (SLES11 SP3)
170 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
171 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
172 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
173 stack overflow. For more information, please refer to bugzilla 17630.
174 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
175 the re-exporting nodes to be mounted with "32bitapi" mount option
177 Severity : enhancement
179 Description: control OST-index in IDIF via ROCOMPAT flag.
180 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
181 last_rcvd file. For new formatted OST device, it will be auto set;
182 for the case of upgrading from old OST device, you can enable it
183 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
184 enabled, for new created OST-object, its IDIF-in-LMA will contain
185 the OST-index; for the existing OST-object, the OSD will convert
186 old format IDIF as new format IDIF with OST-index stored in the
187 LMA EA when accessing such OST-object or via OI scrub. Once such
188 flag is enabled, it cannot be reverted back, so the system cannot
189 be downgraded to the orignal incompatible version.
191 --------------------------------------------------------------------------------
193 07-30-2014 Intel Corporation
195 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
196 for currently supported client and server kernel versions.
197 * Server known to build on patched kernels:
198 2.6.32-431.20.3.el6 (RHEL6.5)
199 3.0.101-0.31 (SLES11 SP3)
200 * Client known to build on unpatched kernels:
201 2.6.32-431.20.3.el6 (RHEL6.5)
202 3.0.101-0.31 (SLES11 SP3)
205 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
206 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
207 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
208 stack overflow. For more information, please refer to bugzilla 17630.
209 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
210 the re-exporting nodes to be mounted with "32bitapi" mount option
212 --------------------------------------------------------------------------------
214 10-31-2013 Intel Corporation
216 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
217 * Server builds for kernels:
218 2.6.32-358.18.1.el6 (RHEL6.4)
219 * Client builds for unpatched kernels:
220 2.6.32-358.18.1.el6 (RHEL6.4)
221 3.0.92-0.8 (SLES11 SP3)
223 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
224 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
225 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
226 stack overflow. For more information, please refer to bugzilla 17630.
227 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
228 the re-exporting nodes to be mounted with "32bitapi" mount option
230 --------------------------------------------------------------------------------
232 05-30-2013 Intel Corporation
234 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
235 * Server builds for kernels:
236 2.6.32-358.6.2.el6 (RHEL6.4)
237 * Client builds for unpatched kernels:
238 2.6.32-358.6.2.el6 (RHEL6.4)
239 3.0.74-0.6.6 (SLES11 SP2)
241 * Recommended e2fsprogs version: 1.42.7.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 --------------------------------------------------------------------------------
250 9-30-2012 Whamcloud, Inc.
252 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
253 * Server support for kernels:
254 2.6.32-279.5.1.el6 (RHEL6)
255 * Client support for unpatched kernels:
256 2.6.32-279.5.1.el6 (RHEL6)
258 2.6.18-238.19.1.el5 (RHEL5)
259 2.6.16.60-0.69.1 (SLES10)
260 * Recommended e2fsprogs version: 1.41.90.wc4
261 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
262 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
263 stack overflow. For more information, please refer to bugzilla 17630.
264 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
265 the re-exporting nodes to be mounted with "32bitapi" mount option
267 Severity : enhancement
269 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
271 Severity : enhancement
273 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
275 Severity : enhancement
277 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
279 Severity : enhancement
281 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
283 Severity : enhancement
285 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
287 --------------------------------------------------------------------------------
289 04-30-2012 Whamcloud, Inc.
291 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
292 * Server support for kernels:
293 2.6.32-220.4.2.el6 (RHEL6)
294 * Client support for unpatched kernels:
295 2.6.32-220.4.2.el6 (RHEL6)
297 2.6.18-238.19.1.el5 (RHEL5)
298 2.6.16.60-0.69.1 (SLES10)
299 * Recommended e2fsprogs version: 1.41.90.wc4
300 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
301 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
302 stack overflow. For more information, please refer to bugzilla 17630.
303 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
304 the re-exporting nodes to be mounted with "32bitapi" mount option
306 Severity : enhancement
308 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
310 Severity : enhancement
312 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
314 Severity : enhancement
316 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
318 Severity : enhancement
320 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
322 --------------------------------------------------------------------------------
324 09-30-2011 Whamcloud, Inc.
326 * Server support for kernels:
327 2.6.18-238.19.1.el5 (RHEL5)
328 2.6.32-131.6.1.el6 (RHEL6)
329 * Client support for unpatched kernels:
330 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
331 2.6.18-238.19.1.el5 (RHEL5)
332 2.6.32-131.6.1.el6 (RHEL6)
333 2.6.29.4-167.fc11 (FC11)
334 2.6.16.60-0.69.1 (SLES10)
336 * Recommended e2fsprogs version: 1.41.90.wc3
337 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
338 Lustre filesystem with 4K stack may cause a stack overflow. For more
339 information, please refer to bugzilla 17630.
340 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
341 mounted with "32bitapi" mount option
343 Severity : enhancement
344 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
346 Severity : enhancement
347 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
349 Severity : enhancement
350 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
353 Frequency : only with SLES10
355 Description: Use OFED "KMP" provided by Novell
356 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
357 Lustre is now built against this package. That means you need to
358 install the ofed-kmp package from Novell for the patchless client.
359 Note that the ofed-kmp that Novell ships may not exactly match the
360 kernel version but should still be compatible.
362 Severity : enhancement
364 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
365 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
367 --------------------------------------------------------------------------------
369 2010-07-15 Oracle, Inc.
371 * Server support for kernels:
372 2.6.18-164.11.1.el5 (RHEL 5)
373 2.6.18-164.11.1.0.1.el5 (OEL 5)
374 * Client support for unpatched kernels:
375 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
376 2.6.18-164.11.1.el5 (RHEL 5),
377 2.6.18-164.11.1.0.1.el5 (OEL 5)
378 2.6.16.60-0.42.8 (SLES 10),
380 2.6.29.4-167.fc11 (FC11)
381 * Recommended e2fsprogs version: 1.41.10-sun2
383 Severity : enhancement
385 Description: Distribute lfs_migrate script with lustre
386 Details : The lfs_migrate mentioned in the manual is now included in the
387 lustre tree in lustre/scripts and will distributed in lustre RPMs.
389 Severity : enhancement
391 Description: Failover nid list added to import proc file.
393 Severity : enhancement
395 Description: "som_preview" mount option required to use SOM
396 Details : The clients will need to use som_preview mount option
400 Description: Lustre does not do 1MB IOs to HW RAID
401 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
403 Severity : enhancement
405 Description: lfs quota output is not very convenient for awk/sed-parsing
406 Details : Some positions in lfs quota output table could be empty or
407 non-empty which made it hard to parse it with scripts, now
408 a dash is put instead of space where there is not supposed
413 Description: Update to OFED 1.5.1
414 Details : OFED 1.5.1 will be used in environments that don't use
415 vendor supplied OFED.
419 Description: Remove lnet. prefix from lctl params display
420 Details : Also renamed lnet.memused to lnet_memused to
425 Description: Use conf_param -d to remove permanant settings.
429 Description: Add list_param -R
430 Details : lctl list_param now supports recursive mode.
433 Bugzilla : 21829, 22312
434 Description: Fix llobdstat.
435 Details : 1. Now updating the write_bytes counter
436 2. Port script fixes from 1.8.
440 Description: decrease the usage of memory on clients.
441 Details : 1. On clients, recycle dentries and inodes unused.
442 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
447 Description: llobdstat fix and enhancement
448 Details : add a counter to set a limit to how many samples will be returned
449 fix a wildcard in the path to limit to obdfilter stats only
453 Description: lfs check only allowed for root
454 Details : Code cleanups and added a regresion test of non-root use of
459 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
460 Details : Fix hr->hr_index overload.
464 Description: Fix FIEMAP functionality.
468 Description: FC11 patchless client support added
472 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
473 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
477 Description: New acc-sm script and various test-framework enhancments.
481 Description: Previously landed patch was reverted.
485 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
489 Description: Add ldiskfs maxdirsize mount option.
493 Description: Fix mkfs.lustre support for 16TB LUNs.
497 Description: Print mmp_check_interval
498 Details : print mmp_check_interval and make it possible to abort
499 mount operation in case it takes too long.
503 Description: Add cascading_rw to lustre mpi tests.
507 Description: Store SOM data in EA.
508 Details : Also several SOM related cleanups.
512 Description: Add refcount to md_open_data
516 Description: Use refcount on lov_request_set to prevent memory corruption.
520 Description: Statahead: use parent dir's i_mutex for synchronization.
524 Description: loadgen improvements.
528 Description: Enable peer health detection in ksocklnd and ko2iblnd.
532 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
536 Description: Fix race on obd_devcie:md_stats
540 Description: Fix SLES 11 client builds.
541 Details : Resolved uninitalized pointer gcc warnings.
545 Description: Allow non-root access for "lfs check".
546 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
550 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
552 Severity : enhancement
554 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
555 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
559 Description: Sanity test_27p fails.
560 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
561 Sanity tests 27 were cleaned up and improved.
565 Description: Unify req format on client/servers
566 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
567 buffers and swabbers approach..
572 Description: drop join file stuff.
574 Severity : enhancement
576 Description: Add 2.6.27 and SLES11 (patchless client) support
580 Description: Fix in ptlrpc_expire_one_request() to print the signed time
583 Severity : enhancement
585 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
588 Severity : enhancement
590 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
592 Severity : enhancement
594 Description: correctly handle big reply message.
595 Details : send LNet event if reply is bigger than buffer and adjust this buffer
600 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
602 Severity : enhancement
604 Description: Add support for OEL5.
608 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
610 Severity : enhancement
612 Description: "weak-modules" support
613 Details : Implement "weak-modules" support which enables kernel modules
614 to be used with any kernel that implements the same kABI. In
615 order to achieve this modules are now installed in
616 /lib/modules/$(uname -r)/updates/kernel on all distributions.
620 Description: client could not reconnect to OST because of an active request.
621 Details : The client now retries to reconnect to the same server, if a
622 connect request failed with EBUSY or -EAGAIN. Test case 35b
623 conf-sanity.sh has been created.
627 Description: Don't allow make backward step on assignin osc next id.
628 Details : race between allocation next id and ll_sync thread can be cause
629 of set wrong osc next id and can be kill valid ost objects.
633 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
637 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
641 Description: Error handling in osc_statfs_interpret() has been improved.
642 Details : Check in osc_statfs_interpret() for EBADR.
646 Description: Do not send statfs() requests to OSTs disabled by administrator.
647 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
649 Severity : enhancement
651 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
653 Severity : enhancement
655 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
657 Severity : enhancement
659 Description: Enhancement for lfs(1) command to use numeric uid/gid.
663 Description: Improved cleanup procedure for conf-sanity test_46a.
667 Description: Do not start bulk transfer, if client tried to reconnect.
668 Details : Do not start bulk transfer, if client tried to reconnect,
669 but failed due to active rpcs. Check for exp_abort_active_req
670 has been implemented.
674 Description: Prevent inconsistences between linux and lustre mount structures.
675 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
676 Make the sleep interruptible.
678 Severity : enhancement
680 Description: provide server to client comms path
681 Details : server to client communications path via new LDLM_SET_INFO rpc
685 Description: truncate starts GFP_FS allocation under transaction causing deadlock
686 Details : ldiskfs_truncate calls grab_cache_page which may start page
687 allocation under an open transaction. This may lead to
688 calling prune_icache with consequent lustre reentrance.
692 Description: send correctly lsm on open replay
693 Details : MDS is trust to LSM size on replay open, but client can set wrong size
698 Description: enable client interface failover
699 Details : When a child reconnects from another NID, properly update export
700 nid hash position and ldlm reverse import.
702 Severity : enhancement
704 Description: Communicate OST degraded/readonly state via statfs to MDS
705 Details : Flags in the statfs returned from OSTs indicate whether the
706 OST is in a degraded RAID state, or if the filesystem has
707 turned read-only after a filesystem error is detected.
712 Description: don't panic if EPROTO was hit when reading symlink
713 Details : correctly handling request reference in error cases.
715 Severity : enhancement
717 Description: Add LustreNetLink, a kernel-userspace communcation path.
719 Severity : enhancement
721 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
726 Description: access to llog context before init.
727 Details : move handling CATALOGS file at osc layer and forbid access to llog
731 Frequency : rare, only when client is writing to OST on local node
733 Description: Avoid deadlock for local client writes
734 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
735 memory freeing context. This allows OST threads to set the
736 PF_MEMALLOC flag on task structures in order to allocate memory
737 from reserved pools and complete IO.
738 Use GFP_HIGHUSER for OST allocations for non-local client writes,
739 so that the OST threads generate memory pressure and allow
740 inactive pages to be reclaimed.
742 Severity : enhancement
744 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
749 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
750 Details : move ldlm namespace creation in setup phase to avoid grab
751 _lprocfs_lock with cli_sem held.
755 Description: Temporarily disable grant shrink.
756 Details : Disable the feature for debugging.
759 Frequency : only when formatting test filesystems
761 Description: Unable to run several mkfs.lustre on loop devices at the same time
762 Details : mkfs.lustre returns error 256 on the concurrent loop devices
763 formatting. The solution is to proper handle the error.
765 Severity : enhancement
767 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
769 Severity : enhancement
771 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
776 Description: resolve race between obd_disconnect and class_disconnect_exports
777 Details : if obd_disconnect will be called to already disconnected export he
778 forget release one reference and osc module can't unloaded.
780 Severity : enhancement
782 Description: Allow tuning service thread via /proc
783 Details : For each service a new
784 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
785 created that can be used to set min/max thread counts, and get the
786 current number of running threads.
791 Descriptoin: don't return error if have partially created objects for file.
792 Details : lov_update_create_set uses set->set_success as index for created
793 objects, so if some requests will be failed, they will have hole
794 at end of array and qos_shrink_lsm for allocate correct lsm.
796 Severity : enhancement
798 Description: Update OFED support to 1.4.1
803 Description: Reduce small size read RPC
804 Details : Set read-ahead limit for every file and only do read-ahead when
805 available read-ahead pages are bigger than 1M to avoid small size
808 Severity : enhancement
810 Description: move AT tunable parameters for more consistent usage
811 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
813 Severity : enhancement
815 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
816 Details : allow skip disconnected ost for send statfs request and hide error
822 Description: fix racy locking of mballoc block bitmaps causing BUG
823 Details : The locking of the mballoc buddy bitmap and the in-memory
824 block bitmap was using two different spin locks in some
825 cases. This made it possible to incorrectly access the
826 mballoc bitmap while another process was modifying it,
827 causing a sanity assertion to fail. While no on-disk corruption
828 was reported, there was some risk of this happening.
831 Frequency : rare, on llog test 6
833 Descriptoin: don't allow connect to already connected import
834 Details : allowing connect to already connected import is hide connecting problem.
839 Descriptoin: allow kill process which wait statahead result
840 Details : for some reasons 'ls' can stick in waiting result from statahead,
841 in this case need way for kill this process.
843 Severity : enhancement
845 Description: Add state history info file, enhance import info file
846 Details : Track import connection state changes in a new osc/mdc proc file;
847 add overview-type data to the osc/mdc import proc file.
849 Severity : enhancement
851 Description: MDS create should not wait for statfs RPC while holding DLM lock.
853 Severity : enhancement
855 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
860 Descriptoin: OST grant too much space to client even there are not enough space.
861 Details : Client will shrink its grant cache to OST if there are no write
862 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
863 this grant cache if there are already not enough avaible space
864 (left_space < total_clients * 32M).
867 Frequency : start MDS on uncleanly shutdowned MDS device
869 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
870 Details : stay in waiting mds<>ost recovery finished produce random bugs
871 due race between two ll_sync thread for one lov target. send
872 ACTIVATE event only if connect realy finished and import have
876 Frequency : rare, connect and disconnect target at same time
878 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
879 Details : don't call obd_disconnect under lov_lock. this long time
880 operation and can block ptlrpcd which answer to connect request.
885 Descriptoin: don't lose wakeup for imp_recovery_waitq
886 Details : recover_import_no_retry or invalidate_import and import_close can
887 both sleep on imp_recovery_waitq, but we was send only one wakeup
891 Frequency : always with long access acl
893 Descriptoin: mds can't pack reply with long acl.
894 Details : mds don't control size of acl but they limited by reint/getattr
897 Severity : enhancement
899 Description: Update to SLES10 kernel-2.6.16.60-0.33.
901 Severity : enhancement
903 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
906 Frequency : start MDS on uncleanly shutdowned MDS device
908 Descriptoin: aborting recovery hang on MDS
909 Details : don't throttle destroy RPCs for the MDT.
912 Frequency : on remount
914 Description: external journal device not working after the remount
915 Details : clear dev_rdonly flag for external journal devices in
921 Description: shutdown vs evict race
922 Details : client_disconnect_export vs connect request race.
923 if client will evicted at this time - we start invalidate
924 thread without referece to import and import can be freed
928 Frequency : rare, need acl's on inode.
930 Description: client can't handle ost additional correctly
931 Details : if ost was added after client connected to mds client can have
932 hit lnet_try_match_md ... to big messages to wide striped files.
933 in this case need teach client to handle config events about add
934 lov target and update client max ea size at that event.
936 Severity : enhancement
938 Description: Changelogs
939 Details : Changelogs are a lightweight mechanism to track filesystem
940 metadata and namespace changes. The changelog is recorded
941 permanently on the MDTs, and is periodically "consumed" / purged
942 when records are no longer needed.
944 Severity : enhancement
946 Description: compact fld format with extents
947 Details : Store range of seq rather than every seq in FLD. Seq
948 controller update FLD rather than clients. In Case of CMD, mdt0
949 has FLD, all other metadata server act as non persistent proxy
950 for FLD queries and cache fld entries in fld cache.
955 Description: don't skip ost target if they assigned to file
956 Details : Drop slow OSCs if we can, but not for requested start idx.
957 This means "if OSC is slow and it is not the requested
958 start OST, then it can be skipped, otherwise skip it only
959 if it is inactive/recovering/out-of-space.
963 Description: more cleanup in mds_lov
964 Details : not send LOV EA under replay, we can't know about they size at this
965 time. Don't allow client connect to mds before any ost connected,
966 for avoid problems with LOV EA size and returning EIO to client.
968 Severity : enhancement
970 Description: Interoperability at server side (Disk interoperability)
972 Severity : enhancement
974 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
976 Severity : enhancement
978 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
980 Severity : enhancement
982 Description: New client IO stack (CLIO).
984 Severity : enhancement
986 Description: Commit on sharing. Eliminate inter-client dependencies between
987 uncommitted transactions by doing transaction commits.
988 Thereby clients may recovery independently.
991 Frequency : Create a symlink file with a very long name
993 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
994 Details : If there is no extra space in the request for early cancels,
995 ldlm_req_handles_avail() returns 0 instead of a negative value.
997 Severity : enhancement
999 Description: Add /proc entry for import status
1000 Details : The mdc, osc, and mgc import directories now have
1001 an import directory that contains useful import data for debugging
1002 connection problems.
1004 Severity : enhancement
1006 Description: Re-disable certain /proc logging
1007 Details : Enable and disable client's offset_stats, extents_stats and
1008 extents_stats_per_process stats logging on the fly.
1011 Frequency : Only on FC kernels 2.6.22+
1013 Description: oops in statahead
1014 Details : Do not drop reference count for the dentry from VFS when lookup,
1015 VFS will do that by itself.
1017 Severity : enhancement
1019 Description: Generic /proc file permissions
1020 Details : Set /Proc file permissions in a more generic way to enable non-
1021 root users operate on some /proc files.
1025 Description: Hitting mdc_commit_close() ASSERTION
1026 Details : Properly handle request reference release in
1027 ll_release_openhandle().
1031 Description: quota recovery deadlock during mds failover
1032 Details : This patch includes att18982, att18236, att18237 in bz14840.
1034 1. fix osts hang when mds does failover with quotaon
1035 2. prevent watchdog storm when osts threads wait for the
1040 Frequency : only patchless client
1041 Description: add workaround for race between add/remove dentry from hash
1043 Severity : enhancement
1045 Description: Allow OST glimpses to return PW locks
1049 Description: LBUG when llog conf file is full
1050 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1054 Description: Prevent import from entering FULL state when server in recovery
1058 Description: service mount cannot take device name with ":"
1059 Details : Only when device name contains ":/" will mount treat it as
1065 Description: replace ptlrpcd with the statahead thread to interpret the async
1066 statahead RPC callback
1070 Frequency : on recovery
1071 Description: I/O failures after umount during fail back
1072 Details : if client reconnected to restarted server we need join to recovery
1073 instead of find server handler is changed and process self
1074 eviction with cancel all locks.
1076 Severity : enhancement
1078 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1080 Severity : enhancement
1082 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1084 Severity : enhancement
1086 Description: Upcall on Lustre log has been dumped
1087 Details : Allow for a user mode script to be called once a Lustre log has
1088 been dumped. It passes the filename of the dumped log to the
1089 script, the location of the script can be specified via
1090 /proc/sys/lnet/debug_log_upcall.
1095 Description: avoid idr_remove called for id which is not allocated.
1096 Details : Move assigment s_dev for clustered nfs to end of initialization,
1097 to avoid problem with error handling.
1102 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1103 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1104 into CONN_USED_HASH and this prodice warning when put connection
1105 again in unused hash.
1108 Severity : enhancement
1110 Description: File striping can now be set to use an arbitrary pool of OSTs.
1112 Severity : enhancement
1114 Description: Export bytes_read/bytes_write count on OSC/OST.
1118 Description: Early reply size mismatch, MGC loses connection
1119 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1120 the connect flags are properly negotiated.
1125 Description: quotas are not honored with O_DIRECT
1126 Details : all writes with the flag O_DIRECT will use grants which leads to
1127 this problem. Now using OBD_BRW_SYNC to guard this.
1131 Description: add quota statistics
1132 Details : 1. sort out quota proc entries and proc code.
1133 2. add quota statistics
1135 Severity : enhancement
1137 Description: enable quota support for HEAD.
1141 Description: Properly propagate oinfo flags from lov to osc for statfs
1142 Details : restore missing copy oi_flags to lov requests.
1144 Severity : enhancement
1146 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1147 : l_getgroups(8), lst(8), routerstat(8)
1148 Details : included man pages for llobdstat(8), llstat(8),
1149 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1151 Severity : enhancement
1153 Description: configure's --enable-quota should check the
1154 : kernel .config for CONFIG_QUOTA
1155 Details : configure is terminated if --enable-quota is passed but
1156 : no quota support is in kernel
1160 Description: Remove portals compatibility
1161 Details : Remove portals compatibility, not interoperable with releases
1166 Description: Resolve device initialization race
1167 Details : Prevent proc handler from accessing devices added to the
1168 obd_devs array but yet be intialized.
1170 Severity : enhancement
1172 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1174 Severity : enhancement
1176 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1181 Description: Using wrong pointer in osc_brw_prep_request
1182 Details : Access to array[-1] can produce panic if kernel compiled with
1183 CONFIG_PAGE_ALLOC enabled
1187 Description: Client runs out of low memory
1188 Details : Consider only lowmem when counting initial number of llap pages
1192 Description: *optional* service tags registration
1193 Details : if the "service tags" package is installed on a Lustre node
1194 When the filesystem is mounted, a local-node service tag will
1195 be created. See http://inventory.sun.com/ for more information
1196 about the Service Tags asset management system.
1200 Description: Kernel BUG tries to release flock
1201 Details : Lustre does not destroy flock lock before last reference goes
1202 away. So always drop flock locks when client is evicted and
1203 perform unlock regardless of successfulness of speaking to MDS.
1207 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1211 Description: sanity test 65a fails if stripecount of -1 is set
1212 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1217 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1218 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1221 Severity : enhancement
1223 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1227 Description: do not process already freed flock
1228 Details : flock can possibly be freed by another thread before it reaches
1229 to ldlm_flock_completion_ast.
1233 Description: LBUG during stress test
1234 Details : Need properly lock accesses the flock deadlock detection list.
1238 Description: oops in page fault handler
1239 Details : kernel page fault handler can return two special 'pages' in error
1240 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1244 Description: timeout with invalidate import.
1245 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1246 should be handled by ptlrpcd. This produce long age waiting and
1247 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1249 Severity : enhancement
1251 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1255 Description: enable MGS and MDT services start separately
1256 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1257 a co-located MGS without starting the MGS, which is a complement
1258 to 'nosvc' mount option.
1262 Frequency : after recovery
1263 Description: precreate to many object's after del orphan.
1264 Details : del orphan st in oscc last_id == next_id and this triger growing
1265 count of precreated objects. Set flag LOW to skip increase count
1266 of precreated objects.
1270 Frequency : rare, on clear nid stats
1271 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1272 Details : when clean nid stats sometimes try destroy live entry,
1273 and this produce panic in free.
1277 Description: Stack overflow during MDS log replay
1278 ease stack pressure by using a thread dealing llog_process.
1282 Description: wait until IO finished before start new when do lock cancel.
1283 Details : VM protocol want old IO finished before start new, in this case
1284 need wait until PG_writeback is cleared until check dirty flag and
1285 call writepages in lock cancel callback.
1287 Severity : enhancement
1289 Description: using special macro for print time and cleanup in includes.
1293 Description: mds_mfd_close() ASSERTION(rc == 0)
1294 Details : In mds_mfd_close(), we need protect inode's writecount change
1295 within its orphan write semaphore to prevent possible races.
1299 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1303 Frequency : rare, on shutdown ost
1304 Description: don't hit live lock with umount ost.
1305 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1306 use shrink_dcache_sb instead.
1310 Description: don't panic with use echo client
1311 Details : echo client pass NULL as client nid pointer and this produce null
1312 pointer dereference.
1316 Description: fix build on ppc32
1317 Details : compile code with -m64 flag produce wrong object file for ppc32.
1321 Description: add message levels for liblustreapi
1325 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1326 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1327 will result in an -EAGAIN and in the exit call path we call
1328 alloc_rr() which will with increasing aggressiveness attempt to
1329 aquire precreated objects on the minimum number of required OSCs.
1333 Description: Use old size assignment to avoid deadlock
1334 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1335 the scheduling while holding a spinlock. We do not need locking
1336 for size in ll_update_inode() because size is only updated from
1337 the MDS for directories or files without objects, so there is no
1338 other place to do the update, and concurrent access to such inodes
1339 are protected by the inode lock.
1343 Description: resolve "_IOWR redefined" build error on SLES10
1347 Description: dump the memory debugging after all modules are unloaded to
1348 suppress false negative in conf_sanity test 39
1350 Severity : enhancement
1352 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1357 Description: fix SLES kernel versioning
1358 Details : the kernel version for our SLES 10 kernel did not include a "-"
1359 before the "smp" at the end. while this was not a problem in
1360 general, it did mean that software trying to use the kernel
1361 version to try to detect a vendor specific kernel would fail.
1362 this was most evident by the OFED build scripts.
1366 Description: Don't update lov_desc members until making sure they are valid
1367 Details : When updating lov_desc members via proc fs, need fix their
1368 validities before doing the real update.
1372 Description: don't put request into delay list while invalidate in flight.
1373 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1374 import in flight. this produce timeout for invalidate and sometimes
1375 can cause stale data.
1379 Frequency : on ppc only
1380 Description: not convert ost objects for directory because it's not exist.
1381 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1385 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1387 Severity : enhancement
1389 Description: Update to support 2.6.22.14 vanilla kernel.
1393 Frequency : rare, on recovery
1394 Description: read procfs can produce deadlock in some situation
1395 Details : Holding lprocfs lock which send rpc can produce block for destroy
1396 obd objects and this also block reconnect with -EALREADY. This
1397 isn't fix all lprocfs bugs - but make it rare.
1399 Severity : enhancement
1401 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1404 Frequency : frequent on X2 node
1406 Description: mdc_set_open_replay_data LBUG
1407 Details : Set replay data for requests that are eligible for replay.
1411 Description: lustre_mgs: operation 101 on unconnected MGS
1412 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1413 MGC, and late on MGC cannot successfully connect to MGS and a lot
1414 of the error messages complaining that MGS is not connected.
1418 Frequency : on network error
1419 Description: panic with double free request if network error
1420 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1421 but it's true only for synchronus enqueue, for async enqueue
1422 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1425 Severity : enhancement
1427 Description: client-side metadata stat-ahead during readdir(directory readahead)
1428 Details : perform client-side metadata stat-ahead when the client detects
1429 readdir and sequential stat of dir entries therein
1432 Frequency : on start mds
1434 Description: Implement get_info(last_id) in obdfilter.
1437 Frequency : occasional
1439 Description: Correctly check stale fid, not start epoch if ost not support SOM
1440 Details : open with flag O_CREATE need set old fid in op_fid3 because
1441 op_fid2 was overwritten with new generated fid, but mds can answer
1442 with one of these two fids and both is not stale. Setattr
1443 incorrectly started an epoch and assume will be called
1444 done_writing, but without SOM done_writing ever being called.
1447 Frequency : rare, depends on device drivers and load
1449 Description: MDS or OSS nodes crash due to stack overflow
1450 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1451 In some cases, in conjunction with device drivers that use a lot
1452 of stack the MDS (or possibly OSS) service threads could overflow
1453 the stack. One change which was identified to consume additional
1454 stack has been reworked to avoid the extra stack usage.
1457 Frequency : occasional
1459 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1460 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1461 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1462 oscc_has_objects and see OSC has no objests and call
1463 oscc_internal_create to resend create request.
1465 Severity : enhancement
1467 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1469 Severity : enhancement
1471 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1474 Frequency : very rare
1476 Description: Fix replay if there is an un-replied request and open
1477 Details : In some cases, older replay request will revert the
1478 mcd->mcd_last_xid on MDS which is used to record the client's
1479 latest sent request.
1481 Severity : enhancement
1483 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1485 Severity : enhancement
1487 Description: Add rhel5 support to HEAD.
1489 Serverity : enhancement
1491 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1496 Description: Don't allow skipping OSTs if index has been specified.
1497 Details : Don't allow skipping OSTs if index has been specified, make
1498 locking in internal create lots better.
1502 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1503 Details : access to bitfield in structure is always rounded to long
1504 and this produce problem with not atomic change any bit.
1508 Description: Lustre make rpms failed.
1509 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1510 builds Lustre rpms from tarball.
1513 Frequency : rare on shutdown ost
1515 Description: If llog cancel was not send before clean_exports phase, this can
1516 produce deadlock in llog code.
1517 Details : If llog thread has last reference to obd and call class_import_put
1518 this produce deadlock because llog_cleanup_commit_master wait when
1519 last llog_commit_thread exited, but this never success because was
1520 called from llog_commit_thread.
1524 Description: allow userland application know is lost one of stripes.
1525 Details : fill lvb_blocks with error code on ost and return it to
1526 application if error flag found.
1530 Description: NULL lov_tgts causing MDS oops
1531 Details : more safe checks for NULL lov_tgts for avoid oops.
1533 Severity : enhancement
1535 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1539 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1540 Details : If an OST is down the MDS will hang indefinitely in
1541 obd_statfs_rqset() waiting for the statfs data. While for
1542 MDS QOS usage of statfs, it should not stuck in waiting.
1544 Severity : enhancement
1546 Description: remote_acl support
1547 Details : Support ACL-based permission check for remote user.
1548 Support setfacl/getfacl for remote user with the utils
1549 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1550 the system "{s,g}etfacl" utils.
1552 Severity : enhancement
1554 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1556 Severity : enhancement
1558 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1562 Description: make mgs_setparam() handle fsname containing dash
1563 Details : fsname containing a dash does not work with lctl conf_param
1565 Severity : enhancement
1567 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1569 Severity : enhancement
1571 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1575 Frequency : rare, at shutdown
1576 Description: access already free / zero obd_namespace.
1577 Details : if client_disconnect_export was called without force flag set,
1578 and exist connect request in flight, this can produce access to
1579 NULL pointer (or already free pointer) when connect_interpret
1580 store ocd flags in obd_namespace.
1584 Frequency : only at startup
1585 Description: not alloc memory with spinlock held.
1586 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1587 if any spinlock held.
1589 Severity : enhancement
1591 Description: make lustre randomly fail allocating memory
1592 Details : Make lustre randomly failed allocating memory for testing purpose.
1594 Severity : enhancement
1596 Description: lost problems with lov objid file
1597 Details : Fixes some scability and access to not inited memory problems
1598 in work with lov objdid file.
1603 Description: lfs find does not continue on file error
1604 Details : Continue other files processing when a file/dir is absent.
1608 Description: Inconsistent usage of lustre_pack_reply()
1609 Details : Standardize the usage of lustre_pack_reply() such that it
1610 always generate a CERROR on failure.
1613 Frequency : occasional
1615 Description: MDS hang or stay in waiting lock
1616 Details : If client receive lock with CBPENDING flag ldlm need send lock
1617 cancel as separate rpc, to avoid situation when cancel request
1618 can't processed due all i/o threads stay in wait lock.
1622 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1626 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1630 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1634 Description: Update to RHEL4 latest kernel.
1636 Severity : enhancement
1638 Description: Build SLES10 patchless client fails
1639 Details : The configure was broken by run ./configure with
1640 --with-linux-obj=.... argument for patchless client. When the
1641 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1642 can't search header adequately. Use absolute path such as
1643 -I($LINUX)/include instead.
1647 Description: interrupt oig_wait produce painc on resend.
1648 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1649 private set, and this produce situation when rq_ptlrpcd_data not
1650 copyed to new allocated request and triggered LBUG on assert
1651 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1652 ptlrpcd set if request is changed and can be safety changed to use
1657 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1658 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1659 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1660 these macros and replace them with their difinitions instead.
1664 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1665 Details : Update which_patch & target file for SLES10 latest kernel.
1668 Description: Remove client patches from SLES 10 kernel.
1669 Details : This causes SLES 10 clients to behave as patchless clients
1670 even on a Lustre-patched (server) kernel.
1672 Severity : enhancement
1674 Description: self-adjustable client's lru lists
1675 Details : use adaptive algorithm for managing client cached locks lru
1676 lists according to current server load, other client's work
1677 pattern, memory activities, etc. Both, server and client
1678 side namespaces provide number of proc tunables for controlling
1681 Severity : enhancement
1683 Description: light-weight GSS support
1684 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1685 and reduce performance overhead.
1687 Severity : enhancement
1689 Description: Linux keyring support
1690 Details : Support using service of Linux keyring for Lustre GSS internal
1691 context refresh/cache mechanism.
1695 Description: Fix errors in lfs documentation
1696 Details : Fixes man pages
1700 Description: don't use GFP_* in generic Lustre code.
1701 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1705 Description: obdclass is limited by single OBD_ALLOC(idarray)
1706 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1710 Description: warnings with build patchless client with vanila 2.6.19 and up
1711 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1712 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1716 Frequency : Only for SLES
1717 Description: sanity_quota fail test_1
1718 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1719 file, which makes TSTID[2] inunique.
1723 Frequency : Only for SLES9
1724 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1725 Details : Update target/ChangeLog/which_patch .
1729 Frequency : Only for test_10 in sanity-quota.sh
1730 Description: a bug in quota test code
1731 Details : checking a test flag in a wrong place
1735 Frequency : Only for test_13 in sanity-quota.sh
1736 Description: a bug in quota test script
1737 Details : shouldn't check overfull allocated quota
1741 Description: jbd statistics
1742 Details : Port older jbd statistics patch for sles10
1744 Severity : enhancement
1746 Description: Add printing inode info into message about error in writepage.
1751 Description: reply_lock_interpret crash due to race with it and lock cancel.
1752 Details : Do not replay locks that are being cancelled. Do not reference
1753 locks by their address during replay, just by their handle.
1757 Frequency : When flocks are used.
1758 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1759 Details : Flock locks might destroy just granted lock if it could be merged
1760 with another existing flock, this is done in completion handler,
1761 so teach ldlm_cli_enquque_fini that this is a valid case for
1767 Description: Oops in read and write path when failing to allocate lock.
1768 Details : Check if lock allocation failed and return error back.
1772 Description: lstripe command fails for valid OST index
1773 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1774 instead of lov->desc.ld_active_tgt_count.
1778 Description: sanity.sh failed test 103
1779 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1784 Description: df doesn't work properly if diskfs blocksize != 4K
1785 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1789 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1790 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1792 Severity : enhancement
1794 Description: Adaptive timeouts
1795 Details : RPC timeouts adapt to changing server load and network
1796 conditions to reduce resend attempts and improve recovery time.
1800 Description: llapi_file_create() does not allow some changes
1801 Details : add llapi_file_open() that allows specifying the mode and
1802 open flags, and also returns an open file handle.
1806 Description: merge and cleanup kernel patches.
1807 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1811 Description: Add journal checksum support.(Kernel part)
1812 Details : The journal checksum feature adds two new flags i.e
1813 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1814 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1815 indicates that the commit block contains the checksum for
1816 the blocks described by the descriptor blocks. Now commit
1817 record can be sent to disk without waiting for descriptor
1818 blocks to be written to disk. This behavior is controlled
1819 using JBD2_FEATURE_ASYNC_COMMIT flag.
1823 Description: OSS needs mutliple precreate threads
1824 Details : Add ability to start more than one create thread per OSS.
1828 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1829 Details : execute lfs setstripe on client
1833 Description: mds_obd_create error creating tmp object
1834 Details : When the user sets quota on root, llog will be affected and can't
1835 create files and write files.
1838 Frequency : Always on ia64 patchless client, and possibly others.
1840 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1841 Details : This allows the patchless client to be loaded on architectures
1842 without this export.
1846 Description: RedHat Update kernel for RHEL5
1847 Details : Add the kernel config file for RHEL5.
1851 Description: RedHat Update kernel for RHEL5
1852 Details : Modify the kernel config file more closer RHEL5.
1856 Description: Build failure against Centos5 (RHEL5)
1857 Details : Define PAGE_SIZE when it isn't present.
1861 Description: test_7 on sanity-quota.sh can't be used on separate machines
1862 Details : add facet to handle it
1866 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1867 Details : with NFS, the anon dentry's parent was set to itself in
1868 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1869 corresponding dentry other than use rec->ur_name.
1871 Severity : enhancement
1873 Description: lfs setstripe enhancement
1874 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1878 Description: enable data checksumming by default
1879 Details : enable checksum by default, allow --disable-checksum
1880 configure option and "-o nochecksum" mount option. Checksums
1881 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1885 Description: lustre support for RHEL5
1886 Details : Add support for RHEL5.
1890 Description: Client eviction due to failover config
1891 Details : after a connection loss, the lustre client should attempt to
1892 reconnect to the last active server first before trying the
1893 other potential connections.
1897 Description: when mds and osts use different quota unit(32bit and 64bit),
1898 quota will be released repeatly.
1899 Details : void sending multiple quota reqs to mds, which will keep the
1900 status between the reqs.
1904 Description: osts not allocated evenly to files
1905 Details : change the condition to increase offset_idx
1908 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1910 Description: Data corruption for OSTs that are formatted larger than 2TB
1912 Details : When generating the bio request for lustre file writes the
1913 sector number would overflow a temporary variable before being
1914 used for the IO. The data reads correctly from Lustre (which
1915 will overflow in a similar manner) but other file data or
1916 filesystem metadata may be corrupted in some cases.
1920 Description: Tune the kernel for good SCSI performance.
1921 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1922 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1927 Description: rewrite ext2-derived code in obdclass/uuid.c
1928 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1929 from scratch preserving functionality.
1933 Description: "lfs find -obd UUID" prints directories
1934 Details : "lfs find -obd UUID" will return all directory names instead
1935 of just file names. It is incorrect because the directories
1936 do not reside on the OSTs.
1940 Description: use i_size_read and i_size_write in 2.6 port
1941 Details : replace inode->i_size access with i_size_read/write()
1944 Frequency : when removing large files
1946 Description: scheduling issue during removal of large Lustre files
1947 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1948 It causes scheduling issues when removing large files (17TB in the
1952 Frequency : only with liblustre clients on XT3
1954 Description: evictions taking too long
1955 Details : allow llrd to evict clients directly on OSTs
1958 Frequency : only on ppc
1960 Description: /proc/fs/lustre/devices broken on ppc
1961 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1962 We should make sure the type of 'index' is loff_t and then cast
1963 explicitly as needed below. Do not assign an explicitly cast
1967 Frequency : only for rhel5
1969 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1970 Details : Modify the target file & which_kernel.
1972 Severity : enhancement
1974 Description: omit set fsid for export NFS
1975 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1976 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1980 Frequency : Always, for kernels after 2.6.16
1981 Description: Fix warning idr_remove called for id=.. which is not allocated.
1982 Details : Last kernels save old s_dev before kill super and not allow
1983 to restore from callback - restore it before call kill_anon_super.
1987 Description: buffer overruns could theoretically occur
1988 Details : llapi_semantic_traverse() modifies the "path" argument by
1989 appending values to the end of the origin string, and a
1990 overrun may occur. Adding buffer overrun check in liblustreapi.
1994 Description: Fix error on 'ls .' at the top of the Lustre mount.
1995 Details : Don't revalidate dentry if it is a root dentry.
1999 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2000 Details : Modify vm-tunables-rhel4.patch.
2004 Description: Kernel config for 2.6.18-vanilla.
2005 Details : Modify targets/2.6-vanilla.target.in.
2006 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2007 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2008 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2009 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2013 Description: improve handling recoverable errors
2014 Details : if request processig with error which can be recoverable on server
2015 request should be resend, otherwise page released from cache and
2020 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2021 Details : Modify target file & which_patch.
2022 A flaw was found in the IA32 system call emulation provided
2023 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2024 value could be stored in the %RAX register, which could trigger an
2025 out-of-bounds system call table access. An untrusted local user
2026 could exploit this flaw to run code in the kernel
2027 (ie a root privilege escalation). (CVE-2007-4573).
2031 Description: O_DIRECT bypasses client statistics.
2032 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2033 read_ahead_stats were updated. copied the stats section from
2034 osc_send_oap_rpc() into async_internal().
2038 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2042 Description: change order of libsysio includes
2043 Details : '#include sysio.h' should always come before '#include xtio.h'
2047 Description: handle "serious error: objid * already exists" more gracefully
2048 Details : If LAST_ID value on disk is smaller than the objects existing in
2049 the O/0/d* directories, it indicates disk corruption and causes an
2050 LBUG(). If the object is 0-length, then we should use the existing
2051 object. This will help to avoid a full fsck in most cases.
2053 Severity : enhancement
2055 Description: adapt the lustre_config script to support the upgrade case
2056 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2057 server targets to 1.6 in parallel.
2061 Description: To avoid grant space > avaible space when the disk is almost
2062 full. Without this patch you might see the error "grant XXXX >
2063 available" or some LBUG about grant, when the disk is almost
2065 Details : In filter_check_grant, for non_grant cache write, we should
2066 check the left space by if (*left > ungranted + bytes), instead
2067 of (*left > ungranted), because only we are sure the left space
2068 is enough for another "bytes", then the ungrant space should be
2069 increase. In client, we should update cl_avail_grant only there
2070 is OBD_MD_FLGRANT in the reply.
2074 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2075 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2076 described in CVE-2007-4573. This problem could allow untrusted
2077 local users to gain root access.
2080 Frequency : when using O_DIRECT and quotas
2082 Description: Incorrect file ownership on O_DIRECT output files
2083 Details : block usage reported by 'lfs quota' does not take into account
2084 files that have been written with O_DIRECT.
2089 Description: touch file failed when fs is not full
2090 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2091 otherwise we can get ENOSP while fs is not full.
2095 Description: parallel lock callbacks
2096 Details : Instead of sending blocking and completion callbacks as separated
2097 requests, adding them to a set and sending in parallel.
2100 Frequency : only for Cray XT3
2101 Bugzilla : 12829/13455
2102 Description: Changing primary group doesn't change the group lustre assigns to
2104 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2105 overridden with the primary group provided by the group upcall,
2106 whereas the supplied fsgid can be trusted if it is in the list of
2107 supplementary groups returned by the group upcall.
2109 Severity : enhancement
2111 Description: Allow masking D_WARNING, D_ERROR messages from console
2112 Details : Console messages can now be disabled via lnet.printk.
2116 Description: User code with malformed file open parameter crashes client node
2117 Details : Before packing join_file req, all the related reference should be
2118 checked carefully in case some malformed flags cause fake
2119 join_file req on client.
2123 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2124 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2125 mode is equal to the granted mode. However, it should also check
2126 that LDLM_FL_CP_REQD is not set.
2130 Description: Heavy nfs access might result in deadlocks
2131 Details : After ELC code landed, it is now improper to enqueue any mds
2132 locks under och_sem, because enqueue might want to decide to
2133 cancel open locks for same inode we are holding och_sem for.
2137 Description: Client eviction while running blogbench
2138 Details : A lot of unlink operations with concurrent I/O can lead to a
2139 deadlock causing evictions. To address the problem, the number of
2140 oustanding OST_DESTROY requests is now throttled to
2141 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2142 callbacks are processed in priority.
2146 Description: enable ACLs on MDS by default
2147 Details : ACLs must be enabled on MDS by default.
2150 Frequency : PPC/PPC64 only
2152 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2153 Details : fix duplicated definitions between asm-ppc64/types.h and
2154 lustre_types.h on PPC.
2157 Frequency : PPC/PPC64 only
2159 Description: asm-ppc/segment.h does not exist
2160 Details : fix compile issue on PPC.
2164 Description: better handle error messages in extents code
2167 Frequency : RHEL4 only
2169 Description: mkfs is very slow on IA64/RHEL4
2170 Details : A performance regression has been discovered in the MPT Fusion
2171 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2172 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2173 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2175 Severity : enhancement
2177 Description: SNMP support enhancement
2178 Details : Adding total number of sampled request for an MDS node in snmp
2181 Severity : enhancement
2183 Description: Optimize ldlm waiting list processing for PR extent locks
2184 Details : When processing waiting list for read extent lock and meeting read
2185 lock that is same or wider to it that is not contended, skip
2186 processing rest of the list and immediatelly return current
2187 status of conflictness, since we are guaranteed there are no
2188 conflicting locks in the rest of the list.
2192 Description: Time out and refuse to reconnect
2193 Details : When the failover node is the primary node, it is possible
2194 to have two identical connections in imp_conn_list. We must
2195 compare not conn's pointers but NIDs, otherwise we can defeat
2196 connection throttling.
2200 Description: port llog fixes from b1_6 into HEAD
2201 Details : Port llog reference couting and some llog cleanups from b1_6
2202 (bug 10800) into HEAD, for protect from panic and access to already
2203 free llog structures.
2207 Description: Detect stride IO mode in read-ahead
2208 Details : When a client does stride read, read-ahead should detect that and
2209 read-ahead pages according to the detected stride pattern.
2213 Description: data checksumming impacts single node performance
2214 Details : add support for several checksum algorithm. Currently, only CRC32
2215 and Adler-32 are supported. The checksum type can be changed on
2216 the fly via /proc/fs/lustre/osc/*/checksum_type.
2220 Description: use adler32 for page checksums
2221 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2226 Description: build for x2 fails
2227 Details : fix compile issue on Cray systems.
2231 Description: Properly match for duplicate locks
2232 Details : Due to different lock order from skiplists code, we need to
2233 traverse entire list for now
2236 Frequency : only on PPC/SLES10
2238 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2239 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2240 be defined. Add a hack in mkfs_lustre.c to work around this compile
2245 Description: LASSERT on MDS when client holding flock lock dies
2246 Details : ldlm pool logic depends on number of granted locks equal to
2247 number of released locks which is not true for flock locks, so
2248 just exclude such locks from consideration.
2252 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2253 Details : Use fsfilt_sync() for both the whole filesystem sync and
2254 individual file sync to eliminate dangerous inode locking
2255 with I_LOCK that can lead to a deadlock.
2259 Description: performance in 1.6.3
2260 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2261 sure that 1MB requests are not fragmented by the block layer.
2263 Severity : enhancement
2265 Description: organize the server-side client stats on per-nid basis
2266 Details : Change the structure of stats under obdfilter and mds to
2274 The "uuid"s file would list the uuids of _active_ exports.
2275 And the clear entry is to clear all stats and stale nids.
2277 Severity : enhancement
2279 Description: eliminate client locks in face of contention
2280 Details : file contention detection and lockless i/o implementation
2281 for contended files.
2285 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2288 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2290 Description: MDS may lose file striping (and hence file data) in some cases
2291 Details : If there are additional extended attributes stored on the MDS,
2292 in particular ACLs, SELinux, or user attributes (if user_xattr
2293 is specified for the client mount options) then there is a risk
2294 of attribute loss. Additionally, the Lustre file striping
2295 needs to be larger than default (e.g. striped over all OSTs),
2296 and an additional attribute must be stored initially in the
2297 inode and then increase in size enough to be moved to the
2298 external attribute block (e.g. ACL growing in size) for file
2303 Description: skiplist implementation simplification
2304 Details : skiplists are used to group compatible locks on granted list
2305 that was implemented as tracking first and last lock of each
2306 lock group the patch changes that to using doubly linked lists
2310 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2311 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2312 because LASSERT on both the data supplied by a client, and the
2313 data on disk is dangerous and incorrect.
2315 Severity : enhancement
2317 Description: Slow truncate/writes to huge files at high offsets.
2318 Details : Directly associate cached pages to lock that protect those pages,
2319 this allows us to quickly find what pages to write and remove
2320 once lock callback is received.
2324 Description: more ldlm soft lockups
2325 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2326 starve other threads from the resource lock for a long time in
2327 case of long waiting queue, so change the debug level from
2328 D_OTHER to the less frequently used D_INFO.
2330 Severity : enhancement
2332 Description: add -gid, -group, -uid, -user options to lfs find
2336 Description: Hung threads in invalidate_inode_pages2_range
2337 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2338 once one is completed. As a result, some RPCs are stuck in the
2339 queue and are never sent.
2343 Description: filter threads hungs on waiting journal commit
2344 Details : Cleanup filter group llog code, then only filter group llog will
2345 be only created in the MDS/OST syncing process.
2349 Description: Procfs and llog threads access destoryed import sometimes.
2350 Details : Sync the import destoryed process with procfs and llog threads by
2351 the import refcount and semaphore.
2353 Severity : enhancement
2355 Description: openlock cache of b1_6 port to HEAD
2360 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2361 Details : If insertion of an extent fails, then discard the inode
2362 preallocation and free data blocks else it can lead to duplicate
2367 Description: don't always update ctime in ext3_xattr_set_handle()
2368 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2369 In some cases the ctime should not be updated, for example for
2370 2.0->1.8 compatibility it is necessary to delete an xattr and it
2371 should not update the ctime.
2375 Bugzilla : 15713/16362
2376 Description: Assertion in iopen_connect_dentry in 1.6.3
2377 Details : looking up an inode via iopen with the wrong generation number can
2378 populate the dcache with a disconneced dentry while the inode
2379 number is in the process of being reallocated. This causes an
2380 assertion failure in iopen since the inode's dentry list contains
2381 both a connected and disconnected dentry.
2385 Description: assertion failure in ldlm_handle2lock()
2386 Details : fix a race between class_handle_unhash() and class_handle2object()
2387 introduced in lustre 1.6.5 by bug 13622.
2392 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2393 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2398 Description: oops in sd_iostats_seq_show()
2399 Details : unloading/reloading the scsi low level driver triggers a kernel
2400 bug when trying to access the sd iostat file.
2405 Description: Kernel panics during QLogic driver reload
2406 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2407 patch, causing memory corruption.
2412 Description: journal_dev option does not work in b1_6
2413 Details : pass mount option during pre-mount.
2415 Severity : enhancement
2417 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2418 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2419 extent information of a file. It can be used to map logical blocks
2420 in a file to physical blocks in the block device.
2424 Description: LDLM soft lockups - improvement
2425 Details : It is be possible to send the lock handle along with each read
2426 or write request because the client is already doing a lock match
2427 itself so there isn't any reason the OST should have to re-do that
2431 Frequency : only on Cray X2
2433 Description: X2 build failures
2434 Details : fix build failures on Cray X2.
2438 Description: xid & resent requests
2439 Details : Initialize RPC XID from clock at startup (randomly if clock is
2442 Severity : enhancement
2444 Description: Add lustre_start utility to start or stop multiple Lustre servers
2449 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2450 Details : In case of memory pressure, list_del() can be called twice on
2451 req->rq_history_list, causing a kernel oops.
2455 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2456 Details : fix stack overflow in the distributed lock manager by defering
2457 export eviction after a failed AST to the elt thread instead of
2458 handling it in the dlm interpret routine.
2462 Description: Convert some comments to new format.
2463 Details : Update documenting comments to match doxygen conventions.
2467 Description: Grammar fixes.
2468 Details : A couple of trivial spelling fixes.
2472 Description: OSD_COUNTERS-mandatory
2473 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2474 to simplify the code.
2478 Description: simplify cmm_device freeing logic.
2479 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2483 Description: Add lockdep support to dt_object_operations locking interface.
2484 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2485 parameter indicating lock ordering. Update mdd code to use new
2490 Description: Introduce failloc constants for lockless IO tests.
2491 Details : Add two new failloc constants to test lockless IO. Only one of
2492 them in implemented---another is checked in yet to be landed
2497 Description: Add lockdep support for inode mutex.
2498 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2499 in the situations where Lustre has to lock more than one inode
2504 Description: Add optional invariants checking support.
2505 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2506 configure switch. This macro is to be used for consistency and
2507 sanity checks that are too expensive to be left in `production'
2512 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2513 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2518 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2519 Details : Use spin_lock_nested() in (the only) situation where more than
2520 one ldlm_lock is locked simultaneously. Also, fix possible
2521 dead-lock in ldlm_lock_change_resource() by enforcing particular
2526 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2527 Details : Instead of specifying each ldlm_lock call-back through separate
2528 parameter, wrap them into struct ldlm_callback_suite.
2532 Description: Kill join_lru obd method and its callers.
2533 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2534 in memory---a requirement implemented through ->o_join_lru() obd
2535 method in HEAD. Remove this method and its users.
2539 Description: Add asynchronous ldlm ENQUEUE completion handler.
2540 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2541 case that stock ldlm_completion_ast() cannot handle as it waits
2542 until lock is granted. Introduce new ldlm_completion_ast_async()
2543 for this. Also comment ldlm_completion_ast().
2547 Description: ldlm_error <-> errno conversion.
2548 Details : Add functions to map (rather arbitrary) between LDLM error codes
2549 and standard errno values. CLIO needs this to prevent LDLM specific
2550 constants from escaping ldlm and osc.
2554 Description: Kill unused ldlm_handle2lock_ns() function.
2555 Details : Kill unused ldlm_handle2lock_ns() function.
2559 Description: Add lu_ref support to ldlm_lock
2560 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2561 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2562 and ldlm_resource. LDLM interface has to be changed, because code that
2563 releases a reference on a lock, has to "know" what reference this is.
2564 In the most frequent case
2566 lock = ldlm_handle2lock(handle);
2568 LDLM_LOCK_PUT(lock);
2570 no changes are required. When any other reference (received _not_ from
2571 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2572 instead of LDLM_LOCK_PUT().
2574 Arguably, changes are pervasive, and interface requires some discipline
2575 for proper use. On the other hand, it was very instrumental in finding
2576 a few leaked lock references.
2580 Description: Add ldlm_lock_addref_try().
2581 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2582 attempts to addref a lock that might be being canceled
2587 Description: Add ldlm_weigh_callback().
2588 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2589 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2590 hard-coded `number of pages' logic.
2594 Description: Add lockdep annotations to llog code.
2595 Details : Use appropriately tagged _nested() locking calls in the places
2596 where llog takes more than one ->lgh_lock lock.
2600 Description: Add loi_kms_set().
2601 Details : Wrap kms updates into a helper function.
2605 Description: Constify instances of struct lsm_operations.
2606 Details : Constify instances of struct lsm_operations.
2610 Description: lu_conf support.
2611 Details : On a server, a file system object is uniquely identified
2612 by a fid, which is sufficient to locate and load all object
2613 state (inode). On a client, on the other hand, more data are
2614 necessary instantiate an object. Change lu_object_find() and
2615 friends to take additional `lu_conf' argument describing object.
2616 Typically this includes layout information.
2620 Description: lu_context fixes.
2621 Details : Introduce new lu_context functions that are needed on the client
2622 side, where some system threads (ptlrpcd) are shared by multiple
2623 modules, and so cannot be stopped during module shutdown.
2627 Description: Add start and stop methods to lu_device_type_operations.
2628 Details : Introduce two new methods in lu_device_type_operations, that are
2629 invoked when first instance of a given type is created and last one
2630 is destroyed respectively. This is need by CLIO.
2634 Description: Add lu_ref support to struct lu_device.
2635 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2636 track leaked references.
2640 Description: Introduce lu_kmem_descr.
2641 Details : lu_kmem_descr and its companion interface allow to create
2642 and destroy a number of kmem caches at once.
2646 Description: Fix lu_object finalization race.
2647 Details : Fix a race between lu_object_find() finding an object and its
2648 concurrent finalization. This race is (most likely) not possible
2649 on the server, but might happen on the client.
2653 Description: Introduce lu_ref interface.
2654 Details : lu_ref is a debugging module allowing to track references to
2655 a given object. It is quite cpu expensive, and has to be
2656 explicitly enabled with --enable-lu_ref. See usage description
2661 Description: Factor lu_site procfs stats into a separate function.
2662 Details : Separate lu_site stats printing code into a separate function
2663 that can be reused on a client.
2667 Description: Constify instances of struct {lu,dt,md}_device_operations.
2668 Details : Constify instances of struct {lu,dt,md}_device_operations.
2672 Description: Introduce struct md_site and move meta-data specific parts of
2673 struct lu_site here.
2674 Details : Move md-specific fields out of struct lu_site into special struct
2675 md_site, so that lu_site can be used on a client.
2679 Description: Kill mdd_lov_destroy().
2680 Details : Remove unused mdd code.
2684 Description: Add st_block checking to multistat.c.
2685 Details : Add st_block checking to multistat.c.
2689 Description: Add lu_ref support to struct obd_device.
2690 Details : Add lu_ref tracking to obd_device.
2694 Description: Kill obd_set_fail_loc().
2695 Details : Remove unused code.
2699 Description: Add special type for ptlrpc_request interpret functions.
2700 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2701 there. Actual usage will be in CLIO.
2705 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2706 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2707 doesn't work with lockdep.
2711 Description: Add rwv.c test program.
2712 Details : New testing program exercising readv(2) and writev(2) (Qian).
2716 Description: Add sendfile.c test program.
2717 Details : New testing program exercising sendfile(2) (Jay).
2721 Description: Ratelimit a message that can be very frequent.
2722 Details : Ratelimit a memory allocation failure message that can
2727 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2728 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2732 Description: Liblustre build fixes.
2733 Details : Liblustre build fixes.
2737 Description: libcfs: add cfs_{need,cond}_resched() interface.
2738 Details : libcfs: add cfs_{need,cond}_resched() definition and
2739 implementations for Linux, NT, and liblustre.
2741 Severity : enhancement
2743 Description: More exported tunables for mballoc
2744 Details : Add support for tunable preallocation window and new tunables for
2745 large/small requests
2749 Description: Detect corruption of block bitmap and checking for preallocations
2750 Details : Checks validity of on-disk block bitmap. Also it does better
2751 checking of number of applied preallocations. When corruption is
2752 found, it turns filesystem readonly to prevent further corruptions.
2756 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2757 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2758 window to 0. If it does hit cache miss, and read-pattern is still
2759 stride-io mode, does not reset the stride window, but also does not
2760 increase the stride window length in this case.
2764 Frequency : only for big-endian servers
2765 Description: Check if system is big-endian while mounting fs with extents feature
2766 Details : Mounting a filesystem with extents feature will fail on big-endian
2767 systems since ext3-based ldiskfs is not supported on big-endian
2768 systems. This can be over-riden with "bigendian_extents" mount option.
2770 Severity : enhancement
2772 Description: The root squash functionality
2773 Details : A security feature, which is to prevent users from being able
2774 to mount lustre on their desktop, run as root, and delete
2775 all of the files in the filesystem. The goal is accomplished by
2776 remapping user id (UID) and group id (GID) of the root user to
2777 a UID and GID specified by the system administartor via Lustre
2778 configuration management server (MGS). The functionality also
2779 allows to specify sets of clients for which the remapping does
2784 Description: Excessive recovery window
2785 Details : With AT enabled, the recovery window can be excessively long (6000+
2786 seconds). To address this problem, we no longer use
2787 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2788 timeout no longer depends on the service time, it is set to
2789 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2790 time via pb_service_time.
2794 Description: Watchdog triggered on MDS failover
2795 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2796 the OSTs know that the MDS "UUID" can be reused for the same export
2797 from a different NID, so we do not need to wait for the export to be
2801 Frequency : rare, only if using MMP with Linux RAID
2803 Description: MMP doesn't work with Linux RAID
2804 Details : While using HA for Lustre servers with Linux RAID, it is possible
2805 that MMP will not detect multiple mounts. To make this work we
2806 need to unplug the device queue in RAID when the MMP block is being
2807 written. Also while reading the MMP block, we should read it from
2808 disk and not the cached one.
2810 Severity : enhancement
2812 Description: open file using fid
2813 Details : A file can be opened using just its fid, like
2814 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2817 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2820 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2821 Details : add proper locking for creation of jbd_4k slab cache
2825 Description: MMP check in ext3_remount() fails without displaying any error
2826 Details : When multiple mount protection fails during remount, proper error
2829 Severity : enhancement
2831 Description: Allow stripe size to be up to 4G-64k
2832 Details : Fix math logic to allow large stripe sizes.
2836 Description: add check for >8TB ldiskfs filesystems
2837 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2838 Don't allow >8TB ldiskfs filesystems to be mounted without
2839 force_over_8tb mount option
2844 Descriptoin: OSS complains "too many blocks" and crashes
2845 Details : In filter_commitrw_write():
2846 1. lquota_chkquota();
2847 2. obdo_from_inode();
2848 3. lquota_pending_commit();
2849 "2." has changed uid and gid in this case, then caused
2850 lquota_chkquota() and lquota_pending_commit() to use
2851 different uid and gid. That is the root of the bug.
2855 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2856 Details : Detect and print out the 'stripe count = -1' case properly.
2860 Description: Out or order replies might be lost on replay
2861 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2862 request with tid smaller than one currently being added, add it
2863 to the start, not end of the list.
2865 --------------------------------------------------------------------------------
2867 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2869 * Support for kernels:
2870 2.6.5-7.283 (SLES 9),
2871 2.6.9-55.EL (RHEL 4),
2872 2.6.16.46-0.14 (SLES 10),
2873 2.6.18.8 vanilla (kernel.org)
2874 * Client support for unpatched kernels:
2875 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2876 2.6.16 - 2.6.22 vanilla (kernel.org)
2877 * Due to recently discovered recovery problems, we do not recommend
2878 using patchless RHEL 4 clients with this or any earlier release.
2879 * Recommended e2fsprogs version: 1.39.cfs8
2880 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2881 * Starting with this release, the ldiskfs backing filesystem required
2882 by Lustre is now in its own package, lustre-ldiskfs. This package
2883 should be installed. It is versioned separately from Lustre and
2884 may be released separately in future.
2888 Description: block reactivating mgc import until all deactivates complete
2889 Details : Fix race when failing back MDT/MGS to itself (testing)
2892 Frequency : at statup only
2894 Description: mds_lov_synchronize race leads to various problems
2895 Details : simultaneous MDT->OST connections at startup can cause the
2896 sync to abort, leaving the OSC in a bad state.
2898 Severity : enhancement
2900 Description: add optional extra BUILD_VERSION info
2901 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2902 to override the lustre version.
2905 Frequency : 2.6.18 servers only
2907 Description: ll_kern_mount() doesn't release the module reference
2908 Details : The ldiskfs module reference count never drops down to 0
2909 because ll_kern_mount() doesn't release the module reference.
2914 Description: server LBUG when using old ost_num_threads parameter
2915 Details : Accept the old ost_num_threads parameter but warn that it
2916 is deprecated, and fix an off-by-one error that caused an LBUG.
2921 Description: Transient SCSI error results in persistent IO issue
2922 Details : iobuf->dr_error is not reinitialized to 0 between two
2926 Frequency : sometimes when underlying device returns I/O errors
2928 Description: OSTs not going read-only during write failures
2929 Details : OSTs are not remounted read-only when the journal commit threads
2930 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2931 instead of the ext3 wrappers.
2935 Description: poor connect scaling with increasing client count
2936 Details : Don't run filter_grant_sanity_check for more than 100 exports
2937 to improve scaling for large numbers of clients.
2940 Frequency : SLES10 only
2942 Description: sanity-quota.sh quotacheck failed: rc = -22
2943 Details : Quotas cannot be enabled on SLES10.
2946 Frequency : liblustre clients only
2948 Description: getdirentries does not give error when run on compute nodes
2949 Details : getdirentries does not fail when the size specified as an argument
2950 is too small to contain at least one entry
2952 Severity : enhancement
2954 Description: Add LNET router traceability for debug purposes
2955 Details : If a checksum failure occurs with a router as part of the
2956 IO path, the NID of the last router that forwarded the bulk data
2957 is printed so it can be identified.
2962 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2963 Details : Due to a race condition, liblustre clients were occasionally
2964 evicted incorrectly.
2966 Severity : enhancement
2968 Description: lfs setstripe use optional parameters instead of postional
2971 Severity : enhancement
2973 Description: Nanosecond timestamp support for ldiskfs
2974 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2975 resolution timestamps. There is not yet support for this at
2976 the Lustre filesystem level.
2979 Frequency : during server recovery
2981 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2982 Details : request with rq_no_resend flag not awake l_wait_event if they get
2986 Frequency : nfs export on patchless client
2988 Description: connectathon hang when test nfs export over patchless client
2989 Details : Disconnected dentry cannot be found with lookup, so we do not need
2990 to unhash it or make it invalid
2993 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2994 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2995 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3000 Description: open req refcounting wrong on reconnect
3001 Details : If reconnect happened between getting open reply from server and
3002 call to mdc_set_replay_data in ll_file_open, we will schedule
3003 replay for unreferenced request that we are about to free.
3004 Subsequent close will crash in variety of ways.
3005 Check that request is still eligible for replay in
3006 mdc_set_replay_data().
3011 Description: disable writes to filesystem when reading health_check file
3012 Details : the default for reading the health_check proc file has changed
3013 to NOT do a journal transaction and write to disk, because this
3014 can cause reads of the /proc file to hang and block HA state
3015 checking on a healthy but otherwise heavily loaded system. It
3016 is possible to return to the previous behaviour during configure
3017 with --enable-health-write.
3019 Severity : enhancement
3021 Description: 64-bit inode version
3022 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3023 to the inode. This will be required for version-based recovery.
3028 Description: MDS fails to start if a duplicate client export is detected
3029 Details : in some rare cases it was possible for a client to connect to
3030 an MDS multiple times. Upon recovery the MDS would detect this
3031 and fail during startup. Handle this more gracefully.
3033 Severity : enhancement
3035 Description: Add -o localflock option to simulate old noflock
3037 Details : This will achieve local-only flock/fcntl locks
3043 Description: log_commit_thread vs filter_destroy race leads to crash
3044 Details : Take import reference before releasing llog record semaphore
3049 Description: Wrong request locking in request set processing
3050 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3051 imp_delayed_list, in this place should be used imp_lock.
3054 Frequency : when reconnection
3056 Description: Grant Leak when osc reconnect to OST
3057 Details : When osc reconnect ost, OST(filter) should check whether it
3058 should grant more space to client by comparing fed_grant and
3059 cl_avail_grant, and return the granted space to client instead
3060 of "new granted" space, because client will call osc_init_grant
3061 to update the client grant space info.
3064 Frequency : when client reconnect to OST
3066 Description: Grant Leak when osc do resend and replay bulk write
3067 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3068 bulk write request, because the grant info while be sync between
3069 OSC and OST when reconnect, and we should ignore the grant info
3070 these of resend/replay write req.
3075 Description: Grant space more than available left space sometimes.
3076 Details : When then OST is about to be full, if two bulk writing from
3077 different clients came to OST. Accord the avaliable space of the
3078 OST, the first req should be permitted, and the second one
3079 should be denied by ENOSPC. But if the seconde arrived before
3080 the first one is committed. The OST might wrongly permit second
3081 writing, which will cause grant space > available space.
3084 Frequency : when client is evicted
3086 Description: Grant might be wrongly erased when osc is evicted by OST
3087 Details : when the import is evicted by server, it will fork another
3088 thread ptlrpc_invalidate_import_thread to invalidate the
3089 import, where the grant will be set to 0. While the original
3090 thread will update the grant it got when connecting. So if
3091 the former happened latter, the grant will be wrongly errased
3092 because of this race.
3097 Description: Checking Stale with correct fid
3098 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3099 to check whether it is stale, because sometimes, we want the
3100 enqueue happened anyway, and op_data.fid2 will not be initialized.
3102 Severity : enhancement
3104 Description: update patchless client
3105 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3108 Frequency : only with 2.4 kernel
3110 Description: random memory corruption
3111 Details : size of struct ll_inode_info is to big for union inode.u and this
3112 can be cause of random memory corruption.
3117 Description: Memory leak in recovery
3118 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3119 It should also check obd_fail before fsfilt_start, otherwise if
3120 fsfilt_start return -EROFS,(failover mds during mds recovery).
3121 then the req will return with repmsg->transno = 0 and rc = EROFS.
3122 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3123 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3124 be freed no matter whether fsfilt_commit success or not.
3127 Frequency : only with huge count clients
3129 Description: Prevents from taking the superblock lock in llap_from_page for
3131 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3132 for llap_from_page call in ll_removepage prevents from taking the
3133 superblock lock for a soon died page.
3138 Description: Not check open intent error before release open handle
3139 Details : in some rare cases, the open intent error is not checked before
3140 release open handle, which may cause
3141 ASSERTION(open_req->rq_transno != 0), because it tries to release
3142 the failed open handle.
3147 Description: Set cat log bitmap only after create log success.
3148 Details : in some rare cases, the cat log bitmap is set too early. and it
3149 should be set only after create log success.
3153 Description: Accessing a block bevice can re-enable I/O when Lustre is
3154 tearing down a device.
3155 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3160 Description: service threads may hog cpus when there are a lot of requests
3161 Details : Insert cond_resched to give other threads a chance to use some CPU
3166 Description: the cat log was not initialized in recovery
3167 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3168 unlink log on mds will not be initialized until mds post
3169 recovery. And also in mds post recovery, the unlink log will
3170 initialization will be done asynchronausly, so there will be race
3171 between add unlink log and unlink log initialization.
3175 Description: brw_stats were being printed incorrectly
3176 Details : brw_stats were being printed as log2 but all of them were not
3177 recorded as log2. Also remove some code duplication arising from
3178 filter_tally_{read,write}.
3182 Frequency : rare, only in recovery.
3183 Description: ASSERTION(req->rq_type != LI_POISON) failed
3184 Details : imp_lock should be held while iterating over imp_sending_list for
3185 prevent destroy request after get timeout in ptlrpc_queue_wait.
3189 Description: replay-single.sh test 52 fails
3190 Details : A lock's skiplist need to be cleanup when it being unlinked
3191 from its resource list.
3195 Description: Short directio read returns full requested size rather than
3197 Details : Direct I/O operations should return actual amount of bytes
3198 transferred rather than requested size.
3200 Severity : enhancement
3202 Description: metadata RPC reduction (e.g. for rm performance)
3203 Details : decrease the amount of synchronous RPC between clients and servers
3204 by canceling conflicing lock before the operation on the client
3205 and packing thier handles into the main operation RPC to server.
3207 Severity : enhancement
3209 Description: Async OSC create to avoid the blocking unnecessarily.
3210 Details : If an OST has no remaining object, system will block on the
3211 creation when it needs to create a new object on this OST. Now,
3212 ways use pre-created objects when available, instead of blocking on
3213 an empty osc while others are not empty. If we must block, we
3214 block for the shortest possible period of time.
3218 Description: improve handling recoverable errors
3219 Details : if request processig with error which can be recoverable on server
3220 request should be resend, otherwise page released from cache and
3223 Severity : enhancement
3225 Description: refine locking for avoid write wrong info into lov_objid file
3226 Details : fix possible races with add new target and write/update data in
3229 --------------------------------------------------------------------------------
3231 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3236 Frequency : on some architectures
3238 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3239 Details : Uninitialized flags sometimes cause configuration commands to
3243 Frequency : patchless clients only
3245 Description: missing __iget() symbol export
3246 Details : The __iget() symbol export is missing. To avoid the need for
3247 this on patchless clients the deathrow inode reaper is turned
3248 off, and we depend on the VM to clean up old inodes. This
3249 dependency was during via the fix for bug 12181.
3251 --------------------------------------------------------------------------------
3253 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3255 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3256 INTEROPERATE with older versions automatically. In many cases a
3257 special upgrade step is needed. Please read the
3258 user documentation before upgrading any part of a live system.
3259 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3260 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3261 * WARNING: Lustre configuration and startup changes are required with
3262 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3266 Severity : enhancement
3268 Description: MountConf
3269 Details : Lustre configuration is now managed via mkfs and mount
3270 commands instead of lmc and lconf. New obd types (MGS, MGC)
3271 are added for dynamic configuration management. See
3272 https://mail.clusterfs.com/wikis/lustre/MountConf for
3275 Severity : enhancement
3277 Description: dynamic OST addition
3278 Details : OSTs can now be added to a live filesystem
3280 Severity : enhancement
3282 Description: startup order invariance
3283 Details : MDTs and OSTs can be started in any order. Clients only
3284 require the MDT to complete startup.
3286 Severity : enhancement
3288 Description: parallel, asynchronous orphan cleanup
3289 Details : orphan cleanup is now performed in separate threads for each
3290 OST, allowing parallel non-blocking operation.
3292 Severity : enhancement
3294 Description: optimized stripe assignment
3295 Details : stripe assignments are now made based on ost space available,
3296 ost previous usage, and OSS previous usage, in order to try
3297 to optimize storage space and networking resources.
3299 Severity : enhancement
3301 Description: Permanently set tunables
3302 Details : All writable /proc/fs/lustre tunables can now be permanently
3303 set on a per-server basis, at mkfs time or on a live system.
3305 Severity : enhancement
3307 Description: Lustre message v2
3308 Details : Add lustre message format v2.
3310 Severity : enhancement
3312 Description: client OST exclusion list
3313 Details : Clients can be started with a list of OSTs that should be
3314 declared "inactive" for known non-responsive OSTs.
3318 Description: SPEC SFS validation failure on NFS v2 over lustre.
3319 Details : Changes the blocksize for regular files to be 2x RPC size,
3320 and not depend on stripe size.
3322 Severity : enhancement
3324 Description: Multiple MD RPCs in flight.
3325 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3326 To avoid overly-overloading MDS, introduce a limit on number of
3327 MDS RPCs in flight for a single client and add /proc controls
3328 to adjust this limit.
3330 Severity : enhancement
3332 Description: client read/write statistics
3333 Details : Add client read/write call usage stats for performance
3334 analysis of user processes.
3335 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3336 file access. extents_stats shows chunk size distribution.
3337 extents_stats_per_process show chunk size distribution per
3340 Severity : enhancement
3342 Description: mds statistics
3343 Details : Add detailed mds operations statistics in
3344 /proc/fs/lustre/mds/*/stats.
3348 Description: Failure of copying files with lustre special EAs.
3349 Details : Client side always return success for setxattr call for lustre
3350 special xattr (currently only "trusted.lov").
3355 Description: Refcount LNET uuids
3356 Details : The global LNET uuid list grew linearly with every startup;
3357 refcount repeated list entries instead of always adding to
3360 Severity : enhancement
3362 Description: Dynamic service threads
3363 Details : Within a small range, start extra service threads
3364 automatically when the request queue builds up.
3367 Frequency : mixed-endian client/server environments
3369 Description: mixed-endian crashes
3370 Details : The new msg_v2 system had some failures in mixed-endian
3373 Severity : enhancement
3375 Description: Easy OST removal
3376 Details : OSTs can be permanently deactivated with e.g. 'lctl
3377 conf_param lustre-OST0001.osc.active=0'
3379 Severity : enhancement
3381 Description: MGS proc entries
3382 Details : Added basic proc entries for the MGS showing what filesystems
3385 Severity : enhancement
3387 Description: provide MGS failover
3388 Details : Added config lock reacquisition after MGS server failover.
3390 Severity : enhancement
3392 Description: add Linux 2.4 support
3393 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3397 Description: a large application tries to do I/O to the same resource and dies
3398 in the middle of it.
3399 Details : Check the req->rq_arrival time after the call to
3400 ost_brw_lock_get(), but before we do anything about
3401 processing it & sending the BULK transfer request. This
3402 should help move old stale pending locks off the queue as
3403 quickly as obd_timeout.
3406 Frequency : when an incorrect nid is specified during startup
3408 Description: ptlrpc connect to non-existant node causes kernel crash
3409 Details : LNET can't be re-entered from an event callback, which
3410 happened when we expire a message after the export has been
3411 cleaned up. Instead, hand the zombie cleanup off to another
3414 Severity : enhancement
3416 Description: plain/inodebits lock performance improvement
3417 Details : Grouping plain/inodebits in granted list by their request modes
3418 and bits policy, thus improving the performance of search through
3422 Frequency : only if OST filesystem is corrupted
3424 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3425 Details : for a short time RPCs with bulk IO are in the replay list,
3426 but replay of bulk IOs is unimplemented. If the OST filesystem
3427 is corrupted due to disk cache incoherency and then replay is
3428 started it is possible to trip an assertion. Avoid putting
3429 committed RPCs into the replay list at all to avoid this issue.
3432 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3434 Description: System hang on startup
3435 Details : This bug allowed the liblustre (e.g. catamount) client to
3436 return to the app before handling all startup RPCs. This
3437 could leave the node unresponsive to lustre network traffic
3438 and manifested as a server ptllnd timeout.
3440 Severity : enhancement
3442 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3443 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3444 boolean to control whether to print peer debug info when a
3445 client's RPC times out.
3448 Frequency : only for kernels with patches from Lustre below 1.4.3
3450 Description: Remove old rdonly API
3451 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3454 Frequency : only for devices with external journals
3456 Description: Set external device read-only also
3457 Details : During a commanded failover stop, we set the disk device
3458 read-only while the server shuts down. We now also set any
3459 external journal device read-only at the same time.
3462 Frequency : when upgrading from 1.4 while trying to change parameters
3464 Description: The wrong (new) MDC name was used when setting parameters for
3465 upgraded MDT's. Also allows changing of OSC (and MDC)
3466 parameters if --writeconf is specified at tunefs upgrade time.
3469 Frequency : when setting specific ost indicies
3471 Description: QOS code breaks on skipped indicies
3472 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3473 created with --index need not be sequential.
3475 Severity : enhancement
3477 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3478 Details : The uninit_groups feature works in conjunction with the kernel
3479 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3480 up the pass1 processing of e2fsck. This is a read-only feature
3481 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3482 mount filesystems that have had this feature enabled.
3484 Severity : enhancement
3486 Description: Improve multi-block allocation algorithm to avoid fragmentation
3487 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3488 allocation locality and avoid filesystem fragmentation.
3490 ------------------------------------------------------------------------------
3492 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3494 * Support for kernels:
3495 2.4.21-47.0.1.EL (RHEL 3)
3496 2.6.5-7.283 (SLES 9)
3497 2.6.9-42.0.10.EL (RHEL 4)
3498 2.6.12.6 vanilla (kernel.org)
3499 2.6.16.27-0.9 (SLES 10)
3500 * Recommended e2fsprogs version: 1.39.cfs5
3502 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3506 Frequency : occasional, depends on client load and configuration
3507 Bugzilla : 12181, 12203
3508 Description: data loss for recently-modified files
3510 Details : In some cases it is possible that recently written or created
3511 files may not be written to disk in a timely manner (this should
3512 normally be within 30s unless client IO load is very high).
3513 The problem appears as zero-length files or files that are a
3514 multiple of 1MB in size after a client crash or client eviction
3515 that are missing data at the end of the file.
3517 This problem is more likely to be hit on clients where files are
3518 repeatedly created and unlinked in the same directory, clients
3519 have a large amount of RAM, have many CPUs, the filesystem has
3520 many OSTs, the clients are rebooted frequently, and/or the files
3521 are not accessed by other nodes after being written.
3523 The presence of the problem can be detected by looking at
3524 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3525 smaller than the second (nr_unused) then dirty files will not
3526 be flushed automatically to disk. "sync; sleep 10" should be
3527 run several times on the node before unmounting it to update
3528 Lustre (this is also safe to run on nodes without this problem).
3530 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3531 that can cause this same problem, so customers using that kernel
3532 also need to update the kernel in addition to Lustre. In order
3533 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3535 It is normal that files written just before a client crash (less
3536 than 30s) may not yet have been flushed to disk, even for local
3540 Frequency : frequent on thin XT3 nodes
3542 Description: UUID collision on thin XT3 Linux nodes
3543 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3544 randomly, since we relied on an insufficiently-seeded PRNG.
3549 Description: OSS hangs after "All ost request buffers busy"
3550 Details : A deadlock between quota and journal operations caused OSS
3551 hangs after printing "All ost request buffers busy."
3554 Frequency : always on liblustre builds
3556 Description: Cleanup compiler warnings on liblustre
3559 Frequency : always on liblustre builds on XT3
3561 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3565 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3566 > 32000 subdirectories
3567 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3568 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3569 subdirectory count crosses 32000. This will aid e2fsck to
3570 correctly handle more than 32000 subdirectories.
3573 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3575 Description: System hang on startup
3576 Details : This bug allowed the liblustre (e.g. catamount) client to
3577 return to the app before handling all startup RPCs. This
3578 could leave the node unresponsive to lustre network traffic
3579 and manifested as a server ptllnd timeout.
3581 Severity : enhancement
3583 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3584 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3585 boolean to control whether to print peer debug info when a
3586 client's RPC times out.
3591 Description: make O_SYNC working on 2.6 kernels
3592 Details : 2.6 kernels use different method for mark pages for write,
3593 so need add a code to lustre for O_SYNC work.
3598 Description: Failure to close file and release space on NFS
3599 Details : Put inode details into lock acquired in ll_intent_file_open.
3600 Use mdc_intent_lock in ll_intent_open to properly
3601 detect all kind of errors unhandled by mdc_enqueue.
3606 Description: proc file read during shutdown sometimes raced obd removal,
3608 Details : Add lock to prevent obd access after proc file removal.
3611 Frequency : Only for files larger than 4GB on 32-bit clients.
3613 Description: improperly doing page alignment of locks
3614 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3615 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3620 Description: RPCs being resent when they shouldn't be
3621 Details : Some RPCs that should not be resent are being resent. This
3622 can cause inconsistencies in the RPC state machine. Do not
3623 resend such requests.
3626 Frequency : rare, only with NFS export
3628 Description: Crash on NFS re-export node
3629 Details : under very unusual load conditions an assertion is hit in
3630 ll_intent_file_open()
3633 Frequency : only if OST filesystem is corrupted
3635 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3636 Details : for a short time RPCs with bulk IO are in the replay list,
3637 but replay of bulk IOs is unimplemented. If the OST filesystem
3638 is corrupted due to disk cache incoherency and then replay is
3639 started it is possible to trip an assertion. Avoid putting
3640 committed RPCs into the replay list at all to avoid this issue.
3645 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3646 Details : Large single O_DIRECT read and write calls can fail to allocate
3647 a sufficiently large buffer to process the request. In case of
3648 allocation failure the allocation is retried with a smaller
3649 buffer and broken into smaller requests.
3651 Severity : enhancement
3653 Description: Add -o localflock option to simulate old noflock behaviour.
3654 Details : This will achieve local-only flock/fcntl locks coherentness.
3659 Description: versioning check is incomplete
3660 Details : Checking the version difference of client vs. server, report
3661 error if the gap is too big.
3666 Description: add support PG_writeback bit
3667 Details : add support for PG_writeback bit for Lustre, for more carefull
3668 work with page cache in 2.6 kernel. This also fix some deadlocks
3669 and remove hack for work O_SYNC with 2.6 kernel.
3671 Severity : enhancement
3673 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3674 Details : The uninit_groups feature works in conjunction with the kernel
3675 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3676 up the pass1 processing of e2fsck. This is a read-only feature
3677 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3678 mount filesystems that have had this feature enabled.
3680 Severity : enhancement
3682 Description: Improve multi-block allocation algorithm to avoid fragmentation
3683 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3684 allocation locality and avoid filesystem fragmentation.
3686 ------------------------------------------------------------------------------
3688 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3690 * Support for kernels:
3691 2.6.9-42.0.3.EL (RHEL 4)
3692 2.6.5-7.276 (SLES 9)
3693 2.4.21-47.0.1.EL (RHEL 3)
3694 2.6.12.6 vanilla (kernel.org)
3695 2.6.16.21-0.8 (SLES10)
3696 * Recommended e2fsprogs version: 1.39.cfs2-0
3698 * The backwards-compatible /proc/sys/portals symlink has been removed
3699 in this release. Before upgrading, please ensure that you change
3700 any configuration scripts or /etc/sysctl.conf files that access
3701 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3702 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3703 in advance of the upgrade on any system running Lustre 1.4.6 or
3704 newer, since /proc/sys/lnet was added in that version.
3705 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3709 Frequency : only when quota is used
3711 Description: avoid scanning export list for quota master
3712 Details : Change the algorithms to avoid scanning export list in order
3713 to improve the efficiency.
3716 Frequency : MDS failover only, very rarely
3718 Description: "went back in time" messages on mds failover
3719 Details : The greatest transno may be lost when the current operation
3720 finishes with an error (transno==0) and the client's last_rcvd
3721 record is over-written. Save the greatest transno in the
3722 mds_last_transno for this case.
3725 Frequency : always for specific kernels and striping counts
3727 Description: client may get "Matching packet too big" without ACL support
3728 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3729 when trying to access files in certain configurations. The
3730 clients should in fact be denied when mounting because they do
3731 not understand ACLs.
3734 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3736 Description: many clients connecting with IO in progress causes connect timeouts
3737 Details : Avoid synchronous journal commits to avoid delays caused by many
3738 clients connecting/disconnecting when bulk IO is in progress.
3739 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3740 OST_IO_PORTAL to avoid delays behind potentially many pending
3744 Frequency : occasionally with multiple writers to a single file
3746 Description: shared writes to file may result in wrong size reported by stat()
3747 Details : Allow growing of kms when extent lock is cancelled
3750 Frequency : always with random mmap IO to multi-striped file
3752 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3753 Details : Only if the hole is at the end of OST object so that kms is too
3754 small. Fix is to increase kms accordingly in ll_nopage.
3757 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3759 Description: writes to a missing object would leak memory on the OST
3760 Details : If there is an inconsistency between the MDS and OST filesystems,
3761 such that the MDS references an object that doesn't exist, writes
3762 to that object will leak memory due to incorrect cleanup in the
3763 error handling path, eventually running out of memory on the OST.
3768 Description: Creating too long symlink causes lustre errors
3769 Details : Check symlink and name lengths before sending requests to MDS.
3772 Frequency : only if flock is enabled (not on by default)
3774 Description: posix locks not released on fd closure on 2.6.9+
3775 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3776 kernels, this caused such locks not to be released on fd close and
3777 then assertions on fs unmount about still used locks.
3780 Frequency : MDS failover only, very rarely
3782 Description: clients may get ASSERTION(granted_lock != NULL)
3783 Details : When request was taking a long time, and a client was resending
3784 a getattr by name lock request. The were multiple lock requests
3785 with the same client lock handle and
3786 mds_getattr_name->fixup_handle_for_resent_request found one of the
3787 lock handles but later failed with ASSERTION(granted_lock != NULL).
3792 Description: handle->h_buffer_credits > 0, assertion failure
3793 Details : h_buffer_credits is zero after truncate, causing assertion
3794 failure. This patch extends the transaction or creates a new
3798 Frequency : NFS re-export or patchless client
3799 Bugzilla : 11179, 10796
3800 Description: Crash on NFS re-export node (__d_move)
3801 Details : We do not want to hash the dentry if we don't have a lock.
3802 But if this dentry is later used in d_move, we'd hit uninitialised
3803 list head d_hash, so we just do this to init d_hash field but
3804 leave dentry unhashed.
3807 Frequency : NFS re-export or patchless client
3809 Description: NFS exports has problem with symbolic link
3810 Details : lustre client didn't properly install dentry when re-exported
3811 to NFS or running patchless client.
3814 Frequency : NFS re-export or patchless client
3816 Description: Various nfs/patchless fixes.
3817 Details : fixes reuse disconected alias for lookup process - this fixes
3818 warning "find_exported_dentry: npd != pd",
3819 fix permission error with open files at nfs.
3820 fix apply umask when do revalidate.
3823 Frequency : occasional
3825 Description: Crash on NFS re-export node
3826 Details : calling clear_page() on the wrong pointer triggered oops in
3827 generic_mapping_read().
3830 Frequency : rarely, using O_DIRECT IO
3832 Description: unaligned directio crashes client with LASSERT
3833 Details : check for unaligned buffers before trying any requests.
3836 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3838 Description: stale data returned from RAID cache
3839 Details : If only a small amount of IO is done to the RAID device before
3840 reading it again it is possible to get stale data from the RAID
3841 cache instead of reading it from disk.
3844 Frequency : always for sles10 kernel
3846 Description: sles10 support
3847 Details : ll_follow_link: compile fixes and using of nd_set_link
3848 under newer kernels.
3851 Frequency : depends on arch, kernel and compiler version, always on sles10
3854 Description: recursive or deep enough symlinks cause stack overflow
3855 Details : getting rid of large stack-allocated variable in
3859 Frequency : depends on hardware
3861 Description: lustre write performance loss in the SLES10 kernel
3862 Details : the performance loss is caused by using of write barriers in the
3863 ext3 code. The SLES10 kernel turns barrier support on by
3864 default. The fix is to undo that change for ldiskfs.
3866 ------------------------------------------------------------------------------
3868 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3870 * Support for kernels:
3871 2.6.9-42.0.3EL (RHEL 4)
3872 2.6.5-7.276 (SLES 9)
3873 2.4.21-47.0.1.EL (RHEL 3)
3874 2.6.12.6 vanilla (kernel.org)
3878 Frequency : quota enabled and large files being deleted
3880 Description: releasing more than 4GB of quota at once hangs OST
3881 Details : If a user deletes more than 4GB of files on a single OST it
3882 will cause the OST to spin in an infinite loop. Release
3883 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3888 Description: statfs data retrieved from /proc may be stale or zero
3889 Details : When reading per-device statfs data from /proc, in the
3890 {kbytes,files}_{total,free,avail} files, it may appear
3891 as zero or be out of date.
3894 Frequency : systems with MD RAID1 external journal devices
3896 Description: lconf's call to blkid is confused by RAID1 journal devices
3897 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3898 of returning all block devices that match the journal UUID.
3901 Frequency : always, for aggregate stripe size over 4GB
3903 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3904 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3905 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3908 Frequency : always if "lfs find" used on a local file/directory
3910 Description: "lfs find" segfaults if used on a local file/directory
3911 Details : The case where a directory component was not specified wasn't
3912 handled correctly. Handle this properly.
3915 Frequency : always on ppc64
3917 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3918 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3919 perform endianness conversion.
3922 Frequency : rarely (truncate to non-zero file size after write under load)
3923 Bugzilla : 10730, 10687
3924 Description: Files padded with zeros to next 4K multiple
3925 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3926 it is possible that files that are truncated to a non-zero size
3927 immediately after being written are filled with zero bytes beyond
3928 the truncated size. No file data is lost.
3930 Severity : enhancement
3932 Description: Allow recovery/failover for liblustre clients.
3933 Details : liblustre clients were unaware of failover configurations until
3936 Severity : enhancement
3938 Description: user file locks should fail when not mounting with flock option
3939 Details : Set up an error-returning stub in ll_file_operations.lock field
3940 to prevent incorrect behaviour when client is mounted without
3941 flock option. Also, set up properly f_op->flock field for
3945 Frequency : always on ia64
3947 Description: "lfs df" loops on printing out MDS statfs information
3948 Details : The obd_ioctl_data was not initialized and in some systems
3949 this caused a failure during the ioctl that did not return
3950 an error. Initialize the struct and return an error on failure.
3953 Frequency : SLES 9 only
3955 Description: Error of copying files with lustre special EAs as root
3956 Details : Client side always return success for setxattr call for lustre
3957 special xattr (currently only "trusted.lov").
3960 Frequency : rarely on clusters with both ia64+i386 clients
3962 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3963 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3964 clients are doing concurrent writes to the same file it is
3965 possible that the ia64 clients may LASSERT because the OST
3966 extent locks are not PAGE_SIZE aligned. Ensure that grown
3967 locks are always aligned on the request boundary.
3970 Frequency : specific use, occasional
3972 Description: Overwriting in use executable truncates on-disk binary image
3973 Details : If one node attempts to overwrite an executable in use by
3974 another node, we now correctly return ETXTBSY instead of
3975 truncating the file.
3977 Severity : enhancement
3979 Description: Async OSC create to avoid the blocking unnecessarily.
3980 Details : If an OST has no remaining object, system will block on the creating
3981 when need to create a new object on this OST. Now, ways use
3982 pre-created objects when available, instead of blocking on an
3983 empty osc while others are not empty. If we must block, we block
3984 for the shortest possible period of time.
3989 Description: chmod on Lustre root is propagated to other clients
3990 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3991 invalid by the follow_mount time.
3994 Frequency : liblustre clients only
3996 Description: Race in 'instant cancel' lock handling could lead to such locks
3997 never to be granted in case of SMP MDS
3998 Details : Do not destroy not yet granted but cbpending locks in
4002 Frequency : replay/resend of open
4004 Description: non null lock assetion failure in mds_intent_policy
4005 Details : Trying to replay/resend lockless open requests resulted in
4006 mds_open() returning 0 with no lock. Now it sets a flag if
4007 a lock is going to be returned.
4009 Severity : enhancement
4011 Description: Checksum enhancements
4012 Details : New checksum enhancements allow for resending RPCs that failed
4015 Severity : enhancement
4017 Description: Tunables on number of dirty pages in cacche
4018 Details : Allow to set limit on number of dirty pages cached.
4023 Description: client crash on unmount - lock still has references
4024 Details : In some error handling cases it was possible to leak a lock
4025 reference on a client while accessing a file. This was not
4026 harmful to the client during operation, but would cause the
4027 client to crash when the filesystem is unmounted.
4030 Frequency : specific case, rare
4032 Description: ETXTBSY on mds though file not in use by client
4033 Details : ETXTBSY is no longer incorrectly returned when attempting to
4034 chmod or chown a directory that the user previously tried to
4035 execute or a currently-executing binary.
4038 Frequency : extremely rare except on liblustre-based clients
4040 Description: Lustre space not freed when files are deleted
4041 Details : Clean up open-unlinked files after client eviction. Previously
4042 the unlink was skipped and the files remained as orphans.
4047 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4048 Details : In some cases it was possible to send a blocking callback to
4049 a client doing a glimpse, even though that client didn't get
4050 a lock granted. When the glimpse lock is cancelled on the OST
4051 the freed lock is left on the waiting list and corrupted the list.
4054 Frequency : all core dumps
4056 Description: Broke core dumps to lustre
4057 Details : Negative dentry may be unhashed if parent does not have UPDATE
4058 lock, but some callers, e.g. do_coredump, expect dentry to be
4059 hashed after successful create, hash it in ll_create_it.
4061 ------------------------------------------------------------------------------
4063 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4065 * Support for kernels:
4066 2.6.9-42.0.2.EL (RHEL 4)
4067 2.6.5-7.276 (SLES 9)
4068 2.4.21-40.EL (RHEL 3)
4069 2.6.12.6 vanilla (kernel.org)
4073 Frequency : always on RHEL 3
4075 Description: Number of open files grows over time
4076 Details : The number of open files grows over time, whether or not
4077 Lustre is started. This was due to a filp leak introduced
4078 by one of our kernel patches.
4080 ------------------------------------------------------------------------------
4082 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4084 * Support for kernels:
4085 2.6.9-42.EL (RHEL 4)
4086 2.6.5-7.276 (SLES 9)
4087 2.4.21-40.EL (RHEL 3)
4088 2.6.12.6 vanilla (kernel.org)
4093 Bugzilla : 5719, 9635, 9792, 9684
4094 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4095 Details : If a server is under heavy load and cannot reply to new
4096 connection requests before the client resends the (re)connect,
4097 the connection handling code can behave badly if two service
4098 threads are concurrently handing separate (re)connections from
4099 the same client. Add better locking to the connection handling
4100 code, and ensure that only a single connection will be processed
4101 for a given client UUID, even if the lock is dropped.
4103 Severity : enhancement
4105 Description: add TCP zero-copy support to kernel
4106 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4107 sends if the hardware supports scatter-gather and checksumming.
4108 This allows socklnd to do client-write and server-read more
4109 efficiently and reduce CPU utilization from skbuf copying.
4112 Frequency : only if NFS exporting from client
4114 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4115 Details : If setting attributes on a file created under NFS that had
4116 never been opened it would be possible to oops the client
4117 if the file had no objects.
4120 Frequency : always for liblustre
4122 Description: liblustre client does MDS+OSTs setattr RPC for each write
4123 Details : When doing a write from a liblustre client, the client
4124 incorrectly issued an RPC to the MDS and each OST the file was
4125 striped over in order to update the timestamps. When writing
4126 with small chunks and many clients this could overwhelm the MDS
4127 with RPCs. In all cases it would slow down the write because
4128 these RPCs are unnecessary.
4130 Severity : enhancement
4132 Description: allow number of MDS service threads to be changed at module load
4133 Details : It is now possible to change the number of MDS service threads
4134 running. Adding "options mds mds_num_threads={N}" to the MDS's
4135 /etc/modprobe.conf will set the number of threads for the next
4136 time Lustre is restarted (assuming the "mds" module is also
4137 reloaded at that time). The default number of threads will
4138 stay the same, 32 for most systems.
4143 Description: OST crash if filesystem is unformatted or corrupt
4144 Details : If an OST is started on a device that has never been formatted
4145 or if the filesystem is corrupt and cannot even mount then the
4146 error handling cleanup routines would dereference a NULL pointer.
4151 Description: NULL pointer deref in llap_from_page.
4152 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4153 incorrect) mapping if the page was truncated/reclaimed while it was
4154 searched for. Check for this condition and skip such pages when
4155 doing readahead. Introduce extra check to llap_from_page() to
4156 verify page->mapping->host is non-NULL (so page is not anonymous).
4159 Frequency : Sometimes when using sys_sendfile
4161 Description: "page not covered by a lock" warnings from ll_readpage
4162 Details : sendfile called ll_readpage without right page locks present.
4163 Now we introduced ll_file_sendfile that does necessary locking
4164 around call to generic_file_sendfile() much like we do in
4168 Frequency : with certain MDS communication failures at client mount time
4170 Description: NULL pointer deref after failed client mount
4171 Details : a client connection request may delayed by the network layer
4172 and not be sent until after the PTLRPC layer has timed out the
4173 request. If the client fails the mount immediately it will try
4174 to clean up before the network times out the request. Add a
4175 reference from the request import to the obd device and delay
4176 the cleanup until the network drops the request.
4179 Frequency : occasionally during client (re)connect
4181 Description: assertion failure during client (re)connect
4182 Details : processing a client connection request may be delayed by the
4183 client or server longer than the client connect timeout. This
4184 causes the client to resend the connection request. If the
4185 original connection request is replied in this interval, the
4186 client may trip an assertion failure in ptlrpc_connect_interpret()
4187 which thought it would be the only running connect process.
4190 Frequency : only with obd_echo servers and clients that are rebooted
4192 Description: kernel BUG accessing uninitialized data structure
4193 Details : When running an obd_echo server it did not start the ping_evictor
4194 thread, and when a client was evicted an uninitialized data
4195 structure was accessed. Start the ping_evictor in the RPC
4196 service startup instead of the OBD startup.
4198 Severity : enhancement
4199 Bugzilla : 10193 (patchless)
4200 Description: Remove dependency on various unexported kernel interfaces.
4201 Details : No longer need reparent_to_init, exit_mm, exit_files,
4202 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4205 Frequency : rare (only users of deprecated and unsupported LDAP config)
4207 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4208 Details : LDAP apparently contains 'lustreName' attributes instead of
4209 'name'. A simple remapping of the name is sufficient.
4212 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4214 Description: waiting_locks_callback trips kernel BUG if client is evicted
4215 Details : Running with the dump_on_timeout debug flag turned on makes
4216 it possible that the waiting_locks_callback() can try to dump
4217 the Lustre kernel debug logs from an interrupt handler. Defer
4218 this log dumping to the expired_lock_main() thread.
4220 Severity : enhancement
4222 Description: Support NFS exporting on 2.6 kernels.
4223 Details : Implement non-rawops metadata methods for NFS server to use without
4224 changing NFS server code.
4227 Frequency : very rare (synthetic metadata workload only)
4229 Description: two racing renames might cause an MDS thread to deadlock
4230 Details : Running the "racer" program may cause one MDS thread to rename
4231 a file from being the source of a rename to being the target of
4232 a rename at exactly the same time that another thread is doing
4233 so, and the second thread has already enqueued these locks after
4234 doing a lookup of the target and is trying to relock them in
4235 order. Ensure that we don't try to re-lock the same resource.
4238 Frequency : only very large systems with liblustre clients
4240 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4241 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4242 all OSTs in parallel. This allows the eviction of stale liblustre
4243 clients to proceed much faster than if they were done in series,
4244 and also offers similar improvements for other set_info RPCs.
4249 Description: excessive CPU usage during initial read phase on client
4250 Details : During the initial read phase on a client, it would agressively
4251 retry readahead on the file, consuming too much CPU and impacting
4252 performance (since 1.4.5.8). Improve the readahead algorithm
4253 to avoid this, and also improve some other common cases (read
4254 of small files in particular, where "small" is files smaller than
4255 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4260 Description: MDS crash when receiving packet with unknown intent.
4261 Details : Do not LBUG in unknown intent case, just return -EFAULT
4263 Severity : enhancement
4264 Bugzilla : 9293, 9385
4265 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4266 Details : Do not serialize getattr (non-intent version) and statfs.
4269 Frequency : occasional, when OST network is overloaded/intermittent
4271 Description: client evicted by OST after bulk IO timeout
4272 Details : If a client sends a bulk IO request (read or write) the OST
4273 may evict the client if it is unresposive to its data GET/PUT
4274 request. This is incorrect if the network is overloaded (takes
4275 too long to transfer the RPC data) or dropped the OST GET/PUT
4276 request. There is no need to evict the client at all, since
4277 the pinger and/or lock callbacks will handle this, and the
4278 client can restart the bulk request.
4281 Frequency : Always when mmapping file with no objects
4283 Description: client crashes when mmapping file with no objects
4284 Details : Check that we actually have objects in a file before doing any
4285 operations on objects in ll_vm_open, ll_vm_close and
4291 Description: Request leak when working with deleted CWD
4292 Details : Introduce advanced request refcount tracking for requests
4293 referenced from lustre intent.
4295 Severity : Enhancement
4297 Description: Cache open file handles on client.
4298 Details : MDS now will return special lock along with openhandle, if
4299 requested and client is allowed to hold openhandle, even if unused,
4300 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4301 closing files for every read/write openration.
4303 Severity : Enhancement
4305 Description: Cache open negative dentries on client when possible.
4306 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4307 negative dentries on lock revocation.
4312 Description: Remounting a client read-only wasn't possible with a zconf mount
4313 Details : It wasn't possible to remount a client read-only with llmount.
4315 Severity : enhancement
4316 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4317 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4318 MPICH is included to add Lustre-specific ADIO interfaces.
4319 This is based closely on the UFS ADIO layer and only differs
4320 in file creation, in order to allow the OST striping to be set.
4321 This is user-contributed code and not supported by CFS.
4326 Description: extended inode attributes (immutable, append-only) work improperly
4327 when 2.4 and 2.6 kernels are used on client/server or vice versa
4328 Details : Introduce kernel-independent values for these flags.
4330 Severity : enhancement
4333 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4334 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4335 granularity. Now, Fractional MB granularity is be supported,
4336 this is very useful for low memory system.
4338 Severity : enhancement
4340 Description: Getattr by fid
4341 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4342 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4343 have locks and avoids getattr rpc requests for it on every stat.
4346 Frequency : Always, for filesystems larger than 2TB
4348 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4349 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4350 filesystems larger than 2TB to be mounted with mballoc enabled.
4353 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4355 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4356 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4357 filesystems larger than 2TB is mounted then the kernel will
4358 silently corrupt the start of the filesystem. CONFIG_LBD is
4359 enabled for all CFS-supported kernels, but the possibility of
4360 this happening with a modified kernel config exists.
4362 Severity : enhancement
4364 Description: add client O_DIRECT support for 2.6 kernels
4365 Details : It is now possible to do O_DIRECT reads and writes to files
4366 in the Lustre client mountpoint on 2.6 kernel clients.
4368 Severity : enhancement
4370 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4371 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4372 parallel, not waiting for response from every OST before sending
4373 a rpc to the next OST.
4378 Description: setattr vs write race when updating file timestamps
4379 Details : Client processes that update a file timestamp into the past
4380 right after writing to the file (e.g. tar) it is possible that
4381 the updated file modification time can be reset to the current
4382 time due to a race between processing the setattr and write RPC.
4384 Severity : enhancement
4386 Description: Bring 'lfs find' closer in line with regular Linux find.
4387 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4388 -print0 options and obtains all the needed info through the lustre
4391 Severity : enhancement
4393 Description: support up to 1024 configured devices on one node
4394 Details : change obd_dev array from statically allocated to dynamically
4395 allocated structs as they are first used to reduce memory usage
4400 Description: Flush dirty partially truncated pages during truncate
4401 Details : Immediatelly flush partially truncated pages in filter_setattr,
4402 this way we completely avoid having any pages in page cache on OST
4403 and can retire ugly workarounds during writes to flush such pages.
4408 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4409 Details : protect inode from truncation within vfs_unlink() context
4410 just take a reference before calling vfs_unlink() and release it
4411 when parent's i_sem is free.
4416 Description: last_id value checked outside lock on OST caused LASSERT failure
4417 Details : If there were multiple MDS->OST object precreate requests in
4418 flight, it was possible that the OST's last object id was checked
4419 outside a lock and incorrectly tripped an assertion. Move checks
4420 inside locks, and discard old precreate requests.
4423 Frequency : always, if extents are used on OSTs
4425 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4426 Details : OSTs using the extents format would not zero the high 16 bits of
4427 the index physical block number. This is not a problem for any
4428 OST filesystems smaller than 16TB, and no kernels support ext3
4429 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4430 new/modified files) and can be fixed for existing filesystems
4431 with e2fsprogs-1.39-cfs1.
4436 Description: import connection selection may be incorrect if timer wraps
4437 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4438 import connections to be ignored if the 32-bit jiffies counter
4439 wraps. Use a 64-bit jiffies counter.
4442 Frequency : very large clusters immediately after boot
4444 Description: LNET request buffers exhausted under heavy short-term load
4445 Details : If a large number of client requests are generated on a service
4446 that has previously never seen so many requests it is possible
4447 that the request buffer growth cannot keep up with the spike in
4448 demand. Instead of dropping incoming requests, they are held in
4449 the LND until the RPC service can accept more requests.
4452 Frequency : Sometimes during replay
4454 Description: Assertion failure in ll_local_open after replay.
4455 Details : If replay happened on an open request reply before we were able
4456 to set replay handler, reply will become not swabbed tripping the
4457 assertion in ll_local_open. Now we set the handler right after
4458 recognising of open request
4461 Frequency : very rare
4463 Description: kernel reports "badness in vsnprintf"
4464 Details : Reading from the "recovery_status" /proc file in small chunks
4465 may cause a negative length in lprocfs_obd_rd_recovery_status()
4466 call to vsnprintf() (which is otherwise harmless). Exit early
4467 if there is no more space in the output buffer.
4469 Severity : enhancement
4471 Description: clear OBD RPC statistics by writing to them
4472 Details : It is now possible to clear the OBD RPC statistics by writing
4473 to the "stats" file.
4478 Description: Client mtime is not the same on different clients after utimes
4479 Details : In some cases, the client was using the utimes() syscall on
4480 a file cached on another node. The clients now validate the
4481 ctime from the MDS + OSTs to determine which one is right.
4486 Description: Inability to activate failout mode
4487 Details : lconf script incorrectly assumed that in pythong string's numeric
4488 value is used in comparisons.
4491 Frequency : always with multiple stripes per file
4493 Description: Inefficient object allocation for mutli-stripe files
4494 Details : When selecting which OSTs to stripe files over, for files with
4495 a stripe count that divides evenly into the number of OSTs,
4496 the MDS is always picking the same starting OST for each file.
4497 Return the OST selection heuristic to the original design.
4502 Description: mount failures may take full timeout to return an error
4503 Details : Under some heavy load conditions it is possible that a
4504 failed mount can wait for the full obd_timeout interval,
4505 possibly several minutes, before reporting an error.
4506 Instead return an error as soon as the status is known.
4508 Frequency : quota enabled and large files being deleted
4510 Description: releasing more than 4GB of quota at once hangs OST
4511 Details : If a user deletes more than 4GB of files on a single OST it
4512 will cause the OST to spin in an infinite loop. Release
4513 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4518 Description: statfs data retrieved from /proc may be stale or zero
4519 Details : When reading per-device statfs data from /proc, in the
4520 {kbytes,files}_{total,free,avail} files, it may appear
4521 as zero or be out of date.
4524 Frequency : systems with MD RAID1 external journal devices
4526 Description: lconf's call to blkid is confused by RAID1 journal devices
4527 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4528 of returning all block devices that match the journal UUID.
4531 Frequency : always, for aggregate stripe size over 4GB
4533 Description: assertion fails when trying to use 4GB stripe size
4534 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4535 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4538 Frequency : always on ppc64
4540 Description: the first write on an ext3 filesystem with mballoc got stuck
4541 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4542 perform endianness conversion.
4544 ------------------------------------------------------------------------------
4546 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4548 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4549 INTEROPERATE with older versions automatically. Please read the
4550 user documentation before upgrading any part of a live system.
4551 * WARNING: Lustre networking configuration changes are required with
4552 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4555 * Support for kernels:
4556 2.6.9-22.0.2.EL (RHEL 4)
4557 2.6.5-7.244 (SLES 9)
4558 2.6.12.6 vanilla (kernel.org)
4561 Severity : enhancement
4562 Bugzilla : 7981/8208
4563 Description: Introduced Lustre Networking (LNET)
4564 Details : LNET is new networking infrastructure for Lustre, it includes
4565 a reorganized network configuration mode (see the user
4566 documentation for full details) as well as support for routing
4567 between different network fabrics. Lustre Networking Devices
4568 (LNDs) for the supported network fabrics have also been
4569 created for this new infrastructure.
4571 Severity : enhancement
4572 Description: Introduced Access control lists
4573 Details : clients can set ACLs on files and directories in order to have
4574 more fine-grained permissions than the standard Unix UGO+RWX.
4575 The MDS must be started with the "-o acl" mount option.
4577 Severity : enhancement
4578 Description: Introduced filesystem quotas
4579 Details : Administrators may now establish per-user quotas on the
4582 Severity : enhancement
4584 Description: Configuration change for the XT3
4585 The PTLLND is now used to run Lustre over Portals on the XT3
4586 The configure option(s) --with-cray-portals are no longer used.
4587 Rather --with-portals=<path-to-portals-includes> is used to
4588 enable building on the XT3. In addition to enable XT3 specific
4589 features the option --enable-cray-xt3 must be used.
4594 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4595 Details : A race between allocating a new llcd and re-getting the llcd_lock
4596 allowed another thread to grab newly-allocated llcd.
4598 Severity : enhancement
4600 Description: 2.6 OST async journal commit and locking fix to improve performance
4601 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4602 2.6 kernels are now async as they already were in 2.4 kernels so
4603 that they can commit concurrently with the network bulk transfer.
4604 For block-allocated files the filter allocation semaphore is held
4605 to avoid filesystem fragmentation during allocation. BKL lock
4606 removed for 2.6 xattr operations where it is no longer needed.
4611 Description: lconf incorrectly determined whether two IP networks could talk
4612 Details : In some more complicated routing and multiple-network
4613 configurations, lconf will avoid trying to make a network
4614 connection to a disjoint part of the IP space. It was doing the
4615 math incorrectly for one set of cases.
4620 Description: Fix for potential infinite loop processing records in an llog.
4621 Details : If an llog record is corrupted/zeroed, it is possible to loop
4622 forever in llog_process(). Validate the llog record length
4623 and skip the remainder of the block on error.
4626 Frequency : occasional (liblustre only)
4628 Description: liblustre could not open files whose last component is a symlink
4629 Details : sysio_path_walk() would incorrectly pass the open intent to
4630 intermediate path components.
4633 Frequency : rare (liblustre only with non-standard tuning)
4634 Bugzilla : 7201 (7350)
4635 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4636 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4637 after it has been released, only do this for referenced locks
4639 Severity : enhancement
4641 Description: specifying an (invalid) directory default stripe_size of -1
4642 would reset the directory default striping
4643 Details : stripe_size -1 was used internally to signal directory stripe
4644 removal, now use "all default" to signal dir stripe removal
4645 as a directory striping of "all default" is not useful
4648 Frequency : common for large clusters running liblustre clients
4650 Description: doing an ls when liblustre clients are running is slow
4651 Details : sending a glimpse AST to a liblustre client waits for every AST
4652 to time out, as liblustre clients will not respond. Since they
4653 cannot cache data we refresh the OST lock LVB from disk instead.
4655 Severity : enhancement
4657 Description: doing an ls at the same time as file IO can be slow
4658 Details : enqueue and other "small" requests can be blocked behind many
4659 large IO requests. Create a new OST IO portal for non-IO
4660 requests so they can be processed faster.
4663 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4665 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4666 created file with mode 000
4667 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4668 allow an ftruncate() as a non-root user to a file with mode 000.
4669 Lustre now respects this flag to disable mode checks when
4670 truncating a file owned by the user
4673 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4676 Description: Revoking locks from clients that went dead or catatonic might take
4678 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4679 cancellation of such locks instant on servers without waiting for
4680 any reply from clients. Clients drops these locks when cancel
4681 notification from server is received without replying.
4684 Frequency : liblustre-only, when liblustre client dies or becomes busy
4686 Description: Doing ls on Linux clients can take a long time with active
4688 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4689 granting such locks to it in the first place if possible. Locks
4690 are taken by proxy on the OST during the read or write and
4691 dropped immediately afterward. Add connect flags handling, do
4692 not grant locks to liblustre clients for glimpse ASTs.
4694 Severity : enhancement
4696 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4697 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4698 sequential reads and behaves badly with applications doing random
4699 reads. Improve it by reading ahead at least read region, and
4700 avoiding excessive large RPC for small reads.
4702 Severity : enhancement
4704 Description: Creating more than 1000 files for a single job may cause a load
4705 imbalance on the OSTs if there are also a large number of OSTs.
4706 Details : qos_prep_create() uses an OST index reseed value that is an
4707 even multiple of the number of available OSTs so that if the
4708 reseed happens in the middle of the object allocation it will
4709 still utilize the OSTs as uniformly as possible.
4714 Description: OST or MDS may oops in ping_evictor_main()
4715 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4716 but doesn't restart at beginning of obd_exports_timed list
4719 Severity : enhancement
4721 Description: improve by-nid export eviction on the MDS and OST
4722 Details : allow multiple exports with the same NID to be evicted at one
4723 time without re-searching the exports list.
4726 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4728 Description: MDS may oops in groups_free()
4729 Details : in rare race conditions a newly allocated group_info struct is
4730 freed again, and this can be NULL. The 2.4 compatibility code
4731 for groups_free() checked for a NULL pointer, but 2.6 did not.
4734 Frequency : common for liblustre clients doing little filesystem IO
4735 Bugzilla : 9352, 7313
4736 Description: server may evict liblustre clients accessing contended locks
4737 Details : if a client is granted a lock or receives a completion AST
4738 with a blocking AST already set it would not reply to the AST
4739 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4742 Frequency : lfs setstripe, only systems with more than 160 OSTs
4744 Description: unable to set striping with a starting offset beyond OST 160
4745 Details : llapi_create_file() incorrectly limited the starting stripe
4746 index to the maximum single-file stripe count.
4749 Frequency : LDAP users only
4751 Description: lconf did not handle in-kernel recovery with LDAP properly
4752 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4754 Severity : enhancement
4756 Description: bind OST threads to NUMA nodes to improve performance
4757 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4758 node and do their allocations there to localize memory access
4760 Severity : enhancement
4762 Description: llmount can determine client NID directly from Myrinet (GM)
4763 Details : the client NID code from gmnalnid was moved directly into
4764 llmount, removing the need to use this or specifying the
4765 client NID explicitly when mounting GM clients with zeroconf
4768 Frequency : if client is started with down MDS
4770 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4771 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4772 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4773 ptlrpc_import_delayed_req() because we didn't check if the
4774 request was interrupted, and we also didn't break out of the
4775 event loop if there was no timeout
4780 Description: data loss during non-page-aligned writes to a single file from
4781 both multiple nodes and multiple threads on one node at same time
4782 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4783 inconsistency led to false short-reads, that were cached and later
4784 used by ->prepare_write() to fill in partially written page,
4785 leading to data loss.
4788 Frequency : always, if lconf --abort_recovery used
4790 Description: lconf --abort_recovery fails with 'Operation not supported'
4791 Details : lconf was attempting to abort recovery on the MDT device and not
4794 Severity : enhancement
4796 Description: remove cleanup logs
4797 Details : replace lconf-generated cleanup logs with lustre internal
4798 cleanup routines. Eliminates the need for client-cleanup and
4801 Severity : enhancement
4803 Description: add support for EAs (user and system) on lustre filesystems
4804 Details : it is now possible to store extended attributes in the Lustre
4805 client filesystem, and with the user_xattr mount option it
4806 is possible to allow users to store EAs on their files also
4808 Severity : enhancement
4810 Description: Add possibility (config option) to show minimal available OST free
4812 Details : When compiled with --enable-mindf configure option, statfs(2)
4813 (and so, df) will return least minimal free space available from
4814 all OSTs as amount of free space on FS, instead of summary of
4815 free spaces of all OSTs.
4817 Severity : enhancement
4819 Description: do not expand extent locks acquired on OST-side
4820 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4821 by server: they are not cached anyway.
4824 Frequency : when mmap is used/binaries executed from Lustre
4826 Description: Unmmap pages before throwing them away from read cache.
4827 Details : llap_shrink cache now attempts to unmap pages before discarding
4828 them (if unmapping failed - do not discard). SLES9 kernel has
4829 extra checks that trigger if this unmapping is not done first.
4834 Description: lconf didn't resolve symlinks before checking to see whether a
4835 given mountpoint was already in use
4838 Frequency : when migrating failover services
4839 Bugzilla : 6395, 9514
4840 Description: When migrating a subset of services from a node (e.g. failback
4841 from a failover service node) the remaining services would
4842 time out and evict clients.
4843 Details : lconf --force (implied by --failover) sets the global obd_timeout
4844 to 5 seconds in order to quickly disconnect, but this caused
4845 other RPCs to time out too quickly. Do not change the global
4846 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4848 Severity : enhancement
4849 Frequency : if MDS is started with down OST
4850 Bugzilla : 9439,5706
4851 Description: Allow startup/shutdown of an MDS without depending on the
4852 availability of the OSTs.
4853 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4854 Add appropriate locking and lov-osc refcounts for safe
4855 cleaning. Add osc abort_inflight calls in case the
4856 synchronize never started.
4859 Frequency : occasional (Cray XT3 only)
4861 Description: root not authorized to access files in CRAY_PORTALS environment
4862 Details : The client process capabilities were not honoured on the MDS in
4863 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4864 been accessed by an authorized user then root was able to access
4865 the file on the local client also. The root user capabilities
4866 are now allowed on the MDS, as this environment has secure UID.
4869 Frequency : occasional
4871 Description: ldiskfs "too long searching" message happens too often
4872 Details : A debugging message (otherwise harmless) prints too often on
4873 the OST console. This has been reduced to only happen when
4874 there are fragmentation problems on the filesystem.
4879 Description: Division by zero in statfs when all OSCs are inactive
4880 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4881 lov_statfs divides by value returned by lov_get_stripecnt().
4885 Bugzilla : 9489, 3273
4886 Description: First write from each client to each OST was only 4kB in size,
4887 to initialize client writeback cache, which caused sub-optimal
4888 RPCs and poor layout on disk for the first written file.
4889 Details : Clients now request an initial cache grant at (re)connect time
4890 and so that they can start streaming writes to the cache right
4891 away and always do full-sized RPCs if there is enough data.
4892 If the OST is rebooted the client also re-establishes its grant
4893 so that client cached writes will be honoured under the grant.
4898 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4899 Details : Now I/O RPCs go to different portal number and (presumably) fast
4900 lock requests (and glimses) and other RPCs get their own service
4901 threads pool that should be able to service those RPCs
4904 Severity : enhancement
4906 Description: Ability to exchange lustre version between client and servers and
4907 issue warnings at client side if client is too old. Also for
4908 liblustre clients there is ability to refuse connection of too old
4910 Details : New 'version' field is added to connect data structure that is
4911 filled with version info. That info is later checked by server and
4915 Frequency : rare, liblustre only.
4916 Bugzilla : 9296, 9581
4917 Description: Two simultaneous writes from liblustre at offset within same page
4918 might proceed at the same time overwriting eachother with stale
4920 Details : I/O lock withing llu_file_prwv was released too early, before data
4921 actually was hitting the wire. Extended lock-holding time until
4922 server acknowledges receiving data.
4925 Frequency : extremely rare. Never observed in practice.
4927 Description: avoid generating lustre_handle cookie of 0.
4928 Details : class_handle_hash() generates handle cookies by incrementing
4929 global counter, and can hit 0 occasionaly (this is unlikely, but
4930 not impossible, because initial value of cookie counter is
4931 selected randonly). Value of 0 is used as a sentinel meaning
4932 "unassigned handle" --- avoid it. Also coalesce two critical
4933 sections in this function into one.
4935 Severity : enhancement
4937 Description: allow liblustre clients to delegate truncate locking to OST
4938 Details : To avoid overhead of locking, liblustre client instructs OST to
4939 take extent lock in ost_punch() on client's behalf. New connection
4940 flag is added to handle backward compatibility.
4942 Severity : enhancement
4943 Bugzilla : 4928, 7341, 9758
4944 Description: allow number of OST service threads to be specified
4945 Details : a module parameter allows the number of OST service threads
4946 to be specified via "options ost ost_num_threads={N}" in the
4947 OSS's /etc/modules.conf or /etc/modprobe.conf.
4951 Bugzilla : 6146, 9635, 9895
4952 Description: servers crash with bad pointer in target_handle_connect()
4953 Details : In rare cases when a client is reconnecting it was possible that
4954 the connection request was the last reference for that export.
4955 We would temporarily drop the export reference and get a new
4956 one, but this may have been the last reference and the export
4957 was just destroyed. Get new reference before dropping old one.
4959 Severity : enhancement
4960 Frequency : if client is started with failover MDS
4962 Description: Allow multiple MDS hostnames in the mount command
4963 Details : Try to read the configuration from all specified MDS
4964 hostnames during a client mount in case the "primary"
4967 Severity : enhancement
4969 Description: Stop sending data to evicted clients as soon as possible.
4970 Details : Check if the client we are about to send or are sending data to
4971 was evicted already. (Check is done every second of waiting,
4972 for which l_wait_event interface was extended to allow checking
4973 of exit condition at specified intervals).
4976 Frequency : rare, normally only when NFS exporting is done from client
4978 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4980 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4983 Frequency : always, if srand() is called before liblustre initialization
4985 Description: Liblustre uses system PRNG disturbing its usage by user application
4986 Details : Introduce internal to lustre fast and high-quality PRNG for
4987 lustre usage and make liblustre and some other places in generic
4988 lustre code to use it.
4990 Severity : enhancement
4991 Bugzilla : 9477, 9557, 9870
4992 Description: Verify that the MDS configuration logs are updated when xml is
4993 Details : Check if the .xml configuration logs are newer than the config
4994 logs stored on the MDS and report an error if this is the case.
4995 Request --write-conf, or allow starting with --old_conf.
4997 Severity : enhancement
4999 Description: Handle symlinks in the path when checking if Lustre is mounted.
5000 Details : Resolve intermediate symlinks when checking if a client has
5001 mounted a filesystem to avoid duplicate client mounts.
5006 Description: lconf can hit an error exception but still return success.
5007 Details : The lconf command catches the Command error exception at the top
5008 level script context and will exit with the associated exit
5009 status, but doesn't ensure that this exit status is non-zero.
5014 Description: failure of ptlrpc thread startup can cause oops
5015 Details : Starting a ptlrpc service thread can fail if there are a large
5016 number of threads or the server memory is very fragmented.
5017 Handle this without oopsing.
5020 Frequency : always, only if liblustre and non-default acceptor port was used
5022 Description: liblustre cannot connect to servers with non-default acceptor port
5023 Details : tcpnal_set_default_params() was not called and was therefore
5024 ignoring the environment varaible TCPNAL_PORT, as well as other
5025 TCPNAL_ environment variables
5030 Description: two objects could be created on the same OST for a single file
5031 Details : If an OST is down, in some cases it was possible to create two
5032 objects on a single OST for a single file. No problems other
5033 than potential performance impact and spurious error messages.
5037 Bugzilla : 5681, 9562
5038 Description: Client may oops in ll_unhash_aliases
5039 Details : Client dcache may become inconsistent in race condition.
5040 In some cases "getcwd" can fail if the current directory is
5046 Description: Inode refcounting problems in NFS export code
5047 Details : link_raw functions used to call d_instantiate without obtaining
5048 extra inode reference first.
5052 Bugzilla : 9942, 9903
5053 Description: Referencing freed requests leading to crash, memleaks with NFS.
5054 Details : We used to require that call to ll_revalidate_it was always
5055 followed by ll_lookup_it. Also with revalidate_special() it is
5056 possible to call ll_revalidate_it() twice for the same dentry
5057 even if first occurence returned success. This fix changes semantic
5058 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5059 reference on a request referred from the intent.
5060 ll_intent_release() then releases such a request.
5063 Frequency : rare, normally benchmark loads only
5065 Description: unlinked inodes were kept in memory on the client
5066 Details : If a client is repeatedly creating and unlinking files it
5067 can accumulate a lot of stale inodes in the inode slab cache.
5068 If there is no other client load running this can cause the
5069 client node to run out of memory. Instead flush old inodes
5070 from client cache that have the same inode number as a new inode.
5073 Frequency : SLES9 2.6.5 kernel and long filenames only
5074 Bugzilla : 9969, 10379
5075 Description: utime reports stale NFS file handle
5076 Details : SLES9 uses out-of-dentry names in some cases, which confused
5077 the lustre dentry revalidation. Change it to always use the
5081 Frequency : rare, unless heavy write-truncate concurrency is continuous
5082 Bugzilla : 4180, 6984, 7171, 9963, 9331
5083 Description: OST becomes very slow and/or deadlocked during object unlink
5084 Details : filter_destroy() was holding onto the parent directory lock
5085 while truncating+unlinking objects. For very large objects this
5086 may block other threads for a long time and slow overall OST
5087 responsiveness. It may also be possible to get a lock ordering
5088 deadlock in this case, or run out of journal credits because of
5089 the combined truncate+unlink. Solution is to do object truncate
5090 first in one transaction without parent lock, and then do the
5091 final unlink in a new transaction with the parent lock. This
5092 reduces the lock hold time dramatically.
5095 Frequency : rare, 2.4 kernels only
5097 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5098 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5099 before unmounting their filesystem in order to avoid deadlock.
5100 The kernel_locked() function in 2.4 kernels only checks whether
5101 the kernel lock is held, not whether it is this process that is
5102 holding it as 2.6 kernels do.
5107 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5108 Details : The client ptlrpc code may be trying to reconnect to a down
5109 server before a previous connection attempt has timed out.
5110 Increase the reconnect interval to be longer than the connection
5111 timeout interval to avoid sending duplicate connections to
5115 Frequency : echo_client brw_test command
5117 Description: fix echo_client to work with OST preallocated code
5118 Details : OST preallocation code (5137) didn't take echo_client IO path
5119 into account: echo_client calls filter methods outside of any
5120 OST thread and, hence, there is no per-thread preallocated
5121 pages and buffers to use. Solution: hijack pga pages for IO. As
5122 a byproduct, this avoids unnecessary data copying.
5126 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5127 Description: Client can oops in mdc_commit_close() after open replay
5128 Details : It was possible for the MDS to return an open request with no
5129 transaction number in mds_finish_transno() if the client was
5130 evicted, but without actually returning an error. Clients
5131 would later try to replay that open and may trip an assertion
5132 Simplify the client close codepath, and always return an error
5133 from the MDS in case the open is not successful.
5136 Frequency : rare, 2.6 OSTs only
5138 Description: OST may deadlock under high load on fragmented files
5139 Details : If there was a heavy load and highly-fragmented OST filesystems
5140 it was possible to have all the OST threads deadlock waiting on
5141 allocation of biovecs, because the biovecs were not released
5142 until the entire RPC IO was completed. Instead, release biovecs
5143 as soon as they are complete to ensure forward IO progress.
5145 Severity : enhancement
5147 Description: Support for specifying external journal device at mount
5148 Details : If an OST or MDS device is formatted with an external journal
5149 device, this device major/minor is stored in the ext3 superblock
5150 and may not be valid for failover. Allow detecting and
5151 specifying the external journal at mount time.
5156 Description: Mounting an MDS with pending unlinked files may cause oops
5157 Details : target_finish_recovery() calls mds_postrecov() which returned
5158 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5159 considers this an error and immediately begins cleaning up the
5160 lov, just after starting the mds_lov process
5162 Severity : enhancement
5164 Description: Implement 'lfs df' to report actual free space on per-OST basis
5165 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5166 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5167 usage of objects; '-h' to report in human readable format.
5169 ------------------------------------------------------------------------------
5171 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5178 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5179 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5180 at startup during orphan unlinking. Instead, skip buddy update
5181 before setup, as it will be regenerated after recovery is complete.
5186 Description: If an OST is inactive, its locks might reference stale inodes.
5187 Details : lov_change_cbdata() must iterate over all namespaces, even if
5188 they are inactive to clear inode references from the lock.
5190 Severity : enhancement
5191 Frequency : occasional, if non-standard max_dirty_mb used
5193 Description: Client will block write RPCs if not enough grant
5194 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5195 then the client will block writes while waiting for another RPC
5196 to complete instead of consuming its dirty limit. With change
5197 we get improved performance when max_dirty_mb is small.
5199 Severity : enhancement
5200 Bugzilla : 3389, 6253
5201 Description: Add support for supplementary groups on the MDS.
5202 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5203 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5204 lookups for user supplementary groups into a cache.
5208 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5209 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5210 for liblustre O_EXCL is enforced by MDS.
5214 Description: readdir never returns NULL in liblustre.
5215 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5216 dentry in struct dirent.
5220 Frequency : liblustre only, depends on application IO pattern
5221 Description: liblustre clients evicted if not contacting servers
5222 Details : Don't put liblustre clients into the ping_evictor list, so
5223 they will not be evicted by the pinger ever.
5225 Severity : enhancement
5227 Description: Add ability to evict clients by NID from MDS.
5228 Details : By echoing "nid:$NID" string into
5229 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5230 $NID would be instantly evicted from this MDS and from all active
5231 OSTs connected to it.
5235 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5236 Details : lookup_it_finish() used to query file size from OSTs that was not
5241 Description: service threads change working directory to that of init
5242 Details : Starting lustre service threads may pin the working directory
5243 of the parent thread, making that filesystem busy. Threads
5244 now change to the working directory of init to avoid this.
5248 Frequency : during shutdown only
5249 Description: shutdown with a failed MDS or OST can cause unmount to hang
5250 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5255 Frequency : 2.6 only
5256 Description: chmod/chown may include an extra supplementary group
5257 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5258 supplementary group and if none is specified this is used.
5261 Bugzilla : 5479 (6816)
5263 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5264 Details : If lookup is in progress on a file that is unlinked we might try
5265 to revalidate the inode and fail in revalidate after lookup is
5266 complete and ll_file_open() enqueues the open again but
5267 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5270 Frequency : always, if lconf --abort_recovery used
5272 Description: lconf --abort_recovery fails with 'Operation not supported'
5273 Details : lconf was attempting to abort recovery on the MDT device and not
5276 ------------------------------------------------------------------------------
5278 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5283 Frequency : rare (only unsupported configurations with a node running as an
5285 Bugzilla : 6514, 5137
5286 Description: Mounting a Lustre file system on a node running as an OST could
5288 Details : OSTs now preallocates memory needed to write out data at
5289 startup, instead of when needed, to avoid having to
5290 allocate memory in possibly low memory situations.
5291 Specifically, if the file system is mounted on on OST,
5292 memory pressure could force it to try to write out data,
5293 which it needed to allocate memory to do. Due to the low
5294 memory, it would be unable to do so and the node would
5295 become unresponsive.
5297 Severity : enhancement
5299 Description: Addition of lconf --service command line option
5300 Details : lconf now accepts a '--service <arg>' option, which is
5301 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5303 Severity : enhancement
5305 Description: Failover mode is now the default for OSTs.
5306 Details : By default, OSTs will now run in failover mode. To return to
5307 the old behaviour, add '--failout' to the lmc line for OSTs.
5309 Severity : enhancement
5311 Description: Health checks are now provided for MDS and OSTs
5312 Details : Additional detailed health check information on MSD and OSTs
5313 is now provided through the procfs health_check value.
5316 Frequency : occasional, depends on IO load
5318 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5319 after numerous create/delete cycles
5320 Details : The ext3 inode allocation policy would not allocate new inodes
5321 very well on the OSTs because there are no new directories
5322 being created. Instead we look for groups with free space if
5323 the parent directories are nearly full.
5328 Description: Network or server problems during mount may cause partially
5329 mounted clients instead of returning an error.
5330 Details : The config llog parsing code may overwrite the error return
5331 code during mount error handling, returning success instead
5337 Description: MDS can fail to allocate large reply buffers
5338 Details : After long uptimes the MDS can fail to allocate large reply
5339 buffers (e.g. zconf client mount config records) due to memory
5340 fragmentation or consumption by the buffer cache. Preallocate
5341 some large reply buffers so that these replies can be sent even
5342 under memory pressure.
5346 Frequency : rare (liblustre)
5347 Description: fsx running with liblustre complained that using truncate() to
5348 extend the file doesn't work. This patch corrects that issue.
5349 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5350 ATTR_SIZE and lsm use in llu_setattr_raw.
5354 Frequency : rare, only 2.6 kernels
5355 Description: Unusual file access patterns on the MDS may result in inode
5356 data being lost in very rare circumstances.
5357 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5358 code in the 2.6 kernel caused the inode and/or EA data not to
5359 be read from disk, causing single-file corruption.
5363 Frequency : rare, only 2.6 filesystems using extents
5364 Description: Heavy concurrent write and delete load may cause data corruption.
5365 Details : It was possible under high-load situations to have an extent
5366 metadata block in the block device cache from a just-unlinked
5367 file overwrite a newly-allocated data block. We now unmap any
5368 metadata buffers that alias just-allocated data blocks.
5372 Frequency : filesystems with default stripe_count larger than 77
5373 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5374 Details : lconf specifies an inode size of 4096 bytes when the default
5375 stripe_count is larger than 77. This conflicts with the default
5376 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5377 this case to avoid pinning too much memory for large EAs.
5379 ------------------------------------------------------------------------------
5381 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5386 Frequency : rare (extremely heavy IO load with hundreds of clients)
5388 Description: Client is evicted, gets IO error writing to file
5389 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5390 caused clients to be evicted for AST timeouts. The fixes in
5391 bug 5192 mean we no longer need to have such short AST timeouts
5392 so ldlm_timeout has been increased.
5395 Frequency : occasional during --force or --failover shutdown under load
5396 Bugzilla : 5949, 4834
5397 Description: Server oops/LBUG if stopped with --force or --failover under load
5398 Details : a collection of import/export refcount and cleanup ordering
5399 issues fixed for safer force cleanup
5402 Frequency : only filesystems larger than 120 OSTs
5403 Bugzilla : 5990, 6223
5404 Description: lfs getstripe would oops on a very large filesystem
5405 Details : lov_getconfig used kfree on vmalloc'd memory
5408 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5409 Bugzilla : 6242, 6243
5410 Description: reading from files that had been truncated to a non-zero size
5411 but never opened returned no data
5412 Details : ll_file_read() reads zeros from no-object files to EOF
5417 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5418 Details : There's little harm in aborting MDS/OSS recovery and letting it
5419 try again, so I removed the LASSERT and return an error instead.
5421 Severity : enhancement
5423 Description: New debugging infrastructure for tracking down data corruption
5424 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5425 (b) cover more of the client-side code path, and (c) try to narrow
5426 down where problems occurred
5430 Bugzilla : 3819, 4364, 4397, 6313
5431 Description: Racing close and eviction MDS could cause assertion in mds_close
5432 Details : It was possible to get multiple mfd references during close and
5433 client eviction, leading to one thread referencing a freed mfd.
5435 Severity: : enhancement
5436 Bugzilla : 3262, 6359
5437 Description: Attempts to reconnect to servers are now more aggressive.
5438 Details : This builds on the enhanced upcall-less recovery that was added
5439 in 1.4.2. When trying to reconnect to servers, clients will
5440 now try each server in the failover group every 10 seconds. By
5441 default, clients would previously try one server every 25 seconds.
5446 Description: After recovery, certain operations trigger a failed
5447 assertion on a client.
5448 Details : Failing over an mds, using lconf -d --failover, while a
5449 client was doing a readdir() call would cause the client to
5450 LBUG after recovery completed and the readdir() was resent.
5452 Severity : enhancement
5454 Description: Default groups are now added by lconf
5455 Details : You can now run lconf --group <servicename> without having to
5456 manually add groups with lmc.
5459 Frequency : occasional
5461 Description: Nodes with an elan id of 0 trigger a failed assertion
5464 Frequency : always when accessing e.g. tty/console device nodes
5466 Description: tty and some other devices nodes cannot be used on lustre
5467 Details : file's private_data field is used by device data and lustre
5468 values in there got lost. New field was added to struct file to
5469 store fs-specific private data.
5472 Frequency : when exporting Lustre via NFS
5474 Description: NFSD failed occasionally when looking up a path component
5475 Details : NFSD is looking up ".." which was broken in ext3 directories
5476 that had grown large enough to become hashed.
5479 Frequency : Clusters with multiple interfaces not on the same subnet
5481 Description: Nodes will repeatedly try to reconnect to an interface which it
5482 cannot reach and report an error to the log.
5483 Details : Extra peer list entries will be created by lconf with some peers
5484 unreachable. lconf now validates the peer before adding it.
5487 Frequency : Only if a default stripe is set on the filesystem root.
5489 Description: Setting a default stripe on the filesystem root prevented the
5490 filesystem from being remounted.
5491 Details : The client was sending extra request flags in the root getattr
5492 request and did not allocate a reply buffer for the dir EA.
5495 Frequency : occasional, higher if lots of files are accessed by one client
5496 Bugzilla : 6159, 6097
5497 Description: Client trips assertion regarding lsm mismatch/magic
5498 Details : While revalidating inodes the VFS looks up inodes with ifind()
5499 and in rare cases can find an inode that is being freed.
5500 The ll_test_inode() code will free the lsm during ifind()
5501 when it finds an existing inode and then the VFS later attaches
5502 this free lsm to a new inode.
5506 Bugzilla : 6422, 7030
5507 Description: MDS deadlock between mkdir and client eviction
5508 Details : Creating a new file via mkdir or mknod (starting a transaction
5509 and getting the ns lock) can deadlock with client eviction
5510 (gets ns lock and trying to finish a synchronous transaction).
5513 Frequency : occasional
5514 Description: While starting a server, the fsfilt_ext3 module could not be
5516 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5517 kernels. Previously, lconf would still use the ext3 name
5518 when trying to load modules. Now, it will correctly use
5519 ext3 on 2.4 and ldiskfs on 2.6.
5521 Severity : enhancement
5522 Description: The default stripe count has been changed to 1
5523 Details : The interpretation of the default stripe count (0, to lfs
5524 or lmc) has been changed to mean striping across a single
5525 OST, rather than all available. For general usage we have
5526 found a stripe count of 1 or 2 works best.
5528 Severity : enhancement
5529 Description: Add support for compiling against Cray portals.
5530 Details : Conditional compiling for some areas that are different
5534 Frequency : occasional
5535 Bugzilla : 6409, 6834
5536 Description: Creating files with an explicit stripe count may lead to
5537 a failed assertion on the MDS
5538 Details : If some OSTs are full or unavailable, creating files may
5539 trigger a failed assertion on the MDS. Now, Lustre will
5540 try to use other servers or return an error to the
5544 Frequency : occasional
5546 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5547 caused lock timeout during readahead (since 1.4.2).
5548 Details : Processes doing readahead might match a lock that hasn't been
5549 granted yet if there are overlapping and conflicting lock
5550 requests. The readahead process waits on ungranted lock
5551 (original lock is CBPENDING), while OST waits for that process
5552 to cancel CBPENDING read lock and eventually evicts client.
5554 Severity : enhancement
5556 Description: Initial enabling of flock support for clients
5557 Details : Implements fcntl advisory locking and file status functions.
5558 This feature is provided as an optional mount flag (default
5559 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5560 locking are implemented yet, and those that are are not guaranteed
5561 to be completely correct in production environments.
5562 mount -t lustre -o [flock|noflock] ...
5565 Frequency : occasional
5567 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5568 assertion in the ext3 JBD (journaling) layer.
5569 Details : The b_committed_data struct is protected by the big kernel lock
5570 in 2.4 kernels, serializing journal_commit_transaction() and
5571 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5572 access to this struct. In 2.6 kernels there is finer grained
5573 locking to improve SMP performance of the JBD layer.
5577 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5582 Description: merge and cleanup kernel patches.
5585 -----------------------------------------------------------------------------
5587 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5589 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5590 versions of Lustre. Please update all servers and clients to
5591 version 1.4.2 or later at the same time. You must also run
5592 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5593 to update the configuration logs.
5595 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5596 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5597 - swab llog records of type '0' so we get proper header size/idx (5861)
5598 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5599 - fix rename of one directory over another leaking an inode (5953)
5600 - avoid SetPageDirty on 2.6 (5981)
5601 - don't re-add just-being-destroyed locks to the waiting list (5653)
5602 - when creating new directories, inherit the parent's custom
5603 striping settings if present parent (3048)
5604 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5605 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5606 - don't hold client locks on temporary worklist from l_lru (5666)
5607 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5608 - automatically evict dead clients (5921)
5609 - Update file size properly in create+truncate+fstat case (6196)
5610 - Do not unhash mountpoint dentries, do not allow removal of
5612 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5613 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5614 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5615 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5616 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5617 - Correctly handle reads of files with no objects (6243)
5618 - lctl recover will also mark a device active if deactivate used (5933)
5620 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5621 - allow --write-conf on an MDS with different nettype than client (5619)
5622 - don't write config llogs to MDS for mounts not from that MDS (5617)
5623 - lconf should create multiple TCP connections from a client (5201)
5624 - init scripts are now turned off by default; run chkconfig --on
5625 lustre and chkconfig --on lustrefs to use them
5626 - upcalls are no longer needed for clients to recover to failover
5628 - add --abort-recovery option to lconf to abort recovery on device
5630 - add support for an arbitrary number of OSTs (3026)
5631 - Quota support protocol changes.
5632 - forward compatibility changes to wire structs (6007)
5633 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5634 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5635 - improved "lustre status" script
5636 - initialize blocksize for non-regular files (6062)
5637 - added --disable-server and --disable-client configure options (5782)
5638 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5639 - Vanilla 2.4.29 support
5640 - increase maximum number of obd devices to 520 (6242)
5641 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5642 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5643 SLES 9 2.6.5 kernels
5644 - limit stripes per file to 160 (the maximum EA size) (6093)
5646 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5649 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5650 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5651 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5652 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5653 - small ext3 extents cleanups and fixes (5733)
5654 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5655 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5656 - avoid needless client->OST connect, fix handle mismatch (5317)
5657 - fix DLM error path that led to out-of-sync client, long delays (5779)
5658 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5659 - fix several locking issues related to i_size (5492,5624,5654,5672)
5660 - don't move pending lock onto export if it is already evicted (5683)
5661 - fix kernel oops when creating .foo in unlinked directory (5548)
5662 - fix deadlock in obdfilter statistics vs. object create (5811)
5663 - use time_{before,after} to avoid timer jiffies wrap (5882)
5664 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5665 - Do not leak request if server was not able to process it (5154)
5666 - If mds_open unable to find parent dir, make that negative lookup(5154)
5667 - don't create new directories with extent-mapping (5909, 5936)
5669 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5670 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5671 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5672 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5673 - support for RHEL4 kernel on i686 (5773)
5674 - provide error messages when incompatible logs are encountered (5898)
5676 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5677 * version 1.4.0.10 (1.4.1 release candidate 1)
5679 - don't keep a lock reference when lock is not granted (4238)
5680 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5681 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5682 - drop import inflight refcount on signal_completed_replay error (5255)
5683 - unlock page after async write error during send (3677)
5684 - handle missing objects in filter_preprw_read properly (5265)
5685 - no transno return for symlink open, don't save no-trasno open (3440)
5686 - don't try to complete elan receive that already failed (4012)
5687 - free RPC server reply state on error (5406)
5688 - clean up thread from ptlrpc_start_thread() on error (5160)
5689 - readahead could read extra page into cache that wasn't ejected (5388)
5690 - prevent races in class_attach/setup/cleanup/detach (5260)
5691 - don't dereference de->d_inode after l_dput of de (5458)
5692 - use "int" for stripe value returned from lock_to_stripe (5544)
5693 - mballoc allocation and error-checking fixes in 2.6 (5504)
5694 - block device patches to fix I/O request sizes in 2.6 (5482)
5695 - look up hostnames for IB nals (5602)
5696 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5697 - don't start multiple acceptors for the same port (5277)
5698 - fix incorrect LASSERT in mds_getattr_name (5635)
5699 - export a proc file for general "ping" checking (5628)
5700 - fix "lfs check" to not block when the MDS is down (5628)
5702 - service request history (4965)
5703 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5704 - create an "evict_client" /proc entry on OSTs, like the MDS has
5705 - fix mount usage message, return errors per mount(8) (5168)
5706 - change grep [] to grep "[]" in tests so they work in more UMLs
5707 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5708 - remove ancient LOV_MAGIC_V0 EA support (5047)
5709 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5710 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5711 - allow readahead allocations to fail when low on memory (5383)
5712 - mmap locking landed again, after considerable improvement (2828)
5713 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5715 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5718 - send OST transaction number in read/write reply to free req (4966)
5719 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5720 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5721 - the watchdog thread now runs as interruptible (5246)
5722 - flock/lockf fixes (but it's still disabled, pending 5135)
5723 - don't use EXT3 constants in llite code (5094)
5724 - memory shortage at startup could cause assertion (5176)
5726 - reorganization of lov code
5727 - single portals codebase
5729 - add extents/mballoc support (5025)
5730 - direct I/O reads in the obdfilter (4048)
5731 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5733 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5736 - send OST transaction number in read/write reply to free req (4966)
5737 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5738 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5739 - the watchdog thread now runs as interruptible (5246)
5740 - handle missing objects in filter_preprw_read properly (5265)
5741 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5742 - drop import inflight refcount on signal_completed_replay error (5255)
5743 - unlock page after async write error during send (3677)
5744 - return original error code on reconstructed replies (3761)
5745 - no transno return for symlink open, don't save no-trasno open (3440)
5747 - add pid to ldlm debugging output (4922)
5748 - bump the watchdog timeouts -- we can't handle 30sec yet
5749 - extra debugging for orphan dentry/inode bug (5259)
5751 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5754 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5755 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5756 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5757 - fix NULL dereference in /proc/sys/portals/routes (4827)
5758 - allow failed mdc_close() operations to be interrupted (4561)
5759 - stop precreate on OST before MDS would time out on it (4778)
5760 - don't send partial-page writes before EOF from client (4410)
5761 - discard client grant for sub-page writes on large-page clients (4520)
5762 - don't free dentries not owned by NFS code, check generation (4806)
5763 - fix lsm leak if mds_create_objects() fails (4801)
5764 - limit debug_daemon file size, always print CERROR messages (4789)
5765 - use transno after validating reply (3892)
5766 - process timed out requests if import state changes (3754)
5767 - update mtime on OST during writes, return in glimpse (4829)
5768 - add mkfsoptions to LDAP (4679)
5769 - use ->max_readahead method instead of zapping global ra (5039)
5770 - don't interrupt __l_wait_event() during strace
5772 - add software watchdogs to catch hung threads quickly (4941)
5773 - make lustrefs init script start after nfs is mounted
5774 - fix CWARN/ERROR duplication (4930)
5775 - return async write errors to application if possible (2248)
5776 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5777 - print NAL number in %x format (4645)
5778 - update barely-supported suse-2.4.21-171 series (4842)
5779 - support for sles 9 %post scripts
5780 - support for building 2.6 kernel-source packages
5781 - support for sles km_* packages
5783 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5786 - ignore -ENOENT errors in osc_destroy (3639)
5787 - notify osc create thread that OSC is being cleaned up (4600)
5788 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5789 - reconstruct ost_handle() like mds_handle() (4657)
5790 - create a new thread to do import eviction to avoid deadlock (3969)
5791 - let lconf resolve symlinked-to devices (4629)
5792 - don't unlink "objects" from directory with default EA (4554)
5793 - hold socknal file ref over connect in case target is down (4394)
5794 - allow more than 32000 subdirectories in a single directory (3244)
5795 - fix blocks count for O_DIRECT writes (3751)
5796 - OST returns ENOSPC from object create when no space left (4539)
5797 - don't send truncate RPC if file size isn't changing (4410)
5798 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5799 - bind to privileged port in socknal and tcpnal (3689)
5801 - rate limit CERROR/CWARN console message to avoid overload (4519)
5802 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5803 - basic mmap support (3918)
5804 - kernel patch series update from b1_4 (4711)
5806 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5809 - avoid crash during MDS cleanup with OST shut down (2775)
5810 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5811 - don't use bad inodes on the MDS (3744)
5812 - dynamic object preallocation to improve recovery speed (4236)
5813 - don't hold spinlock over lock dumping or change debug flags (4401)
5814 - don't zero obd_dev when it is force cleaned (3651)
5815 - print grants to console if they go negative (4431)
5816 - "lctl deactivate" will stop automatic recovery attempts (3406)
5817 - look for existing locks in ldlm_handle_enqueue() (3764)
5818 - don't resolve lock handle twice in recovery avoiding race (4401)
5819 - revalidate should check working dir is a directory (4134)
5821 - don't always mark "slow" obdfilter messages as errors (4418)
5823 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5826 - don't close LustreDB during write_conf until it is done (3860)
5827 - fix typo in lconf for_each_profile (3821)
5828 - allow dumping logs from multiple threads at one time (3820)
5829 - don't allow multiple threads in OSC recovery (3812)
5830 - fix debug_size parameters (3864)
5831 - fix mds_postrecov to initialize import for llog ctxt (3121)
5832 - replace config semaphore with spinlock (3306)
5833 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5834 - don't allow enqueue to complete on a destroyed export (3822)
5835 - down write_lock before checking llog header bitmap (3825)
5836 - recover from lock replay timeout (3764)
5837 - up llog sem before sending rpc (3652)
5838 - reduce ns lock hold times when setting kms (3267)
5839 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5840 - fix NULL deref and obd_dev leak on setup error (3312)
5841 - replace some LBUG about llog ops with error handling (3841)
5842 - don't match INVALID dentries from d_lookup and spin (3784)
5843 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5844 - fix invalid assertion in ptlrpc_set_wait (3880)
5846 - add libwrap support for the TCP acceptor (3996)
5847 - add /proc/sys/portals/routes for non-root route listing (3994)
5848 - allow setting MDS UUID in .xml (2580)
5849 - print the stack of a process that LBUGs (4228)
5851 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5854 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5855 - make sure to unset replay flag from failed open requests (3440)
5856 - if default stripe count is 0, use OST count for inode size (3636)
5857 - update parent mtime/ctime on client for create/unlink (2611)
5858 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5859 - free recovery state on server during a forced cleanup (3571)
5860 - unregister_reply for resent reqs (3063)
5861 - loop back devices mounting and status check on 2.6 (3563)
5862 - fix resource-creation race that can provoke i_size == 0 (3513)
5863 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5864 - more debugging for page-accounting assertion (3746)
5865 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5866 - avoid deadlock after precreation failure (3758)
5867 - fix race and lock order deadlock in orphan handling (3450, 3750)
5868 - add validity checks when grabbing inodes from l_ast_data (3599)
5870 - add /proc/.../recovery_status to obdfilter (3428)
5871 - lightweight CDEBUG infrastructure, debug daemon (3668)
5872 - change default OSC RPC parameters to be better on small clusters
5873 - turn off OST read cache for files smaller than 32MB
5874 - install man pages and include them in rpms (3100)
5875 - add new init script for (un)mounting lustre filesystems (2593)
5876 - run chkconfig in %post for init scripts (3701)
5877 - drop scimac NAL (unmaintained)
5879 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5882 - clean kiobufs before and after use (3485)
5883 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5884 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5885 - change init script to fail more clearly if not run as root (1528)
5886 - allow clients to reconnect during replay (1742)
5887 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5888 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5889 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5890 - fix cond_resched (3554)
5891 - only evict unfinished clients after recovery (3515)
5892 - allow bulk resend, prevent data loss (3570)
5893 - dynamic ptlrpc request buffer allocation (2102)
5894 - don't allow unlinking open directory if it isn't empty (2904)
5895 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5896 - remove extraneous obd dereference causing LASSERT failure (3334)
5897 - don't use get_cycles() when creating temp. files on the mds (3156)
5898 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5899 - handle EEXIST for set-stripe, set proper directory name (3336)
5901 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5902 - fix ksocknal_fmb_callback() error messages (2918)
5904 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5907 - don't copy lvb into (possibly NULL) reply on error (2983)
5908 - don't deref dentry after dput, don't free lvb on error (2922)
5909 - use the kms to determine writeback rpc length (2947)
5910 - increment oti_logcookies when osc is inactive (2948)
5911 - update client's i_blocks count via lvb messages (2543)
5912 - handle intent open/close of special files properly (1557)
5913 - mount MDS with errors=remount-ro, like obdfilter (2009)
5914 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5915 - don't use cancelling-locks' kms values (2947)
5916 - use highest lock extent for kms, not last one (2925)
5917 - don't dereference ERR_PTR() dentry in error handling path (3107)
5918 - fix thread race in portals_debug_dumplog() (3122)
5919 - create lprocfs device entries at setup instead of at attach (1519)
5920 - common AST error handler, don't evict client on completion race (3145)
5921 - zero nameidata in detach_mnt in 2.6 (3118)
5922 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5923 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5924 - initialize RPC timeout timer earlier for 2.6 (3219)
5925 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5926 - print nal/nid for unknown nid (3258)
5927 - additional checks for oscc recovery before doing precreate (3284)
5928 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5929 - don't crash in mdc_close for bad permissions on open (3285)
5930 - zero i_rdev for non-device files (3147)
5931 - clear page->private before handing to FS, better assertion (3119)
5932 - tune the read pipeline (3236)
5933 - fix incorrect decref of invalidated dentry (2350)
5934 - provide read-ahead stats and refine rpc in flight stats (3328)
5935 - don't hold journal transaction open across create RPC (3313)
5936 - update atime on MDS at close time (3265)
5937 - close LDAP connection when recovering to avoid server load (3315)
5938 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5939 - don't leak open file on MDS after open resend (3325)
5940 - serialize filter_precreate and filter_destroy_precreated (3329)
5941 - loop device shouldn't call sync_dev() for nul device (3092)
5942 - clear page cache after eviction (2766)
5943 - resynchronize MDS->OST in background (2824)
5944 - refuse to mount the same filesystem twice on same mountpoint (3394)
5945 - allow llmount to create routes for mounting behind routers (3320)
5946 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5947 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5948 - fix rare problem with rename on htree directories (3417)
5950 - allow default OST striping configuration per directory (1414)
5951 - fix compilation for qswnal for 2.6 kernels (3125)
5952 - increase maximum number of MDS request buffers for large systems
5953 - change liblustreapi to be useful for external progs like lfsck (3098)
5954 - increase local configuration timeout for slow disks (3353)
5955 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5957 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5960 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5961 - don't overwrite extent policy data in reply if lock was blocked (2901)
5962 - drop filter export grants atomically with removal from device (2663)
5963 - del obd_self_export from work_list in class_disconnect_exports (2908)
5964 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5965 - swab reply message in mdc_close, other PPC fixes (2464)
5966 - fix destroying of named logs (2325)
5967 - overwrite old logs when running lconf --write_conf (2264)
5968 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5969 - fix race in target_handle_connect (2898)
5970 - mds_reint_create() should take same inode create lock (2926)
5971 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5972 - don't close files for self_export to avoid uninitialized obd (2936)
5973 - allow MDS with the same name as client node (2939)
5974 - hold dentry reference for closed log files for unlink (2325)
5975 - reserve space for all logs during transactions (2059)
5976 - don't evict page beyond end of stripe extent (2925)
5977 - don't oops on a deleted current working directory (2399)
5978 - handle hard links to targets without a parent properly (2517)
5979 - don't dereference NULL lock when racing during eviction (2867)
5980 - don't grow lock extents when lots of conflicting locks (2919)
5982 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5985 - account for cache space usage on clients to avoid data loss (974)
5986 - lfsck support in lustre kernel code (2349)
5987 - reduce journal credits needed for BRW writes (2370)
5988 - orphan handling to avoid losing space on client/server crashes
5989 - ptlrpcd can be blocked, stopping ALL progress (2477)
5990 - use lock value blocks to assist in proper KMS, faster stat (1021)
5991 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5992 - recovery for initial connections (2355)
5993 - fixes for mds_cleanup_orphans (1934)
5994 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5995 - block all file creations until orphan recovery completes (1901)
5996 - client remove rq_connection from request struct (2423)
5997 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5998 - recovery timer race (2670)
5999 - mdc_close recovey bug (2532)
6000 - ptlrpc cleanup bug (2710)
6001 - mds timeout on local locks (2588)
6002 - namespace lock held during RPCs (2431)
6003 - handle interrupted sync write properly (2503)
6004 - don't try to handle a message that hasn't been replied to (2699)
6005 - client assert failure during cleanup after abort recovery (2701)
6006 - leak mdc device after failed mount (2712)
6007 - ptlrpc_check_set allows timedout requests to complete (2714)
6008 - wait for inflight reqs when ptlrpcd finishes (2710)
6009 - make sure unregistered services are removed from the srv_list
6010 - reset bulk XID's when resending them (caught by 1138 test)
6011 - unregister_bulk after timeout
6012 - fix lconf error (2694)
6013 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6014 - readahead locks pages, leaves pending causing memory pressure (2673)
6015 - increase OST request buffers to 4096 on large machines (2729)
6016 - fix up permission of existing directories in simple_mkdir (2661)
6017 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6018 - don't assign transno to errored transactions (2742)
6019 - don't delete objects on OST if given a bogus objid from MDS (2751)
6020 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6021 - if rq_no_resend, then timeout request after recovery (2432)
6022 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6023 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6024 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6025 - MDS recovery shouldn't delete objects that it has given out (2730)
6026 - if enqueue arrives after completion, don't clobber LVB (2819)
6027 - don't unlock pages twice when trigger_group_io returns error (2814)
6028 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6029 - don't write pages to disk if there was an error (1450)
6030 - don't ping imports that have recovery disabled (2676)
6031 - take buffered bytes into account when balancing socknal conn (2817)
6032 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6033 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6034 - remove little-endian swabbing of llog records (1987)
6035 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6036 - retry reposting request buffers if they fail (1191)
6037 - grow extent at grant time to avoid granting a revoked lock (2809)
6038 - lock revoke doesn't evict page if covered by a second lock (2765)
6039 - disable VM readahead to avoid reading outside lock extents (2805)
6041 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6042 - updated kernel patches for hp-2.4.20 kernel (2681)
6044 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6047 - fix truncated write corruption (2366)
6048 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6050 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6051 - ensure len(uuid) < 37 in lmc (1171)
6052 - fix ia64 OOPS in llog_test (2255)
6053 - zero end of page at obdfilter for partial page writes (2648)
6054 - don't leave stale dentries around after renames (bug 2428)
6055 - fix timeouts when evicting a client with a single lock held (2642)
6056 - set deadline for the initial HELLO message to drain (2634)
6057 - print out dotted-quad IP addresses in the socknal (2302)
6059 - additional debugging for MDS client eviction problem (2443)
6060 - fix mkfsoptions support for osts (2603, 2604)
6062 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6065 - add series for the vanilla 2.6.0 kernel
6066 - add series for the vanilla 2.4.24 kernel
6067 - add series for a cray x86/64 UL kernel drop
6068 - fix xattr patches for the vanilla 2.4.19 series
6070 - generate true UUIDs in lmc (1171)
6071 - have portals stack dumping break in UML (2466)
6072 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6073 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6074 - wake callback waiting threads on client eviction (2460)
6075 - Add --ptldebug and --subsystem to lmc (1719)
6076 - update assertion to allow safe interrupt allocation
6077 - set rq_no_resend for cancel requests (2432)
6078 - recalculate ptlrpcd timeout after resend (2494)
6079 - call vfs_rmdir when removing pending directories (2368)
6080 - fix renaming a file to itself (2429)
6081 - lmc creates a default one-stripe lov (2454)
6082 - expand procfs space to handle large clusters (2326)
6083 - increase UML stack to avoid overflow
6084 - update lconf's list of debug and subsystem masks
6085 - fix lfs find --obd (2510)
6086 - /proc tunable for disabling filter read caching (2591)
6087 - stop rpm packages from altering slapd.conf (2301)
6088 - disable nagle in the socknal under 0conf (2578)
6089 - choose mds inode size based on stripe count (2572)
6090 - fix kernel-source rpm problems (2516)
6092 - add --disable-doc to avoid pdf generation (2421)
6093 - update documentation, tests, type-os, comments
6094 - avoid format warnings on ia64
6095 - remove the TOE NAL
6096 - tiny code cleanups by removing unused fields
6098 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6101 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6102 - remove the most blatant lies from BUILDING file (2371)
6103 - change default debug level to reasonable production setting
6104 - reduce client side cache size to reduce cache flush time
6105 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6106 - make TCP zerocopy and pinger support enabled by default (2476)
6107 - sync writes completed after process exits caused crashes (2319)
6108 - maintain correct mount count on the MDS (2356)
6109 - backout 1557, because 2316 wasn't really fixed
6110 - better file I/O statistics gathering in /proc
6111 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6112 - another kernel patch to fix zero-copy TCP function export
6113 - don't take duplicate lock when processing re-sent getattr (2420)
6114 - lctl uses obd_self_export instead of creating new conn (2353)
6115 - MDS/OST recovery case which requires object creation asserted (2425)
6116 - move lfs from /usr/sbin to /usr/bin in packages
6117 - fix race between mds_client_add and mds_client_free (2417)
6118 - use kmalloc instead of slabs in portals (2430)
6119 - don't create duplicate records when a failover MDS is present (2442)
6120 - remove unnecessary mount age check (2332)
6121 - don't remove directory inodes from locks prematurely (2451)
6122 - don't break if MDS service name is the same as hostname (2103)
6123 - fix races in client write RPC generation when cache full (2482)
6125 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6128 - remove now-unused request->rq_obd (278)
6129 - if an allocation fails, print out how much memory we've used (1933)
6130 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6131 - add the "configurable stack size" patch to most series files (1256)
6132 - ability to write large log records, for 100+ OST configs (2306)
6133 - fix NULL deref when filter_prep fails (2314)
6134 - fix operator precedence error in filter_sync
6135 - dynamic allocation of socknal TX descriptors (2315)
6136 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6137 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6138 - ensure that CWARN messages in Portals always get to the syslog
6139 - __init/__exit are not for prototype decls (ldlm_init/exit)
6140 - x86-64 compile warning fixes
6141 - fix gateway LMC keyword conflict (2318)
6142 - fix MDS lock inversions in getattr/reint paths (1844)
6143 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6144 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6145 - DLM race condition prevented some lock evictions (2328)
6146 - ENOMEM detection and retry on socknal sends (2230)
6147 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6148 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6149 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6150 - better error messages when a client is rejected during recovery (1505)
6151 - avoid cancelling locks which were never granted, after failure (2330)
6152 - fix i_sem/journal inversion in mds_client_add (2333)
6153 - fix truncate/getattr lock cycle deadlock (2334)
6154 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6155 - fix two rare exit paths which could leak an l_lock() ref (2321)
6156 - fencepost error in MDS/OST orphan recovery (2226)
6157 - make log record alignment 8 bytes (1988)
6158 - lstripe now fails when requested offset > ost_count (2237)
6159 - ensure that all kernel series have a complete list.h (1607)
6160 - fix crashes in special-file operations (2316)
6161 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6163 - allow configurable automake binary, for testing new versions
6164 - small update to the lfs documentation
6166 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6168 * fix negative export reference count in fsfilt_sync (2312)
6170 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6171 * release candidate 0.9.1
6173 - orphans are moved into the PENDING directory for possible recovery
6174 - replayed opens now open by fid for orphan/rename safety (1042)
6175 - last close of an orphan inode generates a transno (683)
6176 - chdir() and mount() now pin the directory entry (1020)
6177 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6178 - discard very old requests without processing them (1502)
6179 - remove some common, well-understood CERRORs (1505)
6180 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6181 - clear "grant" flags in OST replies until OST grant code lands (1644)
6182 - fix read performance by not clobbering i_blksize on client (1598)
6183 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6184 - make LRU size a /proc tunable, clears locks when reduced (707)
6185 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6186 - server locks take references on exports now (1558)
6187 - build fixes for 2.4.20-rh trees (1663)
6188 - return an error from lov_create if all OSCs are inactive (1751)
6189 - fix import levels when a reconnect happens without a timeout (1597)
6190 - exit early from mds_open if we get a lookup error (1749)
6191 - partial page read at EOF wouldn't wait for disk before sending (1642)
6192 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6193 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6194 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6195 - write/truncate lock inversion (1639)
6196 - Don't auto-load obdclass, portals modules during cleanup (1495)
6197 - fix timestamps from jumping to "now" (1763)
6198 - extra journal assertions (1648)
6199 - add an extra multiunlink test (1771)
6200 - fix read_record/write_record API (1776)
6201 - fix leak of offset_extent, possible incorrect i_size later (1772)
6202 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6203 - Debugging for the kqswnal_get_idle_tx problems (1820)
6204 - Allow recovery to be attempted multiple times (1536)
6205 - Write out MDS last_rcvd file after it is first created (1600)
6206 - Fix tx_descriptor leak in failed transmit situations (1827)
6207 - ext3 journaling fixes for assertion failure after IO error (1871)
6208 - class_export_put() on freed export after completion AST error (1896)
6209 - Fix revalidate looping in VFS (1322)
6210 - Don't access a freed export during MDS_REINT timeout (1521)
6211 - Add open-unlink recovery support on the MDS (1673,1764)
6212 - Return an error if no MDS data was read from last_rcvd (1946)
6213 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6214 - Don't setup a disk device that doesn't match exported UUID (317)
6215 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6216 - avoid committing NULL handle in force close
6217 - local.sh is now a one-stripe LOV configuration
6218 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6219 - fix ext3 htree duplicate directory entry corruption (1516)
6220 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6221 - update to new LOV EA format (2097)
6222 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6224 2003-06-15 Phil Schwan <phil@clusterfs.com>
6227 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6228 - per-import/export recovery handling (958, 931, 959)
6229 - multiple last-rcvd slots, for serving multiple FSes (949)
6230 - connections are again shared between multiple imp/exports (963, 964)
6231 - "umount -f" would hang if any requests needed to be sent (393, 978)
6232 - avoid pinning large req buffer by copying for queued messages (989)
6233 - add "uuid" to "lctl device" command to help upcalls (991)
6234 - "open" RPCs with transnos would confuse recovery counters (1037)
6235 - do proper endian conversion of all wire messages (288, 340, 891)
6236 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6237 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6238 - fix ptlrpc_connection leak in target_handle_connect (1174)
6239 - fix import refcounting bug in OST and MDS cleanup (1134)
6240 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6241 - fix very unlikely obd_types race condition (501)
6242 - remove osc_open hack for echo_client (1187)
6243 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6244 - a failure in read_inode2 leads to deadlock (1139)
6245 - cancel ack-locks as soon as transaction is committed (1072)
6246 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6247 - make sure to commitrw after any preprw to avoid deadlock (1162)
6248 - failing to execute a file in a lustre FS would lock inode (1203)
6249 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6250 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6251 - obd_uuid2tgt would walk off the end of the list (1255)
6252 - on IA64 the osc would give portals incorrect bulk size (1258)
6253 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6254 - fix lock inversion caused by new llite matching code (1282)
6255 - limit the number of dirty pages on a client to 10MB (1286)
6256 - timed out locks were not being corrected cancelled (1289)
6257 - fix O_DIRECT above 4GB on IA-32 (1292)
6258 * major user-visible changes
6259 - fail out/fail over policy now controlled by the upcall (993)
6261 - add OBD_PING to check server availability and failure (954)
6262 - lustre messages are now sent in sending host order (288, 340, 891)
6263 - add eadatalen to MDS getattr reply (340)
6264 - OST read replies may contain second buffer, with per-page status (593)
6266 2003-03-11 Phil Schwan <phil@clusterfs.com>
6269 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6270 - failed open()s could cause deadlock; fixed (867, 869)
6271 - stop cancelling OST locks when files are closed (481)
6272 - overlapping XID spaces caused network corruption (851, 853)
6273 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6274 - setattr_raw added, to do single-RPC, server-side setattrs
6275 - lmc/lconf syntax change for OST UUIDs
6276 - fix crashy race condition between ptlrpc_free_req and osc_close
6277 - don't use request in mdc_enqueue if we hit a timeout (889)
6278 - don't set the inode i_size for regular files from the MDS (896)
6279 - handle out of order completion AST (842)
6280 - don't LBUG if a lock request times out after receiving AST (913)
6281 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6282 - if a bad lock AST arrives, send an error instead of dropping entirely
6283 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6284 - fix leak in bulk IO when only partially completed (899, 900, 926)
6285 - fix O_DIRECT for ia64 (55)
6286 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6287 - C-z after timeout could hang a process forever; fixed (977)
6289 - client-side I/O cache (678, 924, 929, 941, 970)
6291 - READPAGE and SETATTRs which don't take server-side locks get
6294 2003-02-11 Phil Schwan <phil@clusterfs.com>
6297 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6298 - Fix fsfilt_extN_readpage() to read a full page of directory
6299 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6300 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6301 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6302 - Most intent operations were removed, in favour of a new RPC mode
6303 that does a single RPC to the server and bypasses most of the VFS
6304 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6305 - Aggressively cancel local locks on DLM servers
6306 - mds_reint_unlink sends EA to the client if it's the last nlink.
6307 client uses that EA to unlink OST objects.
6308 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6309 - recursive symlinks were fixed (439)
6310 - fixed NULL deref in DEBUG_REQ
6311 - filter_update_lastobjid no longer calls sync, which annoyed extN
6312 - fixed multi-client small-writes to a single file problem (445)
6313 - fixed mtime updates during file writes (607)
6314 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6315 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6316 - fixed Linux OST size reporting problem (444, 656)
6317 - OST now updates object mtime with writes or setattr (607, 619)
6318 - client verifies file size before zeroing page past EOF (445)
6319 - OST now writes last allocated objid to disk with allocation (108)
6320 - LOV on echo now works (409)
6322 - mds_reint_unlink sends a new buffer, with the EA included. this
6323 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6324 set if a regular file was being unlinked, and it was the last link
6325 - use PtlGet from the target for bulk writes (315)
6326 - OST now updates object mtime with writes or setattr (607, 619)
6327 - LDLM now has a grant-time callback to revalidate locked items, if
6329 - Many MDS operations were reorganized to combat race conditions
6331 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6332 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6334 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6337 - Fully reactivate OST imports after reconnection (512, others)
6338 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6339 - More graceful error handling for truncating on dead OST (515)
6340 - Don't error out unless we're actually accessing dead stripes (474)
6341 - Fix garbage sizes when stripes are missing (410)
6342 - LRU counters were broken, causing constant lock purge (433, 432)
6343 - garbage on read from stripes with failed OSTs (441)
6344 - mark OSCs as active before reconnecting during recovery (438)
6345 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6346 - lfind did not preserve OST order in output (443)
6347 - symlinks cause hung clients, incorrect data (439)
6348 - stop dereferencing request after dropping refcount (457)
6349 - don't LASSERT(spin_is_locked) on non-SMP (455)
6350 - fixes for many rename() bugs
6351 - fstat didn't correctly synchronize attributes (399)
6352 - server must handle lock cancellation during blocking AST prep (487)
6353 - bulk descriptors were free()d too soon (511)
6354 - fix paths in lconf, which would load incorrect modules (451, 507)
6355 - fix confusing lconf 'host not found' error message (386)
6356 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6357 - fix race condition in mdc_blocking_ast() for inode access (526)
6358 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6359 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6360 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6361 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6362 - fix lstripe in several ways (kernel side) (527)
6363 - fix request leak in ldlm_cli_enqueue (262)
6364 - incorrect OSC was marked inactive after OST failure
6365 - call mds_fs_cleanup before unmounting filesystem (524)
6366 - fix races between taking ns_lock and ldlm_lock_change_resource
6367 - fix races updating LOV export open file list
6368 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6369 - fix recovery NULL deref in ldlm_cli_cancel_unused
6370 - fix some DLM races by using new hash table for lock handles (419)
6371 - permit the client to specify desired inodes, at replay
6372 - duplicate requests when we queue them for replay reintegration
6373 - fix last_rcvd offset calculation
6374 - sync after each recovered transaction, so we always make progress
6375 - never, not always, ERESTART requests without transnos
6376 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6378 - skip replay if the MDS says that the client is already connected
6379 - don't check for a recovery-enabled export to match lctl's UUID
6380 - don't INC_USE_COUNT for phantom exports
6381 - don't crash when cleaning up phantom exports (567)
6382 - don't double-finish or set replay data for errored mdc_open requests
6383 - abort requests when they time out, so we don't get old replies
6384 - send/receive replies for AST messages again
6385 - if the client says that it doesn't have the lock, cancel it on the
6387 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6388 mark it as destroyed, it will all work out when decref is called
6389 - fix module use counts (22, 581)
6391 - ASTs now expect a reply (server cancels lock on error reply)
6393 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6396 - fix many simultaneous client startup (392)
6397 - fix dentry->d_it clobbering
6398 - credentials weren't being shipped for readdir/getattr operations
6399 - remove invalid assertions triggered during some concurrent MD
6401 - proper Lustre versions added (336, 389)
6402 - fix memory leak for create error case (398)
6403 - fix LOV locking bug that would get cli/srv out of sync
6404 - fix echo client over LOV (409)
6405 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6406 - fix double-O_EXCL intent crash (424)
6407 - avoid sending multiple lock CANCELs (352)
6409 - MDS can do multi-client recovery (modulo bugs in new code)
6411 - many updates, edits, cleanups
6413 2002-11-18 Phil Schwan <phil@clusterfs.com>
6416 - fix null d_it dereference (346)
6417 - fix full OST/dbench hang (333)
6418 - fix permission problem with file removal (286)
6419 - fix removal of OSCs from LOV when they fail
6420 - fix NULL deref during bulk timeout (214)
6421 - fix problems related to multiple filesystems on one MDS (241)
6422 - fixed serious subtle metadata locking bugs
6423 - free locks on clients when inodes are removed due to memory
6425 - fix inode pointer in lock data (285)
6426 - partial support for multiple MDS on a single host (241)
6427 - data locks weren't cancelled at clear_inode time (290, 311)
6428 - intent locks could lead to unbounded lock growth (205)
6429 - added a maximum lock count, an LRU list, and a flusher
6430 - fix multiple rename (365)
6431 - properly abstracted the echo client
6432 - OSC locked 1 byte too many; fixed
6433 - rewrote brw callback code:
6434 - fixed recovery bugs related to LOVs (306)
6435 - fixed too-many-pages-in-one-write crash (191)
6436 - fixed (again) crash in sync_io_timeout (214)
6437 - probably fixed callback-related race (385)
6439 - Add capability to MDS protocol
6440 - LDLM cancellations and callbacks on different portals
6442 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6445 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6446 - defer open object destruction to close time (601981, 138)
6447 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6448 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6449 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6450 - don't call obd_setattr in ll_file_release for destroyed objects
6451 * protocol change to lustre_msg: move |version| and add |flags|
6452 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6453 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6454 * added replay of create, unlink, link and rename operations during
6455 MDS failover; recovery should be much more robust now
6456 * remove failed OSCs from LOVs (only lov_create uses this so far)
6457 * the lustre-HOWTO was brought (more) up to date (582544)
6459 2002-10-23 Phil Schwan <phil@clusterfs.com>
6462 - in-use dentries weren't being reused properly (617851)
6463 - prevent multiple LDLM setup (599178)
6464 - fix LOV size calculations for truncate (617853)
6465 - fix client handling of MDS intent errors (POSIX)
6466 - fix permission bug in lovstripe.c test (624321)
6467 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6468 - truncate past end of file could corrupt data
6469 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6470 - a race in recovery could return ETIMEDOUT to apps (623947)
6471 - building outside the source directory was fixed
6472 * the lustre-HOWTO was brought (more) up to date (582544)
6473 * major progress was made on recovery functionality
6475 2002-10-10 Phil Schwan <phil@clusterfs.com>
6478 - recovery deadlock fix
6479 - rm -rf causes LBUG fix (617817)
6480 - file open by multiple tasks fix (618962)
6481 - directory permissions bugs (602707 and 620007)
6482 - journal_stop fixed with locking (611313)
6483 - O_APPEND failures resolved (618273, perhaps 614459)
6484 - lconf PATH fix (619770)
6485 - IA64 build fix (621450)
6486 - RPC buffer sizes scale with amount of memory
6488 2002-10-01 Phil Schwan <phil@clusterfs.com>
6491 - locks would be cancelled without throwing away data pages,
6492 resulting in inconsistent data (605627)
6493 - inode attributes were not always being refreshed (605627, 612449)
6494 - lconf now continues to cleanup after lctl reports an error
6495 - MDS now enforces user permissions (602707)
6496 - lprocfs cleanup fixed, but not yet enabled (614157)
6497 - fixed infinite server hang, should a client not respond to an AST
6498 - avoid going into recovery if user calls readlink() with a buffer
6499 that's too small (613941)
6500 - AST RPCs no longer require replies (614867) -- this may be changed
6501 - don't crash server if client sends an IOV that's too big (611336)
6502 - fixed lock conversion deadlock (611892)
6503 - fixed the following of symlinks (614622)
6504 * recovery: the server can remove locks from a client that dies, other
6505 clients can make progress
6506 * more extN patch fixes
6507 * compile-time configurable ptlrpc buffer allocations
6509 - collaborative read cache document
6510 - Lustre Lite Performance CDR document-in-progress
6512 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6515 - fix typo in patch-2.4.18
6517 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6520 - clear ptlrpc request each time in handle_incoming_request()
6521 - unlink of files now destroys the object on the OST
6523 2002-09-19 Peter Braam <braam@clusterfs.com>
6525 * add hard link support
6526 * change obdfile creation method
6527 * kernel patch changed
6529 2002-09-19 Peter Braam <braam@clusterfs.com>
6532 - stack overflow bug in extN fixed
6534 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6536 * documentation updates
6537 - add man pages for config tools
6538 - update tests/README to describe testing with new config tools
6539 - finish metadata API descriptions
6540 * bug fixes and cleanups
6541 - statfs workaround for 16TB limit
6542 - LOV stripe allocation improved, can stripe on subset of OSTs
6543 - LOV file size/IO offset was wrong for files > 4GB in size
6544 - object EA data was being dropped, caused files to be unreadable
6545 - memory overflow with non-LOV OST caused memory corruption
6546 - fixed regression tests to work with new config tools, obdfilter
6547 - fixed bug when directory size became larger than 1 block
6548 - fixed bug (for single client case) when PWD was deleted
6549 - invalidate local directory pages when doing intent-based ops
6550 - avoid LDLM oops when lock callback contained bad data
6552 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6554 * documentation updates
6555 * bug fixes and cleanups
6556 - configuration tools
6559 - 64-bit compile warnings
6560 - 64-bit internal statfs data
6562 * test_brw on persistent OST devices
6564 * lprocfs (disabled)
6566 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6568 * documentation updates
6569 * bug fixes and cleanups
6570 * configuration tools
6572 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6575 * many small fixes to 0_5_4
6576 * io/network handling
6577 * thinkos in MDS operations
6579 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6582 * crucial basic fixes to 0.5.3
6583 * IOR, Iozone work over Elan
6586 2002-08-07 Phil Schwan <phil@clusterfs.com>
6587 * version 0_5_3, our first alpha
6588 * we use the new Portals iovs
6589 * documentation updates
6590 * bug fixes and cleanups
6591 * small changes in the DLM wire protocol
6593 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6594 * version 0_5_1 with some initial stability,
6595 * locking on MD and file I/O.
6596 * documentation updates
6597 * several bug fixes since 0.5.0
6598 * small changes in wire protocol
6600 2002-07-18 Phil Schwan <phil@clusterfs.com>
6602 * delivered as Lustre Light Alpha
6603 * fixed a crash after handling invalid MDS requests
6604 * fixed directory pages for architectures with non-4k pages sizes
6606 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6607 * release version v0_4_4
6608 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6610 * Fixes a number of interruption problems with OST operations.
6611 * Update documentation for portals header changes
6612 * Move all wire protocol structs/defines to lustre_idl.h
6613 * Fixes symlink length bug.
6614 * Add tcpdump to repository.
6616 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6617 * release version v0_4_3
6618 * Fixes statfs for inodes on extN.
6619 * Fixes bug in runtests which would delete /etc/hosts.
6620 * Use 64-bit object IDs wherever possible (not into VFS though)
6621 Remove ost_get_info, which is unused by lustre, and out of date.
6623 2002-07-03 Peter Braam <braam@clusterfs.com>
6624 * release version v0_4_2 Fixes a lookup error (type not passed)
6625 * move forward to head of Portals
6626 * move forward to latest Lustre kernel
6628 2002-06-25 Peter Braam <braam@clusterfs.com>
6629 * release version v0_4_1. Hopefully stable on single node use.