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 2.6.32-696.6.3.el6 (RHEL6.9)
11 3.10.0-514.26.2.el7 (RHEL7.3)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-100 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.74-92.29 (SLES12 SP2)
16 vanilla linux 4.6.7 (ZFS only)
17 * Client known to build on unpatched kernels:
18 2.6.32-431.29.2.el6 (RHEL6.5)
19 2.6.32-504.30.3.el6 (RHEL6.6)
20 2.6.32-573.26.1.el6 (RHEL6.7)
21 2.6.32-642.15.1.el6 (RHEL6.8)
22 2.6.32-696.6.3.el6 (RHEL6.9)
23 3.10.0-514.26.2.el7 (RHEL7.3)
24 3.0.101-0.47.71 (SLES11 SP3)
25 3.0.101-100 (SLES11 SP4)
26 3.12.74-60.64.40 (SLES12 SP1)
27 4.4.74-92.29 (SLES12 SP2)
29 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
30 * Recommended ZFS / SPL version: 0.6.5.9
31 * Tested with ZFS / SPL version: 0.6.5.9
32 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
33 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
34 stack overflow. For more information, please refer to bugzilla 17630.
35 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
36 the re-exporting nodes to be mounted with "32bitapi" mount option
38 --------------------------------------------------------------------------------
40 2017-07-05 Intel Corporation
42 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
43 for currently supported client and server kernel versions.
44 * Server known to build on patched kernels:
45 2.6.32-431.29.2.el6 (RHEL6.5)
46 2.6.32-504.30.3.el6 (RHEL6.6)
47 2.6.32-573.26.1.el6 (RHEL6.7)
48 2.6.32-642.15.1.el6 (RHEL6.8)
49 2.6.32-696.3.1.el6 (RHEL6.9)
50 3.10.0-514.21.1.el7 (RHEL7.3)
51 3.0.101-0.47.71 (SLES11 SP3)
52 3.0.101-107 (SLES11 SP4)
53 3.12.74-60.64.40 (SLES12 SP1)
54 4.4.59-92.17 (SLES12 SP2)
55 vanilla linux 4.6.7 (ZFS only)
56 * Client known to build on unpatched kernels:
57 2.6.32-431.29.2.el6 (RHEL6.5)
58 2.6.32-504.30.3.el6 (RHEL6.6)
59 2.6.32-573.26.1.el6 (RHEL6.7)
60 2.6.32-642.15.1.el6 (RHEL6.8)
61 2.6.32-696.3.1.el6 (RHEL6.9)
62 3.10.0-514.21.1.el7 (RHEL7.3)
63 3.0.101-0.47.71 (SLES11 SP3)
64 3.0.101-107 (SLES11 SP4)
65 3.12.74-60.64.40 (SLES12 SP1)
66 4.4.59-92.17 (SLES12 SP2)
68 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
69 * Recommended ZFS / SPL version: 0.6.5.9
70 * Tested with ZFS / SPL version: 0.6.5.9
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
77 --------------------------------------------------------------------------------
79 12-07-2016 Intel Corporation
81 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
82 for currently supported client and server kernel versions.
83 * Server known to build on patched kernels:
84 2.6.32-431.29.2.el6 (RHEL6.5)
85 2.6.32-504.30.3.el6 (RHEL6.6)
86 2.6.32-573.26.1.el6 (RHEL6.7)
87 2.6.32-642.11.1.el6 (RHEL6.8)
88 3.10.0-514.el7 (RHEL7.3)
89 3.0.101-0.47.71 (SLES11 SP3)
90 3.0.101-80 (SLES11 SP4)
91 3.12.62-60.64.8 (SLES12 SP1)
92 vanilla linux 4.5.7 (ZFS only)
93 * Client known to build on unpatched kernels:
94 2.6.32-431.29.2.el6 (RHEL6.5)
95 2.6.32-504.30.3.el6 (RHEL6.6)
96 2.6.32-573.26.1.el6 (RHEL6.7)
97 2.6.32-642.11.1.el6 (RHEL6.8)
98 3.10.0-514.el7 (RHEL7.3)
99 3.0.101-0.47.71 (SLES11 SP3)
100 3.0.101-80 (SLES11 SP4)
101 3.12.62-60.64.8 (SLES12 SP1)
103 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
104 * Recommended ZFS / SPL version: 0.6.5.8
105 * Tested with ZFS / SPL version: 0.6.5.8
106 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
107 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
108 stack overflow. For more information, please refer to bugzilla 17630.
109 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
110 the re-exporting nodes to be mounted with "32bitapi" mount option
112 --------------------------------------------------------------------------------
114 02-29-2016 Intel Corporation
116 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
117 for currently supported client and server kernel versions.
118 * Server known to build on patched kernels:
119 2.6.32-431.29.2.el6 (RHEL6.5)
120 2.6.32-504.30.3.el6 (RHEL6.6)
121 2.6.32-573.12.1.el6 (RHEL6.7)
122 3.10.0-327.3.1.el7 (RHEL7.2)
123 3.0.101-0.47.71 (SLES11 SP3)
124 3.0.101-68 (SLES11 SP4)
125 vanilla linux 4.2.1 (ZFS only)
126 * Client known to build on unpatched kernels:
127 2.6.32-431.29.2.el6 (RHEL6.5)
128 2.6.32-504.30.3.el6 (RHEL6.6)
129 2.6.32-573.12.1.el6 (RHEL6.7)
130 3.10.0-327.3.1.el7 (RHEL7.2)
131 3.0.101-0.47.71 (SLES11 SP3)
132 3.0.101-68 (SLES11 SP4)
135 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
136 * Recommended ZFS / SPL version: 0.6.4.2
137 * Tested with ZFS / SPL version: 0.6.4.2
138 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
139 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
140 stack overflow. For more information, please refer to bugzilla 17630.
141 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
142 the re-exporting nodes to be mounted with "32bitapi" mount option
144 --------------------------------------------------------------------------------
146 03-10-2015 Intel Corporation
148 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
149 for currently supported client and server kernel versions.
150 * Server known to build on patched kernels:
151 2.6.32-431.29.2.el6 (RHEL6.5)
152 2.6.32-504.8.1.el6 (RHEL6.6)
153 3.0.101-0.46 (SLES11 SP3)
154 * Client known to build on unpatched kernels:
155 2.6.32-431.29.2.el6 (RHEL6.5)
156 2.6.32-504.8.1.el6 (RHEL6.6)
157 3.10.0-123.20.1.el7 (RHEL7)
158 3.0.101-0.46 (SLES11 SP3)
159 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
160 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
161 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
162 stack overflow. For more information, please refer to bugzilla 17630.
163 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
164 the re-exporting nodes to be mounted with "32bitapi" mount option
166 Severity : enhancement
168 Description: control OST-index in IDIF via ROCOMPAT flag.
169 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
170 last_rcvd file. For new formatted OST device, it will be auto set;
171 for the case of upgrading from old OST device, you can enable it
172 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
173 enabled, for new created OST-object, its IDIF-in-LMA will contain
174 the OST-index; for the existing OST-object, the OSD will convert
175 old format IDIF as new format IDIF with OST-index stored in the
176 LMA EA when accessing such OST-object or via OI scrub. Once such
177 flag is enabled, it cannot be reverted back, so the system cannot
178 be downgraded to the orignal incompatible version.
180 --------------------------------------------------------------------------------
182 07-30-2014 Intel Corporation
184 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
185 for currently supported client and server kernel versions.
186 * Server known to build on patched kernels:
187 2.6.32-431.20.3.el6 (RHEL6.5)
188 3.0.101-0.31 (SLES11 SP3)
189 * Client known to build on unpatched kernels:
190 2.6.32-431.20.3.el6 (RHEL6.5)
191 3.0.101-0.31 (SLES11 SP3)
194 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
195 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
196 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
197 stack overflow. For more information, please refer to bugzilla 17630.
198 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
199 the re-exporting nodes to be mounted with "32bitapi" mount option
201 --------------------------------------------------------------------------------
203 10-31-2013 Intel Corporation
205 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
206 * Server builds for kernels:
207 2.6.32-358.18.1.el6 (RHEL6.4)
208 * Client builds for unpatched kernels:
209 2.6.32-358.18.1.el6 (RHEL6.4)
210 3.0.92-0.8 (SLES11 SP3)
212 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
213 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
214 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
215 stack overflow. For more information, please refer to bugzilla 17630.
216 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
217 the re-exporting nodes to be mounted with "32bitapi" mount option
219 --------------------------------------------------------------------------------
221 05-30-2013 Intel Corporation
223 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
224 * Server builds for kernels:
225 2.6.32-358.6.2.el6 (RHEL6.4)
226 * Client builds for unpatched kernels:
227 2.6.32-358.6.2.el6 (RHEL6.4)
228 3.0.74-0.6.6 (SLES11 SP2)
230 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
231 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
232 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
233 stack overflow. For more information, please refer to bugzilla 17630.
234 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
235 the re-exporting nodes to be mounted with "32bitapi" mount option
237 --------------------------------------------------------------------------------
239 9-30-2012 Whamcloud, Inc.
241 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
242 * Server support for kernels:
243 2.6.32-279.5.1.el6 (RHEL6)
244 * Client support for unpatched kernels:
245 2.6.32-279.5.1.el6 (RHEL6)
247 2.6.18-238.19.1.el5 (RHEL5)
248 2.6.16.60-0.69.1 (SLES10)
249 * Recommended e2fsprogs version: 1.41.90.wc4
250 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
251 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
252 stack overflow. For more information, please refer to bugzilla 17630.
253 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
254 the re-exporting nodes to be mounted with "32bitapi" mount option
256 Severity : enhancement
258 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
260 Severity : enhancement
262 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
264 Severity : enhancement
266 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
268 Severity : enhancement
270 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
272 Severity : enhancement
274 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
276 --------------------------------------------------------------------------------
278 04-30-2012 Whamcloud, Inc.
280 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
281 * Server support for kernels:
282 2.6.32-220.4.2.el6 (RHEL6)
283 * Client support for unpatched kernels:
284 2.6.32-220.4.2.el6 (RHEL6)
286 2.6.18-238.19.1.el5 (RHEL5)
287 2.6.16.60-0.69.1 (SLES10)
288 * Recommended e2fsprogs version: 1.41.90.wc4
289 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
290 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
291 stack overflow. For more information, please refer to bugzilla 17630.
292 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
293 the re-exporting nodes to be mounted with "32bitapi" mount option
295 Severity : enhancement
297 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
299 Severity : enhancement
301 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
303 Severity : enhancement
305 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
307 Severity : enhancement
309 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
311 --------------------------------------------------------------------------------
313 09-30-2011 Whamcloud, Inc.
315 * Server support for kernels:
316 2.6.18-238.19.1.el5 (RHEL5)
317 2.6.32-131.6.1.el6 (RHEL6)
318 * Client support for unpatched kernels:
319 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
320 2.6.18-238.19.1.el5 (RHEL5)
321 2.6.32-131.6.1.el6 (RHEL6)
322 2.6.29.4-167.fc11 (FC11)
323 2.6.16.60-0.69.1 (SLES10)
325 * Recommended e2fsprogs version: 1.41.90.wc3
326 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
327 Lustre filesystem with 4K stack may cause a stack overflow. For more
328 information, please refer to bugzilla 17630.
329 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
330 mounted with "32bitapi" mount option
332 Severity : enhancement
333 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
335 Severity : enhancement
336 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
338 Severity : enhancement
339 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
342 Frequency : only with SLES10
344 Description: Use OFED "KMP" provided by Novell
345 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
346 Lustre is now built against this package. That means you need to
347 install the ofed-kmp package from Novell for the patchless client.
348 Note that the ofed-kmp that Novell ships may not exactly match the
349 kernel version but should still be compatible.
351 Severity : enhancement
353 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
354 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
356 --------------------------------------------------------------------------------
358 2010-07-15 Oracle, Inc.
360 * Server support for kernels:
361 2.6.18-164.11.1.el5 (RHEL 5)
362 2.6.18-164.11.1.0.1.el5 (OEL 5)
363 * Client support for unpatched kernels:
364 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
365 2.6.18-164.11.1.el5 (RHEL 5),
366 2.6.18-164.11.1.0.1.el5 (OEL 5)
367 2.6.16.60-0.42.8 (SLES 10),
369 2.6.29.4-167.fc11 (FC11)
370 * Recommended e2fsprogs version: 1.41.10-sun2
372 Severity : enhancement
374 Description: Distribute lfs_migrate script with lustre
375 Details : The lfs_migrate mentioned in the manual is now included in the
376 lustre tree in lustre/scripts and will distributed in lustre RPMs.
378 Severity : enhancement
380 Description: Failover nid list added to import proc file.
382 Severity : enhancement
384 Description: "som_preview" mount option required to use SOM
385 Details : The clients will need to use som_preview mount option
389 Description: Lustre does not do 1MB IOs to HW RAID
390 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
392 Severity : enhancement
394 Description: lfs quota output is not very convenient for awk/sed-parsing
395 Details : Some positions in lfs quota output table could be empty or
396 non-empty which made it hard to parse it with scripts, now
397 a dash is put instead of space where there is not supposed
402 Description: Update to OFED 1.5.1
403 Details : OFED 1.5.1 will be used in environments that don't use
404 vendor supplied OFED.
408 Description: Remove lnet. prefix from lctl params display
409 Details : Also renamed lnet.memused to lnet_memused to
414 Description: Use conf_param -d to remove permanant settings.
418 Description: Add list_param -R
419 Details : lctl list_param now supports recursive mode.
422 Bugzilla : 21829, 22312
423 Description: Fix llobdstat.
424 Details : 1. Now updating the write_bytes counter
425 2. Port script fixes from 1.8.
429 Description: decrease the usage of memory on clients.
430 Details : 1. On clients, recycle dentries and inodes unused.
431 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
436 Description: llobdstat fix and enhancement
437 Details : add a counter to set a limit to how many samples will be returned
438 fix a wildcard in the path to limit to obdfilter stats only
442 Description: lfs check only allowed for root
443 Details : Code cleanups and added a regresion test of non-root use of
448 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
449 Details : Fix hr->hr_index overload.
453 Description: Fix FIEMAP functionality.
457 Description: FC11 patchless client support added
461 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
462 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
466 Description: New acc-sm script and various test-framework enhancments.
470 Description: Previously landed patch was reverted.
474 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
478 Description: Add ldiskfs maxdirsize mount option.
482 Description: Fix mkfs.lustre support for 16TB LUNs.
486 Description: Print mmp_check_interval
487 Details : print mmp_check_interval and make it possible to abort
488 mount operation in case it takes too long.
492 Description: Add cascading_rw to lustre mpi tests.
496 Description: Store SOM data in EA.
497 Details : Also several SOM related cleanups.
501 Description: Add refcount to md_open_data
505 Description: Use refcount on lov_request_set to prevent memory corruption.
509 Description: Statahead: use parent dir's i_mutex for synchronization.
513 Description: loadgen improvements.
517 Description: Enable peer health detection in ksocklnd and ko2iblnd.
521 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
525 Description: Fix race on obd_devcie:md_stats
529 Description: Fix SLES 11 client builds.
530 Details : Resolved uninitalized pointer gcc warnings.
534 Description: Allow non-root access for "lfs check".
535 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
539 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
541 Severity : enhancement
543 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
544 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
548 Description: Sanity test_27p fails.
549 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
550 Sanity tests 27 were cleaned up and improved.
554 Description: Unify req format on client/servers
555 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
556 buffers and swabbers approach..
561 Description: drop join file stuff.
563 Severity : enhancement
565 Description: Add 2.6.27 and SLES11 (patchless client) support
569 Description: Fix in ptlrpc_expire_one_request() to print the signed time
572 Severity : enhancement
574 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
577 Severity : enhancement
579 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
581 Severity : enhancement
583 Description: correctly handle big reply message.
584 Details : send LNet event if reply is bigger than buffer and adjust this buffer
589 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
591 Severity : enhancement
593 Description: Add support for OEL5.
597 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
599 Severity : enhancement
601 Description: "weak-modules" support
602 Details : Implement "weak-modules" support which enables kernel modules
603 to be used with any kernel that implements the same kABI. In
604 order to achieve this modules are now installed in
605 /lib/modules/$(uname -r)/updates/kernel on all distributions.
609 Description: client could not reconnect to OST because of an active request.
610 Details : The client now retries to reconnect to the same server, if a
611 connect request failed with EBUSY or -EAGAIN. Test case 35b
612 conf-sanity.sh has been created.
616 Description: Don't allow make backward step on assignin osc next id.
617 Details : race between allocation next id and ll_sync thread can be cause
618 of set wrong osc next id and can be kill valid ost objects.
622 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
626 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
630 Description: Error handling in osc_statfs_interpret() has been improved.
631 Details : Check in osc_statfs_interpret() for EBADR.
635 Description: Do not send statfs() requests to OSTs disabled by administrator.
636 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
638 Severity : enhancement
640 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
642 Severity : enhancement
644 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
646 Severity : enhancement
648 Description: Enhancement for lfs(1) command to use numeric uid/gid.
652 Description: Improved cleanup procedure for conf-sanity test_46a.
656 Description: Do not start bulk transfer, if client tried to reconnect.
657 Details : Do not start bulk transfer, if client tried to reconnect,
658 but failed due to active rpcs. Check for exp_abort_active_req
659 has been implemented.
663 Description: Prevent inconsistences between linux and lustre mount structures.
664 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
665 Make the sleep interruptible.
667 Severity : enhancement
669 Description: provide server to client comms path
670 Details : server to client communications path via new LDLM_SET_INFO rpc
674 Description: truncate starts GFP_FS allocation under transaction causing deadlock
675 Details : ldiskfs_truncate calls grab_cache_page which may start page
676 allocation under an open transaction. This may lead to
677 calling prune_icache with consequent lustre reentrance.
681 Description: send correctly lsm on open replay
682 Details : MDS is trust to LSM size on replay open, but client can set wrong size
687 Description: enable client interface failover
688 Details : When a child reconnects from another NID, properly update export
689 nid hash position and ldlm reverse import.
691 Severity : enhancement
693 Description: Communicate OST degraded/readonly state via statfs to MDS
694 Details : Flags in the statfs returned from OSTs indicate whether the
695 OST is in a degraded RAID state, or if the filesystem has
696 turned read-only after a filesystem error is detected.
701 Description: don't panic if EPROTO was hit when reading symlink
702 Details : correctly handling request reference in error cases.
704 Severity : enhancement
706 Description: Add LustreNetLink, a kernel-userspace communcation path.
708 Severity : enhancement
710 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
715 Description: access to llog context before init.
716 Details : move handling CATALOGS file at osc layer and forbid access to llog
720 Frequency : rare, only when client is writing to OST on local node
722 Description: Avoid deadlock for local client writes
723 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
724 memory freeing context. This allows OST threads to set the
725 PF_MEMALLOC flag on task structures in order to allocate memory
726 from reserved pools and complete IO.
727 Use GFP_HIGHUSER for OST allocations for non-local client writes,
728 so that the OST threads generate memory pressure and allow
729 inactive pages to be reclaimed.
731 Severity : enhancement
733 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
738 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
739 Details : move ldlm namespace creation in setup phase to avoid grab
740 _lprocfs_lock with cli_sem held.
744 Description: Temporarily disable grant shrink.
745 Details : Disable the feature for debugging.
748 Frequency : only when formatting test filesystems
750 Description: Unable to run several mkfs.lustre on loop devices at the same time
751 Details : mkfs.lustre returns error 256 on the concurrent loop devices
752 formatting. The solution is to proper handle the error.
754 Severity : enhancement
756 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
758 Severity : enhancement
760 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
765 Description: resolve race between obd_disconnect and class_disconnect_exports
766 Details : if obd_disconnect will be called to already disconnected export he
767 forget release one reference and osc module can't unloaded.
769 Severity : enhancement
771 Description: Allow tuning service thread via /proc
772 Details : For each service a new
773 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
774 created that can be used to set min/max thread counts, and get the
775 current number of running threads.
780 Descriptoin: don't return error if have partially created objects for file.
781 Details : lov_update_create_set uses set->set_success as index for created
782 objects, so if some requests will be failed, they will have hole
783 at end of array and qos_shrink_lsm for allocate correct lsm.
785 Severity : enhancement
787 Description: Update OFED support to 1.4.1
792 Description: Reduce small size read RPC
793 Details : Set read-ahead limit for every file and only do read-ahead when
794 available read-ahead pages are bigger than 1M to avoid small size
797 Severity : enhancement
799 Description: move AT tunable parameters for more consistent usage
800 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
802 Severity : enhancement
804 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
805 Details : allow skip disconnected ost for send statfs request and hide error
811 Description: fix racy locking of mballoc block bitmaps causing BUG
812 Details : The locking of the mballoc buddy bitmap and the in-memory
813 block bitmap was using two different spin locks in some
814 cases. This made it possible to incorrectly access the
815 mballoc bitmap while another process was modifying it,
816 causing a sanity assertion to fail. While no on-disk corruption
817 was reported, there was some risk of this happening.
820 Frequency : rare, on llog test 6
822 Descriptoin: don't allow connect to already connected import
823 Details : allowing connect to already connected import is hide connecting problem.
828 Descriptoin: allow kill process which wait statahead result
829 Details : for some reasons 'ls' can stick in waiting result from statahead,
830 in this case need way for kill this process.
832 Severity : enhancement
834 Description: Add state history info file, enhance import info file
835 Details : Track import connection state changes in a new osc/mdc proc file;
836 add overview-type data to the osc/mdc import proc file.
838 Severity : enhancement
840 Description: MDS create should not wait for statfs RPC while holding DLM lock.
842 Severity : enhancement
844 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
849 Descriptoin: OST grant too much space to client even there are not enough space.
850 Details : Client will shrink its grant cache to OST if there are no write
851 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
852 this grant cache if there are already not enough avaible space
853 (left_space < total_clients * 32M).
856 Frequency : start MDS on uncleanly shutdowned MDS device
858 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
859 Details : stay in waiting mds<>ost recovery finished produce random bugs
860 due race between two ll_sync thread for one lov target. send
861 ACTIVATE event only if connect realy finished and import have
865 Frequency : rare, connect and disconnect target at same time
867 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
868 Details : don't call obd_disconnect under lov_lock. this long time
869 operation and can block ptlrpcd which answer to connect request.
874 Descriptoin: don't lose wakeup for imp_recovery_waitq
875 Details : recover_import_no_retry or invalidate_import and import_close can
876 both sleep on imp_recovery_waitq, but we was send only one wakeup
880 Frequency : always with long access acl
882 Descriptoin: mds can't pack reply with long acl.
883 Details : mds don't control size of acl but they limited by reint/getattr
886 Severity : enhancement
888 Description: Update to SLES10 kernel-2.6.16.60-0.33.
890 Severity : enhancement
892 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
895 Frequency : start MDS on uncleanly shutdowned MDS device
897 Descriptoin: aborting recovery hang on MDS
898 Details : don't throttle destroy RPCs for the MDT.
901 Frequency : on remount
903 Description: external journal device not working after the remount
904 Details : clear dev_rdonly flag for external journal devices in
910 Description: shutdown vs evict race
911 Details : client_disconnect_export vs connect request race.
912 if client will evicted at this time - we start invalidate
913 thread without referece to import and import can be freed
917 Frequency : rare, need acl's on inode.
919 Description: client can't handle ost additional correctly
920 Details : if ost was added after client connected to mds client can have
921 hit lnet_try_match_md ... to big messages to wide striped files.
922 in this case need teach client to handle config events about add
923 lov target and update client max ea size at that event.
925 Severity : enhancement
927 Description: Changelogs
928 Details : Changelogs are a lightweight mechanism to track filesystem
929 metadata and namespace changes. The changelog is recorded
930 permanently on the MDTs, and is periodically "consumed" / purged
931 when records are no longer needed.
933 Severity : enhancement
935 Description: compact fld format with extents
936 Details : Store range of seq rather than every seq in FLD. Seq
937 controller update FLD rather than clients. In Case of CMD, mdt0
938 has FLD, all other metadata server act as non persistent proxy
939 for FLD queries and cache fld entries in fld cache.
944 Description: don't skip ost target if they assigned to file
945 Details : Drop slow OSCs if we can, but not for requested start idx.
946 This means "if OSC is slow and it is not the requested
947 start OST, then it can be skipped, otherwise skip it only
948 if it is inactive/recovering/out-of-space.
952 Description: more cleanup in mds_lov
953 Details : not send LOV EA under replay, we can't know about they size at this
954 time. Don't allow client connect to mds before any ost connected,
955 for avoid problems with LOV EA size and returning EIO to client.
957 Severity : enhancement
959 Description: Interoperability at server side (Disk interoperability)
961 Severity : enhancement
963 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
965 Severity : enhancement
967 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
969 Severity : enhancement
971 Description: New client IO stack (CLIO).
973 Severity : enhancement
975 Description: Commit on sharing. Eliminate inter-client dependencies between
976 uncommitted transactions by doing transaction commits.
977 Thereby clients may recovery independently.
980 Frequency : Create a symlink file with a very long name
982 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
983 Details : If there is no extra space in the request for early cancels,
984 ldlm_req_handles_avail() returns 0 instead of a negative value.
986 Severity : enhancement
988 Description: Add /proc entry for import status
989 Details : The mdc, osc, and mgc import directories now have
990 an import directory that contains useful import data for debugging
993 Severity : enhancement
995 Description: Re-disable certain /proc logging
996 Details : Enable and disable client's offset_stats, extents_stats and
997 extents_stats_per_process stats logging on the fly.
1000 Frequency : Only on FC kernels 2.6.22+
1002 Description: oops in statahead
1003 Details : Do not drop reference count for the dentry from VFS when lookup,
1004 VFS will do that by itself.
1006 Severity : enhancement
1008 Description: Generic /proc file permissions
1009 Details : Set /Proc file permissions in a more generic way to enable non-
1010 root users operate on some /proc files.
1014 Description: Hitting mdc_commit_close() ASSERTION
1015 Details : Properly handle request reference release in
1016 ll_release_openhandle().
1020 Description: quota recovery deadlock during mds failover
1021 Details : This patch includes att18982, att18236, att18237 in bz14840.
1023 1. fix osts hang when mds does failover with quotaon
1024 2. prevent watchdog storm when osts threads wait for the
1029 Frequency : only patchless client
1030 Description: add workaround for race between add/remove dentry from hash
1032 Severity : enhancement
1034 Description: Allow OST glimpses to return PW locks
1038 Description: LBUG when llog conf file is full
1039 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1043 Description: Prevent import from entering FULL state when server in recovery
1047 Description: service mount cannot take device name with ":"
1048 Details : Only when device name contains ":/" will mount treat it as
1054 Description: replace ptlrpcd with the statahead thread to interpret the async
1055 statahead RPC callback
1059 Frequency : on recovery
1060 Description: I/O failures after umount during fail back
1061 Details : if client reconnected to restarted server we need join to recovery
1062 instead of find server handler is changed and process self
1063 eviction with cancel all locks.
1065 Severity : enhancement
1067 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1069 Severity : enhancement
1071 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1073 Severity : enhancement
1075 Description: Upcall on Lustre log has been dumped
1076 Details : Allow for a user mode script to be called once a Lustre log has
1077 been dumped. It passes the filename of the dumped log to the
1078 script, the location of the script can be specified via
1079 /proc/sys/lnet/debug_log_upcall.
1084 Description: avoid idr_remove called for id which is not allocated.
1085 Details : Move assigment s_dev for clustered nfs to end of initialization,
1086 to avoid problem with error handling.
1091 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1092 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1093 into CONN_USED_HASH and this prodice warning when put connection
1094 again in unused hash.
1097 Severity : enhancement
1099 Description: File striping can now be set to use an arbitrary pool of OSTs.
1101 Severity : enhancement
1103 Description: Export bytes_read/bytes_write count on OSC/OST.
1107 Description: Early reply size mismatch, MGC loses connection
1108 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1109 the connect flags are properly negotiated.
1114 Description: quotas are not honored with O_DIRECT
1115 Details : all writes with the flag O_DIRECT will use grants which leads to
1116 this problem. Now using OBD_BRW_SYNC to guard this.
1120 Description: add quota statistics
1121 Details : 1. sort out quota proc entries and proc code.
1122 2. add quota statistics
1124 Severity : enhancement
1126 Description: enable quota support for HEAD.
1130 Description: Properly propagate oinfo flags from lov to osc for statfs
1131 Details : restore missing copy oi_flags to lov requests.
1133 Severity : enhancement
1135 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1136 : l_getgroups(8), lst(8), routerstat(8)
1137 Details : included man pages for llobdstat(8), llstat(8),
1138 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1140 Severity : enhancement
1142 Description: configure's --enable-quota should check the
1143 : kernel .config for CONFIG_QUOTA
1144 Details : configure is terminated if --enable-quota is passed but
1145 : no quota support is in kernel
1149 Description: Remove portals compatibility
1150 Details : Remove portals compatibility, not interoperable with releases
1155 Description: Resolve device initialization race
1156 Details : Prevent proc handler from accessing devices added to the
1157 obd_devs array but yet be intialized.
1159 Severity : enhancement
1161 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1163 Severity : enhancement
1165 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1170 Description: Using wrong pointer in osc_brw_prep_request
1171 Details : Access to array[-1] can produce panic if kernel compiled with
1172 CONFIG_PAGE_ALLOC enabled
1176 Description: Client runs out of low memory
1177 Details : Consider only lowmem when counting initial number of llap pages
1181 Description: *optional* service tags registration
1182 Details : if the "service tags" package is installed on a Lustre node
1183 When the filesystem is mounted, a local-node service tag will
1184 be created. See http://inventory.sun.com/ for more information
1185 about the Service Tags asset management system.
1189 Description: Kernel BUG tries to release flock
1190 Details : Lustre does not destroy flock lock before last reference goes
1191 away. So always drop flock locks when client is evicted and
1192 perform unlock regardless of successfulness of speaking to MDS.
1196 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1200 Description: sanity test 65a fails if stripecount of -1 is set
1201 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1206 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1207 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1210 Severity : enhancement
1212 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1216 Description: do not process already freed flock
1217 Details : flock can possibly be freed by another thread before it reaches
1218 to ldlm_flock_completion_ast.
1222 Description: LBUG during stress test
1223 Details : Need properly lock accesses the flock deadlock detection list.
1227 Description: oops in page fault handler
1228 Details : kernel page fault handler can return two special 'pages' in error
1229 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1233 Description: timeout with invalidate import.
1234 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1235 should be handled by ptlrpcd. This produce long age waiting and
1236 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1238 Severity : enhancement
1240 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1244 Description: enable MGS and MDT services start separately
1245 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1246 a co-located MGS without starting the MGS, which is a complement
1247 to 'nosvc' mount option.
1251 Frequency : after recovery
1252 Description: precreate to many object's after del orphan.
1253 Details : del orphan st in oscc last_id == next_id and this triger growing
1254 count of precreated objects. Set flag LOW to skip increase count
1255 of precreated objects.
1259 Frequency : rare, on clear nid stats
1260 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1261 Details : when clean nid stats sometimes try destroy live entry,
1262 and this produce panic in free.
1266 Description: Stack overflow during MDS log replay
1267 ease stack pressure by using a thread dealing llog_process.
1271 Description: wait until IO finished before start new when do lock cancel.
1272 Details : VM protocol want old IO finished before start new, in this case
1273 need wait until PG_writeback is cleared until check dirty flag and
1274 call writepages in lock cancel callback.
1276 Severity : enhancement
1278 Description: using special macro for print time and cleanup in includes.
1282 Description: mds_mfd_close() ASSERTION(rc == 0)
1283 Details : In mds_mfd_close(), we need protect inode's writecount change
1284 within its orphan write semaphore to prevent possible races.
1288 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1292 Frequency : rare, on shutdown ost
1293 Description: don't hit live lock with umount ost.
1294 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1295 use shrink_dcache_sb instead.
1299 Description: don't panic with use echo client
1300 Details : echo client pass NULL as client nid pointer and this produce null
1301 pointer dereference.
1305 Description: fix build on ppc32
1306 Details : compile code with -m64 flag produce wrong object file for ppc32.
1310 Description: add message levels for liblustreapi
1314 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1315 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1316 will result in an -EAGAIN and in the exit call path we call
1317 alloc_rr() which will with increasing aggressiveness attempt to
1318 aquire precreated objects on the minimum number of required OSCs.
1322 Description: Use old size assignment to avoid deadlock
1323 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1324 the scheduling while holding a spinlock. We do not need locking
1325 for size in ll_update_inode() because size is only updated from
1326 the MDS for directories or files without objects, so there is no
1327 other place to do the update, and concurrent access to such inodes
1328 are protected by the inode lock.
1332 Description: resolve "_IOWR redefined" build error on SLES10
1336 Description: dump the memory debugging after all modules are unloaded to
1337 suppress false negative in conf_sanity test 39
1339 Severity : enhancement
1341 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1346 Description: fix SLES kernel versioning
1347 Details : the kernel version for our SLES 10 kernel did not include a "-"
1348 before the "smp" at the end. while this was not a problem in
1349 general, it did mean that software trying to use the kernel
1350 version to try to detect a vendor specific kernel would fail.
1351 this was most evident by the OFED build scripts.
1355 Description: Don't update lov_desc members until making sure they are valid
1356 Details : When updating lov_desc members via proc fs, need fix their
1357 validities before doing the real update.
1361 Description: don't put request into delay list while invalidate in flight.
1362 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1363 import in flight. this produce timeout for invalidate and sometimes
1364 can cause stale data.
1368 Frequency : on ppc only
1369 Description: not convert ost objects for directory because it's not exist.
1370 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1374 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1376 Severity : enhancement
1378 Description: Update to support 2.6.22.14 vanilla kernel.
1382 Frequency : rare, on recovery
1383 Description: read procfs can produce deadlock in some situation
1384 Details : Holding lprocfs lock which send rpc can produce block for destroy
1385 obd objects and this also block reconnect with -EALREADY. This
1386 isn't fix all lprocfs bugs - but make it rare.
1388 Severity : enhancement
1390 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1393 Frequency : frequent on X2 node
1395 Description: mdc_set_open_replay_data LBUG
1396 Details : Set replay data for requests that are eligible for replay.
1400 Description: lustre_mgs: operation 101 on unconnected MGS
1401 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1402 MGC, and late on MGC cannot successfully connect to MGS and a lot
1403 of the error messages complaining that MGS is not connected.
1407 Frequency : on network error
1408 Description: panic with double free request if network error
1409 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1410 but it's true only for synchronus enqueue, for async enqueue
1411 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1414 Severity : enhancement
1416 Description: client-side metadata stat-ahead during readdir(directory readahead)
1417 Details : perform client-side metadata stat-ahead when the client detects
1418 readdir and sequential stat of dir entries therein
1421 Frequency : on start mds
1423 Description: Implement get_info(last_id) in obdfilter.
1426 Frequency : occasional
1428 Description: Correctly check stale fid, not start epoch if ost not support SOM
1429 Details : open with flag O_CREATE need set old fid in op_fid3 because
1430 op_fid2 was overwritten with new generated fid, but mds can answer
1431 with one of these two fids and both is not stale. Setattr
1432 incorrectly started an epoch and assume will be called
1433 done_writing, but without SOM done_writing ever being called.
1436 Frequency : rare, depends on device drivers and load
1438 Description: MDS or OSS nodes crash due to stack overflow
1439 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1440 In some cases, in conjunction with device drivers that use a lot
1441 of stack the MDS (or possibly OSS) service threads could overflow
1442 the stack. One change which was identified to consume additional
1443 stack has been reworked to avoid the extra stack usage.
1446 Frequency : occasional
1448 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1449 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1450 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1451 oscc_has_objects and see OSC has no objests and call
1452 oscc_internal_create to resend create request.
1454 Severity : enhancement
1456 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1458 Severity : enhancement
1460 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1463 Frequency : very rare
1465 Description: Fix replay if there is an un-replied request and open
1466 Details : In some cases, older replay request will revert the
1467 mcd->mcd_last_xid on MDS which is used to record the client's
1468 latest sent request.
1470 Severity : enhancement
1472 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1474 Severity : enhancement
1476 Description: Add rhel5 support to HEAD.
1478 Serverity : enhancement
1480 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1485 Description: Don't allow skipping OSTs if index has been specified.
1486 Details : Don't allow skipping OSTs if index has been specified, make
1487 locking in internal create lots better.
1491 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1492 Details : access to bitfield in structure is always rounded to long
1493 and this produce problem with not atomic change any bit.
1497 Description: Lustre make rpms failed.
1498 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1499 builds Lustre rpms from tarball.
1502 Frequency : rare on shutdown ost
1504 Description: If llog cancel was not send before clean_exports phase, this can
1505 produce deadlock in llog code.
1506 Details : If llog thread has last reference to obd and call class_import_put
1507 this produce deadlock because llog_cleanup_commit_master wait when
1508 last llog_commit_thread exited, but this never success because was
1509 called from llog_commit_thread.
1513 Description: allow userland application know is lost one of stripes.
1514 Details : fill lvb_blocks with error code on ost and return it to
1515 application if error flag found.
1519 Description: NULL lov_tgts causing MDS oops
1520 Details : more safe checks for NULL lov_tgts for avoid oops.
1522 Severity : enhancement
1524 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1528 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1529 Details : If an OST is down the MDS will hang indefinitely in
1530 obd_statfs_rqset() waiting for the statfs data. While for
1531 MDS QOS usage of statfs, it should not stuck in waiting.
1533 Severity : enhancement
1535 Description: remote_acl support
1536 Details : Support ACL-based permission check for remote user.
1537 Support setfacl/getfacl for remote user with the utils
1538 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1539 the system "{s,g}etfacl" utils.
1541 Severity : enhancement
1543 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1545 Severity : enhancement
1547 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1551 Description: make mgs_setparam() handle fsname containing dash
1552 Details : fsname containing a dash does not work with lctl conf_param
1554 Severity : enhancement
1556 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1558 Severity : enhancement
1560 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1564 Frequency : rare, at shutdown
1565 Description: access already free / zero obd_namespace.
1566 Details : if client_disconnect_export was called without force flag set,
1567 and exist connect request in flight, this can produce access to
1568 NULL pointer (or already free pointer) when connect_interpret
1569 store ocd flags in obd_namespace.
1573 Frequency : only at startup
1574 Description: not alloc memory with spinlock held.
1575 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1576 if any spinlock held.
1578 Severity : enhancement
1580 Description: make lustre randomly fail allocating memory
1581 Details : Make lustre randomly failed allocating memory for testing purpose.
1583 Severity : enhancement
1585 Description: lost problems with lov objid file
1586 Details : Fixes some scability and access to not inited memory problems
1587 in work with lov objdid file.
1592 Description: lfs find does not continue on file error
1593 Details : Continue other files processing when a file/dir is absent.
1597 Description: Inconsistent usage of lustre_pack_reply()
1598 Details : Standardize the usage of lustre_pack_reply() such that it
1599 always generate a CERROR on failure.
1602 Frequency : occasional
1604 Description: MDS hang or stay in waiting lock
1605 Details : If client receive lock with CBPENDING flag ldlm need send lock
1606 cancel as separate rpc, to avoid situation when cancel request
1607 can't processed due all i/o threads stay in wait lock.
1611 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1615 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1619 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1623 Description: Update to RHEL4 latest kernel.
1625 Severity : enhancement
1627 Description: Build SLES10 patchless client fails
1628 Details : The configure was broken by run ./configure with
1629 --with-linux-obj=.... argument for patchless client. When the
1630 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1631 can't search header adequately. Use absolute path such as
1632 -I($LINUX)/include instead.
1636 Description: interrupt oig_wait produce painc on resend.
1637 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1638 private set, and this produce situation when rq_ptlrpcd_data not
1639 copyed to new allocated request and triggered LBUG on assert
1640 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1641 ptlrpcd set if request is changed and can be safety changed to use
1646 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1647 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1648 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1649 these macros and replace them with their difinitions instead.
1653 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1654 Details : Update which_patch & target file for SLES10 latest kernel.
1657 Description: Remove client patches from SLES 10 kernel.
1658 Details : This causes SLES 10 clients to behave as patchless clients
1659 even on a Lustre-patched (server) kernel.
1661 Severity : enhancement
1663 Description: self-adjustable client's lru lists
1664 Details : use adaptive algorithm for managing client cached locks lru
1665 lists according to current server load, other client's work
1666 pattern, memory activities, etc. Both, server and client
1667 side namespaces provide number of proc tunables for controlling
1670 Severity : enhancement
1672 Description: light-weight GSS support
1673 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1674 and reduce performance overhead.
1676 Severity : enhancement
1678 Description: Linux keyring support
1679 Details : Support using service of Linux keyring for Lustre GSS internal
1680 context refresh/cache mechanism.
1684 Description: Fix errors in lfs documentation
1685 Details : Fixes man pages
1689 Description: don't use GFP_* in generic Lustre code.
1690 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1694 Description: obdclass is limited by single OBD_ALLOC(idarray)
1695 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1699 Description: warnings with build patchless client with vanila 2.6.19 and up
1700 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1701 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1705 Frequency : Only for SLES
1706 Description: sanity_quota fail test_1
1707 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1708 file, which makes TSTID[2] inunique.
1712 Frequency : Only for SLES9
1713 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1714 Details : Update target/ChangeLog/which_patch .
1718 Frequency : Only for test_10 in sanity-quota.sh
1719 Description: a bug in quota test code
1720 Details : checking a test flag in a wrong place
1724 Frequency : Only for test_13 in sanity-quota.sh
1725 Description: a bug in quota test script
1726 Details : shouldn't check overfull allocated quota
1730 Description: jbd statistics
1731 Details : Port older jbd statistics patch for sles10
1733 Severity : enhancement
1735 Description: Add printing inode info into message about error in writepage.
1740 Description: reply_lock_interpret crash due to race with it and lock cancel.
1741 Details : Do not replay locks that are being cancelled. Do not reference
1742 locks by their address during replay, just by their handle.
1746 Frequency : When flocks are used.
1747 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1748 Details : Flock locks might destroy just granted lock if it could be merged
1749 with another existing flock, this is done in completion handler,
1750 so teach ldlm_cli_enquque_fini that this is a valid case for
1756 Description: Oops in read and write path when failing to allocate lock.
1757 Details : Check if lock allocation failed and return error back.
1761 Description: lstripe command fails for valid OST index
1762 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1763 instead of lov->desc.ld_active_tgt_count.
1767 Description: sanity.sh failed test 103
1768 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1773 Description: df doesn't work properly if diskfs blocksize != 4K
1774 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1778 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1779 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1781 Severity : enhancement
1783 Description: Adaptive timeouts
1784 Details : RPC timeouts adapt to changing server load and network
1785 conditions to reduce resend attempts and improve recovery time.
1789 Description: llapi_file_create() does not allow some changes
1790 Details : add llapi_file_open() that allows specifying the mode and
1791 open flags, and also returns an open file handle.
1795 Description: merge and cleanup kernel patches.
1796 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1800 Description: Add journal checksum support.(Kernel part)
1801 Details : The journal checksum feature adds two new flags i.e
1802 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1803 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1804 indicates that the commit block contains the checksum for
1805 the blocks described by the descriptor blocks. Now commit
1806 record can be sent to disk without waiting for descriptor
1807 blocks to be written to disk. This behavior is controlled
1808 using JBD2_FEATURE_ASYNC_COMMIT flag.
1812 Description: OSS needs mutliple precreate threads
1813 Details : Add ability to start more than one create thread per OSS.
1817 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1818 Details : execute lfs setstripe on client
1822 Description: mds_obd_create error creating tmp object
1823 Details : When the user sets quota on root, llog will be affected and can't
1824 create files and write files.
1827 Frequency : Always on ia64 patchless client, and possibly others.
1829 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1830 Details : This allows the patchless client to be loaded on architectures
1831 without this export.
1835 Description: RedHat Update kernel for RHEL5
1836 Details : Add the kernel config file for RHEL5.
1840 Description: RedHat Update kernel for RHEL5
1841 Details : Modify the kernel config file more closer RHEL5.
1845 Description: Build failure against Centos5 (RHEL5)
1846 Details : Define PAGE_SIZE when it isn't present.
1850 Description: test_7 on sanity-quota.sh can't be used on separate machines
1851 Details : add facet to handle it
1855 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1856 Details : with NFS, the anon dentry's parent was set to itself in
1857 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1858 corresponding dentry other than use rec->ur_name.
1860 Severity : enhancement
1862 Description: lfs setstripe enhancement
1863 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1867 Description: enable data checksumming by default
1868 Details : enable checksum by default, allow --disable-checksum
1869 configure option and "-o nochecksum" mount option. Checksums
1870 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1874 Description: lustre support for RHEL5
1875 Details : Add support for RHEL5.
1879 Description: Client eviction due to failover config
1880 Details : after a connection loss, the lustre client should attempt to
1881 reconnect to the last active server first before trying the
1882 other potential connections.
1886 Description: when mds and osts use different quota unit(32bit and 64bit),
1887 quota will be released repeatly.
1888 Details : void sending multiple quota reqs to mds, which will keep the
1889 status between the reqs.
1893 Description: osts not allocated evenly to files
1894 Details : change the condition to increase offset_idx
1897 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1899 Description: Data corruption for OSTs that are formatted larger than 2TB
1901 Details : When generating the bio request for lustre file writes the
1902 sector number would overflow a temporary variable before being
1903 used for the IO. The data reads correctly from Lustre (which
1904 will overflow in a similar manner) but other file data or
1905 filesystem metadata may be corrupted in some cases.
1909 Description: Tune the kernel for good SCSI performance.
1910 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1911 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1916 Description: rewrite ext2-derived code in obdclass/uuid.c
1917 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1918 from scratch preserving functionality.
1922 Description: "lfs find -obd UUID" prints directories
1923 Details : "lfs find -obd UUID" will return all directory names instead
1924 of just file names. It is incorrect because the directories
1925 do not reside on the OSTs.
1929 Description: use i_size_read and i_size_write in 2.6 port
1930 Details : replace inode->i_size access with i_size_read/write()
1933 Frequency : when removing large files
1935 Description: scheduling issue during removal of large Lustre files
1936 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1937 It causes scheduling issues when removing large files (17TB in the
1941 Frequency : only with liblustre clients on XT3
1943 Description: evictions taking too long
1944 Details : allow llrd to evict clients directly on OSTs
1947 Frequency : only on ppc
1949 Description: /proc/fs/lustre/devices broken on ppc
1950 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1951 We should make sure the type of 'index' is loff_t and then cast
1952 explicitly as needed below. Do not assign an explicitly cast
1956 Frequency : only for rhel5
1958 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1959 Details : Modify the target file & which_kernel.
1961 Severity : enhancement
1963 Description: omit set fsid for export NFS
1964 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1965 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1969 Frequency : Always, for kernels after 2.6.16
1970 Description: Fix warning idr_remove called for id=.. which is not allocated.
1971 Details : Last kernels save old s_dev before kill super and not allow
1972 to restore from callback - restore it before call kill_anon_super.
1976 Description: buffer overruns could theoretically occur
1977 Details : llapi_semantic_traverse() modifies the "path" argument by
1978 appending values to the end of the origin string, and a
1979 overrun may occur. Adding buffer overrun check in liblustreapi.
1983 Description: Fix error on 'ls .' at the top of the Lustre mount.
1984 Details : Don't revalidate dentry if it is a root dentry.
1988 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1989 Details : Modify vm-tunables-rhel4.patch.
1993 Description: Kernel config for 2.6.18-vanilla.
1994 Details : Modify targets/2.6-vanilla.target.in.
1995 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1996 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1997 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1998 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2002 Description: improve handling recoverable errors
2003 Details : if request processig with error which can be recoverable on server
2004 request should be resend, otherwise page released from cache and
2009 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2010 Details : Modify target file & which_patch.
2011 A flaw was found in the IA32 system call emulation provided
2012 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2013 value could be stored in the %RAX register, which could trigger an
2014 out-of-bounds system call table access. An untrusted local user
2015 could exploit this flaw to run code in the kernel
2016 (ie a root privilege escalation). (CVE-2007-4573).
2020 Description: O_DIRECT bypasses client statistics.
2021 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2022 read_ahead_stats were updated. copied the stats section from
2023 osc_send_oap_rpc() into async_internal().
2027 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2031 Description: change order of libsysio includes
2032 Details : '#include sysio.h' should always come before '#include xtio.h'
2036 Description: handle "serious error: objid * already exists" more gracefully
2037 Details : If LAST_ID value on disk is smaller than the objects existing in
2038 the O/0/d* directories, it indicates disk corruption and causes an
2039 LBUG(). If the object is 0-length, then we should use the existing
2040 object. This will help to avoid a full fsck in most cases.
2042 Severity : enhancement
2044 Description: adapt the lustre_config script to support the upgrade case
2045 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2046 server targets to 1.6 in parallel.
2050 Description: To avoid grant space > avaible space when the disk is almost
2051 full. Without this patch you might see the error "grant XXXX >
2052 available" or some LBUG about grant, when the disk is almost
2054 Details : In filter_check_grant, for non_grant cache write, we should
2055 check the left space by if (*left > ungranted + bytes), instead
2056 of (*left > ungranted), because only we are sure the left space
2057 is enough for another "bytes", then the ungrant space should be
2058 increase. In client, we should update cl_avail_grant only there
2059 is OBD_MD_FLGRANT in the reply.
2063 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2064 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2065 described in CVE-2007-4573. This problem could allow untrusted
2066 local users to gain root access.
2069 Frequency : when using O_DIRECT and quotas
2071 Description: Incorrect file ownership on O_DIRECT output files
2072 Details : block usage reported by 'lfs quota' does not take into account
2073 files that have been written with O_DIRECT.
2078 Description: touch file failed when fs is not full
2079 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2080 otherwise we can get ENOSP while fs is not full.
2084 Description: parallel lock callbacks
2085 Details : Instead of sending blocking and completion callbacks as separated
2086 requests, adding them to a set and sending in parallel.
2089 Frequency : only for Cray XT3
2090 Bugzilla : 12829/13455
2091 Description: Changing primary group doesn't change the group lustre assigns to
2093 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2094 overridden with the primary group provided by the group upcall,
2095 whereas the supplied fsgid can be trusted if it is in the list of
2096 supplementary groups returned by the group upcall.
2098 Severity : enhancement
2100 Description: Allow masking D_WARNING, D_ERROR messages from console
2101 Details : Console messages can now be disabled via lnet.printk.
2105 Description: User code with malformed file open parameter crashes client node
2106 Details : Before packing join_file req, all the related reference should be
2107 checked carefully in case some malformed flags cause fake
2108 join_file req on client.
2112 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2113 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2114 mode is equal to the granted mode. However, it should also check
2115 that LDLM_FL_CP_REQD is not set.
2119 Description: Heavy nfs access might result in deadlocks
2120 Details : After ELC code landed, it is now improper to enqueue any mds
2121 locks under och_sem, because enqueue might want to decide to
2122 cancel open locks for same inode we are holding och_sem for.
2126 Description: Client eviction while running blogbench
2127 Details : A lot of unlink operations with concurrent I/O can lead to a
2128 deadlock causing evictions. To address the problem, the number of
2129 oustanding OST_DESTROY requests is now throttled to
2130 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2131 callbacks are processed in priority.
2135 Description: enable ACLs on MDS by default
2136 Details : ACLs must be enabled on MDS by default.
2139 Frequency : PPC/PPC64 only
2141 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2142 Details : fix duplicated definitions between asm-ppc64/types.h and
2143 lustre_types.h on PPC.
2146 Frequency : PPC/PPC64 only
2148 Description: asm-ppc/segment.h does not exist
2149 Details : fix compile issue on PPC.
2153 Description: better handle error messages in extents code
2156 Frequency : RHEL4 only
2158 Description: mkfs is very slow on IA64/RHEL4
2159 Details : A performance regression has been discovered in the MPT Fusion
2160 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2161 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2162 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2164 Severity : enhancement
2166 Description: SNMP support enhancement
2167 Details : Adding total number of sampled request for an MDS node in snmp
2170 Severity : enhancement
2172 Description: Optimize ldlm waiting list processing for PR extent locks
2173 Details : When processing waiting list for read extent lock and meeting read
2174 lock that is same or wider to it that is not contended, skip
2175 processing rest of the list and immediatelly return current
2176 status of conflictness, since we are guaranteed there are no
2177 conflicting locks in the rest of the list.
2181 Description: Time out and refuse to reconnect
2182 Details : When the failover node is the primary node, it is possible
2183 to have two identical connections in imp_conn_list. We must
2184 compare not conn's pointers but NIDs, otherwise we can defeat
2185 connection throttling.
2189 Description: port llog fixes from b1_6 into HEAD
2190 Details : Port llog reference couting and some llog cleanups from b1_6
2191 (bug 10800) into HEAD, for protect from panic and access to already
2192 free llog structures.
2196 Description: Detect stride IO mode in read-ahead
2197 Details : When a client does stride read, read-ahead should detect that and
2198 read-ahead pages according to the detected stride pattern.
2202 Description: data checksumming impacts single node performance
2203 Details : add support for several checksum algorithm. Currently, only CRC32
2204 and Adler-32 are supported. The checksum type can be changed on
2205 the fly via /proc/fs/lustre/osc/*/checksum_type.
2209 Description: use adler32 for page checksums
2210 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2215 Description: build for x2 fails
2216 Details : fix compile issue on Cray systems.
2220 Description: Properly match for duplicate locks
2221 Details : Due to different lock order from skiplists code, we need to
2222 traverse entire list for now
2225 Frequency : only on PPC/SLES10
2227 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2228 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2229 be defined. Add a hack in mkfs_lustre.c to work around this compile
2234 Description: LASSERT on MDS when client holding flock lock dies
2235 Details : ldlm pool logic depends on number of granted locks equal to
2236 number of released locks which is not true for flock locks, so
2237 just exclude such locks from consideration.
2241 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2242 Details : Use fsfilt_sync() for both the whole filesystem sync and
2243 individual file sync to eliminate dangerous inode locking
2244 with I_LOCK that can lead to a deadlock.
2248 Description: performance in 1.6.3
2249 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2250 sure that 1MB requests are not fragmented by the block layer.
2252 Severity : enhancement
2254 Description: organize the server-side client stats on per-nid basis
2255 Details : Change the structure of stats under obdfilter and mds to
2263 The "uuid"s file would list the uuids of _active_ exports.
2264 And the clear entry is to clear all stats and stale nids.
2266 Severity : enhancement
2268 Description: eliminate client locks in face of contention
2269 Details : file contention detection and lockless i/o implementation
2270 for contended files.
2274 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2277 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2279 Description: MDS may lose file striping (and hence file data) in some cases
2280 Details : If there are additional extended attributes stored on the MDS,
2281 in particular ACLs, SELinux, or user attributes (if user_xattr
2282 is specified for the client mount options) then there is a risk
2283 of attribute loss. Additionally, the Lustre file striping
2284 needs to be larger than default (e.g. striped over all OSTs),
2285 and an additional attribute must be stored initially in the
2286 inode and then increase in size enough to be moved to the
2287 external attribute block (e.g. ACL growing in size) for file
2292 Description: skiplist implementation simplification
2293 Details : skiplists are used to group compatible locks on granted list
2294 that was implemented as tracking first and last lock of each
2295 lock group the patch changes that to using doubly linked lists
2299 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2300 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2301 because LASSERT on both the data supplied by a client, and the
2302 data on disk is dangerous and incorrect.
2304 Severity : enhancement
2306 Description: Slow truncate/writes to huge files at high offsets.
2307 Details : Directly associate cached pages to lock that protect those pages,
2308 this allows us to quickly find what pages to write and remove
2309 once lock callback is received.
2313 Description: more ldlm soft lockups
2314 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2315 starve other threads from the resource lock for a long time in
2316 case of long waiting queue, so change the debug level from
2317 D_OTHER to the less frequently used D_INFO.
2319 Severity : enhancement
2321 Description: add -gid, -group, -uid, -user options to lfs find
2325 Description: Hung threads in invalidate_inode_pages2_range
2326 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2327 once one is completed. As a result, some RPCs are stuck in the
2328 queue and are never sent.
2332 Description: filter threads hungs on waiting journal commit
2333 Details : Cleanup filter group llog code, then only filter group llog will
2334 be only created in the MDS/OST syncing process.
2338 Description: Procfs and llog threads access destoryed import sometimes.
2339 Details : Sync the import destoryed process with procfs and llog threads by
2340 the import refcount and semaphore.
2342 Severity : enhancement
2344 Description: openlock cache of b1_6 port to HEAD
2349 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2350 Details : If insertion of an extent fails, then discard the inode
2351 preallocation and free data blocks else it can lead to duplicate
2356 Description: don't always update ctime in ext3_xattr_set_handle()
2357 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2358 In some cases the ctime should not be updated, for example for
2359 2.0->1.8 compatibility it is necessary to delete an xattr and it
2360 should not update the ctime.
2364 Bugzilla : 15713/16362
2365 Description: Assertion in iopen_connect_dentry in 1.6.3
2366 Details : looking up an inode via iopen with the wrong generation number can
2367 populate the dcache with a disconneced dentry while the inode
2368 number is in the process of being reallocated. This causes an
2369 assertion failure in iopen since the inode's dentry list contains
2370 both a connected and disconnected dentry.
2374 Description: assertion failure in ldlm_handle2lock()
2375 Details : fix a race between class_handle_unhash() and class_handle2object()
2376 introduced in lustre 1.6.5 by bug 13622.
2381 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2382 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2387 Description: oops in sd_iostats_seq_show()
2388 Details : unloading/reloading the scsi low level driver triggers a kernel
2389 bug when trying to access the sd iostat file.
2394 Description: Kernel panics during QLogic driver reload
2395 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2396 patch, causing memory corruption.
2401 Description: journal_dev option does not work in b1_6
2402 Details : pass mount option during pre-mount.
2404 Severity : enhancement
2406 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2407 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2408 extent information of a file. It can be used to map logical blocks
2409 in a file to physical blocks in the block device.
2413 Description: LDLM soft lockups - improvement
2414 Details : It is be possible to send the lock handle along with each read
2415 or write request because the client is already doing a lock match
2416 itself so there isn't any reason the OST should have to re-do that
2420 Frequency : only on Cray X2
2422 Description: X2 build failures
2423 Details : fix build failures on Cray X2.
2427 Description: xid & resent requests
2428 Details : Initialize RPC XID from clock at startup (randomly if clock is
2431 Severity : enhancement
2433 Description: Add lustre_start utility to start or stop multiple Lustre servers
2438 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2439 Details : In case of memory pressure, list_del() can be called twice on
2440 req->rq_history_list, causing a kernel oops.
2444 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2445 Details : fix stack overflow in the distributed lock manager by defering
2446 export eviction after a failed AST to the elt thread instead of
2447 handling it in the dlm interpret routine.
2451 Description: Convert some comments to new format.
2452 Details : Update documenting comments to match doxygen conventions.
2456 Description: Grammar fixes.
2457 Details : A couple of trivial spelling fixes.
2461 Description: OSD_COUNTERS-mandatory
2462 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2463 to simplify the code.
2467 Description: simplify cmm_device freeing logic.
2468 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2472 Description: Add lockdep support to dt_object_operations locking interface.
2473 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2474 parameter indicating lock ordering. Update mdd code to use new
2479 Description: Introduce failloc constants for lockless IO tests.
2480 Details : Add two new failloc constants to test lockless IO. Only one of
2481 them in implemented---another is checked in yet to be landed
2486 Description: Add lockdep support for inode mutex.
2487 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2488 in the situations where Lustre has to lock more than one inode
2493 Description: Add optional invariants checking support.
2494 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2495 configure switch. This macro is to be used for consistency and
2496 sanity checks that are too expensive to be left in `production'
2501 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2502 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2507 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2508 Details : Use spin_lock_nested() in (the only) situation where more than
2509 one ldlm_lock is locked simultaneously. Also, fix possible
2510 dead-lock in ldlm_lock_change_resource() by enforcing particular
2515 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2516 Details : Instead of specifying each ldlm_lock call-back through separate
2517 parameter, wrap them into struct ldlm_callback_suite.
2521 Description: Kill join_lru obd method and its callers.
2522 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2523 in memory---a requirement implemented through ->o_join_lru() obd
2524 method in HEAD. Remove this method and its users.
2528 Description: Add asynchronous ldlm ENQUEUE completion handler.
2529 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2530 case that stock ldlm_completion_ast() cannot handle as it waits
2531 until lock is granted. Introduce new ldlm_completion_ast_async()
2532 for this. Also comment ldlm_completion_ast().
2536 Description: ldlm_error <-> errno conversion.
2537 Details : Add functions to map (rather arbitrary) between LDLM error codes
2538 and standard errno values. CLIO needs this to prevent LDLM specific
2539 constants from escaping ldlm and osc.
2543 Description: Kill unused ldlm_handle2lock_ns() function.
2544 Details : Kill unused ldlm_handle2lock_ns() function.
2548 Description: Add lu_ref support to ldlm_lock
2549 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2550 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2551 and ldlm_resource. LDLM interface has to be changed, because code that
2552 releases a reference on a lock, has to "know" what reference this is.
2553 In the most frequent case
2555 lock = ldlm_handle2lock(handle);
2557 LDLM_LOCK_PUT(lock);
2559 no changes are required. When any other reference (received _not_ from
2560 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2561 instead of LDLM_LOCK_PUT().
2563 Arguably, changes are pervasive, and interface requires some discipline
2564 for proper use. On the other hand, it was very instrumental in finding
2565 a few leaked lock references.
2569 Description: Add ldlm_lock_addref_try().
2570 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2571 attempts to addref a lock that might be being canceled
2576 Description: Add ldlm_weigh_callback().
2577 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2578 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2579 hard-coded `number of pages' logic.
2583 Description: Add lockdep annotations to llog code.
2584 Details : Use appropriately tagged _nested() locking calls in the places
2585 where llog takes more than one ->lgh_lock lock.
2589 Description: Add loi_kms_set().
2590 Details : Wrap kms updates into a helper function.
2594 Description: Constify instances of struct lsm_operations.
2595 Details : Constify instances of struct lsm_operations.
2599 Description: lu_conf support.
2600 Details : On a server, a file system object is uniquely identified
2601 by a fid, which is sufficient to locate and load all object
2602 state (inode). On a client, on the other hand, more data are
2603 necessary instantiate an object. Change lu_object_find() and
2604 friends to take additional `lu_conf' argument describing object.
2605 Typically this includes layout information.
2609 Description: lu_context fixes.
2610 Details : Introduce new lu_context functions that are needed on the client
2611 side, where some system threads (ptlrpcd) are shared by multiple
2612 modules, and so cannot be stopped during module shutdown.
2616 Description: Add start and stop methods to lu_device_type_operations.
2617 Details : Introduce two new methods in lu_device_type_operations, that are
2618 invoked when first instance of a given type is created and last one
2619 is destroyed respectively. This is need by CLIO.
2623 Description: Add lu_ref support to struct lu_device.
2624 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2625 track leaked references.
2629 Description: Introduce lu_kmem_descr.
2630 Details : lu_kmem_descr and its companion interface allow to create
2631 and destroy a number of kmem caches at once.
2635 Description: Fix lu_object finalization race.
2636 Details : Fix a race between lu_object_find() finding an object and its
2637 concurrent finalization. This race is (most likely) not possible
2638 on the server, but might happen on the client.
2642 Description: Introduce lu_ref interface.
2643 Details : lu_ref is a debugging module allowing to track references to
2644 a given object. It is quite cpu expensive, and has to be
2645 explicitly enabled with --enable-lu_ref. See usage description
2650 Description: Factor lu_site procfs stats into a separate function.
2651 Details : Separate lu_site stats printing code into a separate function
2652 that can be reused on a client.
2656 Description: Constify instances of struct {lu,dt,md}_device_operations.
2657 Details : Constify instances of struct {lu,dt,md}_device_operations.
2661 Description: Introduce struct md_site and move meta-data specific parts of
2662 struct lu_site here.
2663 Details : Move md-specific fields out of struct lu_site into special struct
2664 md_site, so that lu_site can be used on a client.
2668 Description: Kill mdd_lov_destroy().
2669 Details : Remove unused mdd code.
2673 Description: Add st_block checking to multistat.c.
2674 Details : Add st_block checking to multistat.c.
2678 Description: Add lu_ref support to struct obd_device.
2679 Details : Add lu_ref tracking to obd_device.
2683 Description: Kill obd_set_fail_loc().
2684 Details : Remove unused code.
2688 Description: Add special type for ptlrpc_request interpret functions.
2689 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2690 there. Actual usage will be in CLIO.
2694 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2695 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2696 doesn't work with lockdep.
2700 Description: Add rwv.c test program.
2701 Details : New testing program exercising readv(2) and writev(2) (Qian).
2705 Description: Add sendfile.c test program.
2706 Details : New testing program exercising sendfile(2) (Jay).
2710 Description: Ratelimit a message that can be very frequent.
2711 Details : Ratelimit a memory allocation failure message that can
2716 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2717 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2721 Description: Liblustre build fixes.
2722 Details : Liblustre build fixes.
2726 Description: libcfs: add cfs_{need,cond}_resched() interface.
2727 Details : libcfs: add cfs_{need,cond}_resched() definition and
2728 implementations for Linux, NT, and liblustre.
2730 Severity : enhancement
2732 Description: More exported tunables for mballoc
2733 Details : Add support for tunable preallocation window and new tunables for
2734 large/small requests
2738 Description: Detect corruption of block bitmap and checking for preallocations
2739 Details : Checks validity of on-disk block bitmap. Also it does better
2740 checking of number of applied preallocations. When corruption is
2741 found, it turns filesystem readonly to prevent further corruptions.
2745 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2746 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2747 window to 0. If it does hit cache miss, and read-pattern is still
2748 stride-io mode, does not reset the stride window, but also does not
2749 increase the stride window length in this case.
2753 Frequency : only for big-endian servers
2754 Description: Check if system is big-endian while mounting fs with extents feature
2755 Details : Mounting a filesystem with extents feature will fail on big-endian
2756 systems since ext3-based ldiskfs is not supported on big-endian
2757 systems. This can be over-riden with "bigendian_extents" mount option.
2759 Severity : enhancement
2761 Description: The root squash functionality
2762 Details : A security feature, which is to prevent users from being able
2763 to mount lustre on their desktop, run as root, and delete
2764 all of the files in the filesystem. The goal is accomplished by
2765 remapping user id (UID) and group id (GID) of the root user to
2766 a UID and GID specified by the system administartor via Lustre
2767 configuration management server (MGS). The functionality also
2768 allows to specify sets of clients for which the remapping does
2773 Description: Excessive recovery window
2774 Details : With AT enabled, the recovery window can be excessively long (6000+
2775 seconds). To address this problem, we no longer use
2776 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2777 timeout no longer depends on the service time, it is set to
2778 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2779 time via pb_service_time.
2783 Description: Watchdog triggered on MDS failover
2784 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2785 the OSTs know that the MDS "UUID" can be reused for the same export
2786 from a different NID, so we do not need to wait for the export to be
2790 Frequency : rare, only if using MMP with Linux RAID
2792 Description: MMP doesn't work with Linux RAID
2793 Details : While using HA for Lustre servers with Linux RAID, it is possible
2794 that MMP will not detect multiple mounts. To make this work we
2795 need to unplug the device queue in RAID when the MMP block is being
2796 written. Also while reading the MMP block, we should read it from
2797 disk and not the cached one.
2799 Severity : enhancement
2801 Description: open file using fid
2802 Details : A file can be opened using just its fid, like
2803 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2806 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2809 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2810 Details : add proper locking for creation of jbd_4k slab cache
2814 Description: MMP check in ext3_remount() fails without displaying any error
2815 Details : When multiple mount protection fails during remount, proper error
2818 Severity : enhancement
2820 Description: Allow stripe size to be up to 4G-64k
2821 Details : Fix math logic to allow large stripe sizes.
2825 Description: add check for >8TB ldiskfs filesystems
2826 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2827 Don't allow >8TB ldiskfs filesystems to be mounted without
2828 force_over_8tb mount option
2833 Descriptoin: OSS complains "too many blocks" and crashes
2834 Details : In filter_commitrw_write():
2835 1. lquota_chkquota();
2836 2. obdo_from_inode();
2837 3. lquota_pending_commit();
2838 "2." has changed uid and gid in this case, then caused
2839 lquota_chkquota() and lquota_pending_commit() to use
2840 different uid and gid. That is the root of the bug.
2844 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2845 Details : Detect and print out the 'stripe count = -1' case properly.
2849 Description: Out or order replies might be lost on replay
2850 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2851 request with tid smaller than one currently being added, add it
2852 to the start, not end of the list.
2854 --------------------------------------------------------------------------------
2856 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2858 * Support for kernels:
2859 2.6.5-7.283 (SLES 9),
2860 2.6.9-55.EL (RHEL 4),
2861 2.6.16.46-0.14 (SLES 10),
2862 2.6.18.8 vanilla (kernel.org)
2863 * Client support for unpatched kernels:
2864 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2865 2.6.16 - 2.6.22 vanilla (kernel.org)
2866 * Due to recently discovered recovery problems, we do not recommend
2867 using patchless RHEL 4 clients with this or any earlier release.
2868 * Recommended e2fsprogs version: 1.39.cfs8
2869 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2870 * Starting with this release, the ldiskfs backing filesystem required
2871 by Lustre is now in its own package, lustre-ldiskfs. This package
2872 should be installed. It is versioned separately from Lustre and
2873 may be released separately in future.
2877 Description: block reactivating mgc import until all deactivates complete
2878 Details : Fix race when failing back MDT/MGS to itself (testing)
2881 Frequency : at statup only
2883 Description: mds_lov_synchronize race leads to various problems
2884 Details : simultaneous MDT->OST connections at startup can cause the
2885 sync to abort, leaving the OSC in a bad state.
2887 Severity : enhancement
2889 Description: add optional extra BUILD_VERSION info
2890 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2891 to override the lustre version.
2894 Frequency : 2.6.18 servers only
2896 Description: ll_kern_mount() doesn't release the module reference
2897 Details : The ldiskfs module reference count never drops down to 0
2898 because ll_kern_mount() doesn't release the module reference.
2903 Description: server LBUG when using old ost_num_threads parameter
2904 Details : Accept the old ost_num_threads parameter but warn that it
2905 is deprecated, and fix an off-by-one error that caused an LBUG.
2910 Description: Transient SCSI error results in persistent IO issue
2911 Details : iobuf->dr_error is not reinitialized to 0 between two
2915 Frequency : sometimes when underlying device returns I/O errors
2917 Description: OSTs not going read-only during write failures
2918 Details : OSTs are not remounted read-only when the journal commit threads
2919 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2920 instead of the ext3 wrappers.
2924 Description: poor connect scaling with increasing client count
2925 Details : Don't run filter_grant_sanity_check for more than 100 exports
2926 to improve scaling for large numbers of clients.
2929 Frequency : SLES10 only
2931 Description: sanity-quota.sh quotacheck failed: rc = -22
2932 Details : Quotas cannot be enabled on SLES10.
2935 Frequency : liblustre clients only
2937 Description: getdirentries does not give error when run on compute nodes
2938 Details : getdirentries does not fail when the size specified as an argument
2939 is too small to contain at least one entry
2941 Severity : enhancement
2943 Description: Add LNET router traceability for debug purposes
2944 Details : If a checksum failure occurs with a router as part of the
2945 IO path, the NID of the last router that forwarded the bulk data
2946 is printed so it can be identified.
2951 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2952 Details : Due to a race condition, liblustre clients were occasionally
2953 evicted incorrectly.
2955 Severity : enhancement
2957 Description: lfs setstripe use optional parameters instead of postional
2960 Severity : enhancement
2962 Description: Nanosecond timestamp support for ldiskfs
2963 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2964 resolution timestamps. There is not yet support for this at
2965 the Lustre filesystem level.
2968 Frequency : during server recovery
2970 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2971 Details : request with rq_no_resend flag not awake l_wait_event if they get
2975 Frequency : nfs export on patchless client
2977 Description: connectathon hang when test nfs export over patchless client
2978 Details : Disconnected dentry cannot be found with lookup, so we do not need
2979 to unhash it or make it invalid
2982 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2983 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2984 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2989 Description: open req refcounting wrong on reconnect
2990 Details : If reconnect happened between getting open reply from server and
2991 call to mdc_set_replay_data in ll_file_open, we will schedule
2992 replay for unreferenced request that we are about to free.
2993 Subsequent close will crash in variety of ways.
2994 Check that request is still eligible for replay in
2995 mdc_set_replay_data().
3000 Description: disable writes to filesystem when reading health_check file
3001 Details : the default for reading the health_check proc file has changed
3002 to NOT do a journal transaction and write to disk, because this
3003 can cause reads of the /proc file to hang and block HA state
3004 checking on a healthy but otherwise heavily loaded system. It
3005 is possible to return to the previous behaviour during configure
3006 with --enable-health-write.
3008 Severity : enhancement
3010 Description: 64-bit inode version
3011 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3012 to the inode. This will be required for version-based recovery.
3017 Description: MDS fails to start if a duplicate client export is detected
3018 Details : in some rare cases it was possible for a client to connect to
3019 an MDS multiple times. Upon recovery the MDS would detect this
3020 and fail during startup. Handle this more gracefully.
3022 Severity : enhancement
3024 Description: Add -o localflock option to simulate old noflock
3026 Details : This will achieve local-only flock/fcntl locks
3032 Description: log_commit_thread vs filter_destroy race leads to crash
3033 Details : Take import reference before releasing llog record semaphore
3038 Description: Wrong request locking in request set processing
3039 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3040 imp_delayed_list, in this place should be used imp_lock.
3043 Frequency : when reconnection
3045 Description: Grant Leak when osc reconnect to OST
3046 Details : When osc reconnect ost, OST(filter) should check whether it
3047 should grant more space to client by comparing fed_grant and
3048 cl_avail_grant, and return the granted space to client instead
3049 of "new granted" space, because client will call osc_init_grant
3050 to update the client grant space info.
3053 Frequency : when client reconnect to OST
3055 Description: Grant Leak when osc do resend and replay bulk write
3056 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3057 bulk write request, because the grant info while be sync between
3058 OSC and OST when reconnect, and we should ignore the grant info
3059 these of resend/replay write req.
3064 Description: Grant space more than available left space sometimes.
3065 Details : When then OST is about to be full, if two bulk writing from
3066 different clients came to OST. Accord the avaliable space of the
3067 OST, the first req should be permitted, and the second one
3068 should be denied by ENOSPC. But if the seconde arrived before
3069 the first one is committed. The OST might wrongly permit second
3070 writing, which will cause grant space > available space.
3073 Frequency : when client is evicted
3075 Description: Grant might be wrongly erased when osc is evicted by OST
3076 Details : when the import is evicted by server, it will fork another
3077 thread ptlrpc_invalidate_import_thread to invalidate the
3078 import, where the grant will be set to 0. While the original
3079 thread will update the grant it got when connecting. So if
3080 the former happened latter, the grant will be wrongly errased
3081 because of this race.
3086 Description: Checking Stale with correct fid
3087 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3088 to check whether it is stale, because sometimes, we want the
3089 enqueue happened anyway, and op_data.fid2 will not be initialized.
3091 Severity : enhancement
3093 Description: update patchless client
3094 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3097 Frequency : only with 2.4 kernel
3099 Description: random memory corruption
3100 Details : size of struct ll_inode_info is to big for union inode.u and this
3101 can be cause of random memory corruption.
3106 Description: Memory leak in recovery
3107 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3108 It should also check obd_fail before fsfilt_start, otherwise if
3109 fsfilt_start return -EROFS,(failover mds during mds recovery).
3110 then the req will return with repmsg->transno = 0 and rc = EROFS.
3111 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3112 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3113 be freed no matter whether fsfilt_commit success or not.
3116 Frequency : only with huge count clients
3118 Description: Prevents from taking the superblock lock in llap_from_page for
3120 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3121 for llap_from_page call in ll_removepage prevents from taking the
3122 superblock lock for a soon died page.
3127 Description: Not check open intent error before release open handle
3128 Details : in some rare cases, the open intent error is not checked before
3129 release open handle, which may cause
3130 ASSERTION(open_req->rq_transno != 0), because it tries to release
3131 the failed open handle.
3136 Description: Set cat log bitmap only after create log success.
3137 Details : in some rare cases, the cat log bitmap is set too early. and it
3138 should be set only after create log success.
3142 Description: Accessing a block bevice can re-enable I/O when Lustre is
3143 tearing down a device.
3144 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3149 Description: service threads may hog cpus when there are a lot of requests
3150 Details : Insert cond_resched to give other threads a chance to use some CPU
3155 Description: the cat log was not initialized in recovery
3156 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3157 unlink log on mds will not be initialized until mds post
3158 recovery. And also in mds post recovery, the unlink log will
3159 initialization will be done asynchronausly, so there will be race
3160 between add unlink log and unlink log initialization.
3164 Description: brw_stats were being printed incorrectly
3165 Details : brw_stats were being printed as log2 but all of them were not
3166 recorded as log2. Also remove some code duplication arising from
3167 filter_tally_{read,write}.
3171 Frequency : rare, only in recovery.
3172 Description: ASSERTION(req->rq_type != LI_POISON) failed
3173 Details : imp_lock should be held while iterating over imp_sending_list for
3174 prevent destroy request after get timeout in ptlrpc_queue_wait.
3178 Description: replay-single.sh test 52 fails
3179 Details : A lock's skiplist need to be cleanup when it being unlinked
3180 from its resource list.
3184 Description: Short directio read returns full requested size rather than
3186 Details : Direct I/O operations should return actual amount of bytes
3187 transferred rather than requested size.
3189 Severity : enhancement
3191 Description: metadata RPC reduction (e.g. for rm performance)
3192 Details : decrease the amount of synchronous RPC between clients and servers
3193 by canceling conflicing lock before the operation on the client
3194 and packing thier handles into the main operation RPC to server.
3196 Severity : enhancement
3198 Description: Async OSC create to avoid the blocking unnecessarily.
3199 Details : If an OST has no remaining object, system will block on the
3200 creation when it needs to create a new object on this OST. Now,
3201 ways use pre-created objects when available, instead of blocking on
3202 an empty osc while others are not empty. If we must block, we
3203 block for the shortest possible period of time.
3207 Description: improve handling recoverable errors
3208 Details : if request processig with error which can be recoverable on server
3209 request should be resend, otherwise page released from cache and
3212 Severity : enhancement
3214 Description: refine locking for avoid write wrong info into lov_objid file
3215 Details : fix possible races with add new target and write/update data in
3218 --------------------------------------------------------------------------------
3220 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3225 Frequency : on some architectures
3227 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3228 Details : Uninitialized flags sometimes cause configuration commands to
3232 Frequency : patchless clients only
3234 Description: missing __iget() symbol export
3235 Details : The __iget() symbol export is missing. To avoid the need for
3236 this on patchless clients the deathrow inode reaper is turned
3237 off, and we depend on the VM to clean up old inodes. This
3238 dependency was during via the fix for bug 12181.
3240 --------------------------------------------------------------------------------
3242 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3244 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3245 INTEROPERATE with older versions automatically. In many cases a
3246 special upgrade step is needed. Please read the
3247 user documentation before upgrading any part of a live system.
3248 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3249 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3250 * WARNING: Lustre configuration and startup changes are required with
3251 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3255 Severity : enhancement
3257 Description: MountConf
3258 Details : Lustre configuration is now managed via mkfs and mount
3259 commands instead of lmc and lconf. New obd types (MGS, MGC)
3260 are added for dynamic configuration management. See
3261 https://mail.clusterfs.com/wikis/lustre/MountConf for
3264 Severity : enhancement
3266 Description: dynamic OST addition
3267 Details : OSTs can now be added to a live filesystem
3269 Severity : enhancement
3271 Description: startup order invariance
3272 Details : MDTs and OSTs can be started in any order. Clients only
3273 require the MDT to complete startup.
3275 Severity : enhancement
3277 Description: parallel, asynchronous orphan cleanup
3278 Details : orphan cleanup is now performed in separate threads for each
3279 OST, allowing parallel non-blocking operation.
3281 Severity : enhancement
3283 Description: optimized stripe assignment
3284 Details : stripe assignments are now made based on ost space available,
3285 ost previous usage, and OSS previous usage, in order to try
3286 to optimize storage space and networking resources.
3288 Severity : enhancement
3290 Description: Permanently set tunables
3291 Details : All writable /proc/fs/lustre tunables can now be permanently
3292 set on a per-server basis, at mkfs time or on a live system.
3294 Severity : enhancement
3296 Description: Lustre message v2
3297 Details : Add lustre message format v2.
3299 Severity : enhancement
3301 Description: client OST exclusion list
3302 Details : Clients can be started with a list of OSTs that should be
3303 declared "inactive" for known non-responsive OSTs.
3307 Description: SPEC SFS validation failure on NFS v2 over lustre.
3308 Details : Changes the blocksize for regular files to be 2x RPC size,
3309 and not depend on stripe size.
3311 Severity : enhancement
3313 Description: Multiple MD RPCs in flight.
3314 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3315 To avoid overly-overloading MDS, introduce a limit on number of
3316 MDS RPCs in flight for a single client and add /proc controls
3317 to adjust this limit.
3319 Severity : enhancement
3321 Description: client read/write statistics
3322 Details : Add client read/write call usage stats for performance
3323 analysis of user processes.
3324 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3325 file access. extents_stats shows chunk size distribution.
3326 extents_stats_per_process show chunk size distribution per
3329 Severity : enhancement
3331 Description: mds statistics
3332 Details : Add detailed mds operations statistics in
3333 /proc/fs/lustre/mds/*/stats.
3337 Description: Failure of copying files with lustre special EAs.
3338 Details : Client side always return success for setxattr call for lustre
3339 special xattr (currently only "trusted.lov").
3344 Description: Refcount LNET uuids
3345 Details : The global LNET uuid list grew linearly with every startup;
3346 refcount repeated list entries instead of always adding to
3349 Severity : enhancement
3351 Description: Dynamic service threads
3352 Details : Within a small range, start extra service threads
3353 automatically when the request queue builds up.
3356 Frequency : mixed-endian client/server environments
3358 Description: mixed-endian crashes
3359 Details : The new msg_v2 system had some failures in mixed-endian
3362 Severity : enhancement
3364 Description: Easy OST removal
3365 Details : OSTs can be permanently deactivated with e.g. 'lctl
3366 conf_param lustre-OST0001.osc.active=0'
3368 Severity : enhancement
3370 Description: MGS proc entries
3371 Details : Added basic proc entries for the MGS showing what filesystems
3374 Severity : enhancement
3376 Description: provide MGS failover
3377 Details : Added config lock reacquisition after MGS server failover.
3379 Severity : enhancement
3381 Description: add Linux 2.4 support
3382 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3386 Description: a large application tries to do I/O to the same resource and dies
3387 in the middle of it.
3388 Details : Check the req->rq_arrival time after the call to
3389 ost_brw_lock_get(), but before we do anything about
3390 processing it & sending the BULK transfer request. This
3391 should help move old stale pending locks off the queue as
3392 quickly as obd_timeout.
3395 Frequency : when an incorrect nid is specified during startup
3397 Description: ptlrpc connect to non-existant node causes kernel crash
3398 Details : LNET can't be re-entered from an event callback, which
3399 happened when we expire a message after the export has been
3400 cleaned up. Instead, hand the zombie cleanup off to another
3403 Severity : enhancement
3405 Description: plain/inodebits lock performance improvement
3406 Details : Grouping plain/inodebits in granted list by their request modes
3407 and bits policy, thus improving the performance of search through
3411 Frequency : only if OST filesystem is corrupted
3413 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3414 Details : for a short time RPCs with bulk IO are in the replay list,
3415 but replay of bulk IOs is unimplemented. If the OST filesystem
3416 is corrupted due to disk cache incoherency and then replay is
3417 started it is possible to trip an assertion. Avoid putting
3418 committed RPCs into the replay list at all to avoid this issue.
3421 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3423 Description: System hang on startup
3424 Details : This bug allowed the liblustre (e.g. catamount) client to
3425 return to the app before handling all startup RPCs. This
3426 could leave the node unresponsive to lustre network traffic
3427 and manifested as a server ptllnd timeout.
3429 Severity : enhancement
3431 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3432 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3433 boolean to control whether to print peer debug info when a
3434 client's RPC times out.
3437 Frequency : only for kernels with patches from Lustre below 1.4.3
3439 Description: Remove old rdonly API
3440 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3443 Frequency : only for devices with external journals
3445 Description: Set external device read-only also
3446 Details : During a commanded failover stop, we set the disk device
3447 read-only while the server shuts down. We now also set any
3448 external journal device read-only at the same time.
3451 Frequency : when upgrading from 1.4 while trying to change parameters
3453 Description: The wrong (new) MDC name was used when setting parameters for
3454 upgraded MDT's. Also allows changing of OSC (and MDC)
3455 parameters if --writeconf is specified at tunefs upgrade time.
3458 Frequency : when setting specific ost indicies
3460 Description: QOS code breaks on skipped indicies
3461 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3462 created with --index need not be sequential.
3464 Severity : enhancement
3466 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3467 Details : The uninit_groups feature works in conjunction with the kernel
3468 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3469 up the pass1 processing of e2fsck. This is a read-only feature
3470 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3471 mount filesystems that have had this feature enabled.
3473 Severity : enhancement
3475 Description: Improve multi-block allocation algorithm to avoid fragmentation
3476 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3477 allocation locality and avoid filesystem fragmentation.
3479 ------------------------------------------------------------------------------
3481 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3483 * Support for kernels:
3484 2.4.21-47.0.1.EL (RHEL 3)
3485 2.6.5-7.283 (SLES 9)
3486 2.6.9-42.0.10.EL (RHEL 4)
3487 2.6.12.6 vanilla (kernel.org)
3488 2.6.16.27-0.9 (SLES 10)
3489 * Recommended e2fsprogs version: 1.39.cfs5
3491 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3495 Frequency : occasional, depends on client load and configuration
3496 Bugzilla : 12181, 12203
3497 Description: data loss for recently-modified files
3499 Details : In some cases it is possible that recently written or created
3500 files may not be written to disk in a timely manner (this should
3501 normally be within 30s unless client IO load is very high).
3502 The problem appears as zero-length files or files that are a
3503 multiple of 1MB in size after a client crash or client eviction
3504 that are missing data at the end of the file.
3506 This problem is more likely to be hit on clients where files are
3507 repeatedly created and unlinked in the same directory, clients
3508 have a large amount of RAM, have many CPUs, the filesystem has
3509 many OSTs, the clients are rebooted frequently, and/or the files
3510 are not accessed by other nodes after being written.
3512 The presence of the problem can be detected by looking at
3513 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3514 smaller than the second (nr_unused) then dirty files will not
3515 be flushed automatically to disk. "sync; sleep 10" should be
3516 run several times on the node before unmounting it to update
3517 Lustre (this is also safe to run on nodes without this problem).
3519 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3520 that can cause this same problem, so customers using that kernel
3521 also need to update the kernel in addition to Lustre. In order
3522 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3524 It is normal that files written just before a client crash (less
3525 than 30s) may not yet have been flushed to disk, even for local
3529 Frequency : frequent on thin XT3 nodes
3531 Description: UUID collision on thin XT3 Linux nodes
3532 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3533 randomly, since we relied on an insufficiently-seeded PRNG.
3538 Description: OSS hangs after "All ost request buffers busy"
3539 Details : A deadlock between quota and journal operations caused OSS
3540 hangs after printing "All ost request buffers busy."
3543 Frequency : always on liblustre builds
3545 Description: Cleanup compiler warnings on liblustre
3548 Frequency : always on liblustre builds on XT3
3550 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3554 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3555 > 32000 subdirectories
3556 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3557 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3558 subdirectory count crosses 32000. This will aid e2fsck to
3559 correctly handle more than 32000 subdirectories.
3562 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3564 Description: System hang on startup
3565 Details : This bug allowed the liblustre (e.g. catamount) client to
3566 return to the app before handling all startup RPCs. This
3567 could leave the node unresponsive to lustre network traffic
3568 and manifested as a server ptllnd timeout.
3570 Severity : enhancement
3572 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3573 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3574 boolean to control whether to print peer debug info when a
3575 client's RPC times out.
3580 Description: make O_SYNC working on 2.6 kernels
3581 Details : 2.6 kernels use different method for mark pages for write,
3582 so need add a code to lustre for O_SYNC work.
3587 Description: Failure to close file and release space on NFS
3588 Details : Put inode details into lock acquired in ll_intent_file_open.
3589 Use mdc_intent_lock in ll_intent_open to properly
3590 detect all kind of errors unhandled by mdc_enqueue.
3595 Description: proc file read during shutdown sometimes raced obd removal,
3597 Details : Add lock to prevent obd access after proc file removal.
3600 Frequency : Only for files larger than 4GB on 32-bit clients.
3602 Description: improperly doing page alignment of locks
3603 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3604 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3609 Description: RPCs being resent when they shouldn't be
3610 Details : Some RPCs that should not be resent are being resent. This
3611 can cause inconsistencies in the RPC state machine. Do not
3612 resend such requests.
3615 Frequency : rare, only with NFS export
3617 Description: Crash on NFS re-export node
3618 Details : under very unusual load conditions an assertion is hit in
3619 ll_intent_file_open()
3622 Frequency : only if OST filesystem is corrupted
3624 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3625 Details : for a short time RPCs with bulk IO are in the replay list,
3626 but replay of bulk IOs is unimplemented. If the OST filesystem
3627 is corrupted due to disk cache incoherency and then replay is
3628 started it is possible to trip an assertion. Avoid putting
3629 committed RPCs into the replay list at all to avoid this issue.
3634 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3635 Details : Large single O_DIRECT read and write calls can fail to allocate
3636 a sufficiently large buffer to process the request. In case of
3637 allocation failure the allocation is retried with a smaller
3638 buffer and broken into smaller requests.
3640 Severity : enhancement
3642 Description: Add -o localflock option to simulate old noflock behaviour.
3643 Details : This will achieve local-only flock/fcntl locks coherentness.
3648 Description: versioning check is incomplete
3649 Details : Checking the version difference of client vs. server, report
3650 error if the gap is too big.
3655 Description: add support PG_writeback bit
3656 Details : add support for PG_writeback bit for Lustre, for more carefull
3657 work with page cache in 2.6 kernel. This also fix some deadlocks
3658 and remove hack for work O_SYNC with 2.6 kernel.
3660 Severity : enhancement
3662 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3663 Details : The uninit_groups feature works in conjunction with the kernel
3664 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3665 up the pass1 processing of e2fsck. This is a read-only feature
3666 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3667 mount filesystems that have had this feature enabled.
3669 Severity : enhancement
3671 Description: Improve multi-block allocation algorithm to avoid fragmentation
3672 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3673 allocation locality and avoid filesystem fragmentation.
3675 ------------------------------------------------------------------------------
3677 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3679 * Support for kernels:
3680 2.6.9-42.0.3.EL (RHEL 4)
3681 2.6.5-7.276 (SLES 9)
3682 2.4.21-47.0.1.EL (RHEL 3)
3683 2.6.12.6 vanilla (kernel.org)
3684 2.6.16.21-0.8 (SLES10)
3685 * Recommended e2fsprogs version: 1.39.cfs2-0
3687 * The backwards-compatible /proc/sys/portals symlink has been removed
3688 in this release. Before upgrading, please ensure that you change
3689 any configuration scripts or /etc/sysctl.conf files that access
3690 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3691 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3692 in advance of the upgrade on any system running Lustre 1.4.6 or
3693 newer, since /proc/sys/lnet was added in that version.
3694 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3698 Frequency : only when quota is used
3700 Description: avoid scanning export list for quota master
3701 Details : Change the algorithms to avoid scanning export list in order
3702 to improve the efficiency.
3705 Frequency : MDS failover only, very rarely
3707 Description: "went back in time" messages on mds failover
3708 Details : The greatest transno may be lost when the current operation
3709 finishes with an error (transno==0) and the client's last_rcvd
3710 record is over-written. Save the greatest transno in the
3711 mds_last_transno for this case.
3714 Frequency : always for specific kernels and striping counts
3716 Description: client may get "Matching packet too big" without ACL support
3717 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3718 when trying to access files in certain configurations. The
3719 clients should in fact be denied when mounting because they do
3720 not understand ACLs.
3723 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3725 Description: many clients connecting with IO in progress causes connect timeouts
3726 Details : Avoid synchronous journal commits to avoid delays caused by many
3727 clients connecting/disconnecting when bulk IO is in progress.
3728 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3729 OST_IO_PORTAL to avoid delays behind potentially many pending
3733 Frequency : occasionally with multiple writers to a single file
3735 Description: shared writes to file may result in wrong size reported by stat()
3736 Details : Allow growing of kms when extent lock is cancelled
3739 Frequency : always with random mmap IO to multi-striped file
3741 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3742 Details : Only if the hole is at the end of OST object so that kms is too
3743 small. Fix is to increase kms accordingly in ll_nopage.
3746 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3748 Description: writes to a missing object would leak memory on the OST
3749 Details : If there is an inconsistency between the MDS and OST filesystems,
3750 such that the MDS references an object that doesn't exist, writes
3751 to that object will leak memory due to incorrect cleanup in the
3752 error handling path, eventually running out of memory on the OST.
3757 Description: Creating too long symlink causes lustre errors
3758 Details : Check symlink and name lengths before sending requests to MDS.
3761 Frequency : only if flock is enabled (not on by default)
3763 Description: posix locks not released on fd closure on 2.6.9+
3764 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3765 kernels, this caused such locks not to be released on fd close and
3766 then assertions on fs unmount about still used locks.
3769 Frequency : MDS failover only, very rarely
3771 Description: clients may get ASSERTION(granted_lock != NULL)
3772 Details : When request was taking a long time, and a client was resending
3773 a getattr by name lock request. The were multiple lock requests
3774 with the same client lock handle and
3775 mds_getattr_name->fixup_handle_for_resent_request found one of the
3776 lock handles but later failed with ASSERTION(granted_lock != NULL).
3781 Description: handle->h_buffer_credits > 0, assertion failure
3782 Details : h_buffer_credits is zero after truncate, causing assertion
3783 failure. This patch extends the transaction or creates a new
3787 Frequency : NFS re-export or patchless client
3788 Bugzilla : 11179, 10796
3789 Description: Crash on NFS re-export node (__d_move)
3790 Details : We do not want to hash the dentry if we don't have a lock.
3791 But if this dentry is later used in d_move, we'd hit uninitialised
3792 list head d_hash, so we just do this to init d_hash field but
3793 leave dentry unhashed.
3796 Frequency : NFS re-export or patchless client
3798 Description: NFS exports has problem with symbolic link
3799 Details : lustre client didn't properly install dentry when re-exported
3800 to NFS or running patchless client.
3803 Frequency : NFS re-export or patchless client
3805 Description: Various nfs/patchless fixes.
3806 Details : fixes reuse disconected alias for lookup process - this fixes
3807 warning "find_exported_dentry: npd != pd",
3808 fix permission error with open files at nfs.
3809 fix apply umask when do revalidate.
3812 Frequency : occasional
3814 Description: Crash on NFS re-export node
3815 Details : calling clear_page() on the wrong pointer triggered oops in
3816 generic_mapping_read().
3819 Frequency : rarely, using O_DIRECT IO
3821 Description: unaligned directio crashes client with LASSERT
3822 Details : check for unaligned buffers before trying any requests.
3825 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3827 Description: stale data returned from RAID cache
3828 Details : If only a small amount of IO is done to the RAID device before
3829 reading it again it is possible to get stale data from the RAID
3830 cache instead of reading it from disk.
3833 Frequency : always for sles10 kernel
3835 Description: sles10 support
3836 Details : ll_follow_link: compile fixes and using of nd_set_link
3837 under newer kernels.
3840 Frequency : depends on arch, kernel and compiler version, always on sles10
3843 Description: recursive or deep enough symlinks cause stack overflow
3844 Details : getting rid of large stack-allocated variable in
3848 Frequency : depends on hardware
3850 Description: lustre write performance loss in the SLES10 kernel
3851 Details : the performance loss is caused by using of write barriers in the
3852 ext3 code. The SLES10 kernel turns barrier support on by
3853 default. The fix is to undo that change for ldiskfs.
3855 ------------------------------------------------------------------------------
3857 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3859 * Support for kernels:
3860 2.6.9-42.0.3EL (RHEL 4)
3861 2.6.5-7.276 (SLES 9)
3862 2.4.21-47.0.1.EL (RHEL 3)
3863 2.6.12.6 vanilla (kernel.org)
3867 Frequency : quota enabled and large files being deleted
3869 Description: releasing more than 4GB of quota at once hangs OST
3870 Details : If a user deletes more than 4GB of files on a single OST it
3871 will cause the OST to spin in an infinite loop. Release
3872 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3877 Description: statfs data retrieved from /proc may be stale or zero
3878 Details : When reading per-device statfs data from /proc, in the
3879 {kbytes,files}_{total,free,avail} files, it may appear
3880 as zero or be out of date.
3883 Frequency : systems with MD RAID1 external journal devices
3885 Description: lconf's call to blkid is confused by RAID1 journal devices
3886 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3887 of returning all block devices that match the journal UUID.
3890 Frequency : always, for aggregate stripe size over 4GB
3892 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3893 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3894 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3897 Frequency : always if "lfs find" used on a local file/directory
3899 Description: "lfs find" segfaults if used on a local file/directory
3900 Details : The case where a directory component was not specified wasn't
3901 handled correctly. Handle this properly.
3904 Frequency : always on ppc64
3906 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3907 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3908 perform endianness conversion.
3911 Frequency : rarely (truncate to non-zero file size after write under load)
3912 Bugzilla : 10730, 10687
3913 Description: Files padded with zeros to next 4K multiple
3914 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3915 it is possible that files that are truncated to a non-zero size
3916 immediately after being written are filled with zero bytes beyond
3917 the truncated size. No file data is lost.
3919 Severity : enhancement
3921 Description: Allow recovery/failover for liblustre clients.
3922 Details : liblustre clients were unaware of failover configurations until
3925 Severity : enhancement
3927 Description: user file locks should fail when not mounting with flock option
3928 Details : Set up an error-returning stub in ll_file_operations.lock field
3929 to prevent incorrect behaviour when client is mounted without
3930 flock option. Also, set up properly f_op->flock field for
3934 Frequency : always on ia64
3936 Description: "lfs df" loops on printing out MDS statfs information
3937 Details : The obd_ioctl_data was not initialized and in some systems
3938 this caused a failure during the ioctl that did not return
3939 an error. Initialize the struct and return an error on failure.
3942 Frequency : SLES 9 only
3944 Description: Error of copying files with lustre special EAs as root
3945 Details : Client side always return success for setxattr call for lustre
3946 special xattr (currently only "trusted.lov").
3949 Frequency : rarely on clusters with both ia64+i386 clients
3951 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3952 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3953 clients are doing concurrent writes to the same file it is
3954 possible that the ia64 clients may LASSERT because the OST
3955 extent locks are not PAGE_SIZE aligned. Ensure that grown
3956 locks are always aligned on the request boundary.
3959 Frequency : specific use, occasional
3961 Description: Overwriting in use executable truncates on-disk binary image
3962 Details : If one node attempts to overwrite an executable in use by
3963 another node, we now correctly return ETXTBSY instead of
3964 truncating the file.
3966 Severity : enhancement
3968 Description: Async OSC create to avoid the blocking unnecessarily.
3969 Details : If an OST has no remaining object, system will block on the creating
3970 when need to create a new object on this OST. Now, ways use
3971 pre-created objects when available, instead of blocking on an
3972 empty osc while others are not empty. If we must block, we block
3973 for the shortest possible period of time.
3978 Description: chmod on Lustre root is propagated to other clients
3979 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3980 invalid by the follow_mount time.
3983 Frequency : liblustre clients only
3985 Description: Race in 'instant cancel' lock handling could lead to such locks
3986 never to be granted in case of SMP MDS
3987 Details : Do not destroy not yet granted but cbpending locks in
3991 Frequency : replay/resend of open
3993 Description: non null lock assetion failure in mds_intent_policy
3994 Details : Trying to replay/resend lockless open requests resulted in
3995 mds_open() returning 0 with no lock. Now it sets a flag if
3996 a lock is going to be returned.
3998 Severity : enhancement
4000 Description: Checksum enhancements
4001 Details : New checksum enhancements allow for resending RPCs that failed
4004 Severity : enhancement
4006 Description: Tunables on number of dirty pages in cacche
4007 Details : Allow to set limit on number of dirty pages cached.
4012 Description: client crash on unmount - lock still has references
4013 Details : In some error handling cases it was possible to leak a lock
4014 reference on a client while accessing a file. This was not
4015 harmful to the client during operation, but would cause the
4016 client to crash when the filesystem is unmounted.
4019 Frequency : specific case, rare
4021 Description: ETXTBSY on mds though file not in use by client
4022 Details : ETXTBSY is no longer incorrectly returned when attempting to
4023 chmod or chown a directory that the user previously tried to
4024 execute or a currently-executing binary.
4027 Frequency : extremely rare except on liblustre-based clients
4029 Description: Lustre space not freed when files are deleted
4030 Details : Clean up open-unlinked files after client eviction. Previously
4031 the unlink was skipped and the files remained as orphans.
4036 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4037 Details : In some cases it was possible to send a blocking callback to
4038 a client doing a glimpse, even though that client didn't get
4039 a lock granted. When the glimpse lock is cancelled on the OST
4040 the freed lock is left on the waiting list and corrupted the list.
4043 Frequency : all core dumps
4045 Description: Broke core dumps to lustre
4046 Details : Negative dentry may be unhashed if parent does not have UPDATE
4047 lock, but some callers, e.g. do_coredump, expect dentry to be
4048 hashed after successful create, hash it in ll_create_it.
4050 ------------------------------------------------------------------------------
4052 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4054 * Support for kernels:
4055 2.6.9-42.0.2.EL (RHEL 4)
4056 2.6.5-7.276 (SLES 9)
4057 2.4.21-40.EL (RHEL 3)
4058 2.6.12.6 vanilla (kernel.org)
4062 Frequency : always on RHEL 3
4064 Description: Number of open files grows over time
4065 Details : The number of open files grows over time, whether or not
4066 Lustre is started. This was due to a filp leak introduced
4067 by one of our kernel patches.
4069 ------------------------------------------------------------------------------
4071 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4073 * Support for kernels:
4074 2.6.9-42.EL (RHEL 4)
4075 2.6.5-7.276 (SLES 9)
4076 2.4.21-40.EL (RHEL 3)
4077 2.6.12.6 vanilla (kernel.org)
4082 Bugzilla : 5719, 9635, 9792, 9684
4083 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4084 Details : If a server is under heavy load and cannot reply to new
4085 connection requests before the client resends the (re)connect,
4086 the connection handling code can behave badly if two service
4087 threads are concurrently handing separate (re)connections from
4088 the same client. Add better locking to the connection handling
4089 code, and ensure that only a single connection will be processed
4090 for a given client UUID, even if the lock is dropped.
4092 Severity : enhancement
4094 Description: add TCP zero-copy support to kernel
4095 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4096 sends if the hardware supports scatter-gather and checksumming.
4097 This allows socklnd to do client-write and server-read more
4098 efficiently and reduce CPU utilization from skbuf copying.
4101 Frequency : only if NFS exporting from client
4103 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4104 Details : If setting attributes on a file created under NFS that had
4105 never been opened it would be possible to oops the client
4106 if the file had no objects.
4109 Frequency : always for liblustre
4111 Description: liblustre client does MDS+OSTs setattr RPC for each write
4112 Details : When doing a write from a liblustre client, the client
4113 incorrectly issued an RPC to the MDS and each OST the file was
4114 striped over in order to update the timestamps. When writing
4115 with small chunks and many clients this could overwhelm the MDS
4116 with RPCs. In all cases it would slow down the write because
4117 these RPCs are unnecessary.
4119 Severity : enhancement
4121 Description: allow number of MDS service threads to be changed at module load
4122 Details : It is now possible to change the number of MDS service threads
4123 running. Adding "options mds mds_num_threads={N}" to the MDS's
4124 /etc/modprobe.conf will set the number of threads for the next
4125 time Lustre is restarted (assuming the "mds" module is also
4126 reloaded at that time). The default number of threads will
4127 stay the same, 32 for most systems.
4132 Description: OST crash if filesystem is unformatted or corrupt
4133 Details : If an OST is started on a device that has never been formatted
4134 or if the filesystem is corrupt and cannot even mount then the
4135 error handling cleanup routines would dereference a NULL pointer.
4140 Description: NULL pointer deref in llap_from_page.
4141 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4142 incorrect) mapping if the page was truncated/reclaimed while it was
4143 searched for. Check for this condition and skip such pages when
4144 doing readahead. Introduce extra check to llap_from_page() to
4145 verify page->mapping->host is non-NULL (so page is not anonymous).
4148 Frequency : Sometimes when using sys_sendfile
4150 Description: "page not covered by a lock" warnings from ll_readpage
4151 Details : sendfile called ll_readpage without right page locks present.
4152 Now we introduced ll_file_sendfile that does necessary locking
4153 around call to generic_file_sendfile() much like we do in
4157 Frequency : with certain MDS communication failures at client mount time
4159 Description: NULL pointer deref after failed client mount
4160 Details : a client connection request may delayed by the network layer
4161 and not be sent until after the PTLRPC layer has timed out the
4162 request. If the client fails the mount immediately it will try
4163 to clean up before the network times out the request. Add a
4164 reference from the request import to the obd device and delay
4165 the cleanup until the network drops the request.
4168 Frequency : occasionally during client (re)connect
4170 Description: assertion failure during client (re)connect
4171 Details : processing a client connection request may be delayed by the
4172 client or server longer than the client connect timeout. This
4173 causes the client to resend the connection request. If the
4174 original connection request is replied in this interval, the
4175 client may trip an assertion failure in ptlrpc_connect_interpret()
4176 which thought it would be the only running connect process.
4179 Frequency : only with obd_echo servers and clients that are rebooted
4181 Description: kernel BUG accessing uninitialized data structure
4182 Details : When running an obd_echo server it did not start the ping_evictor
4183 thread, and when a client was evicted an uninitialized data
4184 structure was accessed. Start the ping_evictor in the RPC
4185 service startup instead of the OBD startup.
4187 Severity : enhancement
4188 Bugzilla : 10193 (patchless)
4189 Description: Remove dependency on various unexported kernel interfaces.
4190 Details : No longer need reparent_to_init, exit_mm, exit_files,
4191 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4194 Frequency : rare (only users of deprecated and unsupported LDAP config)
4196 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4197 Details : LDAP apparently contains 'lustreName' attributes instead of
4198 'name'. A simple remapping of the name is sufficient.
4201 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4203 Description: waiting_locks_callback trips kernel BUG if client is evicted
4204 Details : Running with the dump_on_timeout debug flag turned on makes
4205 it possible that the waiting_locks_callback() can try to dump
4206 the Lustre kernel debug logs from an interrupt handler. Defer
4207 this log dumping to the expired_lock_main() thread.
4209 Severity : enhancement
4211 Description: Support NFS exporting on 2.6 kernels.
4212 Details : Implement non-rawops metadata methods for NFS server to use without
4213 changing NFS server code.
4216 Frequency : very rare (synthetic metadata workload only)
4218 Description: two racing renames might cause an MDS thread to deadlock
4219 Details : Running the "racer" program may cause one MDS thread to rename
4220 a file from being the source of a rename to being the target of
4221 a rename at exactly the same time that another thread is doing
4222 so, and the second thread has already enqueued these locks after
4223 doing a lookup of the target and is trying to relock them in
4224 order. Ensure that we don't try to re-lock the same resource.
4227 Frequency : only very large systems with liblustre clients
4229 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4230 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4231 all OSTs in parallel. This allows the eviction of stale liblustre
4232 clients to proceed much faster than if they were done in series,
4233 and also offers similar improvements for other set_info RPCs.
4238 Description: excessive CPU usage during initial read phase on client
4239 Details : During the initial read phase on a client, it would agressively
4240 retry readahead on the file, consuming too much CPU and impacting
4241 performance (since 1.4.5.8). Improve the readahead algorithm
4242 to avoid this, and also improve some other common cases (read
4243 of small files in particular, where "small" is files smaller than
4244 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4249 Description: MDS crash when receiving packet with unknown intent.
4250 Details : Do not LBUG in unknown intent case, just return -EFAULT
4252 Severity : enhancement
4253 Bugzilla : 9293, 9385
4254 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4255 Details : Do not serialize getattr (non-intent version) and statfs.
4258 Frequency : occasional, when OST network is overloaded/intermittent
4260 Description: client evicted by OST after bulk IO timeout
4261 Details : If a client sends a bulk IO request (read or write) the OST
4262 may evict the client if it is unresposive to its data GET/PUT
4263 request. This is incorrect if the network is overloaded (takes
4264 too long to transfer the RPC data) or dropped the OST GET/PUT
4265 request. There is no need to evict the client at all, since
4266 the pinger and/or lock callbacks will handle this, and the
4267 client can restart the bulk request.
4270 Frequency : Always when mmapping file with no objects
4272 Description: client crashes when mmapping file with no objects
4273 Details : Check that we actually have objects in a file before doing any
4274 operations on objects in ll_vm_open, ll_vm_close and
4280 Description: Request leak when working with deleted CWD
4281 Details : Introduce advanced request refcount tracking for requests
4282 referenced from lustre intent.
4284 Severity : Enhancement
4286 Description: Cache open file handles on client.
4287 Details : MDS now will return special lock along with openhandle, if
4288 requested and client is allowed to hold openhandle, even if unused,
4289 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4290 closing files for every read/write openration.
4292 Severity : Enhancement
4294 Description: Cache open negative dentries on client when possible.
4295 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4296 negative dentries on lock revocation.
4301 Description: Remounting a client read-only wasn't possible with a zconf mount
4302 Details : It wasn't possible to remount a client read-only with llmount.
4304 Severity : enhancement
4305 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4306 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4307 MPICH is included to add Lustre-specific ADIO interfaces.
4308 This is based closely on the UFS ADIO layer and only differs
4309 in file creation, in order to allow the OST striping to be set.
4310 This is user-contributed code and not supported by CFS.
4315 Description: extended inode attributes (immutable, append-only) work improperly
4316 when 2.4 and 2.6 kernels are used on client/server or vice versa
4317 Details : Introduce kernel-independent values for these flags.
4319 Severity : enhancement
4322 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4323 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4324 granularity. Now, Fractional MB granularity is be supported,
4325 this is very useful for low memory system.
4327 Severity : enhancement
4329 Description: Getattr by fid
4330 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4331 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4332 have locks and avoids getattr rpc requests for it on every stat.
4335 Frequency : Always, for filesystems larger than 2TB
4337 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4338 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4339 filesystems larger than 2TB to be mounted with mballoc enabled.
4342 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4344 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4345 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4346 filesystems larger than 2TB is mounted then the kernel will
4347 silently corrupt the start of the filesystem. CONFIG_LBD is
4348 enabled for all CFS-supported kernels, but the possibility of
4349 this happening with a modified kernel config exists.
4351 Severity : enhancement
4353 Description: add client O_DIRECT support for 2.6 kernels
4354 Details : It is now possible to do O_DIRECT reads and writes to files
4355 in the Lustre client mountpoint on 2.6 kernel clients.
4357 Severity : enhancement
4359 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4360 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4361 parallel, not waiting for response from every OST before sending
4362 a rpc to the next OST.
4367 Description: setattr vs write race when updating file timestamps
4368 Details : Client processes that update a file timestamp into the past
4369 right after writing to the file (e.g. tar) it is possible that
4370 the updated file modification time can be reset to the current
4371 time due to a race between processing the setattr and write RPC.
4373 Severity : enhancement
4375 Description: Bring 'lfs find' closer in line with regular Linux find.
4376 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4377 -print0 options and obtains all the needed info through the lustre
4380 Severity : enhancement
4382 Description: support up to 1024 configured devices on one node
4383 Details : change obd_dev array from statically allocated to dynamically
4384 allocated structs as they are first used to reduce memory usage
4389 Description: Flush dirty partially truncated pages during truncate
4390 Details : Immediatelly flush partially truncated pages in filter_setattr,
4391 this way we completely avoid having any pages in page cache on OST
4392 and can retire ugly workarounds during writes to flush such pages.
4397 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4398 Details : protect inode from truncation within vfs_unlink() context
4399 just take a reference before calling vfs_unlink() and release it
4400 when parent's i_sem is free.
4405 Description: last_id value checked outside lock on OST caused LASSERT failure
4406 Details : If there were multiple MDS->OST object precreate requests in
4407 flight, it was possible that the OST's last object id was checked
4408 outside a lock and incorrectly tripped an assertion. Move checks
4409 inside locks, and discard old precreate requests.
4412 Frequency : always, if extents are used on OSTs
4414 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4415 Details : OSTs using the extents format would not zero the high 16 bits of
4416 the index physical block number. This is not a problem for any
4417 OST filesystems smaller than 16TB, and no kernels support ext3
4418 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4419 new/modified files) and can be fixed for existing filesystems
4420 with e2fsprogs-1.39-cfs1.
4425 Description: import connection selection may be incorrect if timer wraps
4426 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4427 import connections to be ignored if the 32-bit jiffies counter
4428 wraps. Use a 64-bit jiffies counter.
4431 Frequency : very large clusters immediately after boot
4433 Description: LNET request buffers exhausted under heavy short-term load
4434 Details : If a large number of client requests are generated on a service
4435 that has previously never seen so many requests it is possible
4436 that the request buffer growth cannot keep up with the spike in
4437 demand. Instead of dropping incoming requests, they are held in
4438 the LND until the RPC service can accept more requests.
4441 Frequency : Sometimes during replay
4443 Description: Assertion failure in ll_local_open after replay.
4444 Details : If replay happened on an open request reply before we were able
4445 to set replay handler, reply will become not swabbed tripping the
4446 assertion in ll_local_open. Now we set the handler right after
4447 recognising of open request
4450 Frequency : very rare
4452 Description: kernel reports "badness in vsnprintf"
4453 Details : Reading from the "recovery_status" /proc file in small chunks
4454 may cause a negative length in lprocfs_obd_rd_recovery_status()
4455 call to vsnprintf() (which is otherwise harmless). Exit early
4456 if there is no more space in the output buffer.
4458 Severity : enhancement
4460 Description: clear OBD RPC statistics by writing to them
4461 Details : It is now possible to clear the OBD RPC statistics by writing
4462 to the "stats" file.
4467 Description: Client mtime is not the same on different clients after utimes
4468 Details : In some cases, the client was using the utimes() syscall on
4469 a file cached on another node. The clients now validate the
4470 ctime from the MDS + OSTs to determine which one is right.
4475 Description: Inability to activate failout mode
4476 Details : lconf script incorrectly assumed that in pythong string's numeric
4477 value is used in comparisons.
4480 Frequency : always with multiple stripes per file
4482 Description: Inefficient object allocation for mutli-stripe files
4483 Details : When selecting which OSTs to stripe files over, for files with
4484 a stripe count that divides evenly into the number of OSTs,
4485 the MDS is always picking the same starting OST for each file.
4486 Return the OST selection heuristic to the original design.
4491 Description: mount failures may take full timeout to return an error
4492 Details : Under some heavy load conditions it is possible that a
4493 failed mount can wait for the full obd_timeout interval,
4494 possibly several minutes, before reporting an error.
4495 Instead return an error as soon as the status is known.
4497 Frequency : quota enabled and large files being deleted
4499 Description: releasing more than 4GB of quota at once hangs OST
4500 Details : If a user deletes more than 4GB of files on a single OST it
4501 will cause the OST to spin in an infinite loop. Release
4502 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4507 Description: statfs data retrieved from /proc may be stale or zero
4508 Details : When reading per-device statfs data from /proc, in the
4509 {kbytes,files}_{total,free,avail} files, it may appear
4510 as zero or be out of date.
4513 Frequency : systems with MD RAID1 external journal devices
4515 Description: lconf's call to blkid is confused by RAID1 journal devices
4516 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4517 of returning all block devices that match the journal UUID.
4520 Frequency : always, for aggregate stripe size over 4GB
4522 Description: assertion fails when trying to use 4GB stripe size
4523 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4524 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4527 Frequency : always on ppc64
4529 Description: the first write on an ext3 filesystem with mballoc got stuck
4530 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4531 perform endianness conversion.
4533 ------------------------------------------------------------------------------
4535 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4537 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4538 INTEROPERATE with older versions automatically. Please read the
4539 user documentation before upgrading any part of a live system.
4540 * WARNING: Lustre networking configuration changes are required with
4541 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4544 * Support for kernels:
4545 2.6.9-22.0.2.EL (RHEL 4)
4546 2.6.5-7.244 (SLES 9)
4547 2.6.12.6 vanilla (kernel.org)
4550 Severity : enhancement
4551 Bugzilla : 7981/8208
4552 Description: Introduced Lustre Networking (LNET)
4553 Details : LNET is new networking infrastructure for Lustre, it includes
4554 a reorganized network configuration mode (see the user
4555 documentation for full details) as well as support for routing
4556 between different network fabrics. Lustre Networking Devices
4557 (LNDs) for the supported network fabrics have also been
4558 created for this new infrastructure.
4560 Severity : enhancement
4561 Description: Introduced Access control lists
4562 Details : clients can set ACLs on files and directories in order to have
4563 more fine-grained permissions than the standard Unix UGO+RWX.
4564 The MDS must be started with the "-o acl" mount option.
4566 Severity : enhancement
4567 Description: Introduced filesystem quotas
4568 Details : Administrators may now establish per-user quotas on the
4571 Severity : enhancement
4573 Description: Configuration change for the XT3
4574 The PTLLND is now used to run Lustre over Portals on the XT3
4575 The configure option(s) --with-cray-portals are no longer used.
4576 Rather --with-portals=<path-to-portals-includes> is used to
4577 enable building on the XT3. In addition to enable XT3 specific
4578 features the option --enable-cray-xt3 must be used.
4583 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4584 Details : A race between allocating a new llcd and re-getting the llcd_lock
4585 allowed another thread to grab newly-allocated llcd.
4587 Severity : enhancement
4589 Description: 2.6 OST async journal commit and locking fix to improve performance
4590 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4591 2.6 kernels are now async as they already were in 2.4 kernels so
4592 that they can commit concurrently with the network bulk transfer.
4593 For block-allocated files the filter allocation semaphore is held
4594 to avoid filesystem fragmentation during allocation. BKL lock
4595 removed for 2.6 xattr operations where it is no longer needed.
4600 Description: lconf incorrectly determined whether two IP networks could talk
4601 Details : In some more complicated routing and multiple-network
4602 configurations, lconf will avoid trying to make a network
4603 connection to a disjoint part of the IP space. It was doing the
4604 math incorrectly for one set of cases.
4609 Description: Fix for potential infinite loop processing records in an llog.
4610 Details : If an llog record is corrupted/zeroed, it is possible to loop
4611 forever in llog_process(). Validate the llog record length
4612 and skip the remainder of the block on error.
4615 Frequency : occasional (liblustre only)
4617 Description: liblustre could not open files whose last component is a symlink
4618 Details : sysio_path_walk() would incorrectly pass the open intent to
4619 intermediate path components.
4622 Frequency : rare (liblustre only with non-standard tuning)
4623 Bugzilla : 7201 (7350)
4624 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4625 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4626 after it has been released, only do this for referenced locks
4628 Severity : enhancement
4630 Description: specifying an (invalid) directory default stripe_size of -1
4631 would reset the directory default striping
4632 Details : stripe_size -1 was used internally to signal directory stripe
4633 removal, now use "all default" to signal dir stripe removal
4634 as a directory striping of "all default" is not useful
4637 Frequency : common for large clusters running liblustre clients
4639 Description: doing an ls when liblustre clients are running is slow
4640 Details : sending a glimpse AST to a liblustre client waits for every AST
4641 to time out, as liblustre clients will not respond. Since they
4642 cannot cache data we refresh the OST lock LVB from disk instead.
4644 Severity : enhancement
4646 Description: doing an ls at the same time as file IO can be slow
4647 Details : enqueue and other "small" requests can be blocked behind many
4648 large IO requests. Create a new OST IO portal for non-IO
4649 requests so they can be processed faster.
4652 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4654 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4655 created file with mode 000
4656 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4657 allow an ftruncate() as a non-root user to a file with mode 000.
4658 Lustre now respects this flag to disable mode checks when
4659 truncating a file owned by the user
4662 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4665 Description: Revoking locks from clients that went dead or catatonic might take
4667 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4668 cancellation of such locks instant on servers without waiting for
4669 any reply from clients. Clients drops these locks when cancel
4670 notification from server is received without replying.
4673 Frequency : liblustre-only, when liblustre client dies or becomes busy
4675 Description: Doing ls on Linux clients can take a long time with active
4677 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4678 granting such locks to it in the first place if possible. Locks
4679 are taken by proxy on the OST during the read or write and
4680 dropped immediately afterward. Add connect flags handling, do
4681 not grant locks to liblustre clients for glimpse ASTs.
4683 Severity : enhancement
4685 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4686 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4687 sequential reads and behaves badly with applications doing random
4688 reads. Improve it by reading ahead at least read region, and
4689 avoiding excessive large RPC for small reads.
4691 Severity : enhancement
4693 Description: Creating more than 1000 files for a single job may cause a load
4694 imbalance on the OSTs if there are also a large number of OSTs.
4695 Details : qos_prep_create() uses an OST index reseed value that is an
4696 even multiple of the number of available OSTs so that if the
4697 reseed happens in the middle of the object allocation it will
4698 still utilize the OSTs as uniformly as possible.
4703 Description: OST or MDS may oops in ping_evictor_main()
4704 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4705 but doesn't restart at beginning of obd_exports_timed list
4708 Severity : enhancement
4710 Description: improve by-nid export eviction on the MDS and OST
4711 Details : allow multiple exports with the same NID to be evicted at one
4712 time without re-searching the exports list.
4715 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4717 Description: MDS may oops in groups_free()
4718 Details : in rare race conditions a newly allocated group_info struct is
4719 freed again, and this can be NULL. The 2.4 compatibility code
4720 for groups_free() checked for a NULL pointer, but 2.6 did not.
4723 Frequency : common for liblustre clients doing little filesystem IO
4724 Bugzilla : 9352, 7313
4725 Description: server may evict liblustre clients accessing contended locks
4726 Details : if a client is granted a lock or receives a completion AST
4727 with a blocking AST already set it would not reply to the AST
4728 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4731 Frequency : lfs setstripe, only systems with more than 160 OSTs
4733 Description: unable to set striping with a starting offset beyond OST 160
4734 Details : llapi_create_file() incorrectly limited the starting stripe
4735 index to the maximum single-file stripe count.
4738 Frequency : LDAP users only
4740 Description: lconf did not handle in-kernel recovery with LDAP properly
4741 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4743 Severity : enhancement
4745 Description: bind OST threads to NUMA nodes to improve performance
4746 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4747 node and do their allocations there to localize memory access
4749 Severity : enhancement
4751 Description: llmount can determine client NID directly from Myrinet (GM)
4752 Details : the client NID code from gmnalnid was moved directly into
4753 llmount, removing the need to use this or specifying the
4754 client NID explicitly when mounting GM clients with zeroconf
4757 Frequency : if client is started with down MDS
4759 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4760 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4761 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4762 ptlrpc_import_delayed_req() because we didn't check if the
4763 request was interrupted, and we also didn't break out of the
4764 event loop if there was no timeout
4769 Description: data loss during non-page-aligned writes to a single file from
4770 both multiple nodes and multiple threads on one node at same time
4771 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4772 inconsistency led to false short-reads, that were cached and later
4773 used by ->prepare_write() to fill in partially written page,
4774 leading to data loss.
4777 Frequency : always, if lconf --abort_recovery used
4779 Description: lconf --abort_recovery fails with 'Operation not supported'
4780 Details : lconf was attempting to abort recovery on the MDT device and not
4783 Severity : enhancement
4785 Description: remove cleanup logs
4786 Details : replace lconf-generated cleanup logs with lustre internal
4787 cleanup routines. Eliminates the need for client-cleanup and
4790 Severity : enhancement
4792 Description: add support for EAs (user and system) on lustre filesystems
4793 Details : it is now possible to store extended attributes in the Lustre
4794 client filesystem, and with the user_xattr mount option it
4795 is possible to allow users to store EAs on their files also
4797 Severity : enhancement
4799 Description: Add possibility (config option) to show minimal available OST free
4801 Details : When compiled with --enable-mindf configure option, statfs(2)
4802 (and so, df) will return least minimal free space available from
4803 all OSTs as amount of free space on FS, instead of summary of
4804 free spaces of all OSTs.
4806 Severity : enhancement
4808 Description: do not expand extent locks acquired on OST-side
4809 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4810 by server: they are not cached anyway.
4813 Frequency : when mmap is used/binaries executed from Lustre
4815 Description: Unmmap pages before throwing them away from read cache.
4816 Details : llap_shrink cache now attempts to unmap pages before discarding
4817 them (if unmapping failed - do not discard). SLES9 kernel has
4818 extra checks that trigger if this unmapping is not done first.
4823 Description: lconf didn't resolve symlinks before checking to see whether a
4824 given mountpoint was already in use
4827 Frequency : when migrating failover services
4828 Bugzilla : 6395, 9514
4829 Description: When migrating a subset of services from a node (e.g. failback
4830 from a failover service node) the remaining services would
4831 time out and evict clients.
4832 Details : lconf --force (implied by --failover) sets the global obd_timeout
4833 to 5 seconds in order to quickly disconnect, but this caused
4834 other RPCs to time out too quickly. Do not change the global
4835 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4837 Severity : enhancement
4838 Frequency : if MDS is started with down OST
4839 Bugzilla : 9439,5706
4840 Description: Allow startup/shutdown of an MDS without depending on the
4841 availability of the OSTs.
4842 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4843 Add appropriate locking and lov-osc refcounts for safe
4844 cleaning. Add osc abort_inflight calls in case the
4845 synchronize never started.
4848 Frequency : occasional (Cray XT3 only)
4850 Description: root not authorized to access files in CRAY_PORTALS environment
4851 Details : The client process capabilities were not honoured on the MDS in
4852 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4853 been accessed by an authorized user then root was able to access
4854 the file on the local client also. The root user capabilities
4855 are now allowed on the MDS, as this environment has secure UID.
4858 Frequency : occasional
4860 Description: ldiskfs "too long searching" message happens too often
4861 Details : A debugging message (otherwise harmless) prints too often on
4862 the OST console. This has been reduced to only happen when
4863 there are fragmentation problems on the filesystem.
4868 Description: Division by zero in statfs when all OSCs are inactive
4869 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4870 lov_statfs divides by value returned by lov_get_stripecnt().
4874 Bugzilla : 9489, 3273
4875 Description: First write from each client to each OST was only 4kB in size,
4876 to initialize client writeback cache, which caused sub-optimal
4877 RPCs and poor layout on disk for the first written file.
4878 Details : Clients now request an initial cache grant at (re)connect time
4879 and so that they can start streaming writes to the cache right
4880 away and always do full-sized RPCs if there is enough data.
4881 If the OST is rebooted the client also re-establishes its grant
4882 so that client cached writes will be honoured under the grant.
4887 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4888 Details : Now I/O RPCs go to different portal number and (presumably) fast
4889 lock requests (and glimses) and other RPCs get their own service
4890 threads pool that should be able to service those RPCs
4893 Severity : enhancement
4895 Description: Ability to exchange lustre version between client and servers and
4896 issue warnings at client side if client is too old. Also for
4897 liblustre clients there is ability to refuse connection of too old
4899 Details : New 'version' field is added to connect data structure that is
4900 filled with version info. That info is later checked by server and
4904 Frequency : rare, liblustre only.
4905 Bugzilla : 9296, 9581
4906 Description: Two simultaneous writes from liblustre at offset within same page
4907 might proceed at the same time overwriting eachother with stale
4909 Details : I/O lock withing llu_file_prwv was released too early, before data
4910 actually was hitting the wire. Extended lock-holding time until
4911 server acknowledges receiving data.
4914 Frequency : extremely rare. Never observed in practice.
4916 Description: avoid generating lustre_handle cookie of 0.
4917 Details : class_handle_hash() generates handle cookies by incrementing
4918 global counter, and can hit 0 occasionaly (this is unlikely, but
4919 not impossible, because initial value of cookie counter is
4920 selected randonly). Value of 0 is used as a sentinel meaning
4921 "unassigned handle" --- avoid it. Also coalesce two critical
4922 sections in this function into one.
4924 Severity : enhancement
4926 Description: allow liblustre clients to delegate truncate locking to OST
4927 Details : To avoid overhead of locking, liblustre client instructs OST to
4928 take extent lock in ost_punch() on client's behalf. New connection
4929 flag is added to handle backward compatibility.
4931 Severity : enhancement
4932 Bugzilla : 4928, 7341, 9758
4933 Description: allow number of OST service threads to be specified
4934 Details : a module parameter allows the number of OST service threads
4935 to be specified via "options ost ost_num_threads={N}" in the
4936 OSS's /etc/modules.conf or /etc/modprobe.conf.
4940 Bugzilla : 6146, 9635, 9895
4941 Description: servers crash with bad pointer in target_handle_connect()
4942 Details : In rare cases when a client is reconnecting it was possible that
4943 the connection request was the last reference for that export.
4944 We would temporarily drop the export reference and get a new
4945 one, but this may have been the last reference and the export
4946 was just destroyed. Get new reference before dropping old one.
4948 Severity : enhancement
4949 Frequency : if client is started with failover MDS
4951 Description: Allow multiple MDS hostnames in the mount command
4952 Details : Try to read the configuration from all specified MDS
4953 hostnames during a client mount in case the "primary"
4956 Severity : enhancement
4958 Description: Stop sending data to evicted clients as soon as possible.
4959 Details : Check if the client we are about to send or are sending data to
4960 was evicted already. (Check is done every second of waiting,
4961 for which l_wait_event interface was extended to allow checking
4962 of exit condition at specified intervals).
4965 Frequency : rare, normally only when NFS exporting is done from client
4967 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4969 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4972 Frequency : always, if srand() is called before liblustre initialization
4974 Description: Liblustre uses system PRNG disturbing its usage by user application
4975 Details : Introduce internal to lustre fast and high-quality PRNG for
4976 lustre usage and make liblustre and some other places in generic
4977 lustre code to use it.
4979 Severity : enhancement
4980 Bugzilla : 9477, 9557, 9870
4981 Description: Verify that the MDS configuration logs are updated when xml is
4982 Details : Check if the .xml configuration logs are newer than the config
4983 logs stored on the MDS and report an error if this is the case.
4984 Request --write-conf, or allow starting with --old_conf.
4986 Severity : enhancement
4988 Description: Handle symlinks in the path when checking if Lustre is mounted.
4989 Details : Resolve intermediate symlinks when checking if a client has
4990 mounted a filesystem to avoid duplicate client mounts.
4995 Description: lconf can hit an error exception but still return success.
4996 Details : The lconf command catches the Command error exception at the top
4997 level script context and will exit with the associated exit
4998 status, but doesn't ensure that this exit status is non-zero.
5003 Description: failure of ptlrpc thread startup can cause oops
5004 Details : Starting a ptlrpc service thread can fail if there are a large
5005 number of threads or the server memory is very fragmented.
5006 Handle this without oopsing.
5009 Frequency : always, only if liblustre and non-default acceptor port was used
5011 Description: liblustre cannot connect to servers with non-default acceptor port
5012 Details : tcpnal_set_default_params() was not called and was therefore
5013 ignoring the environment varaible TCPNAL_PORT, as well as other
5014 TCPNAL_ environment variables
5019 Description: two objects could be created on the same OST for a single file
5020 Details : If an OST is down, in some cases it was possible to create two
5021 objects on a single OST for a single file. No problems other
5022 than potential performance impact and spurious error messages.
5026 Bugzilla : 5681, 9562
5027 Description: Client may oops in ll_unhash_aliases
5028 Details : Client dcache may become inconsistent in race condition.
5029 In some cases "getcwd" can fail if the current directory is
5035 Description: Inode refcounting problems in NFS export code
5036 Details : link_raw functions used to call d_instantiate without obtaining
5037 extra inode reference first.
5041 Bugzilla : 9942, 9903
5042 Description: Referencing freed requests leading to crash, memleaks with NFS.
5043 Details : We used to require that call to ll_revalidate_it was always
5044 followed by ll_lookup_it. Also with revalidate_special() it is
5045 possible to call ll_revalidate_it() twice for the same dentry
5046 even if first occurence returned success. This fix changes semantic
5047 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5048 reference on a request referred from the intent.
5049 ll_intent_release() then releases such a request.
5052 Frequency : rare, normally benchmark loads only
5054 Description: unlinked inodes were kept in memory on the client
5055 Details : If a client is repeatedly creating and unlinking files it
5056 can accumulate a lot of stale inodes in the inode slab cache.
5057 If there is no other client load running this can cause the
5058 client node to run out of memory. Instead flush old inodes
5059 from client cache that have the same inode number as a new inode.
5062 Frequency : SLES9 2.6.5 kernel and long filenames only
5063 Bugzilla : 9969, 10379
5064 Description: utime reports stale NFS file handle
5065 Details : SLES9 uses out-of-dentry names in some cases, which confused
5066 the lustre dentry revalidation. Change it to always use the
5070 Frequency : rare, unless heavy write-truncate concurrency is continuous
5071 Bugzilla : 4180, 6984, 7171, 9963, 9331
5072 Description: OST becomes very slow and/or deadlocked during object unlink
5073 Details : filter_destroy() was holding onto the parent directory lock
5074 while truncating+unlinking objects. For very large objects this
5075 may block other threads for a long time and slow overall OST
5076 responsiveness. It may also be possible to get a lock ordering
5077 deadlock in this case, or run out of journal credits because of
5078 the combined truncate+unlink. Solution is to do object truncate
5079 first in one transaction without parent lock, and then do the
5080 final unlink in a new transaction with the parent lock. This
5081 reduces the lock hold time dramatically.
5084 Frequency : rare, 2.4 kernels only
5086 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5087 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5088 before unmounting their filesystem in order to avoid deadlock.
5089 The kernel_locked() function in 2.4 kernels only checks whether
5090 the kernel lock is held, not whether it is this process that is
5091 holding it as 2.6 kernels do.
5096 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5097 Details : The client ptlrpc code may be trying to reconnect to a down
5098 server before a previous connection attempt has timed out.
5099 Increase the reconnect interval to be longer than the connection
5100 timeout interval to avoid sending duplicate connections to
5104 Frequency : echo_client brw_test command
5106 Description: fix echo_client to work with OST preallocated code
5107 Details : OST preallocation code (5137) didn't take echo_client IO path
5108 into account: echo_client calls filter methods outside of any
5109 OST thread and, hence, there is no per-thread preallocated
5110 pages and buffers to use. Solution: hijack pga pages for IO. As
5111 a byproduct, this avoids unnecessary data copying.
5115 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5116 Description: Client can oops in mdc_commit_close() after open replay
5117 Details : It was possible for the MDS to return an open request with no
5118 transaction number in mds_finish_transno() if the client was
5119 evicted, but without actually returning an error. Clients
5120 would later try to replay that open and may trip an assertion
5121 Simplify the client close codepath, and always return an error
5122 from the MDS in case the open is not successful.
5125 Frequency : rare, 2.6 OSTs only
5127 Description: OST may deadlock under high load on fragmented files
5128 Details : If there was a heavy load and highly-fragmented OST filesystems
5129 it was possible to have all the OST threads deadlock waiting on
5130 allocation of biovecs, because the biovecs were not released
5131 until the entire RPC IO was completed. Instead, release biovecs
5132 as soon as they are complete to ensure forward IO progress.
5134 Severity : enhancement
5136 Description: Support for specifying external journal device at mount
5137 Details : If an OST or MDS device is formatted with an external journal
5138 device, this device major/minor is stored in the ext3 superblock
5139 and may not be valid for failover. Allow detecting and
5140 specifying the external journal at mount time.
5145 Description: Mounting an MDS with pending unlinked files may cause oops
5146 Details : target_finish_recovery() calls mds_postrecov() which returned
5147 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5148 considers this an error and immediately begins cleaning up the
5149 lov, just after starting the mds_lov process
5151 Severity : enhancement
5153 Description: Implement 'lfs df' to report actual free space on per-OST basis
5154 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5155 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5156 usage of objects; '-h' to report in human readable format.
5158 ------------------------------------------------------------------------------
5160 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5167 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5168 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5169 at startup during orphan unlinking. Instead, skip buddy update
5170 before setup, as it will be regenerated after recovery is complete.
5175 Description: If an OST is inactive, its locks might reference stale inodes.
5176 Details : lov_change_cbdata() must iterate over all namespaces, even if
5177 they are inactive to clear inode references from the lock.
5179 Severity : enhancement
5180 Frequency : occasional, if non-standard max_dirty_mb used
5182 Description: Client will block write RPCs if not enough grant
5183 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5184 then the client will block writes while waiting for another RPC
5185 to complete instead of consuming its dirty limit. With change
5186 we get improved performance when max_dirty_mb is small.
5188 Severity : enhancement
5189 Bugzilla : 3389, 6253
5190 Description: Add support for supplementary groups on the MDS.
5191 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5192 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5193 lookups for user supplementary groups into a cache.
5197 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5198 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5199 for liblustre O_EXCL is enforced by MDS.
5203 Description: readdir never returns NULL in liblustre.
5204 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5205 dentry in struct dirent.
5209 Frequency : liblustre only, depends on application IO pattern
5210 Description: liblustre clients evicted if not contacting servers
5211 Details : Don't put liblustre clients into the ping_evictor list, so
5212 they will not be evicted by the pinger ever.
5214 Severity : enhancement
5216 Description: Add ability to evict clients by NID from MDS.
5217 Details : By echoing "nid:$NID" string into
5218 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5219 $NID would be instantly evicted from this MDS and from all active
5220 OSTs connected to it.
5224 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5225 Details : lookup_it_finish() used to query file size from OSTs that was not
5230 Description: service threads change working directory to that of init
5231 Details : Starting lustre service threads may pin the working directory
5232 of the parent thread, making that filesystem busy. Threads
5233 now change to the working directory of init to avoid this.
5237 Frequency : during shutdown only
5238 Description: shutdown with a failed MDS or OST can cause unmount to hang
5239 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5244 Frequency : 2.6 only
5245 Description: chmod/chown may include an extra supplementary group
5246 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5247 supplementary group and if none is specified this is used.
5250 Bugzilla : 5479 (6816)
5252 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5253 Details : If lookup is in progress on a file that is unlinked we might try
5254 to revalidate the inode and fail in revalidate after lookup is
5255 complete and ll_file_open() enqueues the open again but
5256 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5259 Frequency : always, if lconf --abort_recovery used
5261 Description: lconf --abort_recovery fails with 'Operation not supported'
5262 Details : lconf was attempting to abort recovery on the MDT device and not
5265 ------------------------------------------------------------------------------
5267 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5272 Frequency : rare (only unsupported configurations with a node running as an
5274 Bugzilla : 6514, 5137
5275 Description: Mounting a Lustre file system on a node running as an OST could
5277 Details : OSTs now preallocates memory needed to write out data at
5278 startup, instead of when needed, to avoid having to
5279 allocate memory in possibly low memory situations.
5280 Specifically, if the file system is mounted on on OST,
5281 memory pressure could force it to try to write out data,
5282 which it needed to allocate memory to do. Due to the low
5283 memory, it would be unable to do so and the node would
5284 become unresponsive.
5286 Severity : enhancement
5288 Description: Addition of lconf --service command line option
5289 Details : lconf now accepts a '--service <arg>' option, which is
5290 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5292 Severity : enhancement
5294 Description: Failover mode is now the default for OSTs.
5295 Details : By default, OSTs will now run in failover mode. To return to
5296 the old behaviour, add '--failout' to the lmc line for OSTs.
5298 Severity : enhancement
5300 Description: Health checks are now provided for MDS and OSTs
5301 Details : Additional detailed health check information on MSD and OSTs
5302 is now provided through the procfs health_check value.
5305 Frequency : occasional, depends on IO load
5307 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5308 after numerous create/delete cycles
5309 Details : The ext3 inode allocation policy would not allocate new inodes
5310 very well on the OSTs because there are no new directories
5311 being created. Instead we look for groups with free space if
5312 the parent directories are nearly full.
5317 Description: Network or server problems during mount may cause partially
5318 mounted clients instead of returning an error.
5319 Details : The config llog parsing code may overwrite the error return
5320 code during mount error handling, returning success instead
5326 Description: MDS can fail to allocate large reply buffers
5327 Details : After long uptimes the MDS can fail to allocate large reply
5328 buffers (e.g. zconf client mount config records) due to memory
5329 fragmentation or consumption by the buffer cache. Preallocate
5330 some large reply buffers so that these replies can be sent even
5331 under memory pressure.
5335 Frequency : rare (liblustre)
5336 Description: fsx running with liblustre complained that using truncate() to
5337 extend the file doesn't work. This patch corrects that issue.
5338 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5339 ATTR_SIZE and lsm use in llu_setattr_raw.
5343 Frequency : rare, only 2.6 kernels
5344 Description: Unusual file access patterns on the MDS may result in inode
5345 data being lost in very rare circumstances.
5346 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5347 code in the 2.6 kernel caused the inode and/or EA data not to
5348 be read from disk, causing single-file corruption.
5352 Frequency : rare, only 2.6 filesystems using extents
5353 Description: Heavy concurrent write and delete load may cause data corruption.
5354 Details : It was possible under high-load situations to have an extent
5355 metadata block in the block device cache from a just-unlinked
5356 file overwrite a newly-allocated data block. We now unmap any
5357 metadata buffers that alias just-allocated data blocks.
5361 Frequency : filesystems with default stripe_count larger than 77
5362 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5363 Details : lconf specifies an inode size of 4096 bytes when the default
5364 stripe_count is larger than 77. This conflicts with the default
5365 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5366 this case to avoid pinning too much memory for large EAs.
5368 ------------------------------------------------------------------------------
5370 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5375 Frequency : rare (extremely heavy IO load with hundreds of clients)
5377 Description: Client is evicted, gets IO error writing to file
5378 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5379 caused clients to be evicted for AST timeouts. The fixes in
5380 bug 5192 mean we no longer need to have such short AST timeouts
5381 so ldlm_timeout has been increased.
5384 Frequency : occasional during --force or --failover shutdown under load
5385 Bugzilla : 5949, 4834
5386 Description: Server oops/LBUG if stopped with --force or --failover under load
5387 Details : a collection of import/export refcount and cleanup ordering
5388 issues fixed for safer force cleanup
5391 Frequency : only filesystems larger than 120 OSTs
5392 Bugzilla : 5990, 6223
5393 Description: lfs getstripe would oops on a very large filesystem
5394 Details : lov_getconfig used kfree on vmalloc'd memory
5397 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5398 Bugzilla : 6242, 6243
5399 Description: reading from files that had been truncated to a non-zero size
5400 but never opened returned no data
5401 Details : ll_file_read() reads zeros from no-object files to EOF
5406 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5407 Details : There's little harm in aborting MDS/OSS recovery and letting it
5408 try again, so I removed the LASSERT and return an error instead.
5410 Severity : enhancement
5412 Description: New debugging infrastructure for tracking down data corruption
5413 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5414 (b) cover more of the client-side code path, and (c) try to narrow
5415 down where problems occurred
5419 Bugzilla : 3819, 4364, 4397, 6313
5420 Description: Racing close and eviction MDS could cause assertion in mds_close
5421 Details : It was possible to get multiple mfd references during close and
5422 client eviction, leading to one thread referencing a freed mfd.
5424 Severity: : enhancement
5425 Bugzilla : 3262, 6359
5426 Description: Attempts to reconnect to servers are now more aggressive.
5427 Details : This builds on the enhanced upcall-less recovery that was added
5428 in 1.4.2. When trying to reconnect to servers, clients will
5429 now try each server in the failover group every 10 seconds. By
5430 default, clients would previously try one server every 25 seconds.
5435 Description: After recovery, certain operations trigger a failed
5436 assertion on a client.
5437 Details : Failing over an mds, using lconf -d --failover, while a
5438 client was doing a readdir() call would cause the client to
5439 LBUG after recovery completed and the readdir() was resent.
5441 Severity : enhancement
5443 Description: Default groups are now added by lconf
5444 Details : You can now run lconf --group <servicename> without having to
5445 manually add groups with lmc.
5448 Frequency : occasional
5450 Description: Nodes with an elan id of 0 trigger a failed assertion
5453 Frequency : always when accessing e.g. tty/console device nodes
5455 Description: tty and some other devices nodes cannot be used on lustre
5456 Details : file's private_data field is used by device data and lustre
5457 values in there got lost. New field was added to struct file to
5458 store fs-specific private data.
5461 Frequency : when exporting Lustre via NFS
5463 Description: NFSD failed occasionally when looking up a path component
5464 Details : NFSD is looking up ".." which was broken in ext3 directories
5465 that had grown large enough to become hashed.
5468 Frequency : Clusters with multiple interfaces not on the same subnet
5470 Description: Nodes will repeatedly try to reconnect to an interface which it
5471 cannot reach and report an error to the log.
5472 Details : Extra peer list entries will be created by lconf with some peers
5473 unreachable. lconf now validates the peer before adding it.
5476 Frequency : Only if a default stripe is set on the filesystem root.
5478 Description: Setting a default stripe on the filesystem root prevented the
5479 filesystem from being remounted.
5480 Details : The client was sending extra request flags in the root getattr
5481 request and did not allocate a reply buffer for the dir EA.
5484 Frequency : occasional, higher if lots of files are accessed by one client
5485 Bugzilla : 6159, 6097
5486 Description: Client trips assertion regarding lsm mismatch/magic
5487 Details : While revalidating inodes the VFS looks up inodes with ifind()
5488 and in rare cases can find an inode that is being freed.
5489 The ll_test_inode() code will free the lsm during ifind()
5490 when it finds an existing inode and then the VFS later attaches
5491 this free lsm to a new inode.
5495 Bugzilla : 6422, 7030
5496 Description: MDS deadlock between mkdir and client eviction
5497 Details : Creating a new file via mkdir or mknod (starting a transaction
5498 and getting the ns lock) can deadlock with client eviction
5499 (gets ns lock and trying to finish a synchronous transaction).
5502 Frequency : occasional
5503 Description: While starting a server, the fsfilt_ext3 module could not be
5505 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5506 kernels. Previously, lconf would still use the ext3 name
5507 when trying to load modules. Now, it will correctly use
5508 ext3 on 2.4 and ldiskfs on 2.6.
5510 Severity : enhancement
5511 Description: The default stripe count has been changed to 1
5512 Details : The interpretation of the default stripe count (0, to lfs
5513 or lmc) has been changed to mean striping across a single
5514 OST, rather than all available. For general usage we have
5515 found a stripe count of 1 or 2 works best.
5517 Severity : enhancement
5518 Description: Add support for compiling against Cray portals.
5519 Details : Conditional compiling for some areas that are different
5523 Frequency : occasional
5524 Bugzilla : 6409, 6834
5525 Description: Creating files with an explicit stripe count may lead to
5526 a failed assertion on the MDS
5527 Details : If some OSTs are full or unavailable, creating files may
5528 trigger a failed assertion on the MDS. Now, Lustre will
5529 try to use other servers or return an error to the
5533 Frequency : occasional
5535 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5536 caused lock timeout during readahead (since 1.4.2).
5537 Details : Processes doing readahead might match a lock that hasn't been
5538 granted yet if there are overlapping and conflicting lock
5539 requests. The readahead process waits on ungranted lock
5540 (original lock is CBPENDING), while OST waits for that process
5541 to cancel CBPENDING read lock and eventually evicts client.
5543 Severity : enhancement
5545 Description: Initial enabling of flock support for clients
5546 Details : Implements fcntl advisory locking and file status functions.
5547 This feature is provided as an optional mount flag (default
5548 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5549 locking are implemented yet, and those that are are not guaranteed
5550 to be completely correct in production environments.
5551 mount -t lustre -o [flock|noflock] ...
5554 Frequency : occasional
5556 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5557 assertion in the ext3 JBD (journaling) layer.
5558 Details : The b_committed_data struct is protected by the big kernel lock
5559 in 2.4 kernels, serializing journal_commit_transaction() and
5560 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5561 access to this struct. In 2.6 kernels there is finer grained
5562 locking to improve SMP performance of the JBD layer.
5566 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5571 Description: merge and cleanup kernel patches.
5574 -----------------------------------------------------------------------------
5576 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5578 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5579 versions of Lustre. Please update all servers and clients to
5580 version 1.4.2 or later at the same time. You must also run
5581 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5582 to update the configuration logs.
5584 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5585 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5586 - swab llog records of type '0' so we get proper header size/idx (5861)
5587 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5588 - fix rename of one directory over another leaking an inode (5953)
5589 - avoid SetPageDirty on 2.6 (5981)
5590 - don't re-add just-being-destroyed locks to the waiting list (5653)
5591 - when creating new directories, inherit the parent's custom
5592 striping settings if present parent (3048)
5593 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5594 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5595 - don't hold client locks on temporary worklist from l_lru (5666)
5596 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5597 - automatically evict dead clients (5921)
5598 - Update file size properly in create+truncate+fstat case (6196)
5599 - Do not unhash mountpoint dentries, do not allow removal of
5601 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5602 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5603 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5604 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5605 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5606 - Correctly handle reads of files with no objects (6243)
5607 - lctl recover will also mark a device active if deactivate used (5933)
5609 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5610 - allow --write-conf on an MDS with different nettype than client (5619)
5611 - don't write config llogs to MDS for mounts not from that MDS (5617)
5612 - lconf should create multiple TCP connections from a client (5201)
5613 - init scripts are now turned off by default; run chkconfig --on
5614 lustre and chkconfig --on lustrefs to use them
5615 - upcalls are no longer needed for clients to recover to failover
5617 - add --abort-recovery option to lconf to abort recovery on device
5619 - add support for an arbitrary number of OSTs (3026)
5620 - Quota support protocol changes.
5621 - forward compatibility changes to wire structs (6007)
5622 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5623 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5624 - improved "lustre status" script
5625 - initialize blocksize for non-regular files (6062)
5626 - added --disable-server and --disable-client configure options (5782)
5627 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5628 - Vanilla 2.4.29 support
5629 - increase maximum number of obd devices to 520 (6242)
5630 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5631 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5632 SLES 9 2.6.5 kernels
5633 - limit stripes per file to 160 (the maximum EA size) (6093)
5635 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5638 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5639 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5640 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5641 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5642 - small ext3 extents cleanups and fixes (5733)
5643 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5644 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5645 - avoid needless client->OST connect, fix handle mismatch (5317)
5646 - fix DLM error path that led to out-of-sync client, long delays (5779)
5647 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5648 - fix several locking issues related to i_size (5492,5624,5654,5672)
5649 - don't move pending lock onto export if it is already evicted (5683)
5650 - fix kernel oops when creating .foo in unlinked directory (5548)
5651 - fix deadlock in obdfilter statistics vs. object create (5811)
5652 - use time_{before,after} to avoid timer jiffies wrap (5882)
5653 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5654 - Do not leak request if server was not able to process it (5154)
5655 - If mds_open unable to find parent dir, make that negative lookup(5154)
5656 - don't create new directories with extent-mapping (5909, 5936)
5658 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5659 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5660 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5661 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5662 - support for RHEL4 kernel on i686 (5773)
5663 - provide error messages when incompatible logs are encountered (5898)
5665 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5666 * version 1.4.0.10 (1.4.1 release candidate 1)
5668 - don't keep a lock reference when lock is not granted (4238)
5669 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5670 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5671 - drop import inflight refcount on signal_completed_replay error (5255)
5672 - unlock page after async write error during send (3677)
5673 - handle missing objects in filter_preprw_read properly (5265)
5674 - no transno return for symlink open, don't save no-trasno open (3440)
5675 - don't try to complete elan receive that already failed (4012)
5676 - free RPC server reply state on error (5406)
5677 - clean up thread from ptlrpc_start_thread() on error (5160)
5678 - readahead could read extra page into cache that wasn't ejected (5388)
5679 - prevent races in class_attach/setup/cleanup/detach (5260)
5680 - don't dereference de->d_inode after l_dput of de (5458)
5681 - use "int" for stripe value returned from lock_to_stripe (5544)
5682 - mballoc allocation and error-checking fixes in 2.6 (5504)
5683 - block device patches to fix I/O request sizes in 2.6 (5482)
5684 - look up hostnames for IB nals (5602)
5685 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5686 - don't start multiple acceptors for the same port (5277)
5687 - fix incorrect LASSERT in mds_getattr_name (5635)
5688 - export a proc file for general "ping" checking (5628)
5689 - fix "lfs check" to not block when the MDS is down (5628)
5691 - service request history (4965)
5692 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5693 - create an "evict_client" /proc entry on OSTs, like the MDS has
5694 - fix mount usage message, return errors per mount(8) (5168)
5695 - change grep [] to grep "[]" in tests so they work in more UMLs
5696 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5697 - remove ancient LOV_MAGIC_V0 EA support (5047)
5698 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5699 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5700 - allow readahead allocations to fail when low on memory (5383)
5701 - mmap locking landed again, after considerable improvement (2828)
5702 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5704 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5707 - send OST transaction number in read/write reply to free req (4966)
5708 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5709 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5710 - the watchdog thread now runs as interruptible (5246)
5711 - flock/lockf fixes (but it's still disabled, pending 5135)
5712 - don't use EXT3 constants in llite code (5094)
5713 - memory shortage at startup could cause assertion (5176)
5715 - reorganization of lov code
5716 - single portals codebase
5718 - add extents/mballoc support (5025)
5719 - direct I/O reads in the obdfilter (4048)
5720 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5722 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5725 - send OST transaction number in read/write reply to free req (4966)
5726 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5727 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5728 - the watchdog thread now runs as interruptible (5246)
5729 - handle missing objects in filter_preprw_read properly (5265)
5730 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5731 - drop import inflight refcount on signal_completed_replay error (5255)
5732 - unlock page after async write error during send (3677)
5733 - return original error code on reconstructed replies (3761)
5734 - no transno return for symlink open, don't save no-trasno open (3440)
5736 - add pid to ldlm debugging output (4922)
5737 - bump the watchdog timeouts -- we can't handle 30sec yet
5738 - extra debugging for orphan dentry/inode bug (5259)
5740 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5743 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5744 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5745 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5746 - fix NULL dereference in /proc/sys/portals/routes (4827)
5747 - allow failed mdc_close() operations to be interrupted (4561)
5748 - stop precreate on OST before MDS would time out on it (4778)
5749 - don't send partial-page writes before EOF from client (4410)
5750 - discard client grant for sub-page writes on large-page clients (4520)
5751 - don't free dentries not owned by NFS code, check generation (4806)
5752 - fix lsm leak if mds_create_objects() fails (4801)
5753 - limit debug_daemon file size, always print CERROR messages (4789)
5754 - use transno after validating reply (3892)
5755 - process timed out requests if import state changes (3754)
5756 - update mtime on OST during writes, return in glimpse (4829)
5757 - add mkfsoptions to LDAP (4679)
5758 - use ->max_readahead method instead of zapping global ra (5039)
5759 - don't interrupt __l_wait_event() during strace
5761 - add software watchdogs to catch hung threads quickly (4941)
5762 - make lustrefs init script start after nfs is mounted
5763 - fix CWARN/ERROR duplication (4930)
5764 - return async write errors to application if possible (2248)
5765 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5766 - print NAL number in %x format (4645)
5767 - update barely-supported suse-2.4.21-171 series (4842)
5768 - support for sles 9 %post scripts
5769 - support for building 2.6 kernel-source packages
5770 - support for sles km_* packages
5772 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5775 - ignore -ENOENT errors in osc_destroy (3639)
5776 - notify osc create thread that OSC is being cleaned up (4600)
5777 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5778 - reconstruct ost_handle() like mds_handle() (4657)
5779 - create a new thread to do import eviction to avoid deadlock (3969)
5780 - let lconf resolve symlinked-to devices (4629)
5781 - don't unlink "objects" from directory with default EA (4554)
5782 - hold socknal file ref over connect in case target is down (4394)
5783 - allow more than 32000 subdirectories in a single directory (3244)
5784 - fix blocks count for O_DIRECT writes (3751)
5785 - OST returns ENOSPC from object create when no space left (4539)
5786 - don't send truncate RPC if file size isn't changing (4410)
5787 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5788 - bind to privileged port in socknal and tcpnal (3689)
5790 - rate limit CERROR/CWARN console message to avoid overload (4519)
5791 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5792 - basic mmap support (3918)
5793 - kernel patch series update from b1_4 (4711)
5795 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5798 - avoid crash during MDS cleanup with OST shut down (2775)
5799 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5800 - don't use bad inodes on the MDS (3744)
5801 - dynamic object preallocation to improve recovery speed (4236)
5802 - don't hold spinlock over lock dumping or change debug flags (4401)
5803 - don't zero obd_dev when it is force cleaned (3651)
5804 - print grants to console if they go negative (4431)
5805 - "lctl deactivate" will stop automatic recovery attempts (3406)
5806 - look for existing locks in ldlm_handle_enqueue() (3764)
5807 - don't resolve lock handle twice in recovery avoiding race (4401)
5808 - revalidate should check working dir is a directory (4134)
5810 - don't always mark "slow" obdfilter messages as errors (4418)
5812 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5815 - don't close LustreDB during write_conf until it is done (3860)
5816 - fix typo in lconf for_each_profile (3821)
5817 - allow dumping logs from multiple threads at one time (3820)
5818 - don't allow multiple threads in OSC recovery (3812)
5819 - fix debug_size parameters (3864)
5820 - fix mds_postrecov to initialize import for llog ctxt (3121)
5821 - replace config semaphore with spinlock (3306)
5822 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5823 - don't allow enqueue to complete on a destroyed export (3822)
5824 - down write_lock before checking llog header bitmap (3825)
5825 - recover from lock replay timeout (3764)
5826 - up llog sem before sending rpc (3652)
5827 - reduce ns lock hold times when setting kms (3267)
5828 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5829 - fix NULL deref and obd_dev leak on setup error (3312)
5830 - replace some LBUG about llog ops with error handling (3841)
5831 - don't match INVALID dentries from d_lookup and spin (3784)
5832 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5833 - fix invalid assertion in ptlrpc_set_wait (3880)
5835 - add libwrap support for the TCP acceptor (3996)
5836 - add /proc/sys/portals/routes for non-root route listing (3994)
5837 - allow setting MDS UUID in .xml (2580)
5838 - print the stack of a process that LBUGs (4228)
5840 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5843 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5844 - make sure to unset replay flag from failed open requests (3440)
5845 - if default stripe count is 0, use OST count for inode size (3636)
5846 - update parent mtime/ctime on client for create/unlink (2611)
5847 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5848 - free recovery state on server during a forced cleanup (3571)
5849 - unregister_reply for resent reqs (3063)
5850 - loop back devices mounting and status check on 2.6 (3563)
5851 - fix resource-creation race that can provoke i_size == 0 (3513)
5852 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5853 - more debugging for page-accounting assertion (3746)
5854 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5855 - avoid deadlock after precreation failure (3758)
5856 - fix race and lock order deadlock in orphan handling (3450, 3750)
5857 - add validity checks when grabbing inodes from l_ast_data (3599)
5859 - add /proc/.../recovery_status to obdfilter (3428)
5860 - lightweight CDEBUG infrastructure, debug daemon (3668)
5861 - change default OSC RPC parameters to be better on small clusters
5862 - turn off OST read cache for files smaller than 32MB
5863 - install man pages and include them in rpms (3100)
5864 - add new init script for (un)mounting lustre filesystems (2593)
5865 - run chkconfig in %post for init scripts (3701)
5866 - drop scimac NAL (unmaintained)
5868 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5871 - clean kiobufs before and after use (3485)
5872 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5873 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5874 - change init script to fail more clearly if not run as root (1528)
5875 - allow clients to reconnect during replay (1742)
5876 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5877 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5878 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5879 - fix cond_resched (3554)
5880 - only evict unfinished clients after recovery (3515)
5881 - allow bulk resend, prevent data loss (3570)
5882 - dynamic ptlrpc request buffer allocation (2102)
5883 - don't allow unlinking open directory if it isn't empty (2904)
5884 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5885 - remove extraneous obd dereference causing LASSERT failure (3334)
5886 - don't use get_cycles() when creating temp. files on the mds (3156)
5887 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5888 - handle EEXIST for set-stripe, set proper directory name (3336)
5890 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5891 - fix ksocknal_fmb_callback() error messages (2918)
5893 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5896 - don't copy lvb into (possibly NULL) reply on error (2983)
5897 - don't deref dentry after dput, don't free lvb on error (2922)
5898 - use the kms to determine writeback rpc length (2947)
5899 - increment oti_logcookies when osc is inactive (2948)
5900 - update client's i_blocks count via lvb messages (2543)
5901 - handle intent open/close of special files properly (1557)
5902 - mount MDS with errors=remount-ro, like obdfilter (2009)
5903 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5904 - don't use cancelling-locks' kms values (2947)
5905 - use highest lock extent for kms, not last one (2925)
5906 - don't dereference ERR_PTR() dentry in error handling path (3107)
5907 - fix thread race in portals_debug_dumplog() (3122)
5908 - create lprocfs device entries at setup instead of at attach (1519)
5909 - common AST error handler, don't evict client on completion race (3145)
5910 - zero nameidata in detach_mnt in 2.6 (3118)
5911 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5912 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5913 - initialize RPC timeout timer earlier for 2.6 (3219)
5914 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5915 - print nal/nid for unknown nid (3258)
5916 - additional checks for oscc recovery before doing precreate (3284)
5917 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5918 - don't crash in mdc_close for bad permissions on open (3285)
5919 - zero i_rdev for non-device files (3147)
5920 - clear page->private before handing to FS, better assertion (3119)
5921 - tune the read pipeline (3236)
5922 - fix incorrect decref of invalidated dentry (2350)
5923 - provide read-ahead stats and refine rpc in flight stats (3328)
5924 - don't hold journal transaction open across create RPC (3313)
5925 - update atime on MDS at close time (3265)
5926 - close LDAP connection when recovering to avoid server load (3315)
5927 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5928 - don't leak open file on MDS after open resend (3325)
5929 - serialize filter_precreate and filter_destroy_precreated (3329)
5930 - loop device shouldn't call sync_dev() for nul device (3092)
5931 - clear page cache after eviction (2766)
5932 - resynchronize MDS->OST in background (2824)
5933 - refuse to mount the same filesystem twice on same mountpoint (3394)
5934 - allow llmount to create routes for mounting behind routers (3320)
5935 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5936 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5937 - fix rare problem with rename on htree directories (3417)
5939 - allow default OST striping configuration per directory (1414)
5940 - fix compilation for qswnal for 2.6 kernels (3125)
5941 - increase maximum number of MDS request buffers for large systems
5942 - change liblustreapi to be useful for external progs like lfsck (3098)
5943 - increase local configuration timeout for slow disks (3353)
5944 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5946 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5949 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5950 - don't overwrite extent policy data in reply if lock was blocked (2901)
5951 - drop filter export grants atomically with removal from device (2663)
5952 - del obd_self_export from work_list in class_disconnect_exports (2908)
5953 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5954 - swab reply message in mdc_close, other PPC fixes (2464)
5955 - fix destroying of named logs (2325)
5956 - overwrite old logs when running lconf --write_conf (2264)
5957 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5958 - fix race in target_handle_connect (2898)
5959 - mds_reint_create() should take same inode create lock (2926)
5960 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5961 - don't close files for self_export to avoid uninitialized obd (2936)
5962 - allow MDS with the same name as client node (2939)
5963 - hold dentry reference for closed log files for unlink (2325)
5964 - reserve space for all logs during transactions (2059)
5965 - don't evict page beyond end of stripe extent (2925)
5966 - don't oops on a deleted current working directory (2399)
5967 - handle hard links to targets without a parent properly (2517)
5968 - don't dereference NULL lock when racing during eviction (2867)
5969 - don't grow lock extents when lots of conflicting locks (2919)
5971 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5974 - account for cache space usage on clients to avoid data loss (974)
5975 - lfsck support in lustre kernel code (2349)
5976 - reduce journal credits needed for BRW writes (2370)
5977 - orphan handling to avoid losing space on client/server crashes
5978 - ptlrpcd can be blocked, stopping ALL progress (2477)
5979 - use lock value blocks to assist in proper KMS, faster stat (1021)
5980 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5981 - recovery for initial connections (2355)
5982 - fixes for mds_cleanup_orphans (1934)
5983 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5984 - block all file creations until orphan recovery completes (1901)
5985 - client remove rq_connection from request struct (2423)
5986 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5987 - recovery timer race (2670)
5988 - mdc_close recovey bug (2532)
5989 - ptlrpc cleanup bug (2710)
5990 - mds timeout on local locks (2588)
5991 - namespace lock held during RPCs (2431)
5992 - handle interrupted sync write properly (2503)
5993 - don't try to handle a message that hasn't been replied to (2699)
5994 - client assert failure during cleanup after abort recovery (2701)
5995 - leak mdc device after failed mount (2712)
5996 - ptlrpc_check_set allows timedout requests to complete (2714)
5997 - wait for inflight reqs when ptlrpcd finishes (2710)
5998 - make sure unregistered services are removed from the srv_list
5999 - reset bulk XID's when resending them (caught by 1138 test)
6000 - unregister_bulk after timeout
6001 - fix lconf error (2694)
6002 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6003 - readahead locks pages, leaves pending causing memory pressure (2673)
6004 - increase OST request buffers to 4096 on large machines (2729)
6005 - fix up permission of existing directories in simple_mkdir (2661)
6006 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6007 - don't assign transno to errored transactions (2742)
6008 - don't delete objects on OST if given a bogus objid from MDS (2751)
6009 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6010 - if rq_no_resend, then timeout request after recovery (2432)
6011 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6012 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6013 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6014 - MDS recovery shouldn't delete objects that it has given out (2730)
6015 - if enqueue arrives after completion, don't clobber LVB (2819)
6016 - don't unlock pages twice when trigger_group_io returns error (2814)
6017 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6018 - don't write pages to disk if there was an error (1450)
6019 - don't ping imports that have recovery disabled (2676)
6020 - take buffered bytes into account when balancing socknal conn (2817)
6021 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6022 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6023 - remove little-endian swabbing of llog records (1987)
6024 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6025 - retry reposting request buffers if they fail (1191)
6026 - grow extent at grant time to avoid granting a revoked lock (2809)
6027 - lock revoke doesn't evict page if covered by a second lock (2765)
6028 - disable VM readahead to avoid reading outside lock extents (2805)
6030 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6031 - updated kernel patches for hp-2.4.20 kernel (2681)
6033 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6036 - fix truncated write corruption (2366)
6037 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6039 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6040 - ensure len(uuid) < 37 in lmc (1171)
6041 - fix ia64 OOPS in llog_test (2255)
6042 - zero end of page at obdfilter for partial page writes (2648)
6043 - don't leave stale dentries around after renames (bug 2428)
6044 - fix timeouts when evicting a client with a single lock held (2642)
6045 - set deadline for the initial HELLO message to drain (2634)
6046 - print out dotted-quad IP addresses in the socknal (2302)
6048 - additional debugging for MDS client eviction problem (2443)
6049 - fix mkfsoptions support for osts (2603, 2604)
6051 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6054 - add series for the vanilla 2.6.0 kernel
6055 - add series for the vanilla 2.4.24 kernel
6056 - add series for a cray x86/64 UL kernel drop
6057 - fix xattr patches for the vanilla 2.4.19 series
6059 - generate true UUIDs in lmc (1171)
6060 - have portals stack dumping break in UML (2466)
6061 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6062 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6063 - wake callback waiting threads on client eviction (2460)
6064 - Add --ptldebug and --subsystem to lmc (1719)
6065 - update assertion to allow safe interrupt allocation
6066 - set rq_no_resend for cancel requests (2432)
6067 - recalculate ptlrpcd timeout after resend (2494)
6068 - call vfs_rmdir when removing pending directories (2368)
6069 - fix renaming a file to itself (2429)
6070 - lmc creates a default one-stripe lov (2454)
6071 - expand procfs space to handle large clusters (2326)
6072 - increase UML stack to avoid overflow
6073 - update lconf's list of debug and subsystem masks
6074 - fix lfs find --obd (2510)
6075 - /proc tunable for disabling filter read caching (2591)
6076 - stop rpm packages from altering slapd.conf (2301)
6077 - disable nagle in the socknal under 0conf (2578)
6078 - choose mds inode size based on stripe count (2572)
6079 - fix kernel-source rpm problems (2516)
6081 - add --disable-doc to avoid pdf generation (2421)
6082 - update documentation, tests, type-os, comments
6083 - avoid format warnings on ia64
6084 - remove the TOE NAL
6085 - tiny code cleanups by removing unused fields
6087 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6090 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6091 - remove the most blatant lies from BUILDING file (2371)
6092 - change default debug level to reasonable production setting
6093 - reduce client side cache size to reduce cache flush time
6094 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6095 - make TCP zerocopy and pinger support enabled by default (2476)
6096 - sync writes completed after process exits caused crashes (2319)
6097 - maintain correct mount count on the MDS (2356)
6098 - backout 1557, because 2316 wasn't really fixed
6099 - better file I/O statistics gathering in /proc
6100 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6101 - another kernel patch to fix zero-copy TCP function export
6102 - don't take duplicate lock when processing re-sent getattr (2420)
6103 - lctl uses obd_self_export instead of creating new conn (2353)
6104 - MDS/OST recovery case which requires object creation asserted (2425)
6105 - move lfs from /usr/sbin to /usr/bin in packages
6106 - fix race between mds_client_add and mds_client_free (2417)
6107 - use kmalloc instead of slabs in portals (2430)
6108 - don't create duplicate records when a failover MDS is present (2442)
6109 - remove unnecessary mount age check (2332)
6110 - don't remove directory inodes from locks prematurely (2451)
6111 - don't break if MDS service name is the same as hostname (2103)
6112 - fix races in client write RPC generation when cache full (2482)
6114 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6117 - remove now-unused request->rq_obd (278)
6118 - if an allocation fails, print out how much memory we've used (1933)
6119 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6120 - add the "configurable stack size" patch to most series files (1256)
6121 - ability to write large log records, for 100+ OST configs (2306)
6122 - fix NULL deref when filter_prep fails (2314)
6123 - fix operator precedence error in filter_sync
6124 - dynamic allocation of socknal TX descriptors (2315)
6125 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6126 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6127 - ensure that CWARN messages in Portals always get to the syslog
6128 - __init/__exit are not for prototype decls (ldlm_init/exit)
6129 - x86-64 compile warning fixes
6130 - fix gateway LMC keyword conflict (2318)
6131 - fix MDS lock inversions in getattr/reint paths (1844)
6132 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6133 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6134 - DLM race condition prevented some lock evictions (2328)
6135 - ENOMEM detection and retry on socknal sends (2230)
6136 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6137 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6138 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6139 - better error messages when a client is rejected during recovery (1505)
6140 - avoid cancelling locks which were never granted, after failure (2330)
6141 - fix i_sem/journal inversion in mds_client_add (2333)
6142 - fix truncate/getattr lock cycle deadlock (2334)
6143 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6144 - fix two rare exit paths which could leak an l_lock() ref (2321)
6145 - fencepost error in MDS/OST orphan recovery (2226)
6146 - make log record alignment 8 bytes (1988)
6147 - lstripe now fails when requested offset > ost_count (2237)
6148 - ensure that all kernel series have a complete list.h (1607)
6149 - fix crashes in special-file operations (2316)
6150 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6152 - allow configurable automake binary, for testing new versions
6153 - small update to the lfs documentation
6155 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6157 * fix negative export reference count in fsfilt_sync (2312)
6159 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6160 * release candidate 0.9.1
6162 - orphans are moved into the PENDING directory for possible recovery
6163 - replayed opens now open by fid for orphan/rename safety (1042)
6164 - last close of an orphan inode generates a transno (683)
6165 - chdir() and mount() now pin the directory entry (1020)
6166 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6167 - discard very old requests without processing them (1502)
6168 - remove some common, well-understood CERRORs (1505)
6169 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6170 - clear "grant" flags in OST replies until OST grant code lands (1644)
6171 - fix read performance by not clobbering i_blksize on client (1598)
6172 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6173 - make LRU size a /proc tunable, clears locks when reduced (707)
6174 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6175 - server locks take references on exports now (1558)
6176 - build fixes for 2.4.20-rh trees (1663)
6177 - return an error from lov_create if all OSCs are inactive (1751)
6178 - fix import levels when a reconnect happens without a timeout (1597)
6179 - exit early from mds_open if we get a lookup error (1749)
6180 - partial page read at EOF wouldn't wait for disk before sending (1642)
6181 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6182 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6183 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6184 - write/truncate lock inversion (1639)
6185 - Don't auto-load obdclass, portals modules during cleanup (1495)
6186 - fix timestamps from jumping to "now" (1763)
6187 - extra journal assertions (1648)
6188 - add an extra multiunlink test (1771)
6189 - fix read_record/write_record API (1776)
6190 - fix leak of offset_extent, possible incorrect i_size later (1772)
6191 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6192 - Debugging for the kqswnal_get_idle_tx problems (1820)
6193 - Allow recovery to be attempted multiple times (1536)
6194 - Write out MDS last_rcvd file after it is first created (1600)
6195 - Fix tx_descriptor leak in failed transmit situations (1827)
6196 - ext3 journaling fixes for assertion failure after IO error (1871)
6197 - class_export_put() on freed export after completion AST error (1896)
6198 - Fix revalidate looping in VFS (1322)
6199 - Don't access a freed export during MDS_REINT timeout (1521)
6200 - Add open-unlink recovery support on the MDS (1673,1764)
6201 - Return an error if no MDS data was read from last_rcvd (1946)
6202 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6203 - Don't setup a disk device that doesn't match exported UUID (317)
6204 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6205 - avoid committing NULL handle in force close
6206 - local.sh is now a one-stripe LOV configuration
6207 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6208 - fix ext3 htree duplicate directory entry corruption (1516)
6209 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6210 - update to new LOV EA format (2097)
6211 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6213 2003-06-15 Phil Schwan <phil@clusterfs.com>
6216 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6217 - per-import/export recovery handling (958, 931, 959)
6218 - multiple last-rcvd slots, for serving multiple FSes (949)
6219 - connections are again shared between multiple imp/exports (963, 964)
6220 - "umount -f" would hang if any requests needed to be sent (393, 978)
6221 - avoid pinning large req buffer by copying for queued messages (989)
6222 - add "uuid" to "lctl device" command to help upcalls (991)
6223 - "open" RPCs with transnos would confuse recovery counters (1037)
6224 - do proper endian conversion of all wire messages (288, 340, 891)
6225 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6226 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6227 - fix ptlrpc_connection leak in target_handle_connect (1174)
6228 - fix import refcounting bug in OST and MDS cleanup (1134)
6229 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6230 - fix very unlikely obd_types race condition (501)
6231 - remove osc_open hack for echo_client (1187)
6232 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6233 - a failure in read_inode2 leads to deadlock (1139)
6234 - cancel ack-locks as soon as transaction is committed (1072)
6235 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6236 - make sure to commitrw after any preprw to avoid deadlock (1162)
6237 - failing to execute a file in a lustre FS would lock inode (1203)
6238 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6239 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6240 - obd_uuid2tgt would walk off the end of the list (1255)
6241 - on IA64 the osc would give portals incorrect bulk size (1258)
6242 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6243 - fix lock inversion caused by new llite matching code (1282)
6244 - limit the number of dirty pages on a client to 10MB (1286)
6245 - timed out locks were not being corrected cancelled (1289)
6246 - fix O_DIRECT above 4GB on IA-32 (1292)
6247 * major user-visible changes
6248 - fail out/fail over policy now controlled by the upcall (993)
6250 - add OBD_PING to check server availability and failure (954)
6251 - lustre messages are now sent in sending host order (288, 340, 891)
6252 - add eadatalen to MDS getattr reply (340)
6253 - OST read replies may contain second buffer, with per-page status (593)
6255 2003-03-11 Phil Schwan <phil@clusterfs.com>
6258 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6259 - failed open()s could cause deadlock; fixed (867, 869)
6260 - stop cancelling OST locks when files are closed (481)
6261 - overlapping XID spaces caused network corruption (851, 853)
6262 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6263 - setattr_raw added, to do single-RPC, server-side setattrs
6264 - lmc/lconf syntax change for OST UUIDs
6265 - fix crashy race condition between ptlrpc_free_req and osc_close
6266 - don't use request in mdc_enqueue if we hit a timeout (889)
6267 - don't set the inode i_size for regular files from the MDS (896)
6268 - handle out of order completion AST (842)
6269 - don't LBUG if a lock request times out after receiving AST (913)
6270 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6271 - if a bad lock AST arrives, send an error instead of dropping entirely
6272 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6273 - fix leak in bulk IO when only partially completed (899, 900, 926)
6274 - fix O_DIRECT for ia64 (55)
6275 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6276 - C-z after timeout could hang a process forever; fixed (977)
6278 - client-side I/O cache (678, 924, 929, 941, 970)
6280 - READPAGE and SETATTRs which don't take server-side locks get
6283 2003-02-11 Phil Schwan <phil@clusterfs.com>
6286 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6287 - Fix fsfilt_extN_readpage() to read a full page of directory
6288 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6289 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6290 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6291 - Most intent operations were removed, in favour of a new RPC mode
6292 that does a single RPC to the server and bypasses most of the VFS
6293 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6294 - Aggressively cancel local locks on DLM servers
6295 - mds_reint_unlink sends EA to the client if it's the last nlink.
6296 client uses that EA to unlink OST objects.
6297 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6298 - recursive symlinks were fixed (439)
6299 - fixed NULL deref in DEBUG_REQ
6300 - filter_update_lastobjid no longer calls sync, which annoyed extN
6301 - fixed multi-client small-writes to a single file problem (445)
6302 - fixed mtime updates during file writes (607)
6303 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6304 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6305 - fixed Linux OST size reporting problem (444, 656)
6306 - OST now updates object mtime with writes or setattr (607, 619)
6307 - client verifies file size before zeroing page past EOF (445)
6308 - OST now writes last allocated objid to disk with allocation (108)
6309 - LOV on echo now works (409)
6311 - mds_reint_unlink sends a new buffer, with the EA included. this
6312 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6313 set if a regular file was being unlinked, and it was the last link
6314 - use PtlGet from the target for bulk writes (315)
6315 - OST now updates object mtime with writes or setattr (607, 619)
6316 - LDLM now has a grant-time callback to revalidate locked items, if
6318 - Many MDS operations were reorganized to combat race conditions
6320 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6321 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6323 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6326 - Fully reactivate OST imports after reconnection (512, others)
6327 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6328 - More graceful error handling for truncating on dead OST (515)
6329 - Don't error out unless we're actually accessing dead stripes (474)
6330 - Fix garbage sizes when stripes are missing (410)
6331 - LRU counters were broken, causing constant lock purge (433, 432)
6332 - garbage on read from stripes with failed OSTs (441)
6333 - mark OSCs as active before reconnecting during recovery (438)
6334 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6335 - lfind did not preserve OST order in output (443)
6336 - symlinks cause hung clients, incorrect data (439)
6337 - stop dereferencing request after dropping refcount (457)
6338 - don't LASSERT(spin_is_locked) on non-SMP (455)
6339 - fixes for many rename() bugs
6340 - fstat didn't correctly synchronize attributes (399)
6341 - server must handle lock cancellation during blocking AST prep (487)
6342 - bulk descriptors were free()d too soon (511)
6343 - fix paths in lconf, which would load incorrect modules (451, 507)
6344 - fix confusing lconf 'host not found' error message (386)
6345 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6346 - fix race condition in mdc_blocking_ast() for inode access (526)
6347 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6348 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6349 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6350 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6351 - fix lstripe in several ways (kernel side) (527)
6352 - fix request leak in ldlm_cli_enqueue (262)
6353 - incorrect OSC was marked inactive after OST failure
6354 - call mds_fs_cleanup before unmounting filesystem (524)
6355 - fix races between taking ns_lock and ldlm_lock_change_resource
6356 - fix races updating LOV export open file list
6357 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6358 - fix recovery NULL deref in ldlm_cli_cancel_unused
6359 - fix some DLM races by using new hash table for lock handles (419)
6360 - permit the client to specify desired inodes, at replay
6361 - duplicate requests when we queue them for replay reintegration
6362 - fix last_rcvd offset calculation
6363 - sync after each recovered transaction, so we always make progress
6364 - never, not always, ERESTART requests without transnos
6365 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6367 - skip replay if the MDS says that the client is already connected
6368 - don't check for a recovery-enabled export to match lctl's UUID
6369 - don't INC_USE_COUNT for phantom exports
6370 - don't crash when cleaning up phantom exports (567)
6371 - don't double-finish or set replay data for errored mdc_open requests
6372 - abort requests when they time out, so we don't get old replies
6373 - send/receive replies for AST messages again
6374 - if the client says that it doesn't have the lock, cancel it on the
6376 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6377 mark it as destroyed, it will all work out when decref is called
6378 - fix module use counts (22, 581)
6380 - ASTs now expect a reply (server cancels lock on error reply)
6382 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6385 - fix many simultaneous client startup (392)
6386 - fix dentry->d_it clobbering
6387 - credentials weren't being shipped for readdir/getattr operations
6388 - remove invalid assertions triggered during some concurrent MD
6390 - proper Lustre versions added (336, 389)
6391 - fix memory leak for create error case (398)
6392 - fix LOV locking bug that would get cli/srv out of sync
6393 - fix echo client over LOV (409)
6394 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6395 - fix double-O_EXCL intent crash (424)
6396 - avoid sending multiple lock CANCELs (352)
6398 - MDS can do multi-client recovery (modulo bugs in new code)
6400 - many updates, edits, cleanups
6402 2002-11-18 Phil Schwan <phil@clusterfs.com>
6405 - fix null d_it dereference (346)
6406 - fix full OST/dbench hang (333)
6407 - fix permission problem with file removal (286)
6408 - fix removal of OSCs from LOV when they fail
6409 - fix NULL deref during bulk timeout (214)
6410 - fix problems related to multiple filesystems on one MDS (241)
6411 - fixed serious subtle metadata locking bugs
6412 - free locks on clients when inodes are removed due to memory
6414 - fix inode pointer in lock data (285)
6415 - partial support for multiple MDS on a single host (241)
6416 - data locks weren't cancelled at clear_inode time (290, 311)
6417 - intent locks could lead to unbounded lock growth (205)
6418 - added a maximum lock count, an LRU list, and a flusher
6419 - fix multiple rename (365)
6420 - properly abstracted the echo client
6421 - OSC locked 1 byte too many; fixed
6422 - rewrote brw callback code:
6423 - fixed recovery bugs related to LOVs (306)
6424 - fixed too-many-pages-in-one-write crash (191)
6425 - fixed (again) crash in sync_io_timeout (214)
6426 - probably fixed callback-related race (385)
6428 - Add capability to MDS protocol
6429 - LDLM cancellations and callbacks on different portals
6431 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6434 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6435 - defer open object destruction to close time (601981, 138)
6436 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6437 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6438 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6439 - don't call obd_setattr in ll_file_release for destroyed objects
6440 * protocol change to lustre_msg: move |version| and add |flags|
6441 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6442 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6443 * added replay of create, unlink, link and rename operations during
6444 MDS failover; recovery should be much more robust now
6445 * remove failed OSCs from LOVs (only lov_create uses this so far)
6446 * the lustre-HOWTO was brought (more) up to date (582544)
6448 2002-10-23 Phil Schwan <phil@clusterfs.com>
6451 - in-use dentries weren't being reused properly (617851)
6452 - prevent multiple LDLM setup (599178)
6453 - fix LOV size calculations for truncate (617853)
6454 - fix client handling of MDS intent errors (POSIX)
6455 - fix permission bug in lovstripe.c test (624321)
6456 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6457 - truncate past end of file could corrupt data
6458 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6459 - a race in recovery could return ETIMEDOUT to apps (623947)
6460 - building outside the source directory was fixed
6461 * the lustre-HOWTO was brought (more) up to date (582544)
6462 * major progress was made on recovery functionality
6464 2002-10-10 Phil Schwan <phil@clusterfs.com>
6467 - recovery deadlock fix
6468 - rm -rf causes LBUG fix (617817)
6469 - file open by multiple tasks fix (618962)
6470 - directory permissions bugs (602707 and 620007)
6471 - journal_stop fixed with locking (611313)
6472 - O_APPEND failures resolved (618273, perhaps 614459)
6473 - lconf PATH fix (619770)
6474 - IA64 build fix (621450)
6475 - RPC buffer sizes scale with amount of memory
6477 2002-10-01 Phil Schwan <phil@clusterfs.com>
6480 - locks would be cancelled without throwing away data pages,
6481 resulting in inconsistent data (605627)
6482 - inode attributes were not always being refreshed (605627, 612449)
6483 - lconf now continues to cleanup after lctl reports an error
6484 - MDS now enforces user permissions (602707)
6485 - lprocfs cleanup fixed, but not yet enabled (614157)
6486 - fixed infinite server hang, should a client not respond to an AST
6487 - avoid going into recovery if user calls readlink() with a buffer
6488 that's too small (613941)
6489 - AST RPCs no longer require replies (614867) -- this may be changed
6490 - don't crash server if client sends an IOV that's too big (611336)
6491 - fixed lock conversion deadlock (611892)
6492 - fixed the following of symlinks (614622)
6493 * recovery: the server can remove locks from a client that dies, other
6494 clients can make progress
6495 * more extN patch fixes
6496 * compile-time configurable ptlrpc buffer allocations
6498 - collaborative read cache document
6499 - Lustre Lite Performance CDR document-in-progress
6501 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6504 - fix typo in patch-2.4.18
6506 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6509 - clear ptlrpc request each time in handle_incoming_request()
6510 - unlink of files now destroys the object on the OST
6512 2002-09-19 Peter Braam <braam@clusterfs.com>
6514 * add hard link support
6515 * change obdfile creation method
6516 * kernel patch changed
6518 2002-09-19 Peter Braam <braam@clusterfs.com>
6521 - stack overflow bug in extN fixed
6523 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6525 * documentation updates
6526 - add man pages for config tools
6527 - update tests/README to describe testing with new config tools
6528 - finish metadata API descriptions
6529 * bug fixes and cleanups
6530 - statfs workaround for 16TB limit
6531 - LOV stripe allocation improved, can stripe on subset of OSTs
6532 - LOV file size/IO offset was wrong for files > 4GB in size
6533 - object EA data was being dropped, caused files to be unreadable
6534 - memory overflow with non-LOV OST caused memory corruption
6535 - fixed regression tests to work with new config tools, obdfilter
6536 - fixed bug when directory size became larger than 1 block
6537 - fixed bug (for single client case) when PWD was deleted
6538 - invalidate local directory pages when doing intent-based ops
6539 - avoid LDLM oops when lock callback contained bad data
6541 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6543 * documentation updates
6544 * bug fixes and cleanups
6545 - configuration tools
6548 - 64-bit compile warnings
6549 - 64-bit internal statfs data
6551 * test_brw on persistent OST devices
6553 * lprocfs (disabled)
6555 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6557 * documentation updates
6558 * bug fixes and cleanups
6559 * configuration tools
6561 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6564 * many small fixes to 0_5_4
6565 * io/network handling
6566 * thinkos in MDS operations
6568 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6571 * crucial basic fixes to 0.5.3
6572 * IOR, Iozone work over Elan
6575 2002-08-07 Phil Schwan <phil@clusterfs.com>
6576 * version 0_5_3, our first alpha
6577 * we use the new Portals iovs
6578 * documentation updates
6579 * bug fixes and cleanups
6580 * small changes in the DLM wire protocol
6582 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6583 * version 0_5_1 with some initial stability,
6584 * locking on MD and file I/O.
6585 * documentation updates
6586 * several bug fixes since 0.5.0
6587 * small changes in wire protocol
6589 2002-07-18 Phil Schwan <phil@clusterfs.com>
6591 * delivered as Lustre Light Alpha
6592 * fixed a crash after handling invalid MDS requests
6593 * fixed directory pages for architectures with non-4k pages sizes
6595 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6596 * release version v0_4_4
6597 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6599 * Fixes a number of interruption problems with OST operations.
6600 * Update documentation for portals header changes
6601 * Move all wire protocol structs/defines to lustre_idl.h
6602 * Fixes symlink length bug.
6603 * Add tcpdump to repository.
6605 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6606 * release version v0_4_3
6607 * Fixes statfs for inodes on extN.
6608 * Fixes bug in runtests which would delete /etc/hosts.
6609 * Use 64-bit object IDs wherever possible (not into VFS though)
6610 Remove ost_get_info, which is unused by lustre, and out of date.
6612 2002-07-03 Peter Braam <braam@clusterfs.com>
6613 * release version v0_4_2 Fixes a lookup error (type not passed)
6614 * move forward to head of Portals
6615 * move forward to latest Lustre kernel
6617 2002-06-25 Peter Braam <braam@clusterfs.com>
6618 * release version v0_4_1. Hopefully stable on single node use.