1 10-02-2017 Intel Corporation
3 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.13.2.el6 (RHEL6.9)
11 3.10.0-693.2.2.el7 (RHEL7.4)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.74-92.38 (SLES12 SP2)
16 4.4.82-6.9 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.13.2.el6 (RHEL6.9)
24 3.10.0-693.2.2.el7 (RHEL7.4)
25 3.0.101-0.47.71 (SLES11 SP3)
26 3.0.101-107 (SLES11 SP4)
27 3.12.74-60.64.40 (SLES12 SP1)
28 4.4.74-92.38 (SLES12 SP2)
29 4.4.82-6.9 (SLES12 SP3)
31 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
32 * Recommended ZFS / SPL version: 0.6.5.9
33 * Tested with ZFS / SPL version: 0.6.5.9
34 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
35 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
36 stack overflow. For more information, please refer to bugzilla 17630.
37 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
38 the re-exporting nodes to be mounted with "32bitapi" mount option
40 --------------------------------------------------------------------------------
42 12-07-2016 Intel Corporation
44 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on patched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.26.1.el6 (RHEL6.7)
50 2.6.32-642.11.1.el6 (RHEL6.8)
51 3.10.0-514.el7 (RHEL7.3)
52 3.0.101-0.47.71 (SLES11 SP3)
53 3.0.101-80 (SLES11 SP4)
54 3.12.62-60.64.8 (SLES12 SP1)
55 vanilla linux 4.5.7 (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 2.6.32-642.11.1.el6 (RHEL6.8)
61 3.10.0-514.el7 (RHEL7.3)
62 3.0.101-0.47.71 (SLES11 SP3)
63 3.0.101-80 (SLES11 SP4)
64 3.12.62-60.64.8 (SLES12 SP1)
66 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
67 * Recommended ZFS / SPL version: 0.6.5.8
68 * Tested with ZFS / SPL version: 0.6.5.8
69 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
70 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
71 stack overflow. For more information, please refer to bugzilla 17630.
72 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
73 the re-exporting nodes to be mounted with "32bitapi" mount option
75 --------------------------------------------------------------------------------
77 02-29-2016 Intel Corporation
79 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
80 for currently supported client and server kernel versions.
81 * Server known to build on patched kernels:
82 2.6.32-431.29.2.el6 (RHEL6.5)
83 2.6.32-504.30.3.el6 (RHEL6.6)
84 2.6.32-573.12.1.el6 (RHEL6.7)
85 3.10.0-327.3.1.el7 (RHEL7.2)
86 3.0.101-0.47.71 (SLES11 SP3)
87 3.0.101-68 (SLES11 SP4)
88 vanilla linux 4.2.1 (ZFS only)
89 * Client known to build on unpatched kernels:
90 2.6.32-431.29.2.el6 (RHEL6.5)
91 2.6.32-504.30.3.el6 (RHEL6.6)
92 2.6.32-573.12.1.el6 (RHEL6.7)
93 3.10.0-327.3.1.el7 (RHEL7.2)
94 3.0.101-0.47.71 (SLES11 SP3)
95 3.0.101-68 (SLES11 SP4)
98 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
99 * Recommended ZFS / SPL version: 0.6.4.2
100 * Tested with ZFS / SPL version: 0.6.4.2
101 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
102 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
103 stack overflow. For more information, please refer to bugzilla 17630.
104 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
105 the re-exporting nodes to be mounted with "32bitapi" mount option
107 --------------------------------------------------------------------------------
109 03-10-2015 Intel Corporation
111 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
112 for currently supported client and server kernel versions.
113 * Server known to build on patched kernels:
114 2.6.32-431.29.2.el6 (RHEL6.5)
115 2.6.32-504.8.1.el6 (RHEL6.6)
116 3.0.101-0.46 (SLES11 SP3)
117 * Client known to build on unpatched kernels:
118 2.6.32-431.29.2.el6 (RHEL6.5)
119 2.6.32-504.8.1.el6 (RHEL6.6)
120 3.10.0-123.20.1.el7 (RHEL7)
121 3.0.101-0.46 (SLES11 SP3)
122 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
129 Severity : enhancement
131 Description: control OST-index in IDIF via ROCOMPAT flag.
132 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
133 last_rcvd file. For new formatted OST device, it will be auto set;
134 for the case of upgrading from old OST device, you can enable it
135 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
136 enabled, for new created OST-object, its IDIF-in-LMA will contain
137 the OST-index; for the existing OST-object, the OSD will convert
138 old format IDIF as new format IDIF with OST-index stored in the
139 LMA EA when accessing such OST-object or via OI scrub. Once such
140 flag is enabled, it cannot be reverted back, so the system cannot
141 be downgraded to the orignal incompatible version.
143 --------------------------------------------------------------------------------
145 07-30-2014 Intel Corporation
147 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
148 for currently supported client and server kernel versions.
149 * Server known to build on patched kernels:
150 2.6.32-431.20.3.el6 (RHEL6.5)
151 3.0.101-0.31 (SLES11 SP3)
152 * Client known to build on unpatched kernels:
153 2.6.32-431.20.3.el6 (RHEL6.5)
154 3.0.101-0.31 (SLES11 SP3)
157 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
158 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
159 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
160 stack overflow. For more information, please refer to bugzilla 17630.
161 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
162 the re-exporting nodes to be mounted with "32bitapi" mount option
164 --------------------------------------------------------------------------------
166 10-31-2013 Intel Corporation
168 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
169 * Server builds for kernels:
170 2.6.32-358.18.1.el6 (RHEL6.4)
171 * Client builds for unpatched kernels:
172 2.6.32-358.18.1.el6 (RHEL6.4)
173 3.0.92-0.8 (SLES11 SP3)
175 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
176 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
177 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
178 stack overflow. For more information, please refer to bugzilla 17630.
179 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
180 the re-exporting nodes to be mounted with "32bitapi" mount option
182 --------------------------------------------------------------------------------
184 05-30-2013 Intel Corporation
186 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
187 * Server builds for kernels:
188 2.6.32-358.6.2.el6 (RHEL6.4)
189 * Client builds for unpatched kernels:
190 2.6.32-358.6.2.el6 (RHEL6.4)
191 3.0.74-0.6.6 (SLES11 SP2)
193 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
194 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
195 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
196 stack overflow. For more information, please refer to bugzilla 17630.
197 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
198 the re-exporting nodes to be mounted with "32bitapi" mount option
200 --------------------------------------------------------------------------------
202 9-30-2012 Whamcloud, Inc.
204 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
205 * Server support for kernels:
206 2.6.32-279.5.1.el6 (RHEL6)
207 * Client support for unpatched kernels:
208 2.6.32-279.5.1.el6 (RHEL6)
210 2.6.18-238.19.1.el5 (RHEL5)
211 2.6.16.60-0.69.1 (SLES10)
212 * Recommended e2fsprogs version: 1.41.90.wc4
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 Severity : enhancement
221 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
223 Severity : enhancement
225 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
227 Severity : enhancement
229 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
231 Severity : enhancement
233 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
235 Severity : enhancement
237 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
239 --------------------------------------------------------------------------------
241 04-30-2012 Whamcloud, Inc.
243 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
244 * Server support for kernels:
245 2.6.32-220.4.2.el6 (RHEL6)
246 * Client support for unpatched kernels:
247 2.6.32-220.4.2.el6 (RHEL6)
249 2.6.18-238.19.1.el5 (RHEL5)
250 2.6.16.60-0.69.1 (SLES10)
251 * Recommended e2fsprogs version: 1.41.90.wc4
252 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
253 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
254 stack overflow. For more information, please refer to bugzilla 17630.
255 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
256 the re-exporting nodes to be mounted with "32bitapi" mount option
258 Severity : enhancement
260 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
262 Severity : enhancement
264 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
266 Severity : enhancement
268 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
270 Severity : enhancement
272 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
274 --------------------------------------------------------------------------------
276 09-30-2011 Whamcloud, Inc.
278 * Server support for kernels:
279 2.6.18-238.19.1.el5 (RHEL5)
280 2.6.32-131.6.1.el6 (RHEL6)
281 * Client support for unpatched kernels:
282 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
283 2.6.18-238.19.1.el5 (RHEL5)
284 2.6.32-131.6.1.el6 (RHEL6)
285 2.6.29.4-167.fc11 (FC11)
286 2.6.16.60-0.69.1 (SLES10)
288 * Recommended e2fsprogs version: 1.41.90.wc3
289 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
290 Lustre filesystem with 4K stack may cause a stack overflow. For more
291 information, please refer to bugzilla 17630.
292 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
293 mounted with "32bitapi" mount option
295 Severity : enhancement
296 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
298 Severity : enhancement
299 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
301 Severity : enhancement
302 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
305 Frequency : only with SLES10
307 Description: Use OFED "KMP" provided by Novell
308 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
309 Lustre is now built against this package. That means you need to
310 install the ofed-kmp package from Novell for the patchless client.
311 Note that the ofed-kmp that Novell ships may not exactly match the
312 kernel version but should still be compatible.
314 Severity : enhancement
316 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
317 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
319 --------------------------------------------------------------------------------
321 2010-07-15 Oracle, Inc.
323 * Server support for kernels:
324 2.6.18-164.11.1.el5 (RHEL 5)
325 2.6.18-164.11.1.0.1.el5 (OEL 5)
326 * Client support for unpatched kernels:
327 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
328 2.6.18-164.11.1.el5 (RHEL 5),
329 2.6.18-164.11.1.0.1.el5 (OEL 5)
330 2.6.16.60-0.42.8 (SLES 10),
332 2.6.29.4-167.fc11 (FC11)
333 * Recommended e2fsprogs version: 1.41.10-sun2
335 Severity : enhancement
337 Description: Distribute lfs_migrate script with lustre
338 Details : The lfs_migrate mentioned in the manual is now included in the
339 lustre tree in lustre/scripts and will distributed in lustre RPMs.
341 Severity : enhancement
343 Description: Failover nid list added to import proc file.
345 Severity : enhancement
347 Description: "som_preview" mount option required to use SOM
348 Details : The clients will need to use som_preview mount option
352 Description: Lustre does not do 1MB IOs to HW RAID
353 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
355 Severity : enhancement
357 Description: lfs quota output is not very convenient for awk/sed-parsing
358 Details : Some positions in lfs quota output table could be empty or
359 non-empty which made it hard to parse it with scripts, now
360 a dash is put instead of space where there is not supposed
365 Description: Update to OFED 1.5.1
366 Details : OFED 1.5.1 will be used in environments that don't use
367 vendor supplied OFED.
371 Description: Remove lnet. prefix from lctl params display
372 Details : Also renamed lnet.memused to lnet_memused to
377 Description: Use conf_param -d to remove permanant settings.
381 Description: Add list_param -R
382 Details : lctl list_param now supports recursive mode.
385 Bugzilla : 21829, 22312
386 Description: Fix llobdstat.
387 Details : 1. Now updating the write_bytes counter
388 2. Port script fixes from 1.8.
392 Description: decrease the usage of memory on clients.
393 Details : 1. On clients, recycle dentries and inodes unused.
394 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
399 Description: llobdstat fix and enhancement
400 Details : add a counter to set a limit to how many samples will be returned
401 fix a wildcard in the path to limit to obdfilter stats only
405 Description: lfs check only allowed for root
406 Details : Code cleanups and added a regresion test of non-root use of
411 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
412 Details : Fix hr->hr_index overload.
416 Description: Fix FIEMAP functionality.
420 Description: FC11 patchless client support added
424 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
425 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
429 Description: New acc-sm script and various test-framework enhancments.
433 Description: Previously landed patch was reverted.
437 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
441 Description: Add ldiskfs maxdirsize mount option.
445 Description: Fix mkfs.lustre support for 16TB LUNs.
449 Description: Print mmp_check_interval
450 Details : print mmp_check_interval and make it possible to abort
451 mount operation in case it takes too long.
455 Description: Add cascading_rw to lustre mpi tests.
459 Description: Store SOM data in EA.
460 Details : Also several SOM related cleanups.
464 Description: Add refcount to md_open_data
468 Description: Use refcount on lov_request_set to prevent memory corruption.
472 Description: Statahead: use parent dir's i_mutex for synchronization.
476 Description: loadgen improvements.
480 Description: Enable peer health detection in ksocklnd and ko2iblnd.
484 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
488 Description: Fix race on obd_devcie:md_stats
492 Description: Fix SLES 11 client builds.
493 Details : Resolved uninitalized pointer gcc warnings.
497 Description: Allow non-root access for "lfs check".
498 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
502 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
504 Severity : enhancement
506 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
507 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
511 Description: Sanity test_27p fails.
512 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
513 Sanity tests 27 were cleaned up and improved.
517 Description: Unify req format on client/servers
518 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
519 buffers and swabbers approach..
524 Description: drop join file stuff.
526 Severity : enhancement
528 Description: Add 2.6.27 and SLES11 (patchless client) support
532 Description: Fix in ptlrpc_expire_one_request() to print the signed time
535 Severity : enhancement
537 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
540 Severity : enhancement
542 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
544 Severity : enhancement
546 Description: correctly handle big reply message.
547 Details : send LNet event if reply is bigger than buffer and adjust this buffer
552 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
554 Severity : enhancement
556 Description: Add support for OEL5.
560 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
562 Severity : enhancement
564 Description: "weak-modules" support
565 Details : Implement "weak-modules" support which enables kernel modules
566 to be used with any kernel that implements the same kABI. In
567 order to achieve this modules are now installed in
568 /lib/modules/$(uname -r)/updates/kernel on all distributions.
572 Description: client could not reconnect to OST because of an active request.
573 Details : The client now retries to reconnect to the same server, if a
574 connect request failed with EBUSY or -EAGAIN. Test case 35b
575 conf-sanity.sh has been created.
579 Description: Don't allow make backward step on assignin osc next id.
580 Details : race between allocation next id and ll_sync thread can be cause
581 of set wrong osc next id and can be kill valid ost objects.
585 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
589 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
593 Description: Error handling in osc_statfs_interpret() has been improved.
594 Details : Check in osc_statfs_interpret() for EBADR.
598 Description: Do not send statfs() requests to OSTs disabled by administrator.
599 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
601 Severity : enhancement
603 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
605 Severity : enhancement
607 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
609 Severity : enhancement
611 Description: Enhancement for lfs(1) command to use numeric uid/gid.
615 Description: Improved cleanup procedure for conf-sanity test_46a.
619 Description: Do not start bulk transfer, if client tried to reconnect.
620 Details : Do not start bulk transfer, if client tried to reconnect,
621 but failed due to active rpcs. Check for exp_abort_active_req
622 has been implemented.
626 Description: Prevent inconsistences between linux and lustre mount structures.
627 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
628 Make the sleep interruptible.
630 Severity : enhancement
632 Description: provide server to client comms path
633 Details : server to client communications path via new LDLM_SET_INFO rpc
637 Description: truncate starts GFP_FS allocation under transaction causing deadlock
638 Details : ldiskfs_truncate calls grab_cache_page which may start page
639 allocation under an open transaction. This may lead to
640 calling prune_icache with consequent lustre reentrance.
644 Description: send correctly lsm on open replay
645 Details : MDS is trust to LSM size on replay open, but client can set wrong size
650 Description: enable client interface failover
651 Details : When a child reconnects from another NID, properly update export
652 nid hash position and ldlm reverse import.
654 Severity : enhancement
656 Description: Communicate OST degraded/readonly state via statfs to MDS
657 Details : Flags in the statfs returned from OSTs indicate whether the
658 OST is in a degraded RAID state, or if the filesystem has
659 turned read-only after a filesystem error is detected.
664 Description: don't panic if EPROTO was hit when reading symlink
665 Details : correctly handling request reference in error cases.
667 Severity : enhancement
669 Description: Add LustreNetLink, a kernel-userspace communcation path.
671 Severity : enhancement
673 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
678 Description: access to llog context before init.
679 Details : move handling CATALOGS file at osc layer and forbid access to llog
683 Frequency : rare, only when client is writing to OST on local node
685 Description: Avoid deadlock for local client writes
686 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
687 memory freeing context. This allows OST threads to set the
688 PF_MEMALLOC flag on task structures in order to allocate memory
689 from reserved pools and complete IO.
690 Use GFP_HIGHUSER for OST allocations for non-local client writes,
691 so that the OST threads generate memory pressure and allow
692 inactive pages to be reclaimed.
694 Severity : enhancement
696 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
701 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
702 Details : move ldlm namespace creation in setup phase to avoid grab
703 _lprocfs_lock with cli_sem held.
707 Description: Temporarily disable grant shrink.
708 Details : Disable the feature for debugging.
711 Frequency : only when formatting test filesystems
713 Description: Unable to run several mkfs.lustre on loop devices at the same time
714 Details : mkfs.lustre returns error 256 on the concurrent loop devices
715 formatting. The solution is to proper handle the error.
717 Severity : enhancement
719 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
721 Severity : enhancement
723 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
728 Description: resolve race between obd_disconnect and class_disconnect_exports
729 Details : if obd_disconnect will be called to already disconnected export he
730 forget release one reference and osc module can't unloaded.
732 Severity : enhancement
734 Description: Allow tuning service thread via /proc
735 Details : For each service a new
736 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
737 created that can be used to set min/max thread counts, and get the
738 current number of running threads.
743 Descriptoin: don't return error if have partially created objects for file.
744 Details : lov_update_create_set uses set->set_success as index for created
745 objects, so if some requests will be failed, they will have hole
746 at end of array and qos_shrink_lsm for allocate correct lsm.
748 Severity : enhancement
750 Description: Update OFED support to 1.4.1
755 Description: Reduce small size read RPC
756 Details : Set read-ahead limit for every file and only do read-ahead when
757 available read-ahead pages are bigger than 1M to avoid small size
760 Severity : enhancement
762 Description: move AT tunable parameters for more consistent usage
763 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
765 Severity : enhancement
767 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
768 Details : allow skip disconnected ost for send statfs request and hide error
774 Description: fix racy locking of mballoc block bitmaps causing BUG
775 Details : The locking of the mballoc buddy bitmap and the in-memory
776 block bitmap was using two different spin locks in some
777 cases. This made it possible to incorrectly access the
778 mballoc bitmap while another process was modifying it,
779 causing a sanity assertion to fail. While no on-disk corruption
780 was reported, there was some risk of this happening.
783 Frequency : rare, on llog test 6
785 Descriptoin: don't allow connect to already connected import
786 Details : allowing connect to already connected import is hide connecting problem.
791 Descriptoin: allow kill process which wait statahead result
792 Details : for some reasons 'ls' can stick in waiting result from statahead,
793 in this case need way for kill this process.
795 Severity : enhancement
797 Description: Add state history info file, enhance import info file
798 Details : Track import connection state changes in a new osc/mdc proc file;
799 add overview-type data to the osc/mdc import proc file.
801 Severity : enhancement
803 Description: MDS create should not wait for statfs RPC while holding DLM lock.
805 Severity : enhancement
807 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
812 Descriptoin: OST grant too much space to client even there are not enough space.
813 Details : Client will shrink its grant cache to OST if there are no write
814 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
815 this grant cache if there are already not enough avaible space
816 (left_space < total_clients * 32M).
819 Frequency : start MDS on uncleanly shutdowned MDS device
821 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
822 Details : stay in waiting mds<>ost recovery finished produce random bugs
823 due race between two ll_sync thread for one lov target. send
824 ACTIVATE event only if connect realy finished and import have
828 Frequency : rare, connect and disconnect target at same time
830 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
831 Details : don't call obd_disconnect under lov_lock. this long time
832 operation and can block ptlrpcd which answer to connect request.
837 Descriptoin: don't lose wakeup for imp_recovery_waitq
838 Details : recover_import_no_retry or invalidate_import and import_close can
839 both sleep on imp_recovery_waitq, but we was send only one wakeup
843 Frequency : always with long access acl
845 Descriptoin: mds can't pack reply with long acl.
846 Details : mds don't control size of acl but they limited by reint/getattr
849 Severity : enhancement
851 Description: Update to SLES10 kernel-2.6.16.60-0.33.
853 Severity : enhancement
855 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
858 Frequency : start MDS on uncleanly shutdowned MDS device
860 Descriptoin: aborting recovery hang on MDS
861 Details : don't throttle destroy RPCs for the MDT.
864 Frequency : on remount
866 Description: external journal device not working after the remount
867 Details : clear dev_rdonly flag for external journal devices in
873 Description: shutdown vs evict race
874 Details : client_disconnect_export vs connect request race.
875 if client will evicted at this time - we start invalidate
876 thread without referece to import and import can be freed
880 Frequency : rare, need acl's on inode.
882 Description: client can't handle ost additional correctly
883 Details : if ost was added after client connected to mds client can have
884 hit lnet_try_match_md ... to big messages to wide striped files.
885 in this case need teach client to handle config events about add
886 lov target and update client max ea size at that event.
888 Severity : enhancement
890 Description: Changelogs
891 Details : Changelogs are a lightweight mechanism to track filesystem
892 metadata and namespace changes. The changelog is recorded
893 permanently on the MDTs, and is periodically "consumed" / purged
894 when records are no longer needed.
896 Severity : enhancement
898 Description: compact fld format with extents
899 Details : Store range of seq rather than every seq in FLD. Seq
900 controller update FLD rather than clients. In Case of CMD, mdt0
901 has FLD, all other metadata server act as non persistent proxy
902 for FLD queries and cache fld entries in fld cache.
907 Description: don't skip ost target if they assigned to file
908 Details : Drop slow OSCs if we can, but not for requested start idx.
909 This means "if OSC is slow and it is not the requested
910 start OST, then it can be skipped, otherwise skip it only
911 if it is inactive/recovering/out-of-space.
915 Description: more cleanup in mds_lov
916 Details : not send LOV EA under replay, we can't know about they size at this
917 time. Don't allow client connect to mds before any ost connected,
918 for avoid problems with LOV EA size and returning EIO to client.
920 Severity : enhancement
922 Description: Interoperability at server side (Disk interoperability)
924 Severity : enhancement
926 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
928 Severity : enhancement
930 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
932 Severity : enhancement
934 Description: New client IO stack (CLIO).
936 Severity : enhancement
938 Description: Commit on sharing. Eliminate inter-client dependencies between
939 uncommitted transactions by doing transaction commits.
940 Thereby clients may recovery independently.
943 Frequency : Create a symlink file with a very long name
945 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
946 Details : If there is no extra space in the request for early cancels,
947 ldlm_req_handles_avail() returns 0 instead of a negative value.
949 Severity : enhancement
951 Description: Add /proc entry for import status
952 Details : The mdc, osc, and mgc import directories now have
953 an import directory that contains useful import data for debugging
956 Severity : enhancement
958 Description: Re-disable certain /proc logging
959 Details : Enable and disable client's offset_stats, extents_stats and
960 extents_stats_per_process stats logging on the fly.
963 Frequency : Only on FC kernels 2.6.22+
965 Description: oops in statahead
966 Details : Do not drop reference count for the dentry from VFS when lookup,
967 VFS will do that by itself.
969 Severity : enhancement
971 Description: Generic /proc file permissions
972 Details : Set /Proc file permissions in a more generic way to enable non-
973 root users operate on some /proc files.
977 Description: Hitting mdc_commit_close() ASSERTION
978 Details : Properly handle request reference release in
979 ll_release_openhandle().
983 Description: quota recovery deadlock during mds failover
984 Details : This patch includes att18982, att18236, att18237 in bz14840.
986 1. fix osts hang when mds does failover with quotaon
987 2. prevent watchdog storm when osts threads wait for the
992 Frequency : only patchless client
993 Description: add workaround for race between add/remove dentry from hash
995 Severity : enhancement
997 Description: Allow OST glimpses to return PW locks
1001 Description: LBUG when llog conf file is full
1002 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1006 Description: Prevent import from entering FULL state when server in recovery
1010 Description: service mount cannot take device name with ":"
1011 Details : Only when device name contains ":/" will mount treat it as
1017 Description: replace ptlrpcd with the statahead thread to interpret the async
1018 statahead RPC callback
1022 Frequency : on recovery
1023 Description: I/O failures after umount during fail back
1024 Details : if client reconnected to restarted server we need join to recovery
1025 instead of find server handler is changed and process self
1026 eviction with cancel all locks.
1028 Severity : enhancement
1030 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1032 Severity : enhancement
1034 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1036 Severity : enhancement
1038 Description: Upcall on Lustre log has been dumped
1039 Details : Allow for a user mode script to be called once a Lustre log has
1040 been dumped. It passes the filename of the dumped log to the
1041 script, the location of the script can be specified via
1042 /proc/sys/lnet/debug_log_upcall.
1047 Description: avoid idr_remove called for id which is not allocated.
1048 Details : Move assigment s_dev for clustered nfs to end of initialization,
1049 to avoid problem with error handling.
1054 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1055 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1056 into CONN_USED_HASH and this prodice warning when put connection
1057 again in unused hash.
1060 Severity : enhancement
1062 Description: File striping can now be set to use an arbitrary pool of OSTs.
1064 Severity : enhancement
1066 Description: Export bytes_read/bytes_write count on OSC/OST.
1070 Description: Early reply size mismatch, MGC loses connection
1071 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1072 the connect flags are properly negotiated.
1077 Description: quotas are not honored with O_DIRECT
1078 Details : all writes with the flag O_DIRECT will use grants which leads to
1079 this problem. Now using OBD_BRW_SYNC to guard this.
1083 Description: add quota statistics
1084 Details : 1. sort out quota proc entries and proc code.
1085 2. add quota statistics
1087 Severity : enhancement
1089 Description: enable quota support for HEAD.
1093 Description: Properly propagate oinfo flags from lov to osc for statfs
1094 Details : restore missing copy oi_flags to lov requests.
1096 Severity : enhancement
1098 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1099 : l_getgroups(8), lst(8), routerstat(8)
1100 Details : included man pages for llobdstat(8), llstat(8),
1101 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1103 Severity : enhancement
1105 Description: configure's --enable-quota should check the
1106 : kernel .config for CONFIG_QUOTA
1107 Details : configure is terminated if --enable-quota is passed but
1108 : no quota support is in kernel
1112 Description: Remove portals compatibility
1113 Details : Remove portals compatibility, not interoperable with releases
1118 Description: Resolve device initialization race
1119 Details : Prevent proc handler from accessing devices added to the
1120 obd_devs array but yet be intialized.
1122 Severity : enhancement
1124 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1126 Severity : enhancement
1128 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1133 Description: Using wrong pointer in osc_brw_prep_request
1134 Details : Access to array[-1] can produce panic if kernel compiled with
1135 CONFIG_PAGE_ALLOC enabled
1139 Description: Client runs out of low memory
1140 Details : Consider only lowmem when counting initial number of llap pages
1144 Description: *optional* service tags registration
1145 Details : if the "service tags" package is installed on a Lustre node
1146 When the filesystem is mounted, a local-node service tag will
1147 be created. See http://inventory.sun.com/ for more information
1148 about the Service Tags asset management system.
1152 Description: Kernel BUG tries to release flock
1153 Details : Lustre does not destroy flock lock before last reference goes
1154 away. So always drop flock locks when client is evicted and
1155 perform unlock regardless of successfulness of speaking to MDS.
1159 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1163 Description: sanity test 65a fails if stripecount of -1 is set
1164 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1169 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1170 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1173 Severity : enhancement
1175 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1179 Description: do not process already freed flock
1180 Details : flock can possibly be freed by another thread before it reaches
1181 to ldlm_flock_completion_ast.
1185 Description: LBUG during stress test
1186 Details : Need properly lock accesses the flock deadlock detection list.
1190 Description: oops in page fault handler
1191 Details : kernel page fault handler can return two special 'pages' in error
1192 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1196 Description: timeout with invalidate import.
1197 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1198 should be handled by ptlrpcd. This produce long age waiting and
1199 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1201 Severity : enhancement
1203 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1207 Description: enable MGS and MDT services start separately
1208 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1209 a co-located MGS without starting the MGS, which is a complement
1210 to 'nosvc' mount option.
1214 Frequency : after recovery
1215 Description: precreate to many object's after del orphan.
1216 Details : del orphan st in oscc last_id == next_id and this triger growing
1217 count of precreated objects. Set flag LOW to skip increase count
1218 of precreated objects.
1222 Frequency : rare, on clear nid stats
1223 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1224 Details : when clean nid stats sometimes try destroy live entry,
1225 and this produce panic in free.
1229 Description: Stack overflow during MDS log replay
1230 ease stack pressure by using a thread dealing llog_process.
1234 Description: wait until IO finished before start new when do lock cancel.
1235 Details : VM protocol want old IO finished before start new, in this case
1236 need wait until PG_writeback is cleared until check dirty flag and
1237 call writepages in lock cancel callback.
1239 Severity : enhancement
1241 Description: using special macro for print time and cleanup in includes.
1245 Description: mds_mfd_close() ASSERTION(rc == 0)
1246 Details : In mds_mfd_close(), we need protect inode's writecount change
1247 within its orphan write semaphore to prevent possible races.
1251 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1255 Frequency : rare, on shutdown ost
1256 Description: don't hit live lock with umount ost.
1257 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1258 use shrink_dcache_sb instead.
1262 Description: don't panic with use echo client
1263 Details : echo client pass NULL as client nid pointer and this produce null
1264 pointer dereference.
1268 Description: fix build on ppc32
1269 Details : compile code with -m64 flag produce wrong object file for ppc32.
1273 Description: add message levels for liblustreapi
1277 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1278 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1279 will result in an -EAGAIN and in the exit call path we call
1280 alloc_rr() which will with increasing aggressiveness attempt to
1281 aquire precreated objects on the minimum number of required OSCs.
1285 Description: Use old size assignment to avoid deadlock
1286 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1287 the scheduling while holding a spinlock. We do not need locking
1288 for size in ll_update_inode() because size is only updated from
1289 the MDS for directories or files without objects, so there is no
1290 other place to do the update, and concurrent access to such inodes
1291 are protected by the inode lock.
1295 Description: resolve "_IOWR redefined" build error on SLES10
1299 Description: dump the memory debugging after all modules are unloaded to
1300 suppress false negative in conf_sanity test 39
1302 Severity : enhancement
1304 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1309 Description: fix SLES kernel versioning
1310 Details : the kernel version for our SLES 10 kernel did not include a "-"
1311 before the "smp" at the end. while this was not a problem in
1312 general, it did mean that software trying to use the kernel
1313 version to try to detect a vendor specific kernel would fail.
1314 this was most evident by the OFED build scripts.
1318 Description: Don't update lov_desc members until making sure they are valid
1319 Details : When updating lov_desc members via proc fs, need fix their
1320 validities before doing the real update.
1324 Description: don't put request into delay list while invalidate in flight.
1325 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1326 import in flight. this produce timeout for invalidate and sometimes
1327 can cause stale data.
1331 Frequency : on ppc only
1332 Description: not convert ost objects for directory because it's not exist.
1333 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1337 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1339 Severity : enhancement
1341 Description: Update to support 2.6.22.14 vanilla kernel.
1345 Frequency : rare, on recovery
1346 Description: read procfs can produce deadlock in some situation
1347 Details : Holding lprocfs lock which send rpc can produce block for destroy
1348 obd objects and this also block reconnect with -EALREADY. This
1349 isn't fix all lprocfs bugs - but make it rare.
1351 Severity : enhancement
1353 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1356 Frequency : frequent on X2 node
1358 Description: mdc_set_open_replay_data LBUG
1359 Details : Set replay data for requests that are eligible for replay.
1363 Description: lustre_mgs: operation 101 on unconnected MGS
1364 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1365 MGC, and late on MGC cannot successfully connect to MGS and a lot
1366 of the error messages complaining that MGS is not connected.
1370 Frequency : on network error
1371 Description: panic with double free request if network error
1372 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1373 but it's true only for synchronus enqueue, for async enqueue
1374 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1377 Severity : enhancement
1379 Description: client-side metadata stat-ahead during readdir(directory readahead)
1380 Details : perform client-side metadata stat-ahead when the client detects
1381 readdir and sequential stat of dir entries therein
1384 Frequency : on start mds
1386 Description: Implement get_info(last_id) in obdfilter.
1389 Frequency : occasional
1391 Description: Correctly check stale fid, not start epoch if ost not support SOM
1392 Details : open with flag O_CREATE need set old fid in op_fid3 because
1393 op_fid2 was overwritten with new generated fid, but mds can answer
1394 with one of these two fids and both is not stale. Setattr
1395 incorrectly started an epoch and assume will be called
1396 done_writing, but without SOM done_writing ever being called.
1399 Frequency : rare, depends on device drivers and load
1401 Description: MDS or OSS nodes crash due to stack overflow
1402 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1403 In some cases, in conjunction with device drivers that use a lot
1404 of stack the MDS (or possibly OSS) service threads could overflow
1405 the stack. One change which was identified to consume additional
1406 stack has been reworked to avoid the extra stack usage.
1409 Frequency : occasional
1411 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1412 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1413 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1414 oscc_has_objects and see OSC has no objests and call
1415 oscc_internal_create to resend create request.
1417 Severity : enhancement
1419 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1421 Severity : enhancement
1423 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1426 Frequency : very rare
1428 Description: Fix replay if there is an un-replied request and open
1429 Details : In some cases, older replay request will revert the
1430 mcd->mcd_last_xid on MDS which is used to record the client's
1431 latest sent request.
1433 Severity : enhancement
1435 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1437 Severity : enhancement
1439 Description: Add rhel5 support to HEAD.
1441 Serverity : enhancement
1443 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1448 Description: Don't allow skipping OSTs if index has been specified.
1449 Details : Don't allow skipping OSTs if index has been specified, make
1450 locking in internal create lots better.
1454 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1455 Details : access to bitfield in structure is always rounded to long
1456 and this produce problem with not atomic change any bit.
1460 Description: Lustre make rpms failed.
1461 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1462 builds Lustre rpms from tarball.
1465 Frequency : rare on shutdown ost
1467 Description: If llog cancel was not send before clean_exports phase, this can
1468 produce deadlock in llog code.
1469 Details : If llog thread has last reference to obd and call class_import_put
1470 this produce deadlock because llog_cleanup_commit_master wait when
1471 last llog_commit_thread exited, but this never success because was
1472 called from llog_commit_thread.
1476 Description: allow userland application know is lost one of stripes.
1477 Details : fill lvb_blocks with error code on ost and return it to
1478 application if error flag found.
1482 Description: NULL lov_tgts causing MDS oops
1483 Details : more safe checks for NULL lov_tgts for avoid oops.
1485 Severity : enhancement
1487 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1491 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1492 Details : If an OST is down the MDS will hang indefinitely in
1493 obd_statfs_rqset() waiting for the statfs data. While for
1494 MDS QOS usage of statfs, it should not stuck in waiting.
1496 Severity : enhancement
1498 Description: remote_acl support
1499 Details : Support ACL-based permission check for remote user.
1500 Support setfacl/getfacl for remote user with the utils
1501 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1502 the system "{s,g}etfacl" utils.
1504 Severity : enhancement
1506 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1508 Severity : enhancement
1510 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1514 Description: make mgs_setparam() handle fsname containing dash
1515 Details : fsname containing a dash does not work with lctl conf_param
1517 Severity : enhancement
1519 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1521 Severity : enhancement
1523 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1527 Frequency : rare, at shutdown
1528 Description: access already free / zero obd_namespace.
1529 Details : if client_disconnect_export was called without force flag set,
1530 and exist connect request in flight, this can produce access to
1531 NULL pointer (or already free pointer) when connect_interpret
1532 store ocd flags in obd_namespace.
1536 Frequency : only at startup
1537 Description: not alloc memory with spinlock held.
1538 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1539 if any spinlock held.
1541 Severity : enhancement
1543 Description: make lustre randomly fail allocating memory
1544 Details : Make lustre randomly failed allocating memory for testing purpose.
1546 Severity : enhancement
1548 Description: lost problems with lov objid file
1549 Details : Fixes some scability and access to not inited memory problems
1550 in work with lov objdid file.
1555 Description: lfs find does not continue on file error
1556 Details : Continue other files processing when a file/dir is absent.
1560 Description: Inconsistent usage of lustre_pack_reply()
1561 Details : Standardize the usage of lustre_pack_reply() such that it
1562 always generate a CERROR on failure.
1565 Frequency : occasional
1567 Description: MDS hang or stay in waiting lock
1568 Details : If client receive lock with CBPENDING flag ldlm need send lock
1569 cancel as separate rpc, to avoid situation when cancel request
1570 can't processed due all i/o threads stay in wait lock.
1574 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1578 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1582 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1586 Description: Update to RHEL4 latest kernel.
1588 Severity : enhancement
1590 Description: Build SLES10 patchless client fails
1591 Details : The configure was broken by run ./configure with
1592 --with-linux-obj=.... argument for patchless client. When the
1593 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1594 can't search header adequately. Use absolute path such as
1595 -I($LINUX)/include instead.
1599 Description: interrupt oig_wait produce painc on resend.
1600 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1601 private set, and this produce situation when rq_ptlrpcd_data not
1602 copyed to new allocated request and triggered LBUG on assert
1603 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1604 ptlrpcd set if request is changed and can be safety changed to use
1609 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1610 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1611 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1612 these macros and replace them with their difinitions instead.
1616 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1617 Details : Update which_patch & target file for SLES10 latest kernel.
1620 Description: Remove client patches from SLES 10 kernel.
1621 Details : This causes SLES 10 clients to behave as patchless clients
1622 even on a Lustre-patched (server) kernel.
1624 Severity : enhancement
1626 Description: self-adjustable client's lru lists
1627 Details : use adaptive algorithm for managing client cached locks lru
1628 lists according to current server load, other client's work
1629 pattern, memory activities, etc. Both, server and client
1630 side namespaces provide number of proc tunables for controlling
1633 Severity : enhancement
1635 Description: light-weight GSS support
1636 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1637 and reduce performance overhead.
1639 Severity : enhancement
1641 Description: Linux keyring support
1642 Details : Support using service of Linux keyring for Lustre GSS internal
1643 context refresh/cache mechanism.
1647 Description: Fix errors in lfs documentation
1648 Details : Fixes man pages
1652 Description: don't use GFP_* in generic Lustre code.
1653 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1657 Description: obdclass is limited by single OBD_ALLOC(idarray)
1658 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1662 Description: warnings with build patchless client with vanila 2.6.19 and up
1663 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1664 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1668 Frequency : Only for SLES
1669 Description: sanity_quota fail test_1
1670 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1671 file, which makes TSTID[2] inunique.
1675 Frequency : Only for SLES9
1676 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1677 Details : Update target/ChangeLog/which_patch .
1681 Frequency : Only for test_10 in sanity-quota.sh
1682 Description: a bug in quota test code
1683 Details : checking a test flag in a wrong place
1687 Frequency : Only for test_13 in sanity-quota.sh
1688 Description: a bug in quota test script
1689 Details : shouldn't check overfull allocated quota
1693 Description: jbd statistics
1694 Details : Port older jbd statistics patch for sles10
1696 Severity : enhancement
1698 Description: Add printing inode info into message about error in writepage.
1703 Description: reply_lock_interpret crash due to race with it and lock cancel.
1704 Details : Do not replay locks that are being cancelled. Do not reference
1705 locks by their address during replay, just by their handle.
1709 Frequency : When flocks are used.
1710 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1711 Details : Flock locks might destroy just granted lock if it could be merged
1712 with another existing flock, this is done in completion handler,
1713 so teach ldlm_cli_enquque_fini that this is a valid case for
1719 Description: Oops in read and write path when failing to allocate lock.
1720 Details : Check if lock allocation failed and return error back.
1724 Description: lstripe command fails for valid OST index
1725 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1726 instead of lov->desc.ld_active_tgt_count.
1730 Description: sanity.sh failed test 103
1731 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1736 Description: df doesn't work properly if diskfs blocksize != 4K
1737 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1741 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1742 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1744 Severity : enhancement
1746 Description: Adaptive timeouts
1747 Details : RPC timeouts adapt to changing server load and network
1748 conditions to reduce resend attempts and improve recovery time.
1752 Description: llapi_file_create() does not allow some changes
1753 Details : add llapi_file_open() that allows specifying the mode and
1754 open flags, and also returns an open file handle.
1758 Description: merge and cleanup kernel patches.
1759 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1763 Description: Add journal checksum support.(Kernel part)
1764 Details : The journal checksum feature adds two new flags i.e
1765 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1766 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1767 indicates that the commit block contains the checksum for
1768 the blocks described by the descriptor blocks. Now commit
1769 record can be sent to disk without waiting for descriptor
1770 blocks to be written to disk. This behavior is controlled
1771 using JBD2_FEATURE_ASYNC_COMMIT flag.
1775 Description: OSS needs mutliple precreate threads
1776 Details : Add ability to start more than one create thread per OSS.
1780 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1781 Details : execute lfs setstripe on client
1785 Description: mds_obd_create error creating tmp object
1786 Details : When the user sets quota on root, llog will be affected and can't
1787 create files and write files.
1790 Frequency : Always on ia64 patchless client, and possibly others.
1792 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1793 Details : This allows the patchless client to be loaded on architectures
1794 without this export.
1798 Description: RedHat Update kernel for RHEL5
1799 Details : Add the kernel config file for RHEL5.
1803 Description: RedHat Update kernel for RHEL5
1804 Details : Modify the kernel config file more closer RHEL5.
1808 Description: Build failure against Centos5 (RHEL5)
1809 Details : Define PAGE_SIZE when it isn't present.
1813 Description: test_7 on sanity-quota.sh can't be used on separate machines
1814 Details : add facet to handle it
1818 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1819 Details : with NFS, the anon dentry's parent was set to itself in
1820 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1821 corresponding dentry other than use rec->ur_name.
1823 Severity : enhancement
1825 Description: lfs setstripe enhancement
1826 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1830 Description: enable data checksumming by default
1831 Details : enable checksum by default, allow --disable-checksum
1832 configure option and "-o nochecksum" mount option. Checksums
1833 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1837 Description: lustre support for RHEL5
1838 Details : Add support for RHEL5.
1842 Description: Client eviction due to failover config
1843 Details : after a connection loss, the lustre client should attempt to
1844 reconnect to the last active server first before trying the
1845 other potential connections.
1849 Description: when mds and osts use different quota unit(32bit and 64bit),
1850 quota will be released repeatly.
1851 Details : void sending multiple quota reqs to mds, which will keep the
1852 status between the reqs.
1856 Description: osts not allocated evenly to files
1857 Details : change the condition to increase offset_idx
1860 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1862 Description: Data corruption for OSTs that are formatted larger than 2TB
1864 Details : When generating the bio request for lustre file writes the
1865 sector number would overflow a temporary variable before being
1866 used for the IO. The data reads correctly from Lustre (which
1867 will overflow in a similar manner) but other file data or
1868 filesystem metadata may be corrupted in some cases.
1872 Description: Tune the kernel for good SCSI performance.
1873 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1874 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1879 Description: rewrite ext2-derived code in obdclass/uuid.c
1880 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1881 from scratch preserving functionality.
1885 Description: "lfs find -obd UUID" prints directories
1886 Details : "lfs find -obd UUID" will return all directory names instead
1887 of just file names. It is incorrect because the directories
1888 do not reside on the OSTs.
1892 Description: use i_size_read and i_size_write in 2.6 port
1893 Details : replace inode->i_size access with i_size_read/write()
1896 Frequency : when removing large files
1898 Description: scheduling issue during removal of large Lustre files
1899 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1900 It causes scheduling issues when removing large files (17TB in the
1904 Frequency : only with liblustre clients on XT3
1906 Description: evictions taking too long
1907 Details : allow llrd to evict clients directly on OSTs
1910 Frequency : only on ppc
1912 Description: /proc/fs/lustre/devices broken on ppc
1913 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1914 We should make sure the type of 'index' is loff_t and then cast
1915 explicitly as needed below. Do not assign an explicitly cast
1919 Frequency : only for rhel5
1921 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1922 Details : Modify the target file & which_kernel.
1924 Severity : enhancement
1926 Description: omit set fsid for export NFS
1927 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1928 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1932 Frequency : Always, for kernels after 2.6.16
1933 Description: Fix warning idr_remove called for id=.. which is not allocated.
1934 Details : Last kernels save old s_dev before kill super and not allow
1935 to restore from callback - restore it before call kill_anon_super.
1939 Description: buffer overruns could theoretically occur
1940 Details : llapi_semantic_traverse() modifies the "path" argument by
1941 appending values to the end of the origin string, and a
1942 overrun may occur. Adding buffer overrun check in liblustreapi.
1946 Description: Fix error on 'ls .' at the top of the Lustre mount.
1947 Details : Don't revalidate dentry if it is a root dentry.
1951 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1952 Details : Modify vm-tunables-rhel4.patch.
1956 Description: Kernel config for 2.6.18-vanilla.
1957 Details : Modify targets/2.6-vanilla.target.in.
1958 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1959 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1960 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1961 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1965 Description: improve handling recoverable errors
1966 Details : if request processig with error which can be recoverable on server
1967 request should be resend, otherwise page released from cache and
1972 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1973 Details : Modify target file & which_patch.
1974 A flaw was found in the IA32 system call emulation provided
1975 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1976 value could be stored in the %RAX register, which could trigger an
1977 out-of-bounds system call table access. An untrusted local user
1978 could exploit this flaw to run code in the kernel
1979 (ie a root privilege escalation). (CVE-2007-4573).
1983 Description: O_DIRECT bypasses client statistics.
1984 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1985 read_ahead_stats were updated. copied the stats section from
1986 osc_send_oap_rpc() into async_internal().
1990 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1994 Description: change order of libsysio includes
1995 Details : '#include sysio.h' should always come before '#include xtio.h'
1999 Description: handle "serious error: objid * already exists" more gracefully
2000 Details : If LAST_ID value on disk is smaller than the objects existing in
2001 the O/0/d* directories, it indicates disk corruption and causes an
2002 LBUG(). If the object is 0-length, then we should use the existing
2003 object. This will help to avoid a full fsck in most cases.
2005 Severity : enhancement
2007 Description: adapt the lustre_config script to support the upgrade case
2008 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2009 server targets to 1.6 in parallel.
2013 Description: To avoid grant space > avaible space when the disk is almost
2014 full. Without this patch you might see the error "grant XXXX >
2015 available" or some LBUG about grant, when the disk is almost
2017 Details : In filter_check_grant, for non_grant cache write, we should
2018 check the left space by if (*left > ungranted + bytes), instead
2019 of (*left > ungranted), because only we are sure the left space
2020 is enough for another "bytes", then the ungrant space should be
2021 increase. In client, we should update cl_avail_grant only there
2022 is OBD_MD_FLGRANT in the reply.
2026 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2027 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2028 described in CVE-2007-4573. This problem could allow untrusted
2029 local users to gain root access.
2032 Frequency : when using O_DIRECT and quotas
2034 Description: Incorrect file ownership on O_DIRECT output files
2035 Details : block usage reported by 'lfs quota' does not take into account
2036 files that have been written with O_DIRECT.
2041 Description: touch file failed when fs is not full
2042 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2043 otherwise we can get ENOSP while fs is not full.
2047 Description: parallel lock callbacks
2048 Details : Instead of sending blocking and completion callbacks as separated
2049 requests, adding them to a set and sending in parallel.
2052 Frequency : only for Cray XT3
2053 Bugzilla : 12829/13455
2054 Description: Changing primary group doesn't change the group lustre assigns to
2056 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2057 overridden with the primary group provided by the group upcall,
2058 whereas the supplied fsgid can be trusted if it is in the list of
2059 supplementary groups returned by the group upcall.
2061 Severity : enhancement
2063 Description: Allow masking D_WARNING, D_ERROR messages from console
2064 Details : Console messages can now be disabled via lnet.printk.
2068 Description: User code with malformed file open parameter crashes client node
2069 Details : Before packing join_file req, all the related reference should be
2070 checked carefully in case some malformed flags cause fake
2071 join_file req on client.
2075 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2076 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2077 mode is equal to the granted mode. However, it should also check
2078 that LDLM_FL_CP_REQD is not set.
2082 Description: Heavy nfs access might result in deadlocks
2083 Details : After ELC code landed, it is now improper to enqueue any mds
2084 locks under och_sem, because enqueue might want to decide to
2085 cancel open locks for same inode we are holding och_sem for.
2089 Description: Client eviction while running blogbench
2090 Details : A lot of unlink operations with concurrent I/O can lead to a
2091 deadlock causing evictions. To address the problem, the number of
2092 oustanding OST_DESTROY requests is now throttled to
2093 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2094 callbacks are processed in priority.
2098 Description: enable ACLs on MDS by default
2099 Details : ACLs must be enabled on MDS by default.
2102 Frequency : PPC/PPC64 only
2104 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2105 Details : fix duplicated definitions between asm-ppc64/types.h and
2106 lustre_types.h on PPC.
2109 Frequency : PPC/PPC64 only
2111 Description: asm-ppc/segment.h does not exist
2112 Details : fix compile issue on PPC.
2116 Description: better handle error messages in extents code
2119 Frequency : RHEL4 only
2121 Description: mkfs is very slow on IA64/RHEL4
2122 Details : A performance regression has been discovered in the MPT Fusion
2123 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2124 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2125 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2127 Severity : enhancement
2129 Description: SNMP support enhancement
2130 Details : Adding total number of sampled request for an MDS node in snmp
2133 Severity : enhancement
2135 Description: Optimize ldlm waiting list processing for PR extent locks
2136 Details : When processing waiting list for read extent lock and meeting read
2137 lock that is same or wider to it that is not contended, skip
2138 processing rest of the list and immediatelly return current
2139 status of conflictness, since we are guaranteed there are no
2140 conflicting locks in the rest of the list.
2144 Description: Time out and refuse to reconnect
2145 Details : When the failover node is the primary node, it is possible
2146 to have two identical connections in imp_conn_list. We must
2147 compare not conn's pointers but NIDs, otherwise we can defeat
2148 connection throttling.
2152 Description: port llog fixes from b1_6 into HEAD
2153 Details : Port llog reference couting and some llog cleanups from b1_6
2154 (bug 10800) into HEAD, for protect from panic and access to already
2155 free llog structures.
2159 Description: Detect stride IO mode in read-ahead
2160 Details : When a client does stride read, read-ahead should detect that and
2161 read-ahead pages according to the detected stride pattern.
2165 Description: data checksumming impacts single node performance
2166 Details : add support for several checksum algorithm. Currently, only CRC32
2167 and Adler-32 are supported. The checksum type can be changed on
2168 the fly via /proc/fs/lustre/osc/*/checksum_type.
2172 Description: use adler32 for page checksums
2173 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2178 Description: build for x2 fails
2179 Details : fix compile issue on Cray systems.
2183 Description: Properly match for duplicate locks
2184 Details : Due to different lock order from skiplists code, we need to
2185 traverse entire list for now
2188 Frequency : only on PPC/SLES10
2190 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2191 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2192 be defined. Add a hack in mkfs_lustre.c to work around this compile
2197 Description: LASSERT on MDS when client holding flock lock dies
2198 Details : ldlm pool logic depends on number of granted locks equal to
2199 number of released locks which is not true for flock locks, so
2200 just exclude such locks from consideration.
2204 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2205 Details : Use fsfilt_sync() for both the whole filesystem sync and
2206 individual file sync to eliminate dangerous inode locking
2207 with I_LOCK that can lead to a deadlock.
2211 Description: performance in 1.6.3
2212 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2213 sure that 1MB requests are not fragmented by the block layer.
2215 Severity : enhancement
2217 Description: organize the server-side client stats on per-nid basis
2218 Details : Change the structure of stats under obdfilter and mds to
2226 The "uuid"s file would list the uuids of _active_ exports.
2227 And the clear entry is to clear all stats and stale nids.
2229 Severity : enhancement
2231 Description: eliminate client locks in face of contention
2232 Details : file contention detection and lockless i/o implementation
2233 for contended files.
2237 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2240 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2242 Description: MDS may lose file striping (and hence file data) in some cases
2243 Details : If there are additional extended attributes stored on the MDS,
2244 in particular ACLs, SELinux, or user attributes (if user_xattr
2245 is specified for the client mount options) then there is a risk
2246 of attribute loss. Additionally, the Lustre file striping
2247 needs to be larger than default (e.g. striped over all OSTs),
2248 and an additional attribute must be stored initially in the
2249 inode and then increase in size enough to be moved to the
2250 external attribute block (e.g. ACL growing in size) for file
2255 Description: skiplist implementation simplification
2256 Details : skiplists are used to group compatible locks on granted list
2257 that was implemented as tracking first and last lock of each
2258 lock group the patch changes that to using doubly linked lists
2262 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2263 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2264 because LASSERT on both the data supplied by a client, and the
2265 data on disk is dangerous and incorrect.
2267 Severity : enhancement
2269 Description: Slow truncate/writes to huge files at high offsets.
2270 Details : Directly associate cached pages to lock that protect those pages,
2271 this allows us to quickly find what pages to write and remove
2272 once lock callback is received.
2276 Description: more ldlm soft lockups
2277 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2278 starve other threads from the resource lock for a long time in
2279 case of long waiting queue, so change the debug level from
2280 D_OTHER to the less frequently used D_INFO.
2282 Severity : enhancement
2284 Description: add -gid, -group, -uid, -user options to lfs find
2288 Description: Hung threads in invalidate_inode_pages2_range
2289 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2290 once one is completed. As a result, some RPCs are stuck in the
2291 queue and are never sent.
2295 Description: filter threads hungs on waiting journal commit
2296 Details : Cleanup filter group llog code, then only filter group llog will
2297 be only created in the MDS/OST syncing process.
2301 Description: Procfs and llog threads access destoryed import sometimes.
2302 Details : Sync the import destoryed process with procfs and llog threads by
2303 the import refcount and semaphore.
2305 Severity : enhancement
2307 Description: openlock cache of b1_6 port to HEAD
2312 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2313 Details : If insertion of an extent fails, then discard the inode
2314 preallocation and free data blocks else it can lead to duplicate
2319 Description: don't always update ctime in ext3_xattr_set_handle()
2320 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2321 In some cases the ctime should not be updated, for example for
2322 2.0->1.8 compatibility it is necessary to delete an xattr and it
2323 should not update the ctime.
2327 Bugzilla : 15713/16362
2328 Description: Assertion in iopen_connect_dentry in 1.6.3
2329 Details : looking up an inode via iopen with the wrong generation number can
2330 populate the dcache with a disconneced dentry while the inode
2331 number is in the process of being reallocated. This causes an
2332 assertion failure in iopen since the inode's dentry list contains
2333 both a connected and disconnected dentry.
2337 Description: assertion failure in ldlm_handle2lock()
2338 Details : fix a race between class_handle_unhash() and class_handle2object()
2339 introduced in lustre 1.6.5 by bug 13622.
2344 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2345 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2350 Description: oops in sd_iostats_seq_show()
2351 Details : unloading/reloading the scsi low level driver triggers a kernel
2352 bug when trying to access the sd iostat file.
2357 Description: Kernel panics during QLogic driver reload
2358 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2359 patch, causing memory corruption.
2364 Description: journal_dev option does not work in b1_6
2365 Details : pass mount option during pre-mount.
2367 Severity : enhancement
2369 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2370 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2371 extent information of a file. It can be used to map logical blocks
2372 in a file to physical blocks in the block device.
2376 Description: LDLM soft lockups - improvement
2377 Details : It is be possible to send the lock handle along with each read
2378 or write request because the client is already doing a lock match
2379 itself so there isn't any reason the OST should have to re-do that
2383 Frequency : only on Cray X2
2385 Description: X2 build failures
2386 Details : fix build failures on Cray X2.
2390 Description: xid & resent requests
2391 Details : Initialize RPC XID from clock at startup (randomly if clock is
2394 Severity : enhancement
2396 Description: Add lustre_start utility to start or stop multiple Lustre servers
2401 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2402 Details : In case of memory pressure, list_del() can be called twice on
2403 req->rq_history_list, causing a kernel oops.
2407 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2408 Details : fix stack overflow in the distributed lock manager by defering
2409 export eviction after a failed AST to the elt thread instead of
2410 handling it in the dlm interpret routine.
2414 Description: Convert some comments to new format.
2415 Details : Update documenting comments to match doxygen conventions.
2419 Description: Grammar fixes.
2420 Details : A couple of trivial spelling fixes.
2424 Description: OSD_COUNTERS-mandatory
2425 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2426 to simplify the code.
2430 Description: simplify cmm_device freeing logic.
2431 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2435 Description: Add lockdep support to dt_object_operations locking interface.
2436 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2437 parameter indicating lock ordering. Update mdd code to use new
2442 Description: Introduce failloc constants for lockless IO tests.
2443 Details : Add two new failloc constants to test lockless IO. Only one of
2444 them in implemented---another is checked in yet to be landed
2449 Description: Add lockdep support for inode mutex.
2450 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2451 in the situations where Lustre has to lock more than one inode
2456 Description: Add optional invariants checking support.
2457 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2458 configure switch. This macro is to be used for consistency and
2459 sanity checks that are too expensive to be left in `production'
2464 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2465 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2470 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2471 Details : Use spin_lock_nested() in (the only) situation where more than
2472 one ldlm_lock is locked simultaneously. Also, fix possible
2473 dead-lock in ldlm_lock_change_resource() by enforcing particular
2478 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2479 Details : Instead of specifying each ldlm_lock call-back through separate
2480 parameter, wrap them into struct ldlm_callback_suite.
2484 Description: Kill join_lru obd method and its callers.
2485 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2486 in memory---a requirement implemented through ->o_join_lru() obd
2487 method in HEAD. Remove this method and its users.
2491 Description: Add asynchronous ldlm ENQUEUE completion handler.
2492 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2493 case that stock ldlm_completion_ast() cannot handle as it waits
2494 until lock is granted. Introduce new ldlm_completion_ast_async()
2495 for this. Also comment ldlm_completion_ast().
2499 Description: ldlm_error <-> errno conversion.
2500 Details : Add functions to map (rather arbitrary) between LDLM error codes
2501 and standard errno values. CLIO needs this to prevent LDLM specific
2502 constants from escaping ldlm and osc.
2506 Description: Kill unused ldlm_handle2lock_ns() function.
2507 Details : Kill unused ldlm_handle2lock_ns() function.
2511 Description: Add lu_ref support to ldlm_lock
2512 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2513 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2514 and ldlm_resource. LDLM interface has to be changed, because code that
2515 releases a reference on a lock, has to "know" what reference this is.
2516 In the most frequent case
2518 lock = ldlm_handle2lock(handle);
2520 LDLM_LOCK_PUT(lock);
2522 no changes are required. When any other reference (received _not_ from
2523 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2524 instead of LDLM_LOCK_PUT().
2526 Arguably, changes are pervasive, and interface requires some discipline
2527 for proper use. On the other hand, it was very instrumental in finding
2528 a few leaked lock references.
2532 Description: Add ldlm_lock_addref_try().
2533 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2534 attempts to addref a lock that might be being canceled
2539 Description: Add ldlm_weigh_callback().
2540 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2541 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2542 hard-coded `number of pages' logic.
2546 Description: Add lockdep annotations to llog code.
2547 Details : Use appropriately tagged _nested() locking calls in the places
2548 where llog takes more than one ->lgh_lock lock.
2552 Description: Add loi_kms_set().
2553 Details : Wrap kms updates into a helper function.
2557 Description: Constify instances of struct lsm_operations.
2558 Details : Constify instances of struct lsm_operations.
2562 Description: lu_conf support.
2563 Details : On a server, a file system object is uniquely identified
2564 by a fid, which is sufficient to locate and load all object
2565 state (inode). On a client, on the other hand, more data are
2566 necessary instantiate an object. Change lu_object_find() and
2567 friends to take additional `lu_conf' argument describing object.
2568 Typically this includes layout information.
2572 Description: lu_context fixes.
2573 Details : Introduce new lu_context functions that are needed on the client
2574 side, where some system threads (ptlrpcd) are shared by multiple
2575 modules, and so cannot be stopped during module shutdown.
2579 Description: Add start and stop methods to lu_device_type_operations.
2580 Details : Introduce two new methods in lu_device_type_operations, that are
2581 invoked when first instance of a given type is created and last one
2582 is destroyed respectively. This is need by CLIO.
2586 Description: Add lu_ref support to struct lu_device.
2587 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2588 track leaked references.
2592 Description: Introduce lu_kmem_descr.
2593 Details : lu_kmem_descr and its companion interface allow to create
2594 and destroy a number of kmem caches at once.
2598 Description: Fix lu_object finalization race.
2599 Details : Fix a race between lu_object_find() finding an object and its
2600 concurrent finalization. This race is (most likely) not possible
2601 on the server, but might happen on the client.
2605 Description: Introduce lu_ref interface.
2606 Details : lu_ref is a debugging module allowing to track references to
2607 a given object. It is quite cpu expensive, and has to be
2608 explicitly enabled with --enable-lu_ref. See usage description
2613 Description: Factor lu_site procfs stats into a separate function.
2614 Details : Separate lu_site stats printing code into a separate function
2615 that can be reused on a client.
2619 Description: Constify instances of struct {lu,dt,md}_device_operations.
2620 Details : Constify instances of struct {lu,dt,md}_device_operations.
2624 Description: Introduce struct md_site and move meta-data specific parts of
2625 struct lu_site here.
2626 Details : Move md-specific fields out of struct lu_site into special struct
2627 md_site, so that lu_site can be used on a client.
2631 Description: Kill mdd_lov_destroy().
2632 Details : Remove unused mdd code.
2636 Description: Add st_block checking to multistat.c.
2637 Details : Add st_block checking to multistat.c.
2641 Description: Add lu_ref support to struct obd_device.
2642 Details : Add lu_ref tracking to obd_device.
2646 Description: Kill obd_set_fail_loc().
2647 Details : Remove unused code.
2651 Description: Add special type for ptlrpc_request interpret functions.
2652 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2653 there. Actual usage will be in CLIO.
2657 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2658 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2659 doesn't work with lockdep.
2663 Description: Add rwv.c test program.
2664 Details : New testing program exercising readv(2) and writev(2) (Qian).
2668 Description: Add sendfile.c test program.
2669 Details : New testing program exercising sendfile(2) (Jay).
2673 Description: Ratelimit a message that can be very frequent.
2674 Details : Ratelimit a memory allocation failure message that can
2679 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2680 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2684 Description: Liblustre build fixes.
2685 Details : Liblustre build fixes.
2689 Description: libcfs: add cfs_{need,cond}_resched() interface.
2690 Details : libcfs: add cfs_{need,cond}_resched() definition and
2691 implementations for Linux, NT, and liblustre.
2693 Severity : enhancement
2695 Description: More exported tunables for mballoc
2696 Details : Add support for tunable preallocation window and new tunables for
2697 large/small requests
2701 Description: Detect corruption of block bitmap and checking for preallocations
2702 Details : Checks validity of on-disk block bitmap. Also it does better
2703 checking of number of applied preallocations. When corruption is
2704 found, it turns filesystem readonly to prevent further corruptions.
2708 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2709 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2710 window to 0. If it does hit cache miss, and read-pattern is still
2711 stride-io mode, does not reset the stride window, but also does not
2712 increase the stride window length in this case.
2716 Frequency : only for big-endian servers
2717 Description: Check if system is big-endian while mounting fs with extents feature
2718 Details : Mounting a filesystem with extents feature will fail on big-endian
2719 systems since ext3-based ldiskfs is not supported on big-endian
2720 systems. This can be over-riden with "bigendian_extents" mount option.
2722 Severity : enhancement
2724 Description: The root squash functionality
2725 Details : A security feature, which is to prevent users from being able
2726 to mount lustre on their desktop, run as root, and delete
2727 all of the files in the filesystem. The goal is accomplished by
2728 remapping user id (UID) and group id (GID) of the root user to
2729 a UID and GID specified by the system administartor via Lustre
2730 configuration management server (MGS). The functionality also
2731 allows to specify sets of clients for which the remapping does
2736 Description: Excessive recovery window
2737 Details : With AT enabled, the recovery window can be excessively long (6000+
2738 seconds). To address this problem, we no longer use
2739 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2740 timeout no longer depends on the service time, it is set to
2741 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2742 time via pb_service_time.
2746 Description: Watchdog triggered on MDS failover
2747 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2748 the OSTs know that the MDS "UUID" can be reused for the same export
2749 from a different NID, so we do not need to wait for the export to be
2753 Frequency : rare, only if using MMP with Linux RAID
2755 Description: MMP doesn't work with Linux RAID
2756 Details : While using HA for Lustre servers with Linux RAID, it is possible
2757 that MMP will not detect multiple mounts. To make this work we
2758 need to unplug the device queue in RAID when the MMP block is being
2759 written. Also while reading the MMP block, we should read it from
2760 disk and not the cached one.
2762 Severity : enhancement
2764 Description: open file using fid
2765 Details : A file can be opened using just its fid, like
2766 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2769 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2772 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2773 Details : add proper locking for creation of jbd_4k slab cache
2777 Description: MMP check in ext3_remount() fails without displaying any error
2778 Details : When multiple mount protection fails during remount, proper error
2781 Severity : enhancement
2783 Description: Allow stripe size to be up to 4G-64k
2784 Details : Fix math logic to allow large stripe sizes.
2788 Description: add check for >8TB ldiskfs filesystems
2789 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2790 Don't allow >8TB ldiskfs filesystems to be mounted without
2791 force_over_8tb mount option
2796 Descriptoin: OSS complains "too many blocks" and crashes
2797 Details : In filter_commitrw_write():
2798 1. lquota_chkquota();
2799 2. obdo_from_inode();
2800 3. lquota_pending_commit();
2801 "2." has changed uid and gid in this case, then caused
2802 lquota_chkquota() and lquota_pending_commit() to use
2803 different uid and gid. That is the root of the bug.
2807 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2808 Details : Detect and print out the 'stripe count = -1' case properly.
2812 Description: Out or order replies might be lost on replay
2813 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2814 request with tid smaller than one currently being added, add it
2815 to the start, not end of the list.
2817 --------------------------------------------------------------------------------
2819 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2821 * Support for kernels:
2822 2.6.5-7.283 (SLES 9),
2823 2.6.9-55.EL (RHEL 4),
2824 2.6.16.46-0.14 (SLES 10),
2825 2.6.18.8 vanilla (kernel.org)
2826 * Client support for unpatched kernels:
2827 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2828 2.6.16 - 2.6.22 vanilla (kernel.org)
2829 * Due to recently discovered recovery problems, we do not recommend
2830 using patchless RHEL 4 clients with this or any earlier release.
2831 * Recommended e2fsprogs version: 1.39.cfs8
2832 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2833 * Starting with this release, the ldiskfs backing filesystem required
2834 by Lustre is now in its own package, lustre-ldiskfs. This package
2835 should be installed. It is versioned separately from Lustre and
2836 may be released separately in future.
2840 Description: block reactivating mgc import until all deactivates complete
2841 Details : Fix race when failing back MDT/MGS to itself (testing)
2844 Frequency : at statup only
2846 Description: mds_lov_synchronize race leads to various problems
2847 Details : simultaneous MDT->OST connections at startup can cause the
2848 sync to abort, leaving the OSC in a bad state.
2850 Severity : enhancement
2852 Description: add optional extra BUILD_VERSION info
2853 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2854 to override the lustre version.
2857 Frequency : 2.6.18 servers only
2859 Description: ll_kern_mount() doesn't release the module reference
2860 Details : The ldiskfs module reference count never drops down to 0
2861 because ll_kern_mount() doesn't release the module reference.
2866 Description: server LBUG when using old ost_num_threads parameter
2867 Details : Accept the old ost_num_threads parameter but warn that it
2868 is deprecated, and fix an off-by-one error that caused an LBUG.
2873 Description: Transient SCSI error results in persistent IO issue
2874 Details : iobuf->dr_error is not reinitialized to 0 between two
2878 Frequency : sometimes when underlying device returns I/O errors
2880 Description: OSTs not going read-only during write failures
2881 Details : OSTs are not remounted read-only when the journal commit threads
2882 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2883 instead of the ext3 wrappers.
2887 Description: poor connect scaling with increasing client count
2888 Details : Don't run filter_grant_sanity_check for more than 100 exports
2889 to improve scaling for large numbers of clients.
2892 Frequency : SLES10 only
2894 Description: sanity-quota.sh quotacheck failed: rc = -22
2895 Details : Quotas cannot be enabled on SLES10.
2898 Frequency : liblustre clients only
2900 Description: getdirentries does not give error when run on compute nodes
2901 Details : getdirentries does not fail when the size specified as an argument
2902 is too small to contain at least one entry
2904 Severity : enhancement
2906 Description: Add LNET router traceability for debug purposes
2907 Details : If a checksum failure occurs with a router as part of the
2908 IO path, the NID of the last router that forwarded the bulk data
2909 is printed so it can be identified.
2914 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2915 Details : Due to a race condition, liblustre clients were occasionally
2916 evicted incorrectly.
2918 Severity : enhancement
2920 Description: lfs setstripe use optional parameters instead of postional
2923 Severity : enhancement
2925 Description: Nanosecond timestamp support for ldiskfs
2926 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2927 resolution timestamps. There is not yet support for this at
2928 the Lustre filesystem level.
2931 Frequency : during server recovery
2933 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2934 Details : request with rq_no_resend flag not awake l_wait_event if they get
2938 Frequency : nfs export on patchless client
2940 Description: connectathon hang when test nfs export over patchless client
2941 Details : Disconnected dentry cannot be found with lookup, so we do not need
2942 to unhash it or make it invalid
2945 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2946 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2947 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2952 Description: open req refcounting wrong on reconnect
2953 Details : If reconnect happened between getting open reply from server and
2954 call to mdc_set_replay_data in ll_file_open, we will schedule
2955 replay for unreferenced request that we are about to free.
2956 Subsequent close will crash in variety of ways.
2957 Check that request is still eligible for replay in
2958 mdc_set_replay_data().
2963 Description: disable writes to filesystem when reading health_check file
2964 Details : the default for reading the health_check proc file has changed
2965 to NOT do a journal transaction and write to disk, because this
2966 can cause reads of the /proc file to hang and block HA state
2967 checking on a healthy but otherwise heavily loaded system. It
2968 is possible to return to the previous behaviour during configure
2969 with --enable-health-write.
2971 Severity : enhancement
2973 Description: 64-bit inode version
2974 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2975 to the inode. This will be required for version-based recovery.
2980 Description: MDS fails to start if a duplicate client export is detected
2981 Details : in some rare cases it was possible for a client to connect to
2982 an MDS multiple times. Upon recovery the MDS would detect this
2983 and fail during startup. Handle this more gracefully.
2985 Severity : enhancement
2987 Description: Add -o localflock option to simulate old noflock
2989 Details : This will achieve local-only flock/fcntl locks
2995 Description: log_commit_thread vs filter_destroy race leads to crash
2996 Details : Take import reference before releasing llog record semaphore
3001 Description: Wrong request locking in request set processing
3002 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3003 imp_delayed_list, in this place should be used imp_lock.
3006 Frequency : when reconnection
3008 Description: Grant Leak when osc reconnect to OST
3009 Details : When osc reconnect ost, OST(filter) should check whether it
3010 should grant more space to client by comparing fed_grant and
3011 cl_avail_grant, and return the granted space to client instead
3012 of "new granted" space, because client will call osc_init_grant
3013 to update the client grant space info.
3016 Frequency : when client reconnect to OST
3018 Description: Grant Leak when osc do resend and replay bulk write
3019 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3020 bulk write request, because the grant info while be sync between
3021 OSC and OST when reconnect, and we should ignore the grant info
3022 these of resend/replay write req.
3027 Description: Grant space more than available left space sometimes.
3028 Details : When then OST is about to be full, if two bulk writing from
3029 different clients came to OST. Accord the avaliable space of the
3030 OST, the first req should be permitted, and the second one
3031 should be denied by ENOSPC. But if the seconde arrived before
3032 the first one is committed. The OST might wrongly permit second
3033 writing, which will cause grant space > available space.
3036 Frequency : when client is evicted
3038 Description: Grant might be wrongly erased when osc is evicted by OST
3039 Details : when the import is evicted by server, it will fork another
3040 thread ptlrpc_invalidate_import_thread to invalidate the
3041 import, where the grant will be set to 0. While the original
3042 thread will update the grant it got when connecting. So if
3043 the former happened latter, the grant will be wrongly errased
3044 because of this race.
3049 Description: Checking Stale with correct fid
3050 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3051 to check whether it is stale, because sometimes, we want the
3052 enqueue happened anyway, and op_data.fid2 will not be initialized.
3054 Severity : enhancement
3056 Description: update patchless client
3057 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3060 Frequency : only with 2.4 kernel
3062 Description: random memory corruption
3063 Details : size of struct ll_inode_info is to big for union inode.u and this
3064 can be cause of random memory corruption.
3069 Description: Memory leak in recovery
3070 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3071 It should also check obd_fail before fsfilt_start, otherwise if
3072 fsfilt_start return -EROFS,(failover mds during mds recovery).
3073 then the req will return with repmsg->transno = 0 and rc = EROFS.
3074 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3075 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3076 be freed no matter whether fsfilt_commit success or not.
3079 Frequency : only with huge count clients
3081 Description: Prevents from taking the superblock lock in llap_from_page for
3083 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3084 for llap_from_page call in ll_removepage prevents from taking the
3085 superblock lock for a soon died page.
3090 Description: Not check open intent error before release open handle
3091 Details : in some rare cases, the open intent error is not checked before
3092 release open handle, which may cause
3093 ASSERTION(open_req->rq_transno != 0), because it tries to release
3094 the failed open handle.
3099 Description: Set cat log bitmap only after create log success.
3100 Details : in some rare cases, the cat log bitmap is set too early. and it
3101 should be set only after create log success.
3105 Description: Accessing a block bevice can re-enable I/O when Lustre is
3106 tearing down a device.
3107 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3112 Description: service threads may hog cpus when there are a lot of requests
3113 Details : Insert cond_resched to give other threads a chance to use some CPU
3118 Description: the cat log was not initialized in recovery
3119 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3120 unlink log on mds will not be initialized until mds post
3121 recovery. And also in mds post recovery, the unlink log will
3122 initialization will be done asynchronausly, so there will be race
3123 between add unlink log and unlink log initialization.
3127 Description: brw_stats were being printed incorrectly
3128 Details : brw_stats were being printed as log2 but all of them were not
3129 recorded as log2. Also remove some code duplication arising from
3130 filter_tally_{read,write}.
3134 Frequency : rare, only in recovery.
3135 Description: ASSERTION(req->rq_type != LI_POISON) failed
3136 Details : imp_lock should be held while iterating over imp_sending_list for
3137 prevent destroy request after get timeout in ptlrpc_queue_wait.
3141 Description: replay-single.sh test 52 fails
3142 Details : A lock's skiplist need to be cleanup when it being unlinked
3143 from its resource list.
3147 Description: Short directio read returns full requested size rather than
3149 Details : Direct I/O operations should return actual amount of bytes
3150 transferred rather than requested size.
3152 Severity : enhancement
3154 Description: metadata RPC reduction (e.g. for rm performance)
3155 Details : decrease the amount of synchronous RPC between clients and servers
3156 by canceling conflicing lock before the operation on the client
3157 and packing thier handles into the main operation RPC to server.
3159 Severity : enhancement
3161 Description: Async OSC create to avoid the blocking unnecessarily.
3162 Details : If an OST has no remaining object, system will block on the
3163 creation when it needs to create a new object on this OST. Now,
3164 ways use pre-created objects when available, instead of blocking on
3165 an empty osc while others are not empty. If we must block, we
3166 block for the shortest possible period of time.
3170 Description: improve handling recoverable errors
3171 Details : if request processig with error which can be recoverable on server
3172 request should be resend, otherwise page released from cache and
3175 Severity : enhancement
3177 Description: refine locking for avoid write wrong info into lov_objid file
3178 Details : fix possible races with add new target and write/update data in
3181 --------------------------------------------------------------------------------
3183 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3188 Frequency : on some architectures
3190 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3191 Details : Uninitialized flags sometimes cause configuration commands to
3195 Frequency : patchless clients only
3197 Description: missing __iget() symbol export
3198 Details : The __iget() symbol export is missing. To avoid the need for
3199 this on patchless clients the deathrow inode reaper is turned
3200 off, and we depend on the VM to clean up old inodes. This
3201 dependency was during via the fix for bug 12181.
3203 --------------------------------------------------------------------------------
3205 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3207 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3208 INTEROPERATE with older versions automatically. In many cases a
3209 special upgrade step is needed. Please read the
3210 user documentation before upgrading any part of a live system.
3211 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3212 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3213 * WARNING: Lustre configuration and startup changes are required with
3214 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3218 Severity : enhancement
3220 Description: MountConf
3221 Details : Lustre configuration is now managed via mkfs and mount
3222 commands instead of lmc and lconf. New obd types (MGS, MGC)
3223 are added for dynamic configuration management. See
3224 https://mail.clusterfs.com/wikis/lustre/MountConf for
3227 Severity : enhancement
3229 Description: dynamic OST addition
3230 Details : OSTs can now be added to a live filesystem
3232 Severity : enhancement
3234 Description: startup order invariance
3235 Details : MDTs and OSTs can be started in any order. Clients only
3236 require the MDT to complete startup.
3238 Severity : enhancement
3240 Description: parallel, asynchronous orphan cleanup
3241 Details : orphan cleanup is now performed in separate threads for each
3242 OST, allowing parallel non-blocking operation.
3244 Severity : enhancement
3246 Description: optimized stripe assignment
3247 Details : stripe assignments are now made based on ost space available,
3248 ost previous usage, and OSS previous usage, in order to try
3249 to optimize storage space and networking resources.
3251 Severity : enhancement
3253 Description: Permanently set tunables
3254 Details : All writable /proc/fs/lustre tunables can now be permanently
3255 set on a per-server basis, at mkfs time or on a live system.
3257 Severity : enhancement
3259 Description: Lustre message v2
3260 Details : Add lustre message format v2.
3262 Severity : enhancement
3264 Description: client OST exclusion list
3265 Details : Clients can be started with a list of OSTs that should be
3266 declared "inactive" for known non-responsive OSTs.
3270 Description: SPEC SFS validation failure on NFS v2 over lustre.
3271 Details : Changes the blocksize for regular files to be 2x RPC size,
3272 and not depend on stripe size.
3274 Severity : enhancement
3276 Description: Multiple MD RPCs in flight.
3277 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3278 To avoid overly-overloading MDS, introduce a limit on number of
3279 MDS RPCs in flight for a single client and add /proc controls
3280 to adjust this limit.
3282 Severity : enhancement
3284 Description: client read/write statistics
3285 Details : Add client read/write call usage stats for performance
3286 analysis of user processes.
3287 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3288 file access. extents_stats shows chunk size distribution.
3289 extents_stats_per_process show chunk size distribution per
3292 Severity : enhancement
3294 Description: mds statistics
3295 Details : Add detailed mds operations statistics in
3296 /proc/fs/lustre/mds/*/stats.
3300 Description: Failure of copying files with lustre special EAs.
3301 Details : Client side always return success for setxattr call for lustre
3302 special xattr (currently only "trusted.lov").
3307 Description: Refcount LNET uuids
3308 Details : The global LNET uuid list grew linearly with every startup;
3309 refcount repeated list entries instead of always adding to
3312 Severity : enhancement
3314 Description: Dynamic service threads
3315 Details : Within a small range, start extra service threads
3316 automatically when the request queue builds up.
3319 Frequency : mixed-endian client/server environments
3321 Description: mixed-endian crashes
3322 Details : The new msg_v2 system had some failures in mixed-endian
3325 Severity : enhancement
3327 Description: Easy OST removal
3328 Details : OSTs can be permanently deactivated with e.g. 'lctl
3329 conf_param lustre-OST0001.osc.active=0'
3331 Severity : enhancement
3333 Description: MGS proc entries
3334 Details : Added basic proc entries for the MGS showing what filesystems
3337 Severity : enhancement
3339 Description: provide MGS failover
3340 Details : Added config lock reacquisition after MGS server failover.
3342 Severity : enhancement
3344 Description: add Linux 2.4 support
3345 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3349 Description: a large application tries to do I/O to the same resource and dies
3350 in the middle of it.
3351 Details : Check the req->rq_arrival time after the call to
3352 ost_brw_lock_get(), but before we do anything about
3353 processing it & sending the BULK transfer request. This
3354 should help move old stale pending locks off the queue as
3355 quickly as obd_timeout.
3358 Frequency : when an incorrect nid is specified during startup
3360 Description: ptlrpc connect to non-existant node causes kernel crash
3361 Details : LNET can't be re-entered from an event callback, which
3362 happened when we expire a message after the export has been
3363 cleaned up. Instead, hand the zombie cleanup off to another
3366 Severity : enhancement
3368 Description: plain/inodebits lock performance improvement
3369 Details : Grouping plain/inodebits in granted list by their request modes
3370 and bits policy, thus improving the performance of search through
3374 Frequency : only if OST filesystem is corrupted
3376 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3377 Details : for a short time RPCs with bulk IO are in the replay list,
3378 but replay of bulk IOs is unimplemented. If the OST filesystem
3379 is corrupted due to disk cache incoherency and then replay is
3380 started it is possible to trip an assertion. Avoid putting
3381 committed RPCs into the replay list at all to avoid this issue.
3384 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3386 Description: System hang on startup
3387 Details : This bug allowed the liblustre (e.g. catamount) client to
3388 return to the app before handling all startup RPCs. This
3389 could leave the node unresponsive to lustre network traffic
3390 and manifested as a server ptllnd timeout.
3392 Severity : enhancement
3394 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3395 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3396 boolean to control whether to print peer debug info when a
3397 client's RPC times out.
3400 Frequency : only for kernels with patches from Lustre below 1.4.3
3402 Description: Remove old rdonly API
3403 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3406 Frequency : only for devices with external journals
3408 Description: Set external device read-only also
3409 Details : During a commanded failover stop, we set the disk device
3410 read-only while the server shuts down. We now also set any
3411 external journal device read-only at the same time.
3414 Frequency : when upgrading from 1.4 while trying to change parameters
3416 Description: The wrong (new) MDC name was used when setting parameters for
3417 upgraded MDT's. Also allows changing of OSC (and MDC)
3418 parameters if --writeconf is specified at tunefs upgrade time.
3421 Frequency : when setting specific ost indicies
3423 Description: QOS code breaks on skipped indicies
3424 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3425 created with --index need not be sequential.
3427 Severity : enhancement
3429 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3430 Details : The uninit_groups feature works in conjunction with the kernel
3431 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3432 up the pass1 processing of e2fsck. This is a read-only feature
3433 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3434 mount filesystems that have had this feature enabled.
3436 Severity : enhancement
3438 Description: Improve multi-block allocation algorithm to avoid fragmentation
3439 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3440 allocation locality and avoid filesystem fragmentation.
3442 ------------------------------------------------------------------------------
3444 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3446 * Support for kernels:
3447 2.4.21-47.0.1.EL (RHEL 3)
3448 2.6.5-7.283 (SLES 9)
3449 2.6.9-42.0.10.EL (RHEL 4)
3450 2.6.12.6 vanilla (kernel.org)
3451 2.6.16.27-0.9 (SLES 10)
3452 * Recommended e2fsprogs version: 1.39.cfs5
3454 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3458 Frequency : occasional, depends on client load and configuration
3459 Bugzilla : 12181, 12203
3460 Description: data loss for recently-modified files
3462 Details : In some cases it is possible that recently written or created
3463 files may not be written to disk in a timely manner (this should
3464 normally be within 30s unless client IO load is very high).
3465 The problem appears as zero-length files or files that are a
3466 multiple of 1MB in size after a client crash or client eviction
3467 that are missing data at the end of the file.
3469 This problem is more likely to be hit on clients where files are
3470 repeatedly created and unlinked in the same directory, clients
3471 have a large amount of RAM, have many CPUs, the filesystem has
3472 many OSTs, the clients are rebooted frequently, and/or the files
3473 are not accessed by other nodes after being written.
3475 The presence of the problem can be detected by looking at
3476 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3477 smaller than the second (nr_unused) then dirty files will not
3478 be flushed automatically to disk. "sync; sleep 10" should be
3479 run several times on the node before unmounting it to update
3480 Lustre (this is also safe to run on nodes without this problem).
3482 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3483 that can cause this same problem, so customers using that kernel
3484 also need to update the kernel in addition to Lustre. In order
3485 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3487 It is normal that files written just before a client crash (less
3488 than 30s) may not yet have been flushed to disk, even for local
3492 Frequency : frequent on thin XT3 nodes
3494 Description: UUID collision on thin XT3 Linux nodes
3495 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3496 randomly, since we relied on an insufficiently-seeded PRNG.
3501 Description: OSS hangs after "All ost request buffers busy"
3502 Details : A deadlock between quota and journal operations caused OSS
3503 hangs after printing "All ost request buffers busy."
3506 Frequency : always on liblustre builds
3508 Description: Cleanup compiler warnings on liblustre
3511 Frequency : always on liblustre builds on XT3
3513 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3517 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3518 > 32000 subdirectories
3519 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3520 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3521 subdirectory count crosses 32000. This will aid e2fsck to
3522 correctly handle more than 32000 subdirectories.
3525 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3527 Description: System hang on startup
3528 Details : This bug allowed the liblustre (e.g. catamount) client to
3529 return to the app before handling all startup RPCs. This
3530 could leave the node unresponsive to lustre network traffic
3531 and manifested as a server ptllnd timeout.
3533 Severity : enhancement
3535 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3536 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3537 boolean to control whether to print peer debug info when a
3538 client's RPC times out.
3543 Description: make O_SYNC working on 2.6 kernels
3544 Details : 2.6 kernels use different method for mark pages for write,
3545 so need add a code to lustre for O_SYNC work.
3550 Description: Failure to close file and release space on NFS
3551 Details : Put inode details into lock acquired in ll_intent_file_open.
3552 Use mdc_intent_lock in ll_intent_open to properly
3553 detect all kind of errors unhandled by mdc_enqueue.
3558 Description: proc file read during shutdown sometimes raced obd removal,
3560 Details : Add lock to prevent obd access after proc file removal.
3563 Frequency : Only for files larger than 4GB on 32-bit clients.
3565 Description: improperly doing page alignment of locks
3566 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3567 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3572 Description: RPCs being resent when they shouldn't be
3573 Details : Some RPCs that should not be resent are being resent. This
3574 can cause inconsistencies in the RPC state machine. Do not
3575 resend such requests.
3578 Frequency : rare, only with NFS export
3580 Description: Crash on NFS re-export node
3581 Details : under very unusual load conditions an assertion is hit in
3582 ll_intent_file_open()
3585 Frequency : only if OST filesystem is corrupted
3587 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3588 Details : for a short time RPCs with bulk IO are in the replay list,
3589 but replay of bulk IOs is unimplemented. If the OST filesystem
3590 is corrupted due to disk cache incoherency and then replay is
3591 started it is possible to trip an assertion. Avoid putting
3592 committed RPCs into the replay list at all to avoid this issue.
3597 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3598 Details : Large single O_DIRECT read and write calls can fail to allocate
3599 a sufficiently large buffer to process the request. In case of
3600 allocation failure the allocation is retried with a smaller
3601 buffer and broken into smaller requests.
3603 Severity : enhancement
3605 Description: Add -o localflock option to simulate old noflock behaviour.
3606 Details : This will achieve local-only flock/fcntl locks coherentness.
3611 Description: versioning check is incomplete
3612 Details : Checking the version difference of client vs. server, report
3613 error if the gap is too big.
3618 Description: add support PG_writeback bit
3619 Details : add support for PG_writeback bit for Lustre, for more carefull
3620 work with page cache in 2.6 kernel. This also fix some deadlocks
3621 and remove hack for work O_SYNC with 2.6 kernel.
3623 Severity : enhancement
3625 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3626 Details : The uninit_groups feature works in conjunction with the kernel
3627 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3628 up the pass1 processing of e2fsck. This is a read-only feature
3629 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3630 mount filesystems that have had this feature enabled.
3632 Severity : enhancement
3634 Description: Improve multi-block allocation algorithm to avoid fragmentation
3635 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3636 allocation locality and avoid filesystem fragmentation.
3638 ------------------------------------------------------------------------------
3640 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3642 * Support for kernels:
3643 2.6.9-42.0.3.EL (RHEL 4)
3644 2.6.5-7.276 (SLES 9)
3645 2.4.21-47.0.1.EL (RHEL 3)
3646 2.6.12.6 vanilla (kernel.org)
3647 2.6.16.21-0.8 (SLES10)
3648 * Recommended e2fsprogs version: 1.39.cfs2-0
3650 * The backwards-compatible /proc/sys/portals symlink has been removed
3651 in this release. Before upgrading, please ensure that you change
3652 any configuration scripts or /etc/sysctl.conf files that access
3653 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3654 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3655 in advance of the upgrade on any system running Lustre 1.4.6 or
3656 newer, since /proc/sys/lnet was added in that version.
3657 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3661 Frequency : only when quota is used
3663 Description: avoid scanning export list for quota master
3664 Details : Change the algorithms to avoid scanning export list in order
3665 to improve the efficiency.
3668 Frequency : MDS failover only, very rarely
3670 Description: "went back in time" messages on mds failover
3671 Details : The greatest transno may be lost when the current operation
3672 finishes with an error (transno==0) and the client's last_rcvd
3673 record is over-written. Save the greatest transno in the
3674 mds_last_transno for this case.
3677 Frequency : always for specific kernels and striping counts
3679 Description: client may get "Matching packet too big" without ACL support
3680 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3681 when trying to access files in certain configurations. The
3682 clients should in fact be denied when mounting because they do
3683 not understand ACLs.
3686 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3688 Description: many clients connecting with IO in progress causes connect timeouts
3689 Details : Avoid synchronous journal commits to avoid delays caused by many
3690 clients connecting/disconnecting when bulk IO is in progress.
3691 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3692 OST_IO_PORTAL to avoid delays behind potentially many pending
3696 Frequency : occasionally with multiple writers to a single file
3698 Description: shared writes to file may result in wrong size reported by stat()
3699 Details : Allow growing of kms when extent lock is cancelled
3702 Frequency : always with random mmap IO to multi-striped file
3704 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3705 Details : Only if the hole is at the end of OST object so that kms is too
3706 small. Fix is to increase kms accordingly in ll_nopage.
3709 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3711 Description: writes to a missing object would leak memory on the OST
3712 Details : If there is an inconsistency between the MDS and OST filesystems,
3713 such that the MDS references an object that doesn't exist, writes
3714 to that object will leak memory due to incorrect cleanup in the
3715 error handling path, eventually running out of memory on the OST.
3720 Description: Creating too long symlink causes lustre errors
3721 Details : Check symlink and name lengths before sending requests to MDS.
3724 Frequency : only if flock is enabled (not on by default)
3726 Description: posix locks not released on fd closure on 2.6.9+
3727 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3728 kernels, this caused such locks not to be released on fd close and
3729 then assertions on fs unmount about still used locks.
3732 Frequency : MDS failover only, very rarely
3734 Description: clients may get ASSERTION(granted_lock != NULL)
3735 Details : When request was taking a long time, and a client was resending
3736 a getattr by name lock request. The were multiple lock requests
3737 with the same client lock handle and
3738 mds_getattr_name->fixup_handle_for_resent_request found one of the
3739 lock handles but later failed with ASSERTION(granted_lock != NULL).
3744 Description: handle->h_buffer_credits > 0, assertion failure
3745 Details : h_buffer_credits is zero after truncate, causing assertion
3746 failure. This patch extends the transaction or creates a new
3750 Frequency : NFS re-export or patchless client
3751 Bugzilla : 11179, 10796
3752 Description: Crash on NFS re-export node (__d_move)
3753 Details : We do not want to hash the dentry if we don't have a lock.
3754 But if this dentry is later used in d_move, we'd hit uninitialised
3755 list head d_hash, so we just do this to init d_hash field but
3756 leave dentry unhashed.
3759 Frequency : NFS re-export or patchless client
3761 Description: NFS exports has problem with symbolic link
3762 Details : lustre client didn't properly install dentry when re-exported
3763 to NFS or running patchless client.
3766 Frequency : NFS re-export or patchless client
3768 Description: Various nfs/patchless fixes.
3769 Details : fixes reuse disconected alias for lookup process - this fixes
3770 warning "find_exported_dentry: npd != pd",
3771 fix permission error with open files at nfs.
3772 fix apply umask when do revalidate.
3775 Frequency : occasional
3777 Description: Crash on NFS re-export node
3778 Details : calling clear_page() on the wrong pointer triggered oops in
3779 generic_mapping_read().
3782 Frequency : rarely, using O_DIRECT IO
3784 Description: unaligned directio crashes client with LASSERT
3785 Details : check for unaligned buffers before trying any requests.
3788 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3790 Description: stale data returned from RAID cache
3791 Details : If only a small amount of IO is done to the RAID device before
3792 reading it again it is possible to get stale data from the RAID
3793 cache instead of reading it from disk.
3796 Frequency : always for sles10 kernel
3798 Description: sles10 support
3799 Details : ll_follow_link: compile fixes and using of nd_set_link
3800 under newer kernels.
3803 Frequency : depends on arch, kernel and compiler version, always on sles10
3806 Description: recursive or deep enough symlinks cause stack overflow
3807 Details : getting rid of large stack-allocated variable in
3811 Frequency : depends on hardware
3813 Description: lustre write performance loss in the SLES10 kernel
3814 Details : the performance loss is caused by using of write barriers in the
3815 ext3 code. The SLES10 kernel turns barrier support on by
3816 default. The fix is to undo that change for ldiskfs.
3818 ------------------------------------------------------------------------------
3820 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3822 * Support for kernels:
3823 2.6.9-42.0.3EL (RHEL 4)
3824 2.6.5-7.276 (SLES 9)
3825 2.4.21-47.0.1.EL (RHEL 3)
3826 2.6.12.6 vanilla (kernel.org)
3830 Frequency : quota enabled and large files being deleted
3832 Description: releasing more than 4GB of quota at once hangs OST
3833 Details : If a user deletes more than 4GB of files on a single OST it
3834 will cause the OST to spin in an infinite loop. Release
3835 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3840 Description: statfs data retrieved from /proc may be stale or zero
3841 Details : When reading per-device statfs data from /proc, in the
3842 {kbytes,files}_{total,free,avail} files, it may appear
3843 as zero or be out of date.
3846 Frequency : systems with MD RAID1 external journal devices
3848 Description: lconf's call to blkid is confused by RAID1 journal devices
3849 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3850 of returning all block devices that match the journal UUID.
3853 Frequency : always, for aggregate stripe size over 4GB
3855 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3856 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3857 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3860 Frequency : always if "lfs find" used on a local file/directory
3862 Description: "lfs find" segfaults if used on a local file/directory
3863 Details : The case where a directory component was not specified wasn't
3864 handled correctly. Handle this properly.
3867 Frequency : always on ppc64
3869 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3870 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3871 perform endianness conversion.
3874 Frequency : rarely (truncate to non-zero file size after write under load)
3875 Bugzilla : 10730, 10687
3876 Description: Files padded with zeros to next 4K multiple
3877 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3878 it is possible that files that are truncated to a non-zero size
3879 immediately after being written are filled with zero bytes beyond
3880 the truncated size. No file data is lost.
3882 Severity : enhancement
3884 Description: Allow recovery/failover for liblustre clients.
3885 Details : liblustre clients were unaware of failover configurations until
3888 Severity : enhancement
3890 Description: user file locks should fail when not mounting with flock option
3891 Details : Set up an error-returning stub in ll_file_operations.lock field
3892 to prevent incorrect behaviour when client is mounted without
3893 flock option. Also, set up properly f_op->flock field for
3897 Frequency : always on ia64
3899 Description: "lfs df" loops on printing out MDS statfs information
3900 Details : The obd_ioctl_data was not initialized and in some systems
3901 this caused a failure during the ioctl that did not return
3902 an error. Initialize the struct and return an error on failure.
3905 Frequency : SLES 9 only
3907 Description: Error of copying files with lustre special EAs as root
3908 Details : Client side always return success for setxattr call for lustre
3909 special xattr (currently only "trusted.lov").
3912 Frequency : rarely on clusters with both ia64+i386 clients
3914 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3915 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3916 clients are doing concurrent writes to the same file it is
3917 possible that the ia64 clients may LASSERT because the OST
3918 extent locks are not PAGE_SIZE aligned. Ensure that grown
3919 locks are always aligned on the request boundary.
3922 Frequency : specific use, occasional
3924 Description: Overwriting in use executable truncates on-disk binary image
3925 Details : If one node attempts to overwrite an executable in use by
3926 another node, we now correctly return ETXTBSY instead of
3927 truncating the file.
3929 Severity : enhancement
3931 Description: Async OSC create to avoid the blocking unnecessarily.
3932 Details : If an OST has no remaining object, system will block on the creating
3933 when need to create a new object on this OST. Now, ways use
3934 pre-created objects when available, instead of blocking on an
3935 empty osc while others are not empty. If we must block, we block
3936 for the shortest possible period of time.
3941 Description: chmod on Lustre root is propagated to other clients
3942 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3943 invalid by the follow_mount time.
3946 Frequency : liblustre clients only
3948 Description: Race in 'instant cancel' lock handling could lead to such locks
3949 never to be granted in case of SMP MDS
3950 Details : Do not destroy not yet granted but cbpending locks in
3954 Frequency : replay/resend of open
3956 Description: non null lock assetion failure in mds_intent_policy
3957 Details : Trying to replay/resend lockless open requests resulted in
3958 mds_open() returning 0 with no lock. Now it sets a flag if
3959 a lock is going to be returned.
3961 Severity : enhancement
3963 Description: Checksum enhancements
3964 Details : New checksum enhancements allow for resending RPCs that failed
3967 Severity : enhancement
3969 Description: Tunables on number of dirty pages in cacche
3970 Details : Allow to set limit on number of dirty pages cached.
3975 Description: client crash on unmount - lock still has references
3976 Details : In some error handling cases it was possible to leak a lock
3977 reference on a client while accessing a file. This was not
3978 harmful to the client during operation, but would cause the
3979 client to crash when the filesystem is unmounted.
3982 Frequency : specific case, rare
3984 Description: ETXTBSY on mds though file not in use by client
3985 Details : ETXTBSY is no longer incorrectly returned when attempting to
3986 chmod or chown a directory that the user previously tried to
3987 execute or a currently-executing binary.
3990 Frequency : extremely rare except on liblustre-based clients
3992 Description: Lustre space not freed when files are deleted
3993 Details : Clean up open-unlinked files after client eviction. Previously
3994 the unlink was skipped and the files remained as orphans.
3999 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4000 Details : In some cases it was possible to send a blocking callback to
4001 a client doing a glimpse, even though that client didn't get
4002 a lock granted. When the glimpse lock is cancelled on the OST
4003 the freed lock is left on the waiting list and corrupted the list.
4006 Frequency : all core dumps
4008 Description: Broke core dumps to lustre
4009 Details : Negative dentry may be unhashed if parent does not have UPDATE
4010 lock, but some callers, e.g. do_coredump, expect dentry to be
4011 hashed after successful create, hash it in ll_create_it.
4013 ------------------------------------------------------------------------------
4015 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4017 * Support for kernels:
4018 2.6.9-42.0.2.EL (RHEL 4)
4019 2.6.5-7.276 (SLES 9)
4020 2.4.21-40.EL (RHEL 3)
4021 2.6.12.6 vanilla (kernel.org)
4025 Frequency : always on RHEL 3
4027 Description: Number of open files grows over time
4028 Details : The number of open files grows over time, whether or not
4029 Lustre is started. This was due to a filp leak introduced
4030 by one of our kernel patches.
4032 ------------------------------------------------------------------------------
4034 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4036 * Support for kernels:
4037 2.6.9-42.EL (RHEL 4)
4038 2.6.5-7.276 (SLES 9)
4039 2.4.21-40.EL (RHEL 3)
4040 2.6.12.6 vanilla (kernel.org)
4045 Bugzilla : 5719, 9635, 9792, 9684
4046 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4047 Details : If a server is under heavy load and cannot reply to new
4048 connection requests before the client resends the (re)connect,
4049 the connection handling code can behave badly if two service
4050 threads are concurrently handing separate (re)connections from
4051 the same client. Add better locking to the connection handling
4052 code, and ensure that only a single connection will be processed
4053 for a given client UUID, even if the lock is dropped.
4055 Severity : enhancement
4057 Description: add TCP zero-copy support to kernel
4058 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4059 sends if the hardware supports scatter-gather and checksumming.
4060 This allows socklnd to do client-write and server-read more
4061 efficiently and reduce CPU utilization from skbuf copying.
4064 Frequency : only if NFS exporting from client
4066 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4067 Details : If setting attributes on a file created under NFS that had
4068 never been opened it would be possible to oops the client
4069 if the file had no objects.
4072 Frequency : always for liblustre
4074 Description: liblustre client does MDS+OSTs setattr RPC for each write
4075 Details : When doing a write from a liblustre client, the client
4076 incorrectly issued an RPC to the MDS and each OST the file was
4077 striped over in order to update the timestamps. When writing
4078 with small chunks and many clients this could overwhelm the MDS
4079 with RPCs. In all cases it would slow down the write because
4080 these RPCs are unnecessary.
4082 Severity : enhancement
4084 Description: allow number of MDS service threads to be changed at module load
4085 Details : It is now possible to change the number of MDS service threads
4086 running. Adding "options mds mds_num_threads={N}" to the MDS's
4087 /etc/modprobe.conf will set the number of threads for the next
4088 time Lustre is restarted (assuming the "mds" module is also
4089 reloaded at that time). The default number of threads will
4090 stay the same, 32 for most systems.
4095 Description: OST crash if filesystem is unformatted or corrupt
4096 Details : If an OST is started on a device that has never been formatted
4097 or if the filesystem is corrupt and cannot even mount then the
4098 error handling cleanup routines would dereference a NULL pointer.
4103 Description: NULL pointer deref in llap_from_page.
4104 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4105 incorrect) mapping if the page was truncated/reclaimed while it was
4106 searched for. Check for this condition and skip such pages when
4107 doing readahead. Introduce extra check to llap_from_page() to
4108 verify page->mapping->host is non-NULL (so page is not anonymous).
4111 Frequency : Sometimes when using sys_sendfile
4113 Description: "page not covered by a lock" warnings from ll_readpage
4114 Details : sendfile called ll_readpage without right page locks present.
4115 Now we introduced ll_file_sendfile that does necessary locking
4116 around call to generic_file_sendfile() much like we do in
4120 Frequency : with certain MDS communication failures at client mount time
4122 Description: NULL pointer deref after failed client mount
4123 Details : a client connection request may delayed by the network layer
4124 and not be sent until after the PTLRPC layer has timed out the
4125 request. If the client fails the mount immediately it will try
4126 to clean up before the network times out the request. Add a
4127 reference from the request import to the obd device and delay
4128 the cleanup until the network drops the request.
4131 Frequency : occasionally during client (re)connect
4133 Description: assertion failure during client (re)connect
4134 Details : processing a client connection request may be delayed by the
4135 client or server longer than the client connect timeout. This
4136 causes the client to resend the connection request. If the
4137 original connection request is replied in this interval, the
4138 client may trip an assertion failure in ptlrpc_connect_interpret()
4139 which thought it would be the only running connect process.
4142 Frequency : only with obd_echo servers and clients that are rebooted
4144 Description: kernel BUG accessing uninitialized data structure
4145 Details : When running an obd_echo server it did not start the ping_evictor
4146 thread, and when a client was evicted an uninitialized data
4147 structure was accessed. Start the ping_evictor in the RPC
4148 service startup instead of the OBD startup.
4150 Severity : enhancement
4151 Bugzilla : 10193 (patchless)
4152 Description: Remove dependency on various unexported kernel interfaces.
4153 Details : No longer need reparent_to_init, exit_mm, exit_files,
4154 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4157 Frequency : rare (only users of deprecated and unsupported LDAP config)
4159 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4160 Details : LDAP apparently contains 'lustreName' attributes instead of
4161 'name'. A simple remapping of the name is sufficient.
4164 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4166 Description: waiting_locks_callback trips kernel BUG if client is evicted
4167 Details : Running with the dump_on_timeout debug flag turned on makes
4168 it possible that the waiting_locks_callback() can try to dump
4169 the Lustre kernel debug logs from an interrupt handler. Defer
4170 this log dumping to the expired_lock_main() thread.
4172 Severity : enhancement
4174 Description: Support NFS exporting on 2.6 kernels.
4175 Details : Implement non-rawops metadata methods for NFS server to use without
4176 changing NFS server code.
4179 Frequency : very rare (synthetic metadata workload only)
4181 Description: two racing renames might cause an MDS thread to deadlock
4182 Details : Running the "racer" program may cause one MDS thread to rename
4183 a file from being the source of a rename to being the target of
4184 a rename at exactly the same time that another thread is doing
4185 so, and the second thread has already enqueued these locks after
4186 doing a lookup of the target and is trying to relock them in
4187 order. Ensure that we don't try to re-lock the same resource.
4190 Frequency : only very large systems with liblustre clients
4192 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4193 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4194 all OSTs in parallel. This allows the eviction of stale liblustre
4195 clients to proceed much faster than if they were done in series,
4196 and also offers similar improvements for other set_info RPCs.
4201 Description: excessive CPU usage during initial read phase on client
4202 Details : During the initial read phase on a client, it would agressively
4203 retry readahead on the file, consuming too much CPU and impacting
4204 performance (since 1.4.5.8). Improve the readahead algorithm
4205 to avoid this, and also improve some other common cases (read
4206 of small files in particular, where "small" is files smaller than
4207 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4212 Description: MDS crash when receiving packet with unknown intent.
4213 Details : Do not LBUG in unknown intent case, just return -EFAULT
4215 Severity : enhancement
4216 Bugzilla : 9293, 9385
4217 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4218 Details : Do not serialize getattr (non-intent version) and statfs.
4221 Frequency : occasional, when OST network is overloaded/intermittent
4223 Description: client evicted by OST after bulk IO timeout
4224 Details : If a client sends a bulk IO request (read or write) the OST
4225 may evict the client if it is unresposive to its data GET/PUT
4226 request. This is incorrect if the network is overloaded (takes
4227 too long to transfer the RPC data) or dropped the OST GET/PUT
4228 request. There is no need to evict the client at all, since
4229 the pinger and/or lock callbacks will handle this, and the
4230 client can restart the bulk request.
4233 Frequency : Always when mmapping file with no objects
4235 Description: client crashes when mmapping file with no objects
4236 Details : Check that we actually have objects in a file before doing any
4237 operations on objects in ll_vm_open, ll_vm_close and
4243 Description: Request leak when working with deleted CWD
4244 Details : Introduce advanced request refcount tracking for requests
4245 referenced from lustre intent.
4247 Severity : Enhancement
4249 Description: Cache open file handles on client.
4250 Details : MDS now will return special lock along with openhandle, if
4251 requested and client is allowed to hold openhandle, even if unused,
4252 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4253 closing files for every read/write openration.
4255 Severity : Enhancement
4257 Description: Cache open negative dentries on client when possible.
4258 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4259 negative dentries on lock revocation.
4264 Description: Remounting a client read-only wasn't possible with a zconf mount
4265 Details : It wasn't possible to remount a client read-only with llmount.
4267 Severity : enhancement
4268 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4269 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4270 MPICH is included to add Lustre-specific ADIO interfaces.
4271 This is based closely on the UFS ADIO layer and only differs
4272 in file creation, in order to allow the OST striping to be set.
4273 This is user-contributed code and not supported by CFS.
4278 Description: extended inode attributes (immutable, append-only) work improperly
4279 when 2.4 and 2.6 kernels are used on client/server or vice versa
4280 Details : Introduce kernel-independent values for these flags.
4282 Severity : enhancement
4285 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4286 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4287 granularity. Now, Fractional MB granularity is be supported,
4288 this is very useful for low memory system.
4290 Severity : enhancement
4292 Description: Getattr by fid
4293 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4294 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4295 have locks and avoids getattr rpc requests for it on every stat.
4298 Frequency : Always, for filesystems larger than 2TB
4300 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4301 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4302 filesystems larger than 2TB to be mounted with mballoc enabled.
4305 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4307 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4308 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4309 filesystems larger than 2TB is mounted then the kernel will
4310 silently corrupt the start of the filesystem. CONFIG_LBD is
4311 enabled for all CFS-supported kernels, but the possibility of
4312 this happening with a modified kernel config exists.
4314 Severity : enhancement
4316 Description: add client O_DIRECT support for 2.6 kernels
4317 Details : It is now possible to do O_DIRECT reads and writes to files
4318 in the Lustre client mountpoint on 2.6 kernel clients.
4320 Severity : enhancement
4322 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4323 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4324 parallel, not waiting for response from every OST before sending
4325 a rpc to the next OST.
4330 Description: setattr vs write race when updating file timestamps
4331 Details : Client processes that update a file timestamp into the past
4332 right after writing to the file (e.g. tar) it is possible that
4333 the updated file modification time can be reset to the current
4334 time due to a race between processing the setattr and write RPC.
4336 Severity : enhancement
4338 Description: Bring 'lfs find' closer in line with regular Linux find.
4339 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4340 -print0 options and obtains all the needed info through the lustre
4343 Severity : enhancement
4345 Description: support up to 1024 configured devices on one node
4346 Details : change obd_dev array from statically allocated to dynamically
4347 allocated structs as they are first used to reduce memory usage
4352 Description: Flush dirty partially truncated pages during truncate
4353 Details : Immediatelly flush partially truncated pages in filter_setattr,
4354 this way we completely avoid having any pages in page cache on OST
4355 and can retire ugly workarounds during writes to flush such pages.
4360 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4361 Details : protect inode from truncation within vfs_unlink() context
4362 just take a reference before calling vfs_unlink() and release it
4363 when parent's i_sem is free.
4368 Description: last_id value checked outside lock on OST caused LASSERT failure
4369 Details : If there were multiple MDS->OST object precreate requests in
4370 flight, it was possible that the OST's last object id was checked
4371 outside a lock and incorrectly tripped an assertion. Move checks
4372 inside locks, and discard old precreate requests.
4375 Frequency : always, if extents are used on OSTs
4377 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4378 Details : OSTs using the extents format would not zero the high 16 bits of
4379 the index physical block number. This is not a problem for any
4380 OST filesystems smaller than 16TB, and no kernels support ext3
4381 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4382 new/modified files) and can be fixed for existing filesystems
4383 with e2fsprogs-1.39-cfs1.
4388 Description: import connection selection may be incorrect if timer wraps
4389 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4390 import connections to be ignored if the 32-bit jiffies counter
4391 wraps. Use a 64-bit jiffies counter.
4394 Frequency : very large clusters immediately after boot
4396 Description: LNET request buffers exhausted under heavy short-term load
4397 Details : If a large number of client requests are generated on a service
4398 that has previously never seen so many requests it is possible
4399 that the request buffer growth cannot keep up with the spike in
4400 demand. Instead of dropping incoming requests, they are held in
4401 the LND until the RPC service can accept more requests.
4404 Frequency : Sometimes during replay
4406 Description: Assertion failure in ll_local_open after replay.
4407 Details : If replay happened on an open request reply before we were able
4408 to set replay handler, reply will become not swabbed tripping the
4409 assertion in ll_local_open. Now we set the handler right after
4410 recognising of open request
4413 Frequency : very rare
4415 Description: kernel reports "badness in vsnprintf"
4416 Details : Reading from the "recovery_status" /proc file in small chunks
4417 may cause a negative length in lprocfs_obd_rd_recovery_status()
4418 call to vsnprintf() (which is otherwise harmless). Exit early
4419 if there is no more space in the output buffer.
4421 Severity : enhancement
4423 Description: clear OBD RPC statistics by writing to them
4424 Details : It is now possible to clear the OBD RPC statistics by writing
4425 to the "stats" file.
4430 Description: Client mtime is not the same on different clients after utimes
4431 Details : In some cases, the client was using the utimes() syscall on
4432 a file cached on another node. The clients now validate the
4433 ctime from the MDS + OSTs to determine which one is right.
4438 Description: Inability to activate failout mode
4439 Details : lconf script incorrectly assumed that in pythong string's numeric
4440 value is used in comparisons.
4443 Frequency : always with multiple stripes per file
4445 Description: Inefficient object allocation for mutli-stripe files
4446 Details : When selecting which OSTs to stripe files over, for files with
4447 a stripe count that divides evenly into the number of OSTs,
4448 the MDS is always picking the same starting OST for each file.
4449 Return the OST selection heuristic to the original design.
4454 Description: mount failures may take full timeout to return an error
4455 Details : Under some heavy load conditions it is possible that a
4456 failed mount can wait for the full obd_timeout interval,
4457 possibly several minutes, before reporting an error.
4458 Instead return an error as soon as the status is known.
4460 Frequency : quota enabled and large files being deleted
4462 Description: releasing more than 4GB of quota at once hangs OST
4463 Details : If a user deletes more than 4GB of files on a single OST it
4464 will cause the OST to spin in an infinite loop. Release
4465 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4470 Description: statfs data retrieved from /proc may be stale or zero
4471 Details : When reading per-device statfs data from /proc, in the
4472 {kbytes,files}_{total,free,avail} files, it may appear
4473 as zero or be out of date.
4476 Frequency : systems with MD RAID1 external journal devices
4478 Description: lconf's call to blkid is confused by RAID1 journal devices
4479 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4480 of returning all block devices that match the journal UUID.
4483 Frequency : always, for aggregate stripe size over 4GB
4485 Description: assertion fails when trying to use 4GB stripe size
4486 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4487 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4490 Frequency : always on ppc64
4492 Description: the first write on an ext3 filesystem with mballoc got stuck
4493 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4494 perform endianness conversion.
4496 ------------------------------------------------------------------------------
4498 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4500 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4501 INTEROPERATE with older versions automatically. Please read the
4502 user documentation before upgrading any part of a live system.
4503 * WARNING: Lustre networking configuration changes are required with
4504 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4507 * Support for kernels:
4508 2.6.9-22.0.2.EL (RHEL 4)
4509 2.6.5-7.244 (SLES 9)
4510 2.6.12.6 vanilla (kernel.org)
4513 Severity : enhancement
4514 Bugzilla : 7981/8208
4515 Description: Introduced Lustre Networking (LNET)
4516 Details : LNET is new networking infrastructure for Lustre, it includes
4517 a reorganized network configuration mode (see the user
4518 documentation for full details) as well as support for routing
4519 between different network fabrics. Lustre Networking Devices
4520 (LNDs) for the supported network fabrics have also been
4521 created for this new infrastructure.
4523 Severity : enhancement
4524 Description: Introduced Access control lists
4525 Details : clients can set ACLs on files and directories in order to have
4526 more fine-grained permissions than the standard Unix UGO+RWX.
4527 The MDS must be started with the "-o acl" mount option.
4529 Severity : enhancement
4530 Description: Introduced filesystem quotas
4531 Details : Administrators may now establish per-user quotas on the
4534 Severity : enhancement
4536 Description: Configuration change for the XT3
4537 The PTLLND is now used to run Lustre over Portals on the XT3
4538 The configure option(s) --with-cray-portals are no longer used.
4539 Rather --with-portals=<path-to-portals-includes> is used to
4540 enable building on the XT3. In addition to enable XT3 specific
4541 features the option --enable-cray-xt3 must be used.
4546 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4547 Details : A race between allocating a new llcd and re-getting the llcd_lock
4548 allowed another thread to grab newly-allocated llcd.
4550 Severity : enhancement
4552 Description: 2.6 OST async journal commit and locking fix to improve performance
4553 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4554 2.6 kernels are now async as they already were in 2.4 kernels so
4555 that they can commit concurrently with the network bulk transfer.
4556 For block-allocated files the filter allocation semaphore is held
4557 to avoid filesystem fragmentation during allocation. BKL lock
4558 removed for 2.6 xattr operations where it is no longer needed.
4563 Description: lconf incorrectly determined whether two IP networks could talk
4564 Details : In some more complicated routing and multiple-network
4565 configurations, lconf will avoid trying to make a network
4566 connection to a disjoint part of the IP space. It was doing the
4567 math incorrectly for one set of cases.
4572 Description: Fix for potential infinite loop processing records in an llog.
4573 Details : If an llog record is corrupted/zeroed, it is possible to loop
4574 forever in llog_process(). Validate the llog record length
4575 and skip the remainder of the block on error.
4578 Frequency : occasional (liblustre only)
4580 Description: liblustre could not open files whose last component is a symlink
4581 Details : sysio_path_walk() would incorrectly pass the open intent to
4582 intermediate path components.
4585 Frequency : rare (liblustre only with non-standard tuning)
4586 Bugzilla : 7201 (7350)
4587 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4588 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4589 after it has been released, only do this for referenced locks
4591 Severity : enhancement
4593 Description: specifying an (invalid) directory default stripe_size of -1
4594 would reset the directory default striping
4595 Details : stripe_size -1 was used internally to signal directory stripe
4596 removal, now use "all default" to signal dir stripe removal
4597 as a directory striping of "all default" is not useful
4600 Frequency : common for large clusters running liblustre clients
4602 Description: doing an ls when liblustre clients are running is slow
4603 Details : sending a glimpse AST to a liblustre client waits for every AST
4604 to time out, as liblustre clients will not respond. Since they
4605 cannot cache data we refresh the OST lock LVB from disk instead.
4607 Severity : enhancement
4609 Description: doing an ls at the same time as file IO can be slow
4610 Details : enqueue and other "small" requests can be blocked behind many
4611 large IO requests. Create a new OST IO portal for non-IO
4612 requests so they can be processed faster.
4615 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4617 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4618 created file with mode 000
4619 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4620 allow an ftruncate() as a non-root user to a file with mode 000.
4621 Lustre now respects this flag to disable mode checks when
4622 truncating a file owned by the user
4625 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4628 Description: Revoking locks from clients that went dead or catatonic might take
4630 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4631 cancellation of such locks instant on servers without waiting for
4632 any reply from clients. Clients drops these locks when cancel
4633 notification from server is received without replying.
4636 Frequency : liblustre-only, when liblustre client dies or becomes busy
4638 Description: Doing ls on Linux clients can take a long time with active
4640 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4641 granting such locks to it in the first place if possible. Locks
4642 are taken by proxy on the OST during the read or write and
4643 dropped immediately afterward. Add connect flags handling, do
4644 not grant locks to liblustre clients for glimpse ASTs.
4646 Severity : enhancement
4648 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4649 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4650 sequential reads and behaves badly with applications doing random
4651 reads. Improve it by reading ahead at least read region, and
4652 avoiding excessive large RPC for small reads.
4654 Severity : enhancement
4656 Description: Creating more than 1000 files for a single job may cause a load
4657 imbalance on the OSTs if there are also a large number of OSTs.
4658 Details : qos_prep_create() uses an OST index reseed value that is an
4659 even multiple of the number of available OSTs so that if the
4660 reseed happens in the middle of the object allocation it will
4661 still utilize the OSTs as uniformly as possible.
4666 Description: OST or MDS may oops in ping_evictor_main()
4667 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4668 but doesn't restart at beginning of obd_exports_timed list
4671 Severity : enhancement
4673 Description: improve by-nid export eviction on the MDS and OST
4674 Details : allow multiple exports with the same NID to be evicted at one
4675 time without re-searching the exports list.
4678 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4680 Description: MDS may oops in groups_free()
4681 Details : in rare race conditions a newly allocated group_info struct is
4682 freed again, and this can be NULL. The 2.4 compatibility code
4683 for groups_free() checked for a NULL pointer, but 2.6 did not.
4686 Frequency : common for liblustre clients doing little filesystem IO
4687 Bugzilla : 9352, 7313
4688 Description: server may evict liblustre clients accessing contended locks
4689 Details : if a client is granted a lock or receives a completion AST
4690 with a blocking AST already set it would not reply to the AST
4691 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4694 Frequency : lfs setstripe, only systems with more than 160 OSTs
4696 Description: unable to set striping with a starting offset beyond OST 160
4697 Details : llapi_create_file() incorrectly limited the starting stripe
4698 index to the maximum single-file stripe count.
4701 Frequency : LDAP users only
4703 Description: lconf did not handle in-kernel recovery with LDAP properly
4704 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4706 Severity : enhancement
4708 Description: bind OST threads to NUMA nodes to improve performance
4709 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4710 node and do their allocations there to localize memory access
4712 Severity : enhancement
4714 Description: llmount can determine client NID directly from Myrinet (GM)
4715 Details : the client NID code from gmnalnid was moved directly into
4716 llmount, removing the need to use this or specifying the
4717 client NID explicitly when mounting GM clients with zeroconf
4720 Frequency : if client is started with down MDS
4722 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4723 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4724 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4725 ptlrpc_import_delayed_req() because we didn't check if the
4726 request was interrupted, and we also didn't break out of the
4727 event loop if there was no timeout
4732 Description: data loss during non-page-aligned writes to a single file from
4733 both multiple nodes and multiple threads on one node at same time
4734 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4735 inconsistency led to false short-reads, that were cached and later
4736 used by ->prepare_write() to fill in partially written page,
4737 leading to data loss.
4740 Frequency : always, if lconf --abort_recovery used
4742 Description: lconf --abort_recovery fails with 'Operation not supported'
4743 Details : lconf was attempting to abort recovery on the MDT device and not
4746 Severity : enhancement
4748 Description: remove cleanup logs
4749 Details : replace lconf-generated cleanup logs with lustre internal
4750 cleanup routines. Eliminates the need for client-cleanup and
4753 Severity : enhancement
4755 Description: add support for EAs (user and system) on lustre filesystems
4756 Details : it is now possible to store extended attributes in the Lustre
4757 client filesystem, and with the user_xattr mount option it
4758 is possible to allow users to store EAs on their files also
4760 Severity : enhancement
4762 Description: Add possibility (config option) to show minimal available OST free
4764 Details : When compiled with --enable-mindf configure option, statfs(2)
4765 (and so, df) will return least minimal free space available from
4766 all OSTs as amount of free space on FS, instead of summary of
4767 free spaces of all OSTs.
4769 Severity : enhancement
4771 Description: do not expand extent locks acquired on OST-side
4772 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4773 by server: they are not cached anyway.
4776 Frequency : when mmap is used/binaries executed from Lustre
4778 Description: Unmmap pages before throwing them away from read cache.
4779 Details : llap_shrink cache now attempts to unmap pages before discarding
4780 them (if unmapping failed - do not discard). SLES9 kernel has
4781 extra checks that trigger if this unmapping is not done first.
4786 Description: lconf didn't resolve symlinks before checking to see whether a
4787 given mountpoint was already in use
4790 Frequency : when migrating failover services
4791 Bugzilla : 6395, 9514
4792 Description: When migrating a subset of services from a node (e.g. failback
4793 from a failover service node) the remaining services would
4794 time out and evict clients.
4795 Details : lconf --force (implied by --failover) sets the global obd_timeout
4796 to 5 seconds in order to quickly disconnect, but this caused
4797 other RPCs to time out too quickly. Do not change the global
4798 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4800 Severity : enhancement
4801 Frequency : if MDS is started with down OST
4802 Bugzilla : 9439,5706
4803 Description: Allow startup/shutdown of an MDS without depending on the
4804 availability of the OSTs.
4805 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4806 Add appropriate locking and lov-osc refcounts for safe
4807 cleaning. Add osc abort_inflight calls in case the
4808 synchronize never started.
4811 Frequency : occasional (Cray XT3 only)
4813 Description: root not authorized to access files in CRAY_PORTALS environment
4814 Details : The client process capabilities were not honoured on the MDS in
4815 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4816 been accessed by an authorized user then root was able to access
4817 the file on the local client also. The root user capabilities
4818 are now allowed on the MDS, as this environment has secure UID.
4821 Frequency : occasional
4823 Description: ldiskfs "too long searching" message happens too often
4824 Details : A debugging message (otherwise harmless) prints too often on
4825 the OST console. This has been reduced to only happen when
4826 there are fragmentation problems on the filesystem.
4831 Description: Division by zero in statfs when all OSCs are inactive
4832 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4833 lov_statfs divides by value returned by lov_get_stripecnt().
4837 Bugzilla : 9489, 3273
4838 Description: First write from each client to each OST was only 4kB in size,
4839 to initialize client writeback cache, which caused sub-optimal
4840 RPCs and poor layout on disk for the first written file.
4841 Details : Clients now request an initial cache grant at (re)connect time
4842 and so that they can start streaming writes to the cache right
4843 away and always do full-sized RPCs if there is enough data.
4844 If the OST is rebooted the client also re-establishes its grant
4845 so that client cached writes will be honoured under the grant.
4850 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4851 Details : Now I/O RPCs go to different portal number and (presumably) fast
4852 lock requests (and glimses) and other RPCs get their own service
4853 threads pool that should be able to service those RPCs
4856 Severity : enhancement
4858 Description: Ability to exchange lustre version between client and servers and
4859 issue warnings at client side if client is too old. Also for
4860 liblustre clients there is ability to refuse connection of too old
4862 Details : New 'version' field is added to connect data structure that is
4863 filled with version info. That info is later checked by server and
4867 Frequency : rare, liblustre only.
4868 Bugzilla : 9296, 9581
4869 Description: Two simultaneous writes from liblustre at offset within same page
4870 might proceed at the same time overwriting eachother with stale
4872 Details : I/O lock withing llu_file_prwv was released too early, before data
4873 actually was hitting the wire. Extended lock-holding time until
4874 server acknowledges receiving data.
4877 Frequency : extremely rare. Never observed in practice.
4879 Description: avoid generating lustre_handle cookie of 0.
4880 Details : class_handle_hash() generates handle cookies by incrementing
4881 global counter, and can hit 0 occasionaly (this is unlikely, but
4882 not impossible, because initial value of cookie counter is
4883 selected randonly). Value of 0 is used as a sentinel meaning
4884 "unassigned handle" --- avoid it. Also coalesce two critical
4885 sections in this function into one.
4887 Severity : enhancement
4889 Description: allow liblustre clients to delegate truncate locking to OST
4890 Details : To avoid overhead of locking, liblustre client instructs OST to
4891 take extent lock in ost_punch() on client's behalf. New connection
4892 flag is added to handle backward compatibility.
4894 Severity : enhancement
4895 Bugzilla : 4928, 7341, 9758
4896 Description: allow number of OST service threads to be specified
4897 Details : a module parameter allows the number of OST service threads
4898 to be specified via "options ost ost_num_threads={N}" in the
4899 OSS's /etc/modules.conf or /etc/modprobe.conf.
4903 Bugzilla : 6146, 9635, 9895
4904 Description: servers crash with bad pointer in target_handle_connect()
4905 Details : In rare cases when a client is reconnecting it was possible that
4906 the connection request was the last reference for that export.
4907 We would temporarily drop the export reference and get a new
4908 one, but this may have been the last reference and the export
4909 was just destroyed. Get new reference before dropping old one.
4911 Severity : enhancement
4912 Frequency : if client is started with failover MDS
4914 Description: Allow multiple MDS hostnames in the mount command
4915 Details : Try to read the configuration from all specified MDS
4916 hostnames during a client mount in case the "primary"
4919 Severity : enhancement
4921 Description: Stop sending data to evicted clients as soon as possible.
4922 Details : Check if the client we are about to send or are sending data to
4923 was evicted already. (Check is done every second of waiting,
4924 for which l_wait_event interface was extended to allow checking
4925 of exit condition at specified intervals).
4928 Frequency : rare, normally only when NFS exporting is done from client
4930 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4932 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4935 Frequency : always, if srand() is called before liblustre initialization
4937 Description: Liblustre uses system PRNG disturbing its usage by user application
4938 Details : Introduce internal to lustre fast and high-quality PRNG for
4939 lustre usage and make liblustre and some other places in generic
4940 lustre code to use it.
4942 Severity : enhancement
4943 Bugzilla : 9477, 9557, 9870
4944 Description: Verify that the MDS configuration logs are updated when xml is
4945 Details : Check if the .xml configuration logs are newer than the config
4946 logs stored on the MDS and report an error if this is the case.
4947 Request --write-conf, or allow starting with --old_conf.
4949 Severity : enhancement
4951 Description: Handle symlinks in the path when checking if Lustre is mounted.
4952 Details : Resolve intermediate symlinks when checking if a client has
4953 mounted a filesystem to avoid duplicate client mounts.
4958 Description: lconf can hit an error exception but still return success.
4959 Details : The lconf command catches the Command error exception at the top
4960 level script context and will exit with the associated exit
4961 status, but doesn't ensure that this exit status is non-zero.
4966 Description: failure of ptlrpc thread startup can cause oops
4967 Details : Starting a ptlrpc service thread can fail if there are a large
4968 number of threads or the server memory is very fragmented.
4969 Handle this without oopsing.
4972 Frequency : always, only if liblustre and non-default acceptor port was used
4974 Description: liblustre cannot connect to servers with non-default acceptor port
4975 Details : tcpnal_set_default_params() was not called and was therefore
4976 ignoring the environment varaible TCPNAL_PORT, as well as other
4977 TCPNAL_ environment variables
4982 Description: two objects could be created on the same OST for a single file
4983 Details : If an OST is down, in some cases it was possible to create two
4984 objects on a single OST for a single file. No problems other
4985 than potential performance impact and spurious error messages.
4989 Bugzilla : 5681, 9562
4990 Description: Client may oops in ll_unhash_aliases
4991 Details : Client dcache may become inconsistent in race condition.
4992 In some cases "getcwd" can fail if the current directory is
4998 Description: Inode refcounting problems in NFS export code
4999 Details : link_raw functions used to call d_instantiate without obtaining
5000 extra inode reference first.
5004 Bugzilla : 9942, 9903
5005 Description: Referencing freed requests leading to crash, memleaks with NFS.
5006 Details : We used to require that call to ll_revalidate_it was always
5007 followed by ll_lookup_it. Also with revalidate_special() it is
5008 possible to call ll_revalidate_it() twice for the same dentry
5009 even if first occurence returned success. This fix changes semantic
5010 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5011 reference on a request referred from the intent.
5012 ll_intent_release() then releases such a request.
5015 Frequency : rare, normally benchmark loads only
5017 Description: unlinked inodes were kept in memory on the client
5018 Details : If a client is repeatedly creating and unlinking files it
5019 can accumulate a lot of stale inodes in the inode slab cache.
5020 If there is no other client load running this can cause the
5021 client node to run out of memory. Instead flush old inodes
5022 from client cache that have the same inode number as a new inode.
5025 Frequency : SLES9 2.6.5 kernel and long filenames only
5026 Bugzilla : 9969, 10379
5027 Description: utime reports stale NFS file handle
5028 Details : SLES9 uses out-of-dentry names in some cases, which confused
5029 the lustre dentry revalidation. Change it to always use the
5033 Frequency : rare, unless heavy write-truncate concurrency is continuous
5034 Bugzilla : 4180, 6984, 7171, 9963, 9331
5035 Description: OST becomes very slow and/or deadlocked during object unlink
5036 Details : filter_destroy() was holding onto the parent directory lock
5037 while truncating+unlinking objects. For very large objects this
5038 may block other threads for a long time and slow overall OST
5039 responsiveness. It may also be possible to get a lock ordering
5040 deadlock in this case, or run out of journal credits because of
5041 the combined truncate+unlink. Solution is to do object truncate
5042 first in one transaction without parent lock, and then do the
5043 final unlink in a new transaction with the parent lock. This
5044 reduces the lock hold time dramatically.
5047 Frequency : rare, 2.4 kernels only
5049 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5050 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5051 before unmounting their filesystem in order to avoid deadlock.
5052 The kernel_locked() function in 2.4 kernels only checks whether
5053 the kernel lock is held, not whether it is this process that is
5054 holding it as 2.6 kernels do.
5059 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5060 Details : The client ptlrpc code may be trying to reconnect to a down
5061 server before a previous connection attempt has timed out.
5062 Increase the reconnect interval to be longer than the connection
5063 timeout interval to avoid sending duplicate connections to
5067 Frequency : echo_client brw_test command
5069 Description: fix echo_client to work with OST preallocated code
5070 Details : OST preallocation code (5137) didn't take echo_client IO path
5071 into account: echo_client calls filter methods outside of any
5072 OST thread and, hence, there is no per-thread preallocated
5073 pages and buffers to use. Solution: hijack pga pages for IO. As
5074 a byproduct, this avoids unnecessary data copying.
5078 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5079 Description: Client can oops in mdc_commit_close() after open replay
5080 Details : It was possible for the MDS to return an open request with no
5081 transaction number in mds_finish_transno() if the client was
5082 evicted, but without actually returning an error. Clients
5083 would later try to replay that open and may trip an assertion
5084 Simplify the client close codepath, and always return an error
5085 from the MDS in case the open is not successful.
5088 Frequency : rare, 2.6 OSTs only
5090 Description: OST may deadlock under high load on fragmented files
5091 Details : If there was a heavy load and highly-fragmented OST filesystems
5092 it was possible to have all the OST threads deadlock waiting on
5093 allocation of biovecs, because the biovecs were not released
5094 until the entire RPC IO was completed. Instead, release biovecs
5095 as soon as they are complete to ensure forward IO progress.
5097 Severity : enhancement
5099 Description: Support for specifying external journal device at mount
5100 Details : If an OST or MDS device is formatted with an external journal
5101 device, this device major/minor is stored in the ext3 superblock
5102 and may not be valid for failover. Allow detecting and
5103 specifying the external journal at mount time.
5108 Description: Mounting an MDS with pending unlinked files may cause oops
5109 Details : target_finish_recovery() calls mds_postrecov() which returned
5110 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5111 considers this an error and immediately begins cleaning up the
5112 lov, just after starting the mds_lov process
5114 Severity : enhancement
5116 Description: Implement 'lfs df' to report actual free space on per-OST basis
5117 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5118 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5119 usage of objects; '-h' to report in human readable format.
5121 ------------------------------------------------------------------------------
5123 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5130 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5131 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5132 at startup during orphan unlinking. Instead, skip buddy update
5133 before setup, as it will be regenerated after recovery is complete.
5138 Description: If an OST is inactive, its locks might reference stale inodes.
5139 Details : lov_change_cbdata() must iterate over all namespaces, even if
5140 they are inactive to clear inode references from the lock.
5142 Severity : enhancement
5143 Frequency : occasional, if non-standard max_dirty_mb used
5145 Description: Client will block write RPCs if not enough grant
5146 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5147 then the client will block writes while waiting for another RPC
5148 to complete instead of consuming its dirty limit. With change
5149 we get improved performance when max_dirty_mb is small.
5151 Severity : enhancement
5152 Bugzilla : 3389, 6253
5153 Description: Add support for supplementary groups on the MDS.
5154 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5155 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5156 lookups for user supplementary groups into a cache.
5160 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5161 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5162 for liblustre O_EXCL is enforced by MDS.
5166 Description: readdir never returns NULL in liblustre.
5167 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5168 dentry in struct dirent.
5172 Frequency : liblustre only, depends on application IO pattern
5173 Description: liblustre clients evicted if not contacting servers
5174 Details : Don't put liblustre clients into the ping_evictor list, so
5175 they will not be evicted by the pinger ever.
5177 Severity : enhancement
5179 Description: Add ability to evict clients by NID from MDS.
5180 Details : By echoing "nid:$NID" string into
5181 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5182 $NID would be instantly evicted from this MDS and from all active
5183 OSTs connected to it.
5187 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5188 Details : lookup_it_finish() used to query file size from OSTs that was not
5193 Description: service threads change working directory to that of init
5194 Details : Starting lustre service threads may pin the working directory
5195 of the parent thread, making that filesystem busy. Threads
5196 now change to the working directory of init to avoid this.
5200 Frequency : during shutdown only
5201 Description: shutdown with a failed MDS or OST can cause unmount to hang
5202 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5207 Frequency : 2.6 only
5208 Description: chmod/chown may include an extra supplementary group
5209 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5210 supplementary group and if none is specified this is used.
5213 Bugzilla : 5479 (6816)
5215 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5216 Details : If lookup is in progress on a file that is unlinked we might try
5217 to revalidate the inode and fail in revalidate after lookup is
5218 complete and ll_file_open() enqueues the open again but
5219 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5222 Frequency : always, if lconf --abort_recovery used
5224 Description: lconf --abort_recovery fails with 'Operation not supported'
5225 Details : lconf was attempting to abort recovery on the MDT device and not
5228 ------------------------------------------------------------------------------
5230 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5235 Frequency : rare (only unsupported configurations with a node running as an
5237 Bugzilla : 6514, 5137
5238 Description: Mounting a Lustre file system on a node running as an OST could
5240 Details : OSTs now preallocates memory needed to write out data at
5241 startup, instead of when needed, to avoid having to
5242 allocate memory in possibly low memory situations.
5243 Specifically, if the file system is mounted on on OST,
5244 memory pressure could force it to try to write out data,
5245 which it needed to allocate memory to do. Due to the low
5246 memory, it would be unable to do so and the node would
5247 become unresponsive.
5249 Severity : enhancement
5251 Description: Addition of lconf --service command line option
5252 Details : lconf now accepts a '--service <arg>' option, which is
5253 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5255 Severity : enhancement
5257 Description: Failover mode is now the default for OSTs.
5258 Details : By default, OSTs will now run in failover mode. To return to
5259 the old behaviour, add '--failout' to the lmc line for OSTs.
5261 Severity : enhancement
5263 Description: Health checks are now provided for MDS and OSTs
5264 Details : Additional detailed health check information on MSD and OSTs
5265 is now provided through the procfs health_check value.
5268 Frequency : occasional, depends on IO load
5270 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5271 after numerous create/delete cycles
5272 Details : The ext3 inode allocation policy would not allocate new inodes
5273 very well on the OSTs because there are no new directories
5274 being created. Instead we look for groups with free space if
5275 the parent directories are nearly full.
5280 Description: Network or server problems during mount may cause partially
5281 mounted clients instead of returning an error.
5282 Details : The config llog parsing code may overwrite the error return
5283 code during mount error handling, returning success instead
5289 Description: MDS can fail to allocate large reply buffers
5290 Details : After long uptimes the MDS can fail to allocate large reply
5291 buffers (e.g. zconf client mount config records) due to memory
5292 fragmentation or consumption by the buffer cache. Preallocate
5293 some large reply buffers so that these replies can be sent even
5294 under memory pressure.
5298 Frequency : rare (liblustre)
5299 Description: fsx running with liblustre complained that using truncate() to
5300 extend the file doesn't work. This patch corrects that issue.
5301 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5302 ATTR_SIZE and lsm use in llu_setattr_raw.
5306 Frequency : rare, only 2.6 kernels
5307 Description: Unusual file access patterns on the MDS may result in inode
5308 data being lost in very rare circumstances.
5309 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5310 code in the 2.6 kernel caused the inode and/or EA data not to
5311 be read from disk, causing single-file corruption.
5315 Frequency : rare, only 2.6 filesystems using extents
5316 Description: Heavy concurrent write and delete load may cause data corruption.
5317 Details : It was possible under high-load situations to have an extent
5318 metadata block in the block device cache from a just-unlinked
5319 file overwrite a newly-allocated data block. We now unmap any
5320 metadata buffers that alias just-allocated data blocks.
5324 Frequency : filesystems with default stripe_count larger than 77
5325 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5326 Details : lconf specifies an inode size of 4096 bytes when the default
5327 stripe_count is larger than 77. This conflicts with the default
5328 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5329 this case to avoid pinning too much memory for large EAs.
5331 ------------------------------------------------------------------------------
5333 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5338 Frequency : rare (extremely heavy IO load with hundreds of clients)
5340 Description: Client is evicted, gets IO error writing to file
5341 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5342 caused clients to be evicted for AST timeouts. The fixes in
5343 bug 5192 mean we no longer need to have such short AST timeouts
5344 so ldlm_timeout has been increased.
5347 Frequency : occasional during --force or --failover shutdown under load
5348 Bugzilla : 5949, 4834
5349 Description: Server oops/LBUG if stopped with --force or --failover under load
5350 Details : a collection of import/export refcount and cleanup ordering
5351 issues fixed for safer force cleanup
5354 Frequency : only filesystems larger than 120 OSTs
5355 Bugzilla : 5990, 6223
5356 Description: lfs getstripe would oops on a very large filesystem
5357 Details : lov_getconfig used kfree on vmalloc'd memory
5360 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5361 Bugzilla : 6242, 6243
5362 Description: reading from files that had been truncated to a non-zero size
5363 but never opened returned no data
5364 Details : ll_file_read() reads zeros from no-object files to EOF
5369 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5370 Details : There's little harm in aborting MDS/OSS recovery and letting it
5371 try again, so I removed the LASSERT and return an error instead.
5373 Severity : enhancement
5375 Description: New debugging infrastructure for tracking down data corruption
5376 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5377 (b) cover more of the client-side code path, and (c) try to narrow
5378 down where problems occurred
5382 Bugzilla : 3819, 4364, 4397, 6313
5383 Description: Racing close and eviction MDS could cause assertion in mds_close
5384 Details : It was possible to get multiple mfd references during close and
5385 client eviction, leading to one thread referencing a freed mfd.
5387 Severity: : enhancement
5388 Bugzilla : 3262, 6359
5389 Description: Attempts to reconnect to servers are now more aggressive.
5390 Details : This builds on the enhanced upcall-less recovery that was added
5391 in 1.4.2. When trying to reconnect to servers, clients will
5392 now try each server in the failover group every 10 seconds. By
5393 default, clients would previously try one server every 25 seconds.
5398 Description: After recovery, certain operations trigger a failed
5399 assertion on a client.
5400 Details : Failing over an mds, using lconf -d --failover, while a
5401 client was doing a readdir() call would cause the client to
5402 LBUG after recovery completed and the readdir() was resent.
5404 Severity : enhancement
5406 Description: Default groups are now added by lconf
5407 Details : You can now run lconf --group <servicename> without having to
5408 manually add groups with lmc.
5411 Frequency : occasional
5413 Description: Nodes with an elan id of 0 trigger a failed assertion
5416 Frequency : always when accessing e.g. tty/console device nodes
5418 Description: tty and some other devices nodes cannot be used on lustre
5419 Details : file's private_data field is used by device data and lustre
5420 values in there got lost. New field was added to struct file to
5421 store fs-specific private data.
5424 Frequency : when exporting Lustre via NFS
5426 Description: NFSD failed occasionally when looking up a path component
5427 Details : NFSD is looking up ".." which was broken in ext3 directories
5428 that had grown large enough to become hashed.
5431 Frequency : Clusters with multiple interfaces not on the same subnet
5433 Description: Nodes will repeatedly try to reconnect to an interface which it
5434 cannot reach and report an error to the log.
5435 Details : Extra peer list entries will be created by lconf with some peers
5436 unreachable. lconf now validates the peer before adding it.
5439 Frequency : Only if a default stripe is set on the filesystem root.
5441 Description: Setting a default stripe on the filesystem root prevented the
5442 filesystem from being remounted.
5443 Details : The client was sending extra request flags in the root getattr
5444 request and did not allocate a reply buffer for the dir EA.
5447 Frequency : occasional, higher if lots of files are accessed by one client
5448 Bugzilla : 6159, 6097
5449 Description: Client trips assertion regarding lsm mismatch/magic
5450 Details : While revalidating inodes the VFS looks up inodes with ifind()
5451 and in rare cases can find an inode that is being freed.
5452 The ll_test_inode() code will free the lsm during ifind()
5453 when it finds an existing inode and then the VFS later attaches
5454 this free lsm to a new inode.
5458 Bugzilla : 6422, 7030
5459 Description: MDS deadlock between mkdir and client eviction
5460 Details : Creating a new file via mkdir or mknod (starting a transaction
5461 and getting the ns lock) can deadlock with client eviction
5462 (gets ns lock and trying to finish a synchronous transaction).
5465 Frequency : occasional
5466 Description: While starting a server, the fsfilt_ext3 module could not be
5468 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5469 kernels. Previously, lconf would still use the ext3 name
5470 when trying to load modules. Now, it will correctly use
5471 ext3 on 2.4 and ldiskfs on 2.6.
5473 Severity : enhancement
5474 Description: The default stripe count has been changed to 1
5475 Details : The interpretation of the default stripe count (0, to lfs
5476 or lmc) has been changed to mean striping across a single
5477 OST, rather than all available. For general usage we have
5478 found a stripe count of 1 or 2 works best.
5480 Severity : enhancement
5481 Description: Add support for compiling against Cray portals.
5482 Details : Conditional compiling for some areas that are different
5486 Frequency : occasional
5487 Bugzilla : 6409, 6834
5488 Description: Creating files with an explicit stripe count may lead to
5489 a failed assertion on the MDS
5490 Details : If some OSTs are full or unavailable, creating files may
5491 trigger a failed assertion on the MDS. Now, Lustre will
5492 try to use other servers or return an error to the
5496 Frequency : occasional
5498 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5499 caused lock timeout during readahead (since 1.4.2).
5500 Details : Processes doing readahead might match a lock that hasn't been
5501 granted yet if there are overlapping and conflicting lock
5502 requests. The readahead process waits on ungranted lock
5503 (original lock is CBPENDING), while OST waits for that process
5504 to cancel CBPENDING read lock and eventually evicts client.
5506 Severity : enhancement
5508 Description: Initial enabling of flock support for clients
5509 Details : Implements fcntl advisory locking and file status functions.
5510 This feature is provided as an optional mount flag (default
5511 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5512 locking are implemented yet, and those that are are not guaranteed
5513 to be completely correct in production environments.
5514 mount -t lustre -o [flock|noflock] ...
5517 Frequency : occasional
5519 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5520 assertion in the ext3 JBD (journaling) layer.
5521 Details : The b_committed_data struct is protected by the big kernel lock
5522 in 2.4 kernels, serializing journal_commit_transaction() and
5523 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5524 access to this struct. In 2.6 kernels there is finer grained
5525 locking to improve SMP performance of the JBD layer.
5529 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5534 Description: merge and cleanup kernel patches.
5537 -----------------------------------------------------------------------------
5539 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5541 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5542 versions of Lustre. Please update all servers and clients to
5543 version 1.4.2 or later at the same time. You must also run
5544 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5545 to update the configuration logs.
5547 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5548 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5549 - swab llog records of type '0' so we get proper header size/idx (5861)
5550 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5551 - fix rename of one directory over another leaking an inode (5953)
5552 - avoid SetPageDirty on 2.6 (5981)
5553 - don't re-add just-being-destroyed locks to the waiting list (5653)
5554 - when creating new directories, inherit the parent's custom
5555 striping settings if present parent (3048)
5556 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5557 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5558 - don't hold client locks on temporary worklist from l_lru (5666)
5559 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5560 - automatically evict dead clients (5921)
5561 - Update file size properly in create+truncate+fstat case (6196)
5562 - Do not unhash mountpoint dentries, do not allow removal of
5564 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5565 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5566 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5567 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5568 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5569 - Correctly handle reads of files with no objects (6243)
5570 - lctl recover will also mark a device active if deactivate used (5933)
5572 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5573 - allow --write-conf on an MDS with different nettype than client (5619)
5574 - don't write config llogs to MDS for mounts not from that MDS (5617)
5575 - lconf should create multiple TCP connections from a client (5201)
5576 - init scripts are now turned off by default; run chkconfig --on
5577 lustre and chkconfig --on lustrefs to use them
5578 - upcalls are no longer needed for clients to recover to failover
5580 - add --abort-recovery option to lconf to abort recovery on device
5582 - add support for an arbitrary number of OSTs (3026)
5583 - Quota support protocol changes.
5584 - forward compatibility changes to wire structs (6007)
5585 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5586 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5587 - improved "lustre status" script
5588 - initialize blocksize for non-regular files (6062)
5589 - added --disable-server and --disable-client configure options (5782)
5590 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5591 - Vanilla 2.4.29 support
5592 - increase maximum number of obd devices to 520 (6242)
5593 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5594 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5595 SLES 9 2.6.5 kernels
5596 - limit stripes per file to 160 (the maximum EA size) (6093)
5598 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5601 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5602 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5603 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5604 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5605 - small ext3 extents cleanups and fixes (5733)
5606 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5607 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5608 - avoid needless client->OST connect, fix handle mismatch (5317)
5609 - fix DLM error path that led to out-of-sync client, long delays (5779)
5610 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5611 - fix several locking issues related to i_size (5492,5624,5654,5672)
5612 - don't move pending lock onto export if it is already evicted (5683)
5613 - fix kernel oops when creating .foo in unlinked directory (5548)
5614 - fix deadlock in obdfilter statistics vs. object create (5811)
5615 - use time_{before,after} to avoid timer jiffies wrap (5882)
5616 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5617 - Do not leak request if server was not able to process it (5154)
5618 - If mds_open unable to find parent dir, make that negative lookup(5154)
5619 - don't create new directories with extent-mapping (5909, 5936)
5621 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5622 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5623 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5624 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5625 - support for RHEL4 kernel on i686 (5773)
5626 - provide error messages when incompatible logs are encountered (5898)
5628 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5629 * version 1.4.0.10 (1.4.1 release candidate 1)
5631 - don't keep a lock reference when lock is not granted (4238)
5632 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5633 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5634 - drop import inflight refcount on signal_completed_replay error (5255)
5635 - unlock page after async write error during send (3677)
5636 - handle missing objects in filter_preprw_read properly (5265)
5637 - no transno return for symlink open, don't save no-trasno open (3440)
5638 - don't try to complete elan receive that already failed (4012)
5639 - free RPC server reply state on error (5406)
5640 - clean up thread from ptlrpc_start_thread() on error (5160)
5641 - readahead could read extra page into cache that wasn't ejected (5388)
5642 - prevent races in class_attach/setup/cleanup/detach (5260)
5643 - don't dereference de->d_inode after l_dput of de (5458)
5644 - use "int" for stripe value returned from lock_to_stripe (5544)
5645 - mballoc allocation and error-checking fixes in 2.6 (5504)
5646 - block device patches to fix I/O request sizes in 2.6 (5482)
5647 - look up hostnames for IB nals (5602)
5648 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5649 - don't start multiple acceptors for the same port (5277)
5650 - fix incorrect LASSERT in mds_getattr_name (5635)
5651 - export a proc file for general "ping" checking (5628)
5652 - fix "lfs check" to not block when the MDS is down (5628)
5654 - service request history (4965)
5655 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5656 - create an "evict_client" /proc entry on OSTs, like the MDS has
5657 - fix mount usage message, return errors per mount(8) (5168)
5658 - change grep [] to grep "[]" in tests so they work in more UMLs
5659 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5660 - remove ancient LOV_MAGIC_V0 EA support (5047)
5661 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5662 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5663 - allow readahead allocations to fail when low on memory (5383)
5664 - mmap locking landed again, after considerable improvement (2828)
5665 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5667 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5670 - send OST transaction number in read/write reply to free req (4966)
5671 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5672 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5673 - the watchdog thread now runs as interruptible (5246)
5674 - flock/lockf fixes (but it's still disabled, pending 5135)
5675 - don't use EXT3 constants in llite code (5094)
5676 - memory shortage at startup could cause assertion (5176)
5678 - reorganization of lov code
5679 - single portals codebase
5681 - add extents/mballoc support (5025)
5682 - direct I/O reads in the obdfilter (4048)
5683 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5685 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5688 - send OST transaction number in read/write reply to free req (4966)
5689 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5690 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5691 - the watchdog thread now runs as interruptible (5246)
5692 - handle missing objects in filter_preprw_read properly (5265)
5693 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5694 - drop import inflight refcount on signal_completed_replay error (5255)
5695 - unlock page after async write error during send (3677)
5696 - return original error code on reconstructed replies (3761)
5697 - no transno return for symlink open, don't save no-trasno open (3440)
5699 - add pid to ldlm debugging output (4922)
5700 - bump the watchdog timeouts -- we can't handle 30sec yet
5701 - extra debugging for orphan dentry/inode bug (5259)
5703 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5706 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5707 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5708 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5709 - fix NULL dereference in /proc/sys/portals/routes (4827)
5710 - allow failed mdc_close() operations to be interrupted (4561)
5711 - stop precreate on OST before MDS would time out on it (4778)
5712 - don't send partial-page writes before EOF from client (4410)
5713 - discard client grant for sub-page writes on large-page clients (4520)
5714 - don't free dentries not owned by NFS code, check generation (4806)
5715 - fix lsm leak if mds_create_objects() fails (4801)
5716 - limit debug_daemon file size, always print CERROR messages (4789)
5717 - use transno after validating reply (3892)
5718 - process timed out requests if import state changes (3754)
5719 - update mtime on OST during writes, return in glimpse (4829)
5720 - add mkfsoptions to LDAP (4679)
5721 - use ->max_readahead method instead of zapping global ra (5039)
5722 - don't interrupt __l_wait_event() during strace
5724 - add software watchdogs to catch hung threads quickly (4941)
5725 - make lustrefs init script start after nfs is mounted
5726 - fix CWARN/ERROR duplication (4930)
5727 - return async write errors to application if possible (2248)
5728 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5729 - print NAL number in %x format (4645)
5730 - update barely-supported suse-2.4.21-171 series (4842)
5731 - support for sles 9 %post scripts
5732 - support for building 2.6 kernel-source packages
5733 - support for sles km_* packages
5735 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5738 - ignore -ENOENT errors in osc_destroy (3639)
5739 - notify osc create thread that OSC is being cleaned up (4600)
5740 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5741 - reconstruct ost_handle() like mds_handle() (4657)
5742 - create a new thread to do import eviction to avoid deadlock (3969)
5743 - let lconf resolve symlinked-to devices (4629)
5744 - don't unlink "objects" from directory with default EA (4554)
5745 - hold socknal file ref over connect in case target is down (4394)
5746 - allow more than 32000 subdirectories in a single directory (3244)
5747 - fix blocks count for O_DIRECT writes (3751)
5748 - OST returns ENOSPC from object create when no space left (4539)
5749 - don't send truncate RPC if file size isn't changing (4410)
5750 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5751 - bind to privileged port in socknal and tcpnal (3689)
5753 - rate limit CERROR/CWARN console message to avoid overload (4519)
5754 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5755 - basic mmap support (3918)
5756 - kernel patch series update from b1_4 (4711)
5758 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5761 - avoid crash during MDS cleanup with OST shut down (2775)
5762 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5763 - don't use bad inodes on the MDS (3744)
5764 - dynamic object preallocation to improve recovery speed (4236)
5765 - don't hold spinlock over lock dumping or change debug flags (4401)
5766 - don't zero obd_dev when it is force cleaned (3651)
5767 - print grants to console if they go negative (4431)
5768 - "lctl deactivate" will stop automatic recovery attempts (3406)
5769 - look for existing locks in ldlm_handle_enqueue() (3764)
5770 - don't resolve lock handle twice in recovery avoiding race (4401)
5771 - revalidate should check working dir is a directory (4134)
5773 - don't always mark "slow" obdfilter messages as errors (4418)
5775 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5778 - don't close LustreDB during write_conf until it is done (3860)
5779 - fix typo in lconf for_each_profile (3821)
5780 - allow dumping logs from multiple threads at one time (3820)
5781 - don't allow multiple threads in OSC recovery (3812)
5782 - fix debug_size parameters (3864)
5783 - fix mds_postrecov to initialize import for llog ctxt (3121)
5784 - replace config semaphore with spinlock (3306)
5785 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5786 - don't allow enqueue to complete on a destroyed export (3822)
5787 - down write_lock before checking llog header bitmap (3825)
5788 - recover from lock replay timeout (3764)
5789 - up llog sem before sending rpc (3652)
5790 - reduce ns lock hold times when setting kms (3267)
5791 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5792 - fix NULL deref and obd_dev leak on setup error (3312)
5793 - replace some LBUG about llog ops with error handling (3841)
5794 - don't match INVALID dentries from d_lookup and spin (3784)
5795 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5796 - fix invalid assertion in ptlrpc_set_wait (3880)
5798 - add libwrap support for the TCP acceptor (3996)
5799 - add /proc/sys/portals/routes for non-root route listing (3994)
5800 - allow setting MDS UUID in .xml (2580)
5801 - print the stack of a process that LBUGs (4228)
5803 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5806 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5807 - make sure to unset replay flag from failed open requests (3440)
5808 - if default stripe count is 0, use OST count for inode size (3636)
5809 - update parent mtime/ctime on client for create/unlink (2611)
5810 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5811 - free recovery state on server during a forced cleanup (3571)
5812 - unregister_reply for resent reqs (3063)
5813 - loop back devices mounting and status check on 2.6 (3563)
5814 - fix resource-creation race that can provoke i_size == 0 (3513)
5815 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5816 - more debugging for page-accounting assertion (3746)
5817 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5818 - avoid deadlock after precreation failure (3758)
5819 - fix race and lock order deadlock in orphan handling (3450, 3750)
5820 - add validity checks when grabbing inodes from l_ast_data (3599)
5822 - add /proc/.../recovery_status to obdfilter (3428)
5823 - lightweight CDEBUG infrastructure, debug daemon (3668)
5824 - change default OSC RPC parameters to be better on small clusters
5825 - turn off OST read cache for files smaller than 32MB
5826 - install man pages and include them in rpms (3100)
5827 - add new init script for (un)mounting lustre filesystems (2593)
5828 - run chkconfig in %post for init scripts (3701)
5829 - drop scimac NAL (unmaintained)
5831 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5834 - clean kiobufs before and after use (3485)
5835 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5836 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5837 - change init script to fail more clearly if not run as root (1528)
5838 - allow clients to reconnect during replay (1742)
5839 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5840 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5841 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5842 - fix cond_resched (3554)
5843 - only evict unfinished clients after recovery (3515)
5844 - allow bulk resend, prevent data loss (3570)
5845 - dynamic ptlrpc request buffer allocation (2102)
5846 - don't allow unlinking open directory if it isn't empty (2904)
5847 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5848 - remove extraneous obd dereference causing LASSERT failure (3334)
5849 - don't use get_cycles() when creating temp. files on the mds (3156)
5850 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5851 - handle EEXIST for set-stripe, set proper directory name (3336)
5853 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5854 - fix ksocknal_fmb_callback() error messages (2918)
5856 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5859 - don't copy lvb into (possibly NULL) reply on error (2983)
5860 - don't deref dentry after dput, don't free lvb on error (2922)
5861 - use the kms to determine writeback rpc length (2947)
5862 - increment oti_logcookies when osc is inactive (2948)
5863 - update client's i_blocks count via lvb messages (2543)
5864 - handle intent open/close of special files properly (1557)
5865 - mount MDS with errors=remount-ro, like obdfilter (2009)
5866 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5867 - don't use cancelling-locks' kms values (2947)
5868 - use highest lock extent for kms, not last one (2925)
5869 - don't dereference ERR_PTR() dentry in error handling path (3107)
5870 - fix thread race in portals_debug_dumplog() (3122)
5871 - create lprocfs device entries at setup instead of at attach (1519)
5872 - common AST error handler, don't evict client on completion race (3145)
5873 - zero nameidata in detach_mnt in 2.6 (3118)
5874 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5875 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5876 - initialize RPC timeout timer earlier for 2.6 (3219)
5877 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5878 - print nal/nid for unknown nid (3258)
5879 - additional checks for oscc recovery before doing precreate (3284)
5880 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5881 - don't crash in mdc_close for bad permissions on open (3285)
5882 - zero i_rdev for non-device files (3147)
5883 - clear page->private before handing to FS, better assertion (3119)
5884 - tune the read pipeline (3236)
5885 - fix incorrect decref of invalidated dentry (2350)
5886 - provide read-ahead stats and refine rpc in flight stats (3328)
5887 - don't hold journal transaction open across create RPC (3313)
5888 - update atime on MDS at close time (3265)
5889 - close LDAP connection when recovering to avoid server load (3315)
5890 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5891 - don't leak open file on MDS after open resend (3325)
5892 - serialize filter_precreate and filter_destroy_precreated (3329)
5893 - loop device shouldn't call sync_dev() for nul device (3092)
5894 - clear page cache after eviction (2766)
5895 - resynchronize MDS->OST in background (2824)
5896 - refuse to mount the same filesystem twice on same mountpoint (3394)
5897 - allow llmount to create routes for mounting behind routers (3320)
5898 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5899 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5900 - fix rare problem with rename on htree directories (3417)
5902 - allow default OST striping configuration per directory (1414)
5903 - fix compilation for qswnal for 2.6 kernels (3125)
5904 - increase maximum number of MDS request buffers for large systems
5905 - change liblustreapi to be useful for external progs like lfsck (3098)
5906 - increase local configuration timeout for slow disks (3353)
5907 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5909 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5913 - don't overwrite extent policy data in reply if lock was blocked (2901)
5914 - drop filter export grants atomically with removal from device (2663)
5915 - del obd_self_export from work_list in class_disconnect_exports (2908)
5916 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5917 - swab reply message in mdc_close, other PPC fixes (2464)
5918 - fix destroying of named logs (2325)
5919 - overwrite old logs when running lconf --write_conf (2264)
5920 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5921 - fix race in target_handle_connect (2898)
5922 - mds_reint_create() should take same inode create lock (2926)
5923 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5924 - don't close files for self_export to avoid uninitialized obd (2936)
5925 - allow MDS with the same name as client node (2939)
5926 - hold dentry reference for closed log files for unlink (2325)
5927 - reserve space for all logs during transactions (2059)
5928 - don't evict page beyond end of stripe extent (2925)
5929 - don't oops on a deleted current working directory (2399)
5930 - handle hard links to targets without a parent properly (2517)
5931 - don't dereference NULL lock when racing during eviction (2867)
5932 - don't grow lock extents when lots of conflicting locks (2919)
5934 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - account for cache space usage on clients to avoid data loss (974)
5938 - lfsck support in lustre kernel code (2349)
5939 - reduce journal credits needed for BRW writes (2370)
5940 - orphan handling to avoid losing space on client/server crashes
5941 - ptlrpcd can be blocked, stopping ALL progress (2477)
5942 - use lock value blocks to assist in proper KMS, faster stat (1021)
5943 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5944 - recovery for initial connections (2355)
5945 - fixes for mds_cleanup_orphans (1934)
5946 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5947 - block all file creations until orphan recovery completes (1901)
5948 - client remove rq_connection from request struct (2423)
5949 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5950 - recovery timer race (2670)
5951 - mdc_close recovey bug (2532)
5952 - ptlrpc cleanup bug (2710)
5953 - mds timeout on local locks (2588)
5954 - namespace lock held during RPCs (2431)
5955 - handle interrupted sync write properly (2503)
5956 - don't try to handle a message that hasn't been replied to (2699)
5957 - client assert failure during cleanup after abort recovery (2701)
5958 - leak mdc device after failed mount (2712)
5959 - ptlrpc_check_set allows timedout requests to complete (2714)
5960 - wait for inflight reqs when ptlrpcd finishes (2710)
5961 - make sure unregistered services are removed from the srv_list
5962 - reset bulk XID's when resending them (caught by 1138 test)
5963 - unregister_bulk after timeout
5964 - fix lconf error (2694)
5965 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5966 - readahead locks pages, leaves pending causing memory pressure (2673)
5967 - increase OST request buffers to 4096 on large machines (2729)
5968 - fix up permission of existing directories in simple_mkdir (2661)
5969 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5970 - don't assign transno to errored transactions (2742)
5971 - don't delete objects on OST if given a bogus objid from MDS (2751)
5972 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5973 - if rq_no_resend, then timeout request after recovery (2432)
5974 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5975 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5976 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5977 - MDS recovery shouldn't delete objects that it has given out (2730)
5978 - if enqueue arrives after completion, don't clobber LVB (2819)
5979 - don't unlock pages twice when trigger_group_io returns error (2814)
5980 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5981 - don't write pages to disk if there was an error (1450)
5982 - don't ping imports that have recovery disabled (2676)
5983 - take buffered bytes into account when balancing socknal conn (2817)
5984 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5985 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5986 - remove little-endian swabbing of llog records (1987)
5987 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5988 - retry reposting request buffers if they fail (1191)
5989 - grow extent at grant time to avoid granting a revoked lock (2809)
5990 - lock revoke doesn't evict page if covered by a second lock (2765)
5991 - disable VM readahead to avoid reading outside lock extents (2805)
5993 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5994 - updated kernel patches for hp-2.4.20 kernel (2681)
5996 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5999 - fix truncated write corruption (2366)
6000 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6002 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6003 - ensure len(uuid) < 37 in lmc (1171)
6004 - fix ia64 OOPS in llog_test (2255)
6005 - zero end of page at obdfilter for partial page writes (2648)
6006 - don't leave stale dentries around after renames (bug 2428)
6007 - fix timeouts when evicting a client with a single lock held (2642)
6008 - set deadline for the initial HELLO message to drain (2634)
6009 - print out dotted-quad IP addresses in the socknal (2302)
6011 - additional debugging for MDS client eviction problem (2443)
6012 - fix mkfsoptions support for osts (2603, 2604)
6014 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6017 - add series for the vanilla 2.6.0 kernel
6018 - add series for the vanilla 2.4.24 kernel
6019 - add series for a cray x86/64 UL kernel drop
6020 - fix xattr patches for the vanilla 2.4.19 series
6022 - generate true UUIDs in lmc (1171)
6023 - have portals stack dumping break in UML (2466)
6024 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6025 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6026 - wake callback waiting threads on client eviction (2460)
6027 - Add --ptldebug and --subsystem to lmc (1719)
6028 - update assertion to allow safe interrupt allocation
6029 - set rq_no_resend for cancel requests (2432)
6030 - recalculate ptlrpcd timeout after resend (2494)
6031 - call vfs_rmdir when removing pending directories (2368)
6032 - fix renaming a file to itself (2429)
6033 - lmc creates a default one-stripe lov (2454)
6034 - expand procfs space to handle large clusters (2326)
6035 - increase UML stack to avoid overflow
6036 - update lconf's list of debug and subsystem masks
6037 - fix lfs find --obd (2510)
6038 - /proc tunable for disabling filter read caching (2591)
6039 - stop rpm packages from altering slapd.conf (2301)
6040 - disable nagle in the socknal under 0conf (2578)
6041 - choose mds inode size based on stripe count (2572)
6042 - fix kernel-source rpm problems (2516)
6044 - add --disable-doc to avoid pdf generation (2421)
6045 - update documentation, tests, type-os, comments
6046 - avoid format warnings on ia64
6047 - remove the TOE NAL
6048 - tiny code cleanups by removing unused fields
6050 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6053 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6054 - remove the most blatant lies from BUILDING file (2371)
6055 - change default debug level to reasonable production setting
6056 - reduce client side cache size to reduce cache flush time
6057 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6058 - make TCP zerocopy and pinger support enabled by default (2476)
6059 - sync writes completed after process exits caused crashes (2319)
6060 - maintain correct mount count on the MDS (2356)
6061 - backout 1557, because 2316 wasn't really fixed
6062 - better file I/O statistics gathering in /proc
6063 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6064 - another kernel patch to fix zero-copy TCP function export
6065 - don't take duplicate lock when processing re-sent getattr (2420)
6066 - lctl uses obd_self_export instead of creating new conn (2353)
6067 - MDS/OST recovery case which requires object creation asserted (2425)
6068 - move lfs from /usr/sbin to /usr/bin in packages
6069 - fix race between mds_client_add and mds_client_free (2417)
6070 - use kmalloc instead of slabs in portals (2430)
6071 - don't create duplicate records when a failover MDS is present (2442)
6072 - remove unnecessary mount age check (2332)
6073 - don't remove directory inodes from locks prematurely (2451)
6074 - don't break if MDS service name is the same as hostname (2103)
6075 - fix races in client write RPC generation when cache full (2482)
6077 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6080 - remove now-unused request->rq_obd (278)
6081 - if an allocation fails, print out how much memory we've used (1933)
6082 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6083 - add the "configurable stack size" patch to most series files (1256)
6084 - ability to write large log records, for 100+ OST configs (2306)
6085 - fix NULL deref when filter_prep fails (2314)
6086 - fix operator precedence error in filter_sync
6087 - dynamic allocation of socknal TX descriptors (2315)
6088 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6089 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6090 - ensure that CWARN messages in Portals always get to the syslog
6091 - __init/__exit are not for prototype decls (ldlm_init/exit)
6092 - x86-64 compile warning fixes
6093 - fix gateway LMC keyword conflict (2318)
6094 - fix MDS lock inversions in getattr/reint paths (1844)
6095 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6096 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6097 - DLM race condition prevented some lock evictions (2328)
6098 - ENOMEM detection and retry on socknal sends (2230)
6099 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6100 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6101 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6102 - better error messages when a client is rejected during recovery (1505)
6103 - avoid cancelling locks which were never granted, after failure (2330)
6104 - fix i_sem/journal inversion in mds_client_add (2333)
6105 - fix truncate/getattr lock cycle deadlock (2334)
6106 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6107 - fix two rare exit paths which could leak an l_lock() ref (2321)
6108 - fencepost error in MDS/OST orphan recovery (2226)
6109 - make log record alignment 8 bytes (1988)
6110 - lstripe now fails when requested offset > ost_count (2237)
6111 - ensure that all kernel series have a complete list.h (1607)
6112 - fix crashes in special-file operations (2316)
6113 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6115 - allow configurable automake binary, for testing new versions
6116 - small update to the lfs documentation
6118 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6120 * fix negative export reference count in fsfilt_sync (2312)
6122 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6123 * release candidate 0.9.1
6125 - orphans are moved into the PENDING directory for possible recovery
6126 - replayed opens now open by fid for orphan/rename safety (1042)
6127 - last close of an orphan inode generates a transno (683)
6128 - chdir() and mount() now pin the directory entry (1020)
6129 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6130 - discard very old requests without processing them (1502)
6131 - remove some common, well-understood CERRORs (1505)
6132 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6133 - clear "grant" flags in OST replies until OST grant code lands (1644)
6134 - fix read performance by not clobbering i_blksize on client (1598)
6135 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6136 - make LRU size a /proc tunable, clears locks when reduced (707)
6137 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6138 - server locks take references on exports now (1558)
6139 - build fixes for 2.4.20-rh trees (1663)
6140 - return an error from lov_create if all OSCs are inactive (1751)
6141 - fix import levels when a reconnect happens without a timeout (1597)
6142 - exit early from mds_open if we get a lookup error (1749)
6143 - partial page read at EOF wouldn't wait for disk before sending (1642)
6144 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6145 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6146 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6147 - write/truncate lock inversion (1639)
6148 - Don't auto-load obdclass, portals modules during cleanup (1495)
6149 - fix timestamps from jumping to "now" (1763)
6150 - extra journal assertions (1648)
6151 - add an extra multiunlink test (1771)
6152 - fix read_record/write_record API (1776)
6153 - fix leak of offset_extent, possible incorrect i_size later (1772)
6154 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6155 - Debugging for the kqswnal_get_idle_tx problems (1820)
6156 - Allow recovery to be attempted multiple times (1536)
6157 - Write out MDS last_rcvd file after it is first created (1600)
6158 - Fix tx_descriptor leak in failed transmit situations (1827)
6159 - ext3 journaling fixes for assertion failure after IO error (1871)
6160 - class_export_put() on freed export after completion AST error (1896)
6161 - Fix revalidate looping in VFS (1322)
6162 - Don't access a freed export during MDS_REINT timeout (1521)
6163 - Add open-unlink recovery support on the MDS (1673,1764)
6164 - Return an error if no MDS data was read from last_rcvd (1946)
6165 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6166 - Don't setup a disk device that doesn't match exported UUID (317)
6167 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6168 - avoid committing NULL handle in force close
6169 - local.sh is now a one-stripe LOV configuration
6170 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6171 - fix ext3 htree duplicate directory entry corruption (1516)
6172 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6173 - update to new LOV EA format (2097)
6174 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6176 2003-06-15 Phil Schwan <phil@clusterfs.com>
6179 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6180 - per-import/export recovery handling (958, 931, 959)
6181 - multiple last-rcvd slots, for serving multiple FSes (949)
6182 - connections are again shared between multiple imp/exports (963, 964)
6183 - "umount -f" would hang if any requests needed to be sent (393, 978)
6184 - avoid pinning large req buffer by copying for queued messages (989)
6185 - add "uuid" to "lctl device" command to help upcalls (991)
6186 - "open" RPCs with transnos would confuse recovery counters (1037)
6187 - do proper endian conversion of all wire messages (288, 340, 891)
6188 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6189 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6190 - fix ptlrpc_connection leak in target_handle_connect (1174)
6191 - fix import refcounting bug in OST and MDS cleanup (1134)
6192 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6193 - fix very unlikely obd_types race condition (501)
6194 - remove osc_open hack for echo_client (1187)
6195 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6196 - a failure in read_inode2 leads to deadlock (1139)
6197 - cancel ack-locks as soon as transaction is committed (1072)
6198 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6199 - make sure to commitrw after any preprw to avoid deadlock (1162)
6200 - failing to execute a file in a lustre FS would lock inode (1203)
6201 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6202 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6203 - obd_uuid2tgt would walk off the end of the list (1255)
6204 - on IA64 the osc would give portals incorrect bulk size (1258)
6205 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6206 - fix lock inversion caused by new llite matching code (1282)
6207 - limit the number of dirty pages on a client to 10MB (1286)
6208 - timed out locks were not being corrected cancelled (1289)
6209 - fix O_DIRECT above 4GB on IA-32 (1292)
6210 * major user-visible changes
6211 - fail out/fail over policy now controlled by the upcall (993)
6213 - add OBD_PING to check server availability and failure (954)
6214 - lustre messages are now sent in sending host order (288, 340, 891)
6215 - add eadatalen to MDS getattr reply (340)
6216 - OST read replies may contain second buffer, with per-page status (593)
6218 2003-03-11 Phil Schwan <phil@clusterfs.com>
6221 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6222 - failed open()s could cause deadlock; fixed (867, 869)
6223 - stop cancelling OST locks when files are closed (481)
6224 - overlapping XID spaces caused network corruption (851, 853)
6225 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6226 - setattr_raw added, to do single-RPC, server-side setattrs
6227 - lmc/lconf syntax change for OST UUIDs
6228 - fix crashy race condition between ptlrpc_free_req and osc_close
6229 - don't use request in mdc_enqueue if we hit a timeout (889)
6230 - don't set the inode i_size for regular files from the MDS (896)
6231 - handle out of order completion AST (842)
6232 - don't LBUG if a lock request times out after receiving AST (913)
6233 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6234 - if a bad lock AST arrives, send an error instead of dropping entirely
6235 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6236 - fix leak in bulk IO when only partially completed (899, 900, 926)
6237 - fix O_DIRECT for ia64 (55)
6238 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6239 - C-z after timeout could hang a process forever; fixed (977)
6241 - client-side I/O cache (678, 924, 929, 941, 970)
6243 - READPAGE and SETATTRs which don't take server-side locks get
6246 2003-02-11 Phil Schwan <phil@clusterfs.com>
6249 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6250 - Fix fsfilt_extN_readpage() to read a full page of directory
6251 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6252 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6253 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6254 - Most intent operations were removed, in favour of a new RPC mode
6255 that does a single RPC to the server and bypasses most of the VFS
6256 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6257 - Aggressively cancel local locks on DLM servers
6258 - mds_reint_unlink sends EA to the client if it's the last nlink.
6259 client uses that EA to unlink OST objects.
6260 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6261 - recursive symlinks were fixed (439)
6262 - fixed NULL deref in DEBUG_REQ
6263 - filter_update_lastobjid no longer calls sync, which annoyed extN
6264 - fixed multi-client small-writes to a single file problem (445)
6265 - fixed mtime updates during file writes (607)
6266 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6267 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6268 - fixed Linux OST size reporting problem (444, 656)
6269 - OST now updates object mtime with writes or setattr (607, 619)
6270 - client verifies file size before zeroing page past EOF (445)
6271 - OST now writes last allocated objid to disk with allocation (108)
6272 - LOV on echo now works (409)
6274 - mds_reint_unlink sends a new buffer, with the EA included. this
6275 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6276 set if a regular file was being unlinked, and it was the last link
6277 - use PtlGet from the target for bulk writes (315)
6278 - OST now updates object mtime with writes or setattr (607, 619)
6279 - LDLM now has a grant-time callback to revalidate locked items, if
6281 - Many MDS operations were reorganized to combat race conditions
6283 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6284 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6286 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6289 - Fully reactivate OST imports after reconnection (512, others)
6290 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6291 - More graceful error handling for truncating on dead OST (515)
6292 - Don't error out unless we're actually accessing dead stripes (474)
6293 - Fix garbage sizes when stripes are missing (410)
6294 - LRU counters were broken, causing constant lock purge (433, 432)
6295 - garbage on read from stripes with failed OSTs (441)
6296 - mark OSCs as active before reconnecting during recovery (438)
6297 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6298 - lfind did not preserve OST order in output (443)
6299 - symlinks cause hung clients, incorrect data (439)
6300 - stop dereferencing request after dropping refcount (457)
6301 - don't LASSERT(spin_is_locked) on non-SMP (455)
6302 - fixes for many rename() bugs
6303 - fstat didn't correctly synchronize attributes (399)
6304 - server must handle lock cancellation during blocking AST prep (487)
6305 - bulk descriptors were free()d too soon (511)
6306 - fix paths in lconf, which would load incorrect modules (451, 507)
6307 - fix confusing lconf 'host not found' error message (386)
6308 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6309 - fix race condition in mdc_blocking_ast() for inode access (526)
6310 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6311 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6312 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6313 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6314 - fix lstripe in several ways (kernel side) (527)
6315 - fix request leak in ldlm_cli_enqueue (262)
6316 - incorrect OSC was marked inactive after OST failure
6317 - call mds_fs_cleanup before unmounting filesystem (524)
6318 - fix races between taking ns_lock and ldlm_lock_change_resource
6319 - fix races updating LOV export open file list
6320 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6321 - fix recovery NULL deref in ldlm_cli_cancel_unused
6322 - fix some DLM races by using new hash table for lock handles (419)
6323 - permit the client to specify desired inodes, at replay
6324 - duplicate requests when we queue them for replay reintegration
6325 - fix last_rcvd offset calculation
6326 - sync after each recovered transaction, so we always make progress
6327 - never, not always, ERESTART requests without transnos
6328 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6330 - skip replay if the MDS says that the client is already connected
6331 - don't check for a recovery-enabled export to match lctl's UUID
6332 - don't INC_USE_COUNT for phantom exports
6333 - don't crash when cleaning up phantom exports (567)
6334 - don't double-finish or set replay data for errored mdc_open requests
6335 - abort requests when they time out, so we don't get old replies
6336 - send/receive replies for AST messages again
6337 - if the client says that it doesn't have the lock, cancel it on the
6339 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6340 mark it as destroyed, it will all work out when decref is called
6341 - fix module use counts (22, 581)
6343 - ASTs now expect a reply (server cancels lock on error reply)
6345 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6348 - fix many simultaneous client startup (392)
6349 - fix dentry->d_it clobbering
6350 - credentials weren't being shipped for readdir/getattr operations
6351 - remove invalid assertions triggered during some concurrent MD
6353 - proper Lustre versions added (336, 389)
6354 - fix memory leak for create error case (398)
6355 - fix LOV locking bug that would get cli/srv out of sync
6356 - fix echo client over LOV (409)
6357 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6358 - fix double-O_EXCL intent crash (424)
6359 - avoid sending multiple lock CANCELs (352)
6361 - MDS can do multi-client recovery (modulo bugs in new code)
6363 - many updates, edits, cleanups
6365 2002-11-18 Phil Schwan <phil@clusterfs.com>
6368 - fix null d_it dereference (346)
6369 - fix full OST/dbench hang (333)
6370 - fix permission problem with file removal (286)
6371 - fix removal of OSCs from LOV when they fail
6372 - fix NULL deref during bulk timeout (214)
6373 - fix problems related to multiple filesystems on one MDS (241)
6374 - fixed serious subtle metadata locking bugs
6375 - free locks on clients when inodes are removed due to memory
6377 - fix inode pointer in lock data (285)
6378 - partial support for multiple MDS on a single host (241)
6379 - data locks weren't cancelled at clear_inode time (290, 311)
6380 - intent locks could lead to unbounded lock growth (205)
6381 - added a maximum lock count, an LRU list, and a flusher
6382 - fix multiple rename (365)
6383 - properly abstracted the echo client
6384 - OSC locked 1 byte too many; fixed
6385 - rewrote brw callback code:
6386 - fixed recovery bugs related to LOVs (306)
6387 - fixed too-many-pages-in-one-write crash (191)
6388 - fixed (again) crash in sync_io_timeout (214)
6389 - probably fixed callback-related race (385)
6391 - Add capability to MDS protocol
6392 - LDLM cancellations and callbacks on different portals
6394 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6397 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6398 - defer open object destruction to close time (601981, 138)
6399 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6400 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6401 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6402 - don't call obd_setattr in ll_file_release for destroyed objects
6403 * protocol change to lustre_msg: move |version| and add |flags|
6404 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6405 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6406 * added replay of create, unlink, link and rename operations during
6407 MDS failover; recovery should be much more robust now
6408 * remove failed OSCs from LOVs (only lov_create uses this so far)
6409 * the lustre-HOWTO was brought (more) up to date (582544)
6411 2002-10-23 Phil Schwan <phil@clusterfs.com>
6414 - in-use dentries weren't being reused properly (617851)
6415 - prevent multiple LDLM setup (599178)
6416 - fix LOV size calculations for truncate (617853)
6417 - fix client handling of MDS intent errors (POSIX)
6418 - fix permission bug in lovstripe.c test (624321)
6419 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6420 - truncate past end of file could corrupt data
6421 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6422 - a race in recovery could return ETIMEDOUT to apps (623947)
6423 - building outside the source directory was fixed
6424 * the lustre-HOWTO was brought (more) up to date (582544)
6425 * major progress was made on recovery functionality
6427 2002-10-10 Phil Schwan <phil@clusterfs.com>
6430 - recovery deadlock fix
6431 - rm -rf causes LBUG fix (617817)
6432 - file open by multiple tasks fix (618962)
6433 - directory permissions bugs (602707 and 620007)
6434 - journal_stop fixed with locking (611313)
6435 - O_APPEND failures resolved (618273, perhaps 614459)
6436 - lconf PATH fix (619770)
6437 - IA64 build fix (621450)
6438 - RPC buffer sizes scale with amount of memory
6440 2002-10-01 Phil Schwan <phil@clusterfs.com>
6443 - locks would be cancelled without throwing away data pages,
6444 resulting in inconsistent data (605627)
6445 - inode attributes were not always being refreshed (605627, 612449)
6446 - lconf now continues to cleanup after lctl reports an error
6447 - MDS now enforces user permissions (602707)
6448 - lprocfs cleanup fixed, but not yet enabled (614157)
6449 - fixed infinite server hang, should a client not respond to an AST
6450 - avoid going into recovery if user calls readlink() with a buffer
6451 that's too small (613941)
6452 - AST RPCs no longer require replies (614867) -- this may be changed
6453 - don't crash server if client sends an IOV that's too big (611336)
6454 - fixed lock conversion deadlock (611892)
6455 - fixed the following of symlinks (614622)
6456 * recovery: the server can remove locks from a client that dies, other
6457 clients can make progress
6458 * more extN patch fixes
6459 * compile-time configurable ptlrpc buffer allocations
6461 - collaborative read cache document
6462 - Lustre Lite Performance CDR document-in-progress
6464 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6467 - fix typo in patch-2.4.18
6469 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6472 - clear ptlrpc request each time in handle_incoming_request()
6473 - unlink of files now destroys the object on the OST
6475 2002-09-19 Peter Braam <braam@clusterfs.com>
6477 * add hard link support
6478 * change obdfile creation method
6479 * kernel patch changed
6481 2002-09-19 Peter Braam <braam@clusterfs.com>
6484 - stack overflow bug in extN fixed
6486 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6488 * documentation updates
6489 - add man pages for config tools
6490 - update tests/README to describe testing with new config tools
6491 - finish metadata API descriptions
6492 * bug fixes and cleanups
6493 - statfs workaround for 16TB limit
6494 - LOV stripe allocation improved, can stripe on subset of OSTs
6495 - LOV file size/IO offset was wrong for files > 4GB in size
6496 - object EA data was being dropped, caused files to be unreadable
6497 - memory overflow with non-LOV OST caused memory corruption
6498 - fixed regression tests to work with new config tools, obdfilter
6499 - fixed bug when directory size became larger than 1 block
6500 - fixed bug (for single client case) when PWD was deleted
6501 - invalidate local directory pages when doing intent-based ops
6502 - avoid LDLM oops when lock callback contained bad data
6504 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6506 * documentation updates
6507 * bug fixes and cleanups
6508 - configuration tools
6511 - 64-bit compile warnings
6512 - 64-bit internal statfs data
6514 * test_brw on persistent OST devices
6516 * lprocfs (disabled)
6518 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6520 * documentation updates
6521 * bug fixes and cleanups
6522 * configuration tools
6524 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6527 * many small fixes to 0_5_4
6528 * io/network handling
6529 * thinkos in MDS operations
6531 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6534 * crucial basic fixes to 0.5.3
6535 * IOR, Iozone work over Elan
6538 2002-08-07 Phil Schwan <phil@clusterfs.com>
6539 * version 0_5_3, our first alpha
6540 * we use the new Portals iovs
6541 * documentation updates
6542 * bug fixes and cleanups
6543 * small changes in the DLM wire protocol
6545 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6546 * version 0_5_1 with some initial stability,
6547 * locking on MD and file I/O.
6548 * documentation updates
6549 * several bug fixes since 0.5.0
6550 * small changes in wire protocol
6552 2002-07-18 Phil Schwan <phil@clusterfs.com>
6554 * delivered as Lustre Light Alpha
6555 * fixed a crash after handling invalid MDS requests
6556 * fixed directory pages for architectures with non-4k pages sizes
6558 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6559 * release version v0_4_4
6560 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6562 * Fixes a number of interruption problems with OST operations.
6563 * Update documentation for portals header changes
6564 * Move all wire protocol structs/defines to lustre_idl.h
6565 * Fixes symlink length bug.
6566 * Add tcpdump to repository.
6568 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6569 * release version v0_4_3
6570 * Fixes statfs for inodes on extN.
6571 * Fixes bug in runtests which would delete /etc/hosts.
6572 * Use 64-bit object IDs wherever possible (not into VFS though)
6573 Remove ost_get_info, which is unused by lustre, and out of date.
6575 2002-07-03 Peter Braam <braam@clusterfs.com>
6576 * release version v0_4_2 Fixes a lookup error (type not passed)
6577 * move forward to head of Portals
6578 * move forward to latest Lustre kernel
6580 2002-06-25 Peter Braam <braam@clusterfs.com>
6581 * release version v0_4_1. Hopefully stable on single node use.