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 2.6.32-573.7.1.el6 (RHEL6.7)
9 3.10.0-229.14.1.el7 (RHEL7.1)
10 3.0.101-0.47.55 (SLES11 SP3)
11 3.0.101-65 (SLES11 SP4)
12 * Client known to build on unpatched kernels:
13 2.6.32-431.29.2.el6 (RHEL6.5)
14 2.6.32-504.30.3.el6 (RHEL6.6)
15 2.6.32-573.7.1.el6 (RHEL6.7)
16 3.10.0-229.14.1.el7 (RHEL7.1)
17 3.0.101-0.47.55 (SLES11 SP3)
18 3.0.101-65 (SLES11 SP4)
20 * Recommended e2fsprogs version: 1.42.13.wc3 or newer
21 * Recommended ZFS / SPL version: 0.6.5.2
22 * Tested with ZFS / SPL version: 0.6.5.2
23 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
24 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
25 stack overflow. For more information, please refer to bugzilla 17630.
26 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
27 the re-exporting nodes to be mounted with "32bitapi" mount option
29 --------------------------------------------------------------------------------
31 03-10-2015 Intel Corporation
33 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
34 for currently supported client and server kernel versions.
35 * Server known to build on patched kernels:
36 2.6.32-431.29.2.el6 (RHEL6.5)
37 2.6.32-504.8.1.el6 (RHEL6.6)
38 3.0.101-0.46 (SLES11 SP3)
39 * Client known to build on unpatched kernels:
40 2.6.32-431.29.2.el6 (RHEL6.5)
41 2.6.32-504.8.1.el6 (RHEL6.6)
42 3.10.0-123.20.1.el7 (RHEL7)
43 3.0.101-0.46 (SLES11 SP3)
44 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
45 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
46 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
47 stack overflow. For more information, please refer to bugzilla 17630.
48 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
49 the re-exporting nodes to be mounted with "32bitapi" mount option
51 Severity : enhancement
53 Description: control OST-index in IDIF via ROCOMPAT flag.
54 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
55 last_rcvd file. For new formatted OST device, it will be auto set;
56 for the case of upgrading from old OST device, you can enable it
57 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
58 enabled, for new created OST-object, its IDIF-in-LMA will contain
59 the OST-index; for the existing OST-object, the OSD will convert
60 old format IDIF as new format IDIF with OST-index stored in the
61 LMA EA when accessing such OST-object or via OI scrub. Once such
62 flag is enabled, it cannot be reverted back, so the system cannot
63 be downgraded to the orignal incompatible version.
65 --------------------------------------------------------------------------------
67 07-30-2014 Intel Corporation
69 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
70 for currently supported client and server kernel versions.
71 * Server known to build on patched kernels:
72 2.6.32-431.20.3.el6 (RHEL6.5)
73 3.0.101-0.31 (SLES11 SP3)
74 * Client known to build on unpatched kernels:
75 2.6.32-431.20.3.el6 (RHEL6.5)
76 3.0.101-0.31 (SLES11 SP3)
79 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
86 --------------------------------------------------------------------------------
88 10-31-2013 Intel Corporation
90 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
91 * Server builds for kernels:
92 2.6.32-358.18.1.el6 (RHEL6.4)
93 * Client builds for unpatched kernels:
94 2.6.32-358.18.1.el6 (RHEL6.4)
95 3.0.92-0.8 (SLES11 SP3)
97 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
98 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
99 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
100 stack overflow. For more information, please refer to bugzilla 17630.
101 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
102 the re-exporting nodes to be mounted with "32bitapi" mount option
104 --------------------------------------------------------------------------------
106 05-30-2013 Intel Corporation
108 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
109 * Server builds for kernels:
110 2.6.32-358.6.2.el6 (RHEL6.4)
111 * Client builds for unpatched kernels:
112 2.6.32-358.6.2.el6 (RHEL6.4)
113 3.0.74-0.6.6 (SLES11 SP2)
115 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
116 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
117 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
118 stack overflow. For more information, please refer to bugzilla 17630.
119 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
120 the re-exporting nodes to be mounted with "32bitapi" mount option
122 --------------------------------------------------------------------------------
124 9-30-2012 Whamcloud, Inc.
126 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
127 * Server support for kernels:
128 2.6.32-279.5.1.el6 (RHEL6)
129 * Client support for unpatched kernels:
130 2.6.32-279.5.1.el6 (RHEL6)
132 2.6.18-238.19.1.el5 (RHEL5)
133 2.6.16.60-0.69.1 (SLES10)
134 * Recommended e2fsprogs version: 1.41.90.wc4
135 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
136 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
137 stack overflow. For more information, please refer to bugzilla 17630.
138 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
139 the re-exporting nodes to be mounted with "32bitapi" mount option
141 Severity : enhancement
143 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
145 Severity : enhancement
147 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
149 Severity : enhancement
151 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
153 Severity : enhancement
155 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
157 Severity : enhancement
159 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
161 --------------------------------------------------------------------------------
163 04-30-2012 Whamcloud, Inc.
165 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
166 * Server support for kernels:
167 2.6.32-220.4.2.el6 (RHEL6)
168 * Client support for unpatched kernels:
169 2.6.32-220.4.2.el6 (RHEL6)
171 2.6.18-238.19.1.el5 (RHEL5)
172 2.6.16.60-0.69.1 (SLES10)
173 * Recommended e2fsprogs version: 1.41.90.wc4
174 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
175 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
176 stack overflow. For more information, please refer to bugzilla 17630.
177 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
178 the re-exporting nodes to be mounted with "32bitapi" mount option
180 Severity : enhancement
182 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
184 Severity : enhancement
186 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
188 Severity : enhancement
190 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
192 Severity : enhancement
194 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
196 --------------------------------------------------------------------------------
198 09-30-2011 Whamcloud, Inc.
200 * Server support for kernels:
201 2.6.18-238.19.1.el5 (RHEL5)
202 2.6.32-131.6.1.el6 (RHEL6)
203 * Client support for unpatched kernels:
204 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
205 2.6.18-238.19.1.el5 (RHEL5)
206 2.6.32-131.6.1.el6 (RHEL6)
207 2.6.29.4-167.fc11 (FC11)
208 2.6.16.60-0.69.1 (SLES10)
210 * Recommended e2fsprogs version: 1.41.90.wc3
211 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
212 Lustre filesystem with 4K stack may cause a stack overflow. For more
213 information, please refer to bugzilla 17630.
214 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
215 mounted with "32bitapi" mount option
217 Severity : enhancement
218 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
220 Severity : enhancement
221 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
223 Severity : enhancement
224 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
227 Frequency : only with SLES10
229 Description: Use OFED "KMP" provided by Novell
230 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
231 Lustre is now built against this package. That means you need to
232 install the ofed-kmp package from Novell for the patchless client.
233 Note that the ofed-kmp that Novell ships may not exactly match the
234 kernel version but should still be compatible.
236 Severity : enhancement
238 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
239 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
241 --------------------------------------------------------------------------------
243 2010-07-15 Oracle, Inc.
245 * Server support for kernels:
246 2.6.18-164.11.1.el5 (RHEL 5)
247 2.6.18-164.11.1.0.1.el5 (OEL 5)
248 * Client support for unpatched kernels:
249 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
250 2.6.18-164.11.1.el5 (RHEL 5),
251 2.6.18-164.11.1.0.1.el5 (OEL 5)
252 2.6.16.60-0.42.8 (SLES 10),
254 2.6.29.4-167.fc11 (FC11)
255 * Recommended e2fsprogs version: 1.41.10-sun2
257 Severity : enhancement
259 Description: Distribute lfs_migrate script with lustre
260 Details : The lfs_migrate mentioned in the manual is now included in the
261 lustre tree in lustre/scripts and will distributed in lustre RPMs.
263 Severity : enhancement
265 Description: Failover nid list added to import proc file.
267 Severity : enhancement
269 Description: "som_preview" mount option required to use SOM
270 Details : The clients will need to use som_preview mount option
274 Description: Lustre does not do 1MB IOs to HW RAID
275 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
277 Severity : enhancement
279 Description: lfs quota output is not very convenient for awk/sed-parsing
280 Details : Some positions in lfs quota output table could be empty or
281 non-empty which made it hard to parse it with scripts, now
282 a dash is put instead of space where there is not supposed
287 Description: Update to OFED 1.5.1
288 Details : OFED 1.5.1 will be used in environments that don't use
289 vendor supplied OFED.
293 Description: Remove lnet. prefix from lctl params display
294 Details : Also renamed lnet.memused to lnet_memused to
299 Description: Use conf_param -d to remove permanant settings.
303 Description: Add list_param -R
304 Details : lctl list_param now supports recursive mode.
307 Bugzilla : 21829, 22312
308 Description: Fix llobdstat.
309 Details : 1. Now updating the write_bytes counter
310 2. Port script fixes from 1.8.
314 Description: decrease the usage of memory on clients.
315 Details : 1. On clients, recycle dentries and inodes unused.
316 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
321 Description: llobdstat fix and enhancement
322 Details : add a counter to set a limit to how many samples will be returned
323 fix a wildcard in the path to limit to obdfilter stats only
327 Description: lfs check only allowed for root
328 Details : Code cleanups and added a regresion test of non-root use of
333 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
334 Details : Fix hr->hr_index overload.
338 Description: Fix FIEMAP functionality.
342 Description: FC11 patchless client support added
346 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
347 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
351 Description: New acc-sm script and various test-framework enhancments.
355 Description: Previously landed patch was reverted.
359 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
363 Description: Add ldiskfs maxdirsize mount option.
367 Description: Fix mkfs.lustre support for 16TB LUNs.
371 Description: Print mmp_check_interval
372 Details : print mmp_check_interval and make it possible to abort
373 mount operation in case it takes too long.
377 Description: Add cascading_rw to lustre mpi tests.
381 Description: Store SOM data in EA.
382 Details : Also several SOM related cleanups.
386 Description: Add refcount to md_open_data
390 Description: Use refcount on lov_request_set to prevent memory corruption.
394 Description: Statahead: use parent dir's i_mutex for synchronization.
398 Description: loadgen improvements.
402 Description: Enable peer health detection in ksocklnd and ko2iblnd.
406 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
410 Description: Fix race on obd_devcie:md_stats
414 Description: Fix SLES 11 client builds.
415 Details : Resolved uninitalized pointer gcc warnings.
419 Description: Allow non-root access for "lfs check".
420 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
424 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
426 Severity : enhancement
428 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
429 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
433 Description: Sanity test_27p fails.
434 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
435 Sanity tests 27 were cleaned up and improved.
439 Description: Unify req format on client/servers
440 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
441 buffers and swabbers approach..
446 Description: drop join file stuff.
448 Severity : enhancement
450 Description: Add 2.6.27 and SLES11 (patchless client) support
454 Description: Fix in ptlrpc_expire_one_request() to print the signed time
457 Severity : enhancement
459 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
462 Severity : enhancement
464 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
466 Severity : enhancement
468 Description: correctly handle big reply message.
469 Details : send LNet event if reply is bigger than buffer and adjust this buffer
474 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
476 Severity : enhancement
478 Description: Add support for OEL5.
482 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
484 Severity : enhancement
486 Description: "weak-modules" support
487 Details : Implement "weak-modules" support which enables kernel modules
488 to be used with any kernel that implements the same kABI. In
489 order to achieve this modules are now installed in
490 /lib/modules/$(uname -r)/updates/kernel on all distributions.
494 Description: client could not reconnect to OST because of an active request.
495 Details : The client now retries to reconnect to the same server, if a
496 connect request failed with EBUSY or -EAGAIN. Test case 35b
497 conf-sanity.sh has been created.
501 Description: Don't allow make backward step on assignin osc next id.
502 Details : race between allocation next id and ll_sync thread can be cause
503 of set wrong osc next id and can be kill valid ost objects.
507 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
511 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
515 Description: Error handling in osc_statfs_interpret() has been improved.
516 Details : Check in osc_statfs_interpret() for EBADR.
520 Description: Do not send statfs() requests to OSTs disabled by administrator.
521 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
523 Severity : enhancement
525 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
527 Severity : enhancement
529 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
531 Severity : enhancement
533 Description: Enhancement for lfs(1) command to use numeric uid/gid.
537 Description: Improved cleanup procedure for conf-sanity test_46a.
541 Description: Do not start bulk transfer, if client tried to reconnect.
542 Details : Do not start bulk transfer, if client tried to reconnect,
543 but failed due to active rpcs. Check for exp_abort_active_req
544 has been implemented.
548 Description: Prevent inconsistences between linux and lustre mount structures.
549 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
550 Make the sleep interruptible.
552 Severity : enhancement
554 Description: provide server to client comms path
555 Details : server to client communications path via new LDLM_SET_INFO rpc
559 Description: truncate starts GFP_FS allocation under transaction causing deadlock
560 Details : ldiskfs_truncate calls grab_cache_page which may start page
561 allocation under an open transaction. This may lead to
562 calling prune_icache with consequent lustre reentrance.
566 Description: send correctly lsm on open replay
567 Details : MDS is trust to LSM size on replay open, but client can set wrong size
572 Description: enable client interface failover
573 Details : When a child reconnects from another NID, properly update export
574 nid hash position and ldlm reverse import.
576 Severity : enhancement
578 Description: Communicate OST degraded/readonly state via statfs to MDS
579 Details : Flags in the statfs returned from OSTs indicate whether the
580 OST is in a degraded RAID state, or if the filesystem has
581 turned read-only after a filesystem error is detected.
586 Description: don't panic if EPROTO was hit when reading symlink
587 Details : correctly handling request reference in error cases.
589 Severity : enhancement
591 Description: Add LustreNetLink, a kernel-userspace communcation path.
593 Severity : enhancement
595 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
600 Description: access to llog context before init.
601 Details : move handling CATALOGS file at osc layer and forbid access to llog
605 Frequency : rare, only when client is writing to OST on local node
607 Description: Avoid deadlock for local client writes
608 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
609 memory freeing context. This allows OST threads to set the
610 PF_MEMALLOC flag on task structures in order to allocate memory
611 from reserved pools and complete IO.
612 Use GFP_HIGHUSER for OST allocations for non-local client writes,
613 so that the OST threads generate memory pressure and allow
614 inactive pages to be reclaimed.
616 Severity : enhancement
618 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
623 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
624 Details : move ldlm namespace creation in setup phase to avoid grab
625 _lprocfs_lock with cli_sem held.
629 Description: Temporarily disable grant shrink.
630 Details : Disable the feature for debugging.
633 Frequency : only when formatting test filesystems
635 Description: Unable to run several mkfs.lustre on loop devices at the same time
636 Details : mkfs.lustre returns error 256 on the concurrent loop devices
637 formatting. The solution is to proper handle the error.
639 Severity : enhancement
641 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
643 Severity : enhancement
645 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
650 Description: resolve race between obd_disconnect and class_disconnect_exports
651 Details : if obd_disconnect will be called to already disconnected export he
652 forget release one reference and osc module can't unloaded.
654 Severity : enhancement
656 Description: Allow tuning service thread via /proc
657 Details : For each service a new
658 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
659 created that can be used to set min/max thread counts, and get the
660 current number of running threads.
665 Descriptoin: don't return error if have partially created objects for file.
666 Details : lov_update_create_set uses set->set_success as index for created
667 objects, so if some requests will be failed, they will have hole
668 at end of array and qos_shrink_lsm for allocate correct lsm.
670 Severity : enhancement
672 Description: Update OFED support to 1.4.1
677 Description: Reduce small size read RPC
678 Details : Set read-ahead limit for every file and only do read-ahead when
679 available read-ahead pages are bigger than 1M to avoid small size
682 Severity : enhancement
684 Description: move AT tunable parameters for more consistent usage
685 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
687 Severity : enhancement
689 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
690 Details : allow skip disconnected ost for send statfs request and hide error
696 Description: fix racy locking of mballoc block bitmaps causing BUG
697 Details : The locking of the mballoc buddy bitmap and the in-memory
698 block bitmap was using two different spin locks in some
699 cases. This made it possible to incorrectly access the
700 mballoc bitmap while another process was modifying it,
701 causing a sanity assertion to fail. While no on-disk corruption
702 was reported, there was some risk of this happening.
705 Frequency : rare, on llog test 6
707 Descriptoin: don't allow connect to already connected import
708 Details : allowing connect to already connected import is hide connecting problem.
713 Descriptoin: allow kill process which wait statahead result
714 Details : for some reasons 'ls' can stick in waiting result from statahead,
715 in this case need way for kill this process.
717 Severity : enhancement
719 Description: Add state history info file, enhance import info file
720 Details : Track import connection state changes in a new osc/mdc proc file;
721 add overview-type data to the osc/mdc import proc file.
723 Severity : enhancement
725 Description: MDS create should not wait for statfs RPC while holding DLM lock.
727 Severity : enhancement
729 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
734 Descriptoin: OST grant too much space to client even there are not enough space.
735 Details : Client will shrink its grant cache to OST if there are no write
736 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
737 this grant cache if there are already not enough avaible space
738 (left_space < total_clients * 32M).
741 Frequency : start MDS on uncleanly shutdowned MDS device
743 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
744 Details : stay in waiting mds<>ost recovery finished produce random bugs
745 due race between two ll_sync thread for one lov target. send
746 ACTIVATE event only if connect realy finished and import have
750 Frequency : rare, connect and disconnect target at same time
752 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
753 Details : don't call obd_disconnect under lov_lock. this long time
754 operation and can block ptlrpcd which answer to connect request.
759 Descriptoin: don't lose wakeup for imp_recovery_waitq
760 Details : recover_import_no_retry or invalidate_import and import_close can
761 both sleep on imp_recovery_waitq, but we was send only one wakeup
765 Frequency : always with long access acl
767 Descriptoin: mds can't pack reply with long acl.
768 Details : mds don't control size of acl but they limited by reint/getattr
771 Severity : enhancement
773 Description: Update to SLES10 kernel-2.6.16.60-0.33.
775 Severity : enhancement
777 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
780 Frequency : start MDS on uncleanly shutdowned MDS device
782 Descriptoin: aborting recovery hang on MDS
783 Details : don't throttle destroy RPCs for the MDT.
786 Frequency : on remount
788 Description: external journal device not working after the remount
789 Details : clear dev_rdonly flag for external journal devices in
795 Description: shutdown vs evict race
796 Details : client_disconnect_export vs connect request race.
797 if client will evicted at this time - we start invalidate
798 thread without referece to import and import can be freed
802 Frequency : rare, need acl's on inode.
804 Description: client can't handle ost additional correctly
805 Details : if ost was added after client connected to mds client can have
806 hit lnet_try_match_md ... to big messages to wide striped files.
807 in this case need teach client to handle config events about add
808 lov target and update client max ea size at that event.
810 Severity : enhancement
812 Description: Changelogs
813 Details : Changelogs are a lightweight mechanism to track filesystem
814 metadata and namespace changes. The changelog is recorded
815 permanently on the MDTs, and is periodically "consumed" / purged
816 when records are no longer needed.
818 Severity : enhancement
820 Description: compact fld format with extents
821 Details : Store range of seq rather than every seq in FLD. Seq
822 controller update FLD rather than clients. In Case of CMD, mdt0
823 has FLD, all other metadata server act as non persistent proxy
824 for FLD queries and cache fld entries in fld cache.
829 Description: don't skip ost target if they assigned to file
830 Details : Drop slow OSCs if we can, but not for requested start idx.
831 This means "if OSC is slow and it is not the requested
832 start OST, then it can be skipped, otherwise skip it only
833 if it is inactive/recovering/out-of-space.
837 Description: more cleanup in mds_lov
838 Details : not send LOV EA under replay, we can't know about they size at this
839 time. Don't allow client connect to mds before any ost connected,
840 for avoid problems with LOV EA size and returning EIO to client.
842 Severity : enhancement
844 Description: Interoperability at server side (Disk interoperability)
846 Severity : enhancement
848 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
850 Severity : enhancement
852 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
854 Severity : enhancement
856 Description: New client IO stack (CLIO).
858 Severity : enhancement
860 Description: Commit on sharing. Eliminate inter-client dependencies between
861 uncommitted transactions by doing transaction commits.
862 Thereby clients may recovery independently.
865 Frequency : Create a symlink file with a very long name
867 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
868 Details : If there is no extra space in the request for early cancels,
869 ldlm_req_handles_avail() returns 0 instead of a negative value.
871 Severity : enhancement
873 Description: Add /proc entry for import status
874 Details : The mdc, osc, and mgc import directories now have
875 an import directory that contains useful import data for debugging
878 Severity : enhancement
880 Description: Re-disable certain /proc logging
881 Details : Enable and disable client's offset_stats, extents_stats and
882 extents_stats_per_process stats logging on the fly.
885 Frequency : Only on FC kernels 2.6.22+
887 Description: oops in statahead
888 Details : Do not drop reference count for the dentry from VFS when lookup,
889 VFS will do that by itself.
891 Severity : enhancement
893 Description: Generic /proc file permissions
894 Details : Set /Proc file permissions in a more generic way to enable non-
895 root users operate on some /proc files.
899 Description: Hitting mdc_commit_close() ASSERTION
900 Details : Properly handle request reference release in
901 ll_release_openhandle().
905 Description: quota recovery deadlock during mds failover
906 Details : This patch includes att18982, att18236, att18237 in bz14840.
908 1. fix osts hang when mds does failover with quotaon
909 2. prevent watchdog storm when osts threads wait for the
914 Frequency : only patchless client
915 Description: add workaround for race between add/remove dentry from hash
917 Severity : enhancement
919 Description: Allow OST glimpses to return PW locks
923 Description: LBUG when llog conf file is full
924 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
928 Description: Prevent import from entering FULL state when server in recovery
932 Description: service mount cannot take device name with ":"
933 Details : Only when device name contains ":/" will mount treat it as
939 Description: replace ptlrpcd with the statahead thread to interpret the async
940 statahead RPC callback
944 Frequency : on recovery
945 Description: I/O failures after umount during fail back
946 Details : if client reconnected to restarted server we need join to recovery
947 instead of find server handler is changed and process self
948 eviction with cancel all locks.
950 Severity : enhancement
952 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
954 Severity : enhancement
956 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
958 Severity : enhancement
960 Description: Upcall on Lustre log has been dumped
961 Details : Allow for a user mode script to be called once a Lustre log has
962 been dumped. It passes the filename of the dumped log to the
963 script, the location of the script can be specified via
964 /proc/sys/lnet/debug_log_upcall.
969 Description: avoid idr_remove called for id which is not allocated.
970 Details : Move assigment s_dev for clustered nfs to end of initialization,
971 to avoid problem with error handling.
976 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
977 Details : When connection is reused this not moved from CONN_UNUSED_HASH
978 into CONN_USED_HASH and this prodice warning when put connection
979 again in unused hash.
982 Severity : enhancement
984 Description: File striping can now be set to use an arbitrary pool of OSTs.
986 Severity : enhancement
988 Description: Export bytes_read/bytes_write count on OSC/OST.
992 Description: Early reply size mismatch, MGC loses connection
993 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
994 the connect flags are properly negotiated.
999 Description: quotas are not honored with O_DIRECT
1000 Details : all writes with the flag O_DIRECT will use grants which leads to
1001 this problem. Now using OBD_BRW_SYNC to guard this.
1005 Description: add quota statistics
1006 Details : 1. sort out quota proc entries and proc code.
1007 2. add quota statistics
1009 Severity : enhancement
1011 Description: enable quota support for HEAD.
1015 Description: Properly propagate oinfo flags from lov to osc for statfs
1016 Details : restore missing copy oi_flags to lov requests.
1018 Severity : enhancement
1020 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1021 : l_getgroups(8), lst(8), routerstat(8)
1022 Details : included man pages for llobdstat(8), llstat(8),
1023 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1025 Severity : enhancement
1027 Description: configure's --enable-quota should check the
1028 : kernel .config for CONFIG_QUOTA
1029 Details : configure is terminated if --enable-quota is passed but
1030 : no quota support is in kernel
1034 Description: Remove portals compatibility
1035 Details : Remove portals compatibility, not interoperable with releases
1040 Description: Resolve device initialization race
1041 Details : Prevent proc handler from accessing devices added to the
1042 obd_devs array but yet be intialized.
1044 Severity : enhancement
1046 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1048 Severity : enhancement
1050 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1055 Description: Using wrong pointer in osc_brw_prep_request
1056 Details : Access to array[-1] can produce panic if kernel compiled with
1057 CONFIG_PAGE_ALLOC enabled
1061 Description: Client runs out of low memory
1062 Details : Consider only lowmem when counting initial number of llap pages
1066 Description: *optional* service tags registration
1067 Details : if the "service tags" package is installed on a Lustre node
1068 When the filesystem is mounted, a local-node service tag will
1069 be created. See http://inventory.sun.com/ for more information
1070 about the Service Tags asset management system.
1074 Description: Kernel BUG tries to release flock
1075 Details : Lustre does not destroy flock lock before last reference goes
1076 away. So always drop flock locks when client is evicted and
1077 perform unlock regardless of successfulness of speaking to MDS.
1081 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1085 Description: sanity test 65a fails if stripecount of -1 is set
1086 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1091 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1092 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1095 Severity : enhancement
1097 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1101 Description: do not process already freed flock
1102 Details : flock can possibly be freed by another thread before it reaches
1103 to ldlm_flock_completion_ast.
1107 Description: LBUG during stress test
1108 Details : Need properly lock accesses the flock deadlock detection list.
1112 Description: oops in page fault handler
1113 Details : kernel page fault handler can return two special 'pages' in error
1114 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1118 Description: timeout with invalidate import.
1119 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1120 should be handled by ptlrpcd. This produce long age waiting and
1121 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1123 Severity : enhancement
1125 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1129 Description: enable MGS and MDT services start separately
1130 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1131 a co-located MGS without starting the MGS, which is a complement
1132 to 'nosvc' mount option.
1136 Frequency : after recovery
1137 Description: precreate to many object's after del orphan.
1138 Details : del orphan st in oscc last_id == next_id and this triger growing
1139 count of precreated objects. Set flag LOW to skip increase count
1140 of precreated objects.
1144 Frequency : rare, on clear nid stats
1145 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1146 Details : when clean nid stats sometimes try destroy live entry,
1147 and this produce panic in free.
1151 Description: Stack overflow during MDS log replay
1152 ease stack pressure by using a thread dealing llog_process.
1156 Description: wait until IO finished before start new when do lock cancel.
1157 Details : VM protocol want old IO finished before start new, in this case
1158 need wait until PG_writeback is cleared until check dirty flag and
1159 call writepages in lock cancel callback.
1161 Severity : enhancement
1163 Description: using special macro for print time and cleanup in includes.
1167 Description: mds_mfd_close() ASSERTION(rc == 0)
1168 Details : In mds_mfd_close(), we need protect inode's writecount change
1169 within its orphan write semaphore to prevent possible races.
1173 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1177 Frequency : rare, on shutdown ost
1178 Description: don't hit live lock with umount ost.
1179 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1180 use shrink_dcache_sb instead.
1184 Description: don't panic with use echo client
1185 Details : echo client pass NULL as client nid pointer and this produce null
1186 pointer dereference.
1190 Description: fix build on ppc32
1191 Details : compile code with -m64 flag produce wrong object file for ppc32.
1195 Description: add message levels for liblustreapi
1199 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1200 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1201 will result in an -EAGAIN and in the exit call path we call
1202 alloc_rr() which will with increasing aggressiveness attempt to
1203 aquire precreated objects on the minimum number of required OSCs.
1207 Description: Use old size assignment to avoid deadlock
1208 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1209 the scheduling while holding a spinlock. We do not need locking
1210 for size in ll_update_inode() because size is only updated from
1211 the MDS for directories or files without objects, so there is no
1212 other place to do the update, and concurrent access to such inodes
1213 are protected by the inode lock.
1217 Description: resolve "_IOWR redefined" build error on SLES10
1221 Description: dump the memory debugging after all modules are unloaded to
1222 suppress false negative in conf_sanity test 39
1224 Severity : enhancement
1226 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1231 Description: fix SLES kernel versioning
1232 Details : the kernel version for our SLES 10 kernel did not include a "-"
1233 before the "smp" at the end. while this was not a problem in
1234 general, it did mean that software trying to use the kernel
1235 version to try to detect a vendor specific kernel would fail.
1236 this was most evident by the OFED build scripts.
1240 Description: Don't update lov_desc members until making sure they are valid
1241 Details : When updating lov_desc members via proc fs, need fix their
1242 validities before doing the real update.
1246 Description: don't put request into delay list while invalidate in flight.
1247 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1248 import in flight. this produce timeout for invalidate and sometimes
1249 can cause stale data.
1253 Frequency : on ppc only
1254 Description: not convert ost objects for directory because it's not exist.
1255 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1259 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1261 Severity : enhancement
1263 Description: Update to support 2.6.22.14 vanilla kernel.
1267 Frequency : rare, on recovery
1268 Description: read procfs can produce deadlock in some situation
1269 Details : Holding lprocfs lock which send rpc can produce block for destroy
1270 obd objects and this also block reconnect with -EALREADY. This
1271 isn't fix all lprocfs bugs - but make it rare.
1273 Severity : enhancement
1275 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1278 Frequency : frequent on X2 node
1280 Description: mdc_set_open_replay_data LBUG
1281 Details : Set replay data for requests that are eligible for replay.
1285 Description: lustre_mgs: operation 101 on unconnected MGS
1286 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1287 MGC, and late on MGC cannot successfully connect to MGS and a lot
1288 of the error messages complaining that MGS is not connected.
1292 Frequency : on network error
1293 Description: panic with double free request if network error
1294 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1295 but it's true only for synchronus enqueue, for async enqueue
1296 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1299 Severity : enhancement
1301 Description: client-side metadata stat-ahead during readdir(directory readahead)
1302 Details : perform client-side metadata stat-ahead when the client detects
1303 readdir and sequential stat of dir entries therein
1306 Frequency : on start mds
1308 Description: Implement get_info(last_id) in obdfilter.
1311 Frequency : occasional
1313 Description: Correctly check stale fid, not start epoch if ost not support SOM
1314 Details : open with flag O_CREATE need set old fid in op_fid3 because
1315 op_fid2 was overwritten with new generated fid, but mds can answer
1316 with one of these two fids and both is not stale. Setattr
1317 incorrectly started an epoch and assume will be called
1318 done_writing, but without SOM done_writing ever being called.
1321 Frequency : rare, depends on device drivers and load
1323 Description: MDS or OSS nodes crash due to stack overflow
1324 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1325 In some cases, in conjunction with device drivers that use a lot
1326 of stack the MDS (or possibly OSS) service threads could overflow
1327 the stack. One change which was identified to consume additional
1328 stack has been reworked to avoid the extra stack usage.
1331 Frequency : occasional
1333 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1334 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1335 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1336 oscc_has_objects and see OSC has no objests and call
1337 oscc_internal_create to resend create request.
1339 Severity : enhancement
1341 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1343 Severity : enhancement
1345 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1348 Frequency : very rare
1350 Description: Fix replay if there is an un-replied request and open
1351 Details : In some cases, older replay request will revert the
1352 mcd->mcd_last_xid on MDS which is used to record the client's
1353 latest sent request.
1355 Severity : enhancement
1357 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1359 Severity : enhancement
1361 Description: Add rhel5 support to HEAD.
1363 Serverity : enhancement
1365 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1370 Description: Don't allow skipping OSTs if index has been specified.
1371 Details : Don't allow skipping OSTs if index has been specified, make
1372 locking in internal create lots better.
1376 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1377 Details : access to bitfield in structure is always rounded to long
1378 and this produce problem with not atomic change any bit.
1382 Description: Lustre make rpms failed.
1383 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1384 builds Lustre rpms from tarball.
1387 Frequency : rare on shutdown ost
1389 Description: If llog cancel was not send before clean_exports phase, this can
1390 produce deadlock in llog code.
1391 Details : If llog thread has last reference to obd and call class_import_put
1392 this produce deadlock because llog_cleanup_commit_master wait when
1393 last llog_commit_thread exited, but this never success because was
1394 called from llog_commit_thread.
1398 Description: allow userland application know is lost one of stripes.
1399 Details : fill lvb_blocks with error code on ost and return it to
1400 application if error flag found.
1404 Description: NULL lov_tgts causing MDS oops
1405 Details : more safe checks for NULL lov_tgts for avoid oops.
1407 Severity : enhancement
1409 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1413 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1414 Details : If an OST is down the MDS will hang indefinitely in
1415 obd_statfs_rqset() waiting for the statfs data. While for
1416 MDS QOS usage of statfs, it should not stuck in waiting.
1418 Severity : enhancement
1420 Description: remote_acl support
1421 Details : Support ACL-based permission check for remote user.
1422 Support setfacl/getfacl for remote user with the utils
1423 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1424 the system "{s,g}etfacl" utils.
1426 Severity : enhancement
1428 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1430 Severity : enhancement
1432 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1436 Description: make mgs_setparam() handle fsname containing dash
1437 Details : fsname containing a dash does not work with lctl conf_param
1439 Severity : enhancement
1441 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1443 Severity : enhancement
1445 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1449 Frequency : rare, at shutdown
1450 Description: access already free / zero obd_namespace.
1451 Details : if client_disconnect_export was called without force flag set,
1452 and exist connect request in flight, this can produce access to
1453 NULL pointer (or already free pointer) when connect_interpret
1454 store ocd flags in obd_namespace.
1458 Frequency : only at startup
1459 Description: not alloc memory with spinlock held.
1460 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1461 if any spinlock held.
1463 Severity : enhancement
1465 Description: make lustre randomly fail allocating memory
1466 Details : Make lustre randomly failed allocating memory for testing purpose.
1468 Severity : enhancement
1470 Description: lost problems with lov objid file
1471 Details : Fixes some scability and access to not inited memory problems
1472 in work with lov objdid file.
1477 Description: lfs find does not continue on file error
1478 Details : Continue other files processing when a file/dir is absent.
1482 Description: Inconsistent usage of lustre_pack_reply()
1483 Details : Standardize the usage of lustre_pack_reply() such that it
1484 always generate a CERROR on failure.
1487 Frequency : occasional
1489 Description: MDS hang or stay in waiting lock
1490 Details : If client receive lock with CBPENDING flag ldlm need send lock
1491 cancel as separate rpc, to avoid situation when cancel request
1492 can't processed due all i/o threads stay in wait lock.
1496 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1500 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1504 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1508 Description: Update to RHEL4 latest kernel.
1510 Severity : enhancement
1512 Description: Build SLES10 patchless client fails
1513 Details : The configure was broken by run ./configure with
1514 --with-linux-obj=.... argument for patchless client. When the
1515 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1516 can't search header adequately. Use absolute path such as
1517 -I($LINUX)/include instead.
1521 Description: interrupt oig_wait produce painc on resend.
1522 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1523 private set, and this produce situation when rq_ptlrpcd_data not
1524 copyed to new allocated request and triggered LBUG on assert
1525 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1526 ptlrpcd set if request is changed and can be safety changed to use
1531 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1532 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1533 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1534 these macros and replace them with their difinitions instead.
1538 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1539 Details : Update which_patch & target file for SLES10 latest kernel.
1542 Description: Remove client patches from SLES 10 kernel.
1543 Details : This causes SLES 10 clients to behave as patchless clients
1544 even on a Lustre-patched (server) kernel.
1546 Severity : enhancement
1548 Description: self-adjustable client's lru lists
1549 Details : use adaptive algorithm for managing client cached locks lru
1550 lists according to current server load, other client's work
1551 pattern, memory activities, etc. Both, server and client
1552 side namespaces provide number of proc tunables for controlling
1555 Severity : enhancement
1557 Description: light-weight GSS support
1558 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1559 and reduce performance overhead.
1561 Severity : enhancement
1563 Description: Linux keyring support
1564 Details : Support using service of Linux keyring for Lustre GSS internal
1565 context refresh/cache mechanism.
1569 Description: Fix errors in lfs documentation
1570 Details : Fixes man pages
1574 Description: don't use GFP_* in generic Lustre code.
1575 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1579 Description: obdclass is limited by single OBD_ALLOC(idarray)
1580 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1584 Description: warnings with build patchless client with vanila 2.6.19 and up
1585 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1586 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1590 Frequency : Only for SLES
1591 Description: sanity_quota fail test_1
1592 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1593 file, which makes TSTID[2] inunique.
1597 Frequency : Only for SLES9
1598 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1599 Details : Update target/ChangeLog/which_patch .
1603 Frequency : Only for test_10 in sanity-quota.sh
1604 Description: a bug in quota test code
1605 Details : checking a test flag in a wrong place
1609 Frequency : Only for test_13 in sanity-quota.sh
1610 Description: a bug in quota test script
1611 Details : shouldn't check overfull allocated quota
1615 Description: jbd statistics
1616 Details : Port older jbd statistics patch for sles10
1618 Severity : enhancement
1620 Description: Add printing inode info into message about error in writepage.
1625 Description: reply_lock_interpret crash due to race with it and lock cancel.
1626 Details : Do not replay locks that are being cancelled. Do not reference
1627 locks by their address during replay, just by their handle.
1631 Frequency : When flocks are used.
1632 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1633 Details : Flock locks might destroy just granted lock if it could be merged
1634 with another existing flock, this is done in completion handler,
1635 so teach ldlm_cli_enquque_fini that this is a valid case for
1641 Description: Oops in read and write path when failing to allocate lock.
1642 Details : Check if lock allocation failed and return error back.
1646 Description: lstripe command fails for valid OST index
1647 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1648 instead of lov->desc.ld_active_tgt_count.
1652 Description: sanity.sh failed test 103
1653 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1658 Description: df doesn't work properly if diskfs blocksize != 4K
1659 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1663 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1664 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1666 Severity : enhancement
1668 Description: Adaptive timeouts
1669 Details : RPC timeouts adapt to changing server load and network
1670 conditions to reduce resend attempts and improve recovery time.
1674 Description: llapi_file_create() does not allow some changes
1675 Details : add llapi_file_open() that allows specifying the mode and
1676 open flags, and also returns an open file handle.
1680 Description: merge and cleanup kernel patches.
1681 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1685 Description: Add journal checksum support.(Kernel part)
1686 Details : The journal checksum feature adds two new flags i.e
1687 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1688 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1689 indicates that the commit block contains the checksum for
1690 the blocks described by the descriptor blocks. Now commit
1691 record can be sent to disk without waiting for descriptor
1692 blocks to be written to disk. This behavior is controlled
1693 using JBD2_FEATURE_ASYNC_COMMIT flag.
1697 Description: OSS needs mutliple precreate threads
1698 Details : Add ability to start more than one create thread per OSS.
1702 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1703 Details : execute lfs setstripe on client
1707 Description: mds_obd_create error creating tmp object
1708 Details : When the user sets quota on root, llog will be affected and can't
1709 create files and write files.
1712 Frequency : Always on ia64 patchless client, and possibly others.
1714 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1715 Details : This allows the patchless client to be loaded on architectures
1716 without this export.
1720 Description: RedHat Update kernel for RHEL5
1721 Details : Add the kernel config file for RHEL5.
1725 Description: RedHat Update kernel for RHEL5
1726 Details : Modify the kernel config file more closer RHEL5.
1730 Description: Build failure against Centos5 (RHEL5)
1731 Details : Define PAGE_SIZE when it isn't present.
1735 Description: test_7 on sanity-quota.sh can't be used on separate machines
1736 Details : add facet to handle it
1740 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1741 Details : with NFS, the anon dentry's parent was set to itself in
1742 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1743 corresponding dentry other than use rec->ur_name.
1745 Severity : enhancement
1747 Description: lfs setstripe enhancement
1748 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1752 Description: enable data checksumming by default
1753 Details : enable checksum by default, allow --disable-checksum
1754 configure option and "-o nochecksum" mount option. Checksums
1755 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1759 Description: lustre support for RHEL5
1760 Details : Add support for RHEL5.
1764 Description: Client eviction due to failover config
1765 Details : after a connection loss, the lustre client should attempt to
1766 reconnect to the last active server first before trying the
1767 other potential connections.
1771 Description: when mds and osts use different quota unit(32bit and 64bit),
1772 quota will be released repeatly.
1773 Details : void sending multiple quota reqs to mds, which will keep the
1774 status between the reqs.
1778 Description: osts not allocated evenly to files
1779 Details : change the condition to increase offset_idx
1782 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1784 Description: Data corruption for OSTs that are formatted larger than 2TB
1786 Details : When generating the bio request for lustre file writes the
1787 sector number would overflow a temporary variable before being
1788 used for the IO. The data reads correctly from Lustre (which
1789 will overflow in a similar manner) but other file data or
1790 filesystem metadata may be corrupted in some cases.
1794 Description: Tune the kernel for good SCSI performance.
1795 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1796 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1801 Description: rewrite ext2-derived code in obdclass/uuid.c
1802 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1803 from scratch preserving functionality.
1807 Description: "lfs find -obd UUID" prints directories
1808 Details : "lfs find -obd UUID" will return all directory names instead
1809 of just file names. It is incorrect because the directories
1810 do not reside on the OSTs.
1814 Description: use i_size_read and i_size_write in 2.6 port
1815 Details : replace inode->i_size access with i_size_read/write()
1818 Frequency : when removing large files
1820 Description: scheduling issue during removal of large Lustre files
1821 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1822 It causes scheduling issues when removing large files (17TB in the
1826 Frequency : only with liblustre clients on XT3
1828 Description: evictions taking too long
1829 Details : allow llrd to evict clients directly on OSTs
1832 Frequency : only on ppc
1834 Description: /proc/fs/lustre/devices broken on ppc
1835 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1836 We should make sure the type of 'index' is loff_t and then cast
1837 explicitly as needed below. Do not assign an explicitly cast
1841 Frequency : only for rhel5
1843 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1844 Details : Modify the target file & which_kernel.
1846 Severity : enhancement
1848 Description: omit set fsid for export NFS
1849 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1850 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1854 Frequency : Always, for kernels after 2.6.16
1855 Description: Fix warning idr_remove called for id=.. which is not allocated.
1856 Details : Last kernels save old s_dev before kill super and not allow
1857 to restore from callback - restore it before call kill_anon_super.
1861 Description: buffer overruns could theoretically occur
1862 Details : llapi_semantic_traverse() modifies the "path" argument by
1863 appending values to the end of the origin string, and a
1864 overrun may occur. Adding buffer overrun check in liblustreapi.
1868 Description: Fix error on 'ls .' at the top of the Lustre mount.
1869 Details : Don't revalidate dentry if it is a root dentry.
1873 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1874 Details : Modify vm-tunables-rhel4.patch.
1878 Description: Kernel config for 2.6.18-vanilla.
1879 Details : Modify targets/2.6-vanilla.target.in.
1880 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1881 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1882 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1883 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1887 Description: improve handling recoverable errors
1888 Details : if request processig with error which can be recoverable on server
1889 request should be resend, otherwise page released from cache and
1894 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1895 Details : Modify target file & which_patch.
1896 A flaw was found in the IA32 system call emulation provided
1897 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1898 value could be stored in the %RAX register, which could trigger an
1899 out-of-bounds system call table access. An untrusted local user
1900 could exploit this flaw to run code in the kernel
1901 (ie a root privilege escalation). (CVE-2007-4573).
1905 Description: O_DIRECT bypasses client statistics.
1906 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1907 read_ahead_stats were updated. copied the stats section from
1908 osc_send_oap_rpc() into async_internal().
1912 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1916 Description: change order of libsysio includes
1917 Details : '#include sysio.h' should always come before '#include xtio.h'
1921 Description: handle "serious error: objid * already exists" more gracefully
1922 Details : If LAST_ID value on disk is smaller than the objects existing in
1923 the O/0/d* directories, it indicates disk corruption and causes an
1924 LBUG(). If the object is 0-length, then we should use the existing
1925 object. This will help to avoid a full fsck in most cases.
1927 Severity : enhancement
1929 Description: adapt the lustre_config script to support the upgrade case
1930 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1931 server targets to 1.6 in parallel.
1935 Description: To avoid grant space > avaible space when the disk is almost
1936 full. Without this patch you might see the error "grant XXXX >
1937 available" or some LBUG about grant, when the disk is almost
1939 Details : In filter_check_grant, for non_grant cache write, we should
1940 check the left space by if (*left > ungranted + bytes), instead
1941 of (*left > ungranted), because only we are sure the left space
1942 is enough for another "bytes", then the ungrant space should be
1943 increase. In client, we should update cl_avail_grant only there
1944 is OBD_MD_FLGRANT in the reply.
1948 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1949 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1950 described in CVE-2007-4573. This problem could allow untrusted
1951 local users to gain root access.
1954 Frequency : when using O_DIRECT and quotas
1956 Description: Incorrect file ownership on O_DIRECT output files
1957 Details : block usage reported by 'lfs quota' does not take into account
1958 files that have been written with O_DIRECT.
1963 Description: touch file failed when fs is not full
1964 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1965 otherwise we can get ENOSP while fs is not full.
1969 Description: parallel lock callbacks
1970 Details : Instead of sending blocking and completion callbacks as separated
1971 requests, adding them to a set and sending in parallel.
1974 Frequency : only for Cray XT3
1975 Bugzilla : 12829/13455
1976 Description: Changing primary group doesn't change the group lustre assigns to
1978 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1979 overridden with the primary group provided by the group upcall,
1980 whereas the supplied fsgid can be trusted if it is in the list of
1981 supplementary groups returned by the group upcall.
1983 Severity : enhancement
1985 Description: Allow masking D_WARNING, D_ERROR messages from console
1986 Details : Console messages can now be disabled via lnet.printk.
1990 Description: User code with malformed file open parameter crashes client node
1991 Details : Before packing join_file req, all the related reference should be
1992 checked carefully in case some malformed flags cause fake
1993 join_file req on client.
1997 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1998 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1999 mode is equal to the granted mode. However, it should also check
2000 that LDLM_FL_CP_REQD is not set.
2004 Description: Heavy nfs access might result in deadlocks
2005 Details : After ELC code landed, it is now improper to enqueue any mds
2006 locks under och_sem, because enqueue might want to decide to
2007 cancel open locks for same inode we are holding och_sem for.
2011 Description: Client eviction while running blogbench
2012 Details : A lot of unlink operations with concurrent I/O can lead to a
2013 deadlock causing evictions. To address the problem, the number of
2014 oustanding OST_DESTROY requests is now throttled to
2015 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2016 callbacks are processed in priority.
2020 Description: enable ACLs on MDS by default
2021 Details : ACLs must be enabled on MDS by default.
2024 Frequency : PPC/PPC64 only
2026 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2027 Details : fix duplicated definitions between asm-ppc64/types.h and
2028 lustre_types.h on PPC.
2031 Frequency : PPC/PPC64 only
2033 Description: asm-ppc/segment.h does not exist
2034 Details : fix compile issue on PPC.
2038 Description: better handle error messages in extents code
2041 Frequency : RHEL4 only
2043 Description: mkfs is very slow on IA64/RHEL4
2044 Details : A performance regression has been discovered in the MPT Fusion
2045 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2046 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2047 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2049 Severity : enhancement
2051 Description: SNMP support enhancement
2052 Details : Adding total number of sampled request for an MDS node in snmp
2055 Severity : enhancement
2057 Description: Optimize ldlm waiting list processing for PR extent locks
2058 Details : When processing waiting list for read extent lock and meeting read
2059 lock that is same or wider to it that is not contended, skip
2060 processing rest of the list and immediatelly return current
2061 status of conflictness, since we are guaranteed there are no
2062 conflicting locks in the rest of the list.
2066 Description: Time out and refuse to reconnect
2067 Details : When the failover node is the primary node, it is possible
2068 to have two identical connections in imp_conn_list. We must
2069 compare not conn's pointers but NIDs, otherwise we can defeat
2070 connection throttling.
2074 Description: port llog fixes from b1_6 into HEAD
2075 Details : Port llog reference couting and some llog cleanups from b1_6
2076 (bug 10800) into HEAD, for protect from panic and access to already
2077 free llog structures.
2081 Description: Detect stride IO mode in read-ahead
2082 Details : When a client does stride read, read-ahead should detect that and
2083 read-ahead pages according to the detected stride pattern.
2087 Description: data checksumming impacts single node performance
2088 Details : add support for several checksum algorithm. Currently, only CRC32
2089 and Adler-32 are supported. The checksum type can be changed on
2090 the fly via /proc/fs/lustre/osc/*/checksum_type.
2094 Description: use adler32 for page checksums
2095 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2100 Description: build for x2 fails
2101 Details : fix compile issue on Cray systems.
2105 Description: Properly match for duplicate locks
2106 Details : Due to different lock order from skiplists code, we need to
2107 traverse entire list for now
2110 Frequency : only on PPC/SLES10
2112 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2113 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2114 be defined. Add a hack in mkfs_lustre.c to work around this compile
2119 Description: LASSERT on MDS when client holding flock lock dies
2120 Details : ldlm pool logic depends on number of granted locks equal to
2121 number of released locks which is not true for flock locks, so
2122 just exclude such locks from consideration.
2126 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2127 Details : Use fsfilt_sync() for both the whole filesystem sync and
2128 individual file sync to eliminate dangerous inode locking
2129 with I_LOCK that can lead to a deadlock.
2133 Description: performance in 1.6.3
2134 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2135 sure that 1MB requests are not fragmented by the block layer.
2137 Severity : enhancement
2139 Description: organize the server-side client stats on per-nid basis
2140 Details : Change the structure of stats under obdfilter and mds to
2148 The "uuid"s file would list the uuids of _active_ exports.
2149 And the clear entry is to clear all stats and stale nids.
2151 Severity : enhancement
2153 Description: eliminate client locks in face of contention
2154 Details : file contention detection and lockless i/o implementation
2155 for contended files.
2159 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2162 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2164 Description: MDS may lose file striping (and hence file data) in some cases
2165 Details : If there are additional extended attributes stored on the MDS,
2166 in particular ACLs, SELinux, or user attributes (if user_xattr
2167 is specified for the client mount options) then there is a risk
2168 of attribute loss. Additionally, the Lustre file striping
2169 needs to be larger than default (e.g. striped over all OSTs),
2170 and an additional attribute must be stored initially in the
2171 inode and then increase in size enough to be moved to the
2172 external attribute block (e.g. ACL growing in size) for file
2177 Description: skiplist implementation simplification
2178 Details : skiplists are used to group compatible locks on granted list
2179 that was implemented as tracking first and last lock of each
2180 lock group the patch changes that to using doubly linked lists
2184 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2185 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2186 because LASSERT on both the data supplied by a client, and the
2187 data on disk is dangerous and incorrect.
2189 Severity : enhancement
2191 Description: Slow truncate/writes to huge files at high offsets.
2192 Details : Directly associate cached pages to lock that protect those pages,
2193 this allows us to quickly find what pages to write and remove
2194 once lock callback is received.
2198 Description: more ldlm soft lockups
2199 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2200 starve other threads from the resource lock for a long time in
2201 case of long waiting queue, so change the debug level from
2202 D_OTHER to the less frequently used D_INFO.
2204 Severity : enhancement
2206 Description: add -gid, -group, -uid, -user options to lfs find
2210 Description: Hung threads in invalidate_inode_pages2_range
2211 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2212 once one is completed. As a result, some RPCs are stuck in the
2213 queue and are never sent.
2217 Description: filter threads hungs on waiting journal commit
2218 Details : Cleanup filter group llog code, then only filter group llog will
2219 be only created in the MDS/OST syncing process.
2223 Description: Procfs and llog threads access destoryed import sometimes.
2224 Details : Sync the import destoryed process with procfs and llog threads by
2225 the import refcount and semaphore.
2227 Severity : enhancement
2229 Description: openlock cache of b1_6 port to HEAD
2234 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2235 Details : If insertion of an extent fails, then discard the inode
2236 preallocation and free data blocks else it can lead to duplicate
2241 Description: don't always update ctime in ext3_xattr_set_handle()
2242 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2243 In some cases the ctime should not be updated, for example for
2244 2.0->1.8 compatibility it is necessary to delete an xattr and it
2245 should not update the ctime.
2249 Bugzilla : 15713/16362
2250 Description: Assertion in iopen_connect_dentry in 1.6.3
2251 Details : looking up an inode via iopen with the wrong generation number can
2252 populate the dcache with a disconneced dentry while the inode
2253 number is in the process of being reallocated. This causes an
2254 assertion failure in iopen since the inode's dentry list contains
2255 both a connected and disconnected dentry.
2259 Description: assertion failure in ldlm_handle2lock()
2260 Details : fix a race between class_handle_unhash() and class_handle2object()
2261 introduced in lustre 1.6.5 by bug 13622.
2266 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2267 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2272 Description: oops in sd_iostats_seq_show()
2273 Details : unloading/reloading the scsi low level driver triggers a kernel
2274 bug when trying to access the sd iostat file.
2279 Description: Kernel panics during QLogic driver reload
2280 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2281 patch, causing memory corruption.
2286 Description: journal_dev option does not work in b1_6
2287 Details : pass mount option during pre-mount.
2289 Severity : enhancement
2291 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2292 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2293 extent information of a file. It can be used to map logical blocks
2294 in a file to physical blocks in the block device.
2298 Description: LDLM soft lockups - improvement
2299 Details : It is be possible to send the lock handle along with each read
2300 or write request because the client is already doing a lock match
2301 itself so there isn't any reason the OST should have to re-do that
2305 Frequency : only on Cray X2
2307 Description: X2 build failures
2308 Details : fix build failures on Cray X2.
2312 Description: xid & resent requests
2313 Details : Initialize RPC XID from clock at startup (randomly if clock is
2316 Severity : enhancement
2318 Description: Add lustre_start utility to start or stop multiple Lustre servers
2323 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2324 Details : In case of memory pressure, list_del() can be called twice on
2325 req->rq_history_list, causing a kernel oops.
2329 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2330 Details : fix stack overflow in the distributed lock manager by defering
2331 export eviction after a failed AST to the elt thread instead of
2332 handling it in the dlm interpret routine.
2336 Description: Convert some comments to new format.
2337 Details : Update documenting comments to match doxygen conventions.
2341 Description: Grammar fixes.
2342 Details : A couple of trivial spelling fixes.
2346 Description: OSD_COUNTERS-mandatory
2347 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2348 to simplify the code.
2352 Description: simplify cmm_device freeing logic.
2353 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2357 Description: Add lockdep support to dt_object_operations locking interface.
2358 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2359 parameter indicating lock ordering. Update mdd code to use new
2364 Description: Introduce failloc constants for lockless IO tests.
2365 Details : Add two new failloc constants to test lockless IO. Only one of
2366 them in implemented---another is checked in yet to be landed
2371 Description: Add lockdep support for inode mutex.
2372 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2373 in the situations where Lustre has to lock more than one inode
2378 Description: Add optional invariants checking support.
2379 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2380 configure switch. This macro is to be used for consistency and
2381 sanity checks that are too expensive to be left in `production'
2386 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2387 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2392 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2393 Details : Use spin_lock_nested() in (the only) situation where more than
2394 one ldlm_lock is locked simultaneously. Also, fix possible
2395 dead-lock in ldlm_lock_change_resource() by enforcing particular
2400 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2401 Details : Instead of specifying each ldlm_lock call-back through separate
2402 parameter, wrap them into struct ldlm_callback_suite.
2406 Description: Kill join_lru obd method and its callers.
2407 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2408 in memory---a requirement implemented through ->o_join_lru() obd
2409 method in HEAD. Remove this method and its users.
2413 Description: Add asynchronous ldlm ENQUEUE completion handler.
2414 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2415 case that stock ldlm_completion_ast() cannot handle as it waits
2416 until lock is granted. Introduce new ldlm_completion_ast_async()
2417 for this. Also comment ldlm_completion_ast().
2421 Description: ldlm_error <-> errno conversion.
2422 Details : Add functions to map (rather arbitrary) between LDLM error codes
2423 and standard errno values. CLIO needs this to prevent LDLM specific
2424 constants from escaping ldlm and osc.
2428 Description: Kill unused ldlm_handle2lock_ns() function.
2429 Details : Kill unused ldlm_handle2lock_ns() function.
2433 Description: Add lu_ref support to ldlm_lock
2434 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2435 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2436 and ldlm_resource. LDLM interface has to be changed, because code that
2437 releases a reference on a lock, has to "know" what reference this is.
2438 In the most frequent case
2440 lock = ldlm_handle2lock(handle);
2442 LDLM_LOCK_PUT(lock);
2444 no changes are required. When any other reference (received _not_ from
2445 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2446 instead of LDLM_LOCK_PUT().
2448 Arguably, changes are pervasive, and interface requires some discipline
2449 for proper use. On the other hand, it was very instrumental in finding
2450 a few leaked lock references.
2454 Description: Add ldlm_lock_addref_try().
2455 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2456 attempts to addref a lock that might be being canceled
2461 Description: Add ldlm_weigh_callback().
2462 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2463 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2464 hard-coded `number of pages' logic.
2468 Description: Add lockdep annotations to llog code.
2469 Details : Use appropriately tagged _nested() locking calls in the places
2470 where llog takes more than one ->lgh_lock lock.
2474 Description: Add loi_kms_set().
2475 Details : Wrap kms updates into a helper function.
2479 Description: Constify instances of struct lsm_operations.
2480 Details : Constify instances of struct lsm_operations.
2484 Description: lu_conf support.
2485 Details : On a server, a file system object is uniquely identified
2486 by a fid, which is sufficient to locate and load all object
2487 state (inode). On a client, on the other hand, more data are
2488 necessary instantiate an object. Change lu_object_find() and
2489 friends to take additional `lu_conf' argument describing object.
2490 Typically this includes layout information.
2494 Description: lu_context fixes.
2495 Details : Introduce new lu_context functions that are needed on the client
2496 side, where some system threads (ptlrpcd) are shared by multiple
2497 modules, and so cannot be stopped during module shutdown.
2501 Description: Add start and stop methods to lu_device_type_operations.
2502 Details : Introduce two new methods in lu_device_type_operations, that are
2503 invoked when first instance of a given type is created and last one
2504 is destroyed respectively. This is need by CLIO.
2508 Description: Add lu_ref support to struct lu_device.
2509 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2510 track leaked references.
2514 Description: Introduce lu_kmem_descr.
2515 Details : lu_kmem_descr and its companion interface allow to create
2516 and destroy a number of kmem caches at once.
2520 Description: Fix lu_object finalization race.
2521 Details : Fix a race between lu_object_find() finding an object and its
2522 concurrent finalization. This race is (most likely) not possible
2523 on the server, but might happen on the client.
2527 Description: Introduce lu_ref interface.
2528 Details : lu_ref is a debugging module allowing to track references to
2529 a given object. It is quite cpu expensive, and has to be
2530 explicitly enabled with --enable-lu_ref. See usage description
2535 Description: Factor lu_site procfs stats into a separate function.
2536 Details : Separate lu_site stats printing code into a separate function
2537 that can be reused on a client.
2541 Description: Constify instances of struct {lu,dt,md}_device_operations.
2542 Details : Constify instances of struct {lu,dt,md}_device_operations.
2546 Description: Introduce struct md_site and move meta-data specific parts of
2547 struct lu_site here.
2548 Details : Move md-specific fields out of struct lu_site into special struct
2549 md_site, so that lu_site can be used on a client.
2553 Description: Kill mdd_lov_destroy().
2554 Details : Remove unused mdd code.
2558 Description: Add st_block checking to multistat.c.
2559 Details : Add st_block checking to multistat.c.
2563 Description: Add lu_ref support to struct obd_device.
2564 Details : Add lu_ref tracking to obd_device.
2568 Description: Kill obd_set_fail_loc().
2569 Details : Remove unused code.
2573 Description: Add special type for ptlrpc_request interpret functions.
2574 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2575 there. Actual usage will be in CLIO.
2579 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2580 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2581 doesn't work with lockdep.
2585 Description: Add rwv.c test program.
2586 Details : New testing program exercising readv(2) and writev(2) (Qian).
2590 Description: Add sendfile.c test program.
2591 Details : New testing program exercising sendfile(2) (Jay).
2595 Description: Ratelimit a message that can be very frequent.
2596 Details : Ratelimit a memory allocation failure message that can
2601 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2602 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2606 Description: Liblustre build fixes.
2607 Details : Liblustre build fixes.
2611 Description: libcfs: add cfs_{need,cond}_resched() interface.
2612 Details : libcfs: add cfs_{need,cond}_resched() definition and
2613 implementations for Linux, NT, and liblustre.
2615 Severity : enhancement
2617 Description: More exported tunables for mballoc
2618 Details : Add support for tunable preallocation window and new tunables for
2619 large/small requests
2623 Description: Detect corruption of block bitmap and checking for preallocations
2624 Details : Checks validity of on-disk block bitmap. Also it does better
2625 checking of number of applied preallocations. When corruption is
2626 found, it turns filesystem readonly to prevent further corruptions.
2630 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2631 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2632 window to 0. If it does hit cache miss, and read-pattern is still
2633 stride-io mode, does not reset the stride window, but also does not
2634 increase the stride window length in this case.
2638 Frequency : only for big-endian servers
2639 Description: Check if system is big-endian while mounting fs with extents feature
2640 Details : Mounting a filesystem with extents feature will fail on big-endian
2641 systems since ext3-based ldiskfs is not supported on big-endian
2642 systems. This can be over-riden with "bigendian_extents" mount option.
2644 Severity : enhancement
2646 Description: The root squash functionality
2647 Details : A security feature, which is to prevent users from being able
2648 to mount lustre on their desktop, run as root, and delete
2649 all of the files in the filesystem. The goal is accomplished by
2650 remapping user id (UID) and group id (GID) of the root user to
2651 a UID and GID specified by the system administartor via Lustre
2652 configuration management server (MGS). The functionality also
2653 allows to specify sets of clients for which the remapping does
2658 Description: Excessive recovery window
2659 Details : With AT enabled, the recovery window can be excessively long (6000+
2660 seconds). To address this problem, we no longer use
2661 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2662 timeout no longer depends on the service time, it is set to
2663 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2664 time via pb_service_time.
2668 Description: Watchdog triggered on MDS failover
2669 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2670 the OSTs know that the MDS "UUID" can be reused for the same export
2671 from a different NID, so we do not need to wait for the export to be
2675 Frequency : rare, only if using MMP with Linux RAID
2677 Description: MMP doesn't work with Linux RAID
2678 Details : While using HA for Lustre servers with Linux RAID, it is possible
2679 that MMP will not detect multiple mounts. To make this work we
2680 need to unplug the device queue in RAID when the MMP block is being
2681 written. Also while reading the MMP block, we should read it from
2682 disk and not the cached one.
2684 Severity : enhancement
2686 Description: open file using fid
2687 Details : A file can be opened using just its fid, like
2688 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2691 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2694 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2695 Details : add proper locking for creation of jbd_4k slab cache
2699 Description: MMP check in ext3_remount() fails without displaying any error
2700 Details : When multiple mount protection fails during remount, proper error
2703 Severity : enhancement
2705 Description: Allow stripe size to be up to 4G-64k
2706 Details : Fix math logic to allow large stripe sizes.
2710 Description: add check for >8TB ldiskfs filesystems
2711 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2712 Don't allow >8TB ldiskfs filesystems to be mounted without
2713 force_over_8tb mount option
2718 Descriptoin: OSS complains "too many blocks" and crashes
2719 Details : In filter_commitrw_write():
2720 1. lquota_chkquota();
2721 2. obdo_from_inode();
2722 3. lquota_pending_commit();
2723 "2." has changed uid and gid in this case, then caused
2724 lquota_chkquota() and lquota_pending_commit() to use
2725 different uid and gid. That is the root of the bug.
2729 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2730 Details : Detect and print out the 'stripe count = -1' case properly.
2734 Description: Out or order replies might be lost on replay
2735 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2736 request with tid smaller than one currently being added, add it
2737 to the start, not end of the list.
2739 --------------------------------------------------------------------------------
2741 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2743 * Support for kernels:
2744 2.6.5-7.283 (SLES 9),
2745 2.6.9-55.EL (RHEL 4),
2746 2.6.16.46-0.14 (SLES 10),
2747 2.6.18.8 vanilla (kernel.org)
2748 * Client support for unpatched kernels:
2749 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2750 2.6.16 - 2.6.22 vanilla (kernel.org)
2751 * Due to recently discovered recovery problems, we do not recommend
2752 using patchless RHEL 4 clients with this or any earlier release.
2753 * Recommended e2fsprogs version: 1.39.cfs8
2754 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2755 * Starting with this release, the ldiskfs backing filesystem required
2756 by Lustre is now in its own package, lustre-ldiskfs. This package
2757 should be installed. It is versioned separately from Lustre and
2758 may be released separately in future.
2762 Description: block reactivating mgc import until all deactivates complete
2763 Details : Fix race when failing back MDT/MGS to itself (testing)
2766 Frequency : at statup only
2768 Description: mds_lov_synchronize race leads to various problems
2769 Details : simultaneous MDT->OST connections at startup can cause the
2770 sync to abort, leaving the OSC in a bad state.
2772 Severity : enhancement
2774 Description: add optional extra BUILD_VERSION info
2775 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2776 to override the lustre version.
2779 Frequency : 2.6.18 servers only
2781 Description: ll_kern_mount() doesn't release the module reference
2782 Details : The ldiskfs module reference count never drops down to 0
2783 because ll_kern_mount() doesn't release the module reference.
2788 Description: server LBUG when using old ost_num_threads parameter
2789 Details : Accept the old ost_num_threads parameter but warn that it
2790 is deprecated, and fix an off-by-one error that caused an LBUG.
2795 Description: Transient SCSI error results in persistent IO issue
2796 Details : iobuf->dr_error is not reinitialized to 0 between two
2800 Frequency : sometimes when underlying device returns I/O errors
2802 Description: OSTs not going read-only during write failures
2803 Details : OSTs are not remounted read-only when the journal commit threads
2804 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2805 instead of the ext3 wrappers.
2809 Description: poor connect scaling with increasing client count
2810 Details : Don't run filter_grant_sanity_check for more than 100 exports
2811 to improve scaling for large numbers of clients.
2814 Frequency : SLES10 only
2816 Description: sanity-quota.sh quotacheck failed: rc = -22
2817 Details : Quotas cannot be enabled on SLES10.
2820 Frequency : liblustre clients only
2822 Description: getdirentries does not give error when run on compute nodes
2823 Details : getdirentries does not fail when the size specified as an argument
2824 is too small to contain at least one entry
2826 Severity : enhancement
2828 Description: Add LNET router traceability for debug purposes
2829 Details : If a checksum failure occurs with a router as part of the
2830 IO path, the NID of the last router that forwarded the bulk data
2831 is printed so it can be identified.
2836 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2837 Details : Due to a race condition, liblustre clients were occasionally
2838 evicted incorrectly.
2840 Severity : enhancement
2842 Description: lfs setstripe use optional parameters instead of postional
2845 Severity : enhancement
2847 Description: Nanosecond timestamp support for ldiskfs
2848 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2849 resolution timestamps. There is not yet support for this at
2850 the Lustre filesystem level.
2853 Frequency : during server recovery
2855 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2856 Details : request with rq_no_resend flag not awake l_wait_event if they get
2860 Frequency : nfs export on patchless client
2862 Description: connectathon hang when test nfs export over patchless client
2863 Details : Disconnected dentry cannot be found with lookup, so we do not need
2864 to unhash it or make it invalid
2867 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2868 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2869 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2874 Description: open req refcounting wrong on reconnect
2875 Details : If reconnect happened between getting open reply from server and
2876 call to mdc_set_replay_data in ll_file_open, we will schedule
2877 replay for unreferenced request that we are about to free.
2878 Subsequent close will crash in variety of ways.
2879 Check that request is still eligible for replay in
2880 mdc_set_replay_data().
2885 Description: disable writes to filesystem when reading health_check file
2886 Details : the default for reading the health_check proc file has changed
2887 to NOT do a journal transaction and write to disk, because this
2888 can cause reads of the /proc file to hang and block HA state
2889 checking on a healthy but otherwise heavily loaded system. It
2890 is possible to return to the previous behaviour during configure
2891 with --enable-health-write.
2893 Severity : enhancement
2895 Description: 64-bit inode version
2896 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2897 to the inode. This will be required for version-based recovery.
2902 Description: MDS fails to start if a duplicate client export is detected
2903 Details : in some rare cases it was possible for a client to connect to
2904 an MDS multiple times. Upon recovery the MDS would detect this
2905 and fail during startup. Handle this more gracefully.
2907 Severity : enhancement
2909 Description: Add -o localflock option to simulate old noflock
2911 Details : This will achieve local-only flock/fcntl locks
2917 Description: log_commit_thread vs filter_destroy race leads to crash
2918 Details : Take import reference before releasing llog record semaphore
2923 Description: Wrong request locking in request set processing
2924 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2925 imp_delayed_list, in this place should be used imp_lock.
2928 Frequency : when reconnection
2930 Description: Grant Leak when osc reconnect to OST
2931 Details : When osc reconnect ost, OST(filter) should check whether it
2932 should grant more space to client by comparing fed_grant and
2933 cl_avail_grant, and return the granted space to client instead
2934 of "new granted" space, because client will call osc_init_grant
2935 to update the client grant space info.
2938 Frequency : when client reconnect to OST
2940 Description: Grant Leak when osc do resend and replay bulk write
2941 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2942 bulk write request, because the grant info while be sync between
2943 OSC and OST when reconnect, and we should ignore the grant info
2944 these of resend/replay write req.
2949 Description: Grant space more than available left space sometimes.
2950 Details : When then OST is about to be full, if two bulk writing from
2951 different clients came to OST. Accord the avaliable space of the
2952 OST, the first req should be permitted, and the second one
2953 should be denied by ENOSPC. But if the seconde arrived before
2954 the first one is commited. The OST might wrongly permit second
2955 writing, which will cause grant space > available space.
2958 Frequency : when client is evicted
2960 Description: Grant might be wrongly erased when osc is evicted by OST
2961 Details : when the import is evicted by server, it will fork another
2962 thread ptlrpc_invalidate_import_thread to invalidate the
2963 import, where the grant will be set to 0. While the original
2964 thread will update the grant it got when connecting. So if
2965 the former happened latter, the grant will be wrongly errased
2966 because of this race.
2971 Description: Checking Stale with correct fid
2972 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2973 to check whether it is stale, because sometimes, we want the
2974 enqueue happened anyway, and op_data.fid2 will not be initialized.
2976 Severity : enhancement
2978 Description: update patchless client
2979 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2982 Frequency : only with 2.4 kernel
2984 Description: random memory corruption
2985 Details : size of struct ll_inode_info is to big for union inode.u and this
2986 can be cause of random memory corruption.
2991 Description: Memory leak in recovery
2992 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2993 It should also check obd_fail before fsfilt_start, otherwise if
2994 fsfilt_start return -EROFS,(failover mds during mds recovery).
2995 then the req will return with repmsg->transno = 0 and rc = EROFS.
2996 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2997 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2998 be freed no matter whether fsfilt_commit success or not.
3001 Frequency : only with huge count clients
3003 Description: Prevents from taking the superblock lock in llap_from_page for
3005 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3006 for llap_from_page call in ll_removepage prevents from taking the
3007 superblock lock for a soon died page.
3012 Description: Not check open intent error before release open handle
3013 Details : in some rare cases, the open intent error is not checked before
3014 release open handle, which may cause
3015 ASSERTION(open_req->rq_transno != 0), because it tries to release
3016 the failed open handle.
3021 Description: Set cat log bitmap only after create log success.
3022 Details : in some rare cases, the cat log bitmap is set too early. and it
3023 should be set only after create log success.
3027 Description: Accessing a block bevice can re-enable I/O when Lustre is
3028 tearing down a device.
3029 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3034 Description: service threads may hog cpus when there are a lot of requests
3035 Details : Insert cond_resched to give other threads a chance to use some CPU
3040 Description: the cat log was not initialized in recovery
3041 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3042 unlink log on mds will not be initialized until mds post
3043 recovery. And also in mds post recovery, the unlink log will
3044 initialization will be done asynchronausly, so there will be race
3045 between add unlink log and unlink log initialization.
3049 Description: brw_stats were being printed incorrectly
3050 Details : brw_stats were being printed as log2 but all of them were not
3051 recorded as log2. Also remove some code duplication arising from
3052 filter_tally_{read,write}.
3056 Frequency : rare, only in recovery.
3057 Description: ASSERTION(req->rq_type != LI_POISON) failed
3058 Details : imp_lock should be held while iterating over imp_sending_list for
3059 prevent destroy request after get timeout in ptlrpc_queue_wait.
3063 Description: replay-single.sh test 52 fails
3064 Details : A lock's skiplist need to be cleanup when it being unlinked
3065 from its resource list.
3069 Description: Short directio read returns full requested size rather than
3071 Details : Direct I/O operations should return actual amount of bytes
3072 transferred rather than requested size.
3074 Severity : enhancement
3076 Description: metadata RPC reduction (e.g. for rm performance)
3077 Details : decrease the amount of synchronous RPC between clients and servers
3078 by canceling conflicing lock before the operation on the client
3079 and packing thier handles into the main operation RPC to server.
3081 Severity : enhancement
3083 Description: Async OSC create to avoid the blocking unnecessarily.
3084 Details : If an OST has no remaining object, system will block on the
3085 creation when it needs to create a new object on this OST. Now,
3086 ways use pre-created objects when available, instead of blocking on
3087 an empty osc while others are not empty. If we must block, we
3088 block for the shortest possible period of time.
3092 Description: improve handling recoverable errors
3093 Details : if request processig with error which can be recoverable on server
3094 request should be resend, otherwise page released from cache and
3097 Severity : enhancement
3099 Description: refine locking for avoid write wrong info into lov_objid file
3100 Details : fix possible races with add new target and write/update data in
3103 --------------------------------------------------------------------------------
3105 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3110 Frequency : on some architectures
3112 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3113 Details : Uninitialized flags sometimes cause configuration commands to
3117 Frequency : patchless clients only
3119 Description: missing __iget() symbol export
3120 Details : The __iget() symbol export is missing. To avoid the need for
3121 this on patchless clients the deathrow inode reaper is turned
3122 off, and we depend on the VM to clean up old inodes. This
3123 dependency was during via the fix for bug 12181.
3125 --------------------------------------------------------------------------------
3127 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3129 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3130 INTEROPERATE with older versions automatically. In many cases a
3131 special upgrade step is needed. Please read the
3132 user documentation before upgrading any part of a live system.
3133 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3134 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3135 * WARNING: Lustre configuration and startup changes are required with
3136 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3140 Severity : enhancement
3142 Description: MountConf
3143 Details : Lustre configuration is now managed via mkfs and mount
3144 commands instead of lmc and lconf. New obd types (MGS, MGC)
3145 are added for dynamic configuration management. See
3146 https://mail.clusterfs.com/wikis/lustre/MountConf for
3149 Severity : enhancement
3151 Description: dynamic OST addition
3152 Details : OSTs can now be added to a live filesystem
3154 Severity : enhancement
3156 Description: startup order invariance
3157 Details : MDTs and OSTs can be started in any order. Clients only
3158 require the MDT to complete startup.
3160 Severity : enhancement
3162 Description: parallel, asynchronous orphan cleanup
3163 Details : orphan cleanup is now performed in separate threads for each
3164 OST, allowing parallel non-blocking operation.
3166 Severity : enhancement
3168 Description: optimized stripe assignment
3169 Details : stripe assignments are now made based on ost space available,
3170 ost previous usage, and OSS previous usage, in order to try
3171 to optimize storage space and networking resources.
3173 Severity : enhancement
3175 Description: Permanently set tunables
3176 Details : All writable /proc/fs/lustre tunables can now be permanently
3177 set on a per-server basis, at mkfs time or on a live system.
3179 Severity : enhancement
3181 Description: Lustre message v2
3182 Details : Add lustre message format v2.
3184 Severity : enhancement
3186 Description: client OST exclusion list
3187 Details : Clients can be started with a list of OSTs that should be
3188 declared "inactive" for known non-responsive OSTs.
3192 Description: SPEC SFS validation failure on NFS v2 over lustre.
3193 Details : Changes the blocksize for regular files to be 2x RPC size,
3194 and not depend on stripe size.
3196 Severity : enhancement
3198 Description: Multiple MD RPCs in flight.
3199 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3200 To avoid overly-overloading MDS, introduce a limit on number of
3201 MDS RPCs in flight for a single client and add /proc controls
3202 to adjust this limit.
3204 Severity : enhancement
3206 Description: client read/write statistics
3207 Details : Add client read/write call usage stats for performance
3208 analysis of user processes.
3209 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3210 file access. extents_stats shows chunk size distribution.
3211 extents_stats_per_process show chunk size distribution per
3214 Severity : enhancement
3216 Description: mds statistics
3217 Details : Add detailed mds operations statistics in
3218 /proc/fs/lustre/mds/*/stats.
3222 Description: Failure of copying files with lustre special EAs.
3223 Details : Client side always return success for setxattr call for lustre
3224 special xattr (currently only "trusted.lov").
3229 Description: Refcount LNET uuids
3230 Details : The global LNET uuid list grew linearly with every startup;
3231 refcount repeated list entries instead of always adding to
3234 Severity : enhancement
3236 Description: Dynamic service threads
3237 Details : Within a small range, start extra service threads
3238 automatically when the request queue builds up.
3241 Frequency : mixed-endian client/server environments
3243 Description: mixed-endian crashes
3244 Details : The new msg_v2 system had some failures in mixed-endian
3247 Severity : enhancement
3249 Description: Easy OST removal
3250 Details : OSTs can be permanently deactivated with e.g. 'lctl
3251 conf_param lustre-OST0001.osc.active=0'
3253 Severity : enhancement
3255 Description: MGS proc entries
3256 Details : Added basic proc entries for the MGS showing what filesystems
3259 Severity : enhancement
3261 Description: provide MGS failover
3262 Details : Added config lock reacquisition after MGS server failover.
3264 Severity : enhancement
3266 Description: add Linux 2.4 support
3267 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3271 Description: a large application tries to do I/O to the same resource and dies
3272 in the middle of it.
3273 Details : Check the req->rq_arrival time after the call to
3274 ost_brw_lock_get(), but before we do anything about
3275 processing it & sending the BULK transfer request. This
3276 should help move old stale pending locks off the queue as
3277 quickly as obd_timeout.
3280 Frequency : when an incorrect nid is specified during startup
3282 Description: ptlrpc connect to non-existant node causes kernel crash
3283 Details : LNET can't be re-entered from an event callback, which
3284 happened when we expire a message after the export has been
3285 cleaned up. Instead, hand the zombie cleanup off to another
3288 Severity : enhancement
3290 Description: plain/inodebits lock performance improvement
3291 Details : Grouping plain/inodebits in granted list by their request modes
3292 and bits policy, thus improving the performance of search through
3296 Frequency : only if OST filesystem is corrupted
3298 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3299 Details : for a short time RPCs with bulk IO are in the replay list,
3300 but replay of bulk IOs is unimplemented. If the OST filesystem
3301 is corrupted due to disk cache incoherency and then replay is
3302 started it is possible to trip an assertion. Avoid putting
3303 committed RPCs into the replay list at all to avoid this issue.
3306 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3308 Description: System hang on startup
3309 Details : This bug allowed the liblustre (e.g. catamount) client to
3310 return to the app before handling all startup RPCs. This
3311 could leave the node unresponsive to lustre network traffic
3312 and manifested as a server ptllnd timeout.
3314 Severity : enhancement
3316 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3317 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3318 boolean to control whether to print peer debug info when a
3319 client's RPC times out.
3322 Frequency : only for kernels with patches from Lustre below 1.4.3
3324 Description: Remove old rdonly API
3325 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3328 Frequency : only for devices with external journals
3330 Description: Set external device read-only also
3331 Details : During a commanded failover stop, we set the disk device
3332 read-only while the server shuts down. We now also set any
3333 external journal device read-only at the same time.
3336 Frequency : when upgrading from 1.4 while trying to change parameters
3338 Description: The wrong (new) MDC name was used when setting parameters for
3339 upgraded MDT's. Also allows changing of OSC (and MDC)
3340 parameters if --writeconf is specified at tunefs upgrade time.
3343 Frequency : when setting specific ost indicies
3345 Description: QOS code breaks on skipped indicies
3346 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3347 created with --index need not be sequential.
3349 Severity : enhancement
3351 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3352 Details : The uninit_groups feature works in conjunction with the kernel
3353 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3354 up the pass1 processing of e2fsck. This is a read-only feature
3355 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3356 mount filesystems that have had this feature enabled.
3358 Severity : enhancement
3360 Description: Improve multi-block allocation algorithm to avoid fragmentation
3361 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3362 allocation locality and avoid filesystem fragmentation.
3364 ------------------------------------------------------------------------------
3366 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3368 * Support for kernels:
3369 2.4.21-47.0.1.EL (RHEL 3)
3370 2.6.5-7.283 (SLES 9)
3371 2.6.9-42.0.10.EL (RHEL 4)
3372 2.6.12.6 vanilla (kernel.org)
3373 2.6.16.27-0.9 (SLES 10)
3374 * Recommended e2fsprogs version: 1.39.cfs5
3376 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3380 Frequency : occasional, depends on client load and configuration
3381 Bugzilla : 12181, 12203
3382 Description: data loss for recently-modified files
3384 Details : In some cases it is possible that recently written or created
3385 files may not be written to disk in a timely manner (this should
3386 normally be within 30s unless client IO load is very high).
3387 The problem appears as zero-length files or files that are a
3388 multiple of 1MB in size after a client crash or client eviction
3389 that are missing data at the end of the file.
3391 This problem is more likely to be hit on clients where files are
3392 repeatedly created and unlinked in the same directory, clients
3393 have a large amount of RAM, have many CPUs, the filesystem has
3394 many OSTs, the clients are rebooted frequently, and/or the files
3395 are not accessed by other nodes after being written.
3397 The presence of the problem can be detected by looking at
3398 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3399 smaller than the second (nr_unused) then dirty files will not
3400 be flushed automatically to disk. "sync; sleep 10" should be
3401 run several times on the node before unmounting it to update
3402 Lustre (this is also safe to run on nodes without this problem).
3404 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3405 that can cause this same problem, so customers using that kernel
3406 also need to update the kernel in addition to Lustre. In order
3407 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3409 It is normal that files written just before a client crash (less
3410 than 30s) may not yet have been flushed to disk, even for local
3414 Frequency : frequent on thin XT3 nodes
3416 Description: UUID collision on thin XT3 Linux nodes
3417 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3418 randomly, since we relied on an insufficiently-seeded PRNG.
3423 Description: OSS hangs after "All ost request buffers busy"
3424 Details : A deadlock between quota and journal operations caused OSS
3425 hangs after printing "All ost request buffers busy."
3428 Frequency : always on liblustre builds
3430 Description: Cleanup compiler warnings on liblustre
3433 Frequency : always on liblustre builds on XT3
3435 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3439 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3440 > 32000 subdirectories
3441 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3442 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3443 subdirectory count crosses 32000. This will aid e2fsck to
3444 correctly handle more than 32000 subdirectories.
3447 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3449 Description: System hang on startup
3450 Details : This bug allowed the liblustre (e.g. catamount) client to
3451 return to the app before handling all startup RPCs. This
3452 could leave the node unresponsive to lustre network traffic
3453 and manifested as a server ptllnd timeout.
3455 Severity : enhancement
3457 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3458 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3459 boolean to control whether to print peer debug info when a
3460 client's RPC times out.
3465 Description: make O_SYNC working on 2.6 kernels
3466 Details : 2.6 kernels use different method for mark pages for write,
3467 so need add a code to lustre for O_SYNC work.
3472 Description: Failure to close file and release space on NFS
3473 Details : Put inode details into lock acquired in ll_intent_file_open.
3474 Use mdc_intent_lock in ll_intent_open to properly
3475 detect all kind of errors unhandled by mdc_enqueue.
3480 Description: proc file read during shutdown sometimes raced obd removal,
3482 Details : Add lock to prevent obd access after proc file removal.
3485 Frequency : Only for files larger than 4GB on 32-bit clients.
3487 Description: improperly doing page alignment of locks
3488 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3489 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3494 Description: RPCs being resent when they shouldn't be
3495 Details : Some RPCs that should not be resent are being resent. This
3496 can cause inconsistencies in the RPC state machine. Do not
3497 resend such requests.
3500 Frequency : rare, only with NFS export
3502 Description: Crash on NFS re-export node
3503 Details : under very unusual load conditions an assertion is hit in
3504 ll_intent_file_open()
3507 Frequency : only if OST filesystem is corrupted
3509 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3510 Details : for a short time RPCs with bulk IO are in the replay list,
3511 but replay of bulk IOs is unimplemented. If the OST filesystem
3512 is corrupted due to disk cache incoherency and then replay is
3513 started it is possible to trip an assertion. Avoid putting
3514 committed RPCs into the replay list at all to avoid this issue.
3519 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3520 Details : Large single O_DIRECT read and write calls can fail to allocate
3521 a sufficiently large buffer to process the request. In case of
3522 allocation failure the allocation is retried with a smaller
3523 buffer and broken into smaller requests.
3525 Severity : enhancement
3527 Description: Add -o localflock option to simulate old noflock behaviour.
3528 Details : This will achieve local-only flock/fcntl locks coherentness.
3533 Description: versioning check is incomplete
3534 Details : Checking the version difference of client vs. server, report
3535 error if the gap is too big.
3540 Description: add support PG_writeback bit
3541 Details : add support for PG_writeback bit for Lustre, for more carefull
3542 work with page cache in 2.6 kernel. This also fix some deadlocks
3543 and remove hack for work O_SYNC with 2.6 kernel.
3545 Severity : enhancement
3547 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3548 Details : The uninit_groups feature works in conjunction with the kernel
3549 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3550 up the pass1 processing of e2fsck. This is a read-only feature
3551 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3552 mount filesystems that have had this feature enabled.
3554 Severity : enhancement
3556 Description: Improve multi-block allocation algorithm to avoid fragmentation
3557 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3558 allocation locality and avoid filesystem fragmentation.
3560 ------------------------------------------------------------------------------
3562 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3564 * Support for kernels:
3565 2.6.9-42.0.3.EL (RHEL 4)
3566 2.6.5-7.276 (SLES 9)
3567 2.4.21-47.0.1.EL (RHEL 3)
3568 2.6.12.6 vanilla (kernel.org)
3569 2.6.16.21-0.8 (SLES10)
3570 * Recommended e2fsprogs version: 1.39.cfs2-0
3572 * The backwards-compatible /proc/sys/portals symlink has been removed
3573 in this release. Before upgrading, please ensure that you change
3574 any configuration scripts or /etc/sysctl.conf files that access
3575 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3576 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3577 in advance of the upgrade on any system running Lustre 1.4.6 or
3578 newer, since /proc/sys/lnet was added in that version.
3579 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3583 Frequency : only when quota is used
3585 Description: avoid scanning export list for quota master
3586 Details : Change the algorithms to avoid scanning export list in order
3587 to improve the efficiency.
3590 Frequency : MDS failover only, very rarely
3592 Description: "went back in time" messages on mds failover
3593 Details : The greatest transno may be lost when the current operation
3594 finishes with an error (transno==0) and the client's last_rcvd
3595 record is over-written. Save the greatest transno in the
3596 mds_last_transno for this case.
3599 Frequency : always for specific kernels and striping counts
3601 Description: client may get "Matching packet too big" without ACL support
3602 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3603 when trying to access files in certain configurations. The
3604 clients should in fact be denied when mounting because they do
3605 not understand ACLs.
3608 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3610 Description: many clients connecting with IO in progress causes connect timeouts
3611 Details : Avoid synchronous journal commits to avoid delays caused by many
3612 clients connecting/disconnecting when bulk IO is in progress.
3613 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3614 OST_IO_PORTAL to avoid delays behind potentially many pending
3618 Frequency : occasionally with multiple writers to a single file
3620 Description: shared writes to file may result in wrong size reported by stat()
3621 Details : Allow growing of kms when extent lock is cancelled
3624 Frequency : always with random mmap IO to multi-striped file
3626 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3627 Details : Only if the hole is at the end of OST object so that kms is too
3628 small. Fix is to increase kms accordingly in ll_nopage.
3631 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3633 Description: writes to a missing object would leak memory on the OST
3634 Details : If there is an inconsistency between the MDS and OST filesystems,
3635 such that the MDS references an object that doesn't exist, writes
3636 to that object will leak memory due to incorrect cleanup in the
3637 error handling path, eventually running out of memory on the OST.
3642 Description: Creating too long symlink causes lustre errors
3643 Details : Check symlink and name lengths before sending requests to MDS.
3646 Frequency : only if flock is enabled (not on by default)
3648 Description: posix locks not released on fd closure on 2.6.9+
3649 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3650 kernels, this caused such locks not to be released on fd close and
3651 then assertions on fs unmount about still used locks.
3654 Frequency : MDS failover only, very rarely
3656 Description: clients may get ASSERTION(granted_lock != NULL)
3657 Details : When request was taking a long time, and a client was resending
3658 a getattr by name lock request. The were multiple lock requests
3659 with the same client lock handle and
3660 mds_getattr_name->fixup_handle_for_resent_request found one of the
3661 lock handles but later failed with ASSERTION(granted_lock != NULL).
3666 Description: handle->h_buffer_credits > 0, assertion failure
3667 Details : h_buffer_credits is zero after truncate, causing assertion
3668 failure. This patch extends the transaction or creates a new
3672 Frequency : NFS re-export or patchless client
3673 Bugzilla : 11179, 10796
3674 Description: Crash on NFS re-export node (__d_move)
3675 Details : We do not want to hash the dentry if we don't have a lock.
3676 But if this dentry is later used in d_move, we'd hit uninitialised
3677 list head d_hash, so we just do this to init d_hash field but
3678 leave dentry unhashed.
3681 Frequency : NFS re-export or patchless client
3683 Description: NFS exports has problem with symbolic link
3684 Details : lustre client didn't properly install dentry when re-exported
3685 to NFS or running patchless client.
3688 Frequency : NFS re-export or patchless client
3690 Description: Various nfs/patchless fixes.
3691 Details : fixes reuse disconected alias for lookup process - this fixes
3692 warning "find_exported_dentry: npd != pd",
3693 fix permission error with open files at nfs.
3694 fix apply umask when do revalidate.
3697 Frequency : occasional
3699 Description: Crash on NFS re-export node
3700 Details : calling clear_page() on the wrong pointer triggered oops in
3701 generic_mapping_read().
3704 Frequency : rarely, using O_DIRECT IO
3706 Description: unaligned directio crashes client with LASSERT
3707 Details : check for unaligned buffers before trying any requests.
3710 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3712 Description: stale data returned from RAID cache
3713 Details : If only a small amount of IO is done to the RAID device before
3714 reading it again it is possible to get stale data from the RAID
3715 cache instead of reading it from disk.
3718 Frequency : always for sles10 kernel
3720 Description: sles10 support
3721 Details : ll_follow_link: compile fixes and using of nd_set_link
3722 under newer kernels.
3725 Frequency : depends on arch, kernel and compiler version, always on sles10
3728 Description: recursive or deep enough symlinks cause stack overflow
3729 Details : getting rid of large stack-allocated variable in
3733 Frequency : depends on hardware
3735 Description: lustre write performance loss in the SLES10 kernel
3736 Details : the performance loss is caused by using of write barriers in the
3737 ext3 code. The SLES10 kernel turns barrier support on by
3738 default. The fix is to undo that change for ldiskfs.
3740 ------------------------------------------------------------------------------
3742 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3744 * Support for kernels:
3745 2.6.9-42.0.3EL (RHEL 4)
3746 2.6.5-7.276 (SLES 9)
3747 2.4.21-47.0.1.EL (RHEL 3)
3748 2.6.12.6 vanilla (kernel.org)
3752 Frequency : quota enabled and large files being deleted
3754 Description: releasing more than 4GB of quota at once hangs OST
3755 Details : If a user deletes more than 4GB of files on a single OST it
3756 will cause the OST to spin in an infinite loop. Release
3757 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3762 Description: statfs data retrieved from /proc may be stale or zero
3763 Details : When reading per-device statfs data from /proc, in the
3764 {kbytes,files}_{total,free,avail} files, it may appear
3765 as zero or be out of date.
3768 Frequency : systems with MD RAID1 external journal devices
3770 Description: lconf's call to blkid is confused by RAID1 journal devices
3771 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3772 of returning all block devices that match the journal UUID.
3775 Frequency : always, for aggregate stripe size over 4GB
3777 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3778 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3779 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3782 Frequency : always if "lfs find" used on a local file/directory
3784 Description: "lfs find" segfaults if used on a local file/directory
3785 Details : The case where a directory component was not specified wasn't
3786 handled correctly. Handle this properly.
3789 Frequency : always on ppc64
3791 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3792 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3793 perform endianness conversion.
3796 Frequency : rarely (truncate to non-zero file size after write under load)
3797 Bugzilla : 10730, 10687
3798 Description: Files padded with zeros to next 4K multiple
3799 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3800 it is possible that files that are truncated to a non-zero size
3801 immediately after being written are filled with zero bytes beyond
3802 the truncated size. No file data is lost.
3804 Severity : enhancement
3806 Description: Allow recovery/failover for liblustre clients.
3807 Details : liblustre clients were unaware of failover configurations until
3810 Severity : enhancement
3812 Description: user file locks should fail when not mounting with flock option
3813 Details : Set up an error-returning stub in ll_file_operations.lock field
3814 to prevent incorrect behaviour when client is mounted without
3815 flock option. Also, set up properly f_op->flock field for
3819 Frequency : always on ia64
3821 Description: "lfs df" loops on printing out MDS statfs information
3822 Details : The obd_ioctl_data was not initialized and in some systems
3823 this caused a failure during the ioctl that did not return
3824 an error. Initialize the struct and return an error on failure.
3827 Frequency : SLES 9 only
3829 Description: Error of copying files with lustre special EAs as root
3830 Details : Client side always return success for setxattr call for lustre
3831 special xattr (currently only "trusted.lov").
3834 Frequency : rarely on clusters with both ia64+i386 clients
3836 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3837 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3838 clients are doing concurrent writes to the same file it is
3839 possible that the ia64 clients may LASSERT because the OST
3840 extent locks are not PAGE_SIZE aligned. Ensure that grown
3841 locks are always aligned on the request boundary.
3844 Frequency : specific use, occasional
3846 Description: Overwriting in use executable truncates on-disk binary image
3847 Details : If one node attempts to overwrite an executable in use by
3848 another node, we now correctly return ETXTBSY instead of
3849 truncating the file.
3851 Severity : enhancement
3853 Description: Async OSC create to avoid the blocking unnecessarily.
3854 Details : If an OST has no remaining object, system will block on the creating
3855 when need to create a new object on this OST. Now, ways use
3856 pre-created objects when available, instead of blocking on an
3857 empty osc while others are not empty. If we must block, we block
3858 for the shortest possible period of time.
3863 Description: chmod on Lustre root is propagated to other clients
3864 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3865 invalid by the follow_mount time.
3868 Frequency : liblustre clients only
3870 Description: Race in 'instant cancel' lock handling could lead to such locks
3871 never to be granted in case of SMP MDS
3872 Details : Do not destroy not yet granted but cbpending locks in
3876 Frequency : replay/resend of open
3878 Description: non null lock assetion failure in mds_intent_policy
3879 Details : Trying to replay/resend lockless open requests resulted in
3880 mds_open() returning 0 with no lock. Now it sets a flag if
3881 a lock is going to be returned.
3883 Severity : enhancement
3885 Description: Checksum enhancements
3886 Details : New checksum enhancements allow for resending RPCs that failed
3889 Severity : enhancement
3891 Description: Tunables on number of dirty pages in cacche
3892 Details : Allow to set limit on number of dirty pages cached.
3897 Description: client crash on unmount - lock still has references
3898 Details : In some error handling cases it was possible to leak a lock
3899 reference on a client while accessing a file. This was not
3900 harmful to the client during operation, but would cause the
3901 client to crash when the filesystem is unmounted.
3904 Frequency : specific case, rare
3906 Description: ETXTBSY on mds though file not in use by client
3907 Details : ETXTBSY is no longer incorrectly returned when attempting to
3908 chmod or chown a directory that the user previously tried to
3909 execute or a currently-executing binary.
3912 Frequency : extremely rare except on liblustre-based clients
3914 Description: Lustre space not freed when files are deleted
3915 Details : Clean up open-unlinked files after client eviction. Previously
3916 the unlink was skipped and the files remained as orphans.
3921 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3922 Details : In some cases it was possible to send a blocking callback to
3923 a client doing a glimpse, even though that client didn't get
3924 a lock granted. When the glimpse lock is cancelled on the OST
3925 the freed lock is left on the waiting list and corrupted the list.
3928 Frequency : all core dumps
3930 Description: Broke core dumps to lustre
3931 Details : Negative dentry may be unhashed if parent does not have UPDATE
3932 lock, but some callers, e.g. do_coredump, expect dentry to be
3933 hashed after successful create, hash it in ll_create_it.
3935 ------------------------------------------------------------------------------
3937 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3939 * Support for kernels:
3940 2.6.9-42.0.2.EL (RHEL 4)
3941 2.6.5-7.276 (SLES 9)
3942 2.4.21-40.EL (RHEL 3)
3943 2.6.12.6 vanilla (kernel.org)
3947 Frequency : always on RHEL 3
3949 Description: Number of open files grows over time
3950 Details : The number of open files grows over time, whether or not
3951 Lustre is started. This was due to a filp leak introduced
3952 by one of our kernel patches.
3954 ------------------------------------------------------------------------------
3956 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3958 * Support for kernels:
3959 2.6.9-42.EL (RHEL 4)
3960 2.6.5-7.276 (SLES 9)
3961 2.4.21-40.EL (RHEL 3)
3962 2.6.12.6 vanilla (kernel.org)
3967 Bugzilla : 5719, 9635, 9792, 9684
3968 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3969 Details : If a server is under heavy load and cannot reply to new
3970 connection requests before the client resends the (re)connect,
3971 the connection handling code can behave badly if two service
3972 threads are concurrently handing separate (re)connections from
3973 the same client. Add better locking to the connection handling
3974 code, and ensure that only a single connection will be processed
3975 for a given client UUID, even if the lock is dropped.
3977 Severity : enhancement
3979 Description: add TCP zero-copy support to kernel
3980 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3981 sends if the hardware supports scatter-gather and checksumming.
3982 This allows socklnd to do client-write and server-read more
3983 efficiently and reduce CPU utilization from skbuf copying.
3986 Frequency : only if NFS exporting from client
3988 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3989 Details : If setting attributes on a file created under NFS that had
3990 never been opened it would be possible to oops the client
3991 if the file had no objects.
3994 Frequency : always for liblustre
3996 Description: liblustre client does MDS+OSTs setattr RPC for each write
3997 Details : When doing a write from a liblustre client, the client
3998 incorrectly issued an RPC to the MDS and each OST the file was
3999 striped over in order to update the timestamps. When writing
4000 with small chunks and many clients this could overwhelm the MDS
4001 with RPCs. In all cases it would slow down the write because
4002 these RPCs are unnecessary.
4004 Severity : enhancement
4006 Description: allow number of MDS service threads to be changed at module load
4007 Details : It is now possible to change the number of MDS service threads
4008 running. Adding "options mds mds_num_threads={N}" to the MDS's
4009 /etc/modprobe.conf will set the number of threads for the next
4010 time Lustre is restarted (assuming the "mds" module is also
4011 reloaded at that time). The default number of threads will
4012 stay the same, 32 for most systems.
4017 Description: OST crash if filesystem is unformatted or corrupt
4018 Details : If an OST is started on a device that has never been formatted
4019 or if the filesystem is corrupt and cannot even mount then the
4020 error handling cleanup routines would dereference a NULL pointer.
4025 Description: NULL pointer deref in llap_from_page.
4026 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4027 incorrect) mapping if the page was truncated/reclaimed while it was
4028 searched for. Check for this condition and skip such pages when
4029 doing readahead. Introduce extra check to llap_from_page() to
4030 verify page->mapping->host is non-NULL (so page is not anonymous).
4033 Frequency : Sometimes when using sys_sendfile
4035 Description: "page not covered by a lock" warnings from ll_readpage
4036 Details : sendfile called ll_readpage without right page locks present.
4037 Now we introduced ll_file_sendfile that does necessary locking
4038 around call to generic_file_sendfile() much like we do in
4042 Frequency : with certain MDS communication failures at client mount time
4044 Description: NULL pointer deref after failed client mount
4045 Details : a client connection request may delayed by the network layer
4046 and not be sent until after the PTLRPC layer has timed out the
4047 request. If the client fails the mount immediately it will try
4048 to clean up before the network times out the request. Add a
4049 reference from the request import to the obd device and delay
4050 the cleanup until the network drops the request.
4053 Frequency : occasionally during client (re)connect
4055 Description: assertion failure during client (re)connect
4056 Details : processing a client connection request may be delayed by the
4057 client or server longer than the client connect timeout. This
4058 causes the client to resend the connection request. If the
4059 original connection request is replied in this interval, the
4060 client may trip an assertion failure in ptlrpc_connect_interpret()
4061 which thought it would be the only running connect process.
4064 Frequency : only with obd_echo servers and clients that are rebooted
4066 Description: kernel BUG accessing uninitialized data structure
4067 Details : When running an obd_echo server it did not start the ping_evictor
4068 thread, and when a client was evicted an uninitialized data
4069 structure was accessed. Start the ping_evictor in the RPC
4070 service startup instead of the OBD startup.
4072 Severity : enhancement
4073 Bugzilla : 10193 (patchless)
4074 Description: Remove dependency on various unexported kernel interfaces.
4075 Details : No longer need reparent_to_init, exit_mm, exit_files,
4076 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4079 Frequency : rare (only users of deprecated and unsupported LDAP config)
4081 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4082 Details : LDAP apparently contains 'lustreName' attributes instead of
4083 'name'. A simple remapping of the name is sufficient.
4086 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4088 Description: waiting_locks_callback trips kernel BUG if client is evicted
4089 Details : Running with the dump_on_timeout debug flag turned on makes
4090 it possible that the waiting_locks_callback() can try to dump
4091 the Lustre kernel debug logs from an interrupt handler. Defer
4092 this log dumping to the expired_lock_main() thread.
4094 Severity : enhancement
4096 Description: Support NFS exporting on 2.6 kernels.
4097 Details : Implement non-rawops metadata methods for NFS server to use without
4098 changing NFS server code.
4101 Frequency : very rare (synthetic metadata workload only)
4103 Description: two racing renames might cause an MDS thread to deadlock
4104 Details : Running the "racer" program may cause one MDS thread to rename
4105 a file from being the source of a rename to being the target of
4106 a rename at exactly the same time that another thread is doing
4107 so, and the second thread has already enqueued these locks after
4108 doing a lookup of the target and is trying to relock them in
4109 order. Ensure that we don't try to re-lock the same resource.
4112 Frequency : only very large systems with liblustre clients
4114 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4115 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4116 all OSTs in parallel. This allows the eviction of stale liblustre
4117 clients to proceed much faster than if they were done in series,
4118 and also offers similar improvements for other set_info RPCs.
4123 Description: excessive CPU usage during initial read phase on client
4124 Details : During the initial read phase on a client, it would agressively
4125 retry readahead on the file, consuming too much CPU and impacting
4126 performance (since 1.4.5.8). Improve the readahead algorithm
4127 to avoid this, and also improve some other common cases (read
4128 of small files in particular, where "small" is files smaller than
4129 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4134 Description: MDS crash when receiving packet with unknown intent.
4135 Details : Do not LBUG in unknown intent case, just return -EFAULT
4137 Severity : enhancement
4138 Bugzilla : 9293, 9385
4139 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4140 Details : Do not serialize getattr (non-intent version) and statfs.
4143 Frequency : occasional, when OST network is overloaded/intermittent
4145 Description: client evicted by OST after bulk IO timeout
4146 Details : If a client sends a bulk IO request (read or write) the OST
4147 may evict the client if it is unresposive to its data GET/PUT
4148 request. This is incorrect if the network is overloaded (takes
4149 too long to transfer the RPC data) or dropped the OST GET/PUT
4150 request. There is no need to evict the client at all, since
4151 the pinger and/or lock callbacks will handle this, and the
4152 client can restart the bulk request.
4155 Frequency : Always when mmapping file with no objects
4157 Description: client crashes when mmapping file with no objects
4158 Details : Check that we actually have objects in a file before doing any
4159 operations on objects in ll_vm_open, ll_vm_close and
4165 Description: Request leak when working with deleted CWD
4166 Details : Introduce advanced request refcount tracking for requests
4167 referenced from lustre intent.
4169 Severity : Enhancement
4171 Description: Cache open file handles on client.
4172 Details : MDS now will return special lock along with openhandle, if
4173 requested and client is allowed to hold openhandle, even if unused,
4174 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4175 closing files for every read/write openration.
4177 Severity : Enhancement
4179 Description: Cache open negative dentries on client when possible.
4180 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4181 negative dentries on lock revocation.
4186 Description: Remounting a client read-only wasn't possible with a zconf mount
4187 Details : It wasn't possible to remount a client read-only with llmount.
4189 Severity : enhancement
4190 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4191 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4192 MPICH is included to add Lustre-specific ADIO interfaces.
4193 This is based closely on the UFS ADIO layer and only differs
4194 in file creation, in order to allow the OST striping to be set.
4195 This is user-contributed code and not supported by CFS.
4200 Description: extended inode attributes (immutable, append-only) work improperly
4201 when 2.4 and 2.6 kernels are used on client/server or vice versa
4202 Details : Introduce kernel-independent values for these flags.
4204 Severity : enhancement
4207 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4208 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4209 granularity. Now, Fractional MB granularity is be supported,
4210 this is very useful for low memory system.
4212 Severity : enhancement
4214 Description: Getattr by fid
4215 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4216 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4217 have locks and avoids getattr rpc requests for it on every stat.
4220 Frequency : Always, for filesystems larger than 2TB
4222 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4223 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4224 filesystems larger than 2TB to be mounted with mballoc enabled.
4227 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4229 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4230 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4231 filesystems larger than 2TB is mounted then the kernel will
4232 silently corrupt the start of the filesystem. CONFIG_LBD is
4233 enabled for all CFS-supported kernels, but the possibility of
4234 this happening with a modified kernel config exists.
4236 Severity : enhancement
4238 Description: add client O_DIRECT support for 2.6 kernels
4239 Details : It is now possible to do O_DIRECT reads and writes to files
4240 in the Lustre client mountpoint on 2.6 kernel clients.
4242 Severity : enhancement
4244 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4245 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4246 parallel, not waiting for response from every OST before sending
4247 a rpc to the next OST.
4252 Description: setattr vs write race when updating file timestamps
4253 Details : Client processes that update a file timestamp into the past
4254 right after writing to the file (e.g. tar) it is possible that
4255 the updated file modification time can be reset to the current
4256 time due to a race between processing the setattr and write RPC.
4258 Severity : enhancement
4260 Description: Bring 'lfs find' closer in line with regular Linux find.
4261 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4262 -print0 options and obtains all the needed info through the lustre
4265 Severity : enhancement
4267 Description: support up to 1024 configured devices on one node
4268 Details : change obd_dev array from statically allocated to dynamically
4269 allocated structs as they are first used to reduce memory usage
4274 Description: Flush dirty partially truncated pages during truncate
4275 Details : Immediatelly flush partially truncated pages in filter_setattr,
4276 this way we completely avoid having any pages in page cache on OST
4277 and can retire ugly workarounds during writes to flush such pages.
4282 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4283 Details : protect inode from truncation within vfs_unlink() context
4284 just take a reference before calling vfs_unlink() and release it
4285 when parent's i_sem is free.
4290 Description: last_id value checked outside lock on OST caused LASSERT failure
4291 Details : If there were multiple MDS->OST object precreate requests in
4292 flight, it was possible that the OST's last object id was checked
4293 outside a lock and incorrectly tripped an assertion. Move checks
4294 inside locks, and discard old precreate requests.
4297 Frequency : always, if extents are used on OSTs
4299 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4300 Details : OSTs using the extents format would not zero the high 16 bits of
4301 the index physical block number. This is not a problem for any
4302 OST filesystems smaller than 16TB, and no kernels support ext3
4303 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4304 new/modified files) and can be fixed for existing filesystems
4305 with e2fsprogs-1.39-cfs1.
4310 Description: import connection selection may be incorrect if timer wraps
4311 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4312 import connections to be ignored if the 32-bit jiffies counter
4313 wraps. Use a 64-bit jiffies counter.
4316 Frequency : very large clusters immediately after boot
4318 Description: LNET request buffers exhausted under heavy short-term load
4319 Details : If a large number of client requests are generated on a service
4320 that has previously never seen so many requests it is possible
4321 that the request buffer growth cannot keep up with the spike in
4322 demand. Instead of dropping incoming requests, they are held in
4323 the LND until the RPC service can accept more requests.
4326 Frequency : Sometimes during replay
4328 Description: Assertion failure in ll_local_open after replay.
4329 Details : If replay happened on an open request reply before we were able
4330 to set replay handler, reply will become not swabbed tripping the
4331 assertion in ll_local_open. Now we set the handler right after
4332 recognising of open request
4335 Frequency : very rare
4337 Description: kernel reports "badness in vsnprintf"
4338 Details : Reading from the "recovery_status" /proc file in small chunks
4339 may cause a negative length in lprocfs_obd_rd_recovery_status()
4340 call to vsnprintf() (which is otherwise harmless). Exit early
4341 if there is no more space in the output buffer.
4343 Severity : enhancement
4345 Description: clear OBD RPC statistics by writing to them
4346 Details : It is now possible to clear the OBD RPC statistics by writing
4347 to the "stats" file.
4352 Description: Client mtime is not the same on different clients after utimes
4353 Details : In some cases, the client was using the utimes() syscall on
4354 a file cached on another node. The clients now validate the
4355 ctime from the MDS + OSTs to determine which one is right.
4360 Description: Inability to activate failout mode
4361 Details : lconf script incorrectly assumed that in pythong string's numeric
4362 value is used in comparisons.
4365 Frequency : always with multiple stripes per file
4367 Description: Inefficient object allocation for mutli-stripe files
4368 Details : When selecting which OSTs to stripe files over, for files with
4369 a stripe count that divides evenly into the number of OSTs,
4370 the MDS is always picking the same starting OST for each file.
4371 Return the OST selection heuristic to the original design.
4376 Description: mount failures may take full timeout to return an error
4377 Details : Under some heavy load conditions it is possible that a
4378 failed mount can wait for the full obd_timeout interval,
4379 possibly several minutes, before reporting an error.
4380 Instead return an error as soon as the status is known.
4382 Frequency : quota enabled and large files being deleted
4384 Description: releasing more than 4GB of quota at once hangs OST
4385 Details : If a user deletes more than 4GB of files on a single OST it
4386 will cause the OST to spin in an infinite loop. Release
4387 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4392 Description: statfs data retrieved from /proc may be stale or zero
4393 Details : When reading per-device statfs data from /proc, in the
4394 {kbytes,files}_{total,free,avail} files, it may appear
4395 as zero or be out of date.
4398 Frequency : systems with MD RAID1 external journal devices
4400 Description: lconf's call to blkid is confused by RAID1 journal devices
4401 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4402 of returning all block devices that match the journal UUID.
4405 Frequency : always, for aggregate stripe size over 4GB
4407 Description: assertion fails when trying to use 4GB stripe size
4408 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4409 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4412 Frequency : always on ppc64
4414 Description: the first write on an ext3 filesystem with mballoc got stuck
4415 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4416 perform endianness conversion.
4418 ------------------------------------------------------------------------------
4420 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4422 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4423 INTEROPERATE with older versions automatically. Please read the
4424 user documentation before upgrading any part of a live system.
4425 * WARNING: Lustre networking configuration changes are required with
4426 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4429 * Support for kernels:
4430 2.6.9-22.0.2.EL (RHEL 4)
4431 2.6.5-7.244 (SLES 9)
4432 2.6.12.6 vanilla (kernel.org)
4435 Severity : enhancement
4436 Bugzilla : 7981/8208
4437 Description: Introduced Lustre Networking (LNET)
4438 Details : LNET is new networking infrastructure for Lustre, it includes
4439 a reorganized network configuration mode (see the user
4440 documentation for full details) as well as support for routing
4441 between different network fabrics. Lustre Networking Devices
4442 (LNDs) for the supported network fabrics have also been
4443 created for this new infrastructure.
4445 Severity : enhancement
4446 Description: Introduced Access control lists
4447 Details : clients can set ACLs on files and directories in order to have
4448 more fine-grained permissions than the standard Unix UGO+RWX.
4449 The MDS must be started with the "-o acl" mount option.
4451 Severity : enhancement
4452 Description: Introduced filesystem quotas
4453 Details : Administrators may now establish per-user quotas on the
4456 Severity : enhancement
4458 Description: Configuration change for the XT3
4459 The PTLLND is now used to run Lustre over Portals on the XT3
4460 The configure option(s) --with-cray-portals are no longer used.
4461 Rather --with-portals=<path-to-portals-includes> is used to
4462 enable building on the XT3. In addition to enable XT3 specific
4463 features the option --enable-cray-xt3 must be used.
4468 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4469 Details : A race between allocating a new llcd and re-getting the llcd_lock
4470 allowed another thread to grab newly-allocated llcd.
4472 Severity : enhancement
4474 Description: 2.6 OST async journal commit and locking fix to improve performance
4475 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4476 2.6 kernels are now async as they already were in 2.4 kernels so
4477 that they can commit concurrently with the network bulk transfer.
4478 For block-allocated files the filter allocation semaphore is held
4479 to avoid filesystem fragmentation during allocation. BKL lock
4480 removed for 2.6 xattr operations where it is no longer needed.
4485 Description: lconf incorrectly determined whether two IP networks could talk
4486 Details : In some more complicated routing and multiple-network
4487 configurations, lconf will avoid trying to make a network
4488 connection to a disjoint part of the IP space. It was doing the
4489 math incorrectly for one set of cases.
4494 Description: Fix for potential infinite loop processing records in an llog.
4495 Details : If an llog record is corrupted/zeroed, it is possible to loop
4496 forever in llog_process(). Validate the llog record length
4497 and skip the remainder of the block on error.
4500 Frequency : occasional (liblustre only)
4502 Description: liblustre could not open files whose last component is a symlink
4503 Details : sysio_path_walk() would incorrectly pass the open intent to
4504 intermediate path components.
4507 Frequency : rare (liblustre only with non-standard tuning)
4508 Bugzilla : 7201 (7350)
4509 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4510 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4511 after it has been released, only do this for referenced locks
4513 Severity : enhancement
4515 Description: specifying an (invalid) directory default stripe_size of -1
4516 would reset the directory default striping
4517 Details : stripe_size -1 was used internally to signal directory stripe
4518 removal, now use "all default" to signal dir stripe removal
4519 as a directory striping of "all default" is not useful
4522 Frequency : common for large clusters running liblustre clients
4524 Description: doing an ls when liblustre clients are running is slow
4525 Details : sending a glimpse AST to a liblustre client waits for every AST
4526 to time out, as liblustre clients will not respond. Since they
4527 cannot cache data we refresh the OST lock LVB from disk instead.
4529 Severity : enhancement
4531 Description: doing an ls at the same time as file IO can be slow
4532 Details : enqueue and other "small" requests can be blocked behind many
4533 large IO requests. Create a new OST IO portal for non-IO
4534 requests so they can be processed faster.
4537 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4539 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4540 created file with mode 000
4541 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4542 allow an ftruncate() as a non-root user to a file with mode 000.
4543 Lustre now respects this flag to disable mode checks when
4544 truncating a file owned by the user
4547 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4550 Description: Revoking locks from clients that went dead or catatonic might take
4552 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4553 cancellation of such locks instant on servers without waiting for
4554 any reply from clients. Clients drops these locks when cancel
4555 notification from server is received without replying.
4558 Frequency : liblustre-only, when liblustre client dies or becomes busy
4560 Description: Doing ls on Linux clients can take a long time with active
4562 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4563 granting such locks to it in the first place if possible. Locks
4564 are taken by proxy on the OST during the read or write and
4565 dropped immediately afterward. Add connect flags handling, do
4566 not grant locks to liblustre clients for glimpse ASTs.
4568 Severity : enhancement
4570 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4571 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4572 sequential reads and behaves badly with applications doing random
4573 reads. Improve it by reading ahead at least read region, and
4574 avoiding excessive large RPC for small reads.
4576 Severity : enhancement
4578 Description: Creating more than 1000 files for a single job may cause a load
4579 imbalance on the OSTs if there are also a large number of OSTs.
4580 Details : qos_prep_create() uses an OST index reseed value that is an
4581 even multiple of the number of available OSTs so that if the
4582 reseed happens in the middle of the object allocation it will
4583 still utilize the OSTs as uniformly as possible.
4588 Description: OST or MDS may oops in ping_evictor_main()
4589 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4590 but doesn't restart at beginning of obd_exports_timed list
4593 Severity : enhancement
4595 Description: improve by-nid export eviction on the MDS and OST
4596 Details : allow multiple exports with the same NID to be evicted at one
4597 time without re-searching the exports list.
4600 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4602 Description: MDS may oops in groups_free()
4603 Details : in rare race conditions a newly allocated group_info struct is
4604 freed again, and this can be NULL. The 2.4 compatibility code
4605 for groups_free() checked for a NULL pointer, but 2.6 did not.
4608 Frequency : common for liblustre clients doing little filesystem IO
4609 Bugzilla : 9352, 7313
4610 Description: server may evict liblustre clients accessing contended locks
4611 Details : if a client is granted a lock or receives a completion AST
4612 with a blocking AST already set it would not reply to the AST
4613 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4616 Frequency : lfs setstripe, only systems with more than 160 OSTs
4618 Description: unable to set striping with a starting offset beyond OST 160
4619 Details : llapi_create_file() incorrectly limited the starting stripe
4620 index to the maximum single-file stripe count.
4623 Frequency : LDAP users only
4625 Description: lconf did not handle in-kernel recovery with LDAP properly
4626 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4628 Severity : enhancement
4630 Description: bind OST threads to NUMA nodes to improve performance
4631 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4632 node and do their allocations there to localize memory access
4634 Severity : enhancement
4636 Description: llmount can determine client NID directly from Myrinet (GM)
4637 Details : the client NID code from gmnalnid was moved directly into
4638 llmount, removing the need to use this or specifying the
4639 client NID explicitly when mounting GM clients with zeroconf
4642 Frequency : if client is started with down MDS
4644 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4645 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4646 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4647 ptlrpc_import_delayed_req() because we didn't check if the
4648 request was interrupted, and we also didn't break out of the
4649 event loop if there was no timeout
4654 Description: data loss during non-page-aligned writes to a single file from
4655 both multiple nodes and multiple threads on one node at same time
4656 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4657 inconsistency led to false short-reads, that were cached and later
4658 used by ->prepare_write() to fill in partially written page,
4659 leading to data loss.
4662 Frequency : always, if lconf --abort_recovery used
4664 Description: lconf --abort_recovery fails with 'Operation not supported'
4665 Details : lconf was attempting to abort recovery on the MDT device and not
4668 Severity : enhancement
4670 Description: remove cleanup logs
4671 Details : replace lconf-generated cleanup logs with lustre internal
4672 cleanup routines. Eliminates the need for client-cleanup and
4675 Severity : enhancement
4677 Description: add support for EAs (user and system) on lustre filesystems
4678 Details : it is now possible to store extended attributes in the Lustre
4679 client filesystem, and with the user_xattr mount option it
4680 is possible to allow users to store EAs on their files also
4682 Severity : enhancement
4684 Description: Add possibility (config option) to show minimal available OST free
4686 Details : When compiled with --enable-mindf configure option, statfs(2)
4687 (and so, df) will return least minimal free space available from
4688 all OSTs as amount of free space on FS, instead of summary of
4689 free spaces of all OSTs.
4691 Severity : enhancement
4693 Description: do not expand extent locks acquired on OST-side
4694 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4695 by server: they are not cached anyway.
4698 Frequency : when mmap is used/binaries executed from Lustre
4700 Description: Unmmap pages before throwing them away from read cache.
4701 Details : llap_shrink cache now attempts to unmap pages before discarding
4702 them (if unmapping failed - do not discard). SLES9 kernel has
4703 extra checks that trigger if this unmapping is not done first.
4708 Description: lconf didn't resolve symlinks before checking to see whether a
4709 given mountpoint was already in use
4712 Frequency : when migrating failover services
4713 Bugzilla : 6395, 9514
4714 Description: When migrating a subset of services from a node (e.g. failback
4715 from a failover service node) the remaining services would
4716 time out and evict clients.
4717 Details : lconf --force (implied by --failover) sets the global obd_timeout
4718 to 5 seconds in order to quickly disconnect, but this caused
4719 other RPCs to time out too quickly. Do not change the global
4720 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4722 Severity : enhancement
4723 Frequency : if MDS is started with down OST
4724 Bugzilla : 9439,5706
4725 Description: Allow startup/shutdown of an MDS without depending on the
4726 availability of the OSTs.
4727 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4728 Add appropriate locking and lov-osc refcounts for safe
4729 cleaning. Add osc abort_inflight calls in case the
4730 synchronize never started.
4733 Frequency : occasional (Cray XT3 only)
4735 Description: root not authorized to access files in CRAY_PORTALS environment
4736 Details : The client process capabilities were not honoured on the MDS in
4737 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4738 been accessed by an authorized user then root was able to access
4739 the file on the local client also. The root user capabilities
4740 are now allowed on the MDS, as this environment has secure UID.
4743 Frequency : occasional
4745 Description: ldiskfs "too long searching" message happens too often
4746 Details : A debugging message (otherwise harmless) prints too often on
4747 the OST console. This has been reduced to only happen when
4748 there are fragmentation problems on the filesystem.
4753 Description: Division by zero in statfs when all OSCs are inactive
4754 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4755 lov_statfs divides by value returned by lov_get_stripecnt().
4759 Bugzilla : 9489, 3273
4760 Description: First write from each client to each OST was only 4kB in size,
4761 to initialize client writeback cache, which caused sub-optimal
4762 RPCs and poor layout on disk for the first written file.
4763 Details : Clients now request an initial cache grant at (re)connect time
4764 and so that they can start streaming writes to the cache right
4765 away and always do full-sized RPCs if there is enough data.
4766 If the OST is rebooted the client also re-establishes its grant
4767 so that client cached writes will be honoured under the grant.
4772 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4773 Details : Now I/O RPCs go to different portal number and (presumably) fast
4774 lock requests (and glimses) and other RPCs get their own service
4775 threads pool that should be able to service those RPCs
4778 Severity : enhancement
4780 Description: Ability to exchange lustre version between client and servers and
4781 issue warnings at client side if client is too old. Also for
4782 liblustre clients there is ability to refuse connection of too old
4784 Details : New 'version' field is added to connect data structure that is
4785 filled with version info. That info is later checked by server and
4789 Frequency : rare, liblustre only.
4790 Bugzilla : 9296, 9581
4791 Description: Two simultaneous writes from liblustre at offset within same page
4792 might proceed at the same time overwriting eachother with stale
4794 Details : I/O lock withing llu_file_prwv was released too early, before data
4795 actually was hitting the wire. Extended lock-holding time until
4796 server acknowledges receiving data.
4799 Frequency : extremely rare. Never observed in practice.
4801 Description: avoid generating lustre_handle cookie of 0.
4802 Details : class_handle_hash() generates handle cookies by incrementing
4803 global counter, and can hit 0 occasionaly (this is unlikely, but
4804 not impossible, because initial value of cookie counter is
4805 selected randonly). Value of 0 is used as a sentinel meaning
4806 "unassigned handle" --- avoid it. Also coalesce two critical
4807 sections in this function into one.
4809 Severity : enhancement
4811 Description: allow liblustre clients to delegate truncate locking to OST
4812 Details : To avoid overhead of locking, liblustre client instructs OST to
4813 take extent lock in ost_punch() on client's behalf. New connection
4814 flag is added to handle backward compatibility.
4816 Severity : enhancement
4817 Bugzilla : 4928, 7341, 9758
4818 Description: allow number of OST service threads to be specified
4819 Details : a module parameter allows the number of OST service threads
4820 to be specified via "options ost ost_num_threads={N}" in the
4821 OSS's /etc/modules.conf or /etc/modprobe.conf.
4825 Bugzilla : 6146, 9635, 9895
4826 Description: servers crash with bad pointer in target_handle_connect()
4827 Details : In rare cases when a client is reconnecting it was possible that
4828 the connection request was the last reference for that export.
4829 We would temporarily drop the export reference and get a new
4830 one, but this may have been the last reference and the export
4831 was just destroyed. Get new reference before dropping old one.
4833 Severity : enhancement
4834 Frequency : if client is started with failover MDS
4836 Description: Allow multiple MDS hostnames in the mount command
4837 Details : Try to read the configuration from all specified MDS
4838 hostnames during a client mount in case the "primary"
4841 Severity : enhancement
4843 Description: Stop sending data to evicted clients as soon as possible.
4844 Details : Check if the client we are about to send or are sending data to
4845 was evicted already. (Check is done every second of waiting,
4846 for which l_wait_event interface was extended to allow checking
4847 of exit condition at specified intervals).
4850 Frequency : rare, normally only when NFS exporting is done from client
4852 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4854 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4857 Frequency : always, if srand() is called before liblustre initialization
4859 Description: Liblustre uses system PRNG disturbing its usage by user application
4860 Details : Introduce internal to lustre fast and high-quality PRNG for
4861 lustre usage and make liblustre and some other places in generic
4862 lustre code to use it.
4864 Severity : enhancement
4865 Bugzilla : 9477, 9557, 9870
4866 Description: Verify that the MDS configuration logs are updated when xml is
4867 Details : Check if the .xml configuration logs are newer than the config
4868 logs stored on the MDS and report an error if this is the case.
4869 Request --write-conf, or allow starting with --old_conf.
4871 Severity : enhancement
4873 Description: Handle symlinks in the path when checking if Lustre is mounted.
4874 Details : Resolve intermediate symlinks when checking if a client has
4875 mounted a filesystem to avoid duplicate client mounts.
4880 Description: lconf can hit an error exception but still return success.
4881 Details : The lconf command catches the Command error exception at the top
4882 level script context and will exit with the associated exit
4883 status, but doesn't ensure that this exit status is non-zero.
4888 Description: failure of ptlrpc thread startup can cause oops
4889 Details : Starting a ptlrpc service thread can fail if there are a large
4890 number of threads or the server memory is very fragmented.
4891 Handle this without oopsing.
4894 Frequency : always, only if liblustre and non-default acceptor port was used
4896 Description: liblustre cannot connect to servers with non-default acceptor port
4897 Details : tcpnal_set_default_params() was not called and was therefore
4898 ignoring the environment varaible TCPNAL_PORT, as well as other
4899 TCPNAL_ environment variables
4904 Description: two objects could be created on the same OST for a single file
4905 Details : If an OST is down, in some cases it was possible to create two
4906 objects on a single OST for a single file. No problems other
4907 than potential performance impact and spurious error messages.
4911 Bugzilla : 5681, 9562
4912 Description: Client may oops in ll_unhash_aliases
4913 Details : Client dcache may become inconsistent in race condition.
4914 In some cases "getcwd" can fail if the current directory is
4920 Description: Inode refcounting problems in NFS export code
4921 Details : link_raw functions used to call d_instantiate without obtaining
4922 extra inode reference first.
4926 Bugzilla : 9942, 9903
4927 Description: Referencing freed requests leading to crash, memleaks with NFS.
4928 Details : We used to require that call to ll_revalidate_it was always
4929 followed by ll_lookup_it. Also with revalidate_special() it is
4930 possible to call ll_revalidate_it() twice for the same dentry
4931 even if first occurence returned success. This fix changes semantic
4932 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4933 reference on a request referred from the intent.
4934 ll_intent_release() then releases such a request.
4937 Frequency : rare, normally benchmark loads only
4939 Description: unlinked inodes were kept in memory on the client
4940 Details : If a client is repeatedly creating and unlinking files it
4941 can accumulate a lot of stale inodes in the inode slab cache.
4942 If there is no other client load running this can cause the
4943 client node to run out of memory. Instead flush old inodes
4944 from client cache that have the same inode number as a new inode.
4947 Frequency : SLES9 2.6.5 kernel and long filenames only
4948 Bugzilla : 9969, 10379
4949 Description: utime reports stale NFS file handle
4950 Details : SLES9 uses out-of-dentry names in some cases, which confused
4951 the lustre dentry revalidation. Change it to always use the
4955 Frequency : rare, unless heavy write-truncate concurrency is continuous
4956 Bugzilla : 4180, 6984, 7171, 9963, 9331
4957 Description: OST becomes very slow and/or deadlocked during object unlink
4958 Details : filter_destroy() was holding onto the parent directory lock
4959 while truncating+unlinking objects. For very large objects this
4960 may block other threads for a long time and slow overall OST
4961 responsiveness. It may also be possible to get a lock ordering
4962 deadlock in this case, or run out of journal credits because of
4963 the combined truncate+unlink. Solution is to do object truncate
4964 first in one transaction without parent lock, and then do the
4965 final unlink in a new transaction with the parent lock. This
4966 reduces the lock hold time dramatically.
4969 Frequency : rare, 2.4 kernels only
4971 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4972 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4973 before unmounting their filesystem in order to avoid deadlock.
4974 The kernel_locked() function in 2.4 kernels only checks whether
4975 the kernel lock is held, not whether it is this process that is
4976 holding it as 2.6 kernels do.
4981 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4982 Details : The client ptlrpc code may be trying to reconnect to a down
4983 server before a previous connection attempt has timed out.
4984 Increase the reconnect interval to be longer than the connection
4985 timeout interval to avoid sending duplicate connections to
4989 Frequency : echo_client brw_test command
4991 Description: fix echo_client to work with OST preallocated code
4992 Details : OST preallocation code (5137) didn't take echo_client IO path
4993 into account: echo_client calls filter methods outside of any
4994 OST thread and, hence, there is no per-thread preallocated
4995 pages and buffers to use. Solution: hijack pga pages for IO. As
4996 a byproduct, this avoids unnecessary data copying.
5000 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5001 Description: Client can oops in mdc_commit_close() after open replay
5002 Details : It was possible for the MDS to return an open request with no
5003 transaction number in mds_finish_transno() if the client was
5004 evicted, but without actually returning an error. Clients
5005 would later try to replay that open and may trip an assertion
5006 Simplify the client close codepath, and always return an error
5007 from the MDS in case the open is not successful.
5010 Frequency : rare, 2.6 OSTs only
5012 Description: OST may deadlock under high load on fragmented files
5013 Details : If there was a heavy load and highly-fragmented OST filesystems
5014 it was possible to have all the OST threads deadlock waiting on
5015 allocation of biovecs, because the biovecs were not released
5016 until the entire RPC IO was completed. Instead, release biovecs
5017 as soon as they are complete to ensure forward IO progress.
5019 Severity : enhancement
5021 Description: Support for specifying external journal device at mount
5022 Details : If an OST or MDS device is formatted with an external journal
5023 device, this device major/minor is stored in the ext3 superblock
5024 and may not be valid for failover. Allow detecting and
5025 specifying the external journal at mount time.
5030 Description: Mounting an MDS with pending unlinked files may cause oops
5031 Details : target_finish_recovery() calls mds_postrecov() which returned
5032 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5033 considers this an error and immediately begins cleaning up the
5034 lov, just after starting the mds_lov process
5036 Severity : enhancement
5038 Description: Implement 'lfs df' to report actual free space on per-OST basis
5039 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5040 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5041 usage of objects; '-h' to report in human readable format.
5043 ------------------------------------------------------------------------------
5045 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5052 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5053 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5054 at startup during orphan unlinking. Instead, skip buddy update
5055 before setup, as it will be regenerated after recovery is complete.
5060 Description: If an OST is inactive, its locks might reference stale inodes.
5061 Details : lov_change_cbdata() must iterate over all namespaces, even if
5062 they are inactive to clear inode references from the lock.
5064 Severity : enhancement
5065 Frequency : occasional, if non-standard max_dirty_mb used
5067 Description: Client will block write RPCs if not enough grant
5068 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5069 then the client will block writes while waiting for another RPC
5070 to complete instead of consuming its dirty limit. With change
5071 we get improved performance when max_dirty_mb is small.
5073 Severity : enhancement
5074 Bugzilla : 3389, 6253
5075 Description: Add support for supplementary groups on the MDS.
5076 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5077 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5078 lookups for user supplementary groups into a cache.
5082 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5083 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5084 for liblustre O_EXCL is enforced by MDS.
5088 Description: readdir never returns NULL in liblustre.
5089 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5090 dentry in struct dirent.
5094 Frequency : liblustre only, depends on application IO pattern
5095 Description: liblustre clients evicted if not contacting servers
5096 Details : Don't put liblustre clients into the ping_evictor list, so
5097 they will not be evicted by the pinger ever.
5099 Severity : enhancement
5101 Description: Add ability to evict clients by NID from MDS.
5102 Details : By echoing "nid:$NID" string into
5103 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5104 $NID would be instantly evicted from this MDS and from all active
5105 OSTs connected to it.
5109 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5110 Details : lookup_it_finish() used to query file size from OSTs that was not
5115 Description: service threads change working directory to that of init
5116 Details : Starting lustre service threads may pin the working directory
5117 of the parent thread, making that filesystem busy. Threads
5118 now change to the working directory of init to avoid this.
5122 Frequency : during shutdown only
5123 Description: shutdown with a failed MDS or OST can cause unmount to hang
5124 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5129 Frequency : 2.6 only
5130 Description: chmod/chown may include an extra supplementary group
5131 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5132 supplementary group and if none is specified this is used.
5135 Bugzilla : 5479 (6816)
5137 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5138 Details : If lookup is in progress on a file that is unlinked we might try
5139 to revalidate the inode and fail in revalidate after lookup is
5140 complete and ll_file_open() enqueues the open again but
5141 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5144 Frequency : always, if lconf --abort_recovery used
5146 Description: lconf --abort_recovery fails with 'Operation not supported'
5147 Details : lconf was attempting to abort recovery on the MDT device and not
5150 ------------------------------------------------------------------------------
5152 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5157 Frequency : rare (only unsupported configurations with a node running as an
5159 Bugzilla : 6514, 5137
5160 Description: Mounting a Lustre file system on a node running as an OST could
5162 Details : OSTs now preallocates memory needed to write out data at
5163 startup, instead of when needed, to avoid having to
5164 allocate memory in possibly low memory situations.
5165 Specifically, if the file system is mounted on on OST,
5166 memory pressure could force it to try to write out data,
5167 which it needed to allocate memory to do. Due to the low
5168 memory, it would be unable to do so and the node would
5169 become unresponsive.
5171 Severity : enhancement
5173 Description: Addition of lconf --service command line option
5174 Details : lconf now accepts a '--service <arg>' option, which is
5175 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5177 Severity : enhancement
5179 Description: Failover mode is now the default for OSTs.
5180 Details : By default, OSTs will now run in failover mode. To return to
5181 the old behaviour, add '--failout' to the lmc line for OSTs.
5183 Severity : enhancement
5185 Description: Health checks are now provided for MDS and OSTs
5186 Details : Additional detailed health check information on MSD and OSTs
5187 is now provided through the procfs health_check value.
5190 Frequency : occasional, depends on IO load
5192 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5193 after numerous create/delete cycles
5194 Details : The ext3 inode allocation policy would not allocate new inodes
5195 very well on the OSTs because there are no new directories
5196 being created. Instead we look for groups with free space if
5197 the parent directories are nearly full.
5202 Description: Network or server problems during mount may cause partially
5203 mounted clients instead of returning an error.
5204 Details : The config llog parsing code may overwrite the error return
5205 code during mount error handling, returning success instead
5211 Description: MDS can fail to allocate large reply buffers
5212 Details : After long uptimes the MDS can fail to allocate large reply
5213 buffers (e.g. zconf client mount config records) due to memory
5214 fragmentation or consumption by the buffer cache. Preallocate
5215 some large reply buffers so that these replies can be sent even
5216 under memory pressure.
5220 Frequency : rare (liblustre)
5221 Description: fsx running with liblustre complained that using truncate() to
5222 extend the file doesn't work. This patch corrects that issue.
5223 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5224 ATTR_SIZE and lsm use in llu_setattr_raw.
5228 Frequency : rare, only 2.6 kernels
5229 Description: Unusual file access patterns on the MDS may result in inode
5230 data being lost in very rare circumstances.
5231 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5232 code in the 2.6 kernel caused the inode and/or EA data not to
5233 be read from disk, causing single-file corruption.
5237 Frequency : rare, only 2.6 filesystems using extents
5238 Description: Heavy concurrent write and delete load may cause data corruption.
5239 Details : It was possible under high-load situations to have an extent
5240 metadata block in the block device cache from a just-unlinked
5241 file overwrite a newly-allocated data block. We now unmap any
5242 metadata buffers that alias just-allocated data blocks.
5246 Frequency : filesystems with default stripe_count larger than 77
5247 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5248 Details : lconf specifies an inode size of 4096 bytes when the default
5249 stripe_count is larger than 77. This conflicts with the default
5250 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5251 this case to avoid pinning too much memory for large EAs.
5253 ------------------------------------------------------------------------------
5255 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5260 Frequency : rare (extremely heavy IO load with hundreds of clients)
5262 Description: Client is evicted, gets IO error writing to file
5263 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5264 caused clients to be evicted for AST timeouts. The fixes in
5265 bug 5192 mean we no longer need to have such short AST timeouts
5266 so ldlm_timeout has been increased.
5269 Frequency : occasional during --force or --failover shutdown under load
5270 Bugzilla : 5949, 4834
5271 Description: Server oops/LBUG if stopped with --force or --failover under load
5272 Details : a collection of import/export refcount and cleanup ordering
5273 issues fixed for safer force cleanup
5276 Frequency : only filesystems larger than 120 OSTs
5277 Bugzilla : 5990, 6223
5278 Description: lfs getstripe would oops on a very large filesystem
5279 Details : lov_getconfig used kfree on vmalloc'd memory
5282 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5283 Bugzilla : 6242, 6243
5284 Description: reading from files that had been truncated to a non-zero size
5285 but never opened returned no data
5286 Details : ll_file_read() reads zeros from no-object files to EOF
5291 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5292 Details : There's little harm in aborting MDS/OSS recovery and letting it
5293 try again, so I removed the LASSERT and return an error instead.
5295 Severity : enhancement
5297 Description: New debugging infrastructure for tracking down data corruption
5298 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5299 (b) cover more of the client-side code path, and (c) try to narrow
5300 down where problems occurred
5304 Bugzilla : 3819, 4364, 4397, 6313
5305 Description: Racing close and eviction MDS could cause assertion in mds_close
5306 Details : It was possible to get multiple mfd references during close and
5307 client eviction, leading to one thread referencing a freed mfd.
5309 Severity: : enhancement
5310 Bugzilla : 3262, 6359
5311 Description: Attempts to reconnect to servers are now more aggressive.
5312 Details : This builds on the enhanced upcall-less recovery that was added
5313 in 1.4.2. When trying to reconnect to servers, clients will
5314 now try each server in the failover group every 10 seconds. By
5315 default, clients would previously try one server every 25 seconds.
5320 Description: After recovery, certain operations trigger a failed
5321 assertion on a client.
5322 Details : Failing over an mds, using lconf -d --failover, while a
5323 client was doing a readdir() call would cause the client to
5324 LBUG after recovery completed and the readdir() was resent.
5326 Severity : enhancement
5328 Description: Default groups are now added by lconf
5329 Details : You can now run lconf --group <servicename> without having to
5330 manually add groups with lmc.
5333 Frequency : occasional
5335 Description: Nodes with an elan id of 0 trigger a failed assertion
5338 Frequency : always when accessing e.g. tty/console device nodes
5340 Description: tty and some other devices nodes cannot be used on lustre
5341 Details : file's private_data field is used by device data and lustre
5342 values in there got lost. New field was added to struct file to
5343 store fs-specific private data.
5346 Frequency : when exporting Lustre via NFS
5348 Description: NFSD failed occasionally when looking up a path component
5349 Details : NFSD is looking up ".." which was broken in ext3 directories
5350 that had grown large enough to become hashed.
5353 Frequency : Clusters with multiple interfaces not on the same subnet
5355 Description: Nodes will repeatedly try to reconnect to an interface which it
5356 cannot reach and report an error to the log.
5357 Details : Extra peer list entries will be created by lconf with some peers
5358 unreachable. lconf now validates the peer before adding it.
5361 Frequency : Only if a default stripe is set on the filesystem root.
5363 Description: Setting a default stripe on the filesystem root prevented the
5364 filesystem from being remounted.
5365 Details : The client was sending extra request flags in the root getattr
5366 request and did not allocate a reply buffer for the dir EA.
5369 Frequency : occasional, higher if lots of files are accessed by one client
5370 Bugzilla : 6159, 6097
5371 Description: Client trips assertion regarding lsm mismatch/magic
5372 Details : While revalidating inodes the VFS looks up inodes with ifind()
5373 and in rare cases can find an inode that is being freed.
5374 The ll_test_inode() code will free the lsm during ifind()
5375 when it finds an existing inode and then the VFS later attaches
5376 this free lsm to a new inode.
5380 Bugzilla : 6422, 7030
5381 Description: MDS deadlock between mkdir and client eviction
5382 Details : Creating a new file via mkdir or mknod (starting a transaction
5383 and getting the ns lock) can deadlock with client eviction
5384 (gets ns lock and trying to finish a synchronous transaction).
5387 Frequency : occasional
5388 Description: While starting a server, the fsfilt_ext3 module could not be
5390 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5391 kernels. Previously, lconf would still use the ext3 name
5392 when trying to load modules. Now, it will correctly use
5393 ext3 on 2.4 and ldiskfs on 2.6.
5395 Severity : enhancement
5396 Description: The default stripe count has been changed to 1
5397 Details : The interpretation of the default stripe count (0, to lfs
5398 or lmc) has been changed to mean striping across a single
5399 OST, rather than all available. For general usage we have
5400 found a stripe count of 1 or 2 works best.
5402 Severity : enhancement
5403 Description: Add support for compiling against Cray portals.
5404 Details : Conditional compiling for some areas that are different
5408 Frequency : occasional
5409 Bugzilla : 6409, 6834
5410 Description: Creating files with an explicit stripe count may lead to
5411 a failed assertion on the MDS
5412 Details : If some OSTs are full or unavailable, creating files may
5413 trigger a failed assertion on the MDS. Now, Lustre will
5414 try to use other servers or return an error to the
5418 Frequency : occasional
5420 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5421 caused lock timeout during readahead (since 1.4.2).
5422 Details : Processes doing readahead might match a lock that hasn't been
5423 granted yet if there are overlapping and conflicting lock
5424 requests. The readahead process waits on ungranted lock
5425 (original lock is CBPENDING), while OST waits for that process
5426 to cancel CBPENDING read lock and eventually evicts client.
5428 Severity : enhancement
5430 Description: Initial enabling of flock support for clients
5431 Details : Implements fcntl advisory locking and file status functions.
5432 This feature is provided as an optional mount flag (default
5433 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5434 locking are implemented yet, and those that are are not guaranteed
5435 to be completely correct in production environments.
5436 mount -t lustre -o [flock|noflock] ...
5439 Frequency : occasional
5441 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5442 assertion in the ext3 JBD (journaling) layer.
5443 Details : The b_committed_data struct is protected by the big kernel lock
5444 in 2.4 kernels, serializing journal_commit_transaction() and
5445 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5446 access to this struct. In 2.6 kernels there is finer grained
5447 locking to improve SMP performance of the JBD layer.
5451 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5456 Description: merge and cleanup kernel patches.
5459 -----------------------------------------------------------------------------
5461 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5463 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5464 versions of Lustre. Please update all servers and clients to
5465 version 1.4.2 or later at the same time. You must also run
5466 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5467 to update the configuration logs.
5469 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5470 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5471 - swab llog records of type '0' so we get proper header size/idx (5861)
5472 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5473 - fix rename of one directory over another leaking an inode (5953)
5474 - avoid SetPageDirty on 2.6 (5981)
5475 - don't re-add just-being-destroyed locks to the waiting list (5653)
5476 - when creating new directories, inherit the parent's custom
5477 striping settings if present parent (3048)
5478 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5479 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5480 - don't hold client locks on temporary worklist from l_lru (5666)
5481 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5482 - automatically evict dead clients (5921)
5483 - Update file size properly in create+truncate+fstat case (6196)
5484 - Do not unhash mountpoint dentries, do not allow removal of
5486 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5487 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5488 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5489 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5490 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5491 - Correctly handle reads of files with no objects (6243)
5492 - lctl recover will also mark a device active if deactivate used (5933)
5494 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5495 - allow --write-conf on an MDS with different nettype than client (5619)
5496 - don't write config llogs to MDS for mounts not from that MDS (5617)
5497 - lconf should create multiple TCP connections from a client (5201)
5498 - init scripts are now turned off by default; run chkconfig --on
5499 lustre and chkconfig --on lustrefs to use them
5500 - upcalls are no longer needed for clients to recover to failover
5502 - add --abort-recovery option to lconf to abort recovery on device
5504 - add support for an arbitrary number of OSTs (3026)
5505 - Quota support protocol changes.
5506 - forward compatibility changes to wire structs (6007)
5507 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5508 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5509 - improved "lustre status" script
5510 - initialize blocksize for non-regular files (6062)
5511 - added --disable-server and --disable-client configure options (5782)
5512 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5513 - Vanilla 2.4.29 support
5514 - increase maximum number of obd devices to 520 (6242)
5515 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5516 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5517 SLES 9 2.6.5 kernels
5518 - limit stripes per file to 160 (the maximum EA size) (6093)
5520 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5523 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5524 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5525 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5526 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5527 - small ext3 extents cleanups and fixes (5733)
5528 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5529 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5530 - avoid needless client->OST connect, fix handle mismatch (5317)
5531 - fix DLM error path that led to out-of-sync client, long delays (5779)
5532 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5533 - fix several locking issues related to i_size (5492,5624,5654,5672)
5534 - don't move pending lock onto export if it is already evicted (5683)
5535 - fix kernel oops when creating .foo in unlinked directory (5548)
5536 - fix deadlock in obdfilter statistics vs. object create (5811)
5537 - use time_{before,after} to avoid timer jiffies wrap (5882)
5538 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5539 - Do not leak request if server was not able to process it (5154)
5540 - If mds_open unable to find parent dir, make that negative lookup(5154)
5541 - don't create new directories with extent-mapping (5909, 5936)
5543 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5544 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5545 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5546 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5547 - support for RHEL4 kernel on i686 (5773)
5548 - provide error messages when incompatible logs are encountered (5898)
5550 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5551 * version 1.4.0.10 (1.4.1 release candidate 1)
5553 - don't keep a lock reference when lock is not granted (4238)
5554 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5555 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5556 - drop import inflight refcount on signal_completed_replay error (5255)
5557 - unlock page after async write error during send (3677)
5558 - handle missing objects in filter_preprw_read properly (5265)
5559 - no transno return for symlink open, don't save no-trasno open (3440)
5560 - don't try to complete elan receive that already failed (4012)
5561 - free RPC server reply state on error (5406)
5562 - clean up thread from ptlrpc_start_thread() on error (5160)
5563 - readahead could read extra page into cache that wasn't ejected (5388)
5564 - prevent races in class_attach/setup/cleanup/detach (5260)
5565 - don't dereference de->d_inode after l_dput of de (5458)
5566 - use "int" for stripe value returned from lock_to_stripe (5544)
5567 - mballoc allocation and error-checking fixes in 2.6 (5504)
5568 - block device patches to fix I/O request sizes in 2.6 (5482)
5569 - look up hostnames for IB nals (5602)
5570 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5571 - don't start multiple acceptors for the same port (5277)
5572 - fix incorrect LASSERT in mds_getattr_name (5635)
5573 - export a proc file for general "ping" checking (5628)
5574 - fix "lfs check" to not block when the MDS is down (5628)
5576 - service request history (4965)
5577 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5578 - create an "evict_client" /proc entry on OSTs, like the MDS has
5579 - fix mount usage message, return errors per mount(8) (5168)
5580 - change grep [] to grep "[]" in tests so they work in more UMLs
5581 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5582 - remove ancient LOV_MAGIC_V0 EA support (5047)
5583 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5584 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5585 - allow readahead allocations to fail when low on memory (5383)
5586 - mmap locking landed again, after considerable improvement (2828)
5587 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5589 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5592 - send OST transaction number in read/write reply to free req (4966)
5593 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5594 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5595 - the watchdog thread now runs as interruptible (5246)
5596 - flock/lockf fixes (but it's still disabled, pending 5135)
5597 - don't use EXT3 constants in llite code (5094)
5598 - memory shortage at startup could cause assertion (5176)
5600 - reorganization of lov code
5601 - single portals codebase
5603 - add extents/mballoc support (5025)
5604 - direct I/O reads in the obdfilter (4048)
5605 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5607 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5610 - send OST transaction number in read/write reply to free req (4966)
5611 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5612 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5613 - the watchdog thread now runs as interruptible (5246)
5614 - handle missing objects in filter_preprw_read properly (5265)
5615 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5616 - drop import inflight refcount on signal_completed_replay error (5255)
5617 - unlock page after async write error during send (3677)
5618 - return original error code on reconstructed replies (3761)
5619 - no transno return for symlink open, don't save no-trasno open (3440)
5621 - add pid to ldlm debugging output (4922)
5622 - bump the watchdog timeouts -- we can't handle 30sec yet
5623 - extra debugging for orphan dentry/inode bug (5259)
5625 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5628 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5629 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5630 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5631 - fix NULL dereference in /proc/sys/portals/routes (4827)
5632 - allow failed mdc_close() operations to be interrupted (4561)
5633 - stop precreate on OST before MDS would time out on it (4778)
5634 - don't send partial-page writes before EOF from client (4410)
5635 - discard client grant for sub-page writes on large-page clients (4520)
5636 - don't free dentries not owned by NFS code, check generation (4806)
5637 - fix lsm leak if mds_create_objects() fails (4801)
5638 - limit debug_daemon file size, always print CERROR messages (4789)
5639 - use transno after validating reply (3892)
5640 - process timed out requests if import state changes (3754)
5641 - update mtime on OST during writes, return in glimpse (4829)
5642 - add mkfsoptions to LDAP (4679)
5643 - use ->max_readahead method instead of zapping global ra (5039)
5644 - don't interrupt __l_wait_event() during strace
5646 - add software watchdogs to catch hung threads quickly (4941)
5647 - make lustrefs init script start after nfs is mounted
5648 - fix CWARN/ERROR duplication (4930)
5649 - return async write errors to application if possible (2248)
5650 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5651 - print NAL number in %x format (4645)
5652 - update barely-supported suse-2.4.21-171 series (4842)
5653 - support for sles 9 %post scripts
5654 - support for building 2.6 kernel-source packages
5655 - support for sles km_* packages
5657 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5660 - ignore -ENOENT errors in osc_destroy (3639)
5661 - notify osc create thread that OSC is being cleaned up (4600)
5662 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5663 - reconstruct ost_handle() like mds_handle() (4657)
5664 - create a new thread to do import eviction to avoid deadlock (3969)
5665 - let lconf resolve symlinked-to devices (4629)
5666 - don't unlink "objects" from directory with default EA (4554)
5667 - hold socknal file ref over connect in case target is down (4394)
5668 - allow more than 32000 subdirectories in a single directory (3244)
5669 - fix blocks count for O_DIRECT writes (3751)
5670 - OST returns ENOSPC from object create when no space left (4539)
5671 - don't send truncate RPC if file size isn't changing (4410)
5672 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5673 - bind to privileged port in socknal and tcpnal (3689)
5675 - rate limit CERROR/CWARN console message to avoid overload (4519)
5676 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5677 - basic mmap support (3918)
5678 - kernel patch series update from b1_4 (4711)
5680 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5683 - avoid crash during MDS cleanup with OST shut down (2775)
5684 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5685 - don't use bad inodes on the MDS (3744)
5686 - dynamic object preallocation to improve recovery speed (4236)
5687 - don't hold spinlock over lock dumping or change debug flags (4401)
5688 - don't zero obd_dev when it is force cleaned (3651)
5689 - print grants to console if they go negative (4431)
5690 - "lctl deactivate" will stop automatic recovery attempts (3406)
5691 - look for existing locks in ldlm_handle_enqueue() (3764)
5692 - don't resolve lock handle twice in recovery avoiding race (4401)
5693 - revalidate should check working dir is a directory (4134)
5695 - don't always mark "slow" obdfilter messages as errors (4418)
5697 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5700 - don't close LustreDB during write_conf until it is done (3860)
5701 - fix typo in lconf for_each_profile (3821)
5702 - allow dumping logs from multiple threads at one time (3820)
5703 - don't allow multiple threads in OSC recovery (3812)
5704 - fix debug_size parameters (3864)
5705 - fix mds_postrecov to initialize import for llog ctxt (3121)
5706 - replace config semaphore with spinlock (3306)
5707 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5708 - don't allow enqueue to complete on a destroyed export (3822)
5709 - down write_lock before checking llog header bitmap (3825)
5710 - recover from lock replay timeout (3764)
5711 - up llog sem before sending rpc (3652)
5712 - reduce ns lock hold times when setting kms (3267)
5713 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5714 - fix NULL deref and obd_dev leak on setup error (3312)
5715 - replace some LBUG about llog ops with error handling (3841)
5716 - don't match INVALID dentries from d_lookup and spin (3784)
5717 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5718 - fix invalid assertion in ptlrpc_set_wait (3880)
5720 - add libwrap support for the TCP acceptor (3996)
5721 - add /proc/sys/portals/routes for non-root route listing (3994)
5722 - allow setting MDS UUID in .xml (2580)
5723 - print the stack of a process that LBUGs (4228)
5725 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5728 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5729 - make sure to unset replay flag from failed open requests (3440)
5730 - if default stripe count is 0, use OST count for inode size (3636)
5731 - update parent mtime/ctime on client for create/unlink (2611)
5732 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5733 - free recovery state on server during a forced cleanup (3571)
5734 - unregister_reply for resent reqs (3063)
5735 - loop back devices mounting and status check on 2.6 (3563)
5736 - fix resource-creation race that can provoke i_size == 0 (3513)
5737 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5738 - more debugging for page-accounting assertion (3746)
5739 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5740 - avoid deadlock after precreation failure (3758)
5741 - fix race and lock order deadlock in orphan handling (3450, 3750)
5742 - add validity checks when grabbing inodes from l_ast_data (3599)
5744 - add /proc/.../recovery_status to obdfilter (3428)
5745 - lightweight CDEBUG infrastructure, debug daemon (3668)
5746 - change default OSC RPC parameters to be better on small clusters
5747 - turn off OST read cache for files smaller than 32MB
5748 - install man pages and include them in rpms (3100)
5749 - add new init script for (un)mounting lustre filesystems (2593)
5750 - run chkconfig in %post for init scripts (3701)
5751 - drop scimac NAL (unmaintained)
5753 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5756 - clean kiobufs before and after use (3485)
5757 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5758 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5759 - change init script to fail more clearly if not run as root (1528)
5760 - allow clients to reconnect during replay (1742)
5761 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5762 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5763 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5764 - fix cond_resched (3554)
5765 - only evict unfinished clients after recovery (3515)
5766 - allow bulk resend, prevent data loss (3570)
5767 - dynamic ptlrpc request buffer allocation (2102)
5768 - don't allow unlinking open directory if it isn't empty (2904)
5769 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5770 - remove extraneous obd dereference causing LASSERT failure (3334)
5771 - don't use get_cycles() when creating temp. files on the mds (3156)
5772 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5773 - handle EEXIST for set-stripe, set proper directory name (3336)
5775 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5776 - fix ksocknal_fmb_callback() error messages (2918)
5778 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5781 - don't copy lvb into (possibly NULL) reply on error (2983)
5782 - don't deref dentry after dput, don't free lvb on error (2922)
5783 - use the kms to determine writeback rpc length (2947)
5784 - increment oti_logcookies when osc is inactive (2948)
5785 - update client's i_blocks count via lvb messages (2543)
5786 - handle intent open/close of special files properly (1557)
5787 - mount MDS with errors=remount-ro, like obdfilter (2009)
5788 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5789 - don't use cancelling-locks' kms values (2947)
5790 - use highest lock extent for kms, not last one (2925)
5791 - don't dereference ERR_PTR() dentry in error handling path (3107)
5792 - fix thread race in portals_debug_dumplog() (3122)
5793 - create lprocfs device entries at setup instead of at attach (1519)
5794 - common AST error handler, don't evict client on completion race (3145)
5795 - zero nameidata in detach_mnt in 2.6 (3118)
5796 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5797 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5798 - initialize RPC timeout timer earlier for 2.6 (3219)
5799 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5800 - print nal/nid for unknown nid (3258)
5801 - additional checks for oscc recovery before doing precreate (3284)
5802 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5803 - don't crash in mdc_close for bad permissions on open (3285)
5804 - zero i_rdev for non-device files (3147)
5805 - clear page->private before handing to FS, better assertion (3119)
5806 - tune the read pipeline (3236)
5807 - fix incorrect decref of invalidated dentry (2350)
5808 - provide read-ahead stats and refine rpc in flight stats (3328)
5809 - don't hold journal transaction open across create RPC (3313)
5810 - update atime on MDS at close time (3265)
5811 - close LDAP connection when recovering to avoid server load (3315)
5812 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5813 - don't leak open file on MDS after open resend (3325)
5814 - serialize filter_precreate and filter_destroy_precreated (3329)
5815 - loop device shouldn't call sync_dev() for nul device (3092)
5816 - clear page cache after eviction (2766)
5817 - resynchronize MDS->OST in background (2824)
5818 - refuse to mount the same filesystem twice on same mountpoint (3394)
5819 - allow llmount to create routes for mounting behind routers (3320)
5820 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5821 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5822 - fix rare problem with rename on htree directories (3417)
5824 - allow default OST striping configuration per directory (1414)
5825 - fix compilation for qswnal for 2.6 kernels (3125)
5826 - increase maximum number of MDS request buffers for large systems
5827 - change liblustreapi to be useful for external progs like lfsck (3098)
5828 - increase local configuration timeout for slow disks (3353)
5829 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5831 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5834 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5835 - don't overwrite extent policy data in reply if lock was blocked (2901)
5836 - drop filter export grants atomically with removal from device (2663)
5837 - del obd_self_export from work_list in class_disconnect_exports (2908)
5838 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5839 - swab reply message in mdc_close, other PPC fixes (2464)
5840 - fix destroying of named logs (2325)
5841 - overwrite old logs when running lconf --write_conf (2264)
5842 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5843 - fix race in target_handle_connect (2898)
5844 - mds_reint_create() should take same inode create lock (2926)
5845 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5846 - don't close files for self_export to avoid uninitialized obd (2936)
5847 - allow MDS with the same name as client node (2939)
5848 - hold dentry reference for closed log files for unlink (2325)
5849 - reserve space for all logs during transactions (2059)
5850 - don't evict page beyond end of stripe extent (2925)
5851 - don't oops on a deleted current working directory (2399)
5852 - handle hard links to targets without a parent properly (2517)
5853 - don't dereference NULL lock when racing during eviction (2867)
5854 - don't grow lock extents when lots of conflicting locks (2919)
5856 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5859 - account for cache space usage on clients to avoid data loss (974)
5860 - lfsck support in lustre kernel code (2349)
5861 - reduce journal credits needed for BRW writes (2370)
5862 - orphan handling to avoid losing space on client/server crashes
5863 - ptlrpcd can be blocked, stopping ALL progress (2477)
5864 - use lock value blocks to assist in proper KMS, faster stat (1021)
5865 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5866 - recovery for initial connections (2355)
5867 - fixes for mds_cleanup_orphans (1934)
5868 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5869 - block all file creations until orphan recovery completes (1901)
5870 - client remove rq_connection from request struct (2423)
5871 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5872 - recovery timer race (2670)
5873 - mdc_close recovey bug (2532)
5874 - ptlrpc cleanup bug (2710)
5875 - mds timeout on local locks (2588)
5876 - namespace lock held during RPCs (2431)
5877 - handle interrupted sync write properly (2503)
5878 - don't try to handle a message that hasn't been replied to (2699)
5879 - client assert failure during cleanup after abort recovery (2701)
5880 - leak mdc device after failed mount (2712)
5881 - ptlrpc_check_set allows timedout requests to complete (2714)
5882 - wait for inflight reqs when ptlrpcd finishes (2710)
5883 - make sure unregistered services are removed from the srv_list
5884 - reset bulk XID's when resending them (caught by 1138 test)
5885 - unregister_bulk after timeout
5886 - fix lconf error (2694)
5887 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5888 - readahead locks pages, leaves pending causing memory pressure (2673)
5889 - increase OST request buffers to 4096 on large machines (2729)
5890 - fix up permission of existing directories in simple_mkdir (2661)
5891 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5892 - don't assign transno to errored transactions (2742)
5893 - don't delete objects on OST if given a bogus objid from MDS (2751)
5894 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5895 - if rq_no_resend, then timeout request after recovery (2432)
5896 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5897 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5898 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5899 - MDS recovery shouldn't delete objects that it has given out (2730)
5900 - if enqueue arrives after completion, don't clobber LVB (2819)
5901 - don't unlock pages twice when trigger_group_io returns error (2814)
5902 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5903 - don't write pages to disk if there was an error (1450)
5904 - don't ping imports that have recovery disabled (2676)
5905 - take buffered bytes into account when balancing socknal conn (2817)
5906 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5907 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5908 - remove little-endian swabbing of llog records (1987)
5909 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5910 - retry reposting request buffers if they fail (1191)
5911 - grow extent at grant time to avoid granting a revoked lock (2809)
5912 - lock revoke doesn't evict page if covered by a second lock (2765)
5913 - disable VM readahead to avoid reading outside lock extents (2805)
5915 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5916 - updated kernel patches for hp-2.4.20 kernel (2681)
5918 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5921 - fix truncated write corruption (2366)
5922 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5924 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5925 - ensure len(uuid) < 37 in lmc (1171)
5926 - fix ia64 OOPS in llog_test (2255)
5927 - zero end of page at obdfilter for partial page writes (2648)
5928 - don't leave stale dentries around after renames (bug 2428)
5929 - fix timeouts when evicting a client with a single lock held (2642)
5930 - set deadline for the initial HELLO message to drain (2634)
5931 - print out dotted-quad IP addresses in the socknal (2302)
5933 - additional debugging for MDS client eviction problem (2443)
5934 - fix mkfsoptions support for osts (2603, 2604)
5936 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5939 - add series for the vanilla 2.6.0 kernel
5940 - add series for the vanilla 2.4.24 kernel
5941 - add series for a cray x86/64 UL kernel drop
5942 - fix xattr patches for the vanilla 2.4.19 series
5944 - generate true UUIDs in lmc (1171)
5945 - have portals stack dumping break in UML (2466)
5946 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5947 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5948 - wake callback waiting threads on client eviction (2460)
5949 - Add --ptldebug and --subsystem to lmc (1719)
5950 - update assertion to allow safe interrupt allocation
5951 - set rq_no_resend for cancel requests (2432)
5952 - recalculate ptlrpcd timeout after resend (2494)
5953 - call vfs_rmdir when removing pending directories (2368)
5954 - fix renaming a file to itself (2429)
5955 - lmc creates a default one-stripe lov (2454)
5956 - expand procfs space to handle large clusters (2326)
5957 - increase UML stack to avoid overflow
5958 - update lconf's list of debug and subsystem masks
5959 - fix lfs find --obd (2510)
5960 - /proc tunable for disabling filter read caching (2591)
5961 - stop rpm packages from altering slapd.conf (2301)
5962 - disable nagle in the socknal under 0conf (2578)
5963 - choose mds inode size based on stripe count (2572)
5964 - fix kernel-source rpm problems (2516)
5966 - add --disable-doc to avoid pdf generation (2421)
5967 - update documentation, tests, type-os, comments
5968 - avoid format warnings on ia64
5969 - remove the TOE NAL
5970 - tiny code cleanups by removing unused fields
5972 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5975 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5976 - remove the most blatant lies from BUILDING file (2371)
5977 - change default debug level to reasonable production setting
5978 - reduce client side cache size to reduce cache flush time
5979 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5980 - make TCP zerocopy and pinger support enabled by default (2476)
5981 - sync writes completed after process exits caused crashes (2319)
5982 - maintain correct mount count on the MDS (2356)
5983 - backout 1557, because 2316 wasn't really fixed
5984 - better file I/O statistics gathering in /proc
5985 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5986 - another kernel patch to fix zero-copy TCP function export
5987 - don't take duplicate lock when processing re-sent getattr (2420)
5988 - lctl uses obd_self_export instead of creating new conn (2353)
5989 - MDS/OST recovery case which requires object creation asserted (2425)
5990 - move lfs from /usr/sbin to /usr/bin in packages
5991 - fix race between mds_client_add and mds_client_free (2417)
5992 - use kmalloc instead of slabs in portals (2430)
5993 - don't create duplicate records when a failover MDS is present (2442)
5994 - remove unnecessary mount age check (2332)
5995 - don't remove directory inodes from locks prematurely (2451)
5996 - don't break if MDS service name is the same as hostname (2103)
5997 - fix races in client write RPC generation when cache full (2482)
5999 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6002 - remove now-unused request->rq_obd (278)
6003 - if an allocation fails, print out how much memory we've used (1933)
6004 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6005 - add the "configurable stack size" patch to most series files (1256)
6006 - ability to write large log records, for 100+ OST configs (2306)
6007 - fix NULL deref when filter_prep fails (2314)
6008 - fix operator precedence error in filter_sync
6009 - dynamic allocation of socknal TX descriptors (2315)
6010 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6011 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6012 - ensure that CWARN messages in Portals always get to the syslog
6013 - __init/__exit are not for prototype decls (ldlm_init/exit)
6014 - x86-64 compile warning fixes
6015 - fix gateway LMC keyword conflict (2318)
6016 - fix MDS lock inversions in getattr/reint paths (1844)
6017 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6018 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6019 - DLM race condition prevented some lock evictions (2328)
6020 - ENOMEM detection and retry on socknal sends (2230)
6021 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6022 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6023 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6024 - better error messages when a client is rejected during recovery (1505)
6025 - avoid cancelling locks which were never granted, after failure (2330)
6026 - fix i_sem/journal inversion in mds_client_add (2333)
6027 - fix truncate/getattr lock cycle deadlock (2334)
6028 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6029 - fix two rare exit paths which could leak an l_lock() ref (2321)
6030 - fencepost error in MDS/OST orphan recovery (2226)
6031 - make log record alignment 8 bytes (1988)
6032 - lstripe now fails when requested offset > ost_count (2237)
6033 - ensure that all kernel series have a complete list.h (1607)
6034 - fix crashes in special-file operations (2316)
6035 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6037 - allow configurable automake binary, for testing new versions
6038 - small update to the lfs documentation
6040 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6042 * fix negative export reference count in fsfilt_sync (2312)
6044 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6045 * release candidate 0.9.1
6047 - orphans are moved into the PENDING directory for possible recovery
6048 - replayed opens now open by fid for orphan/rename safety (1042)
6049 - last close of an orphan inode generates a transno (683)
6050 - chdir() and mount() now pin the directory entry (1020)
6051 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6052 - discard very old requests without processing them (1502)
6053 - remove some common, well-understood CERRORs (1505)
6054 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6055 - clear "grant" flags in OST replies until OST grant code lands (1644)
6056 - fix read performance by not clobbering i_blksize on client (1598)
6057 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6058 - make LRU size a /proc tunable, clears locks when reduced (707)
6059 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6060 - server locks take references on exports now (1558)
6061 - build fixes for 2.4.20-rh trees (1663)
6062 - return an error from lov_create if all OSCs are inactive (1751)
6063 - fix import levels when a reconnect happens without a timeout (1597)
6064 - exit early from mds_open if we get a lookup error (1749)
6065 - partial page read at EOF wouldn't wait for disk before sending (1642)
6066 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6067 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6068 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6069 - write/truncate lock inversion (1639)
6070 - Don't auto-load obdclass, portals modules during cleanup (1495)
6071 - fix timestamps from jumping to "now" (1763)
6072 - extra journal assertions (1648)
6073 - add an extra multiunlink test (1771)
6074 - fix read_record/write_record API (1776)
6075 - fix leak of offset_extent, possible incorrect i_size later (1772)
6076 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6077 - Debugging for the kqswnal_get_idle_tx problems (1820)
6078 - Allow recovery to be attempted multiple times (1536)
6079 - Write out MDS last_rcvd file after it is first created (1600)
6080 - Fix tx_descriptor leak in failed transmit situations (1827)
6081 - ext3 journaling fixes for assertion failure after IO error (1871)
6082 - class_export_put() on freed export after completion AST error (1896)
6083 - Fix revalidate looping in VFS (1322)
6084 - Don't access a freed export during MDS_REINT timeout (1521)
6085 - Add open-unlink recovery support on the MDS (1673,1764)
6086 - Return an error if no MDS data was read from last_rcvd (1946)
6087 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6088 - Don't setup a disk device that doesn't match exported UUID (317)
6089 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6090 - avoid committing NULL handle in force close
6091 - local.sh is now a one-stripe LOV configuration
6092 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6093 - fix ext3 htree duplicate directory entry corruption (1516)
6094 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6095 - update to new LOV EA format (2097)
6096 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6098 2003-06-15 Phil Schwan <phil@clusterfs.com>
6101 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6102 - per-import/export recovery handling (958, 931, 959)
6103 - multiple last-rcvd slots, for serving multiple FSes (949)
6104 - connections are again shared between multiple imp/exports (963, 964)
6105 - "umount -f" would hang if any requests needed to be sent (393, 978)
6106 - avoid pinning large req buffer by copying for queued messages (989)
6107 - add "uuid" to "lctl device" command to help upcalls (991)
6108 - "open" RPCs with transnos would confuse recovery counters (1037)
6109 - do proper endian conversion of all wire messages (288, 340, 891)
6110 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6111 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6112 - fix ptlrpc_connection leak in target_handle_connect (1174)
6113 - fix import refcounting bug in OST and MDS cleanup (1134)
6114 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6115 - fix very unlikely obd_types race condition (501)
6116 - remove osc_open hack for echo_client (1187)
6117 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6118 - a failure in read_inode2 leads to deadlock (1139)
6119 - cancel ack-locks as soon as transaction is committed (1072)
6120 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6121 - make sure to commitrw after any preprw to avoid deadlock (1162)
6122 - failing to execute a file in a lustre FS would lock inode (1203)
6123 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6124 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6125 - obd_uuid2tgt would walk off the end of the list (1255)
6126 - on IA64 the osc would give portals incorrect bulk size (1258)
6127 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6128 - fix lock inversion caused by new llite matching code (1282)
6129 - limit the number of dirty pages on a client to 10MB (1286)
6130 - timed out locks were not being corrected cancelled (1289)
6131 - fix O_DIRECT above 4GB on IA-32 (1292)
6132 * major user-visible changes
6133 - fail out/fail over policy now controlled by the upcall (993)
6135 - add OBD_PING to check server availability and failure (954)
6136 - lustre messages are now sent in sending host order (288, 340, 891)
6137 - add eadatalen to MDS getattr reply (340)
6138 - OST read replies may contain second buffer, with per-page status (593)
6140 2003-03-11 Phil Schwan <phil@clusterfs.com>
6143 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6144 - failed open()s could cause deadlock; fixed (867, 869)
6145 - stop cancelling OST locks when files are closed (481)
6146 - overlapping XID spaces caused network corruption (851, 853)
6147 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6148 - setattr_raw added, to do single-RPC, server-side setattrs
6149 - lmc/lconf syntax change for OST UUIDs
6150 - fix crashy race condition between ptlrpc_free_req and osc_close
6151 - don't use request in mdc_enqueue if we hit a timeout (889)
6152 - don't set the inode i_size for regular files from the MDS (896)
6153 - handle out of order completion AST (842)
6154 - don't LBUG if a lock request times out after receiving AST (913)
6155 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6156 - if a bad lock AST arrives, send an error instead of dropping entirely
6157 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6158 - fix leak in bulk IO when only partially completed (899, 900, 926)
6159 - fix O_DIRECT for ia64 (55)
6160 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6161 - C-z after timeout could hang a process forever; fixed (977)
6163 - client-side I/O cache (678, 924, 929, 941, 970)
6165 - READPAGE and SETATTRs which don't take server-side locks get
6168 2003-02-11 Phil Schwan <phil@clusterfs.com>
6171 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6172 - Fix fsfilt_extN_readpage() to read a full page of directory
6173 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6174 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6175 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6176 - Most intent operations were removed, in favour of a new RPC mode
6177 that does a single RPC to the server and bypasses most of the VFS
6178 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6179 - Aggressively cancel local locks on DLM servers
6180 - mds_reint_unlink sends EA to the client if it's the last nlink.
6181 client uses that EA to unlink OST objects.
6182 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6183 - recursive symlinks were fixed (439)
6184 - fixed NULL deref in DEBUG_REQ
6185 - filter_update_lastobjid no longer calls sync, which annoyed extN
6186 - fixed multi-client small-writes to a single file problem (445)
6187 - fixed mtime updates during file writes (607)
6188 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6189 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6190 - fixed Linux OST size reporting problem (444, 656)
6191 - OST now updates object mtime with writes or setattr (607, 619)
6192 - client verifies file size before zeroing page past EOF (445)
6193 - OST now writes last allocated objid to disk with allocation (108)
6194 - LOV on echo now works (409)
6196 - mds_reint_unlink sends a new buffer, with the EA included. this
6197 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6198 set if a regular file was being unlinked, and it was the last link
6199 - use PtlGet from the target for bulk writes (315)
6200 - OST now updates object mtime with writes or setattr (607, 619)
6201 - LDLM now has a grant-time callback to revalidate locked items, if
6203 - Many MDS operations were reorganized to combat race conditions
6205 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6206 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6208 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6211 - Fully reactivate OST imports after reconnection (512, others)
6212 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6213 - More graceful error handling for truncating on dead OST (515)
6214 - Don't error out unless we're actually accessing dead stripes (474)
6215 - Fix garbage sizes when stripes are missing (410)
6216 - LRU counters were broken, causing constant lock purge (433, 432)
6217 - garbage on read from stripes with failed OSTs (441)
6218 - mark OSCs as active before reconnecting during recovery (438)
6219 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6220 - lfind did not preserve OST order in output (443)
6221 - symlinks cause hung clients, incorrect data (439)
6222 - stop dereferencing request after dropping refcount (457)
6223 - don't LASSERT(spin_is_locked) on non-SMP (455)
6224 - fixes for many rename() bugs
6225 - fstat didn't correctly synchronize attributes (399)
6226 - server must handle lock cancellation during blocking AST prep (487)
6227 - bulk descriptors were free()d too soon (511)
6228 - fix paths in lconf, which would load incorrect modules (451, 507)
6229 - fix confusing lconf 'host not found' error message (386)
6230 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6231 - fix race condition in mdc_blocking_ast() for inode access (526)
6232 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6233 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6234 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6235 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6236 - fix lstripe in several ways (kernel side) (527)
6237 - fix request leak in ldlm_cli_enqueue (262)
6238 - incorrect OSC was marked inactive after OST failure
6239 - call mds_fs_cleanup before unmounting filesystem (524)
6240 - fix races between taking ns_lock and ldlm_lock_change_resource
6241 - fix races updating LOV export open file list
6242 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6243 - fix recovery NULL deref in ldlm_cli_cancel_unused
6244 - fix some DLM races by using new hash table for lock handles (419)
6245 - permit the client to specify desired inodes, at replay
6246 - duplicate requests when we queue them for replay reintegration
6247 - fix last_rcvd offset calculation
6248 - sync after each recovered transaction, so we always make progress
6249 - never, not always, ERESTART requests without transnos
6250 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6252 - skip replay if the MDS says that the client is already connected
6253 - don't check for a recovery-enabled export to match lctl's UUID
6254 - don't INC_USE_COUNT for phantom exports
6255 - don't crash when cleaning up phantom exports (567)
6256 - don't double-finish or set replay data for errored mdc_open requests
6257 - abort requests when they time out, so we don't get old replies
6258 - send/receive replies for AST messages again
6259 - if the client says that it doesn't have the lock, cancel it on the
6261 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6262 mark it as destroyed, it will all work out when decref is called
6263 - fix module use counts (22, 581)
6265 - ASTs now expect a reply (server cancels lock on error reply)
6267 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6270 - fix many simultaneous client startup (392)
6271 - fix dentry->d_it clobbering
6272 - credentials weren't being shipped for readdir/getattr operations
6273 - remove invalid assertions triggered during some concurrent MD
6275 - proper Lustre versions added (336, 389)
6276 - fix memory leak for create error case (398)
6277 - fix LOV locking bug that would get cli/srv out of sync
6278 - fix echo client over LOV (409)
6279 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6280 - fix double-O_EXCL intent crash (424)
6281 - avoid sending multiple lock CANCELs (352)
6283 - MDS can do multi-client recovery (modulo bugs in new code)
6285 - many updates, edits, cleanups
6287 2002-11-18 Phil Schwan <phil@clusterfs.com>
6290 - fix null d_it dereference (346)
6291 - fix full OST/dbench hang (333)
6292 - fix permission problem with file removal (286)
6293 - fix removal of OSCs from LOV when they fail
6294 - fix NULL deref during bulk timeout (214)
6295 - fix problems related to multiple filesystems on one MDS (241)
6296 - fixed serious subtle metadata locking bugs
6297 - free locks on clients when inodes are removed due to memory
6299 - fix inode pointer in lock data (285)
6300 - partial support for multiple MDS on a single host (241)
6301 - data locks weren't cancelled at clear_inode time (290, 311)
6302 - intent locks could lead to unbounded lock growth (205)
6303 - added a maximum lock count, an LRU list, and a flusher
6304 - fix multiple rename (365)
6305 - properly abstracted the echo client
6306 - OSC locked 1 byte too many; fixed
6307 - rewrote brw callback code:
6308 - fixed recovery bugs related to LOVs (306)
6309 - fixed too-many-pages-in-one-write crash (191)
6310 - fixed (again) crash in sync_io_timeout (214)
6311 - probably fixed callback-related race (385)
6313 - Add capability to MDS protocol
6314 - LDLM cancellations and callbacks on different portals
6316 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6319 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6320 - defer open object destruction to close time (601981, 138)
6321 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6322 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6323 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6324 - don't call obd_setattr in ll_file_release for destroyed objects
6325 * protocol change to lustre_msg: move |version| and add |flags|
6326 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6327 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6328 * added replay of create, unlink, link and rename operations during
6329 MDS failover; recovery should be much more robust now
6330 * remove failed OSCs from LOVs (only lov_create uses this so far)
6331 * the lustre-HOWTO was brought (more) up to date (582544)
6333 2002-10-23 Phil Schwan <phil@clusterfs.com>
6336 - in-use dentries weren't being reused properly (617851)
6337 - prevent multiple LDLM setup (599178)
6338 - fix LOV size calculations for truncate (617853)
6339 - fix client handling of MDS intent errors (POSIX)
6340 - fix permission bug in lovstripe.c test (624321)
6341 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6342 - truncate past end of file could corrupt data
6343 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6344 - a race in recovery could return ETIMEDOUT to apps (623947)
6345 - building outside the source directory was fixed
6346 * the lustre-HOWTO was brought (more) up to date (582544)
6347 * major progress was made on recovery functionality
6349 2002-10-10 Phil Schwan <phil@clusterfs.com>
6352 - recovery deadlock fix
6353 - rm -rf causes LBUG fix (617817)
6354 - file open by multiple tasks fix (618962)
6355 - directory permissions bugs (602707 and 620007)
6356 - journal_stop fixed with locking (611313)
6357 - O_APPEND failures resolved (618273, perhaps 614459)
6358 - lconf PATH fix (619770)
6359 - IA64 build fix (621450)
6360 - RPC buffer sizes scale with amount of memory
6362 2002-10-01 Phil Schwan <phil@clusterfs.com>
6365 - locks would be cancelled without throwing away data pages,
6366 resulting in inconsistent data (605627)
6367 - inode attributes were not always being refreshed (605627, 612449)
6368 - lconf now continues to cleanup after lctl reports an error
6369 - MDS now enforces user permissions (602707)
6370 - lprocfs cleanup fixed, but not yet enabled (614157)
6371 - fixed infinite server hang, should a client not respond to an AST
6372 - avoid going into recovery if user calls readlink() with a buffer
6373 that's too small (613941)
6374 - AST RPCs no longer require replies (614867) -- this may be changed
6375 - don't crash server if client sends an IOV that's too big (611336)
6376 - fixed lock conversion deadlock (611892)
6377 - fixed the following of symlinks (614622)
6378 * recovery: the server can remove locks from a client that dies, other
6379 clients can make progress
6380 * more extN patch fixes
6381 * compile-time configurable ptlrpc buffer allocations
6383 - collaborative read cache document
6384 - Lustre Lite Performance CDR document-in-progress
6386 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6389 - fix typo in patch-2.4.18
6391 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6394 - clear ptlrpc request each time in handle_incoming_request()
6395 - unlink of files now destroys the object on the OST
6397 2002-09-19 Peter Braam <braam@clusterfs.com>
6399 * add hard link support
6400 * change obdfile creation method
6401 * kernel patch changed
6403 2002-09-19 Peter Braam <braam@clusterfs.com>
6406 - stack overflow bug in extN fixed
6408 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6410 * documentation updates
6411 - add man pages for config tools
6412 - update tests/README to describe testing with new config tools
6413 - finish metadata API descriptions
6414 * bug fixes and cleanups
6415 - statfs workaround for 16TB limit
6416 - LOV stripe allocation improved, can stripe on subset of OSTs
6417 - LOV file size/IO offset was wrong for files > 4GB in size
6418 - object EA data was being dropped, caused files to be unreadable
6419 - memory overflow with non-LOV OST caused memory corruption
6420 - fixed regression tests to work with new config tools, obdfilter
6421 - fixed bug when directory size became larger than 1 block
6422 - fixed bug (for single client case) when PWD was deleted
6423 - invalidate local directory pages when doing intent-based ops
6424 - avoid LDLM oops when lock callback contained bad data
6426 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6428 * documentation updates
6429 * bug fixes and cleanups
6430 - configuration tools
6433 - 64-bit compile warnings
6434 - 64-bit internal statfs data
6436 * test_brw on persistent OST devices
6438 * lprocfs (disabled)
6440 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6442 * documentation updates
6443 * bug fixes and cleanups
6444 * configuration tools
6446 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6449 * many small fixes to 0_5_4
6450 * io/network handling
6451 * thinkos in MDS operations
6453 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6456 * crucial basic fixes to 0.5.3
6457 * IOR, Iozone work over Elan
6460 2002-08-07 Phil Schwan <phil@clusterfs.com>
6461 * version 0_5_3, our first alpha
6462 * we use the new Portals iovs
6463 * documentation updates
6464 * bug fixes and cleanups
6465 * small changes in the DLM wire protocol
6467 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6468 * version 0_5_1 with some initial stability,
6469 * locking on MD and file I/O.
6470 * documentation updates
6471 * several bug fixes since 0.5.0
6472 * small changes in wire protocol
6474 2002-07-18 Phil Schwan <phil@clusterfs.com>
6476 * delivered as Lustre Light Alpha
6477 * fixed a crash after handling invalid MDS requests
6478 * fixed directory pages for architectures with non-4k pages sizes
6480 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6481 * release version v0_4_4
6482 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6484 * Fixes a number of interruption problems with OST operations.
6485 * Update documentation for portals header changes
6486 * Move all wire protocol structs/defines to lustre_idl.h
6487 * Fixes symlink length bug.
6488 * Add tcpdump to repository.
6490 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6491 * release version v0_4_3
6492 * Fixes statfs for inodes on extN.
6493 * Fixes bug in runtests which would delete /etc/hosts.
6494 * Use 64-bit object IDs wherever possible (not into VFS though)
6495 Remove ost_get_info, which is unused by lustre, and out of date.
6497 2002-07-03 Peter Braam <braam@clusterfs.com>
6498 * release version v0_4_2 Fixes a lookup error (type not passed)
6499 * move forward to head of Portals
6500 * move forward to latest Lustre kernel
6502 2002-06-25 Peter Braam <braam@clusterfs.com>
6503 * release version v0_4_1. Hopefully stable on single node use.