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 * Client known to build on unpatched kernels:
12 2.6.32-431.29.2.el6 (RHEL6.5)
13 2.6.32-504.30.3.el6 (RHEL6.6)
14 2.6.32-573.7.1.el6 (RHEL6.7)
15 3.10.0-229.14.1.el7 (RHEL7.1)
16 3.0.101-0.47.55 (SLES11 SP3)
18 * Recommended e2fsprogs version: 1.42.13.wc3 or newer
19 * Recommended ZFS / SPL version: 0.6.5.2
20 * Tested with ZFS / SPL version: 0.6.5.2
21 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
22 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
23 stack overflow. For more information, please refer to bugzilla 17630.
24 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
25 the re-exporting nodes to be mounted with "32bitapi" mount option
27 --------------------------------------------------------------------------------
29 03-10-2015 Intel Corporation
31 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
32 for currently supported client and server kernel versions.
33 * Server known to build on patched kernels:
34 2.6.32-431.29.2.el6 (RHEL6.5)
35 2.6.32-504.8.1.el6 (RHEL6.6)
36 3.0.101-0.46 (SLES11 SP3)
37 * Client known to build on unpatched kernels:
38 2.6.32-431.29.2.el6 (RHEL6.5)
39 2.6.32-504.8.1.el6 (RHEL6.6)
40 3.10.0-123.20.1.el7 (RHEL7)
41 3.0.101-0.46 (SLES11 SP3)
42 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
43 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
44 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
45 stack overflow. For more information, please refer to bugzilla 17630.
46 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
47 the re-exporting nodes to be mounted with "32bitapi" mount option
49 Severity : enhancement
51 Description: control OST-index in IDIF via ROCOMPAT flag.
52 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
53 last_rcvd file. For new formatted OST device, it will be auto set;
54 for the case of upgrading from old OST device, you can enable it
55 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
56 enabled, for new created OST-object, its IDIF-in-LMA will contain
57 the OST-index; for the existing OST-object, the OSD will convert
58 old format IDIF as new format IDIF with OST-index stored in the
59 LMA EA when accessing such OST-object or via OI scrub. Once such
60 flag is enabled, it cannot be reverted back, so the system cannot
61 be downgraded to the orignal incompatible version.
63 --------------------------------------------------------------------------------
65 07-30-2014 Intel Corporation
67 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
68 for currently supported client and server kernel versions.
69 * Server known to build on patched kernels:
70 2.6.32-431.20.3.el6 (RHEL6.5)
71 3.0.101-0.31 (SLES11 SP3)
72 * Client known to build on unpatched kernels:
73 2.6.32-431.20.3.el6 (RHEL6.5)
74 3.0.101-0.31 (SLES11 SP3)
77 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
78 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
79 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
80 stack overflow. For more information, please refer to bugzilla 17630.
81 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
82 the re-exporting nodes to be mounted with "32bitapi" mount option
84 --------------------------------------------------------------------------------
86 10-31-2013 Intel Corporation
88 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
89 * Server builds for kernels:
90 2.6.32-358.18.1.el6 (RHEL6.4)
91 * Client builds for unpatched kernels:
92 2.6.32-358.18.1.el6 (RHEL6.4)
93 3.0.92-0.8 (SLES11 SP3)
95 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
96 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
97 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
98 stack overflow. For more information, please refer to bugzilla 17630.
99 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
100 the re-exporting nodes to be mounted with "32bitapi" mount option
102 --------------------------------------------------------------------------------
104 05-30-2013 Intel Corporation
106 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
107 * Server builds for kernels:
108 2.6.32-358.6.2.el6 (RHEL6.4)
109 * Client builds for unpatched kernels:
110 2.6.32-358.6.2.el6 (RHEL6.4)
111 3.0.74-0.6.6 (SLES11 SP2)
113 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
114 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
115 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
116 stack overflow. For more information, please refer to bugzilla 17630.
117 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
118 the re-exporting nodes to be mounted with "32bitapi" mount option
120 --------------------------------------------------------------------------------
122 9-30-2012 Whamcloud, Inc.
124 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
125 * Server support for kernels:
126 2.6.32-279.5.1.el6 (RHEL6)
127 * Client support for unpatched kernels:
128 2.6.32-279.5.1.el6 (RHEL6)
130 2.6.18-238.19.1.el5 (RHEL5)
131 2.6.16.60-0.69.1 (SLES10)
132 * Recommended e2fsprogs version: 1.41.90.wc4
133 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
134 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
135 stack overflow. For more information, please refer to bugzilla 17630.
136 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
137 the re-exporting nodes to be mounted with "32bitapi" mount option
139 Severity : enhancement
141 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
143 Severity : enhancement
145 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
147 Severity : enhancement
149 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
151 Severity : enhancement
153 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
155 Severity : enhancement
157 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
159 --------------------------------------------------------------------------------
161 04-30-2012 Whamcloud, Inc.
163 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
164 * Server support for kernels:
165 2.6.32-220.4.2.el6 (RHEL6)
166 * Client support for unpatched kernels:
167 2.6.32-220.4.2.el6 (RHEL6)
169 2.6.18-238.19.1.el5 (RHEL5)
170 2.6.16.60-0.69.1 (SLES10)
171 * Recommended e2fsprogs version: 1.41.90.wc4
172 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
173 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
174 stack overflow. For more information, please refer to bugzilla 17630.
175 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
176 the re-exporting nodes to be mounted with "32bitapi" mount option
178 Severity : enhancement
180 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
182 Severity : enhancement
184 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
186 Severity : enhancement
188 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
190 Severity : enhancement
192 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
194 --------------------------------------------------------------------------------
196 09-30-2011 Whamcloud, Inc.
198 * Server support for kernels:
199 2.6.18-238.19.1.el5 (RHEL5)
200 2.6.32-131.6.1.el6 (RHEL6)
201 * Client support for unpatched kernels:
202 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
203 2.6.18-238.19.1.el5 (RHEL5)
204 2.6.32-131.6.1.el6 (RHEL6)
205 2.6.29.4-167.fc11 (FC11)
206 2.6.16.60-0.69.1 (SLES10)
208 * Recommended e2fsprogs version: 1.41.90.wc3
209 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
210 Lustre filesystem with 4K stack may cause a stack overflow. For more
211 information, please refer to bugzilla 17630.
212 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
213 mounted with "32bitapi" mount option
215 Severity : enhancement
216 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
218 Severity : enhancement
219 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
221 Severity : enhancement
222 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
225 Frequency : only with SLES10
227 Description: Use OFED "KMP" provided by Novell
228 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
229 Lustre is now built against this package. That means you need to
230 install the ofed-kmp package from Novell for the patchless client.
231 Note that the ofed-kmp that Novell ships may not exactly match the
232 kernel version but should still be compatible.
234 Severity : enhancement
236 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
237 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
239 --------------------------------------------------------------------------------
241 2010-07-15 Oracle, Inc.
243 * Server support for kernels:
244 2.6.18-164.11.1.el5 (RHEL 5)
245 2.6.18-164.11.1.0.1.el5 (OEL 5)
246 * Client support for unpatched kernels:
247 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
248 2.6.18-164.11.1.el5 (RHEL 5),
249 2.6.18-164.11.1.0.1.el5 (OEL 5)
250 2.6.16.60-0.42.8 (SLES 10),
252 2.6.29.4-167.fc11 (FC11)
253 * Recommended e2fsprogs version: 1.41.10-sun2
255 Severity : enhancement
257 Description: Distribute lfs_migrate script with lustre
258 Details : The lfs_migrate mentioned in the manual is now included in the
259 lustre tree in lustre/scripts and will distributed in lustre RPMs.
261 Severity : enhancement
263 Description: Failover nid list added to import proc file.
265 Severity : enhancement
267 Description: "som_preview" mount option required to use SOM
268 Details : The clients will need to use som_preview mount option
272 Description: Lustre does not do 1MB IOs to HW RAID
273 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
275 Severity : enhancement
277 Description: lfs quota output is not very convenient for awk/sed-parsing
278 Details : Some positions in lfs quota output table could be empty or
279 non-empty which made it hard to parse it with scripts, now
280 a dash is put instead of space where there is not supposed
285 Description: Update to OFED 1.5.1
286 Details : OFED 1.5.1 will be used in environments that don't use
287 vendor supplied OFED.
291 Description: Remove lnet. prefix from lctl params display
292 Details : Also renamed lnet.memused to lnet_memused to
297 Description: Use conf_param -d to remove permanant settings.
301 Description: Add list_param -R
302 Details : lctl list_param now supports recursive mode.
305 Bugzilla : 21829, 22312
306 Description: Fix llobdstat.
307 Details : 1. Now updating the write_bytes counter
308 2. Port script fixes from 1.8.
312 Description: decrease the usage of memory on clients.
313 Details : 1. On clients, recycle dentries and inodes unused.
314 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
319 Description: llobdstat fix and enhancement
320 Details : add a counter to set a limit to how many samples will be returned
321 fix a wildcard in the path to limit to obdfilter stats only
325 Description: lfs check only allowed for root
326 Details : Code cleanups and added a regresion test of non-root use of
331 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
332 Details : Fix hr->hr_index overload.
336 Description: Fix FIEMAP functionality.
340 Description: FC11 patchless client support added
344 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
345 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
349 Description: New acc-sm script and various test-framework enhancments.
353 Description: Previously landed patch was reverted.
357 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
361 Description: Add ldiskfs maxdirsize mount option.
365 Description: Fix mkfs.lustre support for 16TB LUNs.
369 Description: Print mmp_check_interval
370 Details : print mmp_check_interval and make it possible to abort
371 mount operation in case it takes too long.
375 Description: Add cascading_rw to lustre mpi tests.
379 Description: Store SOM data in EA.
380 Details : Also several SOM related cleanups.
384 Description: Add refcount to md_open_data
388 Description: Use refcount on lov_request_set to prevent memory corruption.
392 Description: Statahead: use parent dir's i_mutex for synchronization.
396 Description: loadgen improvements.
400 Description: Enable peer health detection in ksocklnd and ko2iblnd.
404 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
408 Description: Fix race on obd_devcie:md_stats
412 Description: Fix SLES 11 client builds.
413 Details : Resolved uninitalized pointer gcc warnings.
417 Description: Allow non-root access for "lfs check".
418 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
422 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
424 Severity : enhancement
426 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
427 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
431 Description: Sanity test_27p fails.
432 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
433 Sanity tests 27 were cleaned up and improved.
437 Description: Unify req format on client/servers
438 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
439 buffers and swabbers approach..
444 Description: drop join file stuff.
446 Severity : enhancement
448 Description: Add 2.6.27 and SLES11 (patchless client) support
452 Description: Fix in ptlrpc_expire_one_request() to print the signed time
455 Severity : enhancement
457 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
460 Severity : enhancement
462 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
464 Severity : enhancement
466 Description: correctly handle big reply message.
467 Details : send LNet event if reply is bigger than buffer and adjust this buffer
472 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
474 Severity : enhancement
476 Description: Add support for OEL5.
480 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
482 Severity : enhancement
484 Description: "weak-modules" support
485 Details : Implement "weak-modules" support which enables kernel modules
486 to be used with any kernel that implements the same kABI. In
487 order to achieve this modules are now installed in
488 /lib/modules/$(uname -r)/updates/kernel on all distributions.
492 Description: client could not reconnect to OST because of an active request.
493 Details : The client now retries to reconnect to the same server, if a
494 connect request failed with EBUSY or -EAGAIN. Test case 35b
495 conf-sanity.sh has been created.
499 Description: Don't allow make backward step on assignin osc next id.
500 Details : race between allocation next id and ll_sync thread can be cause
501 of set wrong osc next id and can be kill valid ost objects.
505 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
509 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
513 Description: Error handling in osc_statfs_interpret() has been improved.
514 Details : Check in osc_statfs_interpret() for EBADR.
518 Description: Do not send statfs() requests to OSTs disabled by administrator.
519 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
521 Severity : enhancement
523 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
525 Severity : enhancement
527 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
529 Severity : enhancement
531 Description: Enhancement for lfs(1) command to use numeric uid/gid.
535 Description: Improved cleanup procedure for conf-sanity test_46a.
539 Description: Do not start bulk transfer, if client tried to reconnect.
540 Details : Do not start bulk transfer, if client tried to reconnect,
541 but failed due to active rpcs. Check for exp_abort_active_req
542 has been implemented.
546 Description: Prevent inconsistences between linux and lustre mount structures.
547 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
548 Make the sleep interruptible.
550 Severity : enhancement
552 Description: provide server to client comms path
553 Details : server to client communications path via new LDLM_SET_INFO rpc
557 Description: truncate starts GFP_FS allocation under transaction causing deadlock
558 Details : ldiskfs_truncate calls grab_cache_page which may start page
559 allocation under an open transaction. This may lead to
560 calling prune_icache with consequent lustre reentrance.
564 Description: send correctly lsm on open replay
565 Details : MDS is trust to LSM size on replay open, but client can set wrong size
570 Description: enable client interface failover
571 Details : When a child reconnects from another NID, properly update export
572 nid hash position and ldlm reverse import.
574 Severity : enhancement
576 Description: Communicate OST degraded/readonly state via statfs to MDS
577 Details : Flags in the statfs returned from OSTs indicate whether the
578 OST is in a degraded RAID state, or if the filesystem has
579 turned read-only after a filesystem error is detected.
584 Description: don't panic if EPROTO was hit when reading symlink
585 Details : correctly handling request reference in error cases.
587 Severity : enhancement
589 Description: Add LustreNetLink, a kernel-userspace communcation path.
591 Severity : enhancement
593 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
598 Description: access to llog context before init.
599 Details : move handling CATALOGS file at osc layer and forbid access to llog
603 Frequency : rare, only when client is writing to OST on local node
605 Description: Avoid deadlock for local client writes
606 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
607 memory freeing context. This allows OST threads to set the
608 PF_MEMALLOC flag on task structures in order to allocate memory
609 from reserved pools and complete IO.
610 Use GFP_HIGHUSER for OST allocations for non-local client writes,
611 so that the OST threads generate memory pressure and allow
612 inactive pages to be reclaimed.
614 Severity : enhancement
616 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
621 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
622 Details : move ldlm namespace creation in setup phase to avoid grab
623 _lprocfs_lock with cli_sem held.
627 Description: Temporarily disable grant shrink.
628 Details : Disable the feature for debugging.
631 Frequency : only when formatting test filesystems
633 Description: Unable to run several mkfs.lustre on loop devices at the same time
634 Details : mkfs.lustre returns error 256 on the concurrent loop devices
635 formatting. The solution is to proper handle the error.
637 Severity : enhancement
639 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
641 Severity : enhancement
643 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
648 Description: resolve race between obd_disconnect and class_disconnect_exports
649 Details : if obd_disconnect will be called to already disconnected export he
650 forget release one reference and osc module can't unloaded.
652 Severity : enhancement
654 Description: Allow tuning service thread via /proc
655 Details : For each service a new
656 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
657 created that can be used to set min/max thread counts, and get the
658 current number of running threads.
663 Descriptoin: don't return error if have partially created objects for file.
664 Details : lov_update_create_set uses set->set_success as index for created
665 objects, so if some requests will be failed, they will have hole
666 at end of array and qos_shrink_lsm for allocate correct lsm.
668 Severity : enhancement
670 Description: Update OFED support to 1.4.1
675 Description: Reduce small size read RPC
676 Details : Set read-ahead limit for every file and only do read-ahead when
677 available read-ahead pages are bigger than 1M to avoid small size
680 Severity : enhancement
682 Description: move AT tunable parameters for more consistent usage
683 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
685 Severity : enhancement
687 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
688 Details : allow skip disconnected ost for send statfs request and hide error
694 Description: fix racy locking of mballoc block bitmaps causing BUG
695 Details : The locking of the mballoc buddy bitmap and the in-memory
696 block bitmap was using two different spin locks in some
697 cases. This made it possible to incorrectly access the
698 mballoc bitmap while another process was modifying it,
699 causing a sanity assertion to fail. While no on-disk corruption
700 was reported, there was some risk of this happening.
703 Frequency : rare, on llog test 6
705 Descriptoin: don't allow connect to already connected import
706 Details : allowing connect to already connected import is hide connecting problem.
711 Descriptoin: allow kill process which wait statahead result
712 Details : for some reasons 'ls' can stick in waiting result from statahead,
713 in this case need way for kill this process.
715 Severity : enhancement
717 Description: Add state history info file, enhance import info file
718 Details : Track import connection state changes in a new osc/mdc proc file;
719 add overview-type data to the osc/mdc import proc file.
721 Severity : enhancement
723 Description: MDS create should not wait for statfs RPC while holding DLM lock.
725 Severity : enhancement
727 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
732 Descriptoin: OST grant too much space to client even there are not enough space.
733 Details : Client will shrink its grant cache to OST if there are no write
734 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
735 this grant cache if there are already not enough avaible space
736 (left_space < total_clients * 32M).
739 Frequency : start MDS on uncleanly shutdowned MDS device
741 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
742 Details : stay in waiting mds<>ost recovery finished produce random bugs
743 due race between two ll_sync thread for one lov target. send
744 ACTIVATE event only if connect realy finished and import have
748 Frequency : rare, connect and disconnect target at same time
750 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
751 Details : don't call obd_disconnect under lov_lock. this long time
752 operation and can block ptlrpcd which answer to connect request.
757 Descriptoin: don't lose wakeup for imp_recovery_waitq
758 Details : recover_import_no_retry or invalidate_import and import_close can
759 both sleep on imp_recovery_waitq, but we was send only one wakeup
763 Frequency : always with long access acl
765 Descriptoin: mds can't pack reply with long acl.
766 Details : mds don't control size of acl but they limited by reint/getattr
769 Severity : enhancement
771 Description: Update to SLES10 kernel-2.6.16.60-0.33.
773 Severity : enhancement
775 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
778 Frequency : start MDS on uncleanly shutdowned MDS device
780 Descriptoin: aborting recovery hang on MDS
781 Details : don't throttle destroy RPCs for the MDT.
784 Frequency : on remount
786 Description: external journal device not working after the remount
787 Details : clear dev_rdonly flag for external journal devices in
793 Description: shutdown vs evict race
794 Details : client_disconnect_export vs connect request race.
795 if client will evicted at this time - we start invalidate
796 thread without referece to import and import can be freed
800 Frequency : rare, need acl's on inode.
802 Description: client can't handle ost additional correctly
803 Details : if ost was added after client connected to mds client can have
804 hit lnet_try_match_md ... to big messages to wide striped files.
805 in this case need teach client to handle config events about add
806 lov target and update client max ea size at that event.
808 Severity : enhancement
810 Description: Changelogs
811 Details : Changelogs are a lightweight mechanism to track filesystem
812 metadata and namespace changes. The changelog is recorded
813 permanently on the MDTs, and is periodically "consumed" / purged
814 when records are no longer needed.
816 Severity : enhancement
818 Description: compact fld format with extents
819 Details : Store range of seq rather than every seq in FLD. Seq
820 controller update FLD rather than clients. In Case of CMD, mdt0
821 has FLD, all other metadata server act as non persistent proxy
822 for FLD queries and cache fld entries in fld cache.
827 Description: don't skip ost target if they assigned to file
828 Details : Drop slow OSCs if we can, but not for requested start idx.
829 This means "if OSC is slow and it is not the requested
830 start OST, then it can be skipped, otherwise skip it only
831 if it is inactive/recovering/out-of-space.
835 Description: more cleanup in mds_lov
836 Details : not send LOV EA under replay, we can't know about they size at this
837 time. Don't allow client connect to mds before any ost connected,
838 for avoid problems with LOV EA size and returning EIO to client.
840 Severity : enhancement
842 Description: Interoperability at server side (Disk interoperability)
844 Severity : enhancement
846 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
848 Severity : enhancement
850 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
852 Severity : enhancement
854 Description: New client IO stack (CLIO).
856 Severity : enhancement
858 Description: Commit on sharing. Eliminate inter-client dependencies between
859 uncommitted transactions by doing transaction commits.
860 Thereby clients may recovery independently.
863 Frequency : Create a symlink file with a very long name
865 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
866 Details : If there is no extra space in the request for early cancels,
867 ldlm_req_handles_avail() returns 0 instead of a negative value.
869 Severity : enhancement
871 Description: Add /proc entry for import status
872 Details : The mdc, osc, and mgc import directories now have
873 an import directory that contains useful import data for debugging
876 Severity : enhancement
878 Description: Re-disable certain /proc logging
879 Details : Enable and disable client's offset_stats, extents_stats and
880 extents_stats_per_process stats logging on the fly.
883 Frequency : Only on FC kernels 2.6.22+
885 Description: oops in statahead
886 Details : Do not drop reference count for the dentry from VFS when lookup,
887 VFS will do that by itself.
889 Severity : enhancement
891 Description: Generic /proc file permissions
892 Details : Set /Proc file permissions in a more generic way to enable non-
893 root users operate on some /proc files.
897 Description: Hitting mdc_commit_close() ASSERTION
898 Details : Properly handle request reference release in
899 ll_release_openhandle().
903 Description: quota recovery deadlock during mds failover
904 Details : This patch includes att18982, att18236, att18237 in bz14840.
906 1. fix osts hang when mds does failover with quotaon
907 2. prevent watchdog storm when osts threads wait for the
912 Frequency : only patchless client
913 Description: add workaround for race between add/remove dentry from hash
915 Severity : enhancement
917 Description: Allow OST glimpses to return PW locks
921 Description: LBUG when llog conf file is full
922 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
926 Description: Prevent import from entering FULL state when server in recovery
930 Description: service mount cannot take device name with ":"
931 Details : Only when device name contains ":/" will mount treat it as
937 Description: replace ptlrpcd with the statahead thread to interpret the async
938 statahead RPC callback
942 Frequency : on recovery
943 Description: I/O failures after umount during fail back
944 Details : if client reconnected to restarted server we need join to recovery
945 instead of find server handler is changed and process self
946 eviction with cancel all locks.
948 Severity : enhancement
950 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
952 Severity : enhancement
954 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
956 Severity : enhancement
958 Description: Upcall on Lustre log has been dumped
959 Details : Allow for a user mode script to be called once a Lustre log has
960 been dumped. It passes the filename of the dumped log to the
961 script, the location of the script can be specified via
962 /proc/sys/lnet/debug_log_upcall.
967 Description: avoid idr_remove called for id which is not allocated.
968 Details : Move assigment s_dev for clustered nfs to end of initialization,
969 to avoid problem with error handling.
974 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
975 Details : When connection is reused this not moved from CONN_UNUSED_HASH
976 into CONN_USED_HASH and this prodice warning when put connection
977 again in unused hash.
980 Severity : enhancement
982 Description: File striping can now be set to use an arbitrary pool of OSTs.
984 Severity : enhancement
986 Description: Export bytes_read/bytes_write count on OSC/OST.
990 Description: Early reply size mismatch, MGC loses connection
991 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
992 the connect flags are properly negotiated.
997 Description: quotas are not honored with O_DIRECT
998 Details : all writes with the flag O_DIRECT will use grants which leads to
999 this problem. Now using OBD_BRW_SYNC to guard this.
1003 Description: add quota statistics
1004 Details : 1. sort out quota proc entries and proc code.
1005 2. add quota statistics
1007 Severity : enhancement
1009 Description: enable quota support for HEAD.
1013 Description: Properly propagate oinfo flags from lov to osc for statfs
1014 Details : restore missing copy oi_flags to lov requests.
1016 Severity : enhancement
1018 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1019 : l_getgroups(8), lst(8), routerstat(8)
1020 Details : included man pages for llobdstat(8), llstat(8),
1021 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1023 Severity : enhancement
1025 Description: configure's --enable-quota should check the
1026 : kernel .config for CONFIG_QUOTA
1027 Details : configure is terminated if --enable-quota is passed but
1028 : no quota support is in kernel
1032 Description: Remove portals compatibility
1033 Details : Remove portals compatibility, not interoperable with releases
1038 Description: Resolve device initialization race
1039 Details : Prevent proc handler from accessing devices added to the
1040 obd_devs array but yet be intialized.
1042 Severity : enhancement
1044 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1046 Severity : enhancement
1048 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1053 Description: Using wrong pointer in osc_brw_prep_request
1054 Details : Access to array[-1] can produce panic if kernel compiled with
1055 CONFIG_PAGE_ALLOC enabled
1059 Description: Client runs out of low memory
1060 Details : Consider only lowmem when counting initial number of llap pages
1064 Description: *optional* service tags registration
1065 Details : if the "service tags" package is installed on a Lustre node
1066 When the filesystem is mounted, a local-node service tag will
1067 be created. See http://inventory.sun.com/ for more information
1068 about the Service Tags asset management system.
1072 Description: Kernel BUG tries to release flock
1073 Details : Lustre does not destroy flock lock before last reference goes
1074 away. So always drop flock locks when client is evicted and
1075 perform unlock regardless of successfulness of speaking to MDS.
1079 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1083 Description: sanity test 65a fails if stripecount of -1 is set
1084 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1089 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1090 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1093 Severity : enhancement
1095 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1099 Description: do not process already freed flock
1100 Details : flock can possibly be freed by another thread before it reaches
1101 to ldlm_flock_completion_ast.
1105 Description: LBUG during stress test
1106 Details : Need properly lock accesses the flock deadlock detection list.
1110 Description: oops in page fault handler
1111 Details : kernel page fault handler can return two special 'pages' in error
1112 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1116 Description: timeout with invalidate import.
1117 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1118 should be handled by ptlrpcd. This produce long age waiting and
1119 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1121 Severity : enhancement
1123 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1127 Description: enable MGS and MDT services start separately
1128 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1129 a co-located MGS without starting the MGS, which is a complement
1130 to 'nosvc' mount option.
1134 Frequency : after recovery
1135 Description: precreate to many object's after del orphan.
1136 Details : del orphan st in oscc last_id == next_id and this triger growing
1137 count of precreated objects. Set flag LOW to skip increase count
1138 of precreated objects.
1142 Frequency : rare, on clear nid stats
1143 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1144 Details : when clean nid stats sometimes try destroy live entry,
1145 and this produce panic in free.
1149 Description: Stack overflow during MDS log replay
1150 ease stack pressure by using a thread dealing llog_process.
1154 Description: wait until IO finished before start new when do lock cancel.
1155 Details : VM protocol want old IO finished before start new, in this case
1156 need wait until PG_writeback is cleared until check dirty flag and
1157 call writepages in lock cancel callback.
1159 Severity : enhancement
1161 Description: using special macro for print time and cleanup in includes.
1165 Description: mds_mfd_close() ASSERTION(rc == 0)
1166 Details : In mds_mfd_close(), we need protect inode's writecount change
1167 within its orphan write semaphore to prevent possible races.
1171 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1175 Frequency : rare, on shutdown ost
1176 Description: don't hit live lock with umount ost.
1177 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1178 use shrink_dcache_sb instead.
1182 Description: don't panic with use echo client
1183 Details : echo client pass NULL as client nid pointer and this produce null
1184 pointer dereference.
1188 Description: fix build on ppc32
1189 Details : compile code with -m64 flag produce wrong object file for ppc32.
1193 Description: add message levels for liblustreapi
1197 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1198 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1199 will result in an -EAGAIN and in the exit call path we call
1200 alloc_rr() which will with increasing aggressiveness attempt to
1201 aquire precreated objects on the minimum number of required OSCs.
1205 Description: Use old size assignment to avoid deadlock
1206 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1207 the scheduling while holding a spinlock. We do not need locking
1208 for size in ll_update_inode() because size is only updated from
1209 the MDS for directories or files without objects, so there is no
1210 other place to do the update, and concurrent access to such inodes
1211 are protected by the inode lock.
1215 Description: resolve "_IOWR redefined" build error on SLES10
1219 Description: dump the memory debugging after all modules are unloaded to
1220 suppress false negative in conf_sanity test 39
1222 Severity : enhancement
1224 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1229 Description: fix SLES kernel versioning
1230 Details : the kernel version for our SLES 10 kernel did not include a "-"
1231 before the "smp" at the end. while this was not a problem in
1232 general, it did mean that software trying to use the kernel
1233 version to try to detect a vendor specific kernel would fail.
1234 this was most evident by the OFED build scripts.
1238 Description: Don't update lov_desc members until making sure they are valid
1239 Details : When updating lov_desc members via proc fs, need fix their
1240 validities before doing the real update.
1244 Description: don't put request into delay list while invalidate in flight.
1245 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1246 import in flight. this produce timeout for invalidate and sometimes
1247 can cause stale data.
1251 Frequency : on ppc only
1252 Description: not convert ost objects for directory because it's not exist.
1253 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1257 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1259 Severity : enhancement
1261 Description: Update to support 2.6.22.14 vanilla kernel.
1265 Frequency : rare, on recovery
1266 Description: read procfs can produce deadlock in some situation
1267 Details : Holding lprocfs lock which send rpc can produce block for destroy
1268 obd objects and this also block reconnect with -EALREADY. This
1269 isn't fix all lprocfs bugs - but make it rare.
1271 Severity : enhancement
1273 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1276 Frequency : frequent on X2 node
1278 Description: mdc_set_open_replay_data LBUG
1279 Details : Set replay data for requests that are eligible for replay.
1283 Description: lustre_mgs: operation 101 on unconnected MGS
1284 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1285 MGC, and late on MGC cannot successfully connect to MGS and a lot
1286 of the error messages complaining that MGS is not connected.
1290 Frequency : on network error
1291 Description: panic with double free request if network error
1292 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1293 but it's true only for synchronus enqueue, for async enqueue
1294 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1297 Severity : enhancement
1299 Description: client-side metadata stat-ahead during readdir(directory readahead)
1300 Details : perform client-side metadata stat-ahead when the client detects
1301 readdir and sequential stat of dir entries therein
1304 Frequency : on start mds
1306 Description: Implement get_info(last_id) in obdfilter.
1309 Frequency : occasional
1311 Description: Correctly check stale fid, not start epoch if ost not support SOM
1312 Details : open with flag O_CREATE need set old fid in op_fid3 because
1313 op_fid2 was overwritten with new generated fid, but mds can answer
1314 with one of these two fids and both is not stale. Setattr
1315 incorrectly started an epoch and assume will be called
1316 done_writing, but without SOM done_writing ever being called.
1319 Frequency : rare, depends on device drivers and load
1321 Description: MDS or OSS nodes crash due to stack overflow
1322 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1323 In some cases, in conjunction with device drivers that use a lot
1324 of stack the MDS (or possibly OSS) service threads could overflow
1325 the stack. One change which was identified to consume additional
1326 stack has been reworked to avoid the extra stack usage.
1329 Frequency : occasional
1331 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1332 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1333 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1334 oscc_has_objects and see OSC has no objests and call
1335 oscc_internal_create to resend create request.
1337 Severity : enhancement
1339 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1341 Severity : enhancement
1343 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1346 Frequency : very rare
1348 Description: Fix replay if there is an un-replied request and open
1349 Details : In some cases, older replay request will revert the
1350 mcd->mcd_last_xid on MDS which is used to record the client's
1351 latest sent request.
1353 Severity : enhancement
1355 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1357 Severity : enhancement
1359 Description: Add rhel5 support to HEAD.
1361 Serverity : enhancement
1363 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1368 Description: Don't allow skipping OSTs if index has been specified.
1369 Details : Don't allow skipping OSTs if index has been specified, make
1370 locking in internal create lots better.
1374 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1375 Details : access to bitfield in structure is always rounded to long
1376 and this produce problem with not atomic change any bit.
1380 Description: Lustre make rpms failed.
1381 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1382 builds Lustre rpms from tarball.
1385 Frequency : rare on shutdown ost
1387 Description: If llog cancel was not send before clean_exports phase, this can
1388 produce deadlock in llog code.
1389 Details : If llog thread has last reference to obd and call class_import_put
1390 this produce deadlock because llog_cleanup_commit_master wait when
1391 last llog_commit_thread exited, but this never success because was
1392 called from llog_commit_thread.
1396 Description: allow userland application know is lost one of stripes.
1397 Details : fill lvb_blocks with error code on ost and return it to
1398 application if error flag found.
1402 Description: NULL lov_tgts causing MDS oops
1403 Details : more safe checks for NULL lov_tgts for avoid oops.
1405 Severity : enhancement
1407 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1411 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1412 Details : If an OST is down the MDS will hang indefinitely in
1413 obd_statfs_rqset() waiting for the statfs data. While for
1414 MDS QOS usage of statfs, it should not stuck in waiting.
1416 Severity : enhancement
1418 Description: remote_acl support
1419 Details : Support ACL-based permission check for remote user.
1420 Support setfacl/getfacl for remote user with the utils
1421 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1422 the system "{s,g}etfacl" utils.
1424 Severity : enhancement
1426 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1428 Severity : enhancement
1430 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1434 Description: make mgs_setparam() handle fsname containing dash
1435 Details : fsname containing a dash does not work with lctl conf_param
1437 Severity : enhancement
1439 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1441 Severity : enhancement
1443 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1447 Frequency : rare, at shutdown
1448 Description: access already free / zero obd_namespace.
1449 Details : if client_disconnect_export was called without force flag set,
1450 and exist connect request in flight, this can produce access to
1451 NULL pointer (or already free pointer) when connect_interpret
1452 store ocd flags in obd_namespace.
1456 Frequency : only at startup
1457 Description: not alloc memory with spinlock held.
1458 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1459 if any spinlock held.
1461 Severity : enhancement
1463 Description: make lustre randomly fail allocating memory
1464 Details : Make lustre randomly failed allocating memory for testing purpose.
1466 Severity : enhancement
1468 Description: lost problems with lov objid file
1469 Details : Fixes some scability and access to not inited memory problems
1470 in work with lov objdid file.
1475 Description: lfs find does not continue on file error
1476 Details : Continue other files processing when a file/dir is absent.
1480 Description: Inconsistent usage of lustre_pack_reply()
1481 Details : Standardize the usage of lustre_pack_reply() such that it
1482 always generate a CERROR on failure.
1485 Frequency : occasional
1487 Description: MDS hang or stay in waiting lock
1488 Details : If client receive lock with CBPENDING flag ldlm need send lock
1489 cancel as separate rpc, to avoid situation when cancel request
1490 can't processed due all i/o threads stay in wait lock.
1494 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1498 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1502 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1506 Description: Update to RHEL4 latest kernel.
1508 Severity : enhancement
1510 Description: Build SLES10 patchless client fails
1511 Details : The configure was broken by run ./configure with
1512 --with-linux-obj=.... argument for patchless client. When the
1513 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1514 can't search header adequately. Use absolute path such as
1515 -I($LINUX)/include instead.
1519 Description: interrupt oig_wait produce painc on resend.
1520 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1521 private set, and this produce situation when rq_ptlrpcd_data not
1522 copyed to new allocated request and triggered LBUG on assert
1523 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1524 ptlrpcd set if request is changed and can be safety changed to use
1529 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1530 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1531 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1532 these macros and replace them with their difinitions instead.
1536 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1537 Details : Update which_patch & target file for SLES10 latest kernel.
1540 Description: Remove client patches from SLES 10 kernel.
1541 Details : This causes SLES 10 clients to behave as patchless clients
1542 even on a Lustre-patched (server) kernel.
1544 Severity : enhancement
1546 Description: self-adjustable client's lru lists
1547 Details : use adaptive algorithm for managing client cached locks lru
1548 lists according to current server load, other client's work
1549 pattern, memory activities, etc. Both, server and client
1550 side namespaces provide number of proc tunables for controlling
1553 Severity : enhancement
1555 Description: light-weight GSS support
1556 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1557 and reduce performance overhead.
1559 Severity : enhancement
1561 Description: Linux keyring support
1562 Details : Support using service of Linux keyring for Lustre GSS internal
1563 context refresh/cache mechanism.
1567 Description: Fix errors in lfs documentation
1568 Details : Fixes man pages
1572 Description: don't use GFP_* in generic Lustre code.
1573 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1577 Description: obdclass is limited by single OBD_ALLOC(idarray)
1578 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1582 Description: warnings with build patchless client with vanila 2.6.19 and up
1583 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1584 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1588 Frequency : Only for SLES
1589 Description: sanity_quota fail test_1
1590 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1591 file, which makes TSTID[2] inunique.
1595 Frequency : Only for SLES9
1596 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1597 Details : Update target/ChangeLog/which_patch .
1601 Frequency : Only for test_10 in sanity-quota.sh
1602 Description: a bug in quota test code
1603 Details : checking a test flag in a wrong place
1607 Frequency : Only for test_13 in sanity-quota.sh
1608 Description: a bug in quota test script
1609 Details : shouldn't check overfull allocated quota
1613 Description: jbd statistics
1614 Details : Port older jbd statistics patch for sles10
1616 Severity : enhancement
1618 Description: Add printing inode info into message about error in writepage.
1623 Description: reply_lock_interpret crash due to race with it and lock cancel.
1624 Details : Do not replay locks that are being cancelled. Do not reference
1625 locks by their address during replay, just by their handle.
1629 Frequency : When flocks are used.
1630 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1631 Details : Flock locks might destroy just granted lock if it could be merged
1632 with another existing flock, this is done in completion handler,
1633 so teach ldlm_cli_enquque_fini that this is a valid case for
1639 Description: Oops in read and write path when failing to allocate lock.
1640 Details : Check if lock allocation failed and return error back.
1644 Description: lstripe command fails for valid OST index
1645 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1646 instead of lov->desc.ld_active_tgt_count.
1650 Description: sanity.sh failed test 103
1651 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1656 Description: df doesn't work properly if diskfs blocksize != 4K
1657 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1661 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1662 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1664 Severity : enhancement
1666 Description: Adaptive timeouts
1667 Details : RPC timeouts adapt to changing server load and network
1668 conditions to reduce resend attempts and improve recovery time.
1672 Description: llapi_file_create() does not allow some changes
1673 Details : add llapi_file_open() that allows specifying the mode and
1674 open flags, and also returns an open file handle.
1678 Description: merge and cleanup kernel patches.
1679 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1683 Description: Add journal checksum support.(Kernel part)
1684 Details : The journal checksum feature adds two new flags i.e
1685 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1686 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1687 indicates that the commit block contains the checksum for
1688 the blocks described by the descriptor blocks. Now commit
1689 record can be sent to disk without waiting for descriptor
1690 blocks to be written to disk. This behavior is controlled
1691 using JBD2_FEATURE_ASYNC_COMMIT flag.
1695 Description: OSS needs mutliple precreate threads
1696 Details : Add ability to start more than one create thread per OSS.
1700 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1701 Details : execute lfs setstripe on client
1705 Description: mds_obd_create error creating tmp object
1706 Details : When the user sets quota on root, llog will be affected and can't
1707 create files and write files.
1710 Frequency : Always on ia64 patchless client, and possibly others.
1712 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1713 Details : This allows the patchless client to be loaded on architectures
1714 without this export.
1718 Description: RedHat Update kernel for RHEL5
1719 Details : Add the kernel config file for RHEL5.
1723 Description: RedHat Update kernel for RHEL5
1724 Details : Modify the kernel config file more closer RHEL5.
1728 Description: Build failure against Centos5 (RHEL5)
1729 Details : Define PAGE_SIZE when it isn't present.
1733 Description: test_7 on sanity-quota.sh can't be used on separate machines
1734 Details : add facet to handle it
1738 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1739 Details : with NFS, the anon dentry's parent was set to itself in
1740 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1741 corresponding dentry other than use rec->ur_name.
1743 Severity : enhancement
1745 Description: lfs setstripe enhancement
1746 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1750 Description: enable data checksumming by default
1751 Details : enable checksum by default, allow --disable-checksum
1752 configure option and "-o nochecksum" mount option. Checksums
1753 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1757 Description: lustre support for RHEL5
1758 Details : Add support for RHEL5.
1762 Description: Client eviction due to failover config
1763 Details : after a connection loss, the lustre client should attempt to
1764 reconnect to the last active server first before trying the
1765 other potential connections.
1769 Description: when mds and osts use different quota unit(32bit and 64bit),
1770 quota will be released repeatly.
1771 Details : void sending multiple quota reqs to mds, which will keep the
1772 status between the reqs.
1776 Description: osts not allocated evenly to files
1777 Details : change the condition to increase offset_idx
1780 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1782 Description: Data corruption for OSTs that are formatted larger than 2TB
1784 Details : When generating the bio request for lustre file writes the
1785 sector number would overflow a temporary variable before being
1786 used for the IO. The data reads correctly from Lustre (which
1787 will overflow in a similar manner) but other file data or
1788 filesystem metadata may be corrupted in some cases.
1792 Description: Tune the kernel for good SCSI performance.
1793 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1794 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1799 Description: rewrite ext2-derived code in obdclass/uuid.c
1800 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1801 from scratch preserving functionality.
1805 Description: "lfs find -obd UUID" prints directories
1806 Details : "lfs find -obd UUID" will return all directory names instead
1807 of just file names. It is incorrect because the directories
1808 do not reside on the OSTs.
1812 Description: use i_size_read and i_size_write in 2.6 port
1813 Details : replace inode->i_size access with i_size_read/write()
1816 Frequency : when removing large files
1818 Description: scheduling issue during removal of large Lustre files
1819 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1820 It causes scheduling issues when removing large files (17TB in the
1824 Frequency : only with liblustre clients on XT3
1826 Description: evictions taking too long
1827 Details : allow llrd to evict clients directly on OSTs
1830 Frequency : only on ppc
1832 Description: /proc/fs/lustre/devices broken on ppc
1833 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1834 We should make sure the type of 'index' is loff_t and then cast
1835 explicitly as needed below. Do not assign an explicitly cast
1839 Frequency : only for rhel5
1841 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1842 Details : Modify the target file & which_kernel.
1844 Severity : enhancement
1846 Description: omit set fsid for export NFS
1847 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1848 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1852 Frequency : Always, for kernels after 2.6.16
1853 Description: Fix warning idr_remove called for id=.. which is not allocated.
1854 Details : Last kernels save old s_dev before kill super and not allow
1855 to restore from callback - restore it before call kill_anon_super.
1859 Description: buffer overruns could theoretically occur
1860 Details : llapi_semantic_traverse() modifies the "path" argument by
1861 appending values to the end of the origin string, and a
1862 overrun may occur. Adding buffer overrun check in liblustreapi.
1866 Description: Fix error on 'ls .' at the top of the Lustre mount.
1867 Details : Don't revalidate dentry if it is a root dentry.
1871 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1872 Details : Modify vm-tunables-rhel4.patch.
1876 Description: Kernel config for 2.6.18-vanilla.
1877 Details : Modify targets/2.6-vanilla.target.in.
1878 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1879 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1880 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1881 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1885 Description: improve handling recoverable errors
1886 Details : if request processig with error which can be recoverable on server
1887 request should be resend, otherwise page released from cache and
1892 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1893 Details : Modify target file & which_patch.
1894 A flaw was found in the IA32 system call emulation provided
1895 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1896 value could be stored in the %RAX register, which could trigger an
1897 out-of-bounds system call table access. An untrusted local user
1898 could exploit this flaw to run code in the kernel
1899 (ie a root privilege escalation). (CVE-2007-4573).
1903 Description: O_DIRECT bypasses client statistics.
1904 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1905 read_ahead_stats were updated. copied the stats section from
1906 osc_send_oap_rpc() into async_internal().
1910 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1914 Description: change order of libsysio includes
1915 Details : '#include sysio.h' should always come before '#include xtio.h'
1919 Description: handle "serious error: objid * already exists" more gracefully
1920 Details : If LAST_ID value on disk is smaller than the objects existing in
1921 the O/0/d* directories, it indicates disk corruption and causes an
1922 LBUG(). If the object is 0-length, then we should use the existing
1923 object. This will help to avoid a full fsck in most cases.
1925 Severity : enhancement
1927 Description: adapt the lustre_config script to support the upgrade case
1928 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1929 server targets to 1.6 in parallel.
1933 Description: To avoid grant space > avaible space when the disk is almost
1934 full. Without this patch you might see the error "grant XXXX >
1935 available" or some LBUG about grant, when the disk is almost
1937 Details : In filter_check_grant, for non_grant cache write, we should
1938 check the left space by if (*left > ungranted + bytes), instead
1939 of (*left > ungranted), because only we are sure the left space
1940 is enough for another "bytes", then the ungrant space should be
1941 increase. In client, we should update cl_avail_grant only there
1942 is OBD_MD_FLGRANT in the reply.
1946 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1947 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1948 described in CVE-2007-4573. This problem could allow untrusted
1949 local users to gain root access.
1952 Frequency : when using O_DIRECT and quotas
1954 Description: Incorrect file ownership on O_DIRECT output files
1955 Details : block usage reported by 'lfs quota' does not take into account
1956 files that have been written with O_DIRECT.
1961 Description: touch file failed when fs is not full
1962 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1963 otherwise we can get ENOSP while fs is not full.
1967 Description: parallel lock callbacks
1968 Details : Instead of sending blocking and completion callbacks as separated
1969 requests, adding them to a set and sending in parallel.
1972 Frequency : only for Cray XT3
1973 Bugzilla : 12829/13455
1974 Description: Changing primary group doesn't change the group lustre assigns to
1976 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1977 overridden with the primary group provided by the group upcall,
1978 whereas the supplied fsgid can be trusted if it is in the list of
1979 supplementary groups returned by the group upcall.
1981 Severity : enhancement
1983 Description: Allow masking D_WARNING, D_ERROR messages from console
1984 Details : Console messages can now be disabled via lnet.printk.
1988 Description: User code with malformed file open parameter crashes client node
1989 Details : Before packing join_file req, all the related reference should be
1990 checked carefully in case some malformed flags cause fake
1991 join_file req on client.
1995 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1996 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1997 mode is equal to the granted mode. However, it should also check
1998 that LDLM_FL_CP_REQD is not set.
2002 Description: Heavy nfs access might result in deadlocks
2003 Details : After ELC code landed, it is now improper to enqueue any mds
2004 locks under och_sem, because enqueue might want to decide to
2005 cancel open locks for same inode we are holding och_sem for.
2009 Description: Client eviction while running blogbench
2010 Details : A lot of unlink operations with concurrent I/O can lead to a
2011 deadlock causing evictions. To address the problem, the number of
2012 oustanding OST_DESTROY requests is now throttled to
2013 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2014 callbacks are processed in priority.
2018 Description: enable ACLs on MDS by default
2019 Details : ACLs must be enabled on MDS by default.
2022 Frequency : PPC/PPC64 only
2024 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2025 Details : fix duplicated definitions between asm-ppc64/types.h and
2026 lustre_types.h on PPC.
2029 Frequency : PPC/PPC64 only
2031 Description: asm-ppc/segment.h does not exist
2032 Details : fix compile issue on PPC.
2036 Description: better handle error messages in extents code
2039 Frequency : RHEL4 only
2041 Description: mkfs is very slow on IA64/RHEL4
2042 Details : A performance regression has been discovered in the MPT Fusion
2043 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2044 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2045 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2047 Severity : enhancement
2049 Description: SNMP support enhancement
2050 Details : Adding total number of sampled request for an MDS node in snmp
2053 Severity : enhancement
2055 Description: Optimize ldlm waiting list processing for PR extent locks
2056 Details : When processing waiting list for read extent lock and meeting read
2057 lock that is same or wider to it that is not contended, skip
2058 processing rest of the list and immediatelly return current
2059 status of conflictness, since we are guaranteed there are no
2060 conflicting locks in the rest of the list.
2064 Description: Time out and refuse to reconnect
2065 Details : When the failover node is the primary node, it is possible
2066 to have two identical connections in imp_conn_list. We must
2067 compare not conn's pointers but NIDs, otherwise we can defeat
2068 connection throttling.
2072 Description: port llog fixes from b1_6 into HEAD
2073 Details : Port llog reference couting and some llog cleanups from b1_6
2074 (bug 10800) into HEAD, for protect from panic and access to already
2075 free llog structures.
2079 Description: Detect stride IO mode in read-ahead
2080 Details : When a client does stride read, read-ahead should detect that and
2081 read-ahead pages according to the detected stride pattern.
2085 Description: data checksumming impacts single node performance
2086 Details : add support for several checksum algorithm. Currently, only CRC32
2087 and Adler-32 are supported. The checksum type can be changed on
2088 the fly via /proc/fs/lustre/osc/*/checksum_type.
2092 Description: use adler32 for page checksums
2093 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2098 Description: build for x2 fails
2099 Details : fix compile issue on Cray systems.
2103 Description: Properly match for duplicate locks
2104 Details : Due to different lock order from skiplists code, we need to
2105 traverse entire list for now
2108 Frequency : only on PPC/SLES10
2110 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2111 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2112 be defined. Add a hack in mkfs_lustre.c to work around this compile
2117 Description: LASSERT on MDS when client holding flock lock dies
2118 Details : ldlm pool logic depends on number of granted locks equal to
2119 number of released locks which is not true for flock locks, so
2120 just exclude such locks from consideration.
2124 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2125 Details : Use fsfilt_sync() for both the whole filesystem sync and
2126 individual file sync to eliminate dangerous inode locking
2127 with I_LOCK that can lead to a deadlock.
2131 Description: performance in 1.6.3
2132 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2133 sure that 1MB requests are not fragmented by the block layer.
2135 Severity : enhancement
2137 Description: organize the server-side client stats on per-nid basis
2138 Details : Change the structure of stats under obdfilter and mds to
2146 The "uuid"s file would list the uuids of _active_ exports.
2147 And the clear entry is to clear all stats and stale nids.
2149 Severity : enhancement
2151 Description: eliminate client locks in face of contention
2152 Details : file contention detection and lockless i/o implementation
2153 for contended files.
2157 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2160 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2162 Description: MDS may lose file striping (and hence file data) in some cases
2163 Details : If there are additional extended attributes stored on the MDS,
2164 in particular ACLs, SELinux, or user attributes (if user_xattr
2165 is specified for the client mount options) then there is a risk
2166 of attribute loss. Additionally, the Lustre file striping
2167 needs to be larger than default (e.g. striped over all OSTs),
2168 and an additional attribute must be stored initially in the
2169 inode and then increase in size enough to be moved to the
2170 external attribute block (e.g. ACL growing in size) for file
2175 Description: skiplist implementation simplification
2176 Details : skiplists are used to group compatible locks on granted list
2177 that was implemented as tracking first and last lock of each
2178 lock group the patch changes that to using doubly linked lists
2182 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2183 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2184 because LASSERT on both the data supplied by a client, and the
2185 data on disk is dangerous and incorrect.
2187 Severity : enhancement
2189 Description: Slow truncate/writes to huge files at high offsets.
2190 Details : Directly associate cached pages to lock that protect those pages,
2191 this allows us to quickly find what pages to write and remove
2192 once lock callback is received.
2196 Description: more ldlm soft lockups
2197 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2198 starve other threads from the resource lock for a long time in
2199 case of long waiting queue, so change the debug level from
2200 D_OTHER to the less frequently used D_INFO.
2202 Severity : enhancement
2204 Description: add -gid, -group, -uid, -user options to lfs find
2208 Description: Hung threads in invalidate_inode_pages2_range
2209 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2210 once one is completed. As a result, some RPCs are stuck in the
2211 queue and are never sent.
2215 Description: filter threads hungs on waiting journal commit
2216 Details : Cleanup filter group llog code, then only filter group llog will
2217 be only created in the MDS/OST syncing process.
2221 Description: Procfs and llog threads access destoryed import sometimes.
2222 Details : Sync the import destoryed process with procfs and llog threads by
2223 the import refcount and semaphore.
2225 Severity : enhancement
2227 Description: openlock cache of b1_6 port to HEAD
2232 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2233 Details : If insertion of an extent fails, then discard the inode
2234 preallocation and free data blocks else it can lead to duplicate
2239 Description: don't always update ctime in ext3_xattr_set_handle()
2240 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2241 In some cases the ctime should not be updated, for example for
2242 2.0->1.8 compatibility it is necessary to delete an xattr and it
2243 should not update the ctime.
2247 Bugzilla : 15713/16362
2248 Description: Assertion in iopen_connect_dentry in 1.6.3
2249 Details : looking up an inode via iopen with the wrong generation number can
2250 populate the dcache with a disconneced dentry while the inode
2251 number is in the process of being reallocated. This causes an
2252 assertion failure in iopen since the inode's dentry list contains
2253 both a connected and disconnected dentry.
2257 Description: assertion failure in ldlm_handle2lock()
2258 Details : fix a race between class_handle_unhash() and class_handle2object()
2259 introduced in lustre 1.6.5 by bug 13622.
2264 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2265 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2270 Description: oops in sd_iostats_seq_show()
2271 Details : unloading/reloading the scsi low level driver triggers a kernel
2272 bug when trying to access the sd iostat file.
2277 Description: Kernel panics during QLogic driver reload
2278 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2279 patch, causing memory corruption.
2284 Description: journal_dev option does not work in b1_6
2285 Details : pass mount option during pre-mount.
2287 Severity : enhancement
2289 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2290 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2291 extent information of a file. It can be used to map logical blocks
2292 in a file to physical blocks in the block device.
2296 Description: LDLM soft lockups - improvement
2297 Details : It is be possible to send the lock handle along with each read
2298 or write request because the client is already doing a lock match
2299 itself so there isn't any reason the OST should have to re-do that
2303 Frequency : only on Cray X2
2305 Description: X2 build failures
2306 Details : fix build failures on Cray X2.
2310 Description: xid & resent requests
2311 Details : Initialize RPC XID from clock at startup (randomly if clock is
2314 Severity : enhancement
2316 Description: Add lustre_start utility to start or stop multiple Lustre servers
2321 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2322 Details : In case of memory pressure, list_del() can be called twice on
2323 req->rq_history_list, causing a kernel oops.
2327 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2328 Details : fix stack overflow in the distributed lock manager by defering
2329 export eviction after a failed AST to the elt thread instead of
2330 handling it in the dlm interpret routine.
2334 Description: Convert some comments to new format.
2335 Details : Update documenting comments to match doxygen conventions.
2339 Description: Grammar fixes.
2340 Details : A couple of trivial spelling fixes.
2344 Description: OSD_COUNTERS-mandatory
2345 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2346 to simplify the code.
2350 Description: simplify cmm_device freeing logic.
2351 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2355 Description: Add lockdep support to dt_object_operations locking interface.
2356 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2357 parameter indicating lock ordering. Update mdd code to use new
2362 Description: Introduce failloc constants for lockless IO tests.
2363 Details : Add two new failloc constants to test lockless IO. Only one of
2364 them in implemented---another is checked in yet to be landed
2369 Description: Add lockdep support for inode mutex.
2370 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2371 in the situations where Lustre has to lock more than one inode
2376 Description: Add optional invariants checking support.
2377 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2378 configure switch. This macro is to be used for consistency and
2379 sanity checks that are too expensive to be left in `production'
2384 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2385 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2390 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2391 Details : Use spin_lock_nested() in (the only) situation where more than
2392 one ldlm_lock is locked simultaneously. Also, fix possible
2393 dead-lock in ldlm_lock_change_resource() by enforcing particular
2398 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2399 Details : Instead of specifying each ldlm_lock call-back through separate
2400 parameter, wrap them into struct ldlm_callback_suite.
2404 Description: Kill join_lru obd method and its callers.
2405 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2406 in memory---a requirement implemented through ->o_join_lru() obd
2407 method in HEAD. Remove this method and its users.
2411 Description: Add asynchronous ldlm ENQUEUE completion handler.
2412 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2413 case that stock ldlm_completion_ast() cannot handle as it waits
2414 until lock is granted. Introduce new ldlm_completion_ast_async()
2415 for this. Also comment ldlm_completion_ast().
2419 Description: ldlm_error <-> errno conversion.
2420 Details : Add functions to map (rather arbitrary) between LDLM error codes
2421 and standard errno values. CLIO needs this to prevent LDLM specific
2422 constants from escaping ldlm and osc.
2426 Description: Kill unused ldlm_handle2lock_ns() function.
2427 Details : Kill unused ldlm_handle2lock_ns() function.
2431 Description: Add lu_ref support to ldlm_lock
2432 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2433 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2434 and ldlm_resource. LDLM interface has to be changed, because code that
2435 releases a reference on a lock, has to "know" what reference this is.
2436 In the most frequent case
2438 lock = ldlm_handle2lock(handle);
2440 LDLM_LOCK_PUT(lock);
2442 no changes are required. When any other reference (received _not_ from
2443 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2444 instead of LDLM_LOCK_PUT().
2446 Arguably, changes are pervasive, and interface requires some discipline
2447 for proper use. On the other hand, it was very instrumental in finding
2448 a few leaked lock references.
2452 Description: Add ldlm_lock_addref_try().
2453 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2454 attempts to addref a lock that might be being canceled
2459 Description: Add ldlm_weigh_callback().
2460 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2461 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2462 hard-coded `number of pages' logic.
2466 Description: Add lockdep annotations to llog code.
2467 Details : Use appropriately tagged _nested() locking calls in the places
2468 where llog takes more than one ->lgh_lock lock.
2472 Description: Add loi_kms_set().
2473 Details : Wrap kms updates into a helper function.
2477 Description: Constify instances of struct lsm_operations.
2478 Details : Constify instances of struct lsm_operations.
2482 Description: lu_conf support.
2483 Details : On a server, a file system object is uniquely identified
2484 by a fid, which is sufficient to locate and load all object
2485 state (inode). On a client, on the other hand, more data are
2486 necessary instantiate an object. Change lu_object_find() and
2487 friends to take additional `lu_conf' argument describing object.
2488 Typically this includes layout information.
2492 Description: lu_context fixes.
2493 Details : Introduce new lu_context functions that are needed on the client
2494 side, where some system threads (ptlrpcd) are shared by multiple
2495 modules, and so cannot be stopped during module shutdown.
2499 Description: Add start and stop methods to lu_device_type_operations.
2500 Details : Introduce two new methods in lu_device_type_operations, that are
2501 invoked when first instance of a given type is created and last one
2502 is destroyed respectively. This is need by CLIO.
2506 Description: Add lu_ref support to struct lu_device.
2507 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2508 track leaked references.
2512 Description: Introduce lu_kmem_descr.
2513 Details : lu_kmem_descr and its companion interface allow to create
2514 and destroy a number of kmem caches at once.
2518 Description: Fix lu_object finalization race.
2519 Details : Fix a race between lu_object_find() finding an object and its
2520 concurrent finalization. This race is (most likely) not possible
2521 on the server, but might happen on the client.
2525 Description: Introduce lu_ref interface.
2526 Details : lu_ref is a debugging module allowing to track references to
2527 a given object. It is quite cpu expensive, and has to be
2528 explicitly enabled with --enable-lu_ref. See usage description
2533 Description: Factor lu_site procfs stats into a separate function.
2534 Details : Separate lu_site stats printing code into a separate function
2535 that can be reused on a client.
2539 Description: Constify instances of struct {lu,dt,md}_device_operations.
2540 Details : Constify instances of struct {lu,dt,md}_device_operations.
2544 Description: Introduce struct md_site and move meta-data specific parts of
2545 struct lu_site here.
2546 Details : Move md-specific fields out of struct lu_site into special struct
2547 md_site, so that lu_site can be used on a client.
2551 Description: Kill mdd_lov_destroy().
2552 Details : Remove unused mdd code.
2556 Description: Add st_block checking to multistat.c.
2557 Details : Add st_block checking to multistat.c.
2561 Description: Add lu_ref support to struct obd_device.
2562 Details : Add lu_ref tracking to obd_device.
2566 Description: Kill obd_set_fail_loc().
2567 Details : Remove unused code.
2571 Description: Add special type for ptlrpc_request interpret functions.
2572 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2573 there. Actual usage will be in CLIO.
2577 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2578 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2579 doesn't work with lockdep.
2583 Description: Add rwv.c test program.
2584 Details : New testing program exercising readv(2) and writev(2) (Qian).
2588 Description: Add sendfile.c test program.
2589 Details : New testing program exercising sendfile(2) (Jay).
2593 Description: Ratelimit a message that can be very frequent.
2594 Details : Ratelimit a memory allocation failure message that can
2599 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2600 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2604 Description: Liblustre build fixes.
2605 Details : Liblustre build fixes.
2609 Description: libcfs: add cfs_{need,cond}_resched() interface.
2610 Details : libcfs: add cfs_{need,cond}_resched() definition and
2611 implementations for Linux, NT, and liblustre.
2613 Severity : enhancement
2615 Description: More exported tunables for mballoc
2616 Details : Add support for tunable preallocation window and new tunables for
2617 large/small requests
2621 Description: Detect corruption of block bitmap and checking for preallocations
2622 Details : Checks validity of on-disk block bitmap. Also it does better
2623 checking of number of applied preallocations. When corruption is
2624 found, it turns filesystem readonly to prevent further corruptions.
2628 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2629 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2630 window to 0. If it does hit cache miss, and read-pattern is still
2631 stride-io mode, does not reset the stride window, but also does not
2632 increase the stride window length in this case.
2636 Frequency : only for big-endian servers
2637 Description: Check if system is big-endian while mounting fs with extents feature
2638 Details : Mounting a filesystem with extents feature will fail on big-endian
2639 systems since ext3-based ldiskfs is not supported on big-endian
2640 systems. This can be over-riden with "bigendian_extents" mount option.
2642 Severity : enhancement
2644 Description: The root squash functionality
2645 Details : A security feature, which is to prevent users from being able
2646 to mount lustre on their desktop, run as root, and delete
2647 all of the files in the filesystem. The goal is accomplished by
2648 remapping user id (UID) and group id (GID) of the root user to
2649 a UID and GID specified by the system administartor via Lustre
2650 configuration management server (MGS). The functionality also
2651 allows to specify sets of clients for which the remapping does
2656 Description: Excessive recovery window
2657 Details : With AT enabled, the recovery window can be excessively long (6000+
2658 seconds). To address this problem, we no longer use
2659 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2660 timeout no longer depends on the service time, it is set to
2661 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2662 time via pb_service_time.
2666 Description: Watchdog triggered on MDS failover
2667 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2668 the OSTs know that the MDS "UUID" can be reused for the same export
2669 from a different NID, so we do not need to wait for the export to be
2673 Frequency : rare, only if using MMP with Linux RAID
2675 Description: MMP doesn't work with Linux RAID
2676 Details : While using HA for Lustre servers with Linux RAID, it is possible
2677 that MMP will not detect multiple mounts. To make this work we
2678 need to unplug the device queue in RAID when the MMP block is being
2679 written. Also while reading the MMP block, we should read it from
2680 disk and not the cached one.
2682 Severity : enhancement
2684 Description: open file using fid
2685 Details : A file can be opened using just its fid, like
2686 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2689 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2692 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2693 Details : add proper locking for creation of jbd_4k slab cache
2697 Description: MMP check in ext3_remount() fails without displaying any error
2698 Details : When multiple mount protection fails during remount, proper error
2701 Severity : enhancement
2703 Description: Allow stripe size to be up to 4G-64k
2704 Details : Fix math logic to allow large stripe sizes.
2708 Description: add check for >8TB ldiskfs filesystems
2709 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2710 Don't allow >8TB ldiskfs filesystems to be mounted without
2711 force_over_8tb mount option
2716 Descriptoin: OSS complains "too many blocks" and crashes
2717 Details : In filter_commitrw_write():
2718 1. lquota_chkquota();
2719 2. obdo_from_inode();
2720 3. lquota_pending_commit();
2721 "2." has changed uid and gid in this case, then caused
2722 lquota_chkquota() and lquota_pending_commit() to use
2723 different uid and gid. That is the root of the bug.
2727 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2728 Details : Detect and print out the 'stripe count = -1' case properly.
2732 Description: Out or order replies might be lost on replay
2733 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2734 request with tid smaller than one currently being added, add it
2735 to the start, not end of the list.
2737 --------------------------------------------------------------------------------
2739 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2741 * Support for kernels:
2742 2.6.5-7.283 (SLES 9),
2743 2.6.9-55.EL (RHEL 4),
2744 2.6.16.46-0.14 (SLES 10),
2745 2.6.18.8 vanilla (kernel.org)
2746 * Client support for unpatched kernels:
2747 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2748 2.6.16 - 2.6.22 vanilla (kernel.org)
2749 * Due to recently discovered recovery problems, we do not recommend
2750 using patchless RHEL 4 clients with this or any earlier release.
2751 * Recommended e2fsprogs version: 1.39.cfs8
2752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2753 * Starting with this release, the ldiskfs backing filesystem required
2754 by Lustre is now in its own package, lustre-ldiskfs. This package
2755 should be installed. It is versioned separately from Lustre and
2756 may be released separately in future.
2760 Description: block reactivating mgc import until all deactivates complete
2761 Details : Fix race when failing back MDT/MGS to itself (testing)
2764 Frequency : at statup only
2766 Description: mds_lov_synchronize race leads to various problems
2767 Details : simultaneous MDT->OST connections at startup can cause the
2768 sync to abort, leaving the OSC in a bad state.
2770 Severity : enhancement
2772 Description: add optional extra BUILD_VERSION info
2773 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2774 to override the lustre version.
2777 Frequency : 2.6.18 servers only
2779 Description: ll_kern_mount() doesn't release the module reference
2780 Details : The ldiskfs module reference count never drops down to 0
2781 because ll_kern_mount() doesn't release the module reference.
2786 Description: server LBUG when using old ost_num_threads parameter
2787 Details : Accept the old ost_num_threads parameter but warn that it
2788 is deprecated, and fix an off-by-one error that caused an LBUG.
2793 Description: Transient SCSI error results in persistent IO issue
2794 Details : iobuf->dr_error is not reinitialized to 0 between two
2798 Frequency : sometimes when underlying device returns I/O errors
2800 Description: OSTs not going read-only during write failures
2801 Details : OSTs are not remounted read-only when the journal commit threads
2802 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2803 instead of the ext3 wrappers.
2807 Description: poor connect scaling with increasing client count
2808 Details : Don't run filter_grant_sanity_check for more than 100 exports
2809 to improve scaling for large numbers of clients.
2812 Frequency : SLES10 only
2814 Description: sanity-quota.sh quotacheck failed: rc = -22
2815 Details : Quotas cannot be enabled on SLES10.
2818 Frequency : liblustre clients only
2820 Description: getdirentries does not give error when run on compute nodes
2821 Details : getdirentries does not fail when the size specified as an argument
2822 is too small to contain at least one entry
2824 Severity : enhancement
2826 Description: Add LNET router traceability for debug purposes
2827 Details : If a checksum failure occurs with a router as part of the
2828 IO path, the NID of the last router that forwarded the bulk data
2829 is printed so it can be identified.
2834 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2835 Details : Due to a race condition, liblustre clients were occasionally
2836 evicted incorrectly.
2838 Severity : enhancement
2840 Description: lfs setstripe use optional parameters instead of postional
2843 Severity : enhancement
2845 Description: Nanosecond timestamp support for ldiskfs
2846 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2847 resolution timestamps. There is not yet support for this at
2848 the Lustre filesystem level.
2851 Frequency : during server recovery
2853 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2854 Details : request with rq_no_resend flag not awake l_wait_event if they get
2858 Frequency : nfs export on patchless client
2860 Description: connectathon hang when test nfs export over patchless client
2861 Details : Disconnected dentry cannot be found with lookup, so we do not need
2862 to unhash it or make it invalid
2865 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2866 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2867 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2872 Description: open req refcounting wrong on reconnect
2873 Details : If reconnect happened between getting open reply from server and
2874 call to mdc_set_replay_data in ll_file_open, we will schedule
2875 replay for unreferenced request that we are about to free.
2876 Subsequent close will crash in variety of ways.
2877 Check that request is still eligible for replay in
2878 mdc_set_replay_data().
2883 Description: disable writes to filesystem when reading health_check file
2884 Details : the default for reading the health_check proc file has changed
2885 to NOT do a journal transaction and write to disk, because this
2886 can cause reads of the /proc file to hang and block HA state
2887 checking on a healthy but otherwise heavily loaded system. It
2888 is possible to return to the previous behaviour during configure
2889 with --enable-health-write.
2891 Severity : enhancement
2893 Description: 64-bit inode version
2894 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2895 to the inode. This will be required for version-based recovery.
2900 Description: MDS fails to start if a duplicate client export is detected
2901 Details : in some rare cases it was possible for a client to connect to
2902 an MDS multiple times. Upon recovery the MDS would detect this
2903 and fail during startup. Handle this more gracefully.
2905 Severity : enhancement
2907 Description: Add -o localflock option to simulate old noflock
2909 Details : This will achieve local-only flock/fcntl locks
2915 Description: log_commit_thread vs filter_destroy race leads to crash
2916 Details : Take import reference before releasing llog record semaphore
2921 Description: Wrong request locking in request set processing
2922 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2923 imp_delayed_list, in this place should be used imp_lock.
2926 Frequency : when reconnection
2928 Description: Grant Leak when osc reconnect to OST
2929 Details : When osc reconnect ost, OST(filter) should check whether it
2930 should grant more space to client by comparing fed_grant and
2931 cl_avail_grant, and return the granted space to client instead
2932 of "new granted" space, because client will call osc_init_grant
2933 to update the client grant space info.
2936 Frequency : when client reconnect to OST
2938 Description: Grant Leak when osc do resend and replay bulk write
2939 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2940 bulk write request, because the grant info while be sync between
2941 OSC and OST when reconnect, and we should ignore the grant info
2942 these of resend/replay write req.
2947 Description: Grant space more than available left space sometimes.
2948 Details : When then OST is about to be full, if two bulk writing from
2949 different clients came to OST. Accord the avaliable space of the
2950 OST, the first req should be permitted, and the second one
2951 should be denied by ENOSPC. But if the seconde arrived before
2952 the first one is commited. The OST might wrongly permit second
2953 writing, which will cause grant space > available space.
2956 Frequency : when client is evicted
2958 Description: Grant might be wrongly erased when osc is evicted by OST
2959 Details : when the import is evicted by server, it will fork another
2960 thread ptlrpc_invalidate_import_thread to invalidate the
2961 import, where the grant will be set to 0. While the original
2962 thread will update the grant it got when connecting. So if
2963 the former happened latter, the grant will be wrongly errased
2964 because of this race.
2969 Description: Checking Stale with correct fid
2970 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2971 to check whether it is stale, because sometimes, we want the
2972 enqueue happened anyway, and op_data.fid2 will not be initialized.
2974 Severity : enhancement
2976 Description: update patchless client
2977 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2980 Frequency : only with 2.4 kernel
2982 Description: random memory corruption
2983 Details : size of struct ll_inode_info is to big for union inode.u and this
2984 can be cause of random memory corruption.
2989 Description: Memory leak in recovery
2990 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2991 It should also check obd_fail before fsfilt_start, otherwise if
2992 fsfilt_start return -EROFS,(failover mds during mds recovery).
2993 then the req will return with repmsg->transno = 0 and rc = EROFS.
2994 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2995 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2996 be freed no matter whether fsfilt_commit success or not.
2999 Frequency : only with huge count clients
3001 Description: Prevents from taking the superblock lock in llap_from_page for
3003 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3004 for llap_from_page call in ll_removepage prevents from taking the
3005 superblock lock for a soon died page.
3010 Description: Not check open intent error before release open handle
3011 Details : in some rare cases, the open intent error is not checked before
3012 release open handle, which may cause
3013 ASSERTION(open_req->rq_transno != 0), because it tries to release
3014 the failed open handle.
3019 Description: Set cat log bitmap only after create log success.
3020 Details : in some rare cases, the cat log bitmap is set too early. and it
3021 should be set only after create log success.
3025 Description: Accessing a block bevice can re-enable I/O when Lustre is
3026 tearing down a device.
3027 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3032 Description: service threads may hog cpus when there are a lot of requests
3033 Details : Insert cond_resched to give other threads a chance to use some CPU
3038 Description: the cat log was not initialized in recovery
3039 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3040 unlink log on mds will not be initialized until mds post
3041 recovery. And also in mds post recovery, the unlink log will
3042 initialization will be done asynchronausly, so there will be race
3043 between add unlink log and unlink log initialization.
3047 Description: brw_stats were being printed incorrectly
3048 Details : brw_stats were being printed as log2 but all of them were not
3049 recorded as log2. Also remove some code duplication arising from
3050 filter_tally_{read,write}.
3054 Frequency : rare, only in recovery.
3055 Description: ASSERTION(req->rq_type != LI_POISON) failed
3056 Details : imp_lock should be held while iterating over imp_sending_list for
3057 prevent destroy request after get timeout in ptlrpc_queue_wait.
3061 Description: replay-single.sh test 52 fails
3062 Details : A lock's skiplist need to be cleanup when it being unlinked
3063 from its resource list.
3067 Description: Short directio read returns full requested size rather than
3069 Details : Direct I/O operations should return actual amount of bytes
3070 transferred rather than requested size.
3072 Severity : enhancement
3074 Description: metadata RPC reduction (e.g. for rm performance)
3075 Details : decrease the amount of synchronous RPC between clients and servers
3076 by canceling conflicing lock before the operation on the client
3077 and packing thier handles into the main operation RPC to server.
3079 Severity : enhancement
3081 Description: Async OSC create to avoid the blocking unnecessarily.
3082 Details : If an OST has no remaining object, system will block on the
3083 creation when it needs to create a new object on this OST. Now,
3084 ways use pre-created objects when available, instead of blocking on
3085 an empty osc while others are not empty. If we must block, we
3086 block for the shortest possible period of time.
3090 Description: improve handling recoverable errors
3091 Details : if request processig with error which can be recoverable on server
3092 request should be resend, otherwise page released from cache and
3095 Severity : enhancement
3097 Description: refine locking for avoid write wrong info into lov_objid file
3098 Details : fix possible races with add new target and write/update data in
3101 --------------------------------------------------------------------------------
3103 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3108 Frequency : on some architectures
3110 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3111 Details : Uninitialized flags sometimes cause configuration commands to
3115 Frequency : patchless clients only
3117 Description: missing __iget() symbol export
3118 Details : The __iget() symbol export is missing. To avoid the need for
3119 this on patchless clients the deathrow inode reaper is turned
3120 off, and we depend on the VM to clean up old inodes. This
3121 dependency was during via the fix for bug 12181.
3123 --------------------------------------------------------------------------------
3125 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3127 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3128 INTEROPERATE with older versions automatically. In many cases a
3129 special upgrade step is needed. Please read the
3130 user documentation before upgrading any part of a live system.
3131 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3132 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3133 * WARNING: Lustre configuration and startup changes are required with
3134 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3138 Severity : enhancement
3140 Description: MountConf
3141 Details : Lustre configuration is now managed via mkfs and mount
3142 commands instead of lmc and lconf. New obd types (MGS, MGC)
3143 are added for dynamic configuration management. See
3144 https://mail.clusterfs.com/wikis/lustre/MountConf for
3147 Severity : enhancement
3149 Description: dynamic OST addition
3150 Details : OSTs can now be added to a live filesystem
3152 Severity : enhancement
3154 Description: startup order invariance
3155 Details : MDTs and OSTs can be started in any order. Clients only
3156 require the MDT to complete startup.
3158 Severity : enhancement
3160 Description: parallel, asynchronous orphan cleanup
3161 Details : orphan cleanup is now performed in separate threads for each
3162 OST, allowing parallel non-blocking operation.
3164 Severity : enhancement
3166 Description: optimized stripe assignment
3167 Details : stripe assignments are now made based on ost space available,
3168 ost previous usage, and OSS previous usage, in order to try
3169 to optimize storage space and networking resources.
3171 Severity : enhancement
3173 Description: Permanently set tunables
3174 Details : All writable /proc/fs/lustre tunables can now be permanently
3175 set on a per-server basis, at mkfs time or on a live system.
3177 Severity : enhancement
3179 Description: Lustre message v2
3180 Details : Add lustre message format v2.
3182 Severity : enhancement
3184 Description: client OST exclusion list
3185 Details : Clients can be started with a list of OSTs that should be
3186 declared "inactive" for known non-responsive OSTs.
3190 Description: SPEC SFS validation failure on NFS v2 over lustre.
3191 Details : Changes the blocksize for regular files to be 2x RPC size,
3192 and not depend on stripe size.
3194 Severity : enhancement
3196 Description: Multiple MD RPCs in flight.
3197 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3198 To avoid overly-overloading MDS, introduce a limit on number of
3199 MDS RPCs in flight for a single client and add /proc controls
3200 to adjust this limit.
3202 Severity : enhancement
3204 Description: client read/write statistics
3205 Details : Add client read/write call usage stats for performance
3206 analysis of user processes.
3207 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3208 file access. extents_stats shows chunk size distribution.
3209 extents_stats_per_process show chunk size distribution per
3212 Severity : enhancement
3214 Description: mds statistics
3215 Details : Add detailed mds operations statistics in
3216 /proc/fs/lustre/mds/*/stats.
3220 Description: Failure of copying files with lustre special EAs.
3221 Details : Client side always return success for setxattr call for lustre
3222 special xattr (currently only "trusted.lov").
3227 Description: Refcount LNET uuids
3228 Details : The global LNET uuid list grew linearly with every startup;
3229 refcount repeated list entries instead of always adding to
3232 Severity : enhancement
3234 Description: Dynamic service threads
3235 Details : Within a small range, start extra service threads
3236 automatically when the request queue builds up.
3239 Frequency : mixed-endian client/server environments
3241 Description: mixed-endian crashes
3242 Details : The new msg_v2 system had some failures in mixed-endian
3245 Severity : enhancement
3247 Description: Easy OST removal
3248 Details : OSTs can be permanently deactivated with e.g. 'lctl
3249 conf_param lustre-OST0001.osc.active=0'
3251 Severity : enhancement
3253 Description: MGS proc entries
3254 Details : Added basic proc entries for the MGS showing what filesystems
3257 Severity : enhancement
3259 Description: provide MGS failover
3260 Details : Added config lock reacquisition after MGS server failover.
3262 Severity : enhancement
3264 Description: add Linux 2.4 support
3265 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3269 Description: a large application tries to do I/O to the same resource and dies
3270 in the middle of it.
3271 Details : Check the req->rq_arrival time after the call to
3272 ost_brw_lock_get(), but before we do anything about
3273 processing it & sending the BULK transfer request. This
3274 should help move old stale pending locks off the queue as
3275 quickly as obd_timeout.
3278 Frequency : when an incorrect nid is specified during startup
3280 Description: ptlrpc connect to non-existant node causes kernel crash
3281 Details : LNET can't be re-entered from an event callback, which
3282 happened when we expire a message after the export has been
3283 cleaned up. Instead, hand the zombie cleanup off to another
3286 Severity : enhancement
3288 Description: plain/inodebits lock performance improvement
3289 Details : Grouping plain/inodebits in granted list by their request modes
3290 and bits policy, thus improving the performance of search through
3294 Frequency : only if OST filesystem is corrupted
3296 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3297 Details : for a short time RPCs with bulk IO are in the replay list,
3298 but replay of bulk IOs is unimplemented. If the OST filesystem
3299 is corrupted due to disk cache incoherency and then replay is
3300 started it is possible to trip an assertion. Avoid putting
3301 committed RPCs into the replay list at all to avoid this issue.
3304 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3306 Description: System hang on startup
3307 Details : This bug allowed the liblustre (e.g. catamount) client to
3308 return to the app before handling all startup RPCs. This
3309 could leave the node unresponsive to lustre network traffic
3310 and manifested as a server ptllnd timeout.
3312 Severity : enhancement
3314 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3315 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3316 boolean to control whether to print peer debug info when a
3317 client's RPC times out.
3320 Frequency : only for kernels with patches from Lustre below 1.4.3
3322 Description: Remove old rdonly API
3323 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3326 Frequency : only for devices with external journals
3328 Description: Set external device read-only also
3329 Details : During a commanded failover stop, we set the disk device
3330 read-only while the server shuts down. We now also set any
3331 external journal device read-only at the same time.
3334 Frequency : when upgrading from 1.4 while trying to change parameters
3336 Description: The wrong (new) MDC name was used when setting parameters for
3337 upgraded MDT's. Also allows changing of OSC (and MDC)
3338 parameters if --writeconf is specified at tunefs upgrade time.
3341 Frequency : when setting specific ost indicies
3343 Description: QOS code breaks on skipped indicies
3344 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3345 created with --index need not be sequential.
3347 Severity : enhancement
3349 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3350 Details : The uninit_groups feature works in conjunction with the kernel
3351 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3352 up the pass1 processing of e2fsck. This is a read-only feature
3353 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3354 mount filesystems that have had this feature enabled.
3356 Severity : enhancement
3358 Description: Improve multi-block allocation algorithm to avoid fragmentation
3359 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3360 allocation locality and avoid filesystem fragmentation.
3362 ------------------------------------------------------------------------------
3364 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3366 * Support for kernels:
3367 2.4.21-47.0.1.EL (RHEL 3)
3368 2.6.5-7.283 (SLES 9)
3369 2.6.9-42.0.10.EL (RHEL 4)
3370 2.6.12.6 vanilla (kernel.org)
3371 2.6.16.27-0.9 (SLES 10)
3372 * Recommended e2fsprogs version: 1.39.cfs5
3374 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3378 Frequency : occasional, depends on client load and configuration
3379 Bugzilla : 12181, 12203
3380 Description: data loss for recently-modified files
3382 Details : In some cases it is possible that recently written or created
3383 files may not be written to disk in a timely manner (this should
3384 normally be within 30s unless client IO load is very high).
3385 The problem appears as zero-length files or files that are a
3386 multiple of 1MB in size after a client crash or client eviction
3387 that are missing data at the end of the file.
3389 This problem is more likely to be hit on clients where files are
3390 repeatedly created and unlinked in the same directory, clients
3391 have a large amount of RAM, have many CPUs, the filesystem has
3392 many OSTs, the clients are rebooted frequently, and/or the files
3393 are not accessed by other nodes after being written.
3395 The presence of the problem can be detected by looking at
3396 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3397 smaller than the second (nr_unused) then dirty files will not
3398 be flushed automatically to disk. "sync; sleep 10" should be
3399 run several times on the node before unmounting it to update
3400 Lustre (this is also safe to run on nodes without this problem).
3402 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3403 that can cause this same problem, so customers using that kernel
3404 also need to update the kernel in addition to Lustre. In order
3405 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3407 It is normal that files written just before a client crash (less
3408 than 30s) may not yet have been flushed to disk, even for local
3412 Frequency : frequent on thin XT3 nodes
3414 Description: UUID collision on thin XT3 Linux nodes
3415 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3416 randomly, since we relied on an insufficiently-seeded PRNG.
3421 Description: OSS hangs after "All ost request buffers busy"
3422 Details : A deadlock between quota and journal operations caused OSS
3423 hangs after printing "All ost request buffers busy."
3426 Frequency : always on liblustre builds
3428 Description: Cleanup compiler warnings on liblustre
3431 Frequency : always on liblustre builds on XT3
3433 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3437 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3438 > 32000 subdirectories
3439 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3440 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3441 subdirectory count crosses 32000. This will aid e2fsck to
3442 correctly handle more than 32000 subdirectories.
3445 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3447 Description: System hang on startup
3448 Details : This bug allowed the liblustre (e.g. catamount) client to
3449 return to the app before handling all startup RPCs. This
3450 could leave the node unresponsive to lustre network traffic
3451 and manifested as a server ptllnd timeout.
3453 Severity : enhancement
3455 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3456 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3457 boolean to control whether to print peer debug info when a
3458 client's RPC times out.
3463 Description: make O_SYNC working on 2.6 kernels
3464 Details : 2.6 kernels use different method for mark pages for write,
3465 so need add a code to lustre for O_SYNC work.
3470 Description: Failure to close file and release space on NFS
3471 Details : Put inode details into lock acquired in ll_intent_file_open.
3472 Use mdc_intent_lock in ll_intent_open to properly
3473 detect all kind of errors unhandled by mdc_enqueue.
3478 Description: proc file read during shutdown sometimes raced obd removal,
3480 Details : Add lock to prevent obd access after proc file removal.
3483 Frequency : Only for files larger than 4GB on 32-bit clients.
3485 Description: improperly doing page alignment of locks
3486 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3487 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3492 Description: RPCs being resent when they shouldn't be
3493 Details : Some RPCs that should not be resent are being resent. This
3494 can cause inconsistencies in the RPC state machine. Do not
3495 resend such requests.
3498 Frequency : rare, only with NFS export
3500 Description: Crash on NFS re-export node
3501 Details : under very unusual load conditions an assertion is hit in
3502 ll_intent_file_open()
3505 Frequency : only if OST filesystem is corrupted
3507 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3508 Details : for a short time RPCs with bulk IO are in the replay list,
3509 but replay of bulk IOs is unimplemented. If the OST filesystem
3510 is corrupted due to disk cache incoherency and then replay is
3511 started it is possible to trip an assertion. Avoid putting
3512 committed RPCs into the replay list at all to avoid this issue.
3517 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3518 Details : Large single O_DIRECT read and write calls can fail to allocate
3519 a sufficiently large buffer to process the request. In case of
3520 allocation failure the allocation is retried with a smaller
3521 buffer and broken into smaller requests.
3523 Severity : enhancement
3525 Description: Add -o localflock option to simulate old noflock behaviour.
3526 Details : This will achieve local-only flock/fcntl locks coherentness.
3531 Description: versioning check is incomplete
3532 Details : Checking the version difference of client vs. server, report
3533 error if the gap is too big.
3538 Description: add support PG_writeback bit
3539 Details : add support for PG_writeback bit for Lustre, for more carefull
3540 work with page cache in 2.6 kernel. This also fix some deadlocks
3541 and remove hack for work O_SYNC with 2.6 kernel.
3543 Severity : enhancement
3545 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3546 Details : The uninit_groups feature works in conjunction with the kernel
3547 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3548 up the pass1 processing of e2fsck. This is a read-only feature
3549 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3550 mount filesystems that have had this feature enabled.
3552 Severity : enhancement
3554 Description: Improve multi-block allocation algorithm to avoid fragmentation
3555 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3556 allocation locality and avoid filesystem fragmentation.
3558 ------------------------------------------------------------------------------
3560 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3562 * Support for kernels:
3563 2.6.9-42.0.3.EL (RHEL 4)
3564 2.6.5-7.276 (SLES 9)
3565 2.4.21-47.0.1.EL (RHEL 3)
3566 2.6.12.6 vanilla (kernel.org)
3567 2.6.16.21-0.8 (SLES10)
3568 * Recommended e2fsprogs version: 1.39.cfs2-0
3570 * The backwards-compatible /proc/sys/portals symlink has been removed
3571 in this release. Before upgrading, please ensure that you change
3572 any configuration scripts or /etc/sysctl.conf files that access
3573 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3574 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3575 in advance of the upgrade on any system running Lustre 1.4.6 or
3576 newer, since /proc/sys/lnet was added in that version.
3577 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3581 Frequency : only when quota is used
3583 Description: avoid scanning export list for quota master
3584 Details : Change the algorithms to avoid scanning export list in order
3585 to improve the efficiency.
3588 Frequency : MDS failover only, very rarely
3590 Description: "went back in time" messages on mds failover
3591 Details : The greatest transno may be lost when the current operation
3592 finishes with an error (transno==0) and the client's last_rcvd
3593 record is over-written. Save the greatest transno in the
3594 mds_last_transno for this case.
3597 Frequency : always for specific kernels and striping counts
3599 Description: client may get "Matching packet too big" without ACL support
3600 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3601 when trying to access files in certain configurations. The
3602 clients should in fact be denied when mounting because they do
3603 not understand ACLs.
3606 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3608 Description: many clients connecting with IO in progress causes connect timeouts
3609 Details : Avoid synchronous journal commits to avoid delays caused by many
3610 clients connecting/disconnecting when bulk IO is in progress.
3611 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3612 OST_IO_PORTAL to avoid delays behind potentially many pending
3616 Frequency : occasionally with multiple writers to a single file
3618 Description: shared writes to file may result in wrong size reported by stat()
3619 Details : Allow growing of kms when extent lock is cancelled
3622 Frequency : always with random mmap IO to multi-striped file
3624 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3625 Details : Only if the hole is at the end of OST object so that kms is too
3626 small. Fix is to increase kms accordingly in ll_nopage.
3629 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3631 Description: writes to a missing object would leak memory on the OST
3632 Details : If there is an inconsistency between the MDS and OST filesystems,
3633 such that the MDS references an object that doesn't exist, writes
3634 to that object will leak memory due to incorrect cleanup in the
3635 error handling path, eventually running out of memory on the OST.
3640 Description: Creating too long symlink causes lustre errors
3641 Details : Check symlink and name lengths before sending requests to MDS.
3644 Frequency : only if flock is enabled (not on by default)
3646 Description: posix locks not released on fd closure on 2.6.9+
3647 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3648 kernels, this caused such locks not to be released on fd close and
3649 then assertions on fs unmount about still used locks.
3652 Frequency : MDS failover only, very rarely
3654 Description: clients may get ASSERTION(granted_lock != NULL)
3655 Details : When request was taking a long time, and a client was resending
3656 a getattr by name lock request. The were multiple lock requests
3657 with the same client lock handle and
3658 mds_getattr_name->fixup_handle_for_resent_request found one of the
3659 lock handles but later failed with ASSERTION(granted_lock != NULL).
3664 Description: handle->h_buffer_credits > 0, assertion failure
3665 Details : h_buffer_credits is zero after truncate, causing assertion
3666 failure. This patch extends the transaction or creates a new
3670 Frequency : NFS re-export or patchless client
3671 Bugzilla : 11179, 10796
3672 Description: Crash on NFS re-export node (__d_move)
3673 Details : We do not want to hash the dentry if we don't have a lock.
3674 But if this dentry is later used in d_move, we'd hit uninitialised
3675 list head d_hash, so we just do this to init d_hash field but
3676 leave dentry unhashed.
3679 Frequency : NFS re-export or patchless client
3681 Description: NFS exports has problem with symbolic link
3682 Details : lustre client didn't properly install dentry when re-exported
3683 to NFS or running patchless client.
3686 Frequency : NFS re-export or patchless client
3688 Description: Various nfs/patchless fixes.
3689 Details : fixes reuse disconected alias for lookup process - this fixes
3690 warning "find_exported_dentry: npd != pd",
3691 fix permission error with open files at nfs.
3692 fix apply umask when do revalidate.
3695 Frequency : occasional
3697 Description: Crash on NFS re-export node
3698 Details : calling clear_page() on the wrong pointer triggered oops in
3699 generic_mapping_read().
3702 Frequency : rarely, using O_DIRECT IO
3704 Description: unaligned directio crashes client with LASSERT
3705 Details : check for unaligned buffers before trying any requests.
3708 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3710 Description: stale data returned from RAID cache
3711 Details : If only a small amount of IO is done to the RAID device before
3712 reading it again it is possible to get stale data from the RAID
3713 cache instead of reading it from disk.
3716 Frequency : always for sles10 kernel
3718 Description: sles10 support
3719 Details : ll_follow_link: compile fixes and using of nd_set_link
3720 under newer kernels.
3723 Frequency : depends on arch, kernel and compiler version, always on sles10
3726 Description: recursive or deep enough symlinks cause stack overflow
3727 Details : getting rid of large stack-allocated variable in
3731 Frequency : depends on hardware
3733 Description: lustre write performance loss in the SLES10 kernel
3734 Details : the performance loss is caused by using of write barriers in the
3735 ext3 code. The SLES10 kernel turns barrier support on by
3736 default. The fix is to undo that change for ldiskfs.
3738 ------------------------------------------------------------------------------
3740 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3742 * Support for kernels:
3743 2.6.9-42.0.3EL (RHEL 4)
3744 2.6.5-7.276 (SLES 9)
3745 2.4.21-47.0.1.EL (RHEL 3)
3746 2.6.12.6 vanilla (kernel.org)
3750 Frequency : quota enabled and large files being deleted
3752 Description: releasing more than 4GB of quota at once hangs OST
3753 Details : If a user deletes more than 4GB of files on a single OST it
3754 will cause the OST to spin in an infinite loop. Release
3755 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3760 Description: statfs data retrieved from /proc may be stale or zero
3761 Details : When reading per-device statfs data from /proc, in the
3762 {kbytes,files}_{total,free,avail} files, it may appear
3763 as zero or be out of date.
3766 Frequency : systems with MD RAID1 external journal devices
3768 Description: lconf's call to blkid is confused by RAID1 journal devices
3769 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3770 of returning all block devices that match the journal UUID.
3773 Frequency : always, for aggregate stripe size over 4GB
3775 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3776 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3777 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3780 Frequency : always if "lfs find" used on a local file/directory
3782 Description: "lfs find" segfaults if used on a local file/directory
3783 Details : The case where a directory component was not specified wasn't
3784 handled correctly. Handle this properly.
3787 Frequency : always on ppc64
3789 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3790 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3791 perform endianness conversion.
3794 Frequency : rarely (truncate to non-zero file size after write under load)
3795 Bugzilla : 10730, 10687
3796 Description: Files padded with zeros to next 4K multiple
3797 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3798 it is possible that files that are truncated to a non-zero size
3799 immediately after being written are filled with zero bytes beyond
3800 the truncated size. No file data is lost.
3802 Severity : enhancement
3804 Description: Allow recovery/failover for liblustre clients.
3805 Details : liblustre clients were unaware of failover configurations until
3808 Severity : enhancement
3810 Description: user file locks should fail when not mounting with flock option
3811 Details : Set up an error-returning stub in ll_file_operations.lock field
3812 to prevent incorrect behaviour when client is mounted without
3813 flock option. Also, set up properly f_op->flock field for
3817 Frequency : always on ia64
3819 Description: "lfs df" loops on printing out MDS statfs information
3820 Details : The obd_ioctl_data was not initialized and in some systems
3821 this caused a failure during the ioctl that did not return
3822 an error. Initialize the struct and return an error on failure.
3825 Frequency : SLES 9 only
3827 Description: Error of copying files with lustre special EAs as root
3828 Details : Client side always return success for setxattr call for lustre
3829 special xattr (currently only "trusted.lov").
3832 Frequency : rarely on clusters with both ia64+i386 clients
3834 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3835 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3836 clients are doing concurrent writes to the same file it is
3837 possible that the ia64 clients may LASSERT because the OST
3838 extent locks are not PAGE_SIZE aligned. Ensure that grown
3839 locks are always aligned on the request boundary.
3842 Frequency : specific use, occasional
3844 Description: Overwriting in use executable truncates on-disk binary image
3845 Details : If one node attempts to overwrite an executable in use by
3846 another node, we now correctly return ETXTBSY instead of
3847 truncating the file.
3849 Severity : enhancement
3851 Description: Async OSC create to avoid the blocking unnecessarily.
3852 Details : If an OST has no remaining object, system will block on the creating
3853 when need to create a new object on this OST. Now, ways use
3854 pre-created objects when available, instead of blocking on an
3855 empty osc while others are not empty. If we must block, we block
3856 for the shortest possible period of time.
3861 Description: chmod on Lustre root is propagated to other clients
3862 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3863 invalid by the follow_mount time.
3866 Frequency : liblustre clients only
3868 Description: Race in 'instant cancel' lock handling could lead to such locks
3869 never to be granted in case of SMP MDS
3870 Details : Do not destroy not yet granted but cbpending locks in
3874 Frequency : replay/resend of open
3876 Description: non null lock assetion failure in mds_intent_policy
3877 Details : Trying to replay/resend lockless open requests resulted in
3878 mds_open() returning 0 with no lock. Now it sets a flag if
3879 a lock is going to be returned.
3881 Severity : enhancement
3883 Description: Checksum enhancements
3884 Details : New checksum enhancements allow for resending RPCs that failed
3887 Severity : enhancement
3889 Description: Tunables on number of dirty pages in cacche
3890 Details : Allow to set limit on number of dirty pages cached.
3895 Description: client crash on unmount - lock still has references
3896 Details : In some error handling cases it was possible to leak a lock
3897 reference on a client while accessing a file. This was not
3898 harmful to the client during operation, but would cause the
3899 client to crash when the filesystem is unmounted.
3902 Frequency : specific case, rare
3904 Description: ETXTBSY on mds though file not in use by client
3905 Details : ETXTBSY is no longer incorrectly returned when attempting to
3906 chmod or chown a directory that the user previously tried to
3907 execute or a currently-executing binary.
3910 Frequency : extremely rare except on liblustre-based clients
3912 Description: Lustre space not freed when files are deleted
3913 Details : Clean up open-unlinked files after client eviction. Previously
3914 the unlink was skipped and the files remained as orphans.
3919 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3920 Details : In some cases it was possible to send a blocking callback to
3921 a client doing a glimpse, even though that client didn't get
3922 a lock granted. When the glimpse lock is cancelled on the OST
3923 the freed lock is left on the waiting list and corrupted the list.
3926 Frequency : all core dumps
3928 Description: Broke core dumps to lustre
3929 Details : Negative dentry may be unhashed if parent does not have UPDATE
3930 lock, but some callers, e.g. do_coredump, expect dentry to be
3931 hashed after successful create, hash it in ll_create_it.
3933 ------------------------------------------------------------------------------
3935 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3937 * Support for kernels:
3938 2.6.9-42.0.2.EL (RHEL 4)
3939 2.6.5-7.276 (SLES 9)
3940 2.4.21-40.EL (RHEL 3)
3941 2.6.12.6 vanilla (kernel.org)
3945 Frequency : always on RHEL 3
3947 Description: Number of open files grows over time
3948 Details : The number of open files grows over time, whether or not
3949 Lustre is started. This was due to a filp leak introduced
3950 by one of our kernel patches.
3952 ------------------------------------------------------------------------------
3954 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3956 * Support for kernels:
3957 2.6.9-42.EL (RHEL 4)
3958 2.6.5-7.276 (SLES 9)
3959 2.4.21-40.EL (RHEL 3)
3960 2.6.12.6 vanilla (kernel.org)
3965 Bugzilla : 5719, 9635, 9792, 9684
3966 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3967 Details : If a server is under heavy load and cannot reply to new
3968 connection requests before the client resends the (re)connect,
3969 the connection handling code can behave badly if two service
3970 threads are concurrently handing separate (re)connections from
3971 the same client. Add better locking to the connection handling
3972 code, and ensure that only a single connection will be processed
3973 for a given client UUID, even if the lock is dropped.
3975 Severity : enhancement
3977 Description: add TCP zero-copy support to kernel
3978 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3979 sends if the hardware supports scatter-gather and checksumming.
3980 This allows socklnd to do client-write and server-read more
3981 efficiently and reduce CPU utilization from skbuf copying.
3984 Frequency : only if NFS exporting from client
3986 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3987 Details : If setting attributes on a file created under NFS that had
3988 never been opened it would be possible to oops the client
3989 if the file had no objects.
3992 Frequency : always for liblustre
3994 Description: liblustre client does MDS+OSTs setattr RPC for each write
3995 Details : When doing a write from a liblustre client, the client
3996 incorrectly issued an RPC to the MDS and each OST the file was
3997 striped over in order to update the timestamps. When writing
3998 with small chunks and many clients this could overwhelm the MDS
3999 with RPCs. In all cases it would slow down the write because
4000 these RPCs are unnecessary.
4002 Severity : enhancement
4004 Description: allow number of MDS service threads to be changed at module load
4005 Details : It is now possible to change the number of MDS service threads
4006 running. Adding "options mds mds_num_threads={N}" to the MDS's
4007 /etc/modprobe.conf will set the number of threads for the next
4008 time Lustre is restarted (assuming the "mds" module is also
4009 reloaded at that time). The default number of threads will
4010 stay the same, 32 for most systems.
4015 Description: OST crash if filesystem is unformatted or corrupt
4016 Details : If an OST is started on a device that has never been formatted
4017 or if the filesystem is corrupt and cannot even mount then the
4018 error handling cleanup routines would dereference a NULL pointer.
4023 Description: NULL pointer deref in llap_from_page.
4024 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4025 incorrect) mapping if the page was truncated/reclaimed while it was
4026 searched for. Check for this condition and skip such pages when
4027 doing readahead. Introduce extra check to llap_from_page() to
4028 verify page->mapping->host is non-NULL (so page is not anonymous).
4031 Frequency : Sometimes when using sys_sendfile
4033 Description: "page not covered by a lock" warnings from ll_readpage
4034 Details : sendfile called ll_readpage without right page locks present.
4035 Now we introduced ll_file_sendfile that does necessary locking
4036 around call to generic_file_sendfile() much like we do in
4040 Frequency : with certain MDS communication failures at client mount time
4042 Description: NULL pointer deref after failed client mount
4043 Details : a client connection request may delayed by the network layer
4044 and not be sent until after the PTLRPC layer has timed out the
4045 request. If the client fails the mount immediately it will try
4046 to clean up before the network times out the request. Add a
4047 reference from the request import to the obd device and delay
4048 the cleanup until the network drops the request.
4051 Frequency : occasionally during client (re)connect
4053 Description: assertion failure during client (re)connect
4054 Details : processing a client connection request may be delayed by the
4055 client or server longer than the client connect timeout. This
4056 causes the client to resend the connection request. If the
4057 original connection request is replied in this interval, the
4058 client may trip an assertion failure in ptlrpc_connect_interpret()
4059 which thought it would be the only running connect process.
4062 Frequency : only with obd_echo servers and clients that are rebooted
4064 Description: kernel BUG accessing uninitialized data structure
4065 Details : When running an obd_echo server it did not start the ping_evictor
4066 thread, and when a client was evicted an uninitialized data
4067 structure was accessed. Start the ping_evictor in the RPC
4068 service startup instead of the OBD startup.
4070 Severity : enhancement
4071 Bugzilla : 10193 (patchless)
4072 Description: Remove dependency on various unexported kernel interfaces.
4073 Details : No longer need reparent_to_init, exit_mm, exit_files,
4074 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4077 Frequency : rare (only users of deprecated and unsupported LDAP config)
4079 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4080 Details : LDAP apparently contains 'lustreName' attributes instead of
4081 'name'. A simple remapping of the name is sufficient.
4084 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4086 Description: waiting_locks_callback trips kernel BUG if client is evicted
4087 Details : Running with the dump_on_timeout debug flag turned on makes
4088 it possible that the waiting_locks_callback() can try to dump
4089 the Lustre kernel debug logs from an interrupt handler. Defer
4090 this log dumping to the expired_lock_main() thread.
4092 Severity : enhancement
4094 Description: Support NFS exporting on 2.6 kernels.
4095 Details : Implement non-rawops metadata methods for NFS server to use without
4096 changing NFS server code.
4099 Frequency : very rare (synthetic metadata workload only)
4101 Description: two racing renames might cause an MDS thread to deadlock
4102 Details : Running the "racer" program may cause one MDS thread to rename
4103 a file from being the source of a rename to being the target of
4104 a rename at exactly the same time that another thread is doing
4105 so, and the second thread has already enqueued these locks after
4106 doing a lookup of the target and is trying to relock them in
4107 order. Ensure that we don't try to re-lock the same resource.
4110 Frequency : only very large systems with liblustre clients
4112 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4113 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4114 all OSTs in parallel. This allows the eviction of stale liblustre
4115 clients to proceed much faster than if they were done in series,
4116 and also offers similar improvements for other set_info RPCs.
4121 Description: excessive CPU usage during initial read phase on client
4122 Details : During the initial read phase on a client, it would agressively
4123 retry readahead on the file, consuming too much CPU and impacting
4124 performance (since 1.4.5.8). Improve the readahead algorithm
4125 to avoid this, and also improve some other common cases (read
4126 of small files in particular, where "small" is files smaller than
4127 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4132 Description: MDS crash when receiving packet with unknown intent.
4133 Details : Do not LBUG in unknown intent case, just return -EFAULT
4135 Severity : enhancement
4136 Bugzilla : 9293, 9385
4137 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4138 Details : Do not serialize getattr (non-intent version) and statfs.
4141 Frequency : occasional, when OST network is overloaded/intermittent
4143 Description: client evicted by OST after bulk IO timeout
4144 Details : If a client sends a bulk IO request (read or write) the OST
4145 may evict the client if it is unresposive to its data GET/PUT
4146 request. This is incorrect if the network is overloaded (takes
4147 too long to transfer the RPC data) or dropped the OST GET/PUT
4148 request. There is no need to evict the client at all, since
4149 the pinger and/or lock callbacks will handle this, and the
4150 client can restart the bulk request.
4153 Frequency : Always when mmapping file with no objects
4155 Description: client crashes when mmapping file with no objects
4156 Details : Check that we actually have objects in a file before doing any
4157 operations on objects in ll_vm_open, ll_vm_close and
4163 Description: Request leak when working with deleted CWD
4164 Details : Introduce advanced request refcount tracking for requests
4165 referenced from lustre intent.
4167 Severity : Enhancement
4169 Description: Cache open file handles on client.
4170 Details : MDS now will return special lock along with openhandle, if
4171 requested and client is allowed to hold openhandle, even if unused,
4172 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4173 closing files for every read/write openration.
4175 Severity : Enhancement
4177 Description: Cache open negative dentries on client when possible.
4178 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4179 negative dentries on lock revocation.
4184 Description: Remounting a client read-only wasn't possible with a zconf mount
4185 Details : It wasn't possible to remount a client read-only with llmount.
4187 Severity : enhancement
4188 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4189 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4190 MPICH is included to add Lustre-specific ADIO interfaces.
4191 This is based closely on the UFS ADIO layer and only differs
4192 in file creation, in order to allow the OST striping to be set.
4193 This is user-contributed code and not supported by CFS.
4198 Description: extended inode attributes (immutable, append-only) work improperly
4199 when 2.4 and 2.6 kernels are used on client/server or vice versa
4200 Details : Introduce kernel-independent values for these flags.
4202 Severity : enhancement
4205 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4206 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4207 granularity. Now, Fractional MB granularity is be supported,
4208 this is very useful for low memory system.
4210 Severity : enhancement
4212 Description: Getattr by fid
4213 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4214 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4215 have locks and avoids getattr rpc requests for it on every stat.
4218 Frequency : Always, for filesystems larger than 2TB
4220 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4221 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4222 filesystems larger than 2TB to be mounted with mballoc enabled.
4225 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4227 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4228 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4229 filesystems larger than 2TB is mounted then the kernel will
4230 silently corrupt the start of the filesystem. CONFIG_LBD is
4231 enabled for all CFS-supported kernels, but the possibility of
4232 this happening with a modified kernel config exists.
4234 Severity : enhancement
4236 Description: add client O_DIRECT support for 2.6 kernels
4237 Details : It is now possible to do O_DIRECT reads and writes to files
4238 in the Lustre client mountpoint on 2.6 kernel clients.
4240 Severity : enhancement
4242 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4243 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4244 parallel, not waiting for response from every OST before sending
4245 a rpc to the next OST.
4250 Description: setattr vs write race when updating file timestamps
4251 Details : Client processes that update a file timestamp into the past
4252 right after writing to the file (e.g. tar) it is possible that
4253 the updated file modification time can be reset to the current
4254 time due to a race between processing the setattr and write RPC.
4256 Severity : enhancement
4258 Description: Bring 'lfs find' closer in line with regular Linux find.
4259 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4260 -print0 options and obtains all the needed info through the lustre
4263 Severity : enhancement
4265 Description: support up to 1024 configured devices on one node
4266 Details : change obd_dev array from statically allocated to dynamically
4267 allocated structs as they are first used to reduce memory usage
4272 Description: Flush dirty partially truncated pages during truncate
4273 Details : Immediatelly flush partially truncated pages in filter_setattr,
4274 this way we completely avoid having any pages in page cache on OST
4275 and can retire ugly workarounds during writes to flush such pages.
4280 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4281 Details : protect inode from truncation within vfs_unlink() context
4282 just take a reference before calling vfs_unlink() and release it
4283 when parent's i_sem is free.
4288 Description: last_id value checked outside lock on OST caused LASSERT failure
4289 Details : If there were multiple MDS->OST object precreate requests in
4290 flight, it was possible that the OST's last object id was checked
4291 outside a lock and incorrectly tripped an assertion. Move checks
4292 inside locks, and discard old precreate requests.
4295 Frequency : always, if extents are used on OSTs
4297 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4298 Details : OSTs using the extents format would not zero the high 16 bits of
4299 the index physical block number. This is not a problem for any
4300 OST filesystems smaller than 16TB, and no kernels support ext3
4301 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4302 new/modified files) and can be fixed for existing filesystems
4303 with e2fsprogs-1.39-cfs1.
4308 Description: import connection selection may be incorrect if timer wraps
4309 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4310 import connections to be ignored if the 32-bit jiffies counter
4311 wraps. Use a 64-bit jiffies counter.
4314 Frequency : very large clusters immediately after boot
4316 Description: LNET request buffers exhausted under heavy short-term load
4317 Details : If a large number of client requests are generated on a service
4318 that has previously never seen so many requests it is possible
4319 that the request buffer growth cannot keep up with the spike in
4320 demand. Instead of dropping incoming requests, they are held in
4321 the LND until the RPC service can accept more requests.
4324 Frequency : Sometimes during replay
4326 Description: Assertion failure in ll_local_open after replay.
4327 Details : If replay happened on an open request reply before we were able
4328 to set replay handler, reply will become not swabbed tripping the
4329 assertion in ll_local_open. Now we set the handler right after
4330 recognising of open request
4333 Frequency : very rare
4335 Description: kernel reports "badness in vsnprintf"
4336 Details : Reading from the "recovery_status" /proc file in small chunks
4337 may cause a negative length in lprocfs_obd_rd_recovery_status()
4338 call to vsnprintf() (which is otherwise harmless). Exit early
4339 if there is no more space in the output buffer.
4341 Severity : enhancement
4343 Description: clear OBD RPC statistics by writing to them
4344 Details : It is now possible to clear the OBD RPC statistics by writing
4345 to the "stats" file.
4350 Description: Client mtime is not the same on different clients after utimes
4351 Details : In some cases, the client was using the utimes() syscall on
4352 a file cached on another node. The clients now validate the
4353 ctime from the MDS + OSTs to determine which one is right.
4358 Description: Inability to activate failout mode
4359 Details : lconf script incorrectly assumed that in pythong string's numeric
4360 value is used in comparisons.
4363 Frequency : always with multiple stripes per file
4365 Description: Inefficient object allocation for mutli-stripe files
4366 Details : When selecting which OSTs to stripe files over, for files with
4367 a stripe count that divides evenly into the number of OSTs,
4368 the MDS is always picking the same starting OST for each file.
4369 Return the OST selection heuristic to the original design.
4374 Description: mount failures may take full timeout to return an error
4375 Details : Under some heavy load conditions it is possible that a
4376 failed mount can wait for the full obd_timeout interval,
4377 possibly several minutes, before reporting an error.
4378 Instead return an error as soon as the status is known.
4380 Frequency : quota enabled and large files being deleted
4382 Description: releasing more than 4GB of quota at once hangs OST
4383 Details : If a user deletes more than 4GB of files on a single OST it
4384 will cause the OST to spin in an infinite loop. Release
4385 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4390 Description: statfs data retrieved from /proc may be stale or zero
4391 Details : When reading per-device statfs data from /proc, in the
4392 {kbytes,files}_{total,free,avail} files, it may appear
4393 as zero or be out of date.
4396 Frequency : systems with MD RAID1 external journal devices
4398 Description: lconf's call to blkid is confused by RAID1 journal devices
4399 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4400 of returning all block devices that match the journal UUID.
4403 Frequency : always, for aggregate stripe size over 4GB
4405 Description: assertion fails when trying to use 4GB stripe size
4406 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4407 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4410 Frequency : always on ppc64
4412 Description: the first write on an ext3 filesystem with mballoc got stuck
4413 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4414 perform endianness conversion.
4416 ------------------------------------------------------------------------------
4418 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4420 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4421 INTEROPERATE with older versions automatically. Please read the
4422 user documentation before upgrading any part of a live system.
4423 * WARNING: Lustre networking configuration changes are required with
4424 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4427 * Support for kernels:
4428 2.6.9-22.0.2.EL (RHEL 4)
4429 2.6.5-7.244 (SLES 9)
4430 2.6.12.6 vanilla (kernel.org)
4433 Severity : enhancement
4434 Bugzilla : 7981/8208
4435 Description: Introduced Lustre Networking (LNET)
4436 Details : LNET is new networking infrastructure for Lustre, it includes
4437 a reorganized network configuration mode (see the user
4438 documentation for full details) as well as support for routing
4439 between different network fabrics. Lustre Networking Devices
4440 (LNDs) for the supported network fabrics have also been
4441 created for this new infrastructure.
4443 Severity : enhancement
4444 Description: Introduced Access control lists
4445 Details : clients can set ACLs on files and directories in order to have
4446 more fine-grained permissions than the standard Unix UGO+RWX.
4447 The MDS must be started with the "-o acl" mount option.
4449 Severity : enhancement
4450 Description: Introduced filesystem quotas
4451 Details : Administrators may now establish per-user quotas on the
4454 Severity : enhancement
4456 Description: Configuration change for the XT3
4457 The PTLLND is now used to run Lustre over Portals on the XT3
4458 The configure option(s) --with-cray-portals are no longer used.
4459 Rather --with-portals=<path-to-portals-includes> is used to
4460 enable building on the XT3. In addition to enable XT3 specific
4461 features the option --enable-cray-xt3 must be used.
4466 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4467 Details : A race between allocating a new llcd and re-getting the llcd_lock
4468 allowed another thread to grab newly-allocated llcd.
4470 Severity : enhancement
4472 Description: 2.6 OST async journal commit and locking fix to improve performance
4473 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4474 2.6 kernels are now async as they already were in 2.4 kernels so
4475 that they can commit concurrently with the network bulk transfer.
4476 For block-allocated files the filter allocation semaphore is held
4477 to avoid filesystem fragmentation during allocation. BKL lock
4478 removed for 2.6 xattr operations where it is no longer needed.
4483 Description: lconf incorrectly determined whether two IP networks could talk
4484 Details : In some more complicated routing and multiple-network
4485 configurations, lconf will avoid trying to make a network
4486 connection to a disjoint part of the IP space. It was doing the
4487 math incorrectly for one set of cases.
4492 Description: Fix for potential infinite loop processing records in an llog.
4493 Details : If an llog record is corrupted/zeroed, it is possible to loop
4494 forever in llog_process(). Validate the llog record length
4495 and skip the remainder of the block on error.
4498 Frequency : occasional (liblustre only)
4500 Description: liblustre could not open files whose last component is a symlink
4501 Details : sysio_path_walk() would incorrectly pass the open intent to
4502 intermediate path components.
4505 Frequency : rare (liblustre only with non-standard tuning)
4506 Bugzilla : 7201 (7350)
4507 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4508 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4509 after it has been released, only do this for referenced locks
4511 Severity : enhancement
4513 Description: specifying an (invalid) directory default stripe_size of -1
4514 would reset the directory default striping
4515 Details : stripe_size -1 was used internally to signal directory stripe
4516 removal, now use "all default" to signal dir stripe removal
4517 as a directory striping of "all default" is not useful
4520 Frequency : common for large clusters running liblustre clients
4522 Description: doing an ls when liblustre clients are running is slow
4523 Details : sending a glimpse AST to a liblustre client waits for every AST
4524 to time out, as liblustre clients will not respond. Since they
4525 cannot cache data we refresh the OST lock LVB from disk instead.
4527 Severity : enhancement
4529 Description: doing an ls at the same time as file IO can be slow
4530 Details : enqueue and other "small" requests can be blocked behind many
4531 large IO requests. Create a new OST IO portal for non-IO
4532 requests so they can be processed faster.
4535 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4537 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4538 created file with mode 000
4539 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4540 allow an ftruncate() as a non-root user to a file with mode 000.
4541 Lustre now respects this flag to disable mode checks when
4542 truncating a file owned by the user
4545 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4548 Description: Revoking locks from clients that went dead or catatonic might take
4550 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4551 cancellation of such locks instant on servers without waiting for
4552 any reply from clients. Clients drops these locks when cancel
4553 notification from server is received without replying.
4556 Frequency : liblustre-only, when liblustre client dies or becomes busy
4558 Description: Doing ls on Linux clients can take a long time with active
4560 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4561 granting such locks to it in the first place if possible. Locks
4562 are taken by proxy on the OST during the read or write and
4563 dropped immediately afterward. Add connect flags handling, do
4564 not grant locks to liblustre clients for glimpse ASTs.
4566 Severity : enhancement
4568 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4569 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4570 sequential reads and behaves badly with applications doing random
4571 reads. Improve it by reading ahead at least read region, and
4572 avoiding excessive large RPC for small reads.
4574 Severity : enhancement
4576 Description: Creating more than 1000 files for a single job may cause a load
4577 imbalance on the OSTs if there are also a large number of OSTs.
4578 Details : qos_prep_create() uses an OST index reseed value that is an
4579 even multiple of the number of available OSTs so that if the
4580 reseed happens in the middle of the object allocation it will
4581 still utilize the OSTs as uniformly as possible.
4586 Description: OST or MDS may oops in ping_evictor_main()
4587 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4588 but doesn't restart at beginning of obd_exports_timed list
4591 Severity : enhancement
4593 Description: improve by-nid export eviction on the MDS and OST
4594 Details : allow multiple exports with the same NID to be evicted at one
4595 time without re-searching the exports list.
4598 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4600 Description: MDS may oops in groups_free()
4601 Details : in rare race conditions a newly allocated group_info struct is
4602 freed again, and this can be NULL. The 2.4 compatibility code
4603 for groups_free() checked for a NULL pointer, but 2.6 did not.
4606 Frequency : common for liblustre clients doing little filesystem IO
4607 Bugzilla : 9352, 7313
4608 Description: server may evict liblustre clients accessing contended locks
4609 Details : if a client is granted a lock or receives a completion AST
4610 with a blocking AST already set it would not reply to the AST
4611 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4614 Frequency : lfs setstripe, only systems with more than 160 OSTs
4616 Description: unable to set striping with a starting offset beyond OST 160
4617 Details : llapi_create_file() incorrectly limited the starting stripe
4618 index to the maximum single-file stripe count.
4621 Frequency : LDAP users only
4623 Description: lconf did not handle in-kernel recovery with LDAP properly
4624 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4626 Severity : enhancement
4628 Description: bind OST threads to NUMA nodes to improve performance
4629 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4630 node and do their allocations there to localize memory access
4632 Severity : enhancement
4634 Description: llmount can determine client NID directly from Myrinet (GM)
4635 Details : the client NID code from gmnalnid was moved directly into
4636 llmount, removing the need to use this or specifying the
4637 client NID explicitly when mounting GM clients with zeroconf
4640 Frequency : if client is started with down MDS
4642 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4643 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4644 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4645 ptlrpc_import_delayed_req() because we didn't check if the
4646 request was interrupted, and we also didn't break out of the
4647 event loop if there was no timeout
4652 Description: data loss during non-page-aligned writes to a single file from
4653 both multiple nodes and multiple threads on one node at same time
4654 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4655 inconsistency led to false short-reads, that were cached and later
4656 used by ->prepare_write() to fill in partially written page,
4657 leading to data loss.
4660 Frequency : always, if lconf --abort_recovery used
4662 Description: lconf --abort_recovery fails with 'Operation not supported'
4663 Details : lconf was attempting to abort recovery on the MDT device and not
4666 Severity : enhancement
4668 Description: remove cleanup logs
4669 Details : replace lconf-generated cleanup logs with lustre internal
4670 cleanup routines. Eliminates the need for client-cleanup and
4673 Severity : enhancement
4675 Description: add support for EAs (user and system) on lustre filesystems
4676 Details : it is now possible to store extended attributes in the Lustre
4677 client filesystem, and with the user_xattr mount option it
4678 is possible to allow users to store EAs on their files also
4680 Severity : enhancement
4682 Description: Add possibility (config option) to show minimal available OST free
4684 Details : When compiled with --enable-mindf configure option, statfs(2)
4685 (and so, df) will return least minimal free space available from
4686 all OSTs as amount of free space on FS, instead of summary of
4687 free spaces of all OSTs.
4689 Severity : enhancement
4691 Description: do not expand extent locks acquired on OST-side
4692 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4693 by server: they are not cached anyway.
4696 Frequency : when mmap is used/binaries executed from Lustre
4698 Description: Unmmap pages before throwing them away from read cache.
4699 Details : llap_shrink cache now attempts to unmap pages before discarding
4700 them (if unmapping failed - do not discard). SLES9 kernel has
4701 extra checks that trigger if this unmapping is not done first.
4706 Description: lconf didn't resolve symlinks before checking to see whether a
4707 given mountpoint was already in use
4710 Frequency : when migrating failover services
4711 Bugzilla : 6395, 9514
4712 Description: When migrating a subset of services from a node (e.g. failback
4713 from a failover service node) the remaining services would
4714 time out and evict clients.
4715 Details : lconf --force (implied by --failover) sets the global obd_timeout
4716 to 5 seconds in order to quickly disconnect, but this caused
4717 other RPCs to time out too quickly. Do not change the global
4718 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4720 Severity : enhancement
4721 Frequency : if MDS is started with down OST
4722 Bugzilla : 9439,5706
4723 Description: Allow startup/shutdown of an MDS without depending on the
4724 availability of the OSTs.
4725 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4726 Add appropriate locking and lov-osc refcounts for safe
4727 cleaning. Add osc abort_inflight calls in case the
4728 synchronize never started.
4731 Frequency : occasional (Cray XT3 only)
4733 Description: root not authorized to access files in CRAY_PORTALS environment
4734 Details : The client process capabilities were not honoured on the MDS in
4735 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4736 been accessed by an authorized user then root was able to access
4737 the file on the local client also. The root user capabilities
4738 are now allowed on the MDS, as this environment has secure UID.
4741 Frequency : occasional
4743 Description: ldiskfs "too long searching" message happens too often
4744 Details : A debugging message (otherwise harmless) prints too often on
4745 the OST console. This has been reduced to only happen when
4746 there are fragmentation problems on the filesystem.
4751 Description: Division by zero in statfs when all OSCs are inactive
4752 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4753 lov_statfs divides by value returned by lov_get_stripecnt().
4757 Bugzilla : 9489, 3273
4758 Description: First write from each client to each OST was only 4kB in size,
4759 to initialize client writeback cache, which caused sub-optimal
4760 RPCs and poor layout on disk for the first written file.
4761 Details : Clients now request an initial cache grant at (re)connect time
4762 and so that they can start streaming writes to the cache right
4763 away and always do full-sized RPCs if there is enough data.
4764 If the OST is rebooted the client also re-establishes its grant
4765 so that client cached writes will be honoured under the grant.
4770 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4771 Details : Now I/O RPCs go to different portal number and (presumably) fast
4772 lock requests (and glimses) and other RPCs get their own service
4773 threads pool that should be able to service those RPCs
4776 Severity : enhancement
4778 Description: Ability to exchange lustre version between client and servers and
4779 issue warnings at client side if client is too old. Also for
4780 liblustre clients there is ability to refuse connection of too old
4782 Details : New 'version' field is added to connect data structure that is
4783 filled with version info. That info is later checked by server and
4787 Frequency : rare, liblustre only.
4788 Bugzilla : 9296, 9581
4789 Description: Two simultaneous writes from liblustre at offset within same page
4790 might proceed at the same time overwriting eachother with stale
4792 Details : I/O lock withing llu_file_prwv was released too early, before data
4793 actually was hitting the wire. Extended lock-holding time until
4794 server acknowledges receiving data.
4797 Frequency : extremely rare. Never observed in practice.
4799 Description: avoid generating lustre_handle cookie of 0.
4800 Details : class_handle_hash() generates handle cookies by incrementing
4801 global counter, and can hit 0 occasionaly (this is unlikely, but
4802 not impossible, because initial value of cookie counter is
4803 selected randonly). Value of 0 is used as a sentinel meaning
4804 "unassigned handle" --- avoid it. Also coalesce two critical
4805 sections in this function into one.
4807 Severity : enhancement
4809 Description: allow liblustre clients to delegate truncate locking to OST
4810 Details : To avoid overhead of locking, liblustre client instructs OST to
4811 take extent lock in ost_punch() on client's behalf. New connection
4812 flag is added to handle backward compatibility.
4814 Severity : enhancement
4815 Bugzilla : 4928, 7341, 9758
4816 Description: allow number of OST service threads to be specified
4817 Details : a module parameter allows the number of OST service threads
4818 to be specified via "options ost ost_num_threads={N}" in the
4819 OSS's /etc/modules.conf or /etc/modprobe.conf.
4823 Bugzilla : 6146, 9635, 9895
4824 Description: servers crash with bad pointer in target_handle_connect()
4825 Details : In rare cases when a client is reconnecting it was possible that
4826 the connection request was the last reference for that export.
4827 We would temporarily drop the export reference and get a new
4828 one, but this may have been the last reference and the export
4829 was just destroyed. Get new reference before dropping old one.
4831 Severity : enhancement
4832 Frequency : if client is started with failover MDS
4834 Description: Allow multiple MDS hostnames in the mount command
4835 Details : Try to read the configuration from all specified MDS
4836 hostnames during a client mount in case the "primary"
4839 Severity : enhancement
4841 Description: Stop sending data to evicted clients as soon as possible.
4842 Details : Check if the client we are about to send or are sending data to
4843 was evicted already. (Check is done every second of waiting,
4844 for which l_wait_event interface was extended to allow checking
4845 of exit condition at specified intervals).
4848 Frequency : rare, normally only when NFS exporting is done from client
4850 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4852 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4855 Frequency : always, if srand() is called before liblustre initialization
4857 Description: Liblustre uses system PRNG disturbing its usage by user application
4858 Details : Introduce internal to lustre fast and high-quality PRNG for
4859 lustre usage and make liblustre and some other places in generic
4860 lustre code to use it.
4862 Severity : enhancement
4863 Bugzilla : 9477, 9557, 9870
4864 Description: Verify that the MDS configuration logs are updated when xml is
4865 Details : Check if the .xml configuration logs are newer than the config
4866 logs stored on the MDS and report an error if this is the case.
4867 Request --write-conf, or allow starting with --old_conf.
4869 Severity : enhancement
4871 Description: Handle symlinks in the path when checking if Lustre is mounted.
4872 Details : Resolve intermediate symlinks when checking if a client has
4873 mounted a filesystem to avoid duplicate client mounts.
4878 Description: lconf can hit an error exception but still return success.
4879 Details : The lconf command catches the Command error exception at the top
4880 level script context and will exit with the associated exit
4881 status, but doesn't ensure that this exit status is non-zero.
4886 Description: failure of ptlrpc thread startup can cause oops
4887 Details : Starting a ptlrpc service thread can fail if there are a large
4888 number of threads or the server memory is very fragmented.
4889 Handle this without oopsing.
4892 Frequency : always, only if liblustre and non-default acceptor port was used
4894 Description: liblustre cannot connect to servers with non-default acceptor port
4895 Details : tcpnal_set_default_params() was not called and was therefore
4896 ignoring the environment varaible TCPNAL_PORT, as well as other
4897 TCPNAL_ environment variables
4902 Description: two objects could be created on the same OST for a single file
4903 Details : If an OST is down, in some cases it was possible to create two
4904 objects on a single OST for a single file. No problems other
4905 than potential performance impact and spurious error messages.
4909 Bugzilla : 5681, 9562
4910 Description: Client may oops in ll_unhash_aliases
4911 Details : Client dcache may become inconsistent in race condition.
4912 In some cases "getcwd" can fail if the current directory is
4918 Description: Inode refcounting problems in NFS export code
4919 Details : link_raw functions used to call d_instantiate without obtaining
4920 extra inode reference first.
4924 Bugzilla : 9942, 9903
4925 Description: Referencing freed requests leading to crash, memleaks with NFS.
4926 Details : We used to require that call to ll_revalidate_it was always
4927 followed by ll_lookup_it. Also with revalidate_special() it is
4928 possible to call ll_revalidate_it() twice for the same dentry
4929 even if first occurence returned success. This fix changes semantic
4930 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4931 reference on a request referred from the intent.
4932 ll_intent_release() then releases such a request.
4935 Frequency : rare, normally benchmark loads only
4937 Description: unlinked inodes were kept in memory on the client
4938 Details : If a client is repeatedly creating and unlinking files it
4939 can accumulate a lot of stale inodes in the inode slab cache.
4940 If there is no other client load running this can cause the
4941 client node to run out of memory. Instead flush old inodes
4942 from client cache that have the same inode number as a new inode.
4945 Frequency : SLES9 2.6.5 kernel and long filenames only
4946 Bugzilla : 9969, 10379
4947 Description: utime reports stale NFS file handle
4948 Details : SLES9 uses out-of-dentry names in some cases, which confused
4949 the lustre dentry revalidation. Change it to always use the
4953 Frequency : rare, unless heavy write-truncate concurrency is continuous
4954 Bugzilla : 4180, 6984, 7171, 9963, 9331
4955 Description: OST becomes very slow and/or deadlocked during object unlink
4956 Details : filter_destroy() was holding onto the parent directory lock
4957 while truncating+unlinking objects. For very large objects this
4958 may block other threads for a long time and slow overall OST
4959 responsiveness. It may also be possible to get a lock ordering
4960 deadlock in this case, or run out of journal credits because of
4961 the combined truncate+unlink. Solution is to do object truncate
4962 first in one transaction without parent lock, and then do the
4963 final unlink in a new transaction with the parent lock. This
4964 reduces the lock hold time dramatically.
4967 Frequency : rare, 2.4 kernels only
4969 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4970 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4971 before unmounting their filesystem in order to avoid deadlock.
4972 The kernel_locked() function in 2.4 kernels only checks whether
4973 the kernel lock is held, not whether it is this process that is
4974 holding it as 2.6 kernels do.
4979 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4980 Details : The client ptlrpc code may be trying to reconnect to a down
4981 server before a previous connection attempt has timed out.
4982 Increase the reconnect interval to be longer than the connection
4983 timeout interval to avoid sending duplicate connections to
4987 Frequency : echo_client brw_test command
4989 Description: fix echo_client to work with OST preallocated code
4990 Details : OST preallocation code (5137) didn't take echo_client IO path
4991 into account: echo_client calls filter methods outside of any
4992 OST thread and, hence, there is no per-thread preallocated
4993 pages and buffers to use. Solution: hijack pga pages for IO. As
4994 a byproduct, this avoids unnecessary data copying.
4998 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4999 Description: Client can oops in mdc_commit_close() after open replay
5000 Details : It was possible for the MDS to return an open request with no
5001 transaction number in mds_finish_transno() if the client was
5002 evicted, but without actually returning an error. Clients
5003 would later try to replay that open and may trip an assertion
5004 Simplify the client close codepath, and always return an error
5005 from the MDS in case the open is not successful.
5008 Frequency : rare, 2.6 OSTs only
5010 Description: OST may deadlock under high load on fragmented files
5011 Details : If there was a heavy load and highly-fragmented OST filesystems
5012 it was possible to have all the OST threads deadlock waiting on
5013 allocation of biovecs, because the biovecs were not released
5014 until the entire RPC IO was completed. Instead, release biovecs
5015 as soon as they are complete to ensure forward IO progress.
5017 Severity : enhancement
5019 Description: Support for specifying external journal device at mount
5020 Details : If an OST or MDS device is formatted with an external journal
5021 device, this device major/minor is stored in the ext3 superblock
5022 and may not be valid for failover. Allow detecting and
5023 specifying the external journal at mount time.
5028 Description: Mounting an MDS with pending unlinked files may cause oops
5029 Details : target_finish_recovery() calls mds_postrecov() which returned
5030 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5031 considers this an error and immediately begins cleaning up the
5032 lov, just after starting the mds_lov process
5034 Severity : enhancement
5036 Description: Implement 'lfs df' to report actual free space on per-OST basis
5037 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5038 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5039 usage of objects; '-h' to report in human readable format.
5041 ------------------------------------------------------------------------------
5043 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5050 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5051 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5052 at startup during orphan unlinking. Instead, skip buddy update
5053 before setup, as it will be regenerated after recovery is complete.
5058 Description: If an OST is inactive, its locks might reference stale inodes.
5059 Details : lov_change_cbdata() must iterate over all namespaces, even if
5060 they are inactive to clear inode references from the lock.
5062 Severity : enhancement
5063 Frequency : occasional, if non-standard max_dirty_mb used
5065 Description: Client will block write RPCs if not enough grant
5066 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5067 then the client will block writes while waiting for another RPC
5068 to complete instead of consuming its dirty limit. With change
5069 we get improved performance when max_dirty_mb is small.
5071 Severity : enhancement
5072 Bugzilla : 3389, 6253
5073 Description: Add support for supplementary groups on the MDS.
5074 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5075 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5076 lookups for user supplementary groups into a cache.
5080 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5081 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5082 for liblustre O_EXCL is enforced by MDS.
5086 Description: readdir never returns NULL in liblustre.
5087 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5088 dentry in struct dirent.
5092 Frequency : liblustre only, depends on application IO pattern
5093 Description: liblustre clients evicted if not contacting servers
5094 Details : Don't put liblustre clients into the ping_evictor list, so
5095 they will not be evicted by the pinger ever.
5097 Severity : enhancement
5099 Description: Add ability to evict clients by NID from MDS.
5100 Details : By echoing "nid:$NID" string into
5101 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5102 $NID would be instantly evicted from this MDS and from all active
5103 OSTs connected to it.
5107 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5108 Details : lookup_it_finish() used to query file size from OSTs that was not
5113 Description: service threads change working directory to that of init
5114 Details : Starting lustre service threads may pin the working directory
5115 of the parent thread, making that filesystem busy. Threads
5116 now change to the working directory of init to avoid this.
5120 Frequency : during shutdown only
5121 Description: shutdown with a failed MDS or OST can cause unmount to hang
5122 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5127 Frequency : 2.6 only
5128 Description: chmod/chown may include an extra supplementary group
5129 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5130 supplementary group and if none is specified this is used.
5133 Bugzilla : 5479 (6816)
5135 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5136 Details : If lookup is in progress on a file that is unlinked we might try
5137 to revalidate the inode and fail in revalidate after lookup is
5138 complete and ll_file_open() enqueues the open again but
5139 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5142 Frequency : always, if lconf --abort_recovery used
5144 Description: lconf --abort_recovery fails with 'Operation not supported'
5145 Details : lconf was attempting to abort recovery on the MDT device and not
5148 ------------------------------------------------------------------------------
5150 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5155 Frequency : rare (only unsupported configurations with a node running as an
5157 Bugzilla : 6514, 5137
5158 Description: Mounting a Lustre file system on a node running as an OST could
5160 Details : OSTs now preallocates memory needed to write out data at
5161 startup, instead of when needed, to avoid having to
5162 allocate memory in possibly low memory situations.
5163 Specifically, if the file system is mounted on on OST,
5164 memory pressure could force it to try to write out data,
5165 which it needed to allocate memory to do. Due to the low
5166 memory, it would be unable to do so and the node would
5167 become unresponsive.
5169 Severity : enhancement
5171 Description: Addition of lconf --service command line option
5172 Details : lconf now accepts a '--service <arg>' option, which is
5173 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5175 Severity : enhancement
5177 Description: Failover mode is now the default for OSTs.
5178 Details : By default, OSTs will now run in failover mode. To return to
5179 the old behaviour, add '--failout' to the lmc line for OSTs.
5181 Severity : enhancement
5183 Description: Health checks are now provided for MDS and OSTs
5184 Details : Additional detailed health check information on MSD and OSTs
5185 is now provided through the procfs health_check value.
5188 Frequency : occasional, depends on IO load
5190 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5191 after numerous create/delete cycles
5192 Details : The ext3 inode allocation policy would not allocate new inodes
5193 very well on the OSTs because there are no new directories
5194 being created. Instead we look for groups with free space if
5195 the parent directories are nearly full.
5200 Description: Network or server problems during mount may cause partially
5201 mounted clients instead of returning an error.
5202 Details : The config llog parsing code may overwrite the error return
5203 code during mount error handling, returning success instead
5209 Description: MDS can fail to allocate large reply buffers
5210 Details : After long uptimes the MDS can fail to allocate large reply
5211 buffers (e.g. zconf client mount config records) due to memory
5212 fragmentation or consumption by the buffer cache. Preallocate
5213 some large reply buffers so that these replies can be sent even
5214 under memory pressure.
5218 Frequency : rare (liblustre)
5219 Description: fsx running with liblustre complained that using truncate() to
5220 extend the file doesn't work. This patch corrects that issue.
5221 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5222 ATTR_SIZE and lsm use in llu_setattr_raw.
5226 Frequency : rare, only 2.6 kernels
5227 Description: Unusual file access patterns on the MDS may result in inode
5228 data being lost in very rare circumstances.
5229 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5230 code in the 2.6 kernel caused the inode and/or EA data not to
5231 be read from disk, causing single-file corruption.
5235 Frequency : rare, only 2.6 filesystems using extents
5236 Description: Heavy concurrent write and delete load may cause data corruption.
5237 Details : It was possible under high-load situations to have an extent
5238 metadata block in the block device cache from a just-unlinked
5239 file overwrite a newly-allocated data block. We now unmap any
5240 metadata buffers that alias just-allocated data blocks.
5244 Frequency : filesystems with default stripe_count larger than 77
5245 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5246 Details : lconf specifies an inode size of 4096 bytes when the default
5247 stripe_count is larger than 77. This conflicts with the default
5248 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5249 this case to avoid pinning too much memory for large EAs.
5251 ------------------------------------------------------------------------------
5253 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5258 Frequency : rare (extremely heavy IO load with hundreds of clients)
5260 Description: Client is evicted, gets IO error writing to file
5261 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5262 caused clients to be evicted for AST timeouts. The fixes in
5263 bug 5192 mean we no longer need to have such short AST timeouts
5264 so ldlm_timeout has been increased.
5267 Frequency : occasional during --force or --failover shutdown under load
5268 Bugzilla : 5949, 4834
5269 Description: Server oops/LBUG if stopped with --force or --failover under load
5270 Details : a collection of import/export refcount and cleanup ordering
5271 issues fixed for safer force cleanup
5274 Frequency : only filesystems larger than 120 OSTs
5275 Bugzilla : 5990, 6223
5276 Description: lfs getstripe would oops on a very large filesystem
5277 Details : lov_getconfig used kfree on vmalloc'd memory
5280 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5281 Bugzilla : 6242, 6243
5282 Description: reading from files that had been truncated to a non-zero size
5283 but never opened returned no data
5284 Details : ll_file_read() reads zeros from no-object files to EOF
5289 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5290 Details : There's little harm in aborting MDS/OSS recovery and letting it
5291 try again, so I removed the LASSERT and return an error instead.
5293 Severity : enhancement
5295 Description: New debugging infrastructure for tracking down data corruption
5296 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5297 (b) cover more of the client-side code path, and (c) try to narrow
5298 down where problems occurred
5302 Bugzilla : 3819, 4364, 4397, 6313
5303 Description: Racing close and eviction MDS could cause assertion in mds_close
5304 Details : It was possible to get multiple mfd references during close and
5305 client eviction, leading to one thread referencing a freed mfd.
5307 Severity: : enhancement
5308 Bugzilla : 3262, 6359
5309 Description: Attempts to reconnect to servers are now more aggressive.
5310 Details : This builds on the enhanced upcall-less recovery that was added
5311 in 1.4.2. When trying to reconnect to servers, clients will
5312 now try each server in the failover group every 10 seconds. By
5313 default, clients would previously try one server every 25 seconds.
5318 Description: After recovery, certain operations trigger a failed
5319 assertion on a client.
5320 Details : Failing over an mds, using lconf -d --failover, while a
5321 client was doing a readdir() call would cause the client to
5322 LBUG after recovery completed and the readdir() was resent.
5324 Severity : enhancement
5326 Description: Default groups are now added by lconf
5327 Details : You can now run lconf --group <servicename> without having to
5328 manually add groups with lmc.
5331 Frequency : occasional
5333 Description: Nodes with an elan id of 0 trigger a failed assertion
5336 Frequency : always when accessing e.g. tty/console device nodes
5338 Description: tty and some other devices nodes cannot be used on lustre
5339 Details : file's private_data field is used by device data and lustre
5340 values in there got lost. New field was added to struct file to
5341 store fs-specific private data.
5344 Frequency : when exporting Lustre via NFS
5346 Description: NFSD failed occasionally when looking up a path component
5347 Details : NFSD is looking up ".." which was broken in ext3 directories
5348 that had grown large enough to become hashed.
5351 Frequency : Clusters with multiple interfaces not on the same subnet
5353 Description: Nodes will repeatedly try to reconnect to an interface which it
5354 cannot reach and report an error to the log.
5355 Details : Extra peer list entries will be created by lconf with some peers
5356 unreachable. lconf now validates the peer before adding it.
5359 Frequency : Only if a default stripe is set on the filesystem root.
5361 Description: Setting a default stripe on the filesystem root prevented the
5362 filesystem from being remounted.
5363 Details : The client was sending extra request flags in the root getattr
5364 request and did not allocate a reply buffer for the dir EA.
5367 Frequency : occasional, higher if lots of files are accessed by one client
5368 Bugzilla : 6159, 6097
5369 Description: Client trips assertion regarding lsm mismatch/magic
5370 Details : While revalidating inodes the VFS looks up inodes with ifind()
5371 and in rare cases can find an inode that is being freed.
5372 The ll_test_inode() code will free the lsm during ifind()
5373 when it finds an existing inode and then the VFS later attaches
5374 this free lsm to a new inode.
5378 Bugzilla : 6422, 7030
5379 Description: MDS deadlock between mkdir and client eviction
5380 Details : Creating a new file via mkdir or mknod (starting a transaction
5381 and getting the ns lock) can deadlock with client eviction
5382 (gets ns lock and trying to finish a synchronous transaction).
5385 Frequency : occasional
5386 Description: While starting a server, the fsfilt_ext3 module could not be
5388 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5389 kernels. Previously, lconf would still use the ext3 name
5390 when trying to load modules. Now, it will correctly use
5391 ext3 on 2.4 and ldiskfs on 2.6.
5393 Severity : enhancement
5394 Description: The default stripe count has been changed to 1
5395 Details : The interpretation of the default stripe count (0, to lfs
5396 or lmc) has been changed to mean striping across a single
5397 OST, rather than all available. For general usage we have
5398 found a stripe count of 1 or 2 works best.
5400 Severity : enhancement
5401 Description: Add support for compiling against Cray portals.
5402 Details : Conditional compiling for some areas that are different
5406 Frequency : occasional
5407 Bugzilla : 6409, 6834
5408 Description: Creating files with an explicit stripe count may lead to
5409 a failed assertion on the MDS
5410 Details : If some OSTs are full or unavailable, creating files may
5411 trigger a failed assertion on the MDS. Now, Lustre will
5412 try to use other servers or return an error to the
5416 Frequency : occasional
5418 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5419 caused lock timeout during readahead (since 1.4.2).
5420 Details : Processes doing readahead might match a lock that hasn't been
5421 granted yet if there are overlapping and conflicting lock
5422 requests. The readahead process waits on ungranted lock
5423 (original lock is CBPENDING), while OST waits for that process
5424 to cancel CBPENDING read lock and eventually evicts client.
5426 Severity : enhancement
5428 Description: Initial enabling of flock support for clients
5429 Details : Implements fcntl advisory locking and file status functions.
5430 This feature is provided as an optional mount flag (default
5431 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5432 locking are implemented yet, and those that are are not guaranteed
5433 to be completely correct in production environments.
5434 mount -t lustre -o [flock|noflock] ...
5437 Frequency : occasional
5439 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5440 assertion in the ext3 JBD (journaling) layer.
5441 Details : The b_committed_data struct is protected by the big kernel lock
5442 in 2.4 kernels, serializing journal_commit_transaction() and
5443 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5444 access to this struct. In 2.6 kernels there is finer grained
5445 locking to improve SMP performance of the JBD layer.
5449 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5454 Description: merge and cleanup kernel patches.
5457 -----------------------------------------------------------------------------
5459 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5461 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5462 versions of Lustre. Please update all servers and clients to
5463 version 1.4.2 or later at the same time. You must also run
5464 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5465 to update the configuration logs.
5467 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5468 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5469 - swab llog records of type '0' so we get proper header size/idx (5861)
5470 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5471 - fix rename of one directory over another leaking an inode (5953)
5472 - avoid SetPageDirty on 2.6 (5981)
5473 - don't re-add just-being-destroyed locks to the waiting list (5653)
5474 - when creating new directories, inherit the parent's custom
5475 striping settings if present parent (3048)
5476 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5477 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5478 - don't hold client locks on temporary worklist from l_lru (5666)
5479 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5480 - automatically evict dead clients (5921)
5481 - Update file size properly in create+truncate+fstat case (6196)
5482 - Do not unhash mountpoint dentries, do not allow removal of
5484 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5485 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5486 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5487 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5488 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5489 - Correctly handle reads of files with no objects (6243)
5490 - lctl recover will also mark a device active if deactivate used (5933)
5492 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5493 - allow --write-conf on an MDS with different nettype than client (5619)
5494 - don't write config llogs to MDS for mounts not from that MDS (5617)
5495 - lconf should create multiple TCP connections from a client (5201)
5496 - init scripts are now turned off by default; run chkconfig --on
5497 lustre and chkconfig --on lustrefs to use them
5498 - upcalls are no longer needed for clients to recover to failover
5500 - add --abort-recovery option to lconf to abort recovery on device
5502 - add support for an arbitrary number of OSTs (3026)
5503 - Quota support protocol changes.
5504 - forward compatibility changes to wire structs (6007)
5505 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5506 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5507 - improved "lustre status" script
5508 - initialize blocksize for non-regular files (6062)
5509 - added --disable-server and --disable-client configure options (5782)
5510 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5511 - Vanilla 2.4.29 support
5512 - increase maximum number of obd devices to 520 (6242)
5513 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5514 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5515 SLES 9 2.6.5 kernels
5516 - limit stripes per file to 160 (the maximum EA size) (6093)
5518 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5521 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5522 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5523 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5524 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5525 - small ext3 extents cleanups and fixes (5733)
5526 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5527 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5528 - avoid needless client->OST connect, fix handle mismatch (5317)
5529 - fix DLM error path that led to out-of-sync client, long delays (5779)
5530 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5531 - fix several locking issues related to i_size (5492,5624,5654,5672)
5532 - don't move pending lock onto export if it is already evicted (5683)
5533 - fix kernel oops when creating .foo in unlinked directory (5548)
5534 - fix deadlock in obdfilter statistics vs. object create (5811)
5535 - use time_{before,after} to avoid timer jiffies wrap (5882)
5536 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5537 - Do not leak request if server was not able to process it (5154)
5538 - If mds_open unable to find parent dir, make that negative lookup(5154)
5539 - don't create new directories with extent-mapping (5909, 5936)
5541 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5542 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5543 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5544 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5545 - support for RHEL4 kernel on i686 (5773)
5546 - provide error messages when incompatible logs are encountered (5898)
5548 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5549 * version 1.4.0.10 (1.4.1 release candidate 1)
5551 - don't keep a lock reference when lock is not granted (4238)
5552 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5553 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5554 - drop import inflight refcount on signal_completed_replay error (5255)
5555 - unlock page after async write error during send (3677)
5556 - handle missing objects in filter_preprw_read properly (5265)
5557 - no transno return for symlink open, don't save no-trasno open (3440)
5558 - don't try to complete elan receive that already failed (4012)
5559 - free RPC server reply state on error (5406)
5560 - clean up thread from ptlrpc_start_thread() on error (5160)
5561 - readahead could read extra page into cache that wasn't ejected (5388)
5562 - prevent races in class_attach/setup/cleanup/detach (5260)
5563 - don't dereference de->d_inode after l_dput of de (5458)
5564 - use "int" for stripe value returned from lock_to_stripe (5544)
5565 - mballoc allocation and error-checking fixes in 2.6 (5504)
5566 - block device patches to fix I/O request sizes in 2.6 (5482)
5567 - look up hostnames for IB nals (5602)
5568 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5569 - don't start multiple acceptors for the same port (5277)
5570 - fix incorrect LASSERT in mds_getattr_name (5635)
5571 - export a proc file for general "ping" checking (5628)
5572 - fix "lfs check" to not block when the MDS is down (5628)
5574 - service request history (4965)
5575 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5576 - create an "evict_client" /proc entry on OSTs, like the MDS has
5577 - fix mount usage message, return errors per mount(8) (5168)
5578 - change grep [] to grep "[]" in tests so they work in more UMLs
5579 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5580 - remove ancient LOV_MAGIC_V0 EA support (5047)
5581 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5582 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5583 - allow readahead allocations to fail when low on memory (5383)
5584 - mmap locking landed again, after considerable improvement (2828)
5585 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5587 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5590 - send OST transaction number in read/write reply to free req (4966)
5591 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5592 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5593 - the watchdog thread now runs as interruptible (5246)
5594 - flock/lockf fixes (but it's still disabled, pending 5135)
5595 - don't use EXT3 constants in llite code (5094)
5596 - memory shortage at startup could cause assertion (5176)
5598 - reorganization of lov code
5599 - single portals codebase
5601 - add extents/mballoc support (5025)
5602 - direct I/O reads in the obdfilter (4048)
5603 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5605 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5608 - send OST transaction number in read/write reply to free req (4966)
5609 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5610 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5611 - the watchdog thread now runs as interruptible (5246)
5612 - handle missing objects in filter_preprw_read properly (5265)
5613 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5614 - drop import inflight refcount on signal_completed_replay error (5255)
5615 - unlock page after async write error during send (3677)
5616 - return original error code on reconstructed replies (3761)
5617 - no transno return for symlink open, don't save no-trasno open (3440)
5619 - add pid to ldlm debugging output (4922)
5620 - bump the watchdog timeouts -- we can't handle 30sec yet
5621 - extra debugging for orphan dentry/inode bug (5259)
5623 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5626 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5627 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5628 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5629 - fix NULL dereference in /proc/sys/portals/routes (4827)
5630 - allow failed mdc_close() operations to be interrupted (4561)
5631 - stop precreate on OST before MDS would time out on it (4778)
5632 - don't send partial-page writes before EOF from client (4410)
5633 - discard client grant for sub-page writes on large-page clients (4520)
5634 - don't free dentries not owned by NFS code, check generation (4806)
5635 - fix lsm leak if mds_create_objects() fails (4801)
5636 - limit debug_daemon file size, always print CERROR messages (4789)
5637 - use transno after validating reply (3892)
5638 - process timed out requests if import state changes (3754)
5639 - update mtime on OST during writes, return in glimpse (4829)
5640 - add mkfsoptions to LDAP (4679)
5641 - use ->max_readahead method instead of zapping global ra (5039)
5642 - don't interrupt __l_wait_event() during strace
5644 - add software watchdogs to catch hung threads quickly (4941)
5645 - make lustrefs init script start after nfs is mounted
5646 - fix CWARN/ERROR duplication (4930)
5647 - return async write errors to application if possible (2248)
5648 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5649 - print NAL number in %x format (4645)
5650 - update barely-supported suse-2.4.21-171 series (4842)
5651 - support for sles 9 %post scripts
5652 - support for building 2.6 kernel-source packages
5653 - support for sles km_* packages
5655 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5658 - ignore -ENOENT errors in osc_destroy (3639)
5659 - notify osc create thread that OSC is being cleaned up (4600)
5660 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5661 - reconstruct ost_handle() like mds_handle() (4657)
5662 - create a new thread to do import eviction to avoid deadlock (3969)
5663 - let lconf resolve symlinked-to devices (4629)
5664 - don't unlink "objects" from directory with default EA (4554)
5665 - hold socknal file ref over connect in case target is down (4394)
5666 - allow more than 32000 subdirectories in a single directory (3244)
5667 - fix blocks count for O_DIRECT writes (3751)
5668 - OST returns ENOSPC from object create when no space left (4539)
5669 - don't send truncate RPC if file size isn't changing (4410)
5670 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5671 - bind to privileged port in socknal and tcpnal (3689)
5673 - rate limit CERROR/CWARN console message to avoid overload (4519)
5674 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5675 - basic mmap support (3918)
5676 - kernel patch series update from b1_4 (4711)
5678 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5681 - avoid crash during MDS cleanup with OST shut down (2775)
5682 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5683 - don't use bad inodes on the MDS (3744)
5684 - dynamic object preallocation to improve recovery speed (4236)
5685 - don't hold spinlock over lock dumping or change debug flags (4401)
5686 - don't zero obd_dev when it is force cleaned (3651)
5687 - print grants to console if they go negative (4431)
5688 - "lctl deactivate" will stop automatic recovery attempts (3406)
5689 - look for existing locks in ldlm_handle_enqueue() (3764)
5690 - don't resolve lock handle twice in recovery avoiding race (4401)
5691 - revalidate should check working dir is a directory (4134)
5693 - don't always mark "slow" obdfilter messages as errors (4418)
5695 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5698 - don't close LustreDB during write_conf until it is done (3860)
5699 - fix typo in lconf for_each_profile (3821)
5700 - allow dumping logs from multiple threads at one time (3820)
5701 - don't allow multiple threads in OSC recovery (3812)
5702 - fix debug_size parameters (3864)
5703 - fix mds_postrecov to initialize import for llog ctxt (3121)
5704 - replace config semaphore with spinlock (3306)
5705 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5706 - don't allow enqueue to complete on a destroyed export (3822)
5707 - down write_lock before checking llog header bitmap (3825)
5708 - recover from lock replay timeout (3764)
5709 - up llog sem before sending rpc (3652)
5710 - reduce ns lock hold times when setting kms (3267)
5711 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5712 - fix NULL deref and obd_dev leak on setup error (3312)
5713 - replace some LBUG about llog ops with error handling (3841)
5714 - don't match INVALID dentries from d_lookup and spin (3784)
5715 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5716 - fix invalid assertion in ptlrpc_set_wait (3880)
5718 - add libwrap support for the TCP acceptor (3996)
5719 - add /proc/sys/portals/routes for non-root route listing (3994)
5720 - allow setting MDS UUID in .xml (2580)
5721 - print the stack of a process that LBUGs (4228)
5723 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5726 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5727 - make sure to unset replay flag from failed open requests (3440)
5728 - if default stripe count is 0, use OST count for inode size (3636)
5729 - update parent mtime/ctime on client for create/unlink (2611)
5730 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5731 - free recovery state on server during a forced cleanup (3571)
5732 - unregister_reply for resent reqs (3063)
5733 - loop back devices mounting and status check on 2.6 (3563)
5734 - fix resource-creation race that can provoke i_size == 0 (3513)
5735 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5736 - more debugging for page-accounting assertion (3746)
5737 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5738 - avoid deadlock after precreation failure (3758)
5739 - fix race and lock order deadlock in orphan handling (3450, 3750)
5740 - add validity checks when grabbing inodes from l_ast_data (3599)
5742 - add /proc/.../recovery_status to obdfilter (3428)
5743 - lightweight CDEBUG infrastructure, debug daemon (3668)
5744 - change default OSC RPC parameters to be better on small clusters
5745 - turn off OST read cache for files smaller than 32MB
5746 - install man pages and include them in rpms (3100)
5747 - add new init script for (un)mounting lustre filesystems (2593)
5748 - run chkconfig in %post for init scripts (3701)
5749 - drop scimac NAL (unmaintained)
5751 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5754 - clean kiobufs before and after use (3485)
5755 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5756 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5757 - change init script to fail more clearly if not run as root (1528)
5758 - allow clients to reconnect during replay (1742)
5759 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5760 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5761 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5762 - fix cond_resched (3554)
5763 - only evict unfinished clients after recovery (3515)
5764 - allow bulk resend, prevent data loss (3570)
5765 - dynamic ptlrpc request buffer allocation (2102)
5766 - don't allow unlinking open directory if it isn't empty (2904)
5767 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5768 - remove extraneous obd dereference causing LASSERT failure (3334)
5769 - don't use get_cycles() when creating temp. files on the mds (3156)
5770 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5771 - handle EEXIST for set-stripe, set proper directory name (3336)
5773 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5774 - fix ksocknal_fmb_callback() error messages (2918)
5776 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5779 - don't copy lvb into (possibly NULL) reply on error (2983)
5780 - don't deref dentry after dput, don't free lvb on error (2922)
5781 - use the kms to determine writeback rpc length (2947)
5782 - increment oti_logcookies when osc is inactive (2948)
5783 - update client's i_blocks count via lvb messages (2543)
5784 - handle intent open/close of special files properly (1557)
5785 - mount MDS with errors=remount-ro, like obdfilter (2009)
5786 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5787 - don't use cancelling-locks' kms values (2947)
5788 - use highest lock extent for kms, not last one (2925)
5789 - don't dereference ERR_PTR() dentry in error handling path (3107)
5790 - fix thread race in portals_debug_dumplog() (3122)
5791 - create lprocfs device entries at setup instead of at attach (1519)
5792 - common AST error handler, don't evict client on completion race (3145)
5793 - zero nameidata in detach_mnt in 2.6 (3118)
5794 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5795 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5796 - initialize RPC timeout timer earlier for 2.6 (3219)
5797 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5798 - print nal/nid for unknown nid (3258)
5799 - additional checks for oscc recovery before doing precreate (3284)
5800 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5801 - don't crash in mdc_close for bad permissions on open (3285)
5802 - zero i_rdev for non-device files (3147)
5803 - clear page->private before handing to FS, better assertion (3119)
5804 - tune the read pipeline (3236)
5805 - fix incorrect decref of invalidated dentry (2350)
5806 - provide read-ahead stats and refine rpc in flight stats (3328)
5807 - don't hold journal transaction open across create RPC (3313)
5808 - update atime on MDS at close time (3265)
5809 - close LDAP connection when recovering to avoid server load (3315)
5810 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5811 - don't leak open file on MDS after open resend (3325)
5812 - serialize filter_precreate and filter_destroy_precreated (3329)
5813 - loop device shouldn't call sync_dev() for nul device (3092)
5814 - clear page cache after eviction (2766)
5815 - resynchronize MDS->OST in background (2824)
5816 - refuse to mount the same filesystem twice on same mountpoint (3394)
5817 - allow llmount to create routes for mounting behind routers (3320)
5818 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5819 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5820 - fix rare problem with rename on htree directories (3417)
5822 - allow default OST striping configuration per directory (1414)
5823 - fix compilation for qswnal for 2.6 kernels (3125)
5824 - increase maximum number of MDS request buffers for large systems
5825 - change liblustreapi to be useful for external progs like lfsck (3098)
5826 - increase local configuration timeout for slow disks (3353)
5827 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5829 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5832 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5833 - don't overwrite extent policy data in reply if lock was blocked (2901)
5834 - drop filter export grants atomically with removal from device (2663)
5835 - del obd_self_export from work_list in class_disconnect_exports (2908)
5836 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5837 - swab reply message in mdc_close, other PPC fixes (2464)
5838 - fix destroying of named logs (2325)
5839 - overwrite old logs when running lconf --write_conf (2264)
5840 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5841 - fix race in target_handle_connect (2898)
5842 - mds_reint_create() should take same inode create lock (2926)
5843 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5844 - don't close files for self_export to avoid uninitialized obd (2936)
5845 - allow MDS with the same name as client node (2939)
5846 - hold dentry reference for closed log files for unlink (2325)
5847 - reserve space for all logs during transactions (2059)
5848 - don't evict page beyond end of stripe extent (2925)
5849 - don't oops on a deleted current working directory (2399)
5850 - handle hard links to targets without a parent properly (2517)
5851 - don't dereference NULL lock when racing during eviction (2867)
5852 - don't grow lock extents when lots of conflicting locks (2919)
5854 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5857 - account for cache space usage on clients to avoid data loss (974)
5858 - lfsck support in lustre kernel code (2349)
5859 - reduce journal credits needed for BRW writes (2370)
5860 - orphan handling to avoid losing space on client/server crashes
5861 - ptlrpcd can be blocked, stopping ALL progress (2477)
5862 - use lock value blocks to assist in proper KMS, faster stat (1021)
5863 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5864 - recovery for initial connections (2355)
5865 - fixes for mds_cleanup_orphans (1934)
5866 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5867 - block all file creations until orphan recovery completes (1901)
5868 - client remove rq_connection from request struct (2423)
5869 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5870 - recovery timer race (2670)
5871 - mdc_close recovey bug (2532)
5872 - ptlrpc cleanup bug (2710)
5873 - mds timeout on local locks (2588)
5874 - namespace lock held during RPCs (2431)
5875 - handle interrupted sync write properly (2503)
5876 - don't try to handle a message that hasn't been replied to (2699)
5877 - client assert failure during cleanup after abort recovery (2701)
5878 - leak mdc device after failed mount (2712)
5879 - ptlrpc_check_set allows timedout requests to complete (2714)
5880 - wait for inflight reqs when ptlrpcd finishes (2710)
5881 - make sure unregistered services are removed from the srv_list
5882 - reset bulk XID's when resending them (caught by 1138 test)
5883 - unregister_bulk after timeout
5884 - fix lconf error (2694)
5885 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5886 - readahead locks pages, leaves pending causing memory pressure (2673)
5887 - increase OST request buffers to 4096 on large machines (2729)
5888 - fix up permission of existing directories in simple_mkdir (2661)
5889 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5890 - don't assign transno to errored transactions (2742)
5891 - don't delete objects on OST if given a bogus objid from MDS (2751)
5892 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5893 - if rq_no_resend, then timeout request after recovery (2432)
5894 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5895 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5896 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5897 - MDS recovery shouldn't delete objects that it has given out (2730)
5898 - if enqueue arrives after completion, don't clobber LVB (2819)
5899 - don't unlock pages twice when trigger_group_io returns error (2814)
5900 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5901 - don't write pages to disk if there was an error (1450)
5902 - don't ping imports that have recovery disabled (2676)
5903 - take buffered bytes into account when balancing socknal conn (2817)
5904 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5905 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5906 - remove little-endian swabbing of llog records (1987)
5907 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5908 - retry reposting request buffers if they fail (1191)
5909 - grow extent at grant time to avoid granting a revoked lock (2809)
5910 - lock revoke doesn't evict page if covered by a second lock (2765)
5911 - disable VM readahead to avoid reading outside lock extents (2805)
5913 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5914 - updated kernel patches for hp-2.4.20 kernel (2681)
5916 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5919 - fix truncated write corruption (2366)
5920 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5922 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5923 - ensure len(uuid) < 37 in lmc (1171)
5924 - fix ia64 OOPS in llog_test (2255)
5925 - zero end of page at obdfilter for partial page writes (2648)
5926 - don't leave stale dentries around after renames (bug 2428)
5927 - fix timeouts when evicting a client with a single lock held (2642)
5928 - set deadline for the initial HELLO message to drain (2634)
5929 - print out dotted-quad IP addresses in the socknal (2302)
5931 - additional debugging for MDS client eviction problem (2443)
5932 - fix mkfsoptions support for osts (2603, 2604)
5934 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - add series for the vanilla 2.6.0 kernel
5938 - add series for the vanilla 2.4.24 kernel
5939 - add series for a cray x86/64 UL kernel drop
5940 - fix xattr patches for the vanilla 2.4.19 series
5942 - generate true UUIDs in lmc (1171)
5943 - have portals stack dumping break in UML (2466)
5944 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5945 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5946 - wake callback waiting threads on client eviction (2460)
5947 - Add --ptldebug and --subsystem to lmc (1719)
5948 - update assertion to allow safe interrupt allocation
5949 - set rq_no_resend for cancel requests (2432)
5950 - recalculate ptlrpcd timeout after resend (2494)
5951 - call vfs_rmdir when removing pending directories (2368)
5952 - fix renaming a file to itself (2429)
5953 - lmc creates a default one-stripe lov (2454)
5954 - expand procfs space to handle large clusters (2326)
5955 - increase UML stack to avoid overflow
5956 - update lconf's list of debug and subsystem masks
5957 - fix lfs find --obd (2510)
5958 - /proc tunable for disabling filter read caching (2591)
5959 - stop rpm packages from altering slapd.conf (2301)
5960 - disable nagle in the socknal under 0conf (2578)
5961 - choose mds inode size based on stripe count (2572)
5962 - fix kernel-source rpm problems (2516)
5964 - add --disable-doc to avoid pdf generation (2421)
5965 - update documentation, tests, type-os, comments
5966 - avoid format warnings on ia64
5967 - remove the TOE NAL
5968 - tiny code cleanups by removing unused fields
5970 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5973 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5974 - remove the most blatant lies from BUILDING file (2371)
5975 - change default debug level to reasonable production setting
5976 - reduce client side cache size to reduce cache flush time
5977 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5978 - make TCP zerocopy and pinger support enabled by default (2476)
5979 - sync writes completed after process exits caused crashes (2319)
5980 - maintain correct mount count on the MDS (2356)
5981 - backout 1557, because 2316 wasn't really fixed
5982 - better file I/O statistics gathering in /proc
5983 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5984 - another kernel patch to fix zero-copy TCP function export
5985 - don't take duplicate lock when processing re-sent getattr (2420)
5986 - lctl uses obd_self_export instead of creating new conn (2353)
5987 - MDS/OST recovery case which requires object creation asserted (2425)
5988 - move lfs from /usr/sbin to /usr/bin in packages
5989 - fix race between mds_client_add and mds_client_free (2417)
5990 - use kmalloc instead of slabs in portals (2430)
5991 - don't create duplicate records when a failover MDS is present (2442)
5992 - remove unnecessary mount age check (2332)
5993 - don't remove directory inodes from locks prematurely (2451)
5994 - don't break if MDS service name is the same as hostname (2103)
5995 - fix races in client write RPC generation when cache full (2482)
5997 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6000 - remove now-unused request->rq_obd (278)
6001 - if an allocation fails, print out how much memory we've used (1933)
6002 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6003 - add the "configurable stack size" patch to most series files (1256)
6004 - ability to write large log records, for 100+ OST configs (2306)
6005 - fix NULL deref when filter_prep fails (2314)
6006 - fix operator precedence error in filter_sync
6007 - dynamic allocation of socknal TX descriptors (2315)
6008 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6009 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6010 - ensure that CWARN messages in Portals always get to the syslog
6011 - __init/__exit are not for prototype decls (ldlm_init/exit)
6012 - x86-64 compile warning fixes
6013 - fix gateway LMC keyword conflict (2318)
6014 - fix MDS lock inversions in getattr/reint paths (1844)
6015 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6016 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6017 - DLM race condition prevented some lock evictions (2328)
6018 - ENOMEM detection and retry on socknal sends (2230)
6019 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6020 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6021 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6022 - better error messages when a client is rejected during recovery (1505)
6023 - avoid cancelling locks which were never granted, after failure (2330)
6024 - fix i_sem/journal inversion in mds_client_add (2333)
6025 - fix truncate/getattr lock cycle deadlock (2334)
6026 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6027 - fix two rare exit paths which could leak an l_lock() ref (2321)
6028 - fencepost error in MDS/OST orphan recovery (2226)
6029 - make log record alignment 8 bytes (1988)
6030 - lstripe now fails when requested offset > ost_count (2237)
6031 - ensure that all kernel series have a complete list.h (1607)
6032 - fix crashes in special-file operations (2316)
6033 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6035 - allow configurable automake binary, for testing new versions
6036 - small update to the lfs documentation
6038 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6040 * fix negative export reference count in fsfilt_sync (2312)
6042 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6043 * release candidate 0.9.1
6045 - orphans are moved into the PENDING directory for possible recovery
6046 - replayed opens now open by fid for orphan/rename safety (1042)
6047 - last close of an orphan inode generates a transno (683)
6048 - chdir() and mount() now pin the directory entry (1020)
6049 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6050 - discard very old requests without processing them (1502)
6051 - remove some common, well-understood CERRORs (1505)
6052 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6053 - clear "grant" flags in OST replies until OST grant code lands (1644)
6054 - fix read performance by not clobbering i_blksize on client (1598)
6055 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6056 - make LRU size a /proc tunable, clears locks when reduced (707)
6057 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6058 - server locks take references on exports now (1558)
6059 - build fixes for 2.4.20-rh trees (1663)
6060 - return an error from lov_create if all OSCs are inactive (1751)
6061 - fix import levels when a reconnect happens without a timeout (1597)
6062 - exit early from mds_open if we get a lookup error (1749)
6063 - partial page read at EOF wouldn't wait for disk before sending (1642)
6064 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6065 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6066 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6067 - write/truncate lock inversion (1639)
6068 - Don't auto-load obdclass, portals modules during cleanup (1495)
6069 - fix timestamps from jumping to "now" (1763)
6070 - extra journal assertions (1648)
6071 - add an extra multiunlink test (1771)
6072 - fix read_record/write_record API (1776)
6073 - fix leak of offset_extent, possible incorrect i_size later (1772)
6074 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6075 - Debugging for the kqswnal_get_idle_tx problems (1820)
6076 - Allow recovery to be attempted multiple times (1536)
6077 - Write out MDS last_rcvd file after it is first created (1600)
6078 - Fix tx_descriptor leak in failed transmit situations (1827)
6079 - ext3 journaling fixes for assertion failure after IO error (1871)
6080 - class_export_put() on freed export after completion AST error (1896)
6081 - Fix revalidate looping in VFS (1322)
6082 - Don't access a freed export during MDS_REINT timeout (1521)
6083 - Add open-unlink recovery support on the MDS (1673,1764)
6084 - Return an error if no MDS data was read from last_rcvd (1946)
6085 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6086 - Don't setup a disk device that doesn't match exported UUID (317)
6087 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6088 - avoid committing NULL handle in force close
6089 - local.sh is now a one-stripe LOV configuration
6090 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6091 - fix ext3 htree duplicate directory entry corruption (1516)
6092 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6093 - update to new LOV EA format (2097)
6094 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6096 2003-06-15 Phil Schwan <phil@clusterfs.com>
6099 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6100 - per-import/export recovery handling (958, 931, 959)
6101 - multiple last-rcvd slots, for serving multiple FSes (949)
6102 - connections are again shared between multiple imp/exports (963, 964)
6103 - "umount -f" would hang if any requests needed to be sent (393, 978)
6104 - avoid pinning large req buffer by copying for queued messages (989)
6105 - add "uuid" to "lctl device" command to help upcalls (991)
6106 - "open" RPCs with transnos would confuse recovery counters (1037)
6107 - do proper endian conversion of all wire messages (288, 340, 891)
6108 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6109 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6110 - fix ptlrpc_connection leak in target_handle_connect (1174)
6111 - fix import refcounting bug in OST and MDS cleanup (1134)
6112 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6113 - fix very unlikely obd_types race condition (501)
6114 - remove osc_open hack for echo_client (1187)
6115 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6116 - a failure in read_inode2 leads to deadlock (1139)
6117 - cancel ack-locks as soon as transaction is committed (1072)
6118 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6119 - make sure to commitrw after any preprw to avoid deadlock (1162)
6120 - failing to execute a file in a lustre FS would lock inode (1203)
6121 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6122 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6123 - obd_uuid2tgt would walk off the end of the list (1255)
6124 - on IA64 the osc would give portals incorrect bulk size (1258)
6125 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6126 - fix lock inversion caused by new llite matching code (1282)
6127 - limit the number of dirty pages on a client to 10MB (1286)
6128 - timed out locks were not being corrected cancelled (1289)
6129 - fix O_DIRECT above 4GB on IA-32 (1292)
6130 * major user-visible changes
6131 - fail out/fail over policy now controlled by the upcall (993)
6133 - add OBD_PING to check server availability and failure (954)
6134 - lustre messages are now sent in sending host order (288, 340, 891)
6135 - add eadatalen to MDS getattr reply (340)
6136 - OST read replies may contain second buffer, with per-page status (593)
6138 2003-03-11 Phil Schwan <phil@clusterfs.com>
6141 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6142 - failed open()s could cause deadlock; fixed (867, 869)
6143 - stop cancelling OST locks when files are closed (481)
6144 - overlapping XID spaces caused network corruption (851, 853)
6145 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6146 - setattr_raw added, to do single-RPC, server-side setattrs
6147 - lmc/lconf syntax change for OST UUIDs
6148 - fix crashy race condition between ptlrpc_free_req and osc_close
6149 - don't use request in mdc_enqueue if we hit a timeout (889)
6150 - don't set the inode i_size for regular files from the MDS (896)
6151 - handle out of order completion AST (842)
6152 - don't LBUG if a lock request times out after receiving AST (913)
6153 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6154 - if a bad lock AST arrives, send an error instead of dropping entirely
6155 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6156 - fix leak in bulk IO when only partially completed (899, 900, 926)
6157 - fix O_DIRECT for ia64 (55)
6158 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6159 - C-z after timeout could hang a process forever; fixed (977)
6161 - client-side I/O cache (678, 924, 929, 941, 970)
6163 - READPAGE and SETATTRs which don't take server-side locks get
6166 2003-02-11 Phil Schwan <phil@clusterfs.com>
6169 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6170 - Fix fsfilt_extN_readpage() to read a full page of directory
6171 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6172 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6173 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6174 - Most intent operations were removed, in favour of a new RPC mode
6175 that does a single RPC to the server and bypasses most of the VFS
6176 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6177 - Aggressively cancel local locks on DLM servers
6178 - mds_reint_unlink sends EA to the client if it's the last nlink.
6179 client uses that EA to unlink OST objects.
6180 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6181 - recursive symlinks were fixed (439)
6182 - fixed NULL deref in DEBUG_REQ
6183 - filter_update_lastobjid no longer calls sync, which annoyed extN
6184 - fixed multi-client small-writes to a single file problem (445)
6185 - fixed mtime updates during file writes (607)
6186 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6187 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6188 - fixed Linux OST size reporting problem (444, 656)
6189 - OST now updates object mtime with writes or setattr (607, 619)
6190 - client verifies file size before zeroing page past EOF (445)
6191 - OST now writes last allocated objid to disk with allocation (108)
6192 - LOV on echo now works (409)
6194 - mds_reint_unlink sends a new buffer, with the EA included. this
6195 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6196 set if a regular file was being unlinked, and it was the last link
6197 - use PtlGet from the target for bulk writes (315)
6198 - OST now updates object mtime with writes or setattr (607, 619)
6199 - LDLM now has a grant-time callback to revalidate locked items, if
6201 - Many MDS operations were reorganized to combat race conditions
6203 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6204 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6206 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6209 - Fully reactivate OST imports after reconnection (512, others)
6210 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6211 - More graceful error handling for truncating on dead OST (515)
6212 - Don't error out unless we're actually accessing dead stripes (474)
6213 - Fix garbage sizes when stripes are missing (410)
6214 - LRU counters were broken, causing constant lock purge (433, 432)
6215 - garbage on read from stripes with failed OSTs (441)
6216 - mark OSCs as active before reconnecting during recovery (438)
6217 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6218 - lfind did not preserve OST order in output (443)
6219 - symlinks cause hung clients, incorrect data (439)
6220 - stop dereferencing request after dropping refcount (457)
6221 - don't LASSERT(spin_is_locked) on non-SMP (455)
6222 - fixes for many rename() bugs
6223 - fstat didn't correctly synchronize attributes (399)
6224 - server must handle lock cancellation during blocking AST prep (487)
6225 - bulk descriptors were free()d too soon (511)
6226 - fix paths in lconf, which would load incorrect modules (451, 507)
6227 - fix confusing lconf 'host not found' error message (386)
6228 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6229 - fix race condition in mdc_blocking_ast() for inode access (526)
6230 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6231 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6232 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6233 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6234 - fix lstripe in several ways (kernel side) (527)
6235 - fix request leak in ldlm_cli_enqueue (262)
6236 - incorrect OSC was marked inactive after OST failure
6237 - call mds_fs_cleanup before unmounting filesystem (524)
6238 - fix races between taking ns_lock and ldlm_lock_change_resource
6239 - fix races updating LOV export open file list
6240 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6241 - fix recovery NULL deref in ldlm_cli_cancel_unused
6242 - fix some DLM races by using new hash table for lock handles (419)
6243 - permit the client to specify desired inodes, at replay
6244 - duplicate requests when we queue them for replay reintegration
6245 - fix last_rcvd offset calculation
6246 - sync after each recovered transaction, so we always make progress
6247 - never, not always, ERESTART requests without transnos
6248 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6250 - skip replay if the MDS says that the client is already connected
6251 - don't check for a recovery-enabled export to match lctl's UUID
6252 - don't INC_USE_COUNT for phantom exports
6253 - don't crash when cleaning up phantom exports (567)
6254 - don't double-finish or set replay data for errored mdc_open requests
6255 - abort requests when they time out, so we don't get old replies
6256 - send/receive replies for AST messages again
6257 - if the client says that it doesn't have the lock, cancel it on the
6259 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6260 mark it as destroyed, it will all work out when decref is called
6261 - fix module use counts (22, 581)
6263 - ASTs now expect a reply (server cancels lock on error reply)
6265 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6268 - fix many simultaneous client startup (392)
6269 - fix dentry->d_it clobbering
6270 - credentials weren't being shipped for readdir/getattr operations
6271 - remove invalid assertions triggered during some concurrent MD
6273 - proper Lustre versions added (336, 389)
6274 - fix memory leak for create error case (398)
6275 - fix LOV locking bug that would get cli/srv out of sync
6276 - fix echo client over LOV (409)
6277 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6278 - fix double-O_EXCL intent crash (424)
6279 - avoid sending multiple lock CANCELs (352)
6281 - MDS can do multi-client recovery (modulo bugs in new code)
6283 - many updates, edits, cleanups
6285 2002-11-18 Phil Schwan <phil@clusterfs.com>
6288 - fix null d_it dereference (346)
6289 - fix full OST/dbench hang (333)
6290 - fix permission problem with file removal (286)
6291 - fix removal of OSCs from LOV when they fail
6292 - fix NULL deref during bulk timeout (214)
6293 - fix problems related to multiple filesystems on one MDS (241)
6294 - fixed serious subtle metadata locking bugs
6295 - free locks on clients when inodes are removed due to memory
6297 - fix inode pointer in lock data (285)
6298 - partial support for multiple MDS on a single host (241)
6299 - data locks weren't cancelled at clear_inode time (290, 311)
6300 - intent locks could lead to unbounded lock growth (205)
6301 - added a maximum lock count, an LRU list, and a flusher
6302 - fix multiple rename (365)
6303 - properly abstracted the echo client
6304 - OSC locked 1 byte too many; fixed
6305 - rewrote brw callback code:
6306 - fixed recovery bugs related to LOVs (306)
6307 - fixed too-many-pages-in-one-write crash (191)
6308 - fixed (again) crash in sync_io_timeout (214)
6309 - probably fixed callback-related race (385)
6311 - Add capability to MDS protocol
6312 - LDLM cancellations and callbacks on different portals
6314 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6317 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6318 - defer open object destruction to close time (601981, 138)
6319 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6320 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6321 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6322 - don't call obd_setattr in ll_file_release for destroyed objects
6323 * protocol change to lustre_msg: move |version| and add |flags|
6324 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6325 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6326 * added replay of create, unlink, link and rename operations during
6327 MDS failover; recovery should be much more robust now
6328 * remove failed OSCs from LOVs (only lov_create uses this so far)
6329 * the lustre-HOWTO was brought (more) up to date (582544)
6331 2002-10-23 Phil Schwan <phil@clusterfs.com>
6334 - in-use dentries weren't being reused properly (617851)
6335 - prevent multiple LDLM setup (599178)
6336 - fix LOV size calculations for truncate (617853)
6337 - fix client handling of MDS intent errors (POSIX)
6338 - fix permission bug in lovstripe.c test (624321)
6339 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6340 - truncate past end of file could corrupt data
6341 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6342 - a race in recovery could return ETIMEDOUT to apps (623947)
6343 - building outside the source directory was fixed
6344 * the lustre-HOWTO was brought (more) up to date (582544)
6345 * major progress was made on recovery functionality
6347 2002-10-10 Phil Schwan <phil@clusterfs.com>
6350 - recovery deadlock fix
6351 - rm -rf causes LBUG fix (617817)
6352 - file open by multiple tasks fix (618962)
6353 - directory permissions bugs (602707 and 620007)
6354 - journal_stop fixed with locking (611313)
6355 - O_APPEND failures resolved (618273, perhaps 614459)
6356 - lconf PATH fix (619770)
6357 - IA64 build fix (621450)
6358 - RPC buffer sizes scale with amount of memory
6360 2002-10-01 Phil Schwan <phil@clusterfs.com>
6363 - locks would be cancelled without throwing away data pages,
6364 resulting in inconsistent data (605627)
6365 - inode attributes were not always being refreshed (605627, 612449)
6366 - lconf now continues to cleanup after lctl reports an error
6367 - MDS now enforces user permissions (602707)
6368 - lprocfs cleanup fixed, but not yet enabled (614157)
6369 - fixed infinite server hang, should a client not respond to an AST
6370 - avoid going into recovery if user calls readlink() with a buffer
6371 that's too small (613941)
6372 - AST RPCs no longer require replies (614867) -- this may be changed
6373 - don't crash server if client sends an IOV that's too big (611336)
6374 - fixed lock conversion deadlock (611892)
6375 - fixed the following of symlinks (614622)
6376 * recovery: the server can remove locks from a client that dies, other
6377 clients can make progress
6378 * more extN patch fixes
6379 * compile-time configurable ptlrpc buffer allocations
6381 - collaborative read cache document
6382 - Lustre Lite Performance CDR document-in-progress
6384 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6387 - fix typo in patch-2.4.18
6389 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6392 - clear ptlrpc request each time in handle_incoming_request()
6393 - unlink of files now destroys the object on the OST
6395 2002-09-19 Peter Braam <braam@clusterfs.com>
6397 * add hard link support
6398 * change obdfile creation method
6399 * kernel patch changed
6401 2002-09-19 Peter Braam <braam@clusterfs.com>
6404 - stack overflow bug in extN fixed
6406 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6408 * documentation updates
6409 - add man pages for config tools
6410 - update tests/README to describe testing with new config tools
6411 - finish metadata API descriptions
6412 * bug fixes and cleanups
6413 - statfs workaround for 16TB limit
6414 - LOV stripe allocation improved, can stripe on subset of OSTs
6415 - LOV file size/IO offset was wrong for files > 4GB in size
6416 - object EA data was being dropped, caused files to be unreadable
6417 - memory overflow with non-LOV OST caused memory corruption
6418 - fixed regression tests to work with new config tools, obdfilter
6419 - fixed bug when directory size became larger than 1 block
6420 - fixed bug (for single client case) when PWD was deleted
6421 - invalidate local directory pages when doing intent-based ops
6422 - avoid LDLM oops when lock callback contained bad data
6424 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6426 * documentation updates
6427 * bug fixes and cleanups
6428 - configuration tools
6431 - 64-bit compile warnings
6432 - 64-bit internal statfs data
6434 * test_brw on persistent OST devices
6436 * lprocfs (disabled)
6438 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6440 * documentation updates
6441 * bug fixes and cleanups
6442 * configuration tools
6444 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6447 * many small fixes to 0_5_4
6448 * io/network handling
6449 * thinkos in MDS operations
6451 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6454 * crucial basic fixes to 0.5.3
6455 * IOR, Iozone work over Elan
6458 2002-08-07 Phil Schwan <phil@clusterfs.com>
6459 * version 0_5_3, our first alpha
6460 * we use the new Portals iovs
6461 * documentation updates
6462 * bug fixes and cleanups
6463 * small changes in the DLM wire protocol
6465 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6466 * version 0_5_1 with some initial stability,
6467 * locking on MD and file I/O.
6468 * documentation updates
6469 * several bug fixes since 0.5.0
6470 * small changes in wire protocol
6472 2002-07-18 Phil Schwan <phil@clusterfs.com>
6474 * delivered as Lustre Light Alpha
6475 * fixed a crash after handling invalid MDS requests
6476 * fixed directory pages for architectures with non-4k pages sizes
6478 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6479 * release version v0_4_4
6480 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6482 * Fixes a number of interruption problems with OST operations.
6483 * Update documentation for portals header changes
6484 * Move all wire protocol structs/defines to lustre_idl.h
6485 * Fixes symlink length bug.
6486 * Add tcpdump to repository.
6488 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6489 * release version v0_4_3
6490 * Fixes statfs for inodes on extN.
6491 * Fixes bug in runtests which would delete /etc/hosts.
6492 * Use 64-bit object IDs wherever possible (not into VFS though)
6493 Remove ost_get_info, which is unused by lustre, and out of date.
6495 2002-07-03 Peter Braam <braam@clusterfs.com>
6496 * release version v0_4_2 Fixes a lookup error (type not passed)
6497 * move forward to head of Portals
6498 * move forward to latest Lustre kernel
6500 2002-06-25 Peter Braam <braam@clusterfs.com>
6501 * release version v0_4_1. Hopefully stable on single node use.